id
stringlengths 27
33
| source
stringclasses 1
value | format
stringclasses 1
value | text
stringlengths 13
1.81M
|
|---|---|---|---|
warning/0005/cond-mat0005426.html
|
ar5iv
|
text
|
# Untitled Document
Marketing Percolation
J. Goldenberg<sup>1</sup>, B. Libai<sup>2</sup>, S. Solomon<sup>3</sup>, N. Jan<sup>4</sup>, and D. Stauffer<sup>5</sup>
<sup>1</sup> School of Business and Administration, Mt. Scopus, Hebrew University, Jerusalem 91905, Israel
<sup>2</sup> Davidson Faculty of Industrial Engineering and Management, Technion, Haifa 32000, Israel.
<sup>3</sup> Racah Institute of Physics, Givat Ram, Hebrew University, Jerusalem 91904, Israel
<sup>4</sup> Physics Department, St Francis Xavier University, Antigonish, Nova Scotia, Canada B2G 2W5
<sup>5</sup> Institute for Theoretical Physics, Cologne University, D-50923 Köln, Germany
e-mail: sorin@vms.huji.ac.il
Abstract: A percolation model is presented, with computer simulations for illustrations, to show how the sales of a new product may penetrate the consumer market. We review the traditional approach in the marketing literature, which is based on differential or difference equations similar to the logistic equation (Bass 1969). This mean field approach is contrasted with the discrete percolation on a lattice, with simulations of ”social percolation” (Solomon et al 2000) in two to five dimensions giving power laws instead of exponential growth, and strong fluctuations right at the percolation threshold.
1. Introduction.
If the amount of activity in an academic area reflects its importance, then research on the diffusion of innovations, with over 4,000 diffusion publications since 1940, is one of the most important areas in the social sciences. “No other field of behavioral science research represents more effort by more scholars in more disciplines in more nations” . Marketing’s considerable share of the output in this research stream reflects not only the importance of new products, but also the role of diffusion research in helping managers to better plan their entry strategy, target the right consumer and anticipate demand so as to have an efficient and effective promotion, production and distribution strategy. (We use here the terminology of marketing theory and call them diffusion models, whereas the physics of diffusion is quite a different process.) Diffusion as marketing experts define it is the development (increase) of sales over time (not spatial again), it is viewed as analogous to epidemics with increasing number of ill people. So is the new product: increasing number of adopters is in essence the diffusion process. The growth of new products is a complex process which typically consists of a large body of consumers interacting with each other over a long period of time.
Distressingly, often only aggregate data on adoption (i.e. the sum of all previous sales etc.) is available to researchers for analysis, as is generally the case with market level diffusion models . (Aggregate data means that the sales are measured once in a quarter or a year, without any attention to spatial distribution, and no attention to the individual buyer.) Even when collecting data at the individual level, diffusion research surveys consist of correlated data gathered in one “snap shot” survey of consumers, a methodology that amounts to freezing the diffusion process, making the continuous time-dependent process timeless .
Hence, it is not surprising that much of the theoretical base to the diffusion of innovations is grounded on repeatedly analyzed small number of data sets, in which researchers could actually follow the diffusion process within small social systems, such as the cases of the diffusion of hybrid corn among farmers in Iowa , antibiotics among US physicians or family planning in Korean villages . While the impressive contribution of these studies is evident, new tools should be considered to analyze the fast changing and complex environment of new product growth.
The small set of available individual based data poses another research dilemma: The small number of cases can not offer us an over-view of how collective behavior emerges from changes in individual characteristics. The span of individual level parameters is too small to allow for developing an explanation of their relations to the diffusion parameters or to predict them from the diffusion parameters.
Thus, the modeling of the diffusion of new products lies between two extremes.
Aggregate, or market level, diffusion models, such as the Bass model , an equation similar to what physicists call the logistic equation or Verhulst factor, are based on market level data and assume a large degree of homogeneity in the population of adopters. Basically, the diffusion of innovations models primary premise is based on the assumption that communication between individuals, is central to the new product’s growth.
One of the advantages of diffusion models is that they provide a relatively easy and parsimonious analytic way to look at the whole market and interpret its behavior, yet, still based on rich and empirically based theory. Another advantage is that very often the market level is also the level managers will be mostly interested in. Finally, aggregate models can be estimated with market level data such as number of adoptions in a given year or average price, which are relatively easy to get.
This simplicity is also associated with some critique on the aggregate approach to diffusion. One shortcoming is that the models make strong and simplifying assumptions on the behavior of individuals, for example the lack of heterogeneity among adopters. Also, the ability to test the assumptions these models make with very limited data at the aggregate level can be questioned .
Individual level models, on the other hand, acknowledge differences between consumers (e.g., difference in utility among potential adopters and their affect on adoption). Generally they follow economic theories (e.g. ) and assume that individuals maximize some personal objective function such as utility of the product, and may update their beliefs as more information arrives at the market. Thus, individual level models can be viewed as more behaviorally based than aggregate models.
Aiming at explaining aggregate adoptions in the market level, restrictions on the heterogeneity in behavior among individuals are sometimes introduced, and individual levels models are aggregated to provide an explicit diffusion function at the market level (e.g., ). Yet, the use of market level data to calibrate individual level models is still not very common, partly because the very limited aggregate level data do not really allow individual level testing, as the case in the traditional diffusion models.
Our study synthesizes individual and aggregate level modeling in a way which may help to overcome some of the outlined barriers. We follow diffusion theory and its emphasis on the communication behavior as a driver of new product growth, and generate a variety of possible dynamics to explore their influence on the aggregate level. Percolation enables us to perform sensitivity analysis and examine the effect of changes in the parameters in the individual level on the aggregate level, and thus overcome some of the limitations that follow the use of few data points at the aggregate level.
2: Diffusion Models: A Background
New products (in particular really new products) undergo a diffusion process: From an initial stage (in which there are zero buyers) individuals start to adopt the innovation and buy the product until the relevant market completely adopts it. Diffusion models try to explain and predict diffusion rates as a function of type of innovation, communication channels, nature of the social systems etc. Despite the large number of factors the models are parsimonious. The history of diffusion research in marketing is briefly presented below:
a) 1969: The Bass model
The modeling of the aggregate penetration of new products in the marketing literature generally follows the Bass model . The model follows Rogers’ diffusion of innovations theory of 1962 which emphasizes the role of communication methods: external influence (e.g., advertising, mass media) and internal influence (e.g. WOM = Word Of Mouth), as driving the product adoption pattern. Thus, an individual’s probability of adopting a new product at time $`t`$ (given that s/he had not adopted yet) depends in the Bass model linearly on two forces: a force which is not related to previous adopters and is represented by the parameter of external influence (traditionally denoted as $`p`$), and a force that is related to the number of previous adopters, the parameter of internal influence (denoted as $`q`$). The hazard model that describes the conditional probability of adoption at time $`t`$ is:
$$f(t)/[1F(t)]=p+qF(t)$$
$`(1)`$
where $`f(t)`$ is the probability of adoption at time $`t`$ and $`F(t)`$ describes the cumulative probability of adoption. Generally, $`p`$ represents the effect of external influences, i.e., influence not related to the number of previous adopters, such as advertising. $`q`$ represents the effect of internal influence, coming from previous adopters.
In the marketing practice eq.(1) is used in the form of eq.(2) in which $`n(t)`$ represents the buyers (or adopters) within a specified time interval and $`N(t)`$ is the cumulative number of buyers in a market of $`M`$ possible buyers:
$$n(t)=[p+q(N(t)/M)][MN(t)]$$
$`(2)`$
The Bass model has four main properties: i) It is the most dominant and popular. ii) It fits well many data. iii) After enough data points it is used in practice to forecast sales. iv) However its relevance to a real consumer behavior is questioned in several papers. Its significance (at least to the marketing people) lies also in the fact that the two main parameters can represent internal effects (due to previous adopting population) and external effects (not related to previous adopting population). In that it follows the diffusion of innovations theory, one of the well known theories of social sciences, that attributes the adoption rate of innovations to communication processes such as Word of Mouth from previous adopters (an internal effect) and mass media influence (an external effect).
b) 1978-79: Extensions of the basic Bass models
Modifications to increase the precision of the model in various cases were suggested. As an example consider a new class of flexible diffusion models, which allow non-symmetric patterns, heterogeneous adopters population etc. Those modifications were motivated by the need for better fit to real life data. A typical model from this generation is:
$$n(t)=[a+b(N(t)/c(M(t))^{1+d}][cM(t)N(t)]^{1+e}$$
$`(3)`$
where M is the market potential and A, b, c, d, e are estimated from the data.
c) 80’s and 90’s : more growth models
During the 80’s data on product penetration and diffusion were accumulated and diverse patterns were observed leading to suggestion of models with different penetration curves. Since growth modeling is an important occupation in a lot of fields, the marketing literature benefits from other fields’ achievements. But the main occupation consisted of tailoring a Bass-type model to a specific segment of innovation adoption. For example eq. 4 below was found to fit well adoptions of durables in the agricultural context.
$$n(t)=b[N(t)/M(t)][\mathrm{ln}M/n(t)]$$
$`(4)`$
In many cases these models do not relate to diffusion theory, rather they offer smoothing of a noisy data better then other regression technique. Furthermore, their relevance to marketing is sometimes criticized because they have little direct marketing application.
3. Shortcomings in this approach
Indeed, this research stream produced many extensions incorporating assumption regarding issues such as the effect of marketing mix, competition, repeat purchase and technological substitution (see, for example, reviews ). The prediction ability was reported to be satisfactory for various practical implications. However, it seems that this aggregate modeling approach reaches its limits. For instance, how can the coefficients of the smoothing function be interpreted in the individual level? This does not come straightforward from the Bass model equation, where $`p`$ and $`q`$ are part of the linear combination that governs the hazard rate.
Aggregate diffusion models make very simplifying assumptions that assume homogeneity in the communication behavior of adopters. However, while concern regarding this issue has been expressed throughout the diffusion literature (e.g. ), because of the nature of the very aggregate data available to researchers, limited options were available to those who wanted to examine these assumptions, and their implications.
In this paper we demonstrate how a microscopic presentation (more precisely percolation modeling) can be used to link market level models to individual level behavior. Further, it will allow us to examine the effect of heterogeneity in the communication behavior of adopters on the aggregate adoption level that are typically analyzed in aggregate diffusion models such as the Bass model and its extensions . In short, we replace the prevailing mean field theory by a more microscopic statistical approach, taking into account fluctuations and spatial correlations.
4. The percolation representation of product diffusion
Our technique is at once simple, direct and very powerful: represent in the computer the individual buyers, products and sales as well as the information transfer, and the changes in their current individual status.
Each site $`i`$ of a large lattice is occupied with a random number $`p_i`$ between zero and one, representing the customer’s quality expectation. The quality of a new product is called $`Q`$, and potential customers buy it only if this quality is above their expectations: $`Q>p_i`$ . This standard percolation model has a critical percolation threshold $`p_c`$ such that for $`Q>p_c`$ an infinite cluster of neighboring buyers can be formed, while for $`Q<p_c`$ all clusters of buyers are finite. There is a formal equivalence between this picture and the marketing phenomena: far below a certain quality level the product does not sell at all, while far above that density of buyers, the product reaches most of its potential market. The percolation literature contains much information about the spatial geometry of clusters which could be used for market modelling.
As long as the $`p_i`$ do not change, the cluster structures for different $`Q`$ are correlated. This suggests that one can use the recorded dynamics of one sweep in order to predict the behavior of the subsequent ones, or in general in order to characterize the cluster structure of the market. This line of thought is natural in the context of microscopic simulation but is quite novel in marketing.
Of course in reality even a product which ”makes it” may produce losses if the producer over estimates its market share and keeps producing after this is exhausted. On the other hand, the fluctuations (which the percolation model predicts) may discourage a producer and lead him to discontinue the production (flop) even in conditions in which the product could ”make it”.
In addition to the basic capability to express detailed spatio-temporal knowledge on the market structure and behavior the model above introduces significant conceptual departures from the main features and assumptions of the Bass model.
a) In the Bass model, the fluctuations around the Bass formula are assigned to measurements errors or to repeat purchases (especially close to the peak) in the microscopic simulation the fluctuations are the result of the random irregularities in the connectivity between various parts of the system. In fact fluctuations in the sales rate can appear even if one excludes the possibility of repeated purchases. In particular one can identify strongly connected clusters within which the sales front advances fast separated by regions poor in potential buyers which correspond to stagnation or slowing down in the sales.
b) We could generalize this site percolation picture to a site-bond percolation model, where connections between neighboring buyers are formed only with some bond probability; and then these bond probabilities increase with time if neighboring sites have bought a product. Then, if one allows for the existence of successive product waves (annual issues of the car models, movies in a series like star wars, pink panther, etc.) one obtains smoother curves and larger clusters than in the initial wave. This is due to the emergence of ”battered paths”: herds of buyers with regularly coordinated coherent response to the product. One can say that once the connections between the buyers clusters are established, they are less effective in slowing the propagation of the product sales front.
c) The functional form of the Bass curve is affected too: rather than an exponential increase during the entire pre-saturation region, one reaches a linear (or in general power) region of sales increase once a clear propagation front is formed. Indeed, the surface of a $`d`$-dimensional ball increases as the radius to the power $`d1`$.
d) The products which take off are the ones which happen to be planted in a cluster rich in potential buyers. If the cluster is small, the product sales will halt upon reaching the cluster boundary. If the cluster is large, the sales will be much higher.
e) Note that the usual Bass model is just the exact solution of the model in the extreme case where the ”neighbors” which link to each site are chosen randomly on the lattice (then the effect of common neighbors is negligible as long as the finite size is negligible) and in which instead of having a buy or a refuse to buy at each site one has always a buy (possibly with a different quantity expressing the buyer preference). In that case, which ignores the discrete character of the buying event (and the discrete choices of the discrete buyers) one gets an exponential increase followed by saturation and one has a perfect averaging of the buying rate by the various quantities bought by the buyers at the buying front.
f) In the case when the product quality $`Q`$ and the quality expectation $`p_i`$ change in time , the adaptation of the buyers tastes and the producers offer (in terms of quality and price) has the effect that after a few waves of similar products the system will be roughly at the boundary between the exponential decay and exponential increase of sales. Moreover one observes a certain convergence of tastes of the buyers towards a common behavior or towards separate groups which have convergent behavior within the group and divergent between the groups (large regions which react to a new product in a series coherently within the group and disjoint among the groups).
g) On top of this one can consider modeling the effects of peer pressure: sites which are not potential buyers becoming buyers when many of the neighbors bought the product. Sometimes this is not just a psychological effect: it is related with the utility of the product depending on its use by the other buyers (like in the case of fax, ps, pdf, word files formats).
h) The above simple percolation model with time-independent $`p_i`$ and $`Q`$ was used to produce in our figure various curves for $`n(t)`$, the number of new buyers which are neighbors to site which have already bought the product in the previous time interval. We start with one buyer and then let the buying spread over the lattice with a Leath algorithm : At each time step, all neighbors of all previous buyers decide, once and for all, if they buy. Thus our time steps are microsteps in the sense of Huang , for one spread of one product through the lattice, and not macrosteps in the sense of Solomon et al referring to repeated attempts with different parameters like $`Q`$. We see in fig.1 single examples below, at and above the percolation threshold (no infinite cluster, one fractal infinite cluster, and one compact infinite cluster, respectively). In reality the time resolution may be less fine than in these simulations; this could be taken into account by binning together several consecutive time steps and thus reducing the short-time fluctuations without changing the long-time trends. Fig.2 shows averages over many samples at the percolation threshold; then the fluctuations vanish. Fig.3 shows such results also for higher dimensions $`d`$ where the initial sales grow roughly as $`t^{d1}`$, approaching for $`d\mathrm{}`$ the exponential growth of the traditional theories like eq.1. (For figures 2 and 3 we used the fully dynamic model where $`q`$ and the $`p_i`$ self-organize towards the percolation threshold $`p_c`$ in steps of 0.001; for fixed $`q=p_c`$ and fixed $`p_i`$ as for fig.1, the slopes are smaller and depend on whether we average over all clusters or only over the “infinite” clusters.) For comparison, Fig.4 shows two examples of real markets, for automobiles and for LCD color television sets in the 20th century, indicating, respectively, an exponential increase (or power law with a large exponent like 5) and a linear growth. The first example may be better described by a Bass-type theory, the second better by two-dimensional percolation as in Fig. 3.
i) As is customary for phase transitions in physics, the percolation transition implies that even if the probability distribution of the $`p_i`$ across the lattice is totally uniform, one ends up with localized clusters and sub-clusters of all scales including macroscopic inhomogeneity leading to macroscopic sales rate fluctuations. The fractal clustered character of the market and the bottlenecks are not detectable by the usual polling techniques. A uniform customer distribution leads at the percolation threshold to un-passable barriers and to (almost-) extinction of sales. Similarly, a minor increase of temperature can make the water boiling, without any change in the intermolecular forces.
5. Review of Cluster Geometry
We summarize here some well known percolation properties which could become relevant in a future more quantitative theory of marketing along our outlines. The geometry of percolation clusters has been studied since decades . Their surface should not be defined as the set of empty neighbors of occupied sites since the number of such empty neighbors is proportional to the number of occupied sites. Instead, the fractal dimension $`D`$ is a widespread quantitative measure, defined through $`MR^D`$ for large clusters with radius $`R`$ and $`M`$ occupied sites. On two-dimensional lattices, $`D`$ is about 1.6, 1.9 and exactly 2 for $`q`$ below, at and above $`p_c`$. For the largest cluster at $`p_c`$ one can replace the radius $`R`$ by the linear lattice dimension $`L`$ in the above mass-radius relation. The fluctuations in the mass of the largest cluster are at $`q=p_c`$ about as large as the average mass, even if $`L`$ goes to infinity.
The largest cluster at the percolation threshold, also called the incipient infinite cluster, consists mostly of dangling ends, that means of links through which no current flows if a voltage is applied to two points of the cluster. The current-carrying part of the backbone, varying as $`L^{1.6}`$, and consists mostly of sites which can be removed without cutting the cluster into parts. Red sites are the bottlenecks, removal of which cuts the cluster into separate parts; their number increases only as $`L^{0.75}`$. The number $`\mathrm{}`$ of sites which link within the incipient infinite cluster two sites at Euclidean distance $`r`$ is called the chemical distance and corresponds to the time (microsteps) needed in our model to transfer information from one customer to the other; it varies on average as $`r^{1.1}`$ for large distances.
All these exponents are valid rather generally in two dimensions, not only for nearest-neighbor connections on the square lattice. For example, we could allow information to flow also to the four next-nearest neighbors in addition to the four nearest neighbors. Then the value of $`p_c`$ would change but the above exponents would be the same and thus universal. Only if this distance of neighbors goes to infinity do we expect a behavior more similar to eqs.(1-4). Thus these exponents as opposed to $`p_c=0.592746`$ are rather general quantitative predictions of percolation theory to be compared with future high-precision data from real marketing.
6. Summary
The traditional approach towards marketing theory, in the literature cited here, has been replaced by a percolation model which treats each customer individually instead of averaging over all of them. As a result strong fluctuations are observed, as found in real sales curves. Just as other percolation applications, also the present one can be modified in numerous ways to describe better specific effects.
We thank Z.F. Huang for discussions and K-mart International Center of Marketing and Retailing, Davidson Center, Alexander Goldberg Academic Lectureship Fund, German-Israeli Foundation, NSERC of Canada and SFB 341 of Germany for partial support.
References
1. Rogers, E. M. (1995). The Diffusion of Innovations, 4th, New York: Free Press.
2. Sultan, F., Farley, J. U., and Lehmann, D. R. (1990). ”A Meta Analysis of Applications of Diffusion Models,” Journal of Marketing Research, 27, 70.
3. Mahajan, V., Muller, E., and Bass, F. M. (1990). ”New Product Diffusion Models in Marketing: A Review and Directions for Research,”Journal of Marketing, 54, 1.
4. Ryan B. and Gross, N.C. (1943). ”The Diffusion of Hybrid Seed Corn in Two Iowa Communities,” Rural Sociology, 8, 15.
5. Coleman, J.S., Katz, E., and Menzel, H. (1966). Medical Innovation: A Diffusion Study. Bobbs-Merrill, New York.
6. Rogers E. and Kincaid, D.L. (1981). Communication Networks: A New Paradigm for Research. New York: Free Press.
7. Bass, F. M., (1969). ”A New Product Growth Model for Consumer Durables,” Management Science, 15, 215.
8. Parker, P. M. (1994).”Aggregate Diffusion Models in Marketing: A Critical Review,” International Journal of Forecasting 10, 353.
9. Lancaster, K. (1966). ”A New Approach to Consumer Theory,” Journal of Political Economy 74, 132.
10. Chatterjee, R. and Eliashberg, J. (1990). ” The Innovation Diffusion Process in a Heterogeneous Population: A Micro Modeling Approach,” Management Science, 36, 1057.
11. Bass, F. M., Trichy, K., V., and Jain, D. C. (1994). ”Why the Bass Model Fits Without Decision Variables,” Marketing Science, 13, 203.
12. Mahajan, V. and Peterson, R. A. (1985). Models for Innovation Diffusion, Sage, Newbury Park, CA.
13. Solomon, S. and Weisbuch, G. e-print adapt-org 9909001; Solomon, S. Weisbuch, G., de Arcangelis L., Jan, N. and Stauffer, D. Physica A 277, 239 (2000)
14. Stauffer, D. and Aharony, A. (1994). Introduction to Percolation Theory. Taylor and Francis, London.
15. Sahimi, M. (1994). Applications of Percolation Theory. Taylor and Francis, London.
16. Evertz, H. G. (1993). J. Stat. Phys. 70, 1075.
17. Huang, Z.F. (2000). Int. J. Mod. Phys. C 11, 287 and Eur. Phys. J. B, in press.
18. The Electronic Market Data Book, Electronic Industries Alliance, Washington DC, 2000.
Figure Captions
Fig.1. Examples of simulated sales curves below (line), at (b, +) and above (x) the percolation threshold (part a). Parts b and c show the cluster geometry for the three curves of part a, after the cluster stopped growing or touched the upper boundary.
Fig.2. Averaged sales curves at the percolation threshold for $`L\times L`$ squares with $`L`$ = 100, 200, 500 and 1001 (from left to right): The individual fluctuations are washed out in the average.
Fig.3. Averaged sales curves for two to five dimensions; the straight lines give the theoretically expected slopes $`d1`$ in this log-log plot.
Fig.4. Yearly sales (in thousands) in the USA of automobiles (a) and of LCD color television sets (b), versus year .
|
warning/0005/nlin0005039.html
|
ar5iv
|
text
|
# References
A model for the emergence of cooperation, interdependence and structure in evolving networks
Sanjay Jain<sup>∗†‡1</sup> and Sandeep Krishna<sup>§2</sup>
Centre for Theoretical Studies and <sup>§</sup> Physics Department, Indian Institute of Science,
Bangalore 560 012, India
Santa Fe Institute, 1399 Hyde Park Road, Santa Fe, NM 87501, USA
Jawaharlal Nehru Centre for Advanced Scientific Research, Bangalore 560 064, India
<sup>1</sup> jain@santafe.edu, jain@cts.iisc.ernet.in; <sup>2</sup> sandeep@physics.iisc.ernet.in
Evolution produces complex and structured networks of interacting components in chemical, biological, and social systems. We describe a simple mathematical model for the evolution of an idealized chemical system to study how a network of cooperative molecular species arises and evolves to become more complex and structured. The network is modeled by a directed weighted graph whose positive and negative links represent ‘catalytic’ and ‘inhibitory’ interactions among the molecular species, and which evolves as the least populated species (typically those that go extinct) are replaced by new ones. A small autocatalytic set (ACS), appearing by chance, provides the seed for the spontaneous growth of connectivity and cooperation in the graph. A highly structured chemical organization arises inevitably as the ACS enlarges and percolates through the network in a short, analytically determined time scale. This self-organization does not require the presence of self-replicating species. The network also exhibits catastrophes over long time scales triggered by the chance elimination of ‘keystone’ species, followed by recoveries.
Structured networks of interacting components are a hallmark of several complex systems, for example, the chemical network of molecular species in cells , the web of interdependent biological species in ecosystems , and social and economic networks of interacting agents in societies . The structure of these networks is a product of evolution, shaped partly by the environment and physical constraints and partly by the population (or other) dynamics in the system. For example, imagine a pond on the prebiotic earth containing a set of interacting molecular species with some concentrations. The interactions among the species in the pond affect how the populations evolve with time. If some population goes to zero, or if new molecular species enter the pond from the environment (through storms, floods or tides), the effective chemical network existing in the pond changes. We discuss a mathematical model that attempts to incorporate this interplay between a network, populations, and the environment in a simple and idealized fashion. The model (including an earlier version ) was inspired by the ideas and results in refs. . Related but different models are studied in refs. .
The Model
The system consists of $`s`$ species labeled by the index $`i=1,2,\mathrm{},s`$. The network of interactions between species is specified by the $`s\times s`$ real matrix $`C\{c_{ij}\}`$. The network can be visualised as a directed graph whose nodes represent the species. A non-zero $`c_{ij}`$ is represented by a directed weighted link from node $`j`$ to node $`i`$. If $`c_{ij}>0`$ then the corresponding link is a cooperative link: species $`j`$ catalyzes the production of species $`i`$. If $`c_{ij}<0`$ it is a destructive link: the presence of $`j`$ causes a depletion of $`i`$.
Population dynamics. The model contains another dynamical variable $`𝐱(x_1,\mathrm{}.x_s)`$, where $`x_i`$ stands for the relative population of the $`i^{th}`$ species ($`0x_i1`$, $`_{i=1}^sx_i=1`$). The time evolution of $`𝐱`$ depends upon the interaction coefficients $`C`$, as is usual in population models. The specific evolution rule we consider is
| $`\dot{x}_i`$ | $`=`$ | $`f_i`$ | if $`x_i>0`$ | or | $`f_i0`$, |
| --- | --- | --- | --- | --- | --- |
| | $`=`$ | $`0`$ | if $`x_i=0`$ | and | $`f_i<0`$, |
(1)
$$\text{where }f_i=\underset{j=1}{\overset{s}{}}c_{ij}x_jx_i\underset{k,j=1}{\overset{s}{}}c_{kj}x_j.$$
This is a particularly simple idealization of catalyzed chemical reaction dynamics in a well stirred reactor, as explained later.
Graph dynamics. The dynamics of $`C`$ in turn depends upon $`𝐱`$, as follows: Start with a random graph of $`s`$ nodes: $`c_{ij}`$ is non-zero with probability $`p`$ and zero with probability $`1p`$. If nonzero, $`c_{ij}`$ is chosen randomly in the interval $`[1,1]`$ for $`ij`$ and $`[1,0]`$ for $`i=j`$. Thus a link between two distinct species, if it exists, is just as likely to be cooperative as destructive, and a link from a species to itself can only be inhibitive, i.e., autocatalytic or self-replicating individual species are not allowed. The variable $`𝐱`$ is initialized by choosing each $`x_i`$ randomly between $`0`$ and $`1`$, and then rescaling all $`x_i`$ uniformly such that $`_{i=1}^sx_i=1`$. The evolution of the network proceeds in three steps :
1) Keeping the network fixed, the populations are evolved according to (1) for a time $`T`$ which is large enough for $`𝐱`$ to get reasonably close to its attractor. We denote $`X_ix_i(T)`$.
2) The set of nodes $`i`$ with the least value of $`X_i`$ is determined. We call this the set of ‘least fit’ nodes, identifying the relative population of a species in the attractor (or, more specifically, at $`T`$) with its ‘fitness’ in the environment defined by the graph. One of the least fit nodes is chosen randomly (say $`i_0`$) and removed from the system along with all its links leaving a graph of $`s1`$ species.
3) A new node is added to the graph so that it again has $`s`$ nodes. The links of the added node ($`c_{ii_0}`$ and $`c_{i_0i}`$, for $`i=1,\mathrm{},s`$) are assigned randomly according to the same rule as for the nodes in the initial graph. The new species is given a small relative population $`x_{i_0}=x_0`$ and the other populations are rescaled to keep $`_{i=1}^sx_i=1`$. This process, from step 1 onwards, is iterated many times.
Motivation for model structure. The choice of Eq. (1) is motivated by the rate equations in a well stirred chemical reactor (representing, say, a prebiotic pond) as follows: If species $`j`$ catalyses the ligation of reactants $`A`$ and $`B`$ to form the species $`i`$, $`A+B\stackrel{j}{}i`$, then the rate of growth of the population $`y_i`$ of species $`i`$ will be given by $`\dot{y}_i=k(1+\nu y_j)n_An_B\varphi y_i`$, where $`n_A,n_B`$ are reactant concentrations, $`k`$ is the rate constant for the spontaneous reaction, $`\nu `$ is the catalytic efficiency, and $`\varphi `$ represents a common death rate or dilution flux in the reactor . Assuming the catalysed reaction is much faster than the spontaneous reaction, and the concentrations of the reactants are large and fixed, the rate equation becomes $`\dot{y}_i=cy_j\varphi y_i`$, where $`c`$ is a constant. If species $`i`$ has multiple catalysts, we get $`\dot{y}_i=_{j=1}^sc_{ij}y_j\varphi y_i`$. The first of equations (1) follows from this upon using the definition $`x_i=y_i/_{j=1}^sy_j`$. When negative links are permitted, the second of equations (1) is needed in general to prevent relative populations from going negative. (With negative links, a more realistic chemical interpretation would be obtained if $`\dot{x}_i`$ were proportional to $`x_i`$, but for simplicity we retain the form of (1) in this paper.) (1) may be viewed as defining an artificial chemistry in the spirit of refs. .
The rules for the evolution of the network $`C`$ are intended to capture two key features of natural evolution, namely, selection and novelty. The species that has the least population in the attractor configuration is the one most likely to be eliminated in a large fluctuation in a possible hostile environment. Often, the least value of $`X_i`$ is zero. Thus the model implements selection by eliminating from the network a species that has become extinct or has the least chance of survival . Novelty is introduced in the network in the form of a new species. This species has on average the same connectivity as the initial set of species, but its actual connections with the existing set are drawn randomly. E.g., if a storm brings into a prebiotic pond a new molecular species from the environment, the new species might be statistically similar to the one being eliminated, but its actual catalytic and inhibitory interactions with the surviving species can be quite different. Another common feature of natural evolution is that populations typically evolve on a fast time scale compared to the network. This is captured in the model by having the $`x_i`$ relax to their attractor before the network is updated. The idealization of a fixed total number of species $`s`$ is one that we hope to relax in future work.
The model described above differs from the one studied in in that it allows negative links and varying link strengths, and that the population dynamics, given by (1), is no longer linear. The earlier model had only fixed point attractors; here limit cycles are also observed. Since $`C`$ now has negative entries, the formalism of non-negative matrices no longer applies.
Results
Emergence of cooperation and interdependence. Figure 1 shows a sample run. The same qualitative behaviour is seen in each of several hundred runs performed for $`p`$ values ranging from $`0.00002`$ to $`0.01`$ and for $`s=100,150,200`$. The fact that the ratio of number of cooperative to destructive links at first remains constant at unity (statistically) and then increases by more than an order of magnitude is evidence of the emergence of cooperation. The increase in $`\overline{d}`$ by an order of magnitude is a quantitative measure of the increase of interdependence of species in the network. The increase in the total density of links $`(l_++l_{})/s`$ is another aspect of the increase of complexity of the system. Note that in the model selection rewards only ‘performance’ as measured in terms of relative population; the rules do not select for higher cooperativity per se. Since a new species is equally likely to have positive or negative links with other species, the introduction of novelty is also not biased in favour of cooperativity. The fact that this behaviour is not a consequence of any intrinsic bias in the model that favours the increase of cooperation and interdependence is evidenced by the flat initial region of all the curves.
Autocatalytic sets. The explanation for the above behaviour lies in the formation and growth of certain structures, autocatalytic sets (ACSs), in the graph. An ACS is defined as a set of nodes such that each node has at least one incoming positive link from a node in the set. Thus an ACS has the property of catalytic closure, i.e., it contains a catalyst for each of its members . The simplest example of an ACS is a cycle of positive links. Every ACS is not such a cycle but it can be shown that an ACS must contain a cycle of positive links . In Fig. 1, there is no ACS in the graph until $`n=1903`$. A small ACS (which happens to be a cycle of positive links between two nodes) appears at $`nn_1=1904`$, exactly where the behaviour of the $`s_1`$ curve changes. As time proceeds this ACS becomes more complex and enlarges until at $`nn_2=3643`$ the entire graph becomes an ACS. $`l_+`$ and $`\overline{d}`$ exhibit an increase and $`l_{}`$ a decrease as the ACS comes to occupy a significant part of the graph. After the ACS first appears (at $`n=n_1`$), the set of populated nodes in the attractor configuration ($`s_1`$ in number), is always an ACS (except for certain catastrophic events to be discussed later), which we call the ‘dominant ACS’. The spontaneous appearance of a small ACS at some $`n=n_1`$, its persistence (except for catastrophes), and its growth until it spans the graph at $`n=n_2`$, is seen in each of the several hundred runs mentioned earlier. The growth of the ACS across the graph between $`n_1`$ and $`n_2`$ occurs exponentially (with stochastic fluctuations),
$$s_1(n)s_1(n_1)e^{(nn_1)/\tau _g},\tau _g=2/p.$$
(2)
This agrees with simulations as shown in Figure 2. The average time scale $`\tau _an_1`$ for the first appearance of the ACS is given, for sufficiently small $`p`$, by $`\tau _a4/(p^2s)`$ ($`=1600`$ for $`p=0.005`$ and $`s=100`$).
Upto $`n=n_1`$, the graph has no ACS. It has chains and trees of positive and negative links and possibly loops containing negative links. These latter structures are not robust. For example, consider a chain of two positive links $`123`$. Since catalytic links are pointing to node $`3`$, it will do well populationally compared to nodes $`1`$ and $`2`$. However, since $`1`$ has no incoming catalytic links, its relative population will decline to zero under (1), and it can be picked for replacement in the next graph update. This can disrupt the chain and hence erode the ‘well being’ of node $`3`$ until eventually after some graph updates the latter can also join the ranks of the least fit. Species $`3`$ gets eliminated eventually because it does not feedback into and ‘protect’ species $`1`$ and $`2`$, on whom its ‘well being’ depends. In a graph without an ACS no structure is protected from disruption. Since every node is liable to be replaced sooner or later, the graph remains as random as the initial graph (we have checked that the probability distribution of the number of incoming and outgoing links at a node remains the appropriate binomial for $`n<n_1`$). This explains why $`s_1`$, $`l_\pm `$ and $`\overline{d}`$ hover around their initial values.
The picture changes the moment a small ACS appears in the graph. The key point is that by virtue of catalytic closure, members of the ACS do well collectively in the population dynamics governed by (1). An ACS is a collective self-replicator and beats chains, trees and other non ACS structures in the population game, reducing their $`X_i`$ to zero when it appears. Thus, since graph update proceeds by replacing one of the nodes with $`X_i=0`$ (if present) with a new one, such a replacement being outside the dominant ACS can cause no damage to the links that constitute the ACS. That is why the ACS structure, once it appears, is much more robust than the non-ACS structures discussed earlier. If the new node happens to get an incoming positive link from the dominant ACS, it becomes part of it. Thus the dominant ACS tends to expand in the graph as new nodes get attached to it and $`s_1`$ increases. In $`\mathrm{\Delta }n`$ graph updates the average increase in $`s_1`$, which is the number of added nodes which will get a positive link from one of the $`s_1`$ nodes of the dominant ACS, is $`\mathrm{\Delta }s_1(p/2)s_1\mathrm{\Delta }n`$, for small $`p`$. This proves (2). (Note that the exponential growth described by (2) is not to be confused with the exponential growth of populations $`y_i`$ of species that are part of the ACS. (2) reflects the growth of the ACS across the graph, or the increase in the number of species that constitute the ACS.)
Since the dominant ACS grows by adding positive links from the existing dominant ACS, the number of positive links increases as the ACS grows. On the other hand nodes receiving negative links usually end up being least fit, hence negative links get removed when these nodes are eliminated. Which novelty is captured thus depends upon the existing ‘context’; the network evolves by preferentially capturing links and nodes that ‘latch on’ cooperatively to the existing ACS and disregarding those that do not. The ‘context’ itself arises when the ACS structure first appears; this event transforms the nature of network evolution from random to ‘purposeful’ (in this case directed towards increasing cooperation). Before the ACS appears nothing interesting happens even though selection is operative (the least populated species are being eliminated). It is only after the ACS topological structure appears that selection for cooperation and complexity begins.
Inevitability of autocatalytic sets. Note that the appearance of an ACS, though a chance event, is inevitable. For $`sp1`$, the probability that a graph not containing a two cycle will acquire one at the next time step is $`p^2s/4q`$. Since the probability of occurrence of 3-cycles, etc., is much smaller, the probability distribution of arrival times $`n_1`$ is approximated by $`P(n_1)=q(1q)^{n_11}`$, whose mean $`\tau _a`$ is $`1/q`$. Since this probability declines exponentially after a time scale $`1/q`$, the appearance of an ACS is inevitable, even for arbitrarily small (but finite) $`p`$.
Occasionally in a graph update $`s_1`$ can decrease for various reasons. If the new node forms an ACS of its own with nodes outside the dominant ACS, and the new ACS has a higher population growth rate (as determined by (1)) than the old ACS it drives the species of the latter to extinction and becomes the new dominant ACS. Alternatively the new node could be a ‘destructive parasite’: it receives one or more positive links from and gives one or more negative links to the dominant ACS. Then part or whole of the ACS may join the set of least fit nodes. Structures that diminish the size of the dominant ACS or destroy it appear rarely. For example in Figure 1, destructive parasites appeared $`6`$ times at $`n=3388,3478,3576,3579,3592`$ and $`3613`$. In each case $`s_1`$ decreased by 1.
Emergence of structure. At $`n=n_2`$ the whole graph becomes an ACS; the entire system can collectively self-replicate despite the explicit absence of individual self-replicators. Such a fully autocatalytic set is a very non-random structure. Consider a graph of $`s`$ nodes and let the probability of a positive link existing between any pair of nodes be $`p^{}`$. Such a graph has on average $`m^{}=p^{}(s1)`$ incoming or outgoing positive links per node. For the entire graph to be an ACS, each node must have at least one incoming positive link, i.e., each row of the matrix $`C`$ must contain at least one positive element. Hence the probability, $`P`$, for the entire random graph to be an ACS is
$`P`$ $`=`$ probability that every row has at least one positive entry $`=`$ \[probability that a row has at least one positive entry$`]^s`$ $`=`$ $`[1(`$probability that every entry of a row is $`0)]^s`$ $`=`$ $`[1(1p^{})^{s1}]^s`$ $`=`$ $`[1(1m^{}/(s1))^{s1}]^s`$.
For large $`s`$ and $`m^{}O(1)`$,
$$P(1e^m^{})^s=e^{\alpha s},$$
(3)
where $`\alpha `$ is positive and $`O(1)`$. At $`n=n_2`$, we find in all our runs that $`l_+(n_2)l^{}`$ is greater than $`s`$ but of order $`s`$, i.e., $`m^{}O(1)`$. Thus dynamical evolution in the model via the ACS mechanism converts a random organization into a highly structured one that is exponentially unlikely to appear by chance. In the displayed run at $`n=n_2`$ the graph had $`117`$ positive links. The probability that a random graph with $`s=100`$ nodes and $`m^{}=1.17`$ would be an ACS is given by (3) to be $`10^{16}`$.
Such a structure would take an exponentially long time to arise by pure chance. The reason it arises inevitably in a short time scale in the present model is the following: a small ACS can appear by chance quite readily, and once appeared, it grows exponentially fast across the graph by the mechanism outlined earlier. The dynamical appearance of such a structure may be regarded as the emergence of ‘organizational order’. The appearance of ‘exponentially unlikely’ structures in the prebiotic context has been a puzzle. The fact that in the present model such structures inevitably form in a short time may be relevant for the origin of life problem.
The self-organization time scale in a prebiotic scenario. We now speculate on a possible application to prebiotic chemical evolution. Imagine the molecular species to be small peptide chains with weak catalytic activity in a prebiotic pond alluded to earlier. The pond periodically receives an influx of new molecular species being randomly generated elsewhere in the environment, through tides, storms or floods. Between these influxes of novelty the pond behaves as a well stirred reactor where the populations of existent molecular species evolve according to (a realistic version of) eq. (1) and reach their attractor configuration. Under the assumption that the present model captures what happens in such a pond, the growth timescale (2) for a highly structured almost fully autocatalytic chemical organization in the pond is $`\tau _g=2/p`$ in units of the graph update time step. In this scenario, the latter time unit corresponds to the periodicity of the influx of new molecular species, hence it ranges from one day (for tides) to one year (for floods). Further, in the present model $`p/2`$ is the probability that a random small peptide will catalyse the production of another , and this has been estimated in : $`p/210^510^{10}`$. With $`p/210^8`$, for example, the time scale for a highly structured chemical organization to grow in the pond would be estimated to be of the order of $`10^6`$ to $`10^8`$ years. It is believed that life originated on Earth in a few hundred million years after the oceans condensed. These considerations suggest that it might be worthwhile to empirically pin down the ‘catalytic probability’ $`p`$ (introduced in ) for peptides, catalytic RNA, lipids, etc., on the one hand, and explore chemically more realistic models on the other.
Catastrophes and recoveries in the network dynamics. After $`n=n_2`$ the character of the network evolution changes again. For the first time the least fit node will be one of the ACS members. Most of the time elimination of the node does not affect the ACS significantly and $`s_1`$ fluctuates between $`s`$ and $`s1`$. Sometimes the least fit node could be a ‘keystone’ species which plays an important organizational role in the network despite its low population. When such a node is eliminated many other nodes can get disconnected from the ACS resulting in large dips in $`s_1`$ and $`\overline{d}`$ and subsequently large fluctuations in $`l_+`$ and $`l_{}`$. These large ‘extinction events’ can be seen in Figure 3. Occasionally the ACS can even be destroyed completely. The system recovers on the time scale $`\tau _g`$ after large extinctions if the ACS is not completely destroyed; if it is and the next few updates obliterate the memory of previous structures in the graph, then again a time on average $`\tau _a`$ elapses before an an ACS arises and the self-organization process begins anew. It may be of interest (especially in ecology, economics, and finance), that network dynamics based on a fitness selection and the ‘incremental’ introduction of novelty, as discussed here, can by itself cause catastrophic events without the presence of large external perturbations.
Discussion
We have described an evolutionary model in which the dynamics of species’ populations (fast variables) and the graph of interactions among them (slow variables) are mutually coupled. The network dynamics displays self-organization seeded by the chance but inevitable appearance of a small cooperative structure, namely, an ACS. In a dynamics that penalizes species for low population performance, the collective cooperativity of the ACS members makes the set relatively robust against disruption. New species that happen to latch on cooperatively to this structure preferentially survive, further enlarging the ACS in the process. Eventually the graph acquires a highly non-random structure. We have discussed the time evolution of quantitative measures of cooperation, interdependence and structure of the network, which capture various aspects of the complexity of the system.
It is noteworthy that collectively replicating ACSs arise even though individual species are not self-replicating. Thus the present mechanism is different from the hypercycle , where a template is needed to produce copies of existing species. Unlike the hypercycle, the ACS is not disrupted by parasites and short-circuits and grows in complexity, as evidenced in all our runs. It can be disrupted, however, when it loses a ‘keystone’ species.
It is also worth mentioning one departure from , in that we find that a fully autocatalytic system (or percolating ACS) is not needed apriori for self-organization. In the present model a small ACS, once formed, typically expands (see also ) and eventually percolates the whole network dynamically. This dynamical process might be relevant for economic takeoff and technological growth in societies.
Acknowledgements. We thank J. D. Farmer and W. C. Saslaw for helpful comments on the manuscript. S. J. acknowledges the Associateship of Abdus Salam International Centre for Theoretical Physics.
|
warning/0005/math0005077.html
|
ar5iv
|
text
|
# Kadison’s similarity problem
Similarity problems and length
by Gilles Pisier<sup>*</sup><sup>*</sup>Supported in part by the NSF and the Texas Advanced Research Program 010366-163. This is an expanded version of a lecture given as keynote speaker at the January 2000 ICMAA in Taiwan.
Texas A&M University
College Station, TX 77843, U. S. A.
and
Université Paris VI
Equipe d’Analyse, Case 186, 75252
Paris Cedex 05, France
Abstract. This is a survey of the author’s recent results on the Kadison and Halmos similarity problems and the closely connected notion of “length” of an operator algebra.
We start by recalling a well known conjecture formulated by Kadison \[Ka\] in 1955.
###### Kadison’s similarity problem
Let $`A`$ be a unital $`C^{}`$-algebra and let $`u:AB(H)`$ ($`H`$ Hilbert) be a unital homomorphism (i.e. we have $`u(1)=1`$ and $`u(ab)=u(a)u(b)`$ $`a,bA`$). Show that if $`u`$ is bounded, then $`u`$ is similar to a $``$-homomorphism, i.e. $`\xi :HH`$ invertible such that $`u_\xi :a\xi ^1u(a)\xi `$ is a $``$-homomorphism ($`=C^{}`$-representation).
Explicitly, the conclusion means that
$$\xi ^1u(a^{})\xi =(\xi ^1u(a)\xi )^{},$$
$`aA`$
when this holds, Kadison calls $`u`$ “orthogonalizable”. Many partial results are known, mainly due to Erik Christensen (\[C1–C4\]) and Uffe Haagerup (\[H1\]). In particular, they established (see \[C3\] and \[H1\]) this conjecture for cyclic homomorphisms, i.e. when $`u`$ admits a cyclic vector $`h`$ in $`H`$ (= a vector $`h`$ such that $`\overline{u(A)h}=H`$) or more generally when $`u`$ admits a finite cyclic set $`h_1,\mathrm{},h_n`$ (so that we have $`\overline{u(A)h_1+\mathrm{}+u(A)h_n}=H`$).
In addition, the Kadison conjecture is known in the following cases:
(i) $`A`$ is commutative.
(ii) $`A`$ is the unitization, denoted by $`\stackrel{~}{𝒦}`$, of the $`C^{}`$-algebra, denoted by $`𝒦`$, of all compact operators on $`\mathrm{}_2`$, or more generally when $`A`$ is nuclear (see \[C2\]).
(iii) $`A=B()`$ or more generally when $`A`$ has no tracial states (see \[H1\]).
(iv) $`A=\stackrel{~}{𝒦}B`$ with $`B`$ arbitrary unital $`C^{}`$-algebra.
(v) $`A`$ is a $`II_1`$-factor with Murray and von Neumann’s property $`\mathrm{\Gamma }`$ (see \[C4\]) for instance when $`A`$ is the so-called hyperfinite $`II_1`$-factor (= infinite tensor product of $`2\times 2`$ matrices with normalized trace).
In sharp contrast, the conjecture is still open when $`A`$ is the reduced $`C^{}`$-algebra of the free group with $`2`$ generators, or even when
$$A=\left(\underset{n1}{}M_n\right)_{\mathrm{}}=\{x=(x_n)x_nM_nsupx_n<\mathrm{}\}.$$
Kadison formulated his conjecture as the $`C^{}`$-algebraic version of a well known problem (at the time of his writing): are all uniformly bounded group representations similar to unitary representations (= unitarizable). While a counterexample to that question was soon found (\[EM\], see also \[P7\] for more recent results on this theme), Kadison’s conjecture remained open. Recently, it became entirely clear that his conjecture is equivalent to another important open question, the derivation problem, itself a crucial problem in the cohomology theory of operator algebras (cf. \[SS\]).
###### Derivation problem
Let $`\pi :AB(H)`$ be a $``$-homomorphism (= representation) on a $`C^{}`$-algebra $`A`$. Let $`\delta :AB(H)`$ be a $`\pi `$-derivation (i.e. $`\delta (ab)=\pi (a)\delta (b)+\delta (a)\pi (b)`$). Show that the boundedness of $`\delta `$ (which is actually automatic here) implies that $`\delta `$ is inner, which means: $`TB(H)`$ such that $`\delta (a)=\pi (a)TT\pi (a)`$ $`aA`$. We set
$$\delta _T(a)=\pi (a)TT\pi (a).$$
The connection between the two problems is simple. Intuitively derivations appear as “infinitesimal generators” for homomorphisms. More elementarily, if $`\delta `$ is as above then
$$u(a)=\left(\begin{array}{cc}\pi (a)& \delta (a)\\ 0& \pi (a)\end{array}\right)$$
is a homomorphism into $`B(HH)=M_2(B(H))`$. Kirchberg \[Ki\] recently proved that the $`C^{}`$-algebras which satisfy the derivation problem are exactly the same as those which satisfy Kadison’s conjecture, but it is still open whether this class is that of all $`C^{}`$-algebras!
We now turn to a key notion to study these problems: “complete boundedness” (see \[Pa1\]).
###### Definition
Let $`EB(H)`$ and $`FB(K)`$ be operator spaces, consider a map
$$\begin{array}{ccc}B(H)& & B(K)\\ & & \\ E& \stackrel{u}{}& F\end{array}$$
For any $`n1`$, let $`M_n(E)=\{(x_{ij})_{ijn}x_{ij}E\}`$ be the space of $`n\times n`$ matrices with entries in $`E`$. In particular we have a natural identification $`M_n(B(H))B(\mathrm{}_2^n(H))`$ where $`\mathrm{}_2^n(H)`$ means $`\underset{n\mathrm{times}}{\underset{}{HH\mathrm{}H}}`$. Thus, we may equip $`M_n(B(H))`$ and a fortiori its subspace $`M_n(E)M_n(B(H))`$ with the norm induced by $`B(\mathrm{}_2^n(H)).`$ Then, for any $`n1`$, the linear map $`u:EF`$ allows to define a linear map $`u_n:M_n(E)M_n(F)`$ by setting
$$u_n\left(\begin{array}{cc}& \mathrm{}\\ \mathrm{}& x_{ij}& \mathrm{}\\ & \mathrm{}\end{array}\right)=\left(\begin{array}{cc}& \mathrm{}\\ \mathrm{}& u(x_{ij})& \mathrm{}\\ & \mathrm{}\end{array}\right).$$
A map $`u:EF`$ is called completely bounded (in short c.b.) if
$$\underset{n1}{sup}u_n_{M_n(E)M_n(F)}<\mathrm{}.$$
We define
$$u_{cb}=\underset{n1}{sup}u_n_{M_n(E)M_n(F)}$$
and we denote by $`cb(E,F)`$ the Banach space of all c.b. maps from $`E`$ into $`F`$ equipped with the c.b. norm.
This concept is fundamental in the currently very actively developed theory of operator spaces, see \[P8\].
###### Theorem 1 (Haagerup 1983, \[H1\])
In the situation of Kadison’s similarity problem, $`u`$ is similar to a $``$-homomorphism iff $`u`$ is c.b. Moreover we have
$$u_{cb}=inf\{\xi ^1\xi u_\xi \mathrm{homomorphism}\}.$$
For derivations, the analogous result is the following.
###### Theorem 2 (Christensen 1977, \[C5\])
In the derivation problem, $`\delta `$ is inner iff $`\delta `$ is c.b. Moreover, we have
$$\delta _{cb}=inf\{2T\delta =\delta _T\}.$$
Vern Paulsen generalized Haagerup’s result to the non-self-adjoint case:
###### Theorem 3 (Paulsen 1984, \[Pa2\])
Let $`A`$ be a unital operator algebra (i.e. we assume only that $`A`$ is a closed subalgebra of $`B()`$ with $`IA`$). Consider again a homomorphism $`u:AB(H)`$. Then $`u_{cb}<\mathrm{}`$ iff $`u`$ is similar to a completely contractive homomorphism, i.e. $`\xi :HH`$ invertible such that $`u_\xi :a\xi ^1u(a)\xi `$ satisfies $`u_\xi _{cb}=1`$. Moreover, we have
$$u_{cb}=inf\{\xi \xi ^1u_\xi _{cb}=1\},$$
and this infimum is attained.
It is easy to see that if $`A`$ is a $`C^{}`$-algebra then
$$(u_{cb}=1)(u=1)(u\text{is a }\text{-homomorphism}).$$
This explains why Theorem 3 contains Theorem 1 as a special case. The preceding result leads us naturally to enlarge our investigation to the non-self-adjoint case as follows.
Generalized Similarity Problem. Which unital operator algebras $`A`$ have the following property denoted by (SP) ?
(SP) Any bounded homomorphism $`u:AB(H)`$ ($`H`$ arbitrary Hilbert space) is c.b.
Loosely speaking, this property (SP) could be described as “automatic complete boundedness” in analogy with the field of automatic continuity for homomorphisms between Banach algebras (see \[DW\]).
Example. The most natural example of a non-self-adjoint algebra is the disc algebra $`A=A(D)`$ which can be described as the completion of the set of all polynomials $`P`$ for the norm
$$P_{\mathrm{}}=sup\{|P(z)|zD\}.$$
We consider $`A(D)`$ as an operator subalgebra of the commutative $`C^{}`$-algebra $`C(D)`$.
Consider a fixed operator $`xB(H)`$. Let
$$u^x:PP(x)B(H)$$
be the homomorphism of evaluation at this fixed $`x`$. Then $`u^x`$ is bounded iff $`x`$ is polynomially bounded, i.e. $`C`$ such that
$$P(x)CP_{\mathrm{}}.$$
$`(1)P`$
On the other hand, it follows from Paulsen’s similarity criterion (Theorem 3 above) that $`u^x`$ is c.b. iff $`x`$ is similar to a contraction (which means $`\xi :HH`$ invertible such that $`\xi ^1x\xi 1`$). Indeed, when $`x1`$, von Neumann’s classical inequality shows that (1) holds and actually also (Sz.-Nagy’s dilation) that $`u^x_{cb}=1`$. Thus it is the same to ask whether $`A(D)`$ satisfies (SP) or to ask whether any polynomially bounded operator $`x`$ is similar to a contraction. This was a well known problem originally formulated by Halmos in a landmark 1970 paper \[Ha\]. We have recently given a counterexample as follows.
###### Theorem 4 (1997, \[P1\])
For any $`c>1`$ there is a unital homomorphism $`u:A(D)B(\mathrm{}_2)`$ (necessarily of the form $`PP(x)`$ for some $`x`$ in $`B(\mathrm{}_2)`$) such that $`uc`$ but $`u_{cb}=\mathrm{}`$.
The proof of the polynomial boundedness was simplified in \[Kis1\] and \[DP\].
Although this solves the somewhat prototypical case of $`A(D)`$, it leaves open the following question: is it true that any uniform algebra (i.e. a unital subalgebra of $`C(K)`$ for some compact set $`K`$) which is proper (i.e. $`A`$ separates the points of $`K`$ and $`AC(K)`$) fails (SP)?
See \[Kis2\] for a partial result on this. Actually, when $`K`$ is a domain in $`{}_{}{}^{_|}\mathrm{C}`$ with at least 2 holes it is already unknown in general whether $`u=1`$ implies $`u_{cb}=1`$! The case of a single hole is covered by \[Ag\]. See also \[DoP\] and \[Pa4\] for more on this theme.
Remarks.
(i) See \[Ku, KuT\] for recent progress on conditions for an operator to be similar to a normal operator.
(ii) The recent paper \[KLM\] contains the following striking example: for any $`c>1`$ there is a power bounded operator on $`\mathrm{}_2`$ which is not similar to any operator with powers bounded by $`c`$. The corresponding statement for polynomial boundedness seems open: given $`c>1`$, is there a polynomially bounded operator which is not similar to any operator polynomially bounded by $`c`$ ?
We now turn to the notion of length which seems closely connected to the generalized similarity problem. The “length” that we have in mind is analogous to the following situation: consider a unital semi-group $`S`$ and a unital generating subset $`BS`$, it is natural to say that $`B`$ generates $`S`$ with length $`d`$ if any $`x`$ in $`S`$ can be written as a product $`x=b_1b_2\mathrm{}b_d`$ with each $`b_i`$ in $`B`$. We will use a somewhat “dual” viewpoint on the “length” based on homomorphisms. Our main idea can be illustrated in a rather transparent way on the above simple model of semi-groups as follows. Assume that $`B`$ generates $`S`$ with length $`d`$. Then any homomorphism $`\pi :SB(H)`$ (i.e. $`\pi (st)=\pi (s)\pi (t)`$ and $`\pi (1)=1`$) which is bounded on $`B`$ with $`\underset{bB}{sup}\pi (b)c`$ must be bounded on the whole of $`S`$ with $`\underset{sS}{sup}\pi (s)c^d`$.
Conversely, assume that we know that for some $`\alpha 0`$ and $`\kappa 0`$, all homomorphisms $`\pi :SB(H)`$ satisfy, for some $`c>1`$, the following implication:
$$\underset{bB}{sup}\pi (b)c\underset{sS}{sup}\pi (s)\kappa c^\alpha .$$
Then it is rather easy to see that $`B`$ necessarily generates $`S`$ with length $`[\alpha ]`$ (integral part of $`\alpha `$), so that we can replace $`\alpha `$ by $`[\alpha ]`$ and $`\kappa `$ by 1.
We called this a “dual” viewpoint because it is reminiscent of the fact that the closed convex hull $`C`$ of a subset $`BE`$ of a Banach space is characterized by the implication
$$\underset{bB}{sup}f(b)1\underset{sC}{sup}f(s)1$$
for all continuous real linear forms $`f`$.
Although this is a wild analogy, we feel that our results on the length are a kind of “nonlinear” analog of this very classical duality principle for convex hulls.
In \[P2\], we study various analogs of this concept of length for operator algebras or even for general Banach algebras. Surprisingly little seems to have been known up to now. We will now review the main results of our papers.
###### Definition
An operator algebra $`AB()`$ is said to be of length $`d`$ if there is a constant $`K`$ such that, for any $`n`$ and any $`x`$ in $`M_n(A)`$, there is an integer $`N=N(n,x)`$ and scalar matrices $`\alpha _0M_{n,N}({}_{}{}^{_|}\mathrm{C})`$, $`\alpha _1M_N({}_{}{}^{_|}\mathrm{C}),\mathrm{},\alpha _{d1}M_N({}_{}{}^{_|}\mathrm{C})`$, $`\alpha _dM_{N,n}({}_{}{}^{_|}\mathrm{C})`$ together with diagonal matrices $`D_1,\mathrm{},D_d`$ in $`M_N(A)`$ satisfying
$$\{\begin{array}{cc}& x=\alpha _0D_1\alpha _1D_2\mathrm{}D_d\alpha _d\hfill \\ & \underset{0}{\overset{d}{}}\alpha _i\underset{1}{\overset{d}{}}D_iKx.\hfill \end{array}$$
We denote by $`\mathrm{}(A)`$ the smallest $`d`$ for which this holds and we call it the “length” of $`A`$ (so that $`A`$ has length $`d`$ is indeed the same as $`\mathrm{}(A)d`$).
Equivalently, we may reformulate this using infinite matrices: if we view as usual $`M_n(A)M_{n+1}(A)`$ via the mapping $`x\left(\begin{array}{cc}x& 0\\ 0& 0\end{array}\right)`$, and if we let $`𝒦(A)=\overline{M_n(A)}`$ be the completion of the union with the natural extension of the norm, then it is easy to check that $`\mathrm{}(A)d`$ iff any $`x`$ in $`𝒦(A)`$ can be written as
$$x=\alpha _0D_1\alpha _1\mathrm{}D_d\alpha _d$$
with $`\alpha _i`$ in $`𝒦({}_{}{}^{_|}\mathrm{C})`$ and $`D_i`$ diagonal in $`𝒦(A)`$. (The constant $`K`$ automatically exists by the open mapping theorem.)
Our central result is as follows.
###### Theorem 5 (1999, \[P2\])
A unital operator algebra $`A`$ satisfies (SP) iff $`\mathrm{}(A)<\mathrm{}`$. Moreover, let
$$d(A)=inf\{\alpha 0Kuu_{cb}Ku^\alpha \}$$
(here of course $`u`$ denotes an arbitrary unital homomorphism from $`A`$ to $`B(H)`$), then
$$d(A)=\mathrm{}(A)$$
and the infimum defining $`d(A)`$ is attained.
Proof that $`d(A)\mathrm{}(A)`$. This is the easy direction. The converse is much more involved. Assume $`\mathrm{}(A)d`$. Consider $`x`$ in $`M_n(A)`$. Recall $`u_{cb}=sup\{u_n(x)_{M_n(B(H))}n1x_{M_n(A)}1\}`$. Consider a factorization of the above form:
$$x=\alpha _0D_1\mathrm{}D_d\alpha _d$$
with $`\alpha _i`$ “scalar” and $`D_i`$ “diagonal”. We have then
$$u_n(x)=\alpha _0u_N(D_1)\alpha _1\mathrm{}u_N(D_d)\alpha _d$$
hence
$$u_n(x)\alpha _iu_N(D_i)$$
but clearly since the $`D_i`$’s are diagonal $`u_N(D_i)uD_i`$ hence
$$u_n(x)u^d\alpha _iD_i$$
which yields (recalling the meaning of $`\mathrm{}(A)d`$)
$$u_{cb}Ku^d.$$
Remark 6. Let us briefly return to the derivation problem. If $`A`$ is a $`C^{}`$-algebra, Kirchberg’s argument in \[Ki\], as slightly improved in \[P2\] shows that if we have
$$\delta _{cb}\alpha \delta $$
$`(2)`$
for all $`\pi `$ and all $`\pi `$-derivations $`\delta :AB(H)`$ then we have $`u_{cb}u^\alpha `$ for all $`u`$ as in Theorem 5. Therefore $`\mathrm{}(A)`$ is less or equal to the integral part of $`\alpha `$. This leads us to conjecture that, in the $`C^{}`$-case, the best possible $`\alpha `$ in (2) is always an integer. Also when $`A`$ is an infinite dimensional $`C^{}`$-algebra we have no example of $`A`$ for which the best $`K`$ such that: $`u`$ $`u_{cb}Ku^{d(A)}`$ is $`>1`$, but we believe such examples exist (we suspect $`A=B(H)\mathrm{}_{\mathrm{}}`$ might be such an example).
It is easy to see that if $`IA`$ is a closed two-sided ideal then $`\mathrm{}(A/I)\mathrm{}(A)`$ and also that $`\mathrm{}(A)\mathrm{max}\{\mathrm{}(I),\mathrm{}(A/I)\}.`$ If $`A`$ is a $`C^{}`$-algebra, we have
$$\mathrm{}(A)=\mathrm{max}\{\mathrm{}(I),\mathrm{}(A/I)\}.$$
To show $`\mathrm{}(I)\mathrm{}(A)`$ we merely use the fact (due to Arveson) see e.g. \[Wa\] that there is a “quasi-central approximate unit” in $`I`$, i.e. a net $`(a_i)`$ in the unit ball of $`I`$ such that for any $`x`$ in $`I`$ we have $`xa_ix`$ and $`a_ixx`$ and moreover (quasi-centrality) $`a_iaaa_i0`$ for any $`a`$ in $`A`$.
In particular, for all finite sets $`A_1,\mathrm{},A_n`$ of operator algebras we have
$$\mathrm{}(A_1\mathrm{}A_n)=\mathrm{max}\{\mathrm{}(A_i)1in\}.$$
The case of infinite direct sums is discussed in \[P6\].
Remark 7. Let $`H=\mathrm{}_2`$. One useful way to apply Theorem 5 is as follows: given a $`d`$-linear map $`w:A^dB(H)`$ we may consider all the possible ways to “factorize” $`w`$ so that there exist linear bounded maps $`v_i:AB(H)`$ such that
$$w(a_1,a_2,\mathrm{},a_d)=v_1(a_1)v_2(a_2)\mathrm{}v_d(a_d).$$
$`(a_1,\mathrm{},a_d)A^d`$
Then we set
$$|w|_d=inf\left\{\underset{i=1}{\overset{d}{}}v_i\right\}$$
where the product runs over all possible ways to “factorize” $`w`$ as above. Then let $`v:AB(H)`$ be a linear map. Assume that we have a finite set of $`d`$-linear maps $`w_p`$ as before such that
$$v(a_1a_2\mathrm{}a_d)=\underset{p}{}w_p(a_1,\mathrm{},a_d).$$
$`a_iA(1id)`$
Then we set
$$v_{[d]}=inf\left\{\underset{p}{}\right|w_p|_d\}$$
where the infimum runs over all possible ways to write as $`v=\underset{p}{}w_p`$. Then if $`\mathrm{}(A)d`$, it is a simple exercise to show that for any linear $`v:AB(H)`$ we have
$$v_{cb}Kv_{[d]}.$$
Thus Theorem 5 allows to strengthen the property (SP): not only homomorphisms are c.b. but also all linear maps $`v`$ for which we have $`v_{[d]}<\mathrm{}`$. Actually, it is possible to show that $`w|w|_d`$ is subadditive but we will not really need this. This norm $`_{[d]}`$ is closely connected with the notion of “multilinear c.b. map” introduced by E. Christensen and A. Sinclair (see \[CS1, CS2\]).
Examples. If $`1<dim(A)<\mathrm{}`$, then $`d(A)=1`$, so from now on we assume $`dim(A)=\mathrm{}`$. We can now review the examples of $`C^{}`$-algebras listed previously:
(i) If $`A`$ is commutative $`d(A)=2`$.
(ii) If $`A=\stackrel{~}{𝒦}`$ or if $`A`$ is nuclear, also $`d(A)=2`$.
(iii) If $`A=B(H)`$, then $`d(A)=3`$.
(iv) If $`A=\stackrel{~}{𝒦}B`$ with $`B`$ arbitrary unital $`C^{}`$-algebra then $`2d(A)3`$.
(v) If $`A`$ is a $`II_1`$-factor with property $`\mathrm{\Gamma }`$ then $`3d(A)5`$.
Notes: (i) and (ii) are due to J. Bunce and E. Christensen (see \[C2\]). In (iii) $`3`$ is proved in \[H1\] while $`3`$ is proved in \[P2\] (see below). (iv) is essentially in \[H1\]. Finally, concerning (v), Christensen proved in \[C4\] that $`d(A)44`$, but the estimate was reduced in \[P6\]. It was also observed in \[P6\] that (as pointed out by N. Ozawa) Anderson’s construction in \[An\] remains valid on any $`II_1`$ factor, thus yielding $`d(A)3`$ for any $`II_1`$ factor $`A`$ by the same argument as in \[P2\].
The class of algebras with $`d(A)(=\mathrm{}(A))`$ equal to 2 is closely related to that of “amenable Banach algebras” (see e.g. \[Pi\]). A von Neumann algebra $`MB()`$ is called amenable (= injective) if there is a projection $`P:B()M`$ with $`P=1`$. It is known that a $`C^{}`$-algebra $`A`$ is nuclear ($``$ amenable by \[H2\]) iff for every representation (= $``$-homomorphism) $`\pi :AB(H)`$, the von Neumann algebra $`M_\pi =\pi (A)^{\prime \prime }`$ generated by $`\pi `$ is amenable (= injective). This motivates the following
###### Definition
A $`C^{}`$-algebra is called semi-nuclear if for any representation $`\pi :AB(H)`$ generating a semi-finite von Neumann algebra $`\pi (A)^{\prime \prime }`$, the generated algebra $`\pi (A)^{\prime \prime }`$ is injective.
###### Theorem 8
(\[P2\]) For a $`C^{}`$-algebra $`A`$, $`d(A)2`$ implies that $`A`$ is semi-nuclear.
It is an open problem whether in general semi-nuclear $``$ nuclear. However, if $`A`$ is either the reduced or the full $`C^{}`$-algebra of a discrete group $`G`$, then
$$A\text{ nuclear }A\text{ semi-nuclear }G\text{ amenable.}$$
The preceding result shows that $`d(B(H))>2`$, since otherwise $`B(H)`$ would be semi-nuclear, which contradicts \[An\]. Hence, we have $`d(B(H))3`$. Actually, using the length $`\mathrm{}(B(H))`$ instead, we can obtain a very simple proof that $`d(B(H))=3`$, as follows.
Proof that $`\mathrm{}(B(H))3`$: This very direct proof comes from \[P6\]. Fix $`n1`$. Let $`W_1`$ and $`W_2`$ be any two $`n\times n`$ unitary matrices such that
$$|W_1(i,j)|=|W_2(i,j)|=n^{1/2}.$$
$`i,j`$
Then, for any $`x`$ in the unit ball of $`M_n(B(H))`$ (with $`dimH=\mathrm{}`$) there are diagonal matrices $`D_1,D_2,D_3`$ also in the unit ball of $`M_n(B(H))`$ such that
$$x=D_1W_1D_2W_2D_3.$$
The proof of this is very simple. Let $`S_i`$, $`i=1,\mathrm{},n`$ be isometries on $`H`$ with orthogonal ranges so that
$$S_i^{}S_j=\delta _{ij}I.$$
$`i,j`$
Then let
$$D_1(i,i)=S_i^{}\text{and}D_3(j,j)=S_j$$
and moreover
$$D_2(k,k)=n\underset{i,j}{}\overline{W_1(i,k)}S_ix_{ij}S_j^{}\overline{W_2(k,j)}.$$
It is an easy exercise (left to the reader) to check the announced properties.
By Theorem 1 and Theorem 5, we have:
###### Proposition 9
The Kadison similarity problem has a positive answer for all unital $`C^{}`$-algebras $`A`$ iff there is an integer $`d_0`$ such that $`\mathrm{}(A)d_0`$ for any $`C^{}`$-algebra $`A`$.
Unfortunately, up to now, the highest known value of $`\mathrm{}(A)`$ for a $`C^{}`$-algebra is 3, but we conjecture that there are examples of arbitrarily large length. However, in the non-self-adjoint case, we have recently been able to prove the following.
###### Theorem 10
For any integer $`d1`$, there is a (non-self-adjoint) operator algebra $`A_d`$ such that
$$\mathrm{}(A_d)=d.$$
Problem. Are there uniform algebras with arbitrarily large finite length?
For uniform algebras no example with $`2<\mathrm{}(A)<\mathrm{}`$ is known. However, it is proved in \[P2\] that any proper uniform algebra $`A`$ must satisfy $`\mathrm{}(A)>2`$. It is also unknown whether there are $`Q`$-algebras (= quotients of uniform algebras) $`A`$ with $`2<\mathrm{}(A)<\mathrm{}`$.
Sketch of proof of Theorem 10. The algebras $`A_d`$ are not at all “pathological”, they are the “obvious” ones: the maximal operator algebras generated by a sequence of contractions $`(x_n)`$ to which we impose the relations
$$x_{n_1}x_{n_2}\mathrm{}x_{n_{d+1}}=0$$
$`(_d)`$
for any $`(d+1)`$-tuple of integers $`(n_1,\mathrm{},n_{d+1})`$. However, while the proof that $`d(A_d)d`$ is then quite easy, the fact that $`\mathrm{}(A_d)>d1`$ has turned out to be much harder to prove. The proof given in \[P4\] uses crucially Gaussian random matrices and specifically a recent difficult estimate due to Haagerup and Thorbjørnsen \[HT\]. We will only give a brief description of the argument from \[P4\]. Let $`P=P(X_1,X_2,\mathrm{})`$ be a polynomial of degree $`d`$ in non-commuting (formal) variables $`X_1,X_2,\mathrm{}`$ . We introduce the norm
$$P_{A_d}=sup\{P(x_1,x_2,\mathrm{})\}$$
$`(3)`$
where the supremum runs over all sequences of contractions in $`B(\mathrm{}_2)`$ satisfying $`(_d)`$. It is easy to check that this is a norm of the set of polynomials $`P`$ with degree $`d`$. We denote by $`A_d`$ the completion of the set of $`P`$’s equipped with this norm. Clearly, this defines an operator algebra naturally embedded into $`\underset{x}{}B(H_x)`$ where $`H_x=\mathrm{}_2`$ where $`x=(x_n)_{n1}`$ runs over the set of all possible sequences of contractions satisfying $`(_d)`$. In order to show that $`\mathrm{}(A_d)>d1`$, the next lemma is crucial. To state it we first need a specific notation.
Notation. Let $`H=\mathrm{}_2`$. Let $`m1`$ and $`d1`$ be fixed integers. We will denote by $`C(m,d)`$ the smallest constant $`C`$ for which the following holds: if $`\{x_ii[m]^d\}`$ in $`B(H)`$ satisfies
$$\lambda _ix_i\underset{\genfrac{}{}{0pt}{}{X_iB\left(H\right)}{X_i1}}{sup}\left\{\lambda _iX_{i_1}X_{i_2}\mathrm{}X_{i_d}\right\}$$
$`(4)\lambda _i{}_{}{}^{_|}\mathrm{C}`$
then $`\widehat{x}_kB(H)`$, $`(1km)`$ with $`\widehat{x}_k1`$ such that
$$x_i=C\widehat{x}_{i_1}\widehat{x}_{i_2}\mathrm{}\widehat{x}_{i_d}.$$
$`(4)^{}i[m]^d`$
###### Lemma 11
For any $`m1`$ and $`d1`$, we have
$$\delta _dm^{\frac{d1}{2}}C(m,d)m^{\frac{d1}{2}}$$
where $`\delta _d>0`$ is a constant independent of $`m`$.
Example. In the case $`d=2`$, this means the following: if $`x_{ij}B(H)`$ $`(i,j=1,\mathrm{},m)`$ satisfy
$$\underset{ijm}{}\lambda _{ij}x_{ij}\underset{X_i1}{sup}\lambda _{ij}X_iX_j$$
$`\lambda _{ij}{}_{}{}^{_|}\mathrm{C}`$
then $`x_{ij}`$ can be factorized as
$$x_{ij}=C\widehat{x}_i\widehat{x}_j\text{with}\widehat{x}_i1$$
$`(5)`$
but in general the best possible $`C`$ will be $`\sqrt{m}`$. This case is rather easy to prove given the state of the art. However, already the case $`d=3`$ is more delicate, and as we already mentioned the case of an arbitrary $`d`$ requires the upper estimates given in \[HT\] which are highly non-trivial. An easier proof of the lower bound (which is the difficult part) in Lemma 11 would be most welcome.
Remark 12. Given $`\{x_ii[m]^d\}`$ in $`B(H)`$ satisfying (4), we can define a linear map
$$v:A_dB(H)$$
by setting $`v(X_{i_1}X_{i_2}\mathrm{}X_{i_d})=x_{i_1i_2\mathrm{}i_d}`$ with $`1i_1,i_2,\mathrm{},i_dm`$ and $`v(X_{i_1}X_{i_2}\mathrm{}X_{i_k})=0`$ in all other cases. Then it can be shown, using the factorization of multilinear cb maps of Christensen-Sinclair and Paulsen-Smith (see \[P4\]) that $`v_{cb}`$ is equal to the smallest constant $`C`$ such that (4) holds.
We now wish to sketch how Lemma 11 is used to prove that $`\mathrm{}(A_d)>d1`$. To lighten the exposition, we will restrict to the simplest case: $`d=3`$. So we will show that Lemma 11 implies $`\mathrm{}(A_3)>2`$. We will show that if $`\mathrm{}(A_3)2`$ then $`C(m,d)K\sqrt{m}`$ for some $`K`$, but this will contradict Lemma 11 for $`d=3`$ since $`(d1)/2=1>1/2`$, whence the conclusion that $`\mathrm{}(A_3)>2`$.
Now assume $`\mathrm{}(A_3)2`$. Let $`\{x_{i_1i_2i_3}i[m]^3\}`$ be as in the definition of $`C(m,d)`$ for $`d=3`$. For convenience, we extend the function $`(i_1,i_2,i_3)x_{i_1i_2i_3}`$ to the whole of $`\mathrm{I}\mathrm{N}^3`$ by setting it equal to zero outside $`[1,\mathrm{},m]^3`$.
We will use Remark 7.
Let $`v:A_3B(H)`$ be the linear map defined by $`v(1)=0`$, $`v(X_i)=0`$, $`v(X_{i_1}X_{i_2})=0`$ and finally:
$$v(X_{i_1}X_{i_2}X_{i_3})=x_{i_1i_2i_3}.$$
It is easy to see using (3) and (4) that
$$v1.$$
We claim that (4) implies (with the notation of Remark 7)
$$v_{[2]}2+2\sqrt{m}.$$
We will use the following notation: we consider the disjoint union
$$\mathrm{\Omega }=\varphi \mathrm{I}\mathrm{N}\mathrm{I}\mathrm{N}^2\mathrm{I}\mathrm{N}^3,$$
and we set
$$\begin{array}{cc}\hfill X^\varphi & =1\hfill \\ \hfill X^i& =X_i\text{if}i\mathrm{I}\mathrm{N}\hfill \\ \hfill X^{ij}& =X_iX_j\text{if}(ij)\mathrm{I}\mathrm{N}^2\hfill \\ \hfill X^{ijk}& =X_iX_jX_k\text{if}(ijk)\mathrm{I}\mathrm{N}^3.\hfill \end{array}$$
For $`i\mathrm{\Omega }`$ we set $`|i|=0`$ if $`i=\varphi `$, and $`|i|=k`$ if $`i\mathrm{I}\mathrm{N}^k`$.
With this notation any polynomial $`P`$ in $`A_3`$ can be written as a finite sum
$$P=\underset{i\mathrm{\Omega }}{}\lambda _i(P)X^i$$
with $`\lambda _i(P){}_{}{}^{_|}\mathrm{C}`$. We have then $`v(X^i)=x_i`$ for all $`i`$ in $`\mathrm{\Omega }`$, hence $`P_1,P_2A_3`$
$$v(P_1P_2)=\underset{i,j\mathrm{\Omega }}{}\lambda _i(P_1)\lambda _j(P_2)x_{ij}$$
where $`ij`$ denotes now the multi-index of length $`6`$ obtained by putting $`j`$ after $`i`$. We set $`|ij|=|i|+|j|`$. But since $`x_{ij}=0`$ unless $`|i|+|j|=3`$ we find a decomposition of $`v`$ as follows:
$$v(P_1P_2)=\underset{(\alpha \beta )J}{}w_{\alpha \beta }(P_1,P_2)$$
$`(6)`$
where the sum runs over the set $`J`$ of all pairs $`(\alpha \beta )`$ in such that $`\alpha +\beta =3`$, and where $`w_{\alpha \beta }`$ are bilinear forms on $`A_3\times A_3`$ defined by setting:
$$w_{\alpha \beta }(P_1,P_2)=\underset{\genfrac{}{}{0pt}{}{\left|i\right|=\alpha }{\left|j\right|=\beta }}{}\lambda _i(P_1)\lambda _j(P_2)x_{ij}.$$
Using (4) it is easy to see that if $`(\alpha \beta )`$ is either $`(30)`$ or $`(03)`$ then with the notation of Remark 7
$$|w_{\alpha \beta }|_21.$$
The remaining possibilities in $`J`$ are only (2 1) and (1 2). But if $`(\alpha \beta )=(12)`$ for instance we can write
$$w_{\alpha \beta }(P_1,P_2)=\left(\underset{|i|=1}{}\lambda _i(P_1)e_{1i}I\right)\left(\underset{k=1}{\overset{m}{}}e_{k1}\underset{|j|=2}{}\lambda _j(P_2)x_{kj}\right)$$
(here we identify $`B(H)`$ with $`B(H)\overline{}B(H)`$ and denote by $`(e_{ki})`$ the standard matrix units in $`B(H)`$). Using this, one can check rather easily that if $`(\alpha \beta )=(21)`$ or $`(12)`$ then
$$|w_{\alpha \beta }|_2\sqrt{m}.$$
Thus using (6) we obtain our claim that
$$v_{[2]}2+2\sqrt{m}.$$
Then if we assume $`\mathrm{}(A_3)2`$, Remark 7 ensures that
$$v_{cb}K(2+2\sqrt{m}).$$
Now by Remark 12 this implies that $`\{x_ii[m]^3\}`$ satisfies (4) with $`CK(2+\sqrt{m})`$. Thus we conclude that $`C(m,3)K(2+\sqrt{m})`$ but this obviously contradicts Lemma 11 with $`d=3`$. Thus we have shown, by this contradiction, that $`\mathrm{}(A_3)>2`$.
The notion of length is quite natural in the more general context of a Banach algebra $`B`$ generated by a family of subalgebras $`B_iB`$ $`(iI)`$. For simplicity, we will restrict ourselves to the case of a pair of subalgebras $`B_1B`$, $`B_2B`$. In this case, we say that $`B_1,B_2`$ generate $`B`$ with length $`d`$ if there is a bounded subset of the union $`CB_1B_2`$ such that every $`x`$ in the unit ball of $`B`$ belongs to the closed convex hull of the union $`_{j=1}^dC^j`$ where
$$C^j=\{x_1x_2\mathrm{}x_jx_kCk=1,\mathrm{},j.\}$$
Assuming that this holds, let $`u:B\beta `$ be a continuous homomorphism into another Banach algebra $`\beta `$. It is then easy to check that
$$uK\underset{j=1}{\overset{d}{}}\mathrm{max}\{u_{|B_1},u_{|B_2}\}^j$$
where $`K`$ is a constant (depending only on $`d`$ and the size of the subset $`C`$).
Thus if $`B,\beta `$ and $`u`$ are all unital we obtain (since all the norms are now $`1`$)
$$udK\mathrm{max}\{u_{|B_1},u_{|B_2}\}^d.$$
In the converse direction, assuming again $`B_1,B_2`$ and $`B`$ all unital, let $`\text{alg}(B_1,B_2)`$ denote the algebra generated by $`B_1`$ and $`B_2`$, which we assume is dense in $`B`$. Assume that every unital homomorphism $`u:\text{alg}(B_1,B_2)\beta `$ into an arbitrary unital Banach algebra $`\beta `$ such that $`u_{|B_1}<\mathrm{}`$ and $`u_{|B_2}<\mathrm{}`$ is actually bounded and satisfies
$$uK(\mathrm{max}\{u_{|B_1},u_{|B_2}\})^\alpha $$
where $`K`$ and $`\alpha 0`$ are independent of $`u`$ and $`\beta `$.
Then it follows (see \[P2, §8\]) that $`B_1,B_2`$ generate $`B`$ with length at most equal to the integral part of $`\alpha `$. For example, let $`A`$ be a unital operator algebra, and let $`B=𝒦(A)`$. We may consider the subalgebra $`B_1B`$ formed of all the diagonal matrices (viewing the elements of $`𝒦(A)`$ as bi-infinite matrices with coefficients in $`A`$) and we let $`B_2=𝒦({}_{}{}^{_|}\mathrm{C})`$.
It is then easy to check that $`\mathrm{}(A)d`$ implies that $`B_1,B_2`$ generate $`B`$ with length $`2d+1`$. Conversely, if $`B_1,B_2`$ generate $`B`$ with length $`m`$, then $`\mathrm{}(A)\left[\frac{m+1}{2}\right]`$.
Remark. The slight discrepancy appearing here comes from the fact that in the products appearing in the subset $`C^d`$ we do not specify that the first term of the product must lie in $`B_2`$ or $`B_1`$ while in the corresponding definition of $`\mathrm{}(A)`$ the analogous term must be in $`B_2`$. This difficulty can be circumvented: one should then consider homomorphisms $`u:\text{alg}(B_1,B_2)\beta `$ such that $`u_{|B_2}=1`$ and study the inequality $`uKu_{|B_1}^\alpha `$. See \[P3\] for more variations on this theme.
The case study of $`\mathrm{}(A)`$ suggests to examine many other examples of the same kind, for instance the pair $`B_1=𝒦(A_1)`$, $`B_2=𝒦(A_2)`$ where $`A_1A`$, $`A_2A`$ are two closed subalgebras. In particular, we may consider the case where $`A`$ is the maximal tensor product of two unital $`C^{}`$-algebras $`C_1,C_2`$: namely we take $`A=C_1_{\mathrm{max}}C_2`$ with $`A_1=C_11`$ and $`A_2=1C_2`$. All these cases are studied in \[P3\], to which we refer the reader for several illustrating examples and more information.
References
\[Ag\] J. Agler. Rational dilation on an annulus. Ann. of Math. 121 (1985) 537–563.
\[An\] J. Anderson. Extreme points in sets of positive linear maps on $`()`$. J. Funct. Anal. 31 (1979) 195-217.
\[BP\] D. Blecher and V. Paulsen. Explicit construction of universal operator algebras and applications to polynomial factorization. Proc. Amer. Math. Soc. 112 (1991) 839-850.
\[C1\] E. Christensen. Extensions of derivations. J. Funct. Anal. 27 (1978) 234–247
\[C2\] $`\underset{¯}{}`$. Extensions of derivations II Math. Scand. 50 (1982) 111–122.
\[C3\] $`\underset{¯}{}`$. On non self adjoint representations of operator algebras Amer. J. Math. 103 (1981) 817-834.
\[C4\] $`\underset{¯}{}`$. Similarities of $`II_1`$ factors with property $`\mathrm{\Gamma }`$. Journal Operator Theory 15 (1986) 281-288.
\[C5\] $`\underset{¯}{}`$. Perturbation of operator algebras II. Indiana Math. J. 26 (1977), 891–904.
\[CS1\] E. Christensen and A. Sinclair. Representations of completely bounded multilinear operators. J. Funct. Anal. 72 (1987) 151-181.
\[CS2\] $`\underset{¯}{}`$. A survey of completely bounded operators. Bull. London Math. Soc. 21 (1989) 417-448.
\[DP\] K. Davidson and V. Paulsen. On polynomially bounded operators, J. für die reine und angewandte Math. 487 (1997) 153-170.
\[DoP\] R. Douglas and V. Paulsen. Hilbert modules over function algebras. Pitman Longman 1989.
\[DW\] H. G. Dales and W. H. Woodin. An introduction to independence for analysts. London Mathematical Society Lecture Note Series, 115. Cambridge University Press, Cambridge-New York, 1987.
\[EM\] L. Ehrenpreis and F.I. Mautner. Uniformly bounded representations of groups Proc. Nat. Acad. Sc. U.S.A. 41 (1955) 231-233.
\[H1\] U. Haagerup. Solution of the similarity problem for cyclic representations of $`C^{}`$-algebras. Annals of Math. 118 (1983), 215-240.
\[H2\] $`\underset{¯}{}`$. All nuclear $`C^{}`$-algebras are amenable. Invent. Math. 74 (1983) ) 305–319.
\[Ha\] P. Halmos. Ten problems in Hilbert space. Bull Amer. Math. Soc. (1970)
\[HT\] U. Haagerup and S. Thorbjørnsen. Random matrices and $`K`$-theory for exact $`C^{}`$-algebras. Documenta Math. 4 (1999) 341-450.
\[Ka\] R. Kadison On the orthogonalization of operator representations. Amer. J. Math. 77 (1955) 600-620.
\[Ki\] E. Kirchberg. The derivation and the similarity problem are equivalent. J. Operator Th. 36 (1996) 59-62.
\[Kis1\] S. Kislyakov. Operators that are (dis)similar to a contraction: Pisier’s counterexample in terms of singular integrals. Zap. Nauchn. Semin. S.- Peterburg. Otdel. Mat. Inst. Steklov (POMI).
\[Kis2\] $`\underset{¯}{}`$. Similarity problem for certain martingale uniform algebras. Preprint.
\[KLM\] N. Kalton and C. Le Merdy. Solution of a problem of Peller concerning similarity. J. Op. Theory. To appear.
\[Ku\] S. Kupin. Similarité à un opérateur normal et certains problèmes d’interpolation. Thèse, Université de Bordeaux I, (3 avril 2000).
\[KuT\] S. Kupin and S. Treil. Linear resolvent growth of a weak contraction does not imply its similarity to a normal operator. Preprint 2000, to appear.
\[Pa1\] V. Paulsen. Completely bounded maps and dilations. Pitman Research Notes in Math. 146, Longman, Wiley, New York, 1986.
\[Pa2\] $`\underset{¯}{}`$. Completely bounded homomorphisms of operator algebras. Proc. Amer. Math. Soc. 92 (1984) 225-228.
\[Pa3\] $`\underset{¯}{}`$. Every completely polynomially bounded operator is similar to a contraction. J. Funct. Anal. 55 (1984) 1-17.
\[Pa4\] $`\underset{¯}{}`$. Toward a theory of $`K`$-spectral sets. Surveys of some recent results in operator theory, Vol. I, 221–240, Pitman Res. Notes Math. Ser., 171, Longman Sci. Tech., Harlow, 1988.
\[P1\] G. Pisier. A polynomially bounded operator on Hilbert space which is not similar to a contraction. Journal Amer. Math. Soc. 10 (1997) 351-369.
\[P2\] $`\underset{¯}{}`$. The similarity degree of an operator algebra. St. Petersburg Math. J. 10 (1999) 103-146.
\[P3\] $`\underset{¯}{}`$. Joint similarity problems and the generation of operator algebras with bounded length. Integr. Equ. Op. Th. 31 (1998) 353-370.
\[P4\] $`\underset{¯}{}`$. The similarity degree of an operator algebra II. Math. Zeit. (2000) To appear.
\[P5\] $`\underset{¯}{}`$. Similarity problems and completely bounded maps. Springer Lecture notes 1618 (1995).
\[P6\] $`\underset{¯}{}`$. Remarks on the similarity degree of an operator algebra. Preprint.
\[P7\] $`\underset{¯}{}`$. Are unitarizable groups amenable ? preprint, 1999.
\[P8\] $`\underset{¯}{}`$. Operator spaces and similarity problems, Proc. Berlin ICM 1998, Doc. Math. Extra Vol. I, 429-452.
\[Pi\] J.P. Pier. Amenable Banach algebras. Pitman, Longman 1988.
\[SS\] A. M. Sinclair and R.R. Smith. Hochshild cohomology of von Neumann algebras. L. M. S. Lecture Notes series. Cambridge University Press, Cambridge 1995.
\[Wa\] S. Wassermann. Exact $`C^{}`$-algebras and related topics. Lecture Notes Series, Seoul Nat. Univ. (1994).
Texas A&M University
College Station, TX 77843, U. S. A.
and
Université Paris VI
Equipe d’Analyse, Case 186,
75252 Paris Cedex 05, France
|
warning/0005/astro-ph0005166.html
|
ar5iv
|
text
|
# 1E 0117.2-2837 and the NLR of Narrow-Line Seyfert 1 galaxies
## 1 Introduction
Narrow Line Seyfert 1 (NLS1) galaxies have recently received much attention due to their unusual optical–X-ray properties which are not yet well understood. Photoionization models of the circum-nuclear emission/absorption regions allow us to investigate scenarios to explain the main characteristics of NLS1s, i.e., (i) extremely steep X-ray spectra within the ROSAT energy band, (ii) narrow (FWHM $`<`$ 2000 km/s) Balmer lines and strong FeII emission, and (iii) weak forbidden lines except for some relatively strong high-ionization iron lines. Here, we concentrate on (iii); for a discussion of (i) and (ii) see Komossa & Fink (1997a), and Komossa & Meerschweinchen (2000). In particular, we study the influence of different EUV - soft-X-ray spectral shapes (a giant soft excess, a steep X-ray powerlaw, the presence of a warm absorber) and NLR cloud properties (density, abundances, and distance from the nucleus) on the predicted optical emission-line ratios. First results of this study were presented by Komossa & Janek (1999).
## 2 1E 0117.2-2837
1E 0117.2-2837 (QSO 0117-2837) was discovered as an X-ray source by Einstein and is at a redshift of $`z`$=0.347 (Stocke et al. 1991, Grupe et al. 1999). Its X-ray spectrum is extremely steep (see Komossa & Meerschweinchen 2000 for a detailed X-ray analysis of this source). We have obtained new optical spectra of 1E 0117.2-2837 with the ESO 1.52 m telescope at LaSilla, in September 1995. The optical spectrum reveals several signs of a NLS1 galaxy (we do not distinguish between NL Seyferts and NL quasars, here): weak \[OIII\]$`\lambda `$5007 emission and strong FeII complexes (Fig. 1). After subtraction of the FeII spectrum (see Grupe et al. 1999 for details) we derive FWHM=2100$`\pm 100`$ km/s, FWHM<sub>\[OIII\]</sub>=820$`\pm 150`$ km/s (based on single-component Gaussian fits to the emission lines; this leaves some broad wings as residuals), and the ratio \[OIII\]/H$`\beta `$=0.056.
## 3 The NLR of NLS1 galaxies: results from photoionization modeling
The narrow emission lines, i.e. those originating from the narrow line region (NLR), like \[OIII\]$`\lambda `$5007 and \[OI\]$`\lambda `$6300, are rather weak in 1E 0117.2-2837 and in NLS1 galaxies in general. Occasionally, however, fairly strong high-ionization iron lines are present. We investigate several models to explain these observations, starting with the assumption that the NLR is ‘normal’ (i.e. a typical type 1 Seyfert, as far as distance from the nucleus, gas density and covering factor are concerned). All photoionization calculations were carried out using the Cloudy code (Ferland, 1993).
In the first step, non-solar metal abundances were assumed. Over abundant metals (with respect to the solar value) were shown to delay the complete removal of a BLR multiphase equilibrium (Komossa & Meerschweinchen 2000). Due to their rather strong influence on the cooling, metals, if overabundant, can lead to weaker optical line emission. However, we find the effect to be insufficient to explain the observed line intensities.
As shown in Komossa & Schulz (1997), the weak \[OIII\]$`\lambda `$5007 domain of the line correlations in the usual diagnostic diagrams of Seyfert 2 galaxies can be explained by very steep EUV continua with $`\alpha _{\mathrm{uv}\mathrm{x}}`$ $``$ –2.5. Although, e.g., the NLS1 RX J1225.7+2055 indeed exhibits a rather steep EUV spectrum (determined by a powerlaw connection between the flux at the Lyman limit and 0.1 keV), that of RX J1239.3+2431 is very flat (Greiner et al. 1996).
Placing warm absorbing material along the line of sight to the NLR would make the latter see a continuum that is only modified in the soft X-ray region, with negligible influence on the line emission. The same holds for an intrinsically steep X-ray powerlaw, which only leads to a slight weakening of the low-ionization lines.
In cases where a warm absorber is present, the high-ionization iron lines (\[FeX\] and higher) can be produced within the warm gas itself (see Komossa & Fink 1997a,b,c for details). However, no one-to-one match between the observed coronal lines in the NLS1 NGC 4051 and those predicted to arise from the X-ray warm absorber in this galaxy was found (Komossa & Fink 1997a, and these proceedings), suggesting that in general, the coronal line and warm absorber regions are separate components.
In order to assess the influence of a strong EUV - soft-X-ray excess on optical line emission, we have calculated a sequence of models with an underlying mean Seyfert continuum plus a black body of varying temperature, for a range of densities and distances of the NLR gas from the central continuum source. Although the contribution of a hot bump-component can considerably strengthen the high-ionization iron lines (eg. \[FeX\]$`\lambda `$6374, \[FeVII\]$`\lambda `$6087 and \[FeXIV\]$`\lambda `$5303), reflecting the fact that their ionization potentials are at soft X-ray energies, these models overpredict the \[OIII\] emission.
We conclude that the weakness of forbidden lines in NLS1s must be due to an overall lower emissivity of the NLR. If this is caused by shielding of the NLR from ionising photons, the model must avoid boosting the low-ionization lines like \[OII\]. More likely, the region is gas poor, i.e. less of the impinging photons can be reprocessed into line emission. ‘Normal’ line ratios would then result in weak forbidden lines being undetected. For those objects with strong high-ionization (iron) lines, models favor the dominance of low-density gas and small distances to the ionizing source.
Acknowledgements: We thank Gary Ferland for providing Cloudy. Preprints of this and related papers can be retrieved from our webpage at
http://www.xray.mpe.mpg.de/$``$skomossa/
|
warning/0005/physics0005034.html
|
ar5iv
|
text
|
# Relativity as the quantum mechanics of space-time measurements
## Relativity as the quantum mechanics of space-time measurements
Richard Lieu
Department of Physics, University of Alabama, Huntsville, AL 35899, U.S.A.
email: lieur@cspar.uah.edu
Abstract
Can a simple microscopic model of space and time demonstrate Special Relativity as the macroscopic (aggregate) behavior of an ensemble ? The question will be investigated in three parts. First, it is shown that the Lorentz transformation formally stems from the First Relativity Postulate (FRP) alone if space-time quantization is a fundamental law of physics which must be included as part of the Postulate. An important corollary, however, is that when measuring devices which carry the basic units of lengths and time (e.g. a clock ticking every time quantum) are ‘moving’ uniformly, they appear to be measuring with larger units. Secondly, such an apparent increase in the sizes of the quanta can be attributed to extra fluctuations associated with motion, which are precisely described in terms of a thermally agitated harmonic oscillator by using a temperature parameter. This provides a stringent constraint on the microscopic properties of flat space-time: it is an array of quantized oscillators. Thirdly, since the foregoing development would suggest that the space-time array of an accelerated frame cannot be in thermal equilibrium, (i.e. it will have a distribution of temperatures), the approach is applied to the case of acceleration by the field of any point object, which corresponds to a temperature ‘spike’ in the array. It is shown that the outward transport of energy by phonon conduction implies an inverse-square law of force at low speeds, and the full Schwarzschild metric at high speeds. A prediction of the new theory is that when two inertial observers move too fast relative to each other, or when fields are too strong, anharmonic corrections will modify effects like time dilation, and will lead to asymmetries which implies that the FRP may not be sustainable in this extreme limit.
In contemporary physics the FRP is usually viewed as a portrait of the complete symmetry between inertial observers. Specifically all such observers experience the same laws of physics$`[1]`$. This Postulate, though simply stated, is powerful because application of it to new physical principles could lead to important consequences. The supreme example is when Maxwell’s equation of electromagnetic wave (i.e. light) propagation finally became recognized as a universal law of nature$`[2]`$. In order that the equation be invariant with respect to all inertial frames, their coordinates cannot be related by the Galilean transformation (which assumes space and time are absolute), as first realized by Lorentz$`[3]`$. A way of arriving at the correct transformation simply and logically, as suggested by Einstein$`[4]`$, is to introduce the Second Relativity Postulate (hereafter the SRP). It states that the speed of light is a universal constant, independent of (a) inertial frames, and (b) the motion of the emission source in each frame. Since these two independences, especially (b), do not arise unambiguously from the assertion that Maxwell’s equations must be formally invariant in all frames, the need for an explicit declaration as a postulate is inevitable. Thus, although Einstein did fruitfully integrate Maxwell’s theory with the FRP, it was necessary for him to enlist a separate postulate. Does this mean the statement of symmetry provided by the FRP, though very elegant, only plays the role of compelling and guiding us to invoke further postulates everytime we have to deal with a new physical law ?
Here I suggest another possibility, viz. that the reason for the problems of the FRP is because, unlike Newton’s laws, neither Maxwell’s theory nor Einstein’s SRP operate on a sufficiently fundamental level. The missing piece is simply that microscopically space and time are quantized: there exists naturally imposed units of distances and time intervals as characterized by the parameters $`(x_o,t_o)`$, the minimum uncertainties in one’s ability to measure the coordinates $`(x,t)`$ of an event, which cannot be surpassed irrespective of an instrument’s accuracy. When incorporated with the FRP, the Postulate now reads: the laws of physics as observed through a system of quantized space-time, the latter a ‘grid’ which inevitably controls the outcomes of measurements, is the same for all inertial observers. Thus, if a frame transformation arbitrarily changes $`x_o`$ and $`t_o`$, the reference ‘grid’ will in general be distorted; but if at least the ratio of the two quantities remains constant, the ‘grid’ will be enlarged or reduced proportionately, and so long as no physics exists which depends on absolute scales, such a transformation will still preserve the symmetry between inertial frames. It is worth investigating whether this generalized FRP alone will lead to a unique set of equations which connect the coordinates of different frames, Before doing so, however, I designate the ratio $`x_o/t_o`$ a universal constant $`v_o`$ which has the dimension of a speed.
Let us first examine the Galilean transformation $`G`$. The event data of an inertial observer $`\mathrm{\Sigma }`$ are $`(x,t)`$ with accuracies $`(x_o,t_o)`$. According to $`G`$, such data would appear differently to another observer $`\mathrm{\Sigma }^{}`$ who moves relative to $`\mathrm{\Sigma }`$ with velocity $`𝐯=(v,0,0)`$. More precisely this observer notices that if he were to repeat the measurement following every step adopted by $`\mathrm{\Sigma }`$, his results would be $`(x^{},t^{})`$ with accuracies $`(x_o^{},t_o^{})`$, where $`x^{}=xvt`$, $`t^{}=t`$; and $`x_o^{}=x_o\sqrt{1+v^2/v_o^2}`$, $`t_o^{}=t_o`$. Based on the above perception, $`\mathrm{\Sigma }^{}`$ constructs a physical model of his world view. Would this be the same model as the one derived from experiments performed directly by himself ? Such a question is indeed valid and pertinent if we interpret the forementioned accuracies as limiting ones which form basic units of measurements. The answer is ‘no’ because of two problems: (a) the measurement ‘grid’ is distorted because the ratio $`x_o/t_o`$ changes from $`v_o`$ to $`v_o^{}=x_o^{}/t_o^{}=v_o\sqrt{1+v^2/v_o^2}>v_o`$; (b) a consequence of space-time quantization is that measured distances transform if and only if the unit of distance does (the same applies to time intervals); in this regard $`G`$, which clearly preserves distances but changes $`x_o`$ to $`x_o^{}>x_o`$, is not even a self-consistent transformation.
I proceed to seek a frame transformation which conserves the ratio $`x_o/t_o`$. We start with $`\mathrm{\Sigma }`$ and $`\mathrm{\Sigma }^{}`$ having their spatial (cartesian) coordinate axes parallel to each other, and common space-time origin. The macroscopic homogeneity and isotropy of space-time, a corollary of the FRP, immediately implies a linear relationship between the two sets of 4-D coordinates with $`y`$ and $`z`$ separated from the rest, i.e.
$`t^{}=r(v)[ts(v)x],x^{}=q(v)[xp(v)t],y^{}=y,z^{}=z`$
where the functions $`q`$, $`r`$, and $`s`$ can depend only on $`v`$. By taking into account the fact that an object stationary relative to $`\mathrm{\Sigma }^{}`$ moves relative to $`\mathrm{\Sigma }`$ with speed $`v`$ along $`+x`$, one easily deduces $`p(v)=v`$, so that
$$t^{}=r(v)[ts(v)x],x^{}=q(v)(xvt),y^{}=y,z^{}=z$$
(1)
Next, since the combined operations $`𝐫𝐫^{}`$, $`tt^{}`$ and $`vv`$ leave the situation unchanged, and since $`q`$ and $`r`$ are positive definite scale factors which cannot be sensitive to the sign of $`v`$ (else there would be a preference between $`+x`$ and $`x`$), the inverse transformation reads:
$$t=r(v)[t^{}s(v)x^{}],x=q(v)(x^{}+vt^{}),y=y^{},z=z^{}$$
(2)
If in the x-transformation part of equ (1) one substitutes the expressions for $`x`$ and $`t`$ from equ (2), and the result were to hold for all times, the coefficient of $`t^{}`$ must vanish. This means:
$$q(v)=r(v)$$
(3)
The development thus far has been based on general considerations, the functions $`q(v)`$ and $`s(v)`$ remain arbitrary. Einstein$`[4]`$ determined them by applying the SRP: the speed of a light signal as measured by the two observers must be equal, i.e. $`\mathrm{\Delta }x/\mathrm{\Delta }t=\mathrm{\Delta }x^{}/\mathrm{\Delta }t^{}=c`$, where, from Equs (1) and (3), $`\mathrm{\Delta }t^{}=q(\mathrm{\Delta }ts\mathrm{\Delta }x)`$ and $`\mathrm{\Delta }x^{}=q(\mathrm{\Delta }xv\mathrm{\Delta }t)`$. This leads to $`s=v/c^2`$, and the criterion of a consistently invertible transformation restricts $`q`$ to:
$$q(v)=\frac{1}{\sqrt{1\frac{v^2}{c^2}}},$$
(4)
What happens if, instead of using the SRP, we apply the constancy of $`x_o/t_o=v_o`$ ? Here the two approaches differ, because unlike $`\mathrm{\Delta }x`$, the uncertainties of measurements, $`\delta x`$, do not transform linearly. If $`\mathrm{\Sigma }`$ measures $`(x,t)`$ independently to accuracies $`\delta x`$ and $`\delta t`$ then, defining the primed quantities as before, we have:
$$(\delta x^{})^2=q^2[(\delta x)^2+v^2(\delta t)^2];(\delta t^{})^2=q^2[(\delta t)^2+s^2(\delta x)^2]$$
(5)
For the limiting (quantum) uncertainties the above formulae remain valid with the substitutions:
$$\delta x=x_o,\delta x^{}=x_o^{},\delta t=t_o,\delta t^{}=t_o^{}$$
(6)
The requirement $`\delta x^{}/\delta t^{}=\delta x/\delta t=v_o`$ implies that $`s=\pm v/v_o^2`$, i.e. two classes of transformation are permitted. Now the criterion of invertibility restricts $`q`$, in the case of the first solution, to the form:
$$q(v)=\frac{1}{\sqrt{1\frac{v^2}{v_o^2}}}fors=\frac{v}{v_o^2}$$
(7)
or, as second solution, to:
$$q(v)=\frac{1}{\sqrt{1+\frac{v^2}{v_o^2}}}fors=\frac{v}{v_o^2}$$
(8)
It is easily verified that the second solution re-scales distances and time intervals, but does not transform the quantities which determine units of measurement, i.e. $`x_o^{}`$ remains equal to $`x_o`$ etc. Thus, by applying the reason (b) given above (which explained why $`G`$ is not an acceptable transformation), we can likewise exclude this second solution. The first solution, of course, has the mathematical form of the Lorentz transformation $`L`$ of Special Relativity, even though we evidently undertook a different path to obtain it. Here the re-scaling of coordinates is entirely consistent with changes in the sizes of the two quanta, as I shall demonstrate. Moreover, since it is clear that there is only one known speed sufficiently universal to participate in the present consideration, viz. the speed of light in vacuum $`c`$, I shall without further arguments set:
$$v_o=c$$
(9)
and our results are identical to $`L`$.
Is such an alternative approach to Special Relativity a mere pedagogy ? A crucial difference from Einstein’s theory is that a further elucidation of the transformation scale factor $`q`$, the well known Lorentz factor, is now available. When the positioning and timing measurements of a ‘moving’ observer are recorded from a ‘stationary’ reference frame, the former appears to have adopted larger basic units. The phenomenon can completely be explained by quantizing space-time as harmonic oscillators (cf. black body radiation). To prove this important point, we remind ourselves of the intrinsic uncertainties $`x_o`$ and $`t_o`$ which an experiment performed aboard any inertial platform is subject to. Now an observer notices that his partner in relative motion suffers from the higher uncertainties $`x_o^{}`$ and $`t_o^{}`$. I shall only examine $`x_o^{}`$, since the treatment of $`t_o^{}`$ is analogous. The first part of equ (5) may be re-written, with the help of equs (6), (7), and (9), as:
$$x_o^{}_{}{}^{}2=x_o^2\left(1+\frac{\frac{2v^2}{c^2}}{1\frac{v^2}{c^2}}\right)$$
(10)
It is reasonable to correspond the minimum value, $`x_o`$, with the width of the position distribution of a 1-D harmonic oscillator at ground state (which is a gaussian). Denoting the zero-point energy and oscillator constant by $`ϵ/2`$ and $`\kappa `$, respectively, we have:
$$ϵ=2\kappa x_o^2$$
(11)
I consider such a standpoint because of its remarkable interpretation of equ (10). Suppose motion enlarges $`x_o`$ because at finite $`v`$ an oscillator can populate the excited states, the degree of which is expressed by a temperature parameter $`T`$, just like a system in thermal equilibrium ($`T`$ increases with $`v`$ and $`T=0`$ when $`v=0`$). We may then write $`x_o^{}_{}{}^{}2=x_o^2(1+<x_1^2>/x_o^2)`$, where $`<x_1^2>`$ is the mean-square amplitude due to the occupation of all energy levels higher than the zero-point, and is related to the mean energy of these upper levels,
$$\overline{E}=\frac{ϵe^{\frac{ϵ}{kT}}}{1e^{\frac{ϵ}{kT}}},$$
(12)
by $`\kappa <x_1^2>=\overline{E}`$. Therefore:
$$x_o^{}_{}{}^{}2=x_o^2\left(1+\frac{ϵ}{\kappa x_o^2}\frac{e^{\frac{ϵ}{kT}}}{1e^{\frac{ϵ}{kT}}}\right)=x_o^2\left(1+\frac{2e^{\frac{ϵ}{kT}}}{1e^{\frac{ϵ}{kT}}}\right)$$
(13)
Equs (10) and (13), when taken together, suggest strongly the following association:
$$\frac{v^2}{c^2}=e^{\frac{ϵ}{kT}}\left(=\frac{\overline{E}}{ϵ+\overline{E}}\right).$$
(14)
The two equations are in perfect agreement. Thus the behavior of $`x_o`$ during uniform motion has an exact parallel with the fluctuation enhancement of a thermally agitated quantum oscillator. In this way, we bolster a posteriori the earlier premise of space-time quantization. Moreover, there is now a clear rationale for postulatin the assignment of a temperature to the space-time array of a moving observer, the ground-breaking potential of such an undertaking will become apparent when we address non-inertial behavior. We note also that the thermodynamics$`[5]`$ and quantum mechanics of space-time has been a subject of much investigation (see, e.g., the recent review of Ashtekar), even though this is the first direct attempt in explaining Relativity as macroscopic (aggregate) behavior of a simple quantum ensemble.
Before doing so, I propose the following microscopic model of space-time. It is a 4-D array of ‘nodes’ (or ‘measurement tickmarks’), all adjacent pairs of which are connected by identical harmonic oscillators with natural length equal to the minimum fluctuation $`x_o=ct_o`$, where $`x_o`$ is given by equ (11) (this reflects my earlier indication that the grid structure is controlled by the intrinsic quantum uncertainties). However, even at $`v=0`$, zero-point fluctuations are inevitable, and enlarges the separation to $`x_m=\sqrt{2}x_o`$, which forms the unit of distance for the rest observer $`\mathrm{\Sigma }`$. When there is relative motion ($`v>0`$ and hence $`T>0`$) $`x_m`$ increases to:
$$x_m^{}_{}{}^{}2=x_m^2(1+\frac{<x_1^2>}{x_m^2})$$
(15)
where, as before, $`\kappa <x_1^2>=\overline{E}`$ and $`\overline{E}`$ is given by equ (12). Thus we have:
$$x_m^{}=x_m\left(1+\frac{ϵ}{2\kappa x_o^2}\frac{e^{\frac{ϵ}{kT}}}{1e^{\frac{ϵ}{kT}}}\right)^{\frac{1}{2}}=\frac{x_m}{\sqrt{1\frac{v^2}{c^2}}}$$
(16)
This proves that our first solution of the FRP, equ (7), is fully self-consistent. Specifically the re-scaling of coordinates by the Lorentz factor $`q`$ (equs (7) and (9)) is due solely to the change of natural units with motion, thereby satisfying the basic requirement that a robust space-time quantization scheme should be compatible with the FRP.
The idea of inertial space-time arrays being equilibrium configurations (albeit having different temperatures) is further strengthened by its application to the problem of acceleration, because logical deduction would suggest this should then correspond to a situation where the array is out of equilibrium, and is characterized by a distribution of temperatures. Consider the simple case of a point (delta-function) enhancement in the temperature at the origin of a rest array, which leads to the isotropic conduction of energy in all four directions $`(ct,𝐫)`$.
Again, I will only solve for one spatial coordinate $`x`$, as the other three will follow in a likewise manner (with $`ct`$ replacing $`x`$ for the case of time). The x-axis is obviously a radial direction, and we let the energy propagate outwards from some minimum radius $`x_{min}`$. The transport equation is:
$$\sigma _{th}\frac{dT}{dx}=n\overline{E}\overline{v}.$$
(17)
Here $`x`$ is the distance from $`x=x_{min}`$ to any ‘downstream’ point, measured, of course, using the oscillator lengths at all intermediate points, which are no longer uniformly distributed, as basic units (we shall find that $`x`$ is after all an Euclidean distance). Moreover, $`\sigma _{th}=n\overline{v}\lambda d\overline{E}/dT`$ is the thermal conductivity of phonons: $`n`$ is the linear phonon density (i.e. always one oscillator per unit length) $`\overline{v}`$ is the ‘speed of sound’ in the array<sup>1</sup><sup>1</sup>1The conduction of energy takes place throughout the entire space-time array. Thus, like the oscillations, there is the need to introduce a new ‘time’ axis when defining the propagation speed of these phonons - signature of a fifth dimension., $`\overline{E}(T)`$ is the mean energy of a phonon above the ambient (zero-point) energy, and $`\lambda `$ is the phonon mean free path measured in the same way as $`x`$ is. Since phonons do not interact, this is simply the size of the available array, i.e. $`\lambda =x`$. Thus equ (17), together with the meaning of the various symbols involved<sup>2</sup><sup>2</sup>2Equ (17), which is a standard heat transport equation, may be tested by applying it to a homogeneous ideal classical gas. In this case $`n`$ and $`\overline{v}`$ are respectively the number density and mean speed of the gas particles, and $`\overline{E}=\frac{3}{2}kT`$. Also, unlike phonons, the mean free path is in general much smaller than the total dimension occupied by the gas. Solution of equ (17) then readily leads to the well-known result that the e-folding length for the temperature profile is $`\lambda `$., imply that
$$x\frac{d\overline{E}}{dx}=\overline{E},or\overline{E}=\frac{1}{\alpha x}$$
(18)
where $`\alpha `$ is a constant of integration. Combining equs (14) and (18), one obtains
$$\frac{v^2}{c^2}=\frac{1}{\alpha ϵx+1}$$
(19)
The oscillators indeed have a ‘profile’ of lengths (reflecting a temperature drop with distance) which is reproduced by assigning to every point $`x`$ a local Lorentz frame moving at speed $`v`$. Evidently $`v`$, and hence the enlargement of the length quantum, decreases with $`x`$. This is because the temperature returns gradually to the ambient value of $`T=0`$ as energy is transported downstream.
Now since $`x`$ is a radial distance we may write $`x=rr_o`$ where $`r_o=x_{min}`$. In the limit of $`rr_o`$ ($`vc`$) we have $`v^21/r`$. A careful reader will realize that the situation is the same as either (a) a point mass at the origin causing space-time curvature which attracts all other masses inwards (equ (19) gives the free-fall speed at every position), or (b) the observer is in an accelerated frame, responding to non-gravitational forces which act along the $`+x`$ direction, due again to the fields of a point object. The Principle of Equivalence excludes the absolute certainty of distinguishing between the two possibilities. This means, for the first time, one can derive from more fundamental principles that the far-field potential of any field emanating from a point source is $`1/r`$, or the force obeys inverse-square law.
Further, in the case of (a), agreement with Newtonian gravity is obtained by setting $`\alpha ϵ=c^2/2GM`$, which removes the arbitrariness of the solution. At high speeds the role of $`r_g`$ must be taken into account ($`vc`$ as $`rr_o`$). In the present case of spherical symmetry there is only one free parameter in the problem, i.e. $`r_g`$ must depend on $`\alpha `$. By setting $`r_o=1/\alpha ϵ=2GM/c^2`$, equ (19) reduces to $`v^2/c^2=r_o/r`$, or $`q=(1\frac{r_o}{r})^{\frac{1}{2}}`$. If we bear in mind that the outward energy conduction is isotropic in the space-time array, i.e. the temperature of the oscillators in the $`ct`$ direction is distributed identically to the $`r`$ direction, it will become obvious that the space-time units are constantly decreasing as one moves away from the origin. When this set of variable units is used to measure distances and time intervals, as we did, we have in fact adopted an Euclidean geometry whereby the path of all freely moving objects is a straight line. The above expression for the Lorentz factor $`q`$ contains all the information one needs to construct the full Schwarzschild metric$`[7]`$ of General Relativity. For example, as a falling object approaches the gravitational radius (or event horizon) $`r=r_o`$ time dilation becomes infinite. The forementioned notion of a minimum conduction radius is now clear: phonon energy transport to other parts of the array takes place only beyond the event horizon. Following the earlier arguments, we realize that this relativistic correction applies to point interaction involving any type of fields, which is also a new conclusion.
Lastly I propose a possible test of the theory. The domain within which the quanta of space-time oscillations manifest themselves collectively as Special Relativistic effects is the ‘harmonic limit’. It is well known that the forces which maintain stability of a system may always be approximated by a harmonic oscillator potential in the case of small perturbations about an equilibrium point. If an oscillator’s temperature is too high (meaning here that $`v`$ is too close to $`c`$) anharmonic terms will no longer be negligible, and will correct the Lorentz transformation. Here I inquire the form by which time dilation could be modified.
According to equ (15), the increase of a measurement unit with $`v`$ is due to the term $`<x_1^2>`$, which in the classical (high T) limit may be written as:
$$<x_1^2>=\frac{x^2e^{\beta V(x)}𝑑x}{e^{\beta V(x)}𝑑x}$$
(20)
where $`\beta =1/kT`$ and
$$V(x)=\frac{1}{2}\kappa x^2+\xi x^3+\eta x^4$$
(21)
Hitherto all but the first term of $`V(x)`$ have been ignored. However, in the Taylor series of equ (21) the coefficients are successive derivatives of $`V(x)`$. For a regular array of nodes, it is therefore reasonable to assume, like solid state theory, that $`\kappa /\xi `$ is comparable to $`\xi /\eta `$ etc. In this way one can include all contributions which belong to the next order of small quantities, but not beyond. The result is:
$$\frac{1}{2}\kappa <x_1^2>=\frac{1}{2}kT+\frac{1}{2}\alpha (kT)^2$$
(22)
where $`\alpha =45\xi ^2/\kappa ^312\eta /\kappa ^2`$. Substituting equ (22) into equ (15), and applying equ (11), one obtains the ratio:
$$\frac{x_m^{}}{x_m}=\frac{1}{\sqrt{1\frac{v^2}{c^2}}}\left(1+\frac{\alpha ϵ}{1\frac{v^2}{c^2}}\right)^{\frac{1}{2}}$$
(23)
which predicts that at very high speeds time dilation deviates from a simple dependence on the Lorentz factor.
To date the best direct experiments on time dilation at large $`v`$ remains those which measure the integral energy spectrum of vertical muons at sea level, where the steepening of the power-law index by unity as compared with that of the parent pions and protons is in agreement with Special Relativity$`[8]`$ up to an energy of 10 TeV$`[9]`$, or muon Lorentz factor $``$ 10<sup>5</sup>. This constrains the magnitude of the coefficient of the correction term to $`\alpha ϵ<`$ 10<sup>-10</sup>. Further, equ (23) will have a ‘feedback’ effect on the Lorentz transformation and the FRP. A thorough analysis will be performed in a later work, except to say here that the statement of complete symmetry provided by the FRP might not be sustainable at such high values of $`v`$ because when the $`\alpha ϵ`$ term is not negligible equ (23) confounds the linearly invertible property of $`L`$. This, an issue which has been raised before$`[10]`$, can be treated quantitatively using the present formalism. It is also a limit which can be investigated by observing extreme energy cosmic ray neutrino events at Lorentz factors $`>`$ 10<sup>20</sup>, where the $`\nu +p`$ interaction at $`E_\nu =`$ 10<sup>20</sup> eV is equivalent to the process of $`\nu +p`$ at $`E_p=10^{29}`$ eV. In the latter case the de Broglie wavelength of protons is smaller than the Planck length by more than two orders of magnitude, yet the same effect is not as obvious in the former case. This highlights the possibility of an asymmetry in the frames of reference represented by the two cases.
In conclusion, it is argued that Relativity is the macroscopic manifestation of space-time as a 4-D array (ensemble) of quantized harmonic oscillators. The Lorentz transformation, which connects the space-time arrays of different inertial frames, simply ‘maps’ the nodes of two equilibrium oscillator ensembles which have a relative temperature difference between them. Accelerated frames, however, are no longer associated with equilibrium states, but rather such ensembles have a temperature gradient within them. As an example, for a ‘point- enhancement’ the result is in complete agreement with General Relativity, with the inverse-square law of force as a limiting case. The theory can cope with many new situations, including a prediction of how time dilation might be modified at very high speeds. Other possibilites are changes in the structure of space-time near the event horizon (where oscillations are not harmonic) which may affect the gravitational bending of light, and diffraction of very energetic photons (with wavelengths comparable to $`x_o`$) by the space-time lattice.
I gratefully acknowledge L. Hillman, Y. Takahashi, M. Bonamente and R.B. Hicks for very helpful discussions and critical reading of the manuscript. I am also indebted to W.I. Axford, R.D. Blandford, and J. Malenfant (the editor) for their encouragement.
References
$`[1]`$ Poincare, H., 1905, ‘The Principles of Mathematics’, The Monist, 15, 1 – 24.
$`[2]`$ Hertz, H., 1893, in Electric Waves (translation by D.E. Jones), London Macmillan;
reprinted by Dover Pubs., Inc., 1962.
$`[3]`$ Lorentz, H.A., 1904, ‘Electromagnetic phenomena in a system moving with any
velocity less than that of light’, Proc. Amsterdam Acad Sci, 6, 809 – 831.
$`[4]`$ Einstein, A., 1905, Annalen der Physik, 18, 891.
$`[5]`$ Bardeen, J.W., Carter, B., Hawking, S.W., 1973, Commun Math Phys, 31, 161 – 170.
$`[6]`$ Ashtekar, A., 2000, Annalen Phys, 9, 178 – 198.
$`[7]`$ Schwarzschild, K., 1916, Über das Gravitationsfeld eines Massenpunktes
nach der Einsteinschen Theorie, Sitzber Preuss. Acad. Wiss. Berlin
189 – 196.
$`[8]`$ Hayakawa, S., Nishimura, J., Yamamoto, Y., 1964, Supplement of the
Progress of Theoretical Physics, 32, 104 – 153.
$`[9]`$ Mizutani, K., Misaki, A., Shirai, T., Watanabe, Z., Akashi, M.,
Takahashi, Y., 1978, Il Nuovo Cimento, 48, 429 – 445.
$`[10]`$ Coleman, S., Glashow, S., 1997, Phys Lett, 47, 1788.
|
warning/0005/math0005022.html
|
ar5iv
|
text
|
# Résolutions de Demazure affines et formule de Casselman-Shalika géométrique
## Introduction
Soit $`G`$ un groupe réductif connexe déployé sur un corps fini $`k=𝔽_q`$. Notons $`𝒪=k[[\varpi ]]`$ l’anneau des séries formelles à coefficients dans $`k`$ et $`F`$ le corps des fractions de $`𝒪`$. Soit $`K=G(𝒪)`$ le sous-groupe compact maximal standard de $`G(F)`$. Pour tout cocaractère dominant $`\lambda `$ de $`G`$, il est possible de construire un $`k`$-schéma projectif $`\overline{𝒬}_\lambda `$ dont l’ensemble des $`k`$-points est
$$\overline{𝒬}_\lambda (k)=\underset{\lambda ^{}\lambda }{}K\varpi ^\lambda ^{}K/K,$$
sur lequel agit le groupe $`K`$, vu comme un $`k`$-groupe algébrique de dimension infinie, à travers un quotient de type fini. Cette action induit une stratification en orbites $`\overline{𝒬}_\lambda =_{\lambda ^{}\lambda }𝒬_\lambda ^{}`$ parmi lesquelles $`𝒬_\lambda `$ est l’orbite ouverte. De plus, les $`\overline{𝒬}_\lambda `$ s’organisent en une famille inductive dont la limite est le réduit associé à la Grassmannienne affine $`𝒬`$ définie comme dans .
Le schéma $`\overline{𝒬}_\lambda `$ n’étant pas lisse en général, pour un nombre premier $`\mathrm{}`$ différent de la caractéristique de $`k`$, il est naturel de considérer le complexe d’intersection $`\mathrm{}`$-adique
$$𝒜_\lambda =\mathrm{IC}(\overline{𝒬}_\lambda ,\overline{}_{\mathrm{}}),$$
qui est $`K`$-équivariant. La fonction trace de Frobenius associée à ce faisceau pervers :
$$A_\lambda (x)=\mathrm{Tr}(\mathrm{Fr}_q,(𝒜_\lambda ){}_{x}{}^{}),$$
définie sur l’ensemble des $`k`$-points de $`\overline{𝒬}_\lambda `$, peut être vue comme un élément de l’algèbre de Hecke non ramifiée $``$ des fonctions à valeurs dans $`\overline{}_{\mathrm{}}`$, à support compact dans $`G(F)`$, qui sont invariantes à gauche et à droite par $`K`$. Lorsque $`\lambda `$ parcourt le cône des cocaractères dominants, ces fonctions $`A_\lambda `$ forment une base de $``$.
Soit $`G^{}`$ le groupe défini sur $`\overline{}_{\mathrm{}}`$ dont la donnée radicielle est duale de celle de $`G`$. Dans , Satake a construit un isomorphisme canonique entre l’algèbre $``$ et l’algèbre des fonctions régulières sur $`G^{}`$ qui sont $`\mathrm{Ad}(G^{})`$-invariantes. D’après Lusztig et Kato, voir , , la transformation de Satake de $`A_\lambda `$ est égale, à un signe près, au caractère de $`V(\lambda )`$, la représentation irréductible de plus haut poids $`\lambda `$ de $`G^{}`$. Plus récemment, Ginzburg et Mirkovic, Vilonen ont mis en lumière une équivalence tannakienne entre la catégorie des faisceaux pervers $`K`$-équivariants semi-simples sur $`𝒬`$ munie d’un produit de convolution et la catégorie des représentations algébriques de $`G^{}`$ munie du produit tensoriel. Le théorème de Lusztig-Kato serait le reflet au niveau des objets simples de cette équivalence, via la formule des traces de Grothendieck .
Les termes constants ainsi que les coefficients de Fourier des fonctions $`A_\lambda `$ sont remarquablement simples. Soit $`B=TU`$ un sous-groupe de Borel de $`G`$ et $`\rho `$ la demi-somme des racines de $`T`$ dans $`\mathrm{Lie}(U)`$. D’après Lusztig et Kato, l’intégrale terme constant est égale à
$$_{U(F)}A_\lambda (x\varpi ^\nu )𝑑x=(1)^{2\rho ,\nu }q^{\rho ,\nu }m_\lambda (\nu ),$$
$`m_\lambda (\nu )`$ est la dimension de l’espace de poids $`\nu `$ dans $`V(\lambda )`$. Parallèlement, pour $`\theta :U(F)_{\mathrm{}}^\times `$ un caractère générique, de conducteur $`U(𝒪)`$, Frenkel, Gaitsgory, Kazhdan et Vilonen ont démontré dans , que
$$_{U(F)}A_\lambda (x\varpi ^\nu )\theta (x)𝑑x=0$$
si $`\nu \lambda `$ et
$$_{U(F)}A_\nu (x\varpi ^\nu )\theta (x)𝑑x=(1)^{2\rho ,\nu }q^{\rho ,\nu }.$$
Leur démonstration s’appuie sur la formule explicite de Casselman-Shalika des valeurs des fonctions de Whittaker non ramifiées , .
L’objet principal de notre travail est de démontrer les énoncés géométriques sous-jacents à ces résultats. Pour tout cocaractère $`\nu `$, il est possible de définir un sous-ind-schéma $`S_\nu 𝒬`$ tel que
$$S_\nu (k)=U(F)\varpi ^\nu G(𝒪)/G(𝒪).$$
Il s’agit de démontrer que le complexe
$$\mathrm{R}\mathrm{\Gamma }_c(S_\nu _k\overline{k},𝒜_\lambda )$$
est concentré en degré $`2\rho ,\nu `$ et que l’endomorphisme de Frobenius agit dans son $`\mathrm{H}^{2\rho ,\nu }`$ comme la multiplication par $`q^{\rho ,\nu }`$. Cet énoncé est dû à Mirkovic et Vilonen lorsque le corps de base est $``$ et joue un rôle important dans l’équivalence tannakienne mentionnée plus haut. Il peut aussi être considéré comme une interprétation géométrique partielle du théorème de Lusztig-Kato.
Lorsque $`\nu `$ est dominant, on peut définir un morphisme $`h:S_\nu 𝔾_a`$ tel que $`\theta (x)=\psi (h(x))`$, où $`\psi :k\overline{}_{\mathrm{}}^\times `$ est un caractère additif non trivial de $`k`$. On démontre que le complexe
$$\mathrm{R}\mathrm{\Gamma }_c(S_\nu _k\overline{k},𝒜_\lambda h^{}_\psi )$$
est nul si $`\nu \lambda `$. Dans le cas $`\nu =\lambda `$, il est isomorphe à $`\overline{}_{\mathrm{}}(\rho ,\nu )`$ placé en degré $`2\rho ,\nu `$. Cet énoncé était une conjecture de Frenkel, Gaitsgory, Kazhdan et Vilonen . Comme expliqué dans loc. cit., il fournit une démonstration géométrique du théorème de Casselman-Shalika. Il pourrait également fournir des résultats pour les groupes tordus.
Voici l’organisation de l’article. Après avoir rappelé, dans la section 2, des résultats connus sur la Grassmannienne affine, nous énonçons les résultats principaux (théorèmes 3.1 et 3.2) dans la section 3.
La démonstration de ces théorèmes occupe le reste de l’article. Elle repose sur l’étude de la géométrie de certaines résolutions, formées à partir des variétés $`\overline{𝒬}_\lambda `$ les plus simples, qui correspondent au cas où $`\lambda `$ est minuscule ou quasi-minuscule. Cette stratégie est déjà utilisée dans , où la conjecture de a été démontrée pour le groupe $`\mathrm{GL}_n`$.
De façon plus détaillée, dans les sections 4 et 5, nous démontrons de façon géométrique deux énoncés (lemmes 4.2 et 5.2) concernant les intersections $`S_\nu \overline{𝒬}_\lambda `$, qui sont probablemment bien connus, mais que nous n’avons su trouver, sous cette forme, dans la littérature. Le lemme 5.2 nous permet de démontrer les théorèmes 3.1 et 3.2 dans le cas où $`\nu `$ et $`\lambda `$ sont conjugués par un élément du groupe de Weyl.
Nous signalons, au passage, un énoncé (proposition 4.3) sur les caractéristiques d’Euler-Poincaré $`\chi _c(S_\nu 𝒬_\lambda )`$ qui peut être considéré comme une interprétation géométrique d’un résultat de Lusztig \[19, 6.1\].
Nous étudions ensuite, dans les sections 6–8, la géométrie des variétés $`\overline{𝒬}_\lambda `$ dans les cas les plus simples, c.à.d. lorsque $`\lambda `$ est minuscule (section 6), ou quasi-minuscule (sections 7 et 8).
Si $`\lambda `$ est minuscule, alors $`\overline{𝒬}_\lambda `$ est égal à $`𝒬_\lambda `$ et est isomorphe au schéma $`G/P`$ des sous-groupes de $`G`$ conjugués à un certain sous-groupe parabolique $`P`$; de plus, seuls les $`\nu `$ conjugués à $`\lambda `$ interviennent, si bien que les énoncés 3.1 et 3.2 découlent dans ce cas du lemme 5.2.
Si $`\lambda `$ est quasi-minuscule, alors le point base de $`𝒬`$ est l’unique point singulier de $`\overline{𝒬}_\lambda `$. L’ouvert complémentaire de ce point, l’orbite $`𝒬_\lambda `$, est un fibré en droites au-dessus d’un $`G/P`$. Nous construisons une résolution (lemme 7.3) de $`\overline{𝒬}_\lambda `$ par un fibré en droites projectives au-dessus de $`G/P`$, qui nous permettra de démontrer, dans la section 8, les théorèmes 3.1 et 3.2 dans ce cas.
Dans la section 9, on considère certaines résolutions, qui conduisent à des produits de convolution de la forme $`𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n}`$ où chaque $`\mu _i`$ est minuscule ou quasi-minuscule. L’idée essentielle de la démonstration est que chaque $`𝒜_\lambda `$ apparaît comme facteur direct (avec une certaine multiplicité) d’un tel produit. Cette assertion est ramenée à un énoncé combinatoire, dont nous donnons deux preuves différentes dans la section 10. L’une repose sur un lemme simple sur les systèmes des racines, l’autre est basée sur la théorie des représentations et le modèle des chemins de Littelmann.
Armés de la connaissance explicite des cas minuscule et quasi-minuscule, et des résultats de la section 9, on peut alors démontrer les théorèmes 3.1 et 3.2 en suivant l’argument de . Ceci est le contenu de la section 11.
Nos résultats ont été exposés au séminaire Formes automorphes de l’Université Paris 7 en Février 1999, et au Number theory seminar du Max Planck Institut fuer Mathematik en Juin 1999. En Juillet 1999, Frenkel, Gaitsgory et Vilonen ont annoncé une autre démonstration de la conjecture de , par une voie différente et indépendante .
Pendant la rédaction de ce travail, B.C. Ngô a bénéficié de l’hospitalité du Max Planck Institut fuer Mathematik. Il remercie G. Laumon et M. Rapoport pour d’utiles discussions sur le sujet de cet article.
Nous remercions le rapporteur pour sa lecture attentive du manuscrit.
## 1. Notations
Soit $`k`$ un corps fini à $`q`$ éléments et de caractéristique $`p`$. Notons $`\overline{k}`$ sa clôture algébrique. Nous supposons, pour des raisons de commodité, que $`G`$ est un groupe algébrique semi-simple déployé sur $`k`$, la généralisation aux groupes réductifs étant évidente. Soient $`T`$ un tore maximal déployé de $`G`$, et $`B,B^{}`$ deux sous-groupes de Borel tels que $`BB^{}=T`$. On note $`U`$ (resp. $`U^{}`$) le radical unipotent de $`B`$ (resp. $`B^{}`$). On pose $`𝔤=\mathrm{Lie}(G)`$ et $`𝔥=\mathrm{Lie}(T)`$.
On note $`,`$ l’accouplement naturel entre $`X:=\mathrm{Hom}(T,𝔾_m)`$ et $`X^{}:=\mathrm{Hom}(𝔾_m,T)`$. Soient $`RX`$ le système de racines associé à $`(G,T)`$, $`R_+`$ (resp $`R_{}`$) l’ensemble des racines correspondant à $`B`$ (resp. $`B^{}`$) et $`\mathrm{\Delta }=\{\alpha _1,\mathrm{},\alpha _r\}`$ l’ensemble des racines simples. Pour tout $`\alpha R`$, on désigne par $`U_\alpha `$ le sous-groupe radiciel de $`G`$ correspondant à la racine $`\alpha `$. Soit $`R^{}X^{}`$ le système de racines dual muni de la bijection $`RR^{}`$, $`\alpha \alpha ^{}`$. Notons $`R_+^{}`$ l’ensemble des coracines positives. Soit $`W`$ le groupe de Weyl de $`(G,T)`$.
Soit $`\rho =(1/2)_{\alpha R_+}\alpha `$ la demi-somme des racines positives. Pour toute racine simple $`\alpha \mathrm{\Delta }`$, on a $`\rho ,\alpha ^{}=1`$.
On note $`Q^{}`$ (resp. $`Q_+^{}`$) le sous-groupe (resp. le sous-monoïde) de $`X^{}`$ engendré par $`R^{}`$ (resp. $`R_+^{}`$). On désigne par $`X_+^{}`$ le cône des cocaractères dominants :
$$X_+^{}=\{\lambda X^{}\alpha ,\lambda 0,\alpha R_+\}.$$
On considère l’ordre partiel sur $`X^{}`$ défini comme suit : $`\nu \nu ^{}`$ si et seulement si $`\nu \nu ^{}Q_+^{}`$.
On note $`G^{}`$ le groupe dual, considéré sur $`\overline{}_{\mathrm{}}`$; il est muni des sous-groupes $`T^{}B^{}`$. Pour tout $`\lambda X_+^{}`$, on note
$$\mathrm{\Omega }(\lambda )=\{\nu X^{}wW,w\nu \lambda \};$$
c’est l’ensemble des poids de $`T^{}`$ dans $`V(\lambda )`$, le $`G^{}`$-module simple, sur $`\overline{}_{\mathrm{}}`$, de plus haut poids $`\lambda `$ (voir, par exemple, \[3, Chap.VIII, Ex.7.1\] ou \[13, Prop. 23.1\]).
On notera $`M`$ l’ensemble des éléments minimaux de $`X_+^{}\{0\}`$.
###### Lemme 1.1.
Soit $`\mu M`$. On a l’une des alternatives suivantes.
1. Si $`\alpha ,\mu \{0,\pm 1\}`$ pour tout $`\alpha R`$, alors $`\mu `$ est un élément minimal de $`X_+^{}`$ et l’on a $`\mathrm{\Omega }(\mu )=W\mu `$. Dans ce cas, on dira que $`\mu `$ est un cocaractère minuscule.
2. Sinon, il existe une unique racine $`\gamma `$ telle que $`\gamma ,\mu 2`$; c’est une racine positive maximale, et l’on a $`\mu =\gamma ^{}`$ et $`\mathrm{\Omega }(\mu )=W\mu \{0\}`$. Dans ce cas, on dira que $`\mu `$ est quasi-minuscule.
Démonstration. Compte-tenu des références citées avant le lemme, la première assertion résulte de \[3, Chap.VI, Ex.1.24\]. Voyons la seconde.
Soit $`\gamma R`$ tel que $`\gamma ,\mu 2`$. D’après \[3, Chap.VI, Ex.1.23\] ou \[13, Prop. 23.1\], $`\mu \gamma ^{}`$ est $`W`$-conjugué à un poids dominant $`\mu `$, donc à $`0`$ ou $`\mu `$ (puisque $`\mu M`$). Or, comme $`\gamma ,\mu 2`$, on voit facilement que la norme de $`\mu \gamma ^{}`$ (relativement à un produit scalaire $`W`$-invariant) est strictement inférieure à celle de $`\mu `$. On en déduit que $`\mu =\gamma ^{}`$, et que $`\gamma `$ est l’unique racine telle que $`\gamma ,\mu 2`$. Soient $`R_\gamma `$ (resp. $`R_\gamma ^{}^{}`$) le sous-système de racines irréductible de $`R`$ (resp. $`R^{}`$) contenant $`\gamma `$ (resp. $`\gamma ^{}`$); on dira que les éléments de $`R_\gamma ^{}^{}`$ de longueur minimale sont des coracines courtes. Il est alors bien connu que l’égalité $`\mathrm{\Omega }(\gamma ^{})=W\gamma ^{}\{0\}`$ entraîne que $`\gamma `$ est l’unique racine maximale de $`R_\gamma `$; de façon équivalente, $`\gamma ^{}`$ est l’unique coracine courte dominante de $`R_\gamma ^{}^{}`$ (c.f. \[3, Chap.VIII, 7.22\]. $`\mathrm{}`$
Remarque. L’usage du mot minuscule est ici plus général que celui de , qui se limite au cas où $`R`$ est irréductible, auquel cas un copoids minuscule est nécessairement un copoids fondamental.
Soient $`𝒪=k[[\varpi ]]`$ l’anneau des séries formelles en une variable $`\varpi `$ et $`F=k((\varpi ))`$ son corps des fractions.
Pour chaque élément $`\nu X^{}`$, on note $`\varpi ^\nu T(F)`$ l’image par le cocaractère $`\nu `$, de l’uniformisante $`\varpi F^\times `$. On rappelle (voir \[6, §3.5\] et ) les décompositions de Cartan et d’Iwasawa
$$\begin{array}{ccc}\hfill G(F)& =& _{\lambda X_+^{}}G(𝒪)\varpi ^\lambda G(𝒪);\hfill \\ \hfill G(F)& =& _{\nu X^{}}U(F)\varpi ^\nu G(𝒪).\hfill \end{array}$$
Convention. Sauf mention expresse du contraire, quand on parle de points de $`k`$-schémas, il s’agira des points à valeurs dans une $`k`$-algèbre arbitraire. Par stratification d’un $`k`$-schéma $`X`$, on entend la donnée des sous-schémas localement fermés $`X_\alpha `$ de $`X`$, deux à deux disjoints, tels que $`X=X_\alpha `$.
## 2. La Grassmannienne affine
Rappelons la construction de la Grassmannienne affine $`𝒬`$, tirée de et . Comme dans loc. cit., appelons $`k`$-espaces, resp. $`k`$-groupes, les faisceaux d’ensembles, resp. de groupes, sur la catégorie des $`k`$-algèbres munie de la topologie fidèlement plate et de présentation finie.
Considérons le $`k`$-groupe $`LG`$ et son $`k`$-sous-groupe $`L^0G`$, qui associent à chaque $`k`$-algèbre $`R`$, le groupe $`G(R((\varpi )))`$ et le sous-groupe $`G(R[[\varpi ]])`$. Ces constructions s’appliquent aussi aux sous-groupes $`T`$ et $`U`$ de $`G`$, à la place de $`G`$.
Il est clair que $`L^0G`$ est représenté par le schéma en groupes limite projective des schémas en groupes de type fini $`RG(R[[\varpi ]]/(\varpi ^n))`$. Pour définir une structure d’ind-schéma sur $`LG`$, choisissons une représentation fidèle $`\rho :GSL(V)`$. Notons $`L^{(N)}G(R)`$ l’ensemble des $`gLG(R)`$ tel que l’ordre des pôles de $`\rho (g)`$ et de $`\rho (g^1)`$ n’excède pas $`N`$. D’après loc. cit., $`L^{(N)}G`$ est représentable par un schéma, et le faisceau $`𝒬`$ associé au préfaisceau $`RG(R((\varpi )))/G(R[[\varpi ]])`$ est une limite inductive de schémas projectifs $`𝒬^{(N)}=L^{(N)}G/L^0G`$.
Notons $`L^0G`$ le $`k`$-groupe $`RG(R[\varpi ^1])`$ et $`L^{<0}G`$ le noyau du morphisme $`L^0GG`$ défini par $`\varpi ^10`$. Ce sont des $`k`$-sous-groupes de $`LG`$.
D’après \[2, Prop. 1.11\] et \[17, Prop. 4.6\], on a alors le lemme suivant. Dans loc. cit., $`G`$ est supposé simplement connexe et $`k=`$, mais la démonstration s’étend au cas général. Ce résultat découle aussi d’un théorème de Ramanathan .
###### Lemme 2.1.
Le morphisme de multiplication
$$L^{<0}G\times L^0GLG$$
est une immersion ouverte.
Identifions $`L^{<0}G`$ à l’ouvert $`L^{<0}Ge_0`$, où $`e_0`$ désigne le point base de $`𝒬`$. La Grassmannienne affine $`𝒬`$ est recouverte par les ouverts translatés $`gL^{<0}Ge_0`$ au-dessus desquels la fibration $`LG𝒬`$ est triviale. Ces ouverts trivialisants sont utiles pour étudier de manière plus explicite la géométrie locale de $`𝒬`$. Par exemple, $`L^{<0}G`$ n’est pas réduit en général si bien que $`𝒬`$ ne l’est pas non plus.
Le groupe $`L^0G`$ agit naturellement sur $`𝒬`$. Pour tout $`\lambda X^{}`$, notons $`e_\lambda `$ le point $`\varpi ^\lambda e_0`$ de $`𝒬`$. Pour $`\lambda X_+^{}`$, notons $`𝒬_\lambda `$ l’orbite de $`L^0G`$ passant par $`e_\lambda `$. Notons $`\overline{𝒬}_\lambda `$ l’adhérence de $`𝒬_\lambda `$. Introduisons aussi les sous-groupes $`L^\lambda G:=\varpi ^\lambda L^0G\varpi ^\lambda `$ et $`L^{<\lambda }G:=\varpi ^\lambda L^{<0}G\varpi ^\lambda `$.
Notons $`J`$ l’image inverse du radical unipotent $`U`$ de $`B`$ par l’homomorphisme $`L^0GG`$ défini par $`\varpi 0`$ ; c’est une limite projective de groupes unipotents. Posons $`J^\lambda =JL^\lambda G`$ et $`J^\lambda =JL^{<\lambda }G`$.
Quelques soient $`\alpha R`$ et $`i`$, on désigne par $`U_{\alpha ,i}`$ l’image de l’homomorphisme $`𝔾_aLG`$ défini par $`xU_\alpha (\varpi ^ix)`$. La multiplication définit un isomorphisme
$$\underset{\alpha R_+,\alpha ,\lambda >0}{}\underset{i=0}{\overset{\alpha ,\lambda 1}{}}U_{\alpha ,i}J^\lambda $$
(en choisissant un ordre total sur l’ensemble des facteurs). En particulier, $`J^\lambda `$ est isomorphe à l’espace affine de dimension $`2\rho ,\lambda `$.
###### Lemme 2.2.
Le morphisme naturel $`J^\lambda 𝒬_\lambda `$ défini par $`jje_\lambda `$ est une immersion ouverte.
Démonstration. Il est clair que la multiplication induit un isomorphisme $`J^\lambda \times J^\lambda J`$. Il est aussi clair que la multiplication induit une immersion ouverte $`J\times B^{}L^0G`$. Par ailleurs, $`J^\lambda `$ et $`B^{}`$ sont des sous-groupes de $`L^\lambda G`$ qui fixent $`e_\lambda `$. Le lemme s’en déduit. $`\mathrm{}`$
Il résulte du lemme que l’orbite $`𝒬_\lambda `$ est lisse, irréductible et de dimension $`2\rho ,\lambda `$. Elle est incluse dans un $`𝒬^{(N)}`$ pour $`N`$ assez grand si bien que son adhérence $`\overline{𝒬}_\lambda `$ est un schéma projectif, irréductible et stable par l’action de $`L^0G`$. Il est bien connu, voir \[19, §11\], que $`\overline{𝒬}_\lambda `$ est la réunion des orbites $`𝒬_\lambda ^{}`$ avec $`\lambda ^{}\lambda `$. En particulier, si $`\mu X_+^{}`$ est nul ou bien minuscule, l’orbite $`𝒬_\mu `$ est un schéma projectif lisse.
Notons $`L^{>0}G`$ le noyau de l’homomorphisme $`L^0GG`$; c’est une limite projective de groupes unipotents. Il est clair que pour tout $`\lambda X_+^{}`$, le morphisme
$$(L^{>0}GL^\lambda G)\times (L^{>0}GL^{<\lambda }G)L^{>0}G$$
est un isomorphisme et que
$$L^{>0}GL^{<\lambda }G=\underset{\alpha R_+,\alpha ,\lambda >1}{}\underset{i=1}{\overset{\alpha ,\lambda 1}{}}U_{\alpha ,i}.$$
Soit $`P_\lambda `$ le sous-groupe parabolique de $`G`$ engendré par $`B^{}`$ et par les sous-groupes radiciels $`U_\alpha `$ avec $`\alpha ,\lambda =0`$. Le groupe de Weyl de $`P_\lambda `$ est égal au stabilisateur $`W_\lambda `$ de $`\lambda `$. Notons $`U_\lambda ^+`$ le radical unipotent du parabolique opposé à $`P_\lambda `$. Il est clair que $`P_\lambda L^\lambda G`$ et que $`J^\lambda =U_\lambda ^+(L^{>0}GL^{<\lambda }G)`$.
###### Lemme 2.3.
On a
$$L^0GL^\lambda G=P_\lambda (L^{>0}GL^\lambda G).$$
En particulier, le groupe $`L^0GL^\lambda G`$ est géométriquement connexe, et l’on a $`GL^\lambda G=P_\lambda `$.
Démonstration. Il suffit de démontrer que le morphisme de multiplication
$$(L^{>0}GL^\lambda G)\times P_\lambda L^0GL^\lambda G$$
est un isomorphisme.
Soit $`g`$ un point de $`L^0G`$ qui s’écrit sous la forme $`g=g^+g^{}uwp`$$`g^+(L^{>0}GL^\lambda G)`$, où $`g^{}(L^{>0}GL^{<\lambda }G)`$, où $`u(UwU_\lambda ^+w^1)`$, où $`pP_\lambda `$ et où $`w`$ est de longueur minimale dans sa classe $`wW_\lambda `$.
Supposons que $`gL^\lambda G`$. Puisque $`g^+`$ et $`p`$ appartiennent déjà à ce groupe, il en est de même de $`g^{}uw`$. On a donc
$$\varpi ^\lambda g^{}uw\varpi ^\lambda =(\varpi ^\lambda g^{}u\varpi ^\lambda )\varpi ^{w\lambda \lambda }wL^0G.$$
Puisque $`w`$ appartient à $`L^0G`$, $`g^{}u`$ appartient à $`LU`$ et $`\varpi ^{w\lambda \lambda }LT`$ et compte tenu de la décompostion d’Iwahori \[6, 3.5\], $`\varpi ^\lambda g^{}u\varpi ^\lambda `$ et $`\varpi ^{w\lambda \lambda }`$ appartiennent, tous les deux, à $`L^0G`$. Par ailleurs, $`\varpi ^{w\lambda \lambda }`$ appartient à $`L^0G`$ si et seulement si $`w\lambda =\lambda `$. Comme $`w`$ est de longueur minimale dans sa classe $`wW_\lambda `$, il vient $`w=1`$. Donc, on a $`uU_\lambda ^+`$. Compte tenu de la décomposition $`J^\lambda =U_\lambda ^+(L^{>0}GL^{<\lambda }G)`$ et du fait que $`J^\lambda L^0G=\{1\}`$, on obtient $`g^{}=1`$ et $`u=1`$. $`\mathrm{}`$
Soit $`\mathrm{}`$ un nombre premier différent de la caractéristique de $`k`$. Pour tout $`\lambda X_+^{}`$, notons $`𝒜_\lambda `$ le complexe d’intersection $`\mathrm{}`$-adique de $`\overline{𝒬}_\lambda `$. D’après le lemme précédent, le stabilisateur de chaque $`e_\lambda `$ dans $`L^0G`$ est géométriquement connexe. Par conséquent, tout faisceau pervers sur $`𝒬`$, géométriquement irréductible, $`L^0G`$-équivariant et dont le support est un schéma de type fini, est isomorphe à $`𝒜_\lambda `$ pour un certain $`\lambda X_+^{}`$.
A la suite de Lusztig, Ginzburg, Mirkovic et Vilonen, voir , et , on va définir le produit de convolution $`𝒜_{\lambda _1}𝒜_{\lambda _2}`$ pour $`\lambda _1,\lambda _2X_+^{}`$ comme suit. Considérons les morphismes
$$𝒬\times 𝒬\stackrel{\pi _1}{}LG\times 𝒬\stackrel{\pi _2}{}𝒬\times 𝒬$$
définis par $`\pi _1(g,x)=(ge_0,x)`$ et $`\pi _2(g,x)=(ge_0,gx)`$. Le morphisme $`\pi _1`$ est le morphisme quotient pour l’action de $`L^0G`$ sur $`LG\times 𝒬`$ définie par
$$\alpha _1(h)(g,x)=(gh^1,x).$$
Le morphisme $`\pi _2`$ est le morphisme quotient pour l’action de $`L^0G`$ sur $`LG\times 𝒬`$ définie par
$$\alpha _2(h)(g,x)=(gh^1,hx).$$
Pour tous $`\lambda _1,\lambda _2X_+^{}`$, notons $`\overline{𝒬}_{\lambda _1}\stackrel{~}{\times }\overline{𝒬}_{\lambda _2}`$ le quotient de $`\pi _1^1(\overline{𝒬}_{\lambda _1}\times \overline{𝒬}_{\lambda _2})`$ par $`\alpha _2(L^0G)`$. L’existence de ce quotient est assurée par la locale trivialité du morphisme $`LG𝒬`$. Plus précisément, au-dessus des ouverts de $`\overline{𝒬}_{\lambda _1}`$ de la forme $`gL^{<0}Ge_0\overline{𝒬}_{\lambda _1}`$, les schémas $`\overline{𝒬}_{\lambda _1}\stackrel{~}{\times }\overline{𝒬}_{\lambda _2}`$ et $`\overline{𝒬}_{\lambda _1}\times \overline{𝒬}_{\lambda _2}`$ sont isomorphes. De plus, ces isomorphismes sont clairement compatibles avec la stratification de $`\overline{𝒬}_{\lambda _1}\times \overline{𝒬}_{\lambda _2}`$ par les sous-schémas localement fermés $`𝒬_{\lambda _1^{}}\times 𝒬_{\lambda _2^{}}`$ et celle de $`\overline{𝒬}_{\lambda _1}\stackrel{~}{\times }\overline{𝒬}_{\lambda _2}`$ par les sous-schémas localement fermés
$$𝒬_{\lambda _1^{}}\stackrel{~}{\times }𝒬_{\lambda _2^{}}=\pi _1^1(𝒬_{\lambda _1^{}}\times 𝒬_{\lambda _2^{}})/\alpha _2(L^0G),$$
avec $`\lambda _1^{}\lambda _1`$ et $`\lambda _2^{}\lambda _2`$. La projection sur le second facteur définit un morphisme
$$m:\overline{𝒬}_{\lambda _1}\stackrel{~}{\times }\overline{𝒬}_{\lambda _2}\overline{𝒬}_{\lambda _1+\lambda _2}.$$
On pose
$$𝒜_{\lambda _1}𝒜_{\lambda _2}=\mathrm{R}m_{}(𝒜_{\lambda _1}\stackrel{~}{}𝒜_{\lambda _2}),$$
$`𝒜_{\lambda _1}\stackrel{~}{}𝒜_{\lambda _2}`$ désigne le complexe d’intersection de $`\overline{𝒬}_{\lambda _1}\stackrel{~}{\times }\overline{𝒬}_{\lambda _2}`$.
La construction précédente, généralisée de la manière évidente, permet de définir le produit de convolution itéré
$$𝒜_{\lambda _1}\mathrm{}𝒜_{\lambda _n}$$
pour tous $`\lambda _1,\mathrm{},\lambda _nX_+^{}`$. D’après et , ce produit de convolution est encore un faisceau pervers, somme directe, avec multiplicités, de faisceaux pervers $`𝒜_\lambda `$ avec $`\lambda \lambda _1+\mathrm{}+\lambda _n`$. Nous n’utiliserons ce résultat que dans le cas où les $`\lambda _i`$ appartiennent à l’ensemble $`M`$. Nous proposons une démonstration simple dans ce cas et montrons comment le cas général peut, en fait, se déduire de celui-ci.
## 3. Les énoncés principaux
Rappelons que $`U`$ désigne le radical unipotent du sous-groupe de Borel $`B`$ associé à $`R_+`$. On définit de façon analogue pour $`U`$, à la place de $`G`$, les groupes de lacets $`LU`$, $`L^0U=LUL^0G`$ et $`L^{<0}U=LUL^{<0}G`$. Pour tout $`\nu X^{}`$, on note aussi $`L^\nu U=\varpi ^\nu L^0U\varpi ^\nu `$ et $`L^{<\nu }U=\varpi ^\nu L^{<0}U\varpi ^\nu `$. La multiplication induit un isomorphisme
$$L^\nu U\times L^{<\nu }ULU.$$
Pour tout $`\nu X^{}`$, $`L^{<\nu }U`$ est un sous-groupe fermé de $`L^{<\nu }G`$ si bien qu’on peut identifier $`L^{<\nu }Ue_0`$ à un fermé, noté $`S_\nu `$, de l’ouvert $`\varpi ^\nu L^{<0}Ge_0`$ de $`𝒬`$. En particulier, pour tout $`\lambda X_+^{}`$ et tout $`\nu X^{}`$, $`S_\nu \overline{𝒬}_\lambda `$ est un sous-schéma localement fermé, éventuellement vide, de $`\overline{𝒬}_\lambda `$. D’après la décomposition d’Iwasawa, ces intersections $`S_\nu \overline{𝒬}_\lambda `$ forment une stratification de $`\overline{𝒬}_\lambda `$.
Nous donnerons une nouvelle démonstration du théorème suivant, dû à Mirkovic et Vilonen dans le cas $`k=`$ ().
###### Théorème 3.1.
Quelques soient $`\lambda X_+^{}`$ et $`\nu X^{}`$, le complexe $`\mathrm{R}\mathrm{\Gamma }_c(S_\nu ,𝒜_\lambda )`$ est concentré en degré $`2\rho ,\nu `$. De plus, l’endomorphisme $`\mathrm{Fr}_q`$ agit dans $`\mathrm{H}_c^{2\rho ,\nu }(S_\nu ,𝒜_\lambda )`$ comme $`q^{\rho ,\nu }`$.
Dans l’énoncé précédent on a écrit $`\mathrm{R}\mathrm{\Gamma }_c(S_\nu ,𝒜_\lambda )`$ à la place de
$$\mathrm{R}\mathrm{\Gamma }_c((S_\nu \overline{𝒬}_\lambda )_k\overline{k},𝒜_\lambda ),$$
pour alléger la notation. On utilisera systématiquement cette notation allégée dans la suite, cet abus de notation ne causant aucune ambiguité.
Soient $`\nu X_+^{}`$ et $`\nu ^{}X_+^{}`$. En choisissant un ordre total sur les racines positives, on a un isomorphisme
$$\underset{\alpha R^+}{}\underset{\alpha ,\nu ^{}i<\alpha ,\nu }{}U_{\alpha ,i}=L^{<\nu }UL^\nu ^{}U.$$
Pour $`\nu `$ fixé et pour $`\nu ^{}`$ de plus en plus anti-dominant, ces groupes forment un système inductif dont la limite est $`L^{<\nu }U`$.
Pour toute racine simple $`\alpha \mathrm{\Delta }`$, notons $`u_{\alpha ,i}`$ la projection sur le facteur $`U_{\alpha ,i}`$ et
$$h:L^{<\nu }UL^\nu ^{}U𝔾_a$$
le morphisme $`h(x)=_{\alpha \mathrm{\Delta }}u_{\alpha ,1}(x)`$. Ce morphisme est visiblement compatible aux flèches de transition et induit sur la limite inductive un morphisme $`h:L^{<\nu }U𝔾_a`$, pour tout $`\nu `$ dominant. Compte tenu de l’isomorphisme $`L^{<\nu }U\times L^\nu ULU`$, $`u^{}u^+u`$, on peut définir un morphisme, noté aussi $`h:LU𝔾_a`$, par la relation $`h(u^{}u^+)=h(u^{})`$. Ce morphisme ne dépend pas du $`\nu `$ dominant choisi. On en déduit un morphisme, noté encore $`h:S_\nu 𝔾_a`$.
Fixons un caractère additif non trivial $`\psi :k\overline{}_{\mathrm{}}^\times `$ et notons $`_\psi `$ le faisceau d’Artin-Schreier sur $`𝔾_a`$ associé à $`\psi `$. Le caractère $`\theta :U(F)\overline{}_{\mathrm{}}`$ considéré dans l’introduction est le caractère $`x\psi (h(x))`$.
L’énoncé suivant a été conjecturé par Frenkel, Gaitsgory, Kazhdan et Vilonen .
###### Théorème 3.2.
Pour $`\nu \lambda `$ dans $`X_+^{}`$, le complexe $`\mathrm{R}\mathrm{\Gamma }_c(S_\nu ,𝒜_\lambda h^{}_\psi )`$ est nul. Pour $`\nu =\lambda `$, ce complexe est isomorphe à $`\overline{}_{\mathrm{}}`$, muni de l’action de Frobenius agissant par $`q^{\rho ,\lambda }`$, placé en degré $`2\rho ,\lambda `$.
Ces résultats impliquent les énoncés sur les termes constants et les coefficients de Fourier mentionnés dans l’introduction via le dictionnaire faisceaux-fonctions de Grothendieck .
Nous exposerons les démonstrations de ces deux théorèmes en parallèle dans la suite de l’article.
## 4. L’action du tore $`T`$
Le tore $`T`$ normalise les sous-groupes $`L^0G`$, $`L^{<0}G`$, $`L^{<\nu }U`$ … de $`LG`$ si bien qu’il agit sur tous les objets géométriques qu’on a considérés dans les deux sections précédentes. Cette action fournit un outil précieux pour étudier leur géométrie. Choisissons une fois pour toutes un cocaractère strictement dominant $`\varphi :𝔾_mT`$. On entendra par l’action $`\varphi (𝔾_m)`$, l’action restreinte de $`T`$ à $`𝔾_m`$ via ce cocaractère.
###### Lemme 4.1.
Pour tout $`\nu X^{}`$, le point $`e_\nu `$ est le seul point fixe de l’action $`\varphi (𝔾_m)`$ sur $`S_\nu `$. De plus, c’est un point fixe attractif.
Démonstration. Tout $`xL^{<\nu }U(\overline{k})`$ est de la forme
$$x=\underset{\alpha R_+}{}\underset{i<\alpha ,\nu }{}U_{\alpha ,i}(x_{\alpha ,i}),$$
où les $`x_{\alpha ,i}\overline{k}`$ sont nuls sauf pour un nombre fini d’entre eux. Alors, pour tout $`z\overline{k}^\times `$, on a
$$\varphi (z)xe_\nu =\underset{\alpha R^+}{}\underset{i<\alpha ,\nu }{}U_{\alpha ,i}(z^{\alpha ,\varphi }x_{\alpha ,i})e_\nu .$$
Le lemme résulte donc de l’hypothèse que pour tout $`\alpha R_+`$, l’entier $`\alpha ,\varphi `$ est strictement positif. $`\mathrm{}`$
Ce lemme montre que les $`e_\nu `$ sont les seuls points fixes de l’action $`\varphi (𝔾_m)`$ sur $`𝒬`$. De plus, il entraîne l’énoncé suivant, qui précise \[21, 2.6.11(3)\] et est certainement bien connu.
###### Lemme 4.2.
Si l’intersection $`S_\nu \overline{𝒬}_\lambda `$ n’est pas vide, $`\nu `$ appartient à $`\mathrm{\Omega }(\lambda )`$.
Démonstration. Si un point $`xe_\nu `$, avec $`xL^{<\nu }U(\overline{k})`$, appartient à $`\overline{𝒬}_\lambda (\overline{k})`$, toute l’orbite de $`\varphi (𝔾_m)`$ passant par ce point y appartient aussi. Puisque le point fixe $`e_\nu `$ appartient à l’adhérence de cette orbite et que $`\overline{𝒬}_\lambda `$ est propre, $`e_\nu `$ appartient à $`\overline{𝒬}_\lambda `$ d’où $`\nu \mathrm{\Omega }(\lambda )`$. $`\mathrm{}`$
Signalons au passage l’énoncé suivant qui ne servira pas dans la suite de l’article. Cet énoncé a été découvert lors d’une conversation que l’un de nous a eu avec M. Rapoport.
###### Proposition 4.3.
La caractéristique d’Euler-Poincaré $`\chi _c(S_\nu 𝒬_\lambda )`$ est égale à $`1`$ si $`\nu `$ est conjugué à $`\lambda `$ par un élément de $`W`$ et à $`0`$ sinon.
Démonstration. Dans le premier cas, $`S_\nu 𝒬_\lambda `$ contient un unique point fixe $`e_\nu `$ de l’action $`\varphi (𝔾_m)`$. Dans le second cas, le groupe multiplicatif $`\varphi (𝔾_m)`$ agit sans point fixe. La proposition résulte donc d’un théorème de Bialynicki-Birula \[4, cor. 2\]. $`\mathrm{}`$
Cet énoncé peut être considéré comme une interprétation géométrique d’un résultat de Lusztig \[19, 6.1\]. Reprenons les notations de l’introduction. Soit $`C_\lambda `$ l’élément de l’algèbre de Hecke $``$ défini par
$$C_\lambda =(1)^{2\rho ,\lambda }q^{\rho ,\lambda }𝕀_\lambda ,$$
$`𝕀_\lambda `$ est la fonction caractéristique de $`K\varpi ^\lambda K`$. On sait que
$$(C_\lambda )=(K_{\lambda ,\mu }(q))^1(A_\lambda ),$$
$`(K_{\lambda ,\mu }(q))`$ est la matrice triangulaire formée des polynômes de Kazhdan et Lusztig. Les termes constants normalisés
$$(1)^{2\rho ,\nu }q^{\rho ,\nu }_{U(F)}C_\lambda (x\varpi ^\mu )𝑑x$$
sont donc calculés par la matrice
$$(K_{\lambda ,\mu }(q))^1(m_\lambda (\mu ))$$
d’après le théorème de Lusztig-Kato. Compte tenu de la proposition précédente (et aussi de 3.1), on obtient, en spécialisant $`q1`$,
$$(K_{\lambda ,\mu }(1))^1(m_\lambda (\mu ))=\mathrm{Id},$$
d’où $`K_{\lambda ,\mu }(1)=m_\lambda (\mu )`$.
## 5. Les intersections $`S_{w\lambda }\overline{𝒬}_\lambda `$
Pour $`\lambda X_+^{}`$, on a considéré dans la section 2 le groupe
$$J^\lambda =\underset{\alpha R_+}{}\underset{i=0}{\overset{\alpha ,\lambda 1}{}}U_{\alpha ,i}$$
qui est manifestement un sous-groupe de $`L^0U`$. On a aussi démontré que le morphisme $`J^\lambda \overline{𝒬}_\lambda `$ défini par $`jje_\lambda `$ est une immersion ouverte.
###### Lemme 5.1.
Soit $`\lambda X_+^{}`$. Le morphisme $`jje_\lambda `$ induit un isomorphisme de $`J^\lambda `$ sur l’ouvert $`\varpi ^\lambda L^{<0}Ge_0\overline{𝒬}_\lambda `$ de $`\overline{𝒬}_\lambda `$.
Démonstration. L’image de $`J^\lambda `$ est contenue dans $`\varpi ^\lambda L^{<0}Ge_0\overline{𝒬}_\lambda `$. D’après le lemme 2.2, elle est en fait un ouvert dense de $`\varpi ^\lambda L^{<0}Ge_0\overline{𝒬}_\lambda `$. Par conséquent, $`\varpi ^\lambda J^\lambda \varpi ^\lambda `$ est un ouvert dense dans l’image inverse de $`\varpi ^\lambda L^{<0}Ge_0\overline{𝒬}_\lambda `$ par l’isomorphisme
$$L^{<0}G\varpi ^\lambda L^{<0}Ge_0.$$
Or, $`\varpi ^\lambda J^\lambda \varpi ^\lambda `$ est un sous-groupe fermé de $`L^{<0}G`$ et le lemme s’en déduit. $`\mathrm{}`$
###### Lemme 5.2.
Soit $`\lambda X_+^{}`$. Pour tout $`wW`$, le morphisme
$$wJ^\lambda w^1LUS_{w\lambda }\overline{𝒬}_\lambda $$
défini par $`jje_{w\lambda }`$ est un isomorphisme. Par conséquent, $`S_{w\lambda }\overline{𝒬}_\lambda `$ est isomorphe à l’espace affine de dimension $`\rho ,\lambda +w\lambda `$.
Démonstration. Pour $`w=1`$, l’assertion résulte de manière évidente du lemme précédent, car on a les inclusions
$$J^\lambda e_\lambda S_\lambda \overline{𝒬}_\lambda \varpi ^\lambda L^{<0}Ge_0\overline{𝒬}_\lambda .$$
Pour $`wW`$ quelconque, on peut raisonner comme suit. D’après le lemme précédent, le morphisme
$$wJ^\lambda w^1\varpi ^{w\lambda }L^{<0}Ge_0\overline{𝒬}_\lambda $$
défini par $`jje_{w\lambda }`$ est un isomorphisme. Par ailleurs, la multiplication
$$(wJ^\lambda w^1LU)\times (wJ^\lambda w^1LU^{})wJ^\lambda w^1$$
définit aussi un isomorphisme si bien que pour $`xL^{<w\lambda }U`$ tel que $`x\varpi ^{w\lambda }\overline{𝒬}_\lambda `$, $`x`$ doit s’écrire uniquement sous la forme $`x=x_+x_{}`$ avec $`x_+wJ^\lambda w^1LU`$ et $`x_{}wJ^\lambda w^1LU^{}`$. Or, l’intersection $`LULU^{}`$ est réduite à l’élément neutre de sorte que la seule possibilité est $`x=x_+`$ et $`x_{}=1`$. Ceci prouve la première assertion.
De plus, la multiplication induit un isomorphisme
$`()`$
$$\underset{\alpha R_+w^1R_+}{}\underset{i=0}{\overset{\alpha ,\lambda 1}{}}U_{w\alpha ,i}\stackrel{}{}wJ^\lambda w^1LU.$$
Compte-tenu de l’égalité
$$\underset{\alpha R_+w^1R_+}{}\alpha =\rho +w^1\rho ,$$
on obtient la seconde assertion. $`\mathrm{}`$
On peut déduire de ce lemme l’énoncé 3.1 dans le cas $`\nu =w\lambda `$ ainsi que 3.2 dans le cas $`\nu =\lambda `$. En effet, l’inclusion évidente $`wJ^\lambda w^1LUL^0U`$ implique que $`S_{w\lambda }\overline{𝒬}_\lambda `$ est contenue dans l’orbite ouverte $`𝒬_\lambda `$. La restriction de $`𝒜_\lambda `$ à $`S_{w\lambda }\overline{𝒬}_\lambda `$ est donc égale à :
$$𝒜_\lambda |_{S_{w\lambda }\overline{𝒬}_\lambda }=\overline{}_{\mathrm{}}[\rho ,2\lambda ](\rho ,\lambda ).$$
L’énoncé 3.1 dans le cas $`\nu =w\lambda `$ s’ensuit.
L’inclusion $`J^\lambda L^0U`$ implique par ailleurs que la restriction de $`h`$ à $`J^\lambda `$ est nulle. L’énoncé 3.2 dans le cas $`\nu =\lambda `$ s’ensuit donc aussi.
On aura besoin plus loin de l’énoncé plus général ci-dessous. Pour tout $`\sigma X_+^{}`$, on note $`h_\sigma :LU𝔾_a`$ le morphisme $`h_\sigma (x)=h(\varpi ^\sigma x\varpi ^\sigma )`$ et aussi le morphisme $`h_\sigma :S_\lambda 𝔾_a`$ qui s’en déduit. Du fait que $`\sigma `$ est dominant, la restriction de $`h_\sigma `$ à $`L^0U`$, et a fortiori à $`J^\lambda `$, est nulle. Le lemme suivant résulte donc également de la discussion précédente.
###### Lemme 5.3.
Pour tous $`\lambda ,\sigma X_+^{}`$, on a
$$\mathrm{R}\mathrm{\Gamma }_c(S_\lambda ,𝒜_\lambda h_\sigma ^{}_\psi )=\overline{}_{\mathrm{}}[2\rho ,\lambda ](\rho ,\lambda ).$$
## 6. Minuscules
Nous utilisons les notations fixées dans la section 1. Soit $`\mu `$ un élément minimal, non nul, de $`X_+^{}`$. D’après le lemme 1.1, $`\mu `$ est un copoids minuscule, et l’on a l’énoncé suivant, que nous rappelons ici pour la commodité du lecteur.
###### Lemme 6.1.
Soit $`\mu `$ minuscule. On a $`\mathrm{\Omega }(\mu )=W\mu `$. Pour tout $`\alpha R`$, on a $`\alpha ,\mu \{0,\pm 1\}`$.
Si $`\mu `$ est minuscule, sa minimalité implique que l’orbite $`𝒬_\mu `$ est fermée. Puisque tout élément $`\nu `$ de $`\mathrm{\Omega }(\mu )`$ est conjugué à $`\mu `$ par l’action de $`W`$ les énoncés 3.1 et 3.2 sont donc vérifiés pour $`\lambda =\mu `$ et $`\nu \mathrm{\Omega }(\mu )`$.
On peut décrire explicitement la variété $`𝒬_\mu `$ et les strates $`S_{w\mu }𝒬_\mu `$. On notera $`P`$ le sous-groupe de $`G`$ engendré par $`T`$ et les $`U_\alpha `$ avec $`\alpha ,\mu 0`$; c’est un sous-groupe parabolique contenant $`B^{}`$.
###### Lemme 6.2.
On a un isomorphisme canonique $`𝒬_\mu G/P`$ via lequel $`S_{w\mu }𝒬_\mu `$ s’identifie à $`UwP/P`$.
Démonstration. Compte tenu du lemme 2.3 et de la deuxième assertion du lemme 6.1, on sait que $`L^0GL^\mu G`$ est l’image inverse de $`P`$ par l’homomorphisme $`L^0GG`$. On en déduit l’isomorphisme
$$𝒬_\mu =L^0G/(L^0GL^\mu G)G/P.$$
Compte tenu, de nouveau, de la deuxième assertion du lemme 6.1, on sait que $`J^\mu `$ est égal à $`U_\mu ^+=_{\alpha ,\mu =1}U_\alpha `$ qui est le radical unipotent du sous-groupe parabolique opposé à $`P`$, et par conséquent
$$wJ^\mu w^1LU=wU_\mu ^+w^1U.$$
La seconde assertion du lemme se déduit alors du lemme 5.2. $`\mathrm{}`$
## 7. Quasi-minuscules : étude géométrique
Soit $`\mu `$ un copoids quasi-minuscule, c.à.d. un élément minimal de $`X_+^{}\{0\}`$ minoré par $`0`$. Rappelons que, d’après le lemme 1.1, on a l’énoncé suivant :
###### Lemme 7.1.
Soit $`\mu `$ quasi-minuscule. Alors $`\mu `$ est égal à la coracine $`\gamma ^{}`$ associée à une racine maximale $`\gamma `$. On a $`\mathrm{\Omega }(\mu )=W\mu \{0\}`$. Pour toute racine $`\alpha R\{\pm \gamma \}`$, on a $`\alpha ,\mu \{0,\pm 1\}`$.
Puisque $`0`$ est le seul cocaractère dominant qui minore $`\mu `$, $`\overline{𝒬}_\mu `$ est la réunion de $`𝒬_\mu `$ et du point base $`e_0`$.
Désignons encore par $`P`$ le sous-groupe parabolique de $`G`$ engendré par $`T`$ et par les sous-groupes radiciels $`U_\alpha `$ tels que $`\alpha ,\gamma ^{}0`$. Notons
$$V=𝔥\underset{\alpha R\{\gamma \}}{}𝔤_\alpha $$
$`𝔥`$ est l’algèbre de Lie de $`T`$ et où $`𝔤_\alpha `$ est le sous-espace de poids $`\alpha `$ de $`𝔤`$. D’après le lemme précédent, $`V`$ est la somme des espaces de poids $`\nu `$ dans $`𝔤`$ tels que $`\gamma ,\nu 1`$. Il résulte alors de la définition de $`P`$ que $`V`$ est $`P`$-stable.
Identifions $`𝔤_\gamma `$ au quotient $`𝔤/V`$ muni de sa structure de $`P`$-module. Considérons le fibré en droites
$$𝕃_\gamma =G\times ^P𝔤_\gamma $$
au-dessus de $`G/P`$.
###### Lemme 7.2.
L’orbite $`𝒬_\mu `$ est canoniquement isomorphe à $`𝕃_\gamma `$.
Démonstration. Par définition, le foncteur $`RG(R[\varpi ]/(\varpi ^2))`$ est représenté par le fibré tangent $`TG`$ de $`G`$ qui est isomorphe au produit semi-direct $`G𝔤`$. Compte tenu du lemme 2.3 et de la dernière assertion du lemme 7.1, on sait que $`L^0GL^\mu G`$ est exactement l’image inverse de $`PV`$ par l’homomorphisme canonique $`L^0GG𝔤`$.
Il en résulte l’isomorphisme
$$𝒬_\mu (G𝔤)/(PV)=G\times ^P(𝔤/V)$$
dont le terme de droite n’est autre que $`𝕃_\gamma `$. $`\mathrm{}`$
Le fibré $`𝕃_\gamma `$ se compactifie de façon naturelle en un fibré en droites projectives. En effet, on a
$$𝕃_\gamma \mathrm{Proj}(𝕃_\gamma 𝒪_{G/P})=_\gamma .$$
Soit $`𝕃_\gamma =G\times ^P𝔤_\gamma `$ le fibré dual. On a un isomorphisme naturel
$$\mathrm{Proj}(𝕃_\gamma 𝒪_{G/P})\mathrm{Proj}(𝒪_{G/P}𝕃_\gamma )=_\gamma ,$$
si bien qu’on peut voir $`_\gamma `$ comme la réunion de $`𝕃_\gamma `$ et de $`𝕃_\gamma `$. Notons $`\varphi _{\pm \gamma }`$ le morphisme $`𝕃_{\pm \gamma }G/P`$ et $`ϵ_{\pm \gamma }`$ la section nulle $`G/P𝕃_{\pm \gamma }`$. On a alors
$$_\gamma =𝕃_{\pm \gamma }ϵ_\gamma (G/P).$$
###### Lemme 7.3.
L’isomorphisme $`𝕃_\gamma 𝒬_\mu `$ se prolonge en un morphisme
$$\pi _\gamma :_\gamma \overline{𝒬}_\mu $$
qui envoie $`ϵ_\gamma (G/P)`$ sur le point $`e_0`$.
Démonstration. Il s’agit d’un cas particulier du théorème principal de Zariski. Nous reproduisons l’argument classique pour la commodité du lecteur. Notons $`\mathrm{\Gamma }`$ l’adhérence du graphe de l’isomorphisme $`𝕃_\gamma 𝒬_\mu `$ dans $`_\gamma \times \overline{𝒬}_\mu `$. Puisque le complémentaire de $`𝒬_\mu `$ dans $`\overline{𝒬}_\mu `$ est constitué d’un seul point $`e_0`$, toutes les fibres de la projection $`\mathrm{\Gamma }_\gamma `$ ne contiennent qu’un point. En particulier, cette projection est quasi-finie. Puisque de plus, elle est propre, elle est finie. Mais $`_\gamma `$ est lisse, en particulier normale, donc le morphisme fini, birationnel $`\mathrm{\Gamma }_\gamma `$ doit être un isomorphisme. En inversant cet isomorphisme et en le composant avec l’autre projection $`\mathrm{\Gamma }\overline{𝒬}_\mu `$, on obtient le morphisme $`\pi _\gamma :_\gamma \overline{𝒬}_\mu `$ voulu. $`\mathrm{}`$
Signalons que l’action de $`L^0G`$ sur $`𝒬_\mu `$ se prolonge à $`_\gamma `$ et que la résolution $`\pi _\gamma `$ est équivariante par rapport à cette action. Nous n’utiliserons cette information que dans la remarque située après le corollaire 9.7, laquelle ne sert pas dans le reste de l’article.
On a la description explicite suivante des strates $`S_{w\mu }\overline{𝒬}_\mu `$.
###### Lemme 7.4.
Si $`w\gamma R_+`$ alors
$$S_{w\mu }\overline{𝒬}_\mu =\varphi _\gamma ^1(UwP/P).$$
Si $`w\gamma R_{}`$ alors
$$S_{w\mu }\overline{𝒬}_\mu =ϵ_\gamma (UwP/P).$$
Démonstration. D’après la formule $`()`$ établie dans la démonstration du lemme 5.2, l’on a:
$$wJ^\mu w^1LU=\underset{\alpha R_+w^1R_+}{}\underset{i=0}{\overset{\alpha ,\mu 1}{}}U_{w\alpha ,i}.$$
De plus, comme $`\alpha ,\mu 1`$ pour tout $`\alpha R_+\{\gamma \}`$, d’après le lemme 7.1, on obtient que ce produit est égal à
$$U_{w\gamma ,1}\underset{\alpha R_+w^1R_+}{}U_{w\alpha ,0}$$
si $`w\gamma R_+`$, et à
$$\underset{\alpha R_+w^1R_+}{}U_{w\alpha ,0}$$
sinon. Le lemme s’en déduit. $`\mathrm{}`$
On note $`W_\gamma `$ le stabilisateur de $`\gamma `$ dans $`W`$, et $`\mathrm{\Delta }_\gamma `$ l’ensemble des racines simples conjuguées à $`\gamma `$.
###### Corollaire 7.5.
On a une stratification
$$S_0\overline{𝒬}_\mu =\{e_0\}\underset{\stackrel{wW/W_\gamma }{w\gamma R_{}}}{}\varphi _\gamma ^1(UwP/P)ϵ_\gamma (UwP/P).$$
En particulier, les composantes irréductibles de $`S_0\overline{𝒬}_\mu `$ sont en bijection avec $`\mathrm{\Delta }_\gamma `$ et sont toutes de dimension $`\rho ,\mu `$.
On a aussi la stratification
$$\pi _\gamma ^1(S_0\overline{𝒬}_\mu )=\underset{\stackrel{wW/W_\gamma }{w\gamma R_{}}}{}\varphi _\gamma ^1(UwP/P)\underset{\stackrel{wW/W_\gamma }{w\gamma R_+}}{}ϵ_\gamma (UwP/P).$$
## 8. Quasi-minuscules : étude cohomologique
Les notations de la section précédente restent en vigueur. En particulier, $`\mu =\gamma ^{}`$ est quasi-minuscule. La résolution $`\pi _\gamma :_\gamma \overline{𝒬}_\mu `$ permet de calculer la cohomologie d’intersection locale de $`𝒜_\mu `$ en le point singulier isolé $`e_0`$. L’énoncé suivant est dû à Kazhdan et Lusztig \[16, Lemme 4.5\]. En fait, dans notre situation les hypothèses sont un peu plus faibles, mais leur argument s’applique encore. Nous détaillons la démonstration pour la commodité du lecteur.
###### Lemme 8.1.
Soit $`d=2\rho ,\mu `$ la dimension de $`\overline{𝒬}_\mu `$. Pour $`i0`$, le groupe $`\mathrm{H}^i(𝒜_\mu )_{e_0}`$ est nul. Pour $`i<0`$, on a une suite exacte courte
$$0\mathrm{H}^{i+d2}(G/P)(d/21)\stackrel{c_\gamma }{}\mathrm{H}^{i+d}(G/P)(d/2)\mathrm{H}^i(𝒜_\mu )_{e_0}0,$$
$`c_\gamma H^2(X_\gamma )(1)`$ est la classe de Chern de $`𝕃_\gamma `$.
Démonstration. Notons $`\overline{𝒬}_\mu ^{}`$ l’ouvert de $`\overline{𝒬}_\mu `$
$$\overline{𝒬}_\mu ^{}=\overline{𝒬}_\mu \pi _\gamma ϵ_\gamma (G/P);$$
on a $`\pi _\gamma ^1(\overline{𝒬}_\mu ^{})=𝕃_\gamma `$. Notons $`𝒜_\mu ^{}`$ la restriction de $`𝒜_\mu `$ à cet ouvert. Notons $`ı`$ l’inclusion du point fermé $`ı:\{e_0\}\overline{𝒬}_\mu ^{}`$. La flèche naturelle $`𝒜_\mu ^{}ı_{}ı^{}𝒜_\mu ^{}`$ induit le morphisme de restriction sur la cohomologie (sans support)
$$ı^{}:\mathrm{R}\mathrm{\Gamma }(\overline{𝒬}_\mu ^{},𝒜_\mu ^{})(𝒜_\mu ^{})_{e_0}.$$
On démontre d’abord que $`ı^{}`$ est un isomorphisme.
Pour cela, utilisons le théorème de décomposition de Beilinson, Bernstein, Deligne et Gabber . Puisque $`\pi _\gamma :_\gamma \overline{𝒬}_\mu `$ est un isomorphisme en dehors de $`e_0`$, on a une décomposition
$$\mathrm{R}\pi _{\gamma ,}\overline{}_{\mathrm{}}[d](d/2)=𝒜_\mu 𝒞,$$
$`𝒞`$ est un complexe supporté par le point $`e_0`$.
La section nulle $`ϵ_\gamma :G/P𝕃_\gamma `$ définit le morphisme de restriction
$$\mathrm{R}\mathrm{\Gamma }(𝕃_\gamma ,\overline{}_{\mathrm{}})\mathrm{R}\mathrm{\Gamma }(G/P,\overline{}_{\mathrm{}})$$
qui est un isomorphisme puisque $`𝕃_\gamma `$ est un fibré en droites au-dessus de $`G/P`$. Or, ce morphisme est la somme directe du morphisme identité $`𝒞𝒞`$ avec le morphisme $`ı^{}:\mathrm{R}\mathrm{\Gamma }(\overline{𝒬}_\mu ^{},𝒜_\mu ^{})(𝒜_\mu ^{})_{e_0}`$. Ce dernier est donc lui aussi un isomorphisme.
Pour $`i0`$, l’annulation $`\mathrm{H}^i(𝒜_\mu )_{e_0}=0`$ fait partie des propriétés qui caractérisent le complexe d’intersection $`𝒜_\mu `$. Cette annulation implique, via la suite exacte longue de cohomologie à support, que la flèche
$$\mathrm{H}_c^{i+d}(𝕃_\gamma ^\times )(d/2)\mathrm{H}_c^i(𝒬_\mu ^{},𝒜_\mu ^{})$$
est un isomorphisme dès que $`i>0`$. Par dualité de Poincaré, la flèche
$$\mathrm{H}^i(𝒬_\mu ^{},𝒜_\mu ^{})\mathrm{H}^{i+d}(𝕃_\gamma ^\times )(d/2)$$
est un isomorphisme pour $`i<0`$. Dans la suite exacte longue de Wang
$$\mathrm{H}^{i+d2}(G/P)(i/21)\stackrel{c_\gamma }{}\mathrm{H}^{i+d}(G/P)(i/2)\mathrm{H}^{i+d}(𝕃_\gamma ^\times )(i/2)$$
la flèche $`\mathrm{H}^{i+d2}(G/P)\stackrel{c_\gamma }{}\mathrm{H}^{i+d}(G/P)`$ est injective pour $`i0`$ d’après le théorème de Lefschetz difficile . On en déduit les suites exactes courtes
$$0\mathrm{H}^{i+d2}(G/P)(1)\stackrel{c_\gamma }{}\mathrm{H}^{i+d}(G/P)\mathrm{H}^{i+d}(𝕃_\gamma ^\times )0$$
pour $`i<0`$. Le lemme est démontré. $`\mathrm{}`$
###### Corollaire 8.2.
Soit $`𝒞`$ le facteur supporté par $`e_0`$ dans la décomposition
$$\mathrm{R}\pi _{\gamma ,}\overline{}_{\mathrm{}}[d](d/2)=𝒜_\mu 𝒞.$$
Pour $`i<0`$, on a
$$\mathrm{H}^i(𝒞)=\mathrm{H}^{i+d2}(G/P)(d/21).$$
Pour $`i0`$, on a
$$\mathrm{H}^i(𝒞)=\mathrm{H}^{i+d}(G/P)(d/2).$$
On peut maintenant démontrer l’énoncé 3.1 dans le cas où $`\lambda `$ est un cocaractère quasi-minuscule $`\mu =\gamma ^{}`$. Compte tenu de la discussion qui suit le lemme 5.2, il ne reste plus qu’à traiter le cas $`\nu =0`$.
###### Lemme 8.3.
On a un isomorphisme
$$\mathrm{R}\mathrm{\Gamma }_c(S_0,𝒜_\mu )\overline{}_{\mathrm{}}^{|\mathrm{\Delta }_\gamma |}.$$
Démonstration. D’après le théorème de changement de base pour un morphisme propre, on a
$$\mathrm{R}\mathrm{\Gamma }_c(\pi _\gamma ^1(S_0\overline{𝒬}_\mu ),\overline{}_{\mathrm{}})[d](d/2)=\mathrm{R}\mathrm{\Gamma }_c(S_0,𝒜_\mu )𝒞.$$
Rappelons la stratification obtenue en 7.5
$$\pi _\gamma ^1(S_0\overline{𝒬}_\mu )=\underset{\stackrel{wW/W_\gamma }{w\gamma R_{}}}{}\varphi _\gamma ^1(UwP/P)\underset{\stackrel{wW/W_\gamma }{w\gamma R_+}}{}ϵ_\gamma (UwP/P).$$
D’après le lemme 5.2, chaque strate $`S_{w\mu }\overline{𝒬}_\mu `$ est de dimension
$$\rho ,w\mu +\mu .$$
De plus, d’après le lemme 7.4, l’on a
$$S_{w\mu }\overline{𝒬}_\mu =\{\begin{array}{cc}\varphi _\gamma ^1(UwP/P)\text{ si }w\gamma R_+;\hfill & \\ ϵ_\gamma (UwP/P)\text{ si }w\gamma R_{}.\hfill & \end{array}$$
On en déduit que si $`w\gamma R_{}`$, alors la strate $`\varphi _\gamma ^1(UwP/P)`$ est un espace affine de dimension
$$\rho ,w\mu +\mu +1.$$
On a, dans ce cas, l’inégalité $`dim(\varphi _\gamma ^1(UwP/P))d/2`$ qui devient une égalité si et seulement si $`w\gamma `$ est l’opposé d’une racine simple.
D’autre part, si $`w\gamma R_+`$, alors la strate $`ϵ_\gamma (UwP/P)`$ est un espace affine de dimension
$$\rho ,w\mu +\mu 1.$$
On a, dans ce cas, l’inégalité $`dim(ϵ_\gamma (UwP/P))d/2`$ qui devient une égalité si et seulement si $`w\gamma `$ est une racine simple.
Pour $`i<0`$, on a donc
$$dim\mathrm{H}_c^{i+d}(\pi _\gamma ^1(S_0\overline{𝒬}_\mu ))=dim\mathrm{H}^i(𝒞)$$
et ces deux nombres valent
$$|\{wW/W_\gamma \rho ,w\mu +\mu =(i+d)/21\}|$$
Pour $`i>0`$, on a aussi
$$dim\mathrm{H}_c^{i+d}(\pi _\gamma ^1(S_0\overline{𝒬}_\mu ))=dim\mathrm{H}^i(𝒞)$$
et ces deux nombres valent
$$|\{wW/W_\gamma \rho ,w\mu +\mu =(i+d)/2+1\}|.$$
Pour $`i=0`$, on a
$$dim\mathrm{H}_c^d(\pi _\gamma ^1(S_0\overline{𝒬}_\mu ))=2|\mathrm{\Delta }_\gamma |$$
et
$$dim(\mathrm{H}^0(𝒞))=|\mathrm{\Delta }_\gamma |.$$
Le lemme s’en déduit. $`\mathrm{}`$
Démontrons maintenant l’énoncé 3.2 dans le cas $`\nu =0`$ et $`\lambda =\mu `$ quasi-minuscule. Nous démontrons en fait un énoncé un peu plus général. Rappelons que pour tout $`\sigma X^{}`$, on a défini un morphisme $`h_\sigma :S_0𝔾_a`$, voir 5.3.
###### Lemme 8.4.
Pour tout $`\sigma X_+^{}`$, on a un isomorphisme
$$\mathrm{R}\mathrm{\Gamma }_c(S_0,𝒜_\mu h_\sigma ^{}_\psi )=\overline{}_{\mathrm{}}^{|\mathrm{\Delta }_\gamma ^\sigma |},$$
$`\mathrm{\Delta }_\gamma ^\sigma `$ est l’ensemble des $`\alpha \mathrm{\Delta }_\gamma `$ telles que $`\alpha ,\sigma >0`$.
La démonstration de 8.4 suit le même schéma que celle du lemme 8.3 qui est, par ailleurs, un cas particulier de 8.4. Il suffit, en fait, de démontrer l’énoncé géométrique suivant.
###### Lemme 8.5.
1. Les restrictions de $`h_\sigma \pi _\gamma `$ aux strates $`ϵ_\gamma (UwP/P)`$ avec $`w\gamma R_+`$ sont nulles.
2. Les restrictions aux strates $`\varphi _\gamma ^1(UwP/P)`$ avec $`w\gamma R_{}`$ sont nulles aussi, à l’exception des $`\varphi _\gamma ^1(UwP/P)`$ telles que $`w\gamma `$ est une racine simple orthogonale à $`\sigma `$.
3. Les restrictions à ces dernières sont non nulles et linéaires par rapport à la structure restreinte du fibré en droites $`𝕃_\gamma `$.
Démonstration. La première assertion est évidente parce que toutes les strates $`ϵ_\gamma (UwP/P)`$ sont contenues dans $`\pi _\gamma ^1(e_0)`$. Pour les deux dernières assertions, il suffit naturellement de calculer les restrictions de $`h_\sigma `$ à
$$\varphi _\gamma ^1(UwP/P)ϵ_\gamma (UwP/P)=\varphi _\gamma ^1(UwP/P)ϵ_\gamma (UwP/P).$$
Pour tout $`wW`$, on a un isomorphisme
$$\varphi _\gamma ^1(UwP/P)ϵ_\gamma (UwP/P)=UwP/P\times 𝔾_m$$
puisque tout point $`y\varphi _\gamma ^1(UwP/P)ϵ_\gamma (UwP/P)`$ s’écrit d’une façon unique sous la forme
$$y=uwU_{\gamma ,1}(x)\varpi ^\mu e_0$$
avec $`x𝔾_m`$ et $`uUw^1U_\gamma ^+w`$, où $`U_\gamma ^+`$ est le radical unipotent du parabolique (positif) opposé à $`P`$.
En faisant passer $`w`$ vers la droite, on obtient
$$uwU_{\gamma ,1}(x)\varpi ^\mu e_0=uU_{w\gamma ,1}(x)\varpi ^{w\mu }e_0.$$
Posons $`t=\varpi x`$ et $`\alpha =w\gamma `$ et récrivons le membre de droite avec ces nouvelles notations
$$uU_{w\gamma }(\varpi x)\varpi ^{w\mu }e_0=uU_\alpha (t)t^\alpha ^{}e_0.$$
Rappelons la relation de Steinberg \[29, chap. 3, lemme 19\]
$$t^\alpha ^{}w_\alpha =U_\alpha (t)U_\alpha (t^1)U_\alpha (t)$$
qui vaut pour toute racine $`\alpha `$, pour tout $`t`$ inversible, et où le représentant $`w_\alpha G`$ est indépendant de $`t`$. On a donc
$$uU_\alpha (t)t^\alpha ^{}e_0=U_\alpha (t^1)U_\alpha (t)w_\alpha ^1e_0.$$
Or $`U_\alpha (t)w_\alpha ^1e_0=e_0`$ si bien que
$$uU_\alpha (t)t^\alpha ^{}e_0=uU_{\alpha ,1}(x^1)e_0.$$
Puisque $`uU`$, on a $`h(u)=0`$ de sorte que
$$h(uU_{\alpha ,1}(x^1))=h(U_{\alpha ,1}(x^1)).$$
Si $`\alpha `$ n’est pas une racine simple, on a $`h_\sigma (U_{\alpha ,1}(x^1))=0`$. Si $`\alpha \mathrm{\Delta }`$ mais $`\alpha ,\sigma >0`$, on a aussi $`h_\sigma (U_{\alpha ,1}(x^1))=0`$. Dans le cas où $`\alpha \mathrm{\Delta }`$ est une racine simple orthogonale à $`\sigma `$, on a $`h_\sigma (U_{\alpha ,1}(x^1))=x^1`$. $`\mathrm{}`$
## 9. Convolution
Rappelons que $`M`$ est l’ensemble des éléments minimaux dans $`X_+^{}\{0\}`$. Pour une suite $`\mu _{}=(\mu _1,\mathrm{},\mu _n)`$ d’éléments de $`M`$, on considère le sous-schéma projectif
$$\overline{𝒬}_\mu _{}=\overline{𝒬}_{\mu _1}\stackrel{~}{\times }\mathrm{}\stackrel{~}{\times }\overline{𝒬}_{\mu _n}$$
de $`𝒬^n`$. La projection sur le dernier facteur de $`𝒬^n`$ définit un morphisme propre
$$m_\mu _{}:\overline{𝒬}_\mu _{}\overline{𝒬}_{|\mu _{}|},$$
$`|\mu _{}|=\mu _1+\mathrm{}+\mu _n`$.
Soit $`\nu _{}=(\nu _1,\mathrm{},\nu _n)`$ une suite d’éléments de $`X^{}`$. Pour tout $`i=1,\mathrm{},n`$, posons $`\sigma _i=\nu _1+\mathrm{}+\nu _i`$. Notons $`S_\nu _{}\overline{𝒬}_\mu _{}`$ l’intersection
$$S_\nu _{}\overline{𝒬}_\mu _{}=S_{\sigma _1}\times \mathrm{}\times S_{\sigma _n}\overline{𝒬}_\mu _{}$$
dans $`𝒬^n`$. Il est clair que les $`S_\nu _{}\overline{𝒬}_\mu _{}`$ forment une stratification de $`\overline{𝒬}_\mu _{}`$.
###### Lemme 9.1.
On a un isomorphisme canonique
$$S_\nu _{}\overline{𝒬}_\mu _{}(S_{\nu _1}\overline{𝒬}_{\mu _1})\times \mathrm{}\times (S_{\nu _n}\overline{𝒬}_{\mu _n}).$$
Démonstration. On montre facilement par récurrence que tout point
$$(y_1,\mathrm{},y_n)S_\nu _{}\overline{𝒬}_\mu _{}$$
s’écrit de manière unique sous la forme
$$\begin{array}{ccc}\hfill y_1& =& x_1\varpi ^{\nu _1}e_0\hfill \\ & \mathrm{}& \\ \hfill y_n& =& x_1\varpi ^{\nu _1}\mathrm{}x_n\varpi ^{\nu _n}e_0\hfill \end{array}$$
avec $`x_iL^{<\nu _i}U`$ tel que $`x_i\varpi ^{\nu _i}e_0\overline{𝒬}_{\mu _i}`$. Le lemme s’en déduit. $`\mathrm{}`$
Pour que la strate $`S_\nu _{}\overline{𝒬}_\mu _{}`$ soit non vide, il est donc nécessaire que, pour tout $`i=1,\mathrm{},n`$, $`\nu _i`$ appartienne à $`\mathrm{\Omega }(\mu _i)`$.
###### Corollaire 9.2.
Soient $`\mu _1,\mathrm{},\mu _n`$ des éléments de $`M`$. Pour toute suite $`\nu _{}`$ avec $`\nu _i\mathrm{\Omega }(\mu _i)`$, toutes les composantes de $`S_\nu _{}\overline{𝒬}_\mu _{}`$ sont de dimension $`\rho ,|\nu _{}|+|\mu _{}|`$.
Démonstration. D’après le lemme 5.2 et le corollaire 7.5, chaque $`S_{\nu _i}\overline{𝒬}_{\mu _i}`$ est purement de dimension $`\rho ,\nu _i+\mu _i`$. Le corollaire découle donc du lemme précédent. $`\mathrm{}`$
En fait, pour $`\lambda X_+^{}`$ et $`\nu \mathrm{\Omega }(\lambda )`$ arbitraires, $`S_\nu \overline{𝒬}_\mu `$ est purement de dimension $`\rho ,\nu +\mu `$. Ce résultat est énoncé dans \[22, 4.5\] avec seulement quelques indications de démonstration. Nous avons pu en établir une autre démonstration, en utilisant l’approche en termes de représentations d’algèbres de Lie affines. Signalons aussi qu’on peut déduire cette formule de dimension, sans l’assertion de pure dimension, du théorème 3.1.
L’énoncé suivant est aussi un cas particulier d’un autre lemme énoncé dans \[22, 2.6\]. Nous proposons ici une démonstration un peu différente.
###### Lemme 9.3.
Pour tout $`\lambda X_+^{}`$ avec $`\lambda |\mu _{}|`$, on a
$$dim(m_\mu _{}^1(𝒬_\lambda ))\rho ,\lambda +|\mu _{}|.$$
Autrement dit, le morphisme $`m_\mu _{}`$ est semi-petit.
Démonstration. Puisque $`S_\mu 𝒬_\mu `$ est un ouvert dense de $`𝒬_\mu `$, d’après 5.1 et 5.2, et puisque les fibres de $`m_\mu _{}`$ sont toutes isomorphes les unes aux autres au-dessus de l’orbite $`𝒬_\mu `$, on a
$$dim(m_\mu _{}^1(𝒬_\mu ))=dim(m_\mu _{}^1(S_\mu 𝒬_\mu )).$$
Par conséquent, on a
$$dim(m_\mu _{}^1(𝒬_\mu ))dim(m_\mu _{}^1(S_\mu \overline{𝒬}_\lambda )).$$
Or, on a une stratification
$$m_\mu _{}^1(S_\mu \overline{𝒬}_\lambda )=\underset{|\nu _{}|=\lambda }{}S_\nu _{}\overline{𝒬}_\mu _{},$$
où toutes les strates sont de dimension $`\rho ,\lambda +|\mu _{}|`$, d’où le lemme. $`\mathrm{}`$
###### Proposition 9.4.
Le produit de convolution $`𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n}`$ est un faisceau pervers. Il se décompose en somme directe
$$𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n}=\underset{\lambda |\mu _{}|}{}𝒜_\lambda V_\mu _{}^\lambda ,$$
où les $`V_\mu _{}^\lambda `$ sont des $`\overline{}_{\mathrm{}}`$-espaces vectoriels dont la dimension vaut le nombre de composantes irréductibles de $`m_\mu _{}^1(S_\lambda \overline{𝒬}_\mu _{})`$ qui sont entièrement contenues dans $`m_\mu _{}^1(S_\lambda \overline{𝒬}_\lambda )`$.
Démonstration. Si les $`\mu _i`$ sont tous minuscules, le schéma source est lisse. Comme $`m_\mu _{}`$ est semi-petit, l’image directe $`\mathrm{R}(m_\mu _{}){}_{}{}^{}\overline{}_{\mathrm{}}^{}[dim(\overline{𝒬}_\mu _{})]`$ est perverse.
En général, on a la stratification du schéma source
$$\overline{𝒬}_\mu _{}=\underset{\mu _i^{}\mu _i}{}𝒬_{\mu _1^{}}\stackrel{~}{\times }\mathrm{}\stackrel{~}{\times }𝒬_{\mu _n^{}},$$
où, comme $`\mu _iM`$, chaque $`\mu _i^{}`$ est ou bien égal à $`\mu _i`$, ou bien égal à $`0`$. Dans tous les cas, le lemme précédent s’applique encore à $`\overline{𝒬}_\mu _{}^{}`$ et nous permet d’obtenir l’inégalité
$$dim(m_\mu _{}^1(𝒬_\lambda )\overline{𝒬}_\mu _{}^{})\rho ,\lambda +|\mu _{}^{}|$$
pour tout $`\lambda |\mu _{}^{}|`$.
Par ailleurs, $`\overline{𝒬}_\mu _{}`$ muni de la stratification par les $`𝒬_{\mu _1^{}}\stackrel{~}{\times }\mathrm{}\stackrel{~}{\times }𝒬_{\mu _n^{}}`$, est localement isomorphe à $`\overline{𝒬}_{\mu _1}\times \mathrm{}\times \overline{𝒬}_{\mu _1}`$ muni de la stratification par les $`𝒬_{\mu _1^{}}\times \mathrm{}\times 𝒬_{\mu _n^{}}`$, voir la section 2. Par conséquent, pour $`\mu _{}^{}<\mu _{}`$,
$$\mathrm{H}^i(\mathrm{IC}(\overline{𝒬}_\mu _{})|_{𝒬_{\mu _1^{}}\stackrel{~}{\times }\mathrm{}\stackrel{~}{\times }𝒬_{\mu _n^{}}})$$
s’annule dès que $`i2\rho ,|\mu _{}^{}|`$.
Il résulte de ces deux dernières assertions que $`\mathrm{R}(m_\mu _{}){}_{}{}^{}\mathrm{IC}(\overline{𝒬}_\mu _{})`$ est un faisceau pervers. Sa décomposition , a priori sur $`\overline{k}`$, doit avoir la forme
$$𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n}=\underset{\lambda |\mu _{}|}{}𝒜_\lambda V_\mu _{}^\lambda $$
où les $`V_\mu _{}^\lambda `$ sont des espaces vectoriels de dimension finie sur $`\overline{}_{\mathrm{}}`$, parce que tous ses facteurs directs sont aussi $`L^0G`$-équivariants.
Les espaces vectoriels $`V_\mu _{}^\lambda `$ admettent une base canonique indexée par les composantes irréductibles de $`m_\mu _{}^1(𝒬_\lambda )`$ de dimension exactement $`\rho ,|\mu _{}|\lambda `$. D’après la démonstration du lemme 9.3, celles-ci correspondent bijectivement aux composantes irréductibles de $`m_\mu _{}^1(S_\lambda \overline{𝒬}_\mu _{})`$ qui sont entièrement contenues dans $`m_\mu _{}^1(S_\lambda \overline{𝒬}_\lambda )`$.
Puisque ces composantes sont toutes définies sur $`k`$, la décomposition est en fait valable sur $`k`$. $`\mathrm{}`$
Soit $`\mu _{}`$ comme précédemment, c.à.d. $`\mu _{}`$ est une suite $`(\mu _1,\mathrm{},\mu _n)`$ d’éléments de $`M`$. A la suite de Littelmann , on appellera $`\mu _{}`$-chemin une donnée combinatoire $`\chi `$ du type suivant
* une suite de sommets $`\sigma _1,\mathrm{},\sigma _n`$ dans $`X^{}`$ tels que pour tout $`i=1,\mathrm{},n`$, on a $`\nu _i=\sigma _i\sigma _{i1}\mathrm{\Omega }(\mu _i)`$ ;
* des applications
$$p_i:[0,1]X^{}_{}$$
vérifiant
+ si $`\sigma _{i1}\sigma _i`$, alors
$$p_i(t)=(1t)\sigma _{i1}+t\sigma _i$$
+ si $`\sigma _{i1}=\sigma _i`$, alors
$$p_i(t)=\{\begin{array}{cc}\sigma _{i1}t\alpha _i^{}\hfill & \mathrm{pour}0t1/2,\hfill \\ \sigma _{i1}+(t1)\alpha _i^{}\hfill & \mathrm{pour}1/2t1,\hfill \end{array}$$
$`\alpha _i^{}`$ est une coracine simple conjuguée à $`\mu _i`$.
En mettant bout à bout les images des $`p_i`$, on obtient un chemin dans $`X^{}_{}`$ allant de $`0`$ à $`\sigma _n`$. Le $`\mu _{}`$-chemin $`\chi `$ est dit dominant s’il est entièrement contenu dans la chambre dominante $`(X^{}_{})_+`$.
D’après le lemme 5.2, chaque $`S_{w\mu _i}\overline{𝒬}_{\mu _i}`$ est irréductible. De plus, d’après le corollaire 7.5, si $`\mu _i`$ est quasi-minuscule, disons $`\mu _i=\gamma _i^{}`$, et si $`\nu =0`$, alors l’ensemble des composantes irréductibles de $`S_0\overline{𝒬}_{\mu _i}`$ est en bijection canonique avec l’ensemble $`\mathrm{\Delta }_\gamma `$ des racines simples $`\alpha `$ conjuguées à $`\gamma `$. Compte tenu du lemme 9.1, pour tout $`\nu \mathrm{\Omega }(|\mu _{}|)`$, l’ensemble des composantes irréductibles de $`\pi ^1(S_\nu \overline{𝒬}_{|\mu _{}|})`$ est en bijection canonique avec l’ensemble des $`\mu _{}`$-chemins $`\chi `$ allant de $`0`$ à $`\nu `$. Notons $`C_\chi `$ la composante correspondante à $`\chi `$.
###### Lemme 9.5.
Soient $`\nu \mathrm{\Omega }(|\mu _{}|)`$ dominant et $`\chi `$ un $`\mu _{}`$-chemin dominant allant de $`0`$ à $`\nu `$. Alors la composante $`C_\chi `$ est contenue dans $`\pi ^1(S_\nu \overline{𝒬}_\nu )`$.
Démonstration. Notons $`I(\chi )`$ l’ensemble des indices $`i=1,\mathrm{},n`$ tels que $`\sigma _{i1}=\sigma _i`$.
Si $`iI(\chi )`$, $`\nu _i`$ est non nul et est donc conjugué à $`\mu _i`$. D’après le lemme 5.2, un point $`p_iS_{\nu _i}\overline{𝒬}_{\mu _i}`$ s’écrit de manière unique sous la forme $`p_i=u_i\varpi ^{\nu _i}e_0`$ avec $`u_iwJ^{\mu _i}w^1LU`$. En particulier, $`u_iL^0U`$.
Si $`iI(\chi )`$, alors $`\mu _i`$ est quasi-minuscule, disons $`\mu _i=\gamma _i^{}`$, et l’hypothèse $`\chi `$ dominant implique $`\alpha _i,\sigma _{i1}1`$. D’après le corollaire 7.5, la composante irréductible de $`S_0\overline{𝒬}_{\gamma _i^{}}`$ correspondant à $`\alpha _i=w_i\gamma _i`$, contient comme ouvert dense le $`𝔾_m`$-torseur trivial
$$\varphi _{\gamma _i}^1(Uw_iP_i/P_i)ϵ_{\gamma _i}^1(Uw_iP_i/P_i)$$
au-dessus de $`Uw_iP_i/P_i`$. D’après la démonstration du lemme 8.5, pour $`iI(\chi )`$ chaque point
$$p_i\varphi _{\gamma _i}^1(Uw_iP_i/P_i)ϵ_{\gamma _i}^1(Uw_iP_i/P_i)$$
s’écrit de manière unique sous la forme $`uU_{\alpha _i,1}(x)e_0`$ avec $`uUw^1U_{\gamma _i}^+w`$ et $`x𝔾_m`$. Ici, $`U_{\gamma _i}^+`$ est le radical unipotent du sous-groupe parabolique opposé à $`P_i`$. Posons $`u_i=uU_{\alpha _i,1}(x)`$.
Dans ce cas, $`u_iL^{<0}U`$. Toutefois, l’unicité de l’expression $`p_i=u_ie_0`$ suffit pour l’argument donné dans la démonstration du lemme 9.1. Le morphisme qui envoie le point
$$(p_1,\mathrm{},p_n)\underset{iI(\chi )}{}(S_{\nu _i}\overline{𝒬}_{\mu _i})\underset{iI(\chi )}{}(\varphi _{\gamma _i}^1(Uw_iP_i/P_i)ϵ_{\gamma _i}^1(Uw_iP_i/P_i)),$$
sur le point
$$(y_1,\mathrm{},y_n)S_\nu _{}\overline{𝒬}_\mu _{}$$
défini par
$$\begin{array}{ccc}\hfill y_1& =& u_1\varpi ^{\nu _1}e_0\hfill \\ & \mathrm{}& \\ \hfill y_n& =& u_1\varpi ^{\nu _1}u_2\varpi ^{\nu _2}\mathrm{}u_n\varpi ^{\nu _n}e_0.\hfill \end{array}$$
induit un isomorphisme du schéma source sur un ouvert dense de $`C_\chi `$.
Pour $`iI(\chi )`$, on a $`u_iL^0U`$ de sorte que $`\varpi ^{\sigma _{i1}}u_i\varpi ^{\sigma _{i1}}`$ appartient aussi à $`L^0U`$ vu que $`\sigma _{i1}`$ est dominant.
Pour $`iI(\chi )`$, on a $`u_i=uU_{\alpha _i,1}(x)`$ et par conséquent,
$$\varpi ^{\sigma _{i1}}u_i\varpi ^{\sigma _{i1}}=\varpi ^{\sigma _{i1}}u\varpi ^{\sigma _{i1}}U_{\alpha _i,\alpha _i,\sigma _{i1}1}(x).$$
Cet élément appartient aussi à $`L^0U`$ parce que $`\alpha _i,\sigma _{i1}1`$.
Il s’ensuit que $`y_nS_\nu \overline{𝒬}_\nu `$ de sorte qu’on a un ouvert dense de $`C_\chi `$ contenu dans $`m_\mu _{}^1(S_\nu \overline{𝒬}_\nu )`$. Or, $`S_\nu \overline{𝒬}_\nu `$ est fermé dans $`S_\nu \overline{𝒬}_{|\mu _{}|}`$ si bien que la composante $`C_\chi `$ toute entière est contenue dans $`m_\mu _{}^1(S_\nu \overline{𝒬}_\nu )`$. $`\mathrm{}`$
Il n’est pas difficile de démontrer qu’inversement, si le $`\mu _{}`$-chemin $`\chi `$ n’est pas dominant, alors $`C_\chi \pi ^1(S_\nu 𝒬_\nu )`$. Nous laissons cette assertion aux soins du lecteur car elle n’est pas logiquement nécessaire pour la suite. Il nous suffira seulement de savoir que la multiplicité $`dim(V_{|\mu _{}|}^\nu )`$ de $`𝒜_\nu `$ dans $`𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n}`$ vaut, au moins, le nombre de $`\mu _{}`$-chemins dominants allant de $`0`$ à $`\nu `$.
###### Proposition 9.6.
Pour tout $`\lambda X_+^{}`$, $`𝒜_\lambda `$ est facteur direct d’un produit de convolution de la forme
$$𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n},$$
avec $`\mu _1,\mathrm{},\mu _nM`$.
Compte tenu de 9.4 et de 9.5, il suffit de démontrer qu’il existe un $`\mu _{}`$-chemin dominant allant de $`0`$ à $`\nu `$. On démontrera cet énoncé combinatoire dans la section 10.
Signalons le corollaire suivant dont on ne se servira pas dans la suite de l’article. Cet énoncé se trouve déjà dans et .
###### Corollaire 9.7.
Pour tous $`\lambda ,\lambda ^{}X_+^{}`$, le produit de convolution $`𝒜_\lambda 𝒜_\lambda ^{}`$ est un faisceau pervers.
Démonstration. Si $`𝒜_\lambda `$, resp. $`𝒜_\lambda ^{}`$, est un facteur direct de $`𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n}`$, resp. de $`𝒜_{\mu _1^{}}\mathrm{}𝒜_{\mu _n^{}^{}}`$, alors $`𝒜_\lambda 𝒜_\lambda ^{}`$ est un facteur direct de
$$𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n}𝒜_{\mu _1^{}}\mathrm{}𝒜_{\mu _n^{}^{}},$$
qui est pervers d’après 9.4. $`\mathrm{}`$
Cette technique permet aussi de généralier la démonstration de \[24, cor. 4.3.2\] à tout groupe réductif. Cet énoncé a été démontré auparavant par Ginzburg, Mirkovic et Vilonen , du moins lorsque le corps $`k`$ est le corps des nombres complexes.
## 10. Combinatoire
Nous proposons deux démonstrations de 9.6. L’une repose sur un lemme sur les systèmes de racines, qui paraît intéressant en soi. L’autre est basée sur la théorie des représentations et le modèle des chemins de Littelmann ; elle présente des similitudes remarquables avec des résultats géométriques des sections 8 et 9, et de cette manière, est une bonne illustration de l’équivalence tannakienne de et .
Preuve combinatoire. Rappelons que $`M`$ désigne l’ensemble des cocaractères minuscules et quasi-minuscules. Si $`\mu =\gamma ^{}`$ est quasi-minuscule, $`\mathrm{\Delta }_\gamma `$ désigne l’ensemble des racines simples conjuguées à $`\gamma `$.
Compte tenu de 9.4 et 9.5, la proposition 9.6 découle de l’énoncé suivant.
###### Lemme 10.1.
Soit $`\lambda X_+^{}`$. Si $`\lambda M`$, il existe une coracine courte $`\beta ^{}`$ telle que $`\lambda \beta ^{}X_+^{}`$.
Démonstration. Comme $`\lambda M`$, il résulte du lemme 1.1 qu’il existe une racine $`\alpha `$ telle que $`\alpha ,\lambda 2`$. Comme $`\lambda `$ est dominant, ceci entraîne que $`\beta ,\lambda 2`$, où $`\beta `$ est la racine maximale (longue!) du sous-système de racines irréductible de $`R`$ contenant $`\alpha `$. On observe que $`\lambda \beta ^{}`$ est un poids de $`V(\lambda )`$, donc appartient à $`\mathrm{\Omega }(\lambda )`$.
Soit $`(,)`$ un produit scalaire $`W`$-invariant sur $`X^{}_{}`$, normalisé par la condition que $`(\alpha ^{},\alpha ^{})=2`$ pour toute coracine courte $`\alpha ^{}`$. Pour tout $`\chi X^{}`$, on posera $`|\chi |^2:=(\chi ,\chi )`$.
Alors, un calcul facile montre que $`|\lambda i\beta ^{}|<|\lambda |`$ pour $`0<i<\beta ,\lambda `$. Par conséquent, comme $`\beta ,\lambda 2`$, on obtient que $`\lambda \beta ^{}`$ n’appartient pas à $`W\lambda `$. Notons $`\lambda ^{}`$ le cocaractère dominant conjugué à $`\lambda \beta ^{}`$, il appartient aussi à $`\mathrm{\Omega }(\lambda )`$. On a ainsi
$$\lambda \beta ^{}\lambda ^{}<\lambda ,$$
et donc $`\lambda ^{}=\lambda \eta `$ et $`\beta ^{}=\eta +\nu `$, avec $`\eta ,\nu Q_+^{}`$ et $`\eta 0`$.
Alors, pour tout $`\alpha R`$, on a
$$(\lambda ,\alpha ^{})=\alpha ,\lambda (\alpha ^{},\alpha ^{})/2.$$
et par conséquent, $`(\lambda ,\delta )0`$ pour tout $`\delta Q_+^{}`$.
Alors, de l’égalité
$$|\lambda \beta ^{}|^2=|\lambda \eta |^2,$$
on déduit que
$$|\beta ^{}|^2|\eta |^2=2(\lambda ,\nu )0,$$
d’où $`|\eta |^2|\beta ^{}|^2=2`$. Ceci entraîne que $`\eta `$ est une coracine courte, et le lemme est démontré. $`\mathrm{}`$
Preuve basée sur la théorie des représentations. Voici un cas très particulier et bien connu de la règle de Littlewood-Richardson, voir pour le cas général. On rappelle que, pour $`\lambda X_+^{}`$, $`V(\lambda )`$ désigne le module simple de plus haut poids $`\lambda `$ pour le groupe $`G^{}`$ défini sur $`\overline{}_{\mathrm{}}`$.
###### Lemme 10.2.
Soient $`\mu M`$ et $`\lambda X_+^{}`$.
1. Si $`\mu `$ est minuscule, alors
$$V(\mu )V(\lambda )=\underset{\stackrel{\nu W\mu }{\nu +\lambda X_+^{}}}{}V(\lambda +\nu ).$$
2. Si $`\mu `$ est quasi-minuscule, alors
$$V(\mu )V(\lambda )=\underset{\stackrel{\nu W\mu }{\nu +\lambda X_+^{}}}{}V(\lambda +\nu )\underset{\stackrel{\alpha \mathrm{\Delta }_\gamma }{\lambda +\frac{1}{2}\alpha ^{}X_+^{}}}{}V(\lambda ).$$
Démonstration. Ceci est bien connu, voir par exemple \[14, Lemma 5A.9\] ou \[8, 4.2.1\] pour le point 1) et \[25, 3.7-3.8\] pour le point 2).
En utilisant le modèle des chemins de Littelmann , on peut aussi argumenter comme suit. D’après loc. cit., le $`G^{}`$-module $`V(\lambda )`$ admet une base paramétrée par certains chemins. En particulier, pour $`\mu `$ minuscule, $`V(\mu )`$ admet une base $`\{v_{p_{w\mu }}\}_{wW/W_\mu }`$$`p_{w\mu }`$ est le chemin défini par
$$p_{w\mu }(t)=tw\mu ,\mathrm{pour}\mathrm{tout}0t1;$$
le poids de $`v_{p_{w\mu }}`$ étant $`p_{w\mu }(1)=w\mu `$.
Pour $`\mu =\gamma ^{}`$ quasi-minuscule, $`V(\mu )`$ admet une base
$$\{v_{p_{w\mu }}\}_{wW/W_\mu }\{v_{p_\alpha }\}_{\alpha \mathrm{\Delta }_\gamma },$$
où pour toute racine simple $`\alpha \mathrm{\Delta }_\gamma `$, $`p_\alpha `$ est le chemin
$$p_\alpha (t)=\{\begin{array}{cc}t\alpha ^{}\hfill & \mathrm{pour}\mathrm{tout}0t1/2;\hfill \\ (t1)\alpha ^{}\hfill & \mathrm{pour}\mathrm{tout}1/2t1;\hfill \end{array}$$
le poids de $`v_{p_\alpha }`$ étant $`p_\alpha (1)=0`$.
D’après loc. cit., $`V(\mu )V(\lambda )`$ est la somme directe des $`V(\lambda +\chi (1))`$, où $`\chi `$ parcourt l’ensemble des chemins dans $`V(\mu )`$ tels que le translaté $`\lambda +\chi ([0,1])`$ soit entièrement contenu dans la chambre dominante. Pour les chemins $`p_{w\mu }`$, cela équivaut à la condition $`\lambda +w\mu X_+^{}`$. Pour les chemins $`p_\alpha `$ avec $`\alpha \mathrm{\Delta }_\gamma `$, cela équivaut à la condition $`\lambda +\frac{1}{2}\alpha ^{}X_+^{}`$. $`\mathrm{}`$
Il résulte de ce lemme que $`V(\lambda )`$ est un facteur direct d’un produit tensoriel $`V(\mu _1)\mathrm{}V(\mu _n)`$ si et seulement s’il existe un $`\mu _{}`$-chemin dominant allant de $`0`$ à $`\lambda `$. Il suffit donc de démontrer le lemme suivant.
###### Lemme 10.3.
Pour tout $`\lambda X_+^{}`$, $`V(\lambda )`$ est facteur direct d’un produit tensoriel de la forme $`V(\mu _1)\mathrm{}V(\mu _n)`$ avec $`\mu _1,\mathrm{}\mu _nM`$.
Démonstration. Démontrons d’abord que la représentation
$$\rho _M:G\underset{\mu M}{}EndV(\mu )$$
est fidèle. D’abord, il est bien connu, et facile de voir, que pour tout $`\xi X_+^{}`$, le sous-groupe de $`X^{}=\mathrm{Hom}(T^{},𝔾_m)`$ engendré par les poids de $`V(\xi )`$ est le sous-groupe engendré par $`Q^{}`$ et $`\xi `$. D’autre part, on déduit de \[3, Chap.VI, Ex.2.5\], que $`M`$ contient un système de représentants de $`X^{}/Q^{}`$. Il en résulte que la restriction de $`\rho _M`$ au tore maximal $`T^{}`$ est fidèle, et donc que $`\rho _M`$ est fidèle.
On en déduit que l’homomorphisme d’algèbres
$$\mathrm{Sym}(\underset{\mu M}{}V(\mu )V(\mu )^{})\overline{}_{\mathrm{}}[G^{}],$$
$`\overline{}_{\mathrm{}}[G^{}]`$ désigne l’algèbre des fonctions régulières sur $`G^{}`$, est surjectif. D’après le théorème de Peter-Weyl, tout module $`V(\lambda )`$ intervient comme facteur direct de l’algèbre $`\overline{}_{\mathrm{}}[G^{}]`$, d’où le lemme. $`\mathrm{}`$
## 11. Fin des démonstrations
On conserve les notations de la section 9. En particulier, soient $`\lambda X_+^{}`$, $`\nu \mathrm{\Omega }(\lambda )`$ et $`\mu _{}=(\mu _1,\mathrm{},\mu _n)`$ une suite d’éléments de $`M`$ telle que $`𝒜_\lambda `$ soit un facteur direct de $`𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n}`$ (c.f. Proposition 9.6).
Preuve du théorème 3.1. Compte-tenu des hypothèses ci-dessus, pour démontrer 3.1, il suffit de démontrer que le complexe
$$\mathrm{R}\mathrm{\Gamma }_c(S_\nu ,𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n})$$
est concentré en degré $`2\rho ,\nu `$ et que l’endomorphisme de Frobenius $`\mathrm{Fr}_q`$ agit dans $`\mathrm{H}_c^{2\rho ,\nu }(S_\nu ,𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n})`$ comme la multiplication par $`q^{\rho ,\nu }`$. D’après le théorème de changement de base pour un morphisme propre, on a
$$\mathrm{R}\mathrm{\Gamma }_c(S_\nu ,𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n})=\mathrm{R}\mathrm{\Gamma }_c(m_\mu _{}^1(S_\nu \overline{𝒬}_{|\mu _{}|}),\mathrm{IC}(\overline{𝒬}_\mu _{})).$$
Rappelons qu’on a la stratification
$$m_\mu _{}^1(S_\nu \overline{𝒬}_{|\mu _{}|})=\underset{|\nu _{}|=\nu }{}S_\nu _{}\overline{𝒬}_\mu _{}$$
et, d’après le lemme 9.1, on a un isomorphisme
$$S_\nu _{}\overline{𝒬}_\mu _{}(S_{\nu _1}\overline{𝒬}_{\mu _1})\times \mathrm{}\times (S_{\nu _n}\overline{𝒬}_{\mu _n}),$$
$`\nu _{}=(\nu _1,\mathrm{},\nu _n)`$. De plus, cet isomorphisme est induit par l’isomorphisme provenant de la locale trivialité
$$\begin{array}{c}(\varpi ^{\nu _1}L^{<0}Ge_0\overline{𝒬}_{\mu _1})\times \mathrm{}\times (\varpi ^{\nu _n}L^{<0}Ge_0\overline{𝒬}_{\mu _n})\\ (\varpi ^{\nu _1}L^{<0}Ge_0\overline{𝒬}_{\mu _1})\stackrel{~}{\times }\mathrm{}\stackrel{~}{\times }(\varpi ^{\nu _n}L^{<0}Ge_0\overline{𝒬}_{\mu _n})\end{array}$$
si bien qu’on a
$$\mathrm{R}\mathrm{\Gamma }_c(S_\nu _{}\overline{𝒬}_\mu _{},\mathrm{IC}(\overline{𝒬}_\mu _{}))=\underset{i=1}{\overset{n}{}}\mathrm{R}\mathrm{\Gamma }_c(S_{\nu _i}\overline{𝒬}_{\mu _i},𝒜_{\mu _i}).$$
L’assertion à démontrer résulte maintenant de 5.2 et de 8.4. $`\mathrm{}`$
Preuve du théorème 3.2. Rappelons que le cas plus facile $`\nu =\lambda `$ a été démontré dans la discussion qui suit le lemme 5.2. On démontre maintenant le cas plus difficile $`\nu \lambda `$.
La suite $`\mu _{}`$ a été choisie de sorte que la multiplicité $`V_\mu _{}^\lambda `$ de $`𝒜_\lambda `$ dans la décomposition 9.4 :
$$𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n}=\underset{\stackrel{\xi X_+^{}}{\xi \mu _1+\mathrm{}+\mu _n}}{}𝒜_\xi V_\mu _{}^\xi $$
est non nulle. On déduit de cette décomposition l’égalité
$$\begin{array}{cc}& \mathrm{R}\mathrm{\Gamma }_c(S_\nu ,𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n}h^{}_\psi )\hfill \\ =& \underset{\stackrel{\xi X_+^{}}{\xi \mu _1+\mathrm{}+\mu _n}}{}\mathrm{R}\mathrm{\Gamma }_c(S_\nu ,𝒜_\xi h^{}_\psi )V_\mu _{}^\xi .\hfill \end{array}$$
Du fait que $`V_\mu _{}^\lambda 0`$ et que $`\lambda \nu `$, pour démontrer que
$$\mathrm{R}\mathrm{\Gamma }_c(S_\nu ,𝒜_\lambda h^{}_\psi )=0$$
il suffit de démontrer que la flèche facteur direct
$$\begin{array}{cc}& \mathrm{R}\mathrm{\Gamma }_c(S_\nu ,𝒜_\nu h^{}_\psi )V_\mu _{}^\nu \hfill \\ & \mathrm{R}\mathrm{\Gamma }_c(S_\nu ,𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n}h^{}_\psi )\hfill \end{array}$$
est un quasi-isomorphisme.
Or, d’après la discussion qui suit le lemme 5.2, on sait que
$$\mathrm{R}\mathrm{\Gamma }_c(S_\nu ,𝒜_\nu h^{}_\psi )V_\mu _{}^\nu =V_\mu _{}^\nu [2\rho ,\nu ](\rho ,\nu ).$$
Il suffit par conséquent de démontrer que pour $`i2\rho ,\nu `$, on a
$$\mathrm{H}_c^i(S_\nu ,𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n}h^{}_\psi )=0$$
et que pour $`i=2\rho ,\nu `$, on a
$$dim(V_\mu _{}^\nu )dim(H_c^i(S_\nu ,𝒜_{\mu _1}\mathrm{}𝒜_{\mu _n}h^{}_\psi )).$$
Rappelons qu’on a la stratification
$$m_\mu _{}^1(S_\nu \overline{𝒬}_{|\mu _{}|})=\underset{|\nu _{}|=\nu }{}S_\nu _{}\overline{𝒬}_\mu _{}$$
et que chaque point
$$(y_1,\mathrm{},y_n)S_\nu _{}\overline{𝒬}_\mu _{}$$
s’écrit de manière unique sous la forme
$$\begin{array}{ccc}\hfill y_1& =& x_1\varpi ^{\nu _1}e_0\hfill \\ & \mathrm{}& \\ \hfill y_n& =& x_1\varpi ^{\nu _1}\mathrm{}x_n\varpi ^{\nu _n}e_0\hfill \end{array}$$
avec $`x_iL^{<\nu _i}U`$ tels que $`x_i\varpi ^{\nu _i}e_0\overline{𝒬}_{\mu _i}`$. Pour $`\sigma X^{}`$, notons $`h_\sigma :LU𝔾_a`$ le morphisme défini par $`h_\sigma (x)=h(\varpi ^\sigma x\varpi ^\sigma )`$ ainsi que ses restriction aux $`L^{<\nu }U`$ et $`S_\nu `$. Il est clair que
$$h(y_n)=h(x_1)+h_{\sigma _1}(x_2)+\mathrm{}+h_{\sigma _{n1}}(x_n).$$
Joint à l’argument de locale trivialité déjà utilisé dans la preuve de 3.1, on obtient l’égalité
$$\begin{array}{cc}& \mathrm{R}\mathrm{\Gamma }_c((S_\nu _{}𝒬_\mu _{}),\mathrm{IC}(\overline{𝒬}_\mu _{})h^{}_\psi )\hfill \\ =& _{i=1}^n\mathrm{R}\mathrm{\Gamma }_c((S_{\nu _i}𝒬_{\mu _i}),𝒜_{\mu _i}h_{\sigma _{i1}}^{}_\psi ).\hfill \end{array}$$
###### Lemme 11.1.
Si $`\sigma X_+^{}`$, alors on a
$$\mathrm{R}\mathrm{\Gamma }_c(S_\nu ,𝒜_\lambda h_\sigma ^{}_\psi )=0.$$
Démonstration. Soit $`\alpha \mathrm{\Delta }`$ une racine simple telle que $`\alpha ,\sigma `$ soit strictement négatif. Le sous-groupe $`𝔾_a=U_{\alpha ,\alpha ,\sigma 1}`$ est alors contenu dans $`L^0U`$, donc agit de manière équivariante sur le couple $`(S_\nu ,𝒜_\lambda )`$. Or, la restriction de $`h_\sigma `$ à ce sous-groupe induit l’identité de $`𝔾_a`$. Il suffit maintenant d’appliquer \[23, lemme 3.3\]. $`\mathrm{}`$
On en déduit l’annulation
$$\mathrm{R}\mathrm{\Gamma }_c((S_\nu _{}𝒬_\mu _{}),\mathrm{IC}(\overline{𝒬}_\mu _{})h^{}_\psi )=0$$
pour les suites $`\nu _{}`$ dont au moins une des sommes partielles $`\sigma _i`$ n’est pas dominante.
Soit maintenant $`\nu _{}`$ une suite avec $`\nu _i\mathrm{\Omega }(\mu _i)`$ telle que toutes les sommes partielles $`\sigma _i=\nu _1+\mathrm{}+\nu _i`$ sont dominantes. On dira qu’un $`\mu _{}`$-chemin est de type $`\nu _{}`$ s’il a pour sommets $`0,\sigma _1,\mathrm{},\sigma _n`$. Observons que la condition $`\alpha ,\sigma 1`$ apparaissant dans le lemme 8.4 équivaut à la condition que $`\alpha ^{}/2+\sigma `$ soit dominant.
En mettant ensemble les lemmes 5.3 et 8.4, on arrive à l’assertion suivante. Pour $`i2\rho ,\nu `$, on a
$$\mathrm{H}_c^i(S_\nu _{}\overline{𝒬}_\mu _{},\mathrm{IC}(\overline{𝒬}_\mu _{})h^{}_\psi )=0$$
et pour $`i=2\rho ,\nu `$, on a
$$\begin{array}{cc}& dim(\mathrm{H}_c^i(S_\nu _{}\overline{𝒬}_\mu _{},\mathrm{IC}(\overline{𝒬}_\mu _{})h^{}_\psi ))\hfill \\ =& |\{\mu _{}\text{-chemins dominants de type }\nu _{}\}|.\hfill \end{array}$$
Par ailleurs, compte tenu de 9.4 et de 9.5, on a l’inégalité
$$dim(V_\mu _{}^\nu )|\{\mu _{}\text{-chemins dominants allant de 0 à }\nu \}|.$$
La démonstration du théorème 3.2 est terminée. $`\mathrm{}`$
CNRS, UMR 7539, LAGA,
Institut Galilée,
Université Paris-Nord,
93430 Villetaneuse, France.
Courriers électroniques :
```
ngo@math.univ-paris13.fr
polo@math.univ-paris13.fr
```
|
warning/0005/hep-th0005145.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
The five known superstring theories as well as the low–energy 11–dimensional supergravity are known to be related through a web of dualities, and it is believed that all these theories are simply different limits of an underlying 11–dimensional quantum theory known as $`M`$–theory, whose fundamental degrees of freedom are as yet unknown, but that can be defined as the strong coupling limit of Type IIA string theory . Let us first recall that $`M`$–theory compactified on a circle is described by Type IIA string theory at finite string coupling. It is by now a well known conjecture that $`M`$–theory compactified on a lightlike circle admits a nonperturbative description in terms of the degrees of freedom of a collection of $`D0`$–branes .
Matrix theory encodes a great deal of information about the structure of both $`M`$–theory and 11–dimensional supergravity (some reviews are ). One knows how to identify supergravitons, membranes and fivebranes in Matrix theory , and the interactions between these objects in Matrix theory have been found to agree with supergravity in a variety of situations. In particular, for general Matrix configurations, it was shown in that the supergravity potential between an arbitrary pair of $`M`$–theory objects arising from the exchange of quanta with zero longitudinal momentum is exactly reproduced by terms in the one–loop Matrix theory potential. These results were also used to describe a formulation of Matrix theory in a general metric and 3–form background, via a matrix sigma model type of action . Such matrix sigma model actions had also been advocated for earlier in . A different type of approach to the 3–form background is, e.g. .
A question that naturally arises is that if we have a formulation of Matrix theory in curved background fields, that should somehow yield a matrix formulation of Type II string theory in curved background fields, and in particular in the presence of R–R fields. Moreover, due to the second quantized nature of the Matrix theory formalism, we should be able to obtain in this way a description of multiple interacting strings in both NS–NS and R–R curved backgrounds. This would be quite interesting, as even for a single fundamental superstring the action in a general background including arbitrary R–R fields is not yet well understood.
Due to the relation between Type IIA string theory and $`M`$–theory, it is possible to construct a matrix theory formulation of superstring theory which is known as matrix string theory . Such a formulation is achieved once one understands toroidal compactifications of Matrix theory , for then the particular case of the $`𝐒^1`$ compactification will lead to the matrix formulation of the Type IIA superstring – as $`M`$–theory compactified on a circle yields the IIA theory, where the IIA string is obtained from the wrapped $`M2`$–brane . Recall that this matrix string theory is a supersymmetric gauge theory that not only contains all of the DLCQ IIA superstring theory, but also contains extra degrees of freedom which represent nonperturbative objects in string theory. These nonperturbative degrees of freedom represent the inclusion of $`D`$–brane states, and also give us a prescription to include nonperturbative corrections in calculations of diverse processes in perturbative string theory.
Because we know a great deal about Type IIA string theory, matrix string theory is a very good laboratory to test Matrix theory. Of course ideally we would like to have a microscopic definition of $`M`$–theory which would be covariant and defined in arbitrarily curved backgrounds. But due to the nonabelian character of the theory such is not an easy goal. Information from the abelian limit of the theory may then prove to be of great value in trying to deal with such issues, and a precious source of information on this abelian limit is undoubtably the Type II theory. In flat space the matrix string theory action has been lifted from the cylinder to its branched coverings and a precise connection with the Green–Schwarz action in light–cone gauge has been achieved , with the interesting result that the full moduli space of the IIA theory is recovered within matrix string theory only in the large $`N`$ limit. Scattering amplitudes have been reproduced within the matrix string formalism in , for reviews on several issues see . More recently, the issue of a spacetime covariant formulation of matrix string theory has been addressed in , but this is a matter which is far from settled.
This paper concerns the generalization of matrix string theory when in the presence of weakly curved background fields. In particular, we want to address the question of how to describe multiple interacting strings in NS–NS and R–R curved backgrounds. Indeed, because it is known how to describe the linear couplings of Matrix theory to a curved 11–dimensional background, we shall also be able to find the linear couplings of matrix string theory to a curved 10–dimensional background. This could be of great interest not only in trying to improve our knowledge of string theory in R–R backgrounds, but also when comparing to the IIA theory in the abelian limit we could expect for new information on how to construct Matrix theory in a general curved background. In summary, we are looking for a matrix string sigma model type of action,
$`𝒮`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau ({\displaystyle \frac{1}{2}}g_{\mu \nu }^{IIA}(X)I_g^{\mu \nu }+\varphi (X)I_\varphi +B_{\mu \nu }(X)I_s^{\mu \nu }+\stackrel{~}{B}_{\mu \nu \lambda \rho \sigma \tau }(X)I_5^{\mu \nu \lambda \rho \sigma \tau }`$ (1)
$`+C_\mu (X)I_0^\mu +\stackrel{~}{C}_{\mu \nu \lambda \rho \sigma \tau \xi }(X)I_6^{\mu \nu \lambda \rho \sigma \tau \xi }+C_{\mu \nu \lambda }(X)I_2^{\mu \nu \lambda }+\stackrel{~}{C}_{\mu \nu \lambda \rho \sigma }(X)I_4^{\mu \nu \lambda \rho \sigma }),`$
and we shall precisely explain in this paper how to construct this action by specifying both the $`I`$ tensor couplings as well as the inclusion of spacetime dependence in the (weak) background fields. The explicit form of all these tensor couplings is presented in section 4.2.
We shall begin in section 2 with a brief review of the work done for the case of Matrix theory in weak background fields . We shall recall that there is a definite proposal on how to supplement the flat space Matrix action with linear couplings between the background fields – the supergraviton, the membrane and the fivebrane – and the respective Matrix descriptions for the supergravity stress–energy tensor, membrane current and fivebrane current. Moreover we shall also recall that through the Sen–Seiberg limiting procedure, this action can be reduced to an action for multiple $`D0`$–branes in weakly curved Type IIA background fields. By $`T`$–duality this can be extended to any Type II $`D`$–brane. Then, in section 3 we present a brief review of the Dijkgraaf–Verlinde–Verlinde (DVV) reduction of Matrix theory to matrix string theory , via both the so–called $`911`$ flip and also the $`T`$$`S`$$`T`$ chain of dualities. This will be of fundamental use in the sections that follows, as we shall be generalizing that procedure to the curved background situation.
In the following sections we perform the DVV reduction to the multiple $`D0`$–brane action in order to find the matrix string theory action for multiple fundamental strings in curved but weak NS–NS and R–R backgrounds. As we just said, this is a generalization of the work by DVV. These sections deals with a great deal of algebra, and we will be schematic in presenting our results. The matrix sigma model obtained in this way gives a definite prescription on how to deal with R–R fields with an explicit spacetime dependence in Type II string theory. Due to the nonabelian nature of the action, it also gives a second quantized description of Type II string theory in such backgrounds. We shall obtain the matrix string sigma model both via the $`911`$ flip (described in section 4) and the chain of $`T`$ and $`S`$ dualities (described in section 5), and further check their equivalence explicitly by obtaining the same results in both cases. In order to do so, we will need to discuss in section 5 the implementation of $`S`$–duality in the composite operators of the 2–dimensional world–volume supersymmetric gauge theory describing the Type IIB $`D`$–string. We shall obtain the $`S`$–duality transformations for the world–volume fields from the equivalence with the $`911`$ flip, and we shall see that these transformation properties are indeed quite simple, as should be expected.
In section 6 we compare the result to the known Green–Schwarz sigma model action (for one string) . This is done by extracting the free string limit (the IR limit of the gauge theory) of the matrix string theory action. This will be a qualitative match only, as we shall not construct the precise lifting of the matrix string action to the Green–Schwarz action. We then use this comparison in order to discuss about possible, non–linear background curvature corrections to the matrix string action (involving many strings), and therefore to the Matrix theory action. Again this is a qualitative analysis, but it gives us further insight into the goal of constructing Matrix theory in arbitrary curved backgrounds. Then, in section 7, we briefly discuss the exponentiation of the noncommutative vertex operators we obtained in order to build coherent states of fundamental strings and so obtain the full non–linear matrix string sigma model. As such a construction is not clear at this stage, we turn to an illustration of the nonabelian character of our action with an example, namely multiple fundamental strings in a non–trivial R–R flux, where the strings are polarized into nonabelian configurations due to the background field. This means that Myers’ dielectric effect for $`D`$–branes has an analogue for fundamental strings. We also speculate on a possible relation between this effect and string theory noncommutative background geometries, where this could provide a very interesting example of target space noncommutativity in the presence of R–R fields (as opposed to recent discussions of world–volume noncommutativity in the presence of NS–NS fields, e.g., ). We conclude in section 8 with some open problems for future research.
## 2 Matrix Theory in Weakly Curved Backgrounds
We begin with a short review of the results obtained for Matrix theory in weakly curved background fields , and also for the action of multiple $`D0`$–branes in weak Type IIA backgrounds as well as for the action of multiple $`Dp`$–branes in Type II weak background fields .
### 2.1 Results for Matrix Theory
In this section we briefly review the results in dealing with the construction of a Matrix theory action in weak $`M`$–theory backgrounds. As we shall see, due to the $`911`$ flip in the DVV construction of matrix string theory, we will have particular interest in the tensors that appear in this Matrix theory action.
The proposal in question actually concerns the terms in the action of Matrix theory which are linear in the background fields . If we consider a general Matrix theory background, with metric $`g_{IJ}=\eta _{IJ}+h_{IJ}`$ and 3–form field $`A_{IJK}`$, then the linear effects of this background can be described by supplementing the flat space Matrix theory action,
$$S_{Flat}=\frac{1}{R}𝑑t\mathrm{𝐓𝐫}\left(\frac{1}{2}D_tX^iD_tX^i\frac{1}{2}\underset{i<j}{}[X^i,X^j]^2\frac{1}{2}\mathrm{\Theta }D_t\mathrm{\Theta }+\frac{1}{2}\mathrm{\Theta }\gamma ^i[X_i,\mathrm{\Theta }]\right),$$
(2)
with additional linear coupling terms of the form,
$`S_{Weak}`$ $`=`$ $`{\displaystyle }dt{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \underset{i_1,\mathrm{},i_n}{}}{\displaystyle \frac{1}{n!}}\{{\displaystyle \frac{1}{2}}T^{IJ(i_1\mathrm{}i_n)}_{i_1}\mathrm{}_{i_n}h_{IJ}(0)+J^{IJK(i_1\mathrm{}i_n)}_{i_1}\mathrm{}_{i_n}A_{IJK}(0)`$ (3)
$`+M^{IJKLMN(i_1\mathrm{}i_n)}_{i_1}\mathrm{}_{i_n}\stackrel{~}{A}_{IJKLMN}(0)+\mathrm{Fermionic}\mathrm{Terms}\},`$
where $`\stackrel{~}{A}`$ is the dual 6–form field which satisfies at linear order,
$$d\stackrel{~}{A}=dA.$$
(4)
The previous matrix expressions $`T^{IJ(i_1\mathrm{}i_n)}`$, $`J^{IJK(i_1\mathrm{}i_n)}`$ and $`M^{IJKLMN(i_1\mathrm{}i_n)}`$ are the Matrix theory forms of the multipole moments of the stress–energy tensor, membrane current and 5–brane current of 11–dimensional supergravity. Explicit forms for the bosonic parts of these moments were first given in , and those results were later extended to quadratic fermionic terms (and also some quartic fermionic terms) in . The complete results in are reproduced in the Appendix.
With these definitions the previous expressions yield a formulation of Matrix theory in a weak background metric to first order in $`h_{IJ}`$, the 3–form $`A_{IJK}`$, and all their higher derivatives. It was moreover argued in that if the Matrix theory conjecture is true in flat space, then this formulation must be correct at least to order $`𝒪(^4h,^4A)`$. It was also conjectured in that paper that this form may work to all orders in derivatives of the background fields, and in a general background. One should observe however that it is not known how to incorporate dependence of the background on the compact coordinate $`X^{}`$.
### 2.2 Results for Multiple D–branes
We proceed by reviewing how the previous results can be used to construct actions for multiple $`D0`$–branes and in general for multiple $`Dp`$–branes in Type II string theory, in the approximation of weak background fields. Of particular interest to our goal in this paper is the case of the $`D0`$–brane action, due to the duality sequence in the DVV construction of matrix string theory and its associated $`911`$ flip.
To start, we shall recall from how one obtains the action for multiple $`D0`$–branes in background fields, as this will later prove its interest when we try to do the same for the matrix string action. We begin with $`M`$–theory on a background metric, $`g_{IJ}=\eta _{IJ}+h_{IJ}`$, in a frame where there is a compact coordinate $`X^{}`$ of size $`R`$, which becomes lightlike in the flat space limit, $`g_{IJ}\eta _{IJ}`$. From the Sen–Seiberg limit we know that this theory can be described as a limit of a family of spacelike compactified theories. If we define an $`\stackrel{~}{M}`$–theory with background metric $`\stackrel{~}{g}_{IJ}=\eta _{IJ}+\stackrel{~}{h}_{IJ}`$, in a frame with a spacelike compact coordinate $`X^{11}`$ of size $`R_{11}`$, then the DLCQ limit of the original $`M`$–theory is found by boosting the $`\stackrel{~}{M}`$–theory along $`X^{11}`$, and then taking the limit $`R_{11}0`$. Knowing the boost we can trivially Lorentz relate the metric $`\stackrel{~}{g}_{IJ}`$ in the $`\stackrel{~}{M}`$–theory with the metric $`g_{IJ}`$ in the $`M`$–theory. Moreover, in the DLCQ description the $`M`$–theory is in light–cone coordinates, $`X^\pm =\frac{1}{\sqrt{2}}(X^0\pm X^{11})`$, and so it is easy to relate the metric $`\stackrel{~}{g}_{IJ}`$ to the light–cone metric $`g_{IJ}`$.
Of course our final goal is more than we have just obtained. We would like to relate the Type IIA string theory background fields to the DLCQ $`M`$–theory ones. But this is now straightforward. $`\stackrel{~}{M}`$–theory on a small spacelike circle of radius $`R_{11}`$ is known to be equivalent to Type IIA string theory with background fields given to leading order by,
$`h_{\mu \nu }^{IIA}`$ $`=`$ $`\stackrel{~}{h}_{\mu \nu }+{\displaystyle \frac{1}{2}}\eta _{\mu \nu }\stackrel{~}{h}_{\mathrm{11\hspace{0.33em}11}},`$
$`C_\mu `$ $`=`$ $`\stackrel{~}{h}_{11\mu },`$
$`\varphi `$ $`=`$ $`{\displaystyle \frac{3}{4}}\stackrel{~}{h}_{\mathrm{11\hspace{0.33em}11}}.`$ (5)
All we have left to do is to relate the $`\stackrel{~}{h}_{IJ}`$ metric to the $`h_{IJ}`$ one through the previously explained procedure. In order to describe nontrivial background antisymmetric tensor fields, one should also include the connections between the IIA background fields and the $`M`$–theory background 3–form field. The action for multiple $`D0`$–branes can now be obtained by direct comparison with the one for Matrix theory just described in the previous subsection. Indeed , one can first write the $`D0`$–brane action in terms of some unknown quantities coupling to the background fields. These quantities will be denoted by $`I_x`$ and will couple linearly to each of the background fields, so that to leading order the action for $`N`$ $`D0`$–branes is written as:
$`S_{D0branes}=S_{Flat}`$ $`+`$ $`{\displaystyle }dt{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{1}{n!}}\{{\displaystyle \frac{1}{2}}(_{k_1}\mathrm{}_{k_n}h_{\mu \nu }^{IIA})I_h^{\mu \nu (k_1\mathrm{}k_n)}+(_{k_1}\mathrm{}_{k_n}\varphi )I_\varphi ^{(k_1\mathrm{}k_n)}`$ (6)
$`+`$ $`(_{k_1}\mathrm{}_{k_n}C_\mu )I_0^{\mu (k_1\mathrm{}k_n)}+(_{k_1}\mathrm{}_{k_n}\stackrel{~}{C}_{\mu \nu \lambda \rho \sigma \tau \xi })I_6^{\mu \nu \lambda \rho \sigma \tau \xi (k_1\mathrm{}k_n)}`$
$`+`$ $`(_{k_1}\mathrm{}_{k_n}B_{\mu \nu })I_s^{\mu \nu (k_1\mathrm{}k_n)}+(_{k_1}\mathrm{}_{k_n}\stackrel{~}{B}_{\mu \nu \lambda \rho \sigma \tau })I_5^{\mu \nu \lambda \rho \sigma \tau (k_1\mathrm{}k_n)}`$
$`+`$ $`(_{k_1}\mathrm{}_{k_n}C_{\mu \nu \lambda })I_2^{\mu \nu \lambda (k_1\mathrm{}k_n)}+(_{k_1}\mathrm{}_{k_n}\stackrel{~}{C}_{\mu \nu \lambda \rho \sigma })I_4^{\mu \nu \lambda \rho \sigma (k_1\mathrm{}k_n)}\}.`$
Replacing in this action the background fields of the Type IIA string theory by the background fields of DLCQ $`M`$–theory according to the previous relations, one can then compare the previous action for $`D0`$–branes to the Matrix theory action and deduce the expressions for the string theory couplings $`I_x`$. These are :
$`I_h^{00}=T^{++}+T^++(I_h^{00})_8+𝒪(X^{12}),`$
$`I_h^{0i}=T^{+i}+T^i+𝒪(X^{10}),`$
$`I_h^{ij}=T^{ij}+(I_h^{ij})_8+𝒪(X^{12}),`$
$`I_\varphi =T^{++}{\displaystyle \frac{1}{3}}(T^++T^{ii})+(I_\varphi )_8+𝒪(X^{12}),`$
$`I_s^{0i}=3J^{+i}+𝒪(X^8),`$
$`I_s^{ij}=3J^{+ij}3J^{ij}+𝒪(X^{10}),`$
$`I_0^0=T^{++},`$
$`I_0^i=T^{+i},`$
$`I_2^{0ij}=J^{+ij}+𝒪(X^{10}),`$
$`I_2^{ijk}=J^{ijk}+𝒪(X^8),`$
$`I_4^{0ijkl}=6M^{+ijkl}+𝒪(X^8),`$
$`I_4^{ijklm}=6M^{ijklm}+𝒪(X^{10}),`$
$`I_6^{0ijklmn}=S^{+ijklmn}+𝒪(X^{10}),`$
$`I_6^{ijklmnp}=S^{ijklmnp}+𝒪(X^{12}).`$ (7)
By $`T`$–duality of background supergravity fields and $`T`$–duality of world–volume fields, the previous action for $`N`$ $`D0`$–branes can be transformed into an action for $`N`$ Type II $`Dp`$–branes, as was discussed in . This also allows for a discussion of nonabelian terms in the Born–Infeld action. For further discussion we refer the reader to the original references .
## 3 Matrix String Theory
According to the DVV formulation of matrix string theory , one can reduce the Matrix theory action to an action for IIA matrix strings in two different ways. One way is by performing the so–called $`911`$ flip, where one exchanges the role of the $`9^{th}`$ and $`11^{th}`$ directions of $`M`$–theory. Another way is via a set of dualities on the background fields. Moreover, the coordinate flip should clearly be equivalent to this specific chain of dualities. In here, one starts by $`T`$–dualizing and then takes an $`S`$–duality followed by another $`T`$–duality. The starting point is the Type IIA theory, with $`N_{11}`$ yielding the $`D`$–particle number. After the $`T`$–duality along $`R_9^{IIA}`$ one reaches Type IIB, where $`N_{11}`$ now equals the $`D`$–string number. The Type IIB $`S`$–duality leads to $`N_{11}`$ equaling the $`F`$–string number, and the final $`T`$–duality along $`R_9^{IIB}`$ leads back to Type IIA, with $`N_{11}`$ now being equal to the $`F`$–string momenta.
In order to see that this exactly matches the simple $`911`$ flip on the compact coordinates, let us follow these dualities with a slightly greater detail . If we compactify $`M`$–theory on $`𝐒_{R_9}^1\times 𝐒_{R_{11}}^1`$, with $`R_{11}`$ the spacelike compact direction which becomes lightlike in the Sen–Seiberg limit, we will have the parameters,
$$R_{11}=g_s\mathrm{}_s,\mathrm{}_P^3=g_s\mathrm{}_s^3,$$
(8)
and also $`R_9`$ for the remaining spacelike compact direction. The $`911`$ flip simply leads to the IIA theory with parameters,
$$R_9=g_s^{}\mathrm{}_s,\mathrm{}_P^3=g_s^{}\mathrm{}_s^3,$$
(9)
where now the remaining spacelike compact direction is $`R_{11}`$. On the other hand, given our starting point and $`T`$–dualizing along $`R_9^{IIA}`$, one obtains the following Type IIB parameters,
$`g_s^{IIB}`$ $`=`$ $`{\displaystyle \frac{\mathrm{}_s}{R_9^{IIA}}}g_s^{IIA}={\displaystyle \frac{\mathrm{}_s}{R_9^{IIA}}}{\displaystyle \frac{R_{11}}{\mathrm{}_s}}={\displaystyle \frac{R_{11}}{R_9^{IIA}}},`$ (10)
$`R_9^{IIB}`$ $`=`$ $`{\displaystyle \frac{\alpha ^{}}{R_9^{IIA}}}={\displaystyle \frac{\mathrm{}_s^2}{R_9^{IIA}}}.`$ (11)
A further IIB $`S`$–duality leads to
$`g_{}^{}{}_{s}{}^{IIB}`$ $`=`$ $`{\displaystyle \frac{1}{g_s^{IIB}}}={\displaystyle \frac{R_9^{IIA}}{R_{11}}},`$ (12)
$`R_{}^{}{}_{9}{}^{IIB}`$ $`=`$ $`{\displaystyle \frac{1}{g_s^{IIB}}}R_9^{IIB}=\left({\displaystyle \frac{R_{11}}{R_9^{IIA}}}\right)^1{\displaystyle \frac{\mathrm{}_s^2}{R_9^{IIA}}}.`$ (13)
In the expressions above for the radius, recall that under $`T`$–duality it is the Einstein frame metric that is invariant. The string frame metric gets transformed with a $`g_s`$ factor. Finally, we finish the chain of dualities by $`T`$–dualizing back to the IIA theory along $`R_{}^{}{}_{9}{}^{IIB}`$. We end up with the parameters,
$`g_{}^{}{}_{s}{}^{IIA}`$ $`=`$ $`{\displaystyle \frac{\mathrm{}_s}{R_{}^{}{}_{9}{}^{IIB}}}g_{}^{}{}_{s}{}^{IIB}={\displaystyle \frac{R_9}{\mathrm{}_s}},`$ (14)
$`R_{}^{}{}_{9}{}^{IIA}`$ $`=`$ $`{\displaystyle \frac{\mathrm{}_s^2}{R_{}^{}{}_{9}{}^{IIB}}}=R_{11},`$ (15)
which are exactly the same as the ones obtained via the $`911`$ flip.
Given that, as we have just seen, the chain of dualities is equivalent to the $`911`$ flip, we shall now obtain the matrix string action from the Matrix action by following the most straightforward path, i.e., we shall simply perform the flip to the Matrix theory action . With the dimensionfull parameters made explicit, in order to produce the correct dimensions for the fields, the Matrix action in a flat background is written as,
$$S=𝑑t\mathrm{𝐓𝐫}\left(\frac{1}{2R}\dot{X_i}\dot{X_i}+\frac{RM_P^6}{8\pi ^2}\underset{i<j}{}[X^i,X^j]^2+\frac{iM_P^3}{4\pi }\theta ^T\dot{\theta }\frac{RM_P^6}{8\pi ^2}\theta ^T\gamma _i[X^i,\theta ]\right),$$
(16)
with $`R=2\pi \mathrm{}_P^3`$ and $`M_P`$ is the Planck mass.
We further consider the theory compactified along the $`9^{th}`$ direction. Therefore, defining $`\widehat{R}_9=\frac{\alpha ^{}}{R_9}`$, one $`T`$–dualizes according to the standard procedure and obtains:
$`S^{}`$ $`=`$ $`{\displaystyle }dt{\displaystyle \frac{1}{2\pi \widehat{R}_9}}{\displaystyle _0^{2\pi \widehat{R}_9}}d\widehat{x}\mathrm{𝐓𝐫}({\displaystyle \frac{1}{2R}}\dot{X_i}\dot{X_i}+{\displaystyle \frac{1}{2R}}(2\pi \alpha ^{})^2\dot{A}^2+{\displaystyle \frac{RM_P^6}{8\pi ^2}}{\displaystyle \underset{i<j}{}}[X^i,X^j]^2`$
$`{\displaystyle \frac{RM_P^6}{8\pi ^2}}(2\pi \alpha ^{})^2(D_{\widehat{x}}X^i)^2+{\displaystyle \frac{iM_P^3}{4\pi }}\theta ^T\dot{\theta }{\displaystyle \frac{RM_P^6}{8\pi ^2}}\theta ^T\gamma _i[X^i,\theta ]i{\displaystyle \frac{RM_P^6}{8\pi ^2}}(2\pi \alpha ^{})\theta ^T\gamma _9D_{\widehat{x}}\theta ).`$
The implementation of the $`911`$ flip is quite simple, as one just has to notice the change in parameters so that $`R_9=g_s\mathrm{}_s`$ and $`\widehat{R}_9=\frac{\mathrm{}_s}{g_s}`$. Consequently,
$`S^{}`$ $`=`$ $`{\displaystyle }dt{\displaystyle \frac{g_s}{2\pi \mathrm{}_s}}{\displaystyle _0^{2\pi \frac{\mathrm{}_s}{g_s}}}d\widehat{x}\mathrm{𝐓𝐫}({\displaystyle \frac{1}{2R}}\dot{X_i}\dot{X_i}+{\displaystyle \frac{2\pi ^2\mathrm{}_s^4}{R}}\dot{A}^2+{\displaystyle \frac{RM_P^6}{8\pi ^2}}{\displaystyle \underset{i<j}{}}[X^i,X^j]^2`$
$`{\displaystyle \frac{1}{2}}RM_P^6\mathrm{}_s^4(D_{\widehat{x}}X^i)^2+{\displaystyle \frac{iM_P^3}{4\pi }}\theta ^T\dot{\theta }{\displaystyle \frac{RM_P^6}{8\pi ^2}}\theta ^T\gamma _i[X^i,\theta ]{\displaystyle \frac{iRM_P^6\mathrm{}_s^2}{4\pi }}\theta ^T\gamma _9D_{\widehat{x}}\theta ).`$
One can rescale the world–sheet coordinates, from $`(\widehat{x},t)`$ to $`(\sigma ,\tau )`$, such that $`0<\sigma <2\pi `$ and so that the coordinates on the cylinder become dimensionless. For that one changes $`\widehat{x}=\frac{\mathrm{}_s}{g_s}\sigma `$ (and therefore $`D_{\widehat{x}}=\frac{g_s}{\mathrm{}_s}D`$ <sup>1</sup><sup>1</sup>1Throughout, derivatives with no explicit subscript shall refer to the cylinder world–sheet index $`\sigma `$, i.e., $`DD_\sigma `$ and $`_\sigma `$.). We also have to rescale time on the world-sheet $`t=\frac{\mathrm{}_s^2}{R}\tau `$. Moreover, we shall deal with dimensionless background target fields such that they will be measured in string units, i.e., rescale $`(X,\theta )`$ to $`(\mathrm{}_sX,\mathrm{}_s\theta )`$. All this done, we are left with the rescaled action,
$`S^{}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\tau d\sigma \mathrm{𝐓𝐫}({\displaystyle \frac{1}{2}}\dot{X_i}\dot{X_i}+2\pi ^2g_s^2\dot{A}^2+{\displaystyle \frac{1}{8\pi ^2g_s^2}}{\displaystyle \underset{i<j}{}}[X^i,X^j]^2`$
$`{\displaystyle \frac{1}{2}}(DX^i)^2+{\displaystyle \frac{i}{4\pi R_9}}\theta ^T\dot{\theta }{\displaystyle \frac{1}{8\pi ^2g_sR_9}}\theta ^T\gamma _i[X^i,\theta ]{\displaystyle \frac{i}{4\pi R_9}}\theta ^T\gamma _9D\theta ).`$
In order to cast this action into a more familiar looking one, we simply have to perform one further rescaling of the background fermions, $`\theta \sqrt{4\pi R_9}\theta `$, and change the notation for the string coupling constant as $`g_s\frac{g_s}{2\pi }`$. Then the $`911`$ flip is concluded and we have obtained the DVV reduction of the Matrix theory action,
$`S`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\tau d\sigma \mathrm{𝐓𝐫}({\displaystyle \frac{1}{2}}((\dot{X_i})^2(DX^i)^2)+{\displaystyle \frac{1}{2g_s^2}}{\displaystyle \underset{i<j}{}}[X^i,X^j]^2+{\displaystyle \frac{1}{2}}g_s^2\dot{A}^2`$ (17)
$`+i(\theta ^T\dot{\theta }\theta ^T\gamma _9D\theta ){\displaystyle \frac{1}{g_s}}\theta ^T\gamma _i[X^i,\theta ]).`$
This action is second quantized in the sense that it describes multiple interacting strings. One can further consider the special case of free strings, recovered in the infra–red limit with $`g_s=0`$. In this limit the above two dimensional gauge theory becomes strongly coupled – as the Yang–Mills gauge coupling is related to the string coupling as $`g_{YM}\frac{1}{g_s}`$ – and a non–trivial conformal field theory will describe the IR fixed point . One can observe that in this limit, $`g_s0`$, the world–sheet gauge field drops out, and moreover all matrices are diagonalized, i.e., they will commute,
$$[X^i,X^j]=0,[X^i,\theta ]=0.$$
In this conformal field theory limit the previous Matrix string action reduces to,
$$S=\frac{1}{2\pi }𝑑\tau 𝑑\sigma \left(\frac{1}{2}(_\mu X^i)^2+i\theta ^T\rho ^\mu _\mu \theta \right),$$
(18)
where $`\{\mu \}`$ are world–sheet indices. This action can be exactly mapped to the light–cone Green-Schwarz action for the Type II superstring .
We have gone through a lengthy review of the DVV reduction, in order to set pace and notation for the section that follows. In there, we shall follow the same procedure applied to the full set of multipole moments of the 11–dimensional supercurrents for the stress tensor $`T^{IJ}`$, membrane current $`J^{IJK}`$ and fivebrane current $`M^{IJKLMN}`$. These “DVV reduced” tensors will be the basis for the matrix string theory action in a weakly curved background.
## 4 Reduction via the 9–11 Flip
In order to write down the matrix string theory action in weak background fields, one needs to know the DVV reduction of the Matrix theory stress tensor, membrane current and 5–brane current. This should be clear from the fact that the matrix string theory action is obtained via a DVV reduction of the Matrix theory action (as explained in the previous section), and the fact that in weak background fields the Matrix theory action is constructed precisely with the use of these tensors and currents (as explained in section 2, in particular in expression (3)). We will begin in here by applying the DVV reduction using the $`911`$ flip, just as described previously. Later, we shall look at the sequence of dualities, and compare both procedures.
Let us start by specifying the conventions for the following of this section. Time derivatives are taken with respect to Minkowski time $`t`$. All expressions have been written in a gauge with $`A_0=0`$. Gauge invariance can be restored by replacing $`\dot{X}`$ with $`D_tX`$. Indices $`i,j,\mathrm{}`$, run from 1 through 9, while indices $`a,b,\mathrm{}`$, run from 0 through 9. In these expressions we use the definitions $`F_{0i}=\dot{X}^i,F_{ij}=i[X^i,X^j]`$. A Matrix form for the transverse 5–brane current components $`M^{+ijklm},M^{ijklmn}`$ is as yet unknown. There are also fermionic components of the supercurrent which couple to background fermion fields in the supergravity theory. We will not discuss these couplings in this paper, but the Matrix theory form of the currents is determined in . Moreover, there is also a 6–brane current appearing in Matrix theory related to nontrivial 11–dimensional background metrics.
### 4.1 Matrix Theory Tensors
We shall briefly describe the DVV reduction of the first component of the stress tensor, referring the specifics to the full description in the previous section. Then, we shall simply present the results for the other components in a schematic form (in the Appendix).
The zeroth moment of the $`T^{++}`$ component of the Matrix stress tensor is given by,
$$T^{++}=\frac{1}{R}\mathrm{𝐒𝐓𝐫}\left(\text{1}\text{ }\text{1}\right),$$
(19)
where STr indicates a trace which is symmetrized over all orderings of terms of the forms $`F_{ab}`$, $`\theta `$ and $`[X^i,\theta ]`$. We shall denote by the same name, $`T^{++}`$, the time integrated component which appears in the curved Matrix theory action. It is to this integrated term that we will apply the DVV reduction. In this term there is no need to introduce explicit dimensionfull parameters – there are no background fields – but we shall do it automatically in all the following terms, just as we did for the Matrix theory action in the previous section. As the theory is further compactified along the $`9^{th}`$ direction, we $`T`$–dualize to obtain, after the $`911`$ flip,
$$T^{++}=𝑑t\frac{1}{R}\frac{g_s}{2\pi \mathrm{}_s}_0^{2\pi \frac{\mathrm{}_s}{g_s}}𝑑\widehat{x}\mathrm{𝐒𝐓𝐫}\left(\text{1}\text{ }\text{1}\right).$$
Rescaling of world–sheet coordinates, background fields, and coupling constants (most of them trivial for this component), we are left with the final result,
$$T^{++}=\frac{1}{2\pi }\left(\frac{\mathrm{}_s}{R}\right)^2𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\text{1}\text{ }\text{1}\right).$$
(20)
Moreover, we shall later be interested in the conformal field theory limit of these tensors. So, we further observe that the free string limit can be easily taken as,
$$\underset{g_s0}{lim}T^{++}=T^{++}.$$
(21)
We can proceed along the same line for the following components. The zeroth moment of the $`T^{+i}`$ component of the Matrix stress tensor is given by,
$$T^{+i}=\frac{1}{R}\mathrm{𝐒𝐓𝐫}\left(\dot{X_i}\right).$$
(22)
Under $`T`$–duality for the $`911`$ flip, one obtains for $`i9`$,
$$T^{+i}=𝑑t\frac{1}{R}\frac{g_s}{2\pi \mathrm{}_s}_0^{2\pi \frac{\mathrm{}_s}{g_s}}𝑑\widehat{x}\mathrm{𝐒𝐓𝐫}\left(\dot{X_i}\right).$$
After the needed rescalings of world–sheet coordinates, background fields, and coupling constants, we are left with the final result,
$$T^{+i}=\frac{1}{2\pi }\left(\frac{\mathrm{}_s}{R}\right)𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\dot{X_i}\right).$$
(23)
As to the free string limit, it can be taken as,
$$\underset{g_s0}{lim}T^{+i}=T^{+i}.$$
(24)
Under $`T`$–duality for the $`911`$ flip, one obtains for $`i=9`$,
$$T^{+9}=𝑑t\frac{1}{R}\frac{g_s}{\mathrm{}_s}_0^{2\pi \frac{\mathrm{}_s}{g_s}}𝑑\widehat{x}\mathrm{𝐒𝐓𝐫}\left(\mathrm{}_s^2\dot{A}\right).$$
After the needed rescalings of world–sheet coordinates, background fields, and coupling constants, we are left with the final result,
$$T^{+9}=\frac{1}{2\pi }\left(\frac{\mathrm{}_s}{R}\right)𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(g_s\dot{A}\right).$$
(25)
As to the free string limit, it can be taken as,
$$\underset{g_s0}{lim}T^{+9}=0.$$
(26)
The procedure is always the same, for all the components. It should be clear to the reader how to obtain all the results, which are presented schematically in the Appendix. A few comments can be made, about the structures we have derived. First, as was trivially expected, the string coupling appears as expected, i.e., a factor of $`g_s`$ for each factor of $`\dot{A}`$, and factor of $`\frac{1}{g_s}`$ for each factor of $`[X,X]`$, or for each factor of $`[X,\theta ]`$. Moreover, every tensor (and the action in section 3, also) has an overall normalization factor of $`\frac{1}{2\pi }`$. Second, and more importantly, we observe that if one counts operator insertions of background coordinates into the currents as $`\dot{X}`$, $`\dot{A}`$, $`\theta `$, $`[X,X]`$ and $`[X,\theta ]`$ each counting as one operator insertion, then the order of the currents depends on the number of insertions as follows. For zero insertions, it is order $`𝒪=(\frac{\mathrm{}_s}{R})^2`$; for one insertion, it is order $`𝒪=(\frac{\mathrm{}_s}{R})`$; for two insertions, it is order $`𝒪=1`$; for three insertions, it is order $`𝒪=(\frac{R}{\mathrm{}_s})`$; for four insertions, it is order $`𝒪=(\frac{R}{\mathrm{}_s})^2`$; and so on, for $`n`$ insertions it is order $`𝒪=(\frac{R}{\mathrm{}_s})^{n2}`$.
### 4.2 Matrix String Theory Tensors
We have thus performed the analysis of the Matrix theory expressions for the stress tensor, the membrane current and the 5–brane current. As previously explained in section 2, one can obtain the matrix string theory action in terms of other tensors: the sources $`I_p`$ of $`Dp`$–brane currents for $`p=2n`$, the sources $`I_s`$ and $`I_5`$ associated with fundamental string and $`NS5`$–brane currents respectively, and also the sources $`I_h`$ and $`I_\varphi `$ of background metric and background dilaton fields. These currents $`I`$ can moreover be expressed as linear combinations of the Matrix theory expressions for $`T`$, $`J`$ and $`M`$. In previous work, the results for the lowest dimension operators appearing in the monopole (integrated) $`D0`$–brane currents were obtained .
Now, because of the $`911`$ flip, we are dealing with $`M`$–theory on spacelike $`R_9`$ instead of $`M`$–theory on spacelike $`R_{11}`$ as before. This means that the $`I`$ tensors are not necessarily related to the $`T`$, $`J`$ and $`M`$ tensors in the same way as in the case of the $`D0`$–brane action that was described briefly in section 2. We begin by addressing such a question, in order to derive the correct expressions for the $`I`$ linear tensor couplings. The original $`M`$–theory where the $`D0`$–brane couplings were derived was spacelike compactified along $`R_{11}`$, so that in light–cone coordinates we would be dealing with $`X^\pm X^0\pm X^{11}`$ and a further compact coordinate $`X^9`$. With the $`911`$ flip we are now led to an $`\widehat{M}`$–theory compactified along $`R_9`$, and where in light–cone gauge the coordinates are now $`\widehat{X}^\pm \widehat{X}^0\pm \widehat{X}^9`$ and the compact coordinate $`\widehat{X}^{11}`$. Clearly we have two frames, the “11” frame in the original $`M`$–theory, and the “9” frame in the flipped $`\widehat{M}`$–theory. The relations we presented briefly in section 2 concerning the relation between the $`I`$ tensors and the Matrix theory tensors $`T`$, $`J`$ and $`M`$, are still valid in the flipped “9” frame, but now relating the $`I`$ tensors to the Matrix tensors in this frame, i.e., $`\widehat{T}`$, $`\widehat{J}`$ and $`\widehat{M}`$. If we then relate these “9” frame Matrix tensors back to the “11” frame Matrix tensors, we will be able to express the matrix string theory couplings $`I`$ in terms of the just derived DVV reduced Matrix tensors $`T`$, $`J`$ and $`M`$. So, all one needs to do is a simple change of coordinates.
To begin with a simple example, let us look at the $`I_s^{ij}`$ component of the matrix fundamental string current, which is given by:
$$I_s^{ij}\widehat{J}^{+ij}\widehat{J}^{ij}.$$
(27)
This expression holds in the “9” frame. Relating the $`\widehat{J}`$ tensor to the $`J`$ tensor in the “11” frame, one obtains,
$$I_s^{ij}\widehat{J}^{+ij}\widehat{J}^{ij}J^{9ij}.$$
(28)
On the other hand, at the level of background fields, one knows how to relate the NS 2–form $`B_{\mu \nu }`$ to the $`M`$–theory 3–form $`A_{IJK}`$, via $`B_{ij}A_{9ij}`$, where 9 is the spacelike compact direction involved in the Sen–Seiberg limit. The coupling we have just derived above is then precisely what one would expect.
The procedure is always the same, and it should be straightforward for the reader to reproduce the results, which we now present schematically. Observe that as we change from the “9” frame to the “11” frame, there is a mixing of different orders in the currents, i.e., there will be tensors in different powers of $`𝒪\left((\frac{R}{\mathrm{}_s})^{n2}\right)`$. We will neglect some of these tensors in the following expressions, and only keep the orders of interest to us. The components of the matrix string current associated with the background metric field are:
$`I_h^{00}`$ $`=`$ $`\widehat{T}^{++}+\widehat{T}^++(I_h^{00})_8+𝒪(\widehat{X}^{12})=T^{++}+T^++\mathrm{}`$
$`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}(\left({\displaystyle \frac{\mathrm{}_s}{R}}\right)^2\text{1}\text{ }\text{1}+{\displaystyle \frac{1}{2}}\dot{X_i}^2+{\displaystyle \frac{1}{2}}(DX^i)^2+{\displaystyle \frac{1}{2}}g_s^2\dot{A}^2{\displaystyle \frac{1}{2g_s^2}}{\displaystyle \underset{i<j}{}}[X^i,X^j]^2`$
$`+{\displaystyle \frac{1}{g_s}}\theta \gamma _i[X^i,\theta ]+i\theta \gamma ^9D\theta )+\mathrm{},`$
$`I_h^{0i}`$ $`=`$ $`\widehat{T}^{+i}+\widehat{T}^i+𝒪(\widehat{X}^{10})=T^{+i}+T^i+\mathrm{}`$
$`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}\{\left({\displaystyle \frac{\mathrm{}_s}{R}}\right)\dot{X_i}+\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)[{\displaystyle \frac{1}{2}}\dot{X_i}(\dot{X_j})^2+{\displaystyle \frac{1}{2}}g_s^2\dot{X_i}\dot{A}^2{\displaystyle \frac{1}{2g_s^2}}\dot{X_i}{\displaystyle \underset{j<k}{}}[X^j,X^k]^2`$
$`+{\displaystyle \frac{1}{2}}\dot{X_i}(DX^j)^2{\displaystyle \frac{1}{g_2^2}}[X^i,X^j][X^j,X^k]\dot{X_k}DX^iDX^k\dot{X_k}+i[X^i,X^j]DX^j\dot{A}`$
$`{\displaystyle \frac{1}{2g_s}}\theta _\alpha \dot{X_k}[X_j,\theta _\beta ]\{\gamma ^k\delta _{ij}+\gamma ^i\delta _{jk}2\gamma ^j\delta _{ki}\}_{\alpha \beta }{\displaystyle \frac{1}{2}}\dot{A}\theta \gamma ^9[X_i,\theta ]+i\dot{X_i}\theta \gamma ^9D\theta `$
$`{\displaystyle \frac{i}{2}}g_s\dot{A}\theta \gamma ^iD\theta {\displaystyle \frac{i}{4g_s^2}}\theta _\alpha [X^k,X^j][X^l,\theta _\beta ]\{\gamma ^{[ikjl]}+2\gamma ^{[jl]}\delta _{ki}+4\delta _{ki}\delta _{jl}\}_{\alpha \beta }`$
$`{\displaystyle \frac{1}{2g_s}}\theta _\alpha DX^k[X^j,\theta _\beta ]\{\gamma ^{[ik9j]}+\gamma ^{[9j]}\delta _{ki}\}_{\alpha \beta }+{\displaystyle \frac{1}{4g_s}}\theta _\alpha [X^k,X^j]D\theta _\beta \{\gamma ^{[ikj9]}+2\gamma ^{[j9]}\delta _{ki}\}_{\alpha \beta }`$
$`iDX^i\theta D\theta +\mathrm{}]\}+\mathrm{},`$
$`I_h^{ij}`$ $`=`$ $`\widehat{T}^{ij}+(I_h^{ij})_8+𝒪(\widehat{X}^{12})=T^{ij}+\mathrm{}`$ (29)
$`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}(\dot{X_i}\dot{X_j}DX^iDX^j{\displaystyle \frac{1}{g_s^2}}[X^i,X^k][X^k,X^j]`$
$`{\displaystyle \frac{1}{2g_s}}\theta \gamma ^i[X_j,\theta ]{\displaystyle \frac{1}{2g_s}}\theta \gamma ^j[X_i,\theta ])+\mathrm{},`$
where $`(I_h^{00})_8=\frac{3}{2}\widehat{T}^{}+\mathrm{}`$ and $`(I_h^{ij})_8=2\widehat{T}^{}+\mathrm{}`$, and we know the matrix string form of $`\widehat{T}^{}`$ from (B). The conformal field theory limit of these tensors is simply:
$`\underset{g_s0}{lim}I_h^{00}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\left(\frac{\mathrm{}_s}{R}\right)^2\text{1}\text{ }\text{1}+\frac{1}{2}\dot{X_i}^2+\frac{1}{2}(X^i)^2+i\theta \gamma ^9\theta \right)}+\mathrm{},`$
$`\underset{g_s0}{lim}I_h^{0i}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}(\left({\displaystyle \frac{\mathrm{}_s}{R}}\right)\dot{X_i}+\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)\{{\displaystyle \frac{1}{2}}\dot{X_i}(\dot{X_j})^2+{\displaystyle \frac{1}{2}}\dot{X_i}(X^j)^2X^iX^k\dot{X_k}`$
$`+i\dot{X_i}\theta \gamma ^9\theta iX^i\theta \theta \})+\mathrm{},`$
$`\underset{g_s0}{lim}I_h^{ij}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\dot{X_i}\dot{X_j}X^iX^j\right)}+\mathrm{}.`$ (30)
The $`I_\varphi `$ matrix string current associated to the dilaton is given by,
$`I_\varphi `$ $`=`$ $`\widehat{T}^{++}{\displaystyle \frac{1}{3}}(\widehat{T}^++\widehat{T}^{ii})+(I_\varphi )_8+𝒪(\widehat{X}^{12})=T^{++}+{\displaystyle \frac{1}{3}}(T^++T^{ii})+\mathrm{}`$ (31)
$`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\left(\frac{\mathrm{}_s}{R}\right)^2\text{1}\text{ }\text{1}+\frac{1}{2}\dot{X_i}^2\frac{1}{2}(DX^i)^2+\frac{1}{2}g_s^2\dot{A}^2+\frac{1}{2g_s^2}\underset{i<j}{}[X^i,X^j]^2\right)}`$
$`+\mathrm{},`$
where $`(I_\varphi )_8=\frac{1}{2}\widehat{T}^{}+\mathrm{}`$ and we know the matrix string form of $`\widehat{T}^{}`$ from (B). This current has the following conformal field theory limit,
$$\underset{g_s0}{lim}I_\varphi =\frac{1}{2\pi }𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\left(\frac{\mathrm{}_s}{R}\right)^2\text{1}\text{ }\text{1}+\frac{1}{2}\dot{X_i}^2\frac{1}{2}(X^i)^2\right)+\mathrm{}.$$
(32)
The components of the matrix fundamental string current are:
$`I_s^{0i}`$ $`=`$ $`3\widehat{J}^{+i}+𝒪(\widehat{X}^8)=3J^{+i9}+3J^{i9}+\mathrm{}`$
$`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{1}{2}}\left({\displaystyle \frac{\mathrm{}_s}{R}}\right)DX^i+\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)\{{\displaystyle \frac{1}{2}}\dot{X^i}\dot{X^k}DX^k{\displaystyle \frac{i}{2}}\dot{A}\dot{X^k}[X^k,X^i]+{\displaystyle \frac{g_s^2}{4}}\dot{A}^2DX^i`$
$`{\displaystyle \frac{1}{4}}(\dot{X^k})^2DX^i{\displaystyle \frac{1}{4g_s^2}}DX^i{\displaystyle \underset{k<l}{}}[X^k,X^l]^2{\displaystyle \frac{1}{4}}DX^i(DX^k)^2`$
$`{\displaystyle \frac{1}{2g_s^2}}[X^i,X^k][X^k,X^l]DX^l+{\displaystyle \frac{1}{4g_s}}\theta _\alpha \dot{X^k}[X^m,\theta _\beta ]\{\gamma ^{[ki9m]}+\gamma ^{[9m]}\delta _{ki}\}_{\alpha \beta }`$
$`{\displaystyle \frac{1}{4}}\dot{A}\theta _\alpha [X^m,\theta _\beta ]\{\gamma ^{[im]}+2\delta _{im}\}_{\alpha \beta }+{\displaystyle \frac{i}{2}}\dot{X^i}\theta D\theta {\displaystyle \frac{i}{4}}\dot{A}\theta \gamma ^{[i9]}D\theta `$
$`+{\displaystyle \frac{3i}{4g_s^2}}\theta _\alpha [X^k,X^l][X^m,\theta _\beta ]\{\gamma ^{[9kl]}\delta _{mi}+2\gamma ^{[li9]}\delta _{km}+2\gamma ^9\delta _{il}\delta _{km}\}_{\alpha \beta }`$
$`{\displaystyle \frac{3}{2g_s}}DX^k\theta \gamma ^k[X^i,\theta ]{\displaystyle \frac{3}{2g_s}}DX^k\theta \gamma ^i[X^k,\theta ]`$
$`+{\displaystyle \frac{3i}{4g_s}}\theta _\alpha [X^k,X^l]D\theta _\beta \{\gamma ^{[ikl]}+2\gamma ^l\delta _{ik}\}_{\alpha \beta }iDX^i\theta \gamma ^9D\theta +\mathrm{}\})+\mathrm{},`$
$`I_s^{ij}`$ $`=`$ $`3\widehat{J}^{+ij}3\widehat{J}^{ij}+𝒪(\widehat{X}^{10})=3J^{9ij}+\mathrm{}`$ (33)
$`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{1}{2}\dot{X^i}DX^j+\frac{1}{2}\dot{X^j}DX^i\frac{i}{2}\dot{A}[X^i,X^j]+\frac{1}{4g_s}\theta \gamma ^{[ij9l]}[X_l,\theta ]\right)}.`$
The conformal field theory limit of these tensors is:
$`\underset{g_s0}{lim}I_s^{0i}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{1}{2}}\left({\displaystyle \frac{\mathrm{}_s}{R}}\right)X^i+\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)\{{\displaystyle \frac{1}{2}}\dot{X^i}\dot{X^k}X^k`$
$`{\displaystyle \frac{1}{4}}(\dot{X^k})^2X^i{\displaystyle \frac{1}{4}}X^i(X^k)^2+{\displaystyle \frac{i}{2}}\dot{X^i}\theta \theta iX^i\theta \gamma ^9\theta \})+\mathrm{},`$
$`\underset{g_s0}{lim}I_s^{ij}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{1}{2}\dot{X^i}X^j+\frac{1}{2}\dot{X^j}X^i\right)}+\mathrm{}.`$ (34)
Let us now move to the R–R fields. The components of the matrix string $`D0`$–brane current are:
$`I_0^0`$ $`=`$ $`\widehat{T}^{++}=T^{+9}+T^9+\mathrm{}`$
$`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}(\left({\displaystyle \frac{\mathrm{}_s}{R}}\right)g_s\dot{A}+\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)\{{\displaystyle \frac{1}{2}}g_s\dot{A}(\dot{X^i})^2+{\displaystyle \frac{1}{2}}g_s^3\dot{A}^3{\displaystyle \frac{1}{2g_s}}\dot{A}{\displaystyle \underset{i<j}{}}[X^i,X^j]^2`$
$`{\displaystyle \frac{i}{g_s}}DX^i[X^i,X^j]\dot{X^j}{\displaystyle \frac{1}{2}}g_s(DX^i)^2\dot{A}{\displaystyle \frac{1}{2g_s}}\dot{X^i}\theta \gamma ^9[X^i,\theta ]+\dot{A}\theta \gamma ^i[X^i,\theta ]`$
$`{\displaystyle \frac{i}{2}}\dot{X^i}\theta \gamma ^iD\theta {\displaystyle \frac{i}{4g_s^2}}[X^i,X^j]\theta \gamma ^{[9ijk]}[X^k,\theta ]+{\displaystyle \frac{1}{2g_s}}\theta _\alpha DX^i[X^j,\theta _\beta ]\{\gamma ^{[ij]}+2\delta _{ij}\}_{\alpha \beta }`$
$`+{\displaystyle \frac{i}{2}}DX^i\theta \gamma ^{[i9]}D\theta +\mathrm{}\})+\mathrm{},`$
$`I_0^i`$ $`=`$ $`\widehat{T}^{+i}=T^{9i}+\mathrm{}`$ (35)
$`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(g_s\dot{X_i}\dot{A}+\frac{i}{g_s}[X^i,X^k]DX^k\frac{i}{2}\theta \gamma ^iD\theta \frac{1}{2g_s}\theta \gamma ^9[X_i,\theta ]\right)}+\mathrm{}.`$
With the conformal field theory limit:
$`\underset{g_s0}{lim}I_0^0`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{i}{2}\dot{X^i}\theta \gamma ^i\theta +\frac{i}{2}X^i\theta \gamma ^{[i9]}\theta \right)}+\mathrm{},`$
$`\underset{g_s0}{lim}I_0^i`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{i}{2}\theta \gamma ^i\theta \right)}+\mathrm{}.`$ (36)
Observe that these expressions for the $`D0`$–brane current are exact in the hatted frame, unlike all the other expressions for the matrix string theory currents, which are given up to higher order terms in the coordinate fields, $`𝒪(\widehat{X}^n)`$.
The components of the matrix string theory $`D2`$–brane current are:
$`I_2^{0ij}`$ $`=`$ $`\widehat{J}^{+ij}+𝒪(\widehat{X}^{10})=J^{+ij}+J^{ij}+\mathrm{}`$
$`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}(\left({\displaystyle \frac{\mathrm{}_s}{R}}\right){\displaystyle \frac{i}{6g_s}}[X^i,X^j]+\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)\{{\displaystyle \frac{i}{6g_s}}\dot{X^i}\dot{X^k}[X^k,X^j]`$
$`{\displaystyle \frac{i}{6g_s}}\dot{X^j}\dot{X^k}[X^k,X^i]{\displaystyle \frac{1}{6}}g_s\dot{A}\dot{X^i}DX^j+{\displaystyle \frac{1}{6}}g_s\dot{A}\dot{X^j}DX^i`$
$`{\displaystyle \frac{i}{12g_s}}(\dot{X^k})^2[X^i,X^j]{\displaystyle \frac{i}{12}}g_s\dot{A}^2[X^i,X^j]{\displaystyle \frac{i}{12g_s^3}}[X^i,X^j]{\displaystyle \underset{k<l}{}}[X^k,X^l]^2`$
$`+{\displaystyle \frac{i}{12g_s}}[X^i,X^j](DX^k)^2{\displaystyle \frac{i}{6g_s^3}}[X^i,X^k][X^k,X^l][X^l,X^j]{\displaystyle \frac{i}{6g_s}}DX^iDX^k[X^k,X^j]`$
$`+{\displaystyle \frac{i}{6g_s}}DX^jDX^k[X^k,X^i]+{\displaystyle \frac{1}{12g_s}}\theta _\alpha \dot{X^k}[X^m,\theta _\beta ]\{\gamma ^{[kijm]}+\gamma ^{[jm]}\delta _{ki}\gamma ^{[im]}\delta _{kj}`$
$`+2\delta _{jm}\delta _{ki}2\delta _{im}\delta _{kj}\}_{\alpha \beta }`$
$`+{\displaystyle \frac{1}{12}}\dot{A}\theta \gamma ^{[9ijm]}[X^m,\theta ]+{\displaystyle \frac{i}{12}}\theta _\alpha \dot{X^k}D\theta _\beta \{\gamma ^{[kij9]}+\gamma ^{[j9]}\delta _{ki}\gamma ^{[i9]}\delta _{kj}\}_{\alpha \beta }`$
$`+{\displaystyle \frac{i}{4g_s^2}}\theta _\alpha [X^k,X^l][X^m,\theta _\beta ]\{\gamma ^{[jkl]}\delta _{mi}\gamma ^{[ikl]}\delta _{mj}+2\gamma ^{[lij]}\delta _{km}+2\gamma ^l\delta _{jk}\delta _{im}`$
$`2\gamma ^l\delta _{ik}\delta _{jm}+2\gamma ^j\delta _{il}\delta _{km}2\gamma ^i\delta _{jl}\delta _{km}\}_{\alpha \beta }`$
$`+{\displaystyle \frac{1}{2g_s}}\theta _\alpha DX^k[X^m,\theta _\beta ]\{\gamma ^{[jk9]}\delta _{mi}\gamma ^{[ik9]}\delta _{mj}+\gamma ^{[9ij]}\delta _{km}+\gamma ^9\delta _{jk}\delta _{im}\gamma ^9\delta _{ik}\delta _{jm}\}_{\alpha \beta }`$
$`{\displaystyle \frac{i}{2}}\theta _\alpha DX^lD\theta _\beta \{\gamma ^{[lij]}+\gamma ^j\delta _{il}\gamma ^i\delta _{jl}\}_{\alpha \beta }+\mathrm{}\})+\mathrm{},`$
$`I_2^{ijk}`$ $`=`$ $`\widehat{J}^{ijk}+𝒪(\widehat{X}^8)=J^{ijk}+\mathrm{}`$ (37)
$`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{i}{6g_s}}\dot{X^i}[X^j,X^k]{\displaystyle \frac{i}{6g_s}}\dot{X^j}[X^k,X^i]{\displaystyle \frac{i}{6g_s}}\dot{X^k}[X^i,X^j]`$
$`+{\displaystyle \frac{1}{12g_s}}\theta \gamma ^{[ijkl]}[X_l,\theta ]+{\displaystyle \frac{i}{12}}\theta \gamma ^{[ijk9]}D\theta )+\mathrm{}.`$
The conformal field theory limit for the membrane current is:
$`\underset{g_s0}{lim}I_2^{0ij}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{i}{12}}\theta _\alpha \dot{X^k}\theta _\beta \{\gamma ^{[kij9]}+\gamma ^{[j9]}\delta _{ki}\gamma ^{[i9]}\delta _{kj}\}_{\alpha \beta }`$
$`{\displaystyle \frac{i}{2}}\theta _\alpha X^l\theta _\beta \{\gamma ^{[lij]}+\gamma ^j\delta _{il}\gamma ^i\delta _{jl}\}_{\alpha \beta })+\mathrm{},`$
$`\underset{g_s0}{lim}I_2^{ijk}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{i}{12}\theta \gamma ^{[ijk9]}\theta \right)}+\mathrm{}.`$ (38)
Moving towards the $`D4`$–brane current, the components are given by:
$`I_4^{0ijkl}`$ $`=`$ $`6\widehat{M}^{+ijkl}+𝒪(\widehat{X}^8)=6M^{i9jkl}+\mathrm{}`$
$`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{30i}{g_s}}\dot{X}^{[i}[X^j,X^k]DX^{l]}{\displaystyle \frac{15}{2g_s}}\dot{A}[X^{[i},X^j][X^l,X^{k]}]`$
$`+5i\theta \dot{X}^{[i}\gamma ^{jkl]}D\theta 5\dot{A}\theta \gamma ^{[ijl}[X^{k]},\theta ]{\displaystyle \frac{15}{g_s}}\theta \dot{X}^{[i}\gamma ^{|9|kl}[X^{j]},\theta ]`$
$`+{\displaystyle \frac{15i}{2g_s^2}}\theta [X^{[i},X^j]\gamma ^{kl]9}\gamma ^n[X^n,\theta ]{\displaystyle \frac{15}{2g_s}}\theta [X^{[i},X^j]\gamma ^{kl]9}\gamma ^9D\theta `$
$`+{\displaystyle \frac{5}{g_s}}\theta DX^{[i}\gamma ^{jlk]}\gamma ^n[X^n,\theta ]+5i\theta DX^{[i}\gamma ^{jlk]}\gamma ^9D\theta )+\mathrm{},`$
$`I_4^{ijklm}`$ $`=`$ $`6\widehat{M}^{ijklm}+𝒪(\widehat{X}^{10})=6M^{ijklm}+\mathrm{}`$ (39)
$`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{15}{2g_s^2}}\dot{X}^{[i}[X^j,X^k][X^l,X^{m]}]+{\displaystyle \frac{5}{g_s}}\theta \dot{X}^{[i}\gamma ^{jkl}[X^{m]},\theta ]`$
$`+{\displaystyle \frac{5i}{2g_s^2}}\theta [X^{[i},X^j]\gamma ^{klm]}\gamma ^n[X^n,\theta ]{\displaystyle \frac{5}{2g_s}}\theta [X^{[i},X^j]\gamma ^{klm]}\gamma ^9D\theta )+\mathrm{}.`$
The conformal field theory limit for the 4–brane current is:
$`\underset{g_s0}{lim}I_4^{0ijkl}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\mathrm{\hspace{0.17em}5}i\left(\theta \dot{X}^{[i}\gamma ^{jkl]}\theta +\theta X^{[i}\gamma ^{jlk]}\gamma ^9\theta \right)}+\mathrm{},`$
$`\underset{g_s0}{lim}I_4^{ijklm}`$ $`=`$ $`𝒪(X^{10})+\mathrm{}.`$ (40)
Next we analyze the $`D6`$–brane current in matrix string theory. The components of this current are given by:
$`I_6^{0ijklmn}`$ $`=`$ $`\widehat{S}^{+ijklmn}+𝒪(\widehat{X}^{10})=S^{+ijklmn}+\mathrm{}`$
$`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{i}{g_s^3}[X^{[i},X^j][X^k,X^l][X^m,X^{n]}]\right)}+\mathrm{},`$
$`I_6^{ijklmnp}`$ $`=`$ $`\widehat{S}^{ijklmnp}+𝒪(\widehat{X}^{12})=S^{ijklmnp}+\mathrm{}`$ (41)
$`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)^2{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{7i}{g_s^3}[X^{[i},X^j][X^k,X^l][X^m,X^n]\dot{X}^{p]}+𝒪(\theta ^2,\theta ^4)\right)}`$
$`+\mathrm{}.`$
As to the conformal field theory limit for the 6–brane current, it is:
$`\underset{g_s0}{lim}I_6^{0ijklmn}`$ $`=`$ $`𝒪(X^{10})+\mathrm{},`$
$`\underset{g_s0}{lim}I_6^{ijklmnp}`$ $`=`$ $`𝒪(X^{12})+\mathrm{}.`$ (42)
A Matrix theory form for the transverse $`M5`$–brane current components $`M^{+ijklm}`$ and $`M^{ijklmn}`$ is not known (though it is believed that these operators identically vanish in this light–front gauge). Therefore, we cannot determine the $`NS5`$–brane current components $`I_5^{ijklmn}`$ and $`I_5^{0ijklm}`$ (though most likely these operators will vanish in the Type IIA description as well, at least to the lowest order we are considering here). For further discussions on these points, see .
With these matrix string currents, the sigma model action for matrix string theory in weakly curved backgrounds is then simply written as:
$`𝒮`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau ({\displaystyle \frac{1}{2}}g_{\mu \nu }^{IIA}(X)I_g^{\mu \nu }+\varphi (X)I_\varphi +B_{\mu \nu }(X)I_s^{\mu \nu }+\stackrel{~}{B}_{\mu \nu \lambda \rho \sigma \tau }(X)I_5^{\mu \nu \lambda \rho \sigma \tau }`$ (43)
$`+C_\mu (X)I_0^\mu +\stackrel{~}{C}_{\mu \nu \lambda \rho \sigma \tau \xi }(X)I_6^{\mu \nu \lambda \rho \sigma \tau \xi }+C_{\mu \nu \lambda }(X)I_2^{\mu \nu \lambda }+\stackrel{~}{C}_{\mu \nu \lambda \rho \sigma }(X)I_4^{\mu \nu \lambda \rho \sigma }),`$
where the notation is as follows. The metric is $`g_{\mu \nu }=\eta _{\mu \nu }+h_{\mu \nu }`$, so that the first term includes naturally, to linear order in $`h_{\mu \nu }`$, the term relative to the matrix string action in flat space and the linear coupling term $`h_{\mu \nu }(X)I_h^{\mu \nu }`$ previously derived. Also, we have seen that the currents $`I`$ that we derived were integrated currents, including an explicit world–sheet integration. In (43) this world–sheet integration, as well as the $`\frac{1}{2\pi }`$ factor, have been brought out of the expressions for the current, in order to stress that we have obtained a two dimensional matrix gauged sigma model field theory. Finally, recall from what is the prescription to include explicit spacetime dependence in all the NS–NS and R–R fields. We should take the following definition:
$$\varphi (X)I_\varphi \underset{n=0}{\overset{\mathrm{}}{}}\frac{1}{n!}(_{k_1}\mathrm{}_{k_n}\varphi )(0)I_\varphi ^{(k_1\mathrm{}k_n)},$$
(44)
where $`I_\varphi ^{(k_1\mathrm{}k_n)}`$ are the higher moments of the matrix string current for the dilaton. Similarly for all the other fields. Observe that in this work we have just analyzed zeroth moments of the Matrix and matrix string currents. Moreover, we shall assume for the remainder of the paper that the background fields satisfy the source–free equations of motion of Type IIA supergravity. In this case, the dual fields $`\stackrel{~}{C}^{D6}`$, $`\stackrel{~}{B}^{NS5}`$ and $`\stackrel{~}{C}^{D4}`$ are well defined $`(p+1)`$–form fields given at linear order by,
$$d\stackrel{~}{C}^{D6}=dC^{D0},d\stackrel{~}{B}^{NS5}=dB,d\stackrel{~}{C}^{D4}=dC^{D2}.$$
(45)
In one sentence, the tensors we have just derived allow us to build a matrix sigma model for the IIA string. Recall from the standard sigma model approach that the background fields are an infinite number of couplings from the point of view of the world–sheet quantum field theory. If the target space has characteristic curvature $`R`$, then derivatives of the metric will be of order $`\frac{1}{R}`$, and so the effective dimensionless coupling in the theory will be $`\frac{\sqrt{\alpha ^{}}}{R}=\frac{\mathrm{}_s}{R}`$, quite similar to what we have obtained (even though previously $`R`$ was simply the radius of the compact dimension). For $`R\mathrm{}_s`$ the effective coupling is small and perturbation theory on the world–sheet is useful. In this regime, the string is effectively point–like, and one can also use the low energy effective field theory to deal with the problem. To these results there will naturally be stringy corrections. They can be obtained from the multi–loop corrections to the world–sheet beta functions, as a power series in our effective coupling $`\frac{\mathrm{}_s}{R}`$.
One expects that a similar story should take place in the case of the matrix sigma model we have derived, that describes multiple string interactions in non–trivial background fields. Indeed, it would be quite interesting to derive matrix beta functions, and therefore matrix equations of motion for the background fields. While at the level of Matrix theory one could expect that a large $`N`$ renormalization group analysis should be required , here at the level of matrix string theory it should only be a direct generalization of the one string sigma model field theory.
## 5 Reduction via the $`T`$$`S`$$`T`$ Duality Sequence
If one recalls section 3, performing the DVV reduction via the $`911`$ flip should be equivalent to a specific set of dualities, namely $`T`$–duality from IIA to IIB, $`S`$–duality of IIB, and then $`T`$–duality back to the IIA theory. We would now like to explicitly check such a procedure in the presence of non–trivial backgrounds. This would amount to a further check of the internal structure of string dualities. Basically, all one needs to do is DVV reduce according to the sequence of dualities as applied to the background fields and to the world–volume fields, and observe that one will obtain the same result as in the previous section. The transformation of the background fields under $`T`$ and $`S`$ dualities is well known. As to the transformation of the world–volume fields, it is well known for the case of $`T`$–duality as discussed for the general case in . For $`S`$–duality, it is not known how the world–volume fields transform, as we are dealing with a nonabelian gauge theory. For the case of the $`D3`$–brane, this has been discussed recently in . In here, we shall be dealing with a $`D1`$–brane. We will obtain the $`S`$–duality transformations for the world–volume fields of the 2–dimensional gauge theory by demanding consistency with the whole structure of matrix string theory. Also, we shall work this out explicitly only for a couple of terms, not for the whole series of components of the several matrix string supercurrents. Moreover, we will neglect all terms involving fermions throughout this section. From the previous section we already known how they appear in the tensor structures that compose matrix string theory, so that we can always take them from there when they are required in the following sections. However, for the purpose of checking the duality sequence it should be enough to look at the bosonic part of the action alone. Extending our results to all the components and including fermions should present no obvious obstacles.
We begin by following closely , in particular their discussion of implementing $`T`$ and $`S`$ dualities for the linear supergravity backgrounds, and somewhat the implementation of $`T`$–duality at the world–volume level. Then, due to the matching between the duality sequence and the $`911`$ flip, we determine the $`S`$–duality transformation rules at the level of the 2–dimensional world–volume theory.
Let us focus on $`T`$–duality, and how it acts both on the supergravity background fields and on the fields that live on the $`D`$–brane. Recall that when we begin the sequence of dualities we are looking at the world–volume theory of $`D0`$–branes. $`T`$–duality acting on arbitrary backgrounds independent of the compact directions is well known in string theory . The standard $`T`$–duality rules can be linearized and these are the transformations we shall be interested in, given that we are working in weak background fields. Using barred indices $`\overline{\alpha },\overline{\beta },\mathrm{}`$, for the compact directions in which a $`T`$–duality is performed and indices $`\mu ,\nu ,\mathrm{}`$, for the remaining $`10p`$ spacetime dimensions (including 0), we can write the action of $`T`$–duality in the linear supergravity background fields as ,
$`h_{\mu \nu }`$ $``$ $`h_{\mu \nu }`$
$`B_{\mu \nu }`$ $``$ $`B_{\mu \nu }`$
$`h_{\mu \overline{\alpha }}`$ $``$ $`B_{\mu \overline{\alpha }}`$
$`h_{\overline{\alpha }\mu }`$ $``$ $`B_{\overline{\alpha }\mu }`$
$`h_{\overline{\alpha }\overline{\beta }}`$ $``$ $`h_{\overline{\alpha }\overline{\beta }}`$
$`B_{\overline{\alpha }\overline{\beta }}`$ $``$ $`B_{\overline{\alpha }\overline{\beta }}`$
$`\varphi `$ $``$ $`\varphi {\displaystyle \frac{1}{2}}{\displaystyle \underset{\overline{\alpha }}{}}h_{\overline{\alpha }\overline{\alpha }}`$
$`C_{\mu _1\mathrm{}\mu _{qk}\overline{\alpha }_1\mathrm{}\overline{\alpha }_k}^{(q)}`$ $``$ $`{\displaystyle \frac{1}{(nk)!}}ϵ^{\overline{\alpha }_1\mathrm{}\overline{\alpha }_k}C_{\mu _1\mathrm{}\mu _{qk}\overline{\alpha }_{k+1}\mathrm{}\overline{\alpha }_n}^{(q2k+n)}`$ (46)
where the $`(q)`$ superscript indicates the $`q`$–form R–R field associated to a $`D(q1)`$–brane.
The implementation being clear at the background level, let us look at the world–volume level. The low energy effective field theory living on the world–volume of a $`Dp`$–brane in flat background space is the dimensional reduction of 10–dimensional SYM theory to the $`p+1`$ world–volume dimensions. One thing one can do is to retain 10–dimensional notation for all the $`Dp`$–brane world–volume theories and reinterpret the resulting expressions appropriately for each case, i.e., if we choose $`a,b,\mathrm{}`$, as world–volume indices and $`i,j,\mathrm{}`$, as indices transverse to the brane, then we would reinterpret expressions such as $`F_{ai}D_aX^i`$ and $`F_{ij}i[X^i,X^j]`$. We therefore see that the action of $`T`$–duality on expressions which have been written in terms of the 10–dimensional notation simply amounts to an adequate reintrepertation of such a notation. There is only one point one should take into account, namely we should be careful when considering transverse fields $`X^i`$ associated with a compact direction. The precise way in which one should deal with such fields has been described in . Briefly stated, transverse fields associated with a compact direction can be Fourier decomposed so that they are $`T`$–dual to the momentum modes of the dual gauge field that lives on the $`T`$–dual brane. The Matrix theory expressions for the moments of the 11–dimensional supergravity currents that we have used in the previous section can all be written easily in the 10–dimensional language, and this has been done in . We refer to the appendix for those expressions.
Given the $`D0`$–brane action in weak background fields, we can now write down the $`T`$–dual action for the Type IIB $`D`$–string. This has in fact been done for any $`Dp`$–brane , as briefly discussed in section 2. The $`D`$–string action is therefore,
$`S_{NSNS}^{D1}`$ $`=`$ $`(\varphi {\displaystyle \frac{1}{2}}h_{\widehat{a}\widehat{a}})I_\varphi +{\displaystyle \frac{1}{2}}h_{00}I_h^{00}+{\displaystyle \frac{1}{2}}h_{ij}I_h^{ij}{\displaystyle \frac{1}{2}}h_{\widehat{a}\widehat{b}}I_h^{\widehat{a}\widehat{b}}+h_{0i}I_h^{0i}+2h_{\widehat{a}i}I_s^{\widehat{a}i}2h_{0\widehat{a}}I_s^{0\widehat{a}}`$ (47)
$`+B_{ij}I_s^{ij}B_{\widehat{a}\widehat{b}}I_s^{\widehat{a}\widehat{b}}+2B_{0i}I_s^{0i}+B_{\widehat{a}i}I_h^{\widehat{a}i}B_{0\widehat{a}}I_h^{0\widehat{a}}`$
$`+\mathrm{Higher}\mathrm{moment}\mathrm{terms}+\mathrm{Nonlinear}\mathrm{terms},`$
$`S_{RR}^{D1}`$ $`=`$ $`{\displaystyle }d^2\sigma ϵ^{a_0a_1}{\displaystyle \underset{q}{}}{\displaystyle \underset{n=\mathrm{Max}(0,q2)}{\overset{n=\mathrm{Min}(q,8)}{}}}{\displaystyle \frac{(1)^{\frac{n(n1)}{2}}(2nq+1)!!}{n!(qn)!(nq+2)!}}\mathrm{𝐒𝐓𝐫}\{C_{a_0\mathrm{}a_{qn1}i_1\mathrm{}i_n}^{(q)}`$ (48)
$`F_{(a_{qn}\mathrm{}a_1i_1\mathrm{}i_n)}^{n+\frac{2q}{2}}\}+\mathrm{Higher}\mathrm{moment}\mathrm{terms}+\mathrm{Nonlinear}\mathrm{terms},`$
where the indices in curved brackets are to be assigned pairwise to the corresponding product of $`F`$’s and then symmetrized over all orderings. Indices $`0,\widehat{a},\mathrm{}`$, live on the 2–dimensional world–volume, while indices $`i,j,\mathrm{}`$, are transverse to the $`D`$–string. The $`I`$ currents in these expressions should not be confused with the $`I`$ currents derived in the previous section. In here we started with the $`D0`$–brane currents mentioned in section 2, which in turn can be determined in terms of the Matrix currents $`T`$, $`J`$ and $`M`$. As we wrote the previous expressions, the 0–brane currents $`I`$ are then to be reinterpreted as 10–dimensional expressions reduced to the 2–dimensional world–volume of the $`D1`$–brane. As to the higher moment terms, the expressions will be just like the ones above, but with the appropriate inclusion of arbitrary derivatives of each background field.
It is therefore clear that in order to explicitly write the $`D`$–string action, it would be useful to start with the 10–dimensional expressions for the $`I`$ currents. These can be obtained from the expressions in , and they are as follows. Observe that we write down the expressions in the previously explained 10–dimensional notation, and so when reducing to the 2–dimensional world–volume one should take into consideration the compact direction carefully, as was mentioned before.
The $`I_h^{00}`$ component of the matrix string current for background metric field, can be written in 10–dimensional notation as (we dropped a factor of $`1/R`$ from all the expressions that follow for the $`I`$ currents):
$`I_h^{00}`$ $`=`$ $`T^{++}+T^++(I_h^{00})_8+𝒪(X^{12})`$ (49)
$`=`$ $`\mathrm{𝐒𝐓𝐫}(\text{1}\text{ }\text{1}+F^{0\mu }F^0{}_{\mu }{}^{}+{\displaystyle \frac{1}{4}}F_{\mu \nu }F^{\mu \nu }+𝒪(\mathrm{\Theta }^2))+(I_h^{00})_8+𝒪(X^{12}).`$
All components of these matrix string currents are straightforward to write down, so we simply present them. For the NS–NS sector, the matrix string current components can be written in 10–dimensional notation as:
$`I_h^{00}`$ $`=`$ $`T^{++}+T^++(I_h^{00})_8+𝒪(X^{12})`$
$`=`$ $`\mathrm{𝐒𝐓𝐫}(\text{1}\text{ }\text{1}+F^{0\mu }F^0{}_{\mu }{}^{}+{\displaystyle \frac{1}{4}}F_{\mu \nu }F^{\mu \nu }+𝒪(\mathrm{\Theta }^2))+(I_h^{00})_8+𝒪(X^{12}),`$
$`I_h^{0i}`$ $`=`$ $`T^{+i}+T^i+𝒪(X^{10})`$
$`=`$ $`\mathrm{𝐒𝐓𝐫}(F^{0i}+F^{0\mu }F_{\mu \nu }F^{\nu i}+{\displaystyle \frac{1}{4}}F^{0i}F_{\mu \nu }F^{\mu \nu }+𝒪(\mathrm{\Theta }^2,\mathrm{\Theta }^4))+𝒪(X^{10}),`$
$`I_h^{ij}`$ $`=`$ $`T^{ij}+(I_h^{ij})_8+𝒪(X^{12})=\mathrm{𝐒𝐓𝐫}(F^{i\mu }F_\mu {}_{}{}^{j}+𝒪(\mathrm{\Theta }^2))+(I_h^{ij})_8+𝒪(X^{12}),`$
$`I_\varphi `$ $`=`$ $`T^{++}{\displaystyle \frac{1}{3}}(T^++T^{ii})+(I_\varphi )_8+𝒪(X^{12})`$
$`=`$ $`\mathrm{𝐒𝐓𝐫}(\text{1}\text{ }\text{1}+{\displaystyle \frac{1}{4}}F_{\mu \nu }F^{\mu \nu }+𝒪(\mathrm{\Theta }^2))+(I_\varphi )_8+𝒪(X^{12}),`$
$`I_s^{0i}`$ $`=`$ $`3J^{+i}+𝒪(X^8)={\displaystyle \frac{1}{2}}\mathrm{𝐒𝐓𝐫}(F^{0\mu }F_\mu {}_{}{}^{i}+𝒪(\mathrm{\Theta }^2))+𝒪(X^8),`$
$`I_s^{ij}`$ $`=`$ $`3J^{+ij}3J^{ij}+𝒪(X^{10})`$ (50)
$`=`$ $`{\displaystyle \frac{1}{2}}\mathrm{𝐒𝐓𝐫}(F^{ij}+F^{i\mu }F_{\mu \nu }F^{\nu j}+{\displaystyle \frac{1}{4}}F^{ij}F_{\mu \nu }F^{\mu \nu }+𝒪(\mathrm{\Theta }^2,\mathrm{\Theta }^4))+𝒪(X^{10}).`$
Moving to the R–R fields, one can write the matrix string currents in the 10–dimensional notation as:
$`I_0^0`$ $`=`$ $`T^{++}=\mathrm{𝐒𝐓𝐫}(\text{1}\text{ }\text{1}),`$
$`I_0^i`$ $`=`$ $`T^{+i}=\mathrm{𝐒𝐓𝐫}(F^{0i}),`$
$`I_2^{0ij}`$ $`=`$ $`J^{+ij}+𝒪(X^{10})={\displaystyle \frac{1}{6}}\mathrm{𝐒𝐓𝐫}(F^{ij})+𝒪(X^{10}),`$
$`I_2^{ijk}`$ $`=`$ $`J^{ijk}+𝒪(X^8)={\displaystyle \frac{1}{6}}\mathrm{𝐒𝐓𝐫}(F^{0i}F^{jk}+F^{0j}F^{ki}+F^{0k}F^{ij}+𝒪(\mathrm{\Theta }^2))+𝒪(X^8),`$
$`I_4^{0ijkl}`$ $`=`$ $`6M^{+ijkl}+𝒪(X^8)={\displaystyle \frac{1}{2}}\mathrm{𝐒𝐓𝐫}(F^{ij}F^{kl}+F^{ik}F^{lj}+F^{il}F^{jk}+𝒪(\mathrm{\Theta }^2))+𝒪(X^8),`$
$`I_4^{ijklm}`$ $`=`$ $`6M^{ijklm}+𝒪(X^{10})={\displaystyle \frac{15}{2}}\mathrm{𝐒𝐓𝐫}(F^{0[i}F^{jk}F^{lm]}+𝒪(\mathrm{\Theta }^2))+𝒪(X^{10}),`$
$`I_6^{0ijklmn}`$ $`=`$ $`S^{+ijklmn}+𝒪(X^{10})=\mathrm{𝐒𝐓𝐫}\left(F_{[ij}F_{kl}F_{mn]}\right)+𝒪(X^{10}),`$
$`I_6^{ijklmnp}`$ $`=`$ $`S^{ijklmnp}+𝒪(X^{12})=7\mathrm{𝐒𝐓𝐫}\left(F_{[ij}F_{kl}F_{mn}\dot{X}_{p]}+𝒪(\theta ^2,\theta ^4)\right)+𝒪(X^{12}).`$ (51)
As we discussed in the previous section, finding a Matrix theory form for the transverse $`M5`$–brane current components $`M^{+ijklm}`$ and $`M^{ijklmn}`$ is a matter which is yet not quite fully understood.
All these straighten out, we can proceed and explicitly write down (47) and (48) for this case of the IIB $`D`$–string. From the previous expressions, one obtains:
$`S_{NSNS}^{D1}`$ $`=`$ $`(\varphi {\displaystyle \frac{1}{2}}h_a{}_{}{}^{a})\mathrm{𝐒𝐓𝐫}(\text{1}\text{ }\text{1}+{\displaystyle \frac{1}{4}}F_{\mu \nu }F^{\mu \nu }+\mathrm{})`$ (52)
$`+{\displaystyle \frac{1}{2}}h_{ij}\mathrm{𝐒𝐓𝐫}\left(F^{i\mu }F_\mu {}_{}{}^{j}+\mathrm{}\right){\displaystyle \frac{1}{2}}h_{ab}\mathrm{𝐒𝐓𝐫}\left(F^{a\mu }F_\mu {}_{}{}^{b}+\mathrm{}\right)`$
$`h_{ai}\mathrm{𝐒𝐓𝐫}\left(F^{ai}+F^{a\mu }F_{\mu \nu }F^{\nu i}+{\displaystyle \frac{1}{4}}F^{ai}F_{\mu \nu }F^{\mu \nu }+\mathrm{}\right)`$
$`{\displaystyle \frac{1}{2}}B_{ij}\mathrm{𝐒𝐓𝐫}\left(F^{ij}+F^{i\mu }F_{\mu \nu }F^{\nu j}+{\displaystyle \frac{1}{4}}F^{ij}F_{\mu \nu }F^{\mu \nu }+\mathrm{}\right)`$
$`+{\displaystyle \frac{1}{2}}B_{ab}\mathrm{𝐒𝐓𝐫}\left(F^{ab}+F^{a\mu }F_{\mu \nu }F^{\nu b}+{\displaystyle \frac{1}{4}}F^{ab}F_{\mu \nu }F^{\mu \nu }+\mathrm{}\right)`$
$`+B_{ai}\mathrm{𝐒𝐓𝐫}\left(F^{a\mu }F_\mu {}_{}{}^{i}+\mathrm{}\right)+\mathrm{},`$
$`S_{RR}^{D1}`$ $`=`$ $`{\displaystyle \frac{1}{2}}ϵ^{ab}C^{D(1)}\mathrm{𝐒𝐓𝐫}(F_{ab}+\mathrm{})+{\displaystyle \frac{1}{2}}ϵ^{ab}C_{ab}^{D1}\mathrm{𝐒𝐓𝐫}(\text{1}\text{ }\text{1}+\mathrm{})+{\displaystyle \frac{1}{2}}ϵ^{ab}C_{ai}^{D1}\mathrm{𝐒𝐓𝐫}\left(F_b{}_{}{}^{i}+\mathrm{}\right)`$ (53)
$`+{\displaystyle \frac{1}{2}}ϵ^{ab}C_{ij}^{D1}\mathrm{𝐒𝐓𝐫}(F_a{}_{}{}^{i}F_{b}^{}{}_{}{}^{j}{\displaystyle \frac{1}{2}}F_{ab}F^{ij}+\mathrm{})+\mathrm{},`$
where in the previous expressions one should still take into consideration that the tensors must be appropriately reduced to 2–dimensional world–volume notation via the identifications $`F_{ai}D_aX^i`$ and $`F_{ij}i[X^i,X^j]`$, and the proper treatment of the compact coordinate $`X^9`$ according to . Integration over the cylindrical world–sheet $`\{\tau ,\sigma \}`$ is implicit. Of course the action (52) and (53) is quite interesting on its own, as it yields the gauged matrix sigma model for the Type IIB $`D`$–string in weakly curved backgrounds.
The issue of $`T`$–duality along $`R_9^{IIA}`$ solved, let us now deal with the IIB $`S`$–duality transformation. The $`SL(2,𝐙)`$ duality symmetry of Type IIB string theory maps a $`(p,q)`$–string into another $`(p^{},q^{})`$–string. In here, we shall focus on the usual $`𝐙_2`$ subgroup of the $`S`$–duality group generated by the transformation which exchanges the NS–NS and R–R 2–form fields, and in particular maps the $`D`$–string into the fundamental string. As in the case of $`T`$–duality, the action of this subgroup of $`S`$–duality on arbitrary IIB supergravity background fields is well known . At linear order the transformation is ,
$`\varphi `$ $``$ $`\varphi `$
$`C^{(0)}`$ $``$ $`C^{(0)}`$
$`B_{\mu \nu }`$ $``$ $`C_{\mu \nu }^{(2)}`$
$`C_{\mu \nu }^{(2)}`$ $``$ $`B_{\mu \nu }`$
$`h_{\mu \nu }`$ $``$ $`h_{\mu \nu }`$
$`C^{(4)}`$ $``$ $`C^{(4)}.`$ (54)
These transformations are written in the Einstein frame, even though we have been working in the string frame. This is fine as the terms that we are considering from the string action, in this paper, are the same in both frames.
The problem we face concerns the implementation of $`S`$–duality at the $`D`$–string world–volume level. In fact, the $`S`$–duality transformation properties of the world–volume operators in the 2–dimensional $`U(N)`$ gauge theory are not known. What we shall see, is that in the end the $`S`$–duality transformations are not so surprising, and they turn out to be quite simple as there will be no change in the composite operators. But for the moment, let us assume that they could be anything.
In here, for the $`D1`$–brane, because we know what the result is from matrix string theory we can predict the precise transformation properties of all the operators that appear in the action. For the moment, we shall perform the IIB $`S`$–duality transformation of the $`D`$–string action (52) and (53) in the following way. We apply the linear $`S`$–duality transformations (5) to the background fields, and we denote $`S`$–duals of the world–volume fields by simply putting a tilde over them. Next, as we $`T`$–dualize back to the IIA theory, we can then compare to the previous section and obtain predictions for all these “tilded” operators. $`S`$–dualizing the $`D`$–string action in this way, one obtains:
$`S_{\stackrel{~}{NSNS}}^{F1}`$ $`=`$ $`(\varphi {\displaystyle \frac{1}{2}}h_a{}_{}{}^{a})\mathrm{𝐒𝐓𝐫}(\text{1}\text{ }\text{1}+{\displaystyle \frac{1}{4}}\stackrel{~}{F_{\mu \nu }F^{\mu \nu }}+\mathrm{})`$ (55)
$`+{\displaystyle \frac{1}{2}}h_{ij}\mathrm{𝐒𝐓𝐫}\left(\stackrel{~}{F^{i\mu }F_\mu ^j}+\mathrm{}\right){\displaystyle \frac{1}{2}}h_{ab}\mathrm{𝐒𝐓𝐫}\left(\stackrel{~}{F^{a\mu }F_\mu ^b}+\mathrm{}\right)`$
$`h_{ai}\mathrm{𝐒𝐓𝐫}\left(\stackrel{~}{F^{ai}}+\stackrel{~}{F^{a\mu }F_{\mu \nu }F^{\nu i}}+{\displaystyle \frac{1}{4}}\stackrel{~}{F^{ai}F_{\mu \nu }F^{\mu \nu }}+\mathrm{}\right)`$
$`+{\displaystyle \frac{1}{2}}C_{ij}\mathrm{𝐒𝐓𝐫}\left(\stackrel{~}{F^{ij}}+\stackrel{~}{F^{i\mu }F_{\mu \nu }F^{\nu j}}+{\displaystyle \frac{1}{4}}\stackrel{~}{F^{ij}F_{\mu \nu }F^{\mu \nu }}+\mathrm{}\right)`$
$`{\displaystyle \frac{1}{2}}C_{ab}\mathrm{𝐒𝐓𝐫}\left(\stackrel{~}{F^{ab}}+\stackrel{~}{F^{a\mu }F_{\mu \nu }F^{\nu b}}+{\displaystyle \frac{1}{4}}\stackrel{~}{F^{ab}F_{\mu \nu }F^{\mu \nu }}+\mathrm{}\right)`$
$`C_{ai}\mathrm{𝐒𝐓𝐫}\left(\stackrel{~}{F^{a\mu }F_\mu ^i}+\mathrm{}\right)+\mathrm{},`$
$`S_{\stackrel{~}{RR}}^{F1}`$ $`=`$ $`{\displaystyle \frac{1}{2}}ϵ^{ab}C^{D(1)}\mathrm{𝐒𝐓𝐫}\left(\stackrel{~}{F_{ab}}+\mathrm{}\right)+{\displaystyle \frac{1}{2}}ϵ^{ab}B_{ab}\mathrm{𝐒𝐓𝐫}\left(\text{1}\text{ }\text{1}+\mathrm{}\right)+{\displaystyle \frac{1}{2}}ϵ^{ab}B_{ai}\mathrm{𝐒𝐓𝐫}\left(\stackrel{~}{F_b^i}+\mathrm{}\right)`$ (56)
$`+{\displaystyle \frac{1}{2}}ϵ^{ab}B_{ij}\mathrm{𝐒𝐓𝐫}\left(\stackrel{~}{F_a{}_{}{}^{i}F_{b}^{}^j}{\displaystyle \frac{1}{2}}\stackrel{~}{F_{ab}F^{ij}}+\mathrm{}\right)+\mathrm{},`$
where we also denoted the action subscripts with a tilde, as we have a mixing of the NS–NS and R–R sectors under $`S`$–duality. A note on notation: in case it is not clear, in the previous expressions the tilde is over the whole composite operator.
Proceeding, we are left with a final $`T`$–duality along $`R_9^{IIB}`$ that leads back to the Type IIA theory, and therefore to matrix string theory. The rules for $`T`$–duality have been previously explained and used, so we just apply them to the previous expressions to obtain the matrix string theory action in weak background fields. Two points should still be stressed. We have been slightly abusive of notation in the previous expressions by allowing more than one compact direction, as we indexed tensors as $`\widehat{a},\widehat{b},\mathrm{}`$. Of course in the case we are dealing with there is only one, $`\sigma `$. Moreover, we still have to write the world–volume tensors in the 2–dimensional world–sheet notation. Once we make the notation completely rigorous, we are left with (recall that in the matrix string limit $`R_90`$),
$`𝒮`$ $`=`$ $`\varphi \mathrm{𝐒𝐓𝐫}\left(\text{1}\text{ }\text{1}+{\displaystyle \frac{1}{2}}\stackrel{~}{\dot{X}_i^2}{\displaystyle \frac{1}{2}}\stackrel{~}{(DX^i)^2}+{\displaystyle \frac{1}{2}}g_s^2\stackrel{~}{\dot{A}^2}+{\displaystyle \frac{1}{2g_s^2}}{\displaystyle \underset{i<j}{}}\stackrel{~}{[X^i,X^j]^2}+\mathrm{}\right)`$ (57)
$`+{\displaystyle \frac{1}{2}}h_{00}\mathrm{𝐒𝐓𝐫}\left(\text{1}\text{ }\text{1}+{\displaystyle \frac{1}{2}}\stackrel{~}{\dot{X}_i^2}+{\displaystyle \frac{1}{2}}\stackrel{~}{(DX^i)^2}+{\displaystyle \frac{1}{2}}g_s^2\stackrel{~}{\dot{A}^2}{\displaystyle \frac{1}{2g_s^2}}{\displaystyle \underset{i<j}{}}\stackrel{~}{[X^i,X^j]^2}+\mathrm{}\right)`$
$`+{\displaystyle \frac{1}{2}}h_{ij}\mathrm{𝐒𝐓𝐫}\left(\stackrel{~}{\dot{X}^i\dot{X}^j}\stackrel{~}{(DX^i)(DX^j)}{\displaystyle \frac{1}{g_s^2}}\stackrel{~}{[X^i,X^k][X^k,X^j]}+\mathrm{}\right)`$
$`+h_{0i}\mathrm{𝐒𝐓𝐫}(\stackrel{~}{\dot{X}^i}+{\displaystyle \frac{1}{2}}g_s^2\stackrel{~}{\dot{A}^2\dot{X}^i}i\stackrel{~}{\dot{A}(DX^j)[X^j,X^i]}+{\displaystyle \frac{1}{2}}\stackrel{~}{(\dot{X}^j)^2\dot{X}^i}\stackrel{~}{\dot{X}^j(DX^j)(DX^i)}`$
$`+{\displaystyle \frac{1}{2}}\stackrel{~}{(DX^j)^2\dot{X}^i}{\displaystyle \frac{1}{g_s^2}}\stackrel{~}{\dot{X}^j[X^j,X^k][X^k,X^i]}{\displaystyle \frac{1}{2g_s^2}}\stackrel{~}{\dot{X}^i{\displaystyle \underset{j<k}{}}[X^j,X^k]^2}+\mathrm{})`$
$`+B_{0i}\mathrm{𝐒𝐓𝐫}({\displaystyle \frac{1}{2}}\stackrel{~}{(DX^i)}+\mathrm{})+B_{ij}\mathrm{𝐒𝐓𝐫}({\displaystyle \frac{1}{2}}\stackrel{~}{\dot{X}^i(DX^j)}+{\displaystyle \frac{1}{2}}\stackrel{~}{\dot{X}^j(DX^i)}{\displaystyle \frac{i}{2}}\stackrel{~}{\dot{A}[X^i,X^j]}`$
$`+\mathrm{})+C_i\mathrm{𝐒𝐓𝐫}(g_s\stackrel{~}{\dot{A}\dot{X}^i}{\displaystyle \frac{i}{g_s}}\stackrel{~}{(DX^j)[X^j,X^i]}+\mathrm{})+C_0\mathrm{𝐒𝐓𝐫}(g_s\stackrel{~}{\dot{A}}+{\displaystyle \frac{1}{2}}g_s^3\stackrel{~}{\dot{A}^3}`$
$`{\displaystyle \frac{1}{2}}g_s\stackrel{~}{\dot{A}(DX^i)^2}+{\displaystyle \frac{1}{2}}g_s\stackrel{~}{\dot{A}(\dot{X}^i)^2}+{\displaystyle \frac{i}{g_s}}\stackrel{~}{\dot{X}^i[X^i,X^j]DX^j}{\displaystyle \frac{1}{2g_s}}\stackrel{~}{\dot{A}{\displaystyle \underset{i<j}{}}[X^i,X^j]^2}+\mathrm{})`$
$`+\mathrm{}`$
This completes the sequence of DVV dualities. The final result should be equivalent to performing the $`911`$ flip of section 4. Comparing this result to the one in section 4, (43), one will obtain the $`S`$–duality transformation rules for the tensor operators that live on the $`D`$–string world–volume. One should only keep in mind that in this section we have dropped a factor of $`1/R`$ from all the expressions, and that one may need to correct for units as comparing to the previous section.
From the expression (57) we have results for the $`I_\varphi `$, $`I_h^{00}`$, $`I_h^{ij}`$, $`I_h^{0i}`$, $`I_s^{0i}`$, $`I_s^{ij}`$, $`I_0^i`$ and $`I_0^0`$ tensors coming from the $`T`$$`S`$$`T`$ chain of dualities. It should be a straightforward exercise to include all other matrix string tensors in this result. For our purposes this is enough. Comparing back to what we have obtained for those same tensors in section 4 – where we used the $`911`$ flip to DVV reduce – we obtain an interesting result: the 2–dimensional $`D`$–string world–volume composite operators are invariant under the target space IIB $`S`$–duality operation. It is indeed somewhat expected that these operators should not change, as from the field theory side we expect non–trivial $`S`$–duality properties only for the $`𝒩=4`$, $`d=4`$ gauge theory, i.e., we only expect to see non–trivial transformation laws for the operators that live on the world–volume of the $`D3`$–brane. For the $`D1`$–brane, the operators are kept fixed under the target transformation.
## 6 Green–Schwarz Action in a Curved Background
Given that the matrix string theory action we have built has been firmly established at the level of string duality, we would further like to confirm it by looking at its conformal field theory limit, where we obtain the free string case. We expect that when we take the $`g_s0`$ limit, our action should match the Green–Schwarz sigma model for the Type IIA superstring , once we consider this latter one in the same weak background approximation that we are considering in here.
### 6.1 The Green–Schwarz Action
We begin with a short review of the results obtained for the IIA superstring in , so that we can establish a bridge between the results of that paper and our notation. We want to compare the Green–Schwarz sigma model to our matrix sigma model, and for that all one needs to do is to consider the matrix sigma model in the free string limit which was presented throughout section 4 for all the tensor fields. One should also take into consideration the weak field approximation in the sigma model for one string. It should be stressed that we shall be looking at schematic and qualitative results only, throughout this section. This is because establishing a precise map from the matrix string theory in the IR to the light–cone Green–Schwarz action requires a precise lifting of the IR matrix action from the cylinder to its branched coverings, and so to any given Riemann surface. The procedure is described at length in for the flat background situation. Completing such a procedure for this curved situation is an interesting project for the future.
The massless spectra of the Type IIA closed superstring includes the metric, $`g_{\mu \nu }`$, the NS $`B`$–field, $`B_{\mu \nu }`$, and the dilaton, $`\varphi `$, whose origin is the bosonic sector of the superstring action. It also includes the $`D0`$–brane 1–form, $`C_\mu `$, and the $`D2`$–brane 3–form, $`C_{\mu \nu \rho }`$, whose origin is the fermionic sector of the same superstring action. The covariant superstring action can be written while in the presence of couplings to the background fields of $`𝒩=2`$ 10–dimensional supergravity, as was shown in . In here we are interested in the form of this action in light–cone gauge, which was also derived in : If one chooses light–cone gauge, and furthermore assumes the supergravity background fields to be non–trivial only in the eight transverse directions (so that the background spacetimes decomposes as $`^{10}=𝐑^{(1,1)}\times ^8`$), then the NS–NS sector of the IIA superstring action is written as ,
$`_{NSNS}`$ $`=`$ $`g_{ij}(X)\sigma ^{ab}_aX^i_bX^j+4\pi \alpha ^{}B_{ij}(X)ϵ^{ab}_aX^i_bX^j2i\theta ^T\gamma ^0\rho ^{\widehat{0}}\gamma ^{}\rho ^a\widehat{D}_a\theta `$ (58)
$`+{\displaystyle \frac{1}{64}}\widehat{R}_{ijkl}\theta ^T\gamma ^0\rho ^{\widehat{0}}\gamma ^{ij}\rho ^a(1+\rho _3)\theta \theta ^T\gamma ^0\rho ^{\widehat{0}}\gamma ^{kl}\rho ^a(1\rho _3)\theta +\mathrm{},`$
where we have the following relations,
$`\widehat{R}^i_{jkl}`$ $`=`$ $`_k\widehat{\mathrm{\Gamma }}_{jl}^i\mathrm{},\widehat{\mathrm{\Gamma }}_{jl}^i=\mathrm{\Gamma }_{jl}^i[g]+2\pi \alpha ^{}H^i{}_{jl}{}^{},H_{ijk}=3_{[i}B_{jk]},`$
$`\widehat{D}_a`$ $`=`$ $`_a{\displaystyle \frac{1}{4}}\gamma ^{\widehat{j}\widehat{k}}\widehat{\omega }_i^{\widehat{j}\widehat{k}}_aX^i,\widehat{\omega }_{..i}=\omega _{..i}+2\pi \alpha ^{}\rho _3H_{..i},\gamma ^{ij}\gamma ^{ij}\gamma ^{},`$ (59)
and moreover: $`i,j,k,\mathrm{}`$, are transverse spacetime indices; $`a,b,\mathrm{}`$ are world–sheet indices; hatted indices correspond to tangent frame indices; $`\sigma _{ab}`$ is the world–sheet metric; and we have introduced two–dimensional world–sheet Dirac matrices $`\rho _{\widehat{a}}`$ as,
$$\rho ^{\widehat{0}}=\left(\begin{array}{cc}0& i\\ i& 0\end{array}\right),\rho ^{\widehat{1}}=\left(\begin{array}{cc}0& i\\ i& 0\end{array}\right),\rho _3=\left(\begin{array}{cc}1& 0\\ 0& 1\end{array}\right).$$
(60)
One can also write down the Lagrangian for the R–R sector of the Type IIA superstring action ,
$`_{RR}`$ $`=`$ $`{\displaystyle \frac{i}{(\alpha ^{})^{3/2}}}\theta ^T\gamma ^0\rho ^{\widehat{0}}\gamma ^{(i}\mathrm{\Gamma }^\mathrm{\Lambda }\gamma ^{j)}(1\rho _3)\theta (^aX^i_aX^j+{\displaystyle \frac{i}{6}}\theta ^T\gamma ^0\rho ^{\widehat{0}}\gamma ^{ni}\rho ^a\theta _aX^j_n`$ (61)
$`{\displaystyle \frac{1}{144}}\theta ^T\gamma ^0\rho ^{\widehat{0}}\gamma ^{ni}\rho ^a\theta \theta ^T\gamma ^0\rho ^{\widehat{0}}\gamma ^{mj}\rho _a\theta _n_m)C_\mathrm{\Lambda }(X)`$
$`+{\displaystyle \frac{i}{(\alpha ^{})^{3/2}}}\theta ^T\gamma ^0\rho ^{\widehat{0}}\gamma ^{[i}\mathrm{\Gamma }^\mathrm{\Lambda }\gamma ^{j]}(1\rho _3)\theta (ϵ^{ab}_aX^i_bX^j+{\displaystyle \frac{i}{6}}\theta ^T\gamma ^0\rho ^{\widehat{0}}\gamma ^{ni}\rho ^a\rho _3\theta _aX^j_n`$
$`{\displaystyle \frac{1}{144}}\theta ^T\gamma ^0\rho ^{\widehat{0}}\gamma ^{ni}\rho ^a\theta \theta ^T\gamma ^0\rho ^{\widehat{0}}\gamma ^{mj}\rho _a\rho _3\theta _n_m)C_\mathrm{\Lambda }(X)+\mathrm{},`$
where we have $`\{(\mathrm{\Gamma }^\mathrm{\Lambda },C_\mathrm{\Lambda })\}=\{(\gamma ^i,C_i),(\gamma ^{ijk},C_{ijk})\}`$ for the background $`D0`$ and $`D2`$–brane currents. Indices $`i,j,\mathrm{}`$, are contracted via the metric $`g_{ij}`$, and the dots in (61) refer to higher–derivative terms (which have no contribution to the light–cone vertex operators) . This completes the information on the Green–Schwarz action that we shall be interested in.
To begin the comparison with the abelian limit of our results from section 4, we look at the NS–NS sector. In the weak background field limit we are considering, the Riemann curvature terms drops out from (58) and all we need to check is for the existence of the abelian couplings,
$$h_{ij}(X)^aX^i_aX^j,$$
(62)
and
$$B_{ij}(X)ϵ^{ab}_aX^i_bX^j.$$
(63)
Comparing back to (4.2) and (4.2), one immediately observes that these terms indeed appear in the abelian conformal limit of our matrix string action. The term involving the Riemann tensor is present in a weak field approximation only via its piece in $`^2h`$. One can realize however that, being schematically of the form $`\widehat{R}_{ijkl}\theta ^T\mathrm{\Gamma }^{ij}\theta \theta ^T\mathrm{\Gamma }^{kl}\theta `$, it is of order $`𝒪=(\frac{R}{\mathrm{}_s})^2`$. At the Matrix theory level we have four fermion terms of this type in the $`T^i`$ component of the Matrix stress tensor and in the $`J^{ij}`$ component of the Matrix membrane current, and therefore we have terms of this type in several components of the matrix string tensor couplings. However, none of these terms seems to have the required index structure to couple to the Riemann curvature term, and besides we are looking for tensors that will couple to a term of the type $`_i_jh_{kl}`$. The reason for this is that such couplings will actually arise from higher moment terms. Indeed, one can see from the appendix that there are two fermion contributions to the first moment terms of $`T_{\mathrm{Fermion}}^{ij(l)}`$, which will couple to a term in $`h`$. Similarly there will be four fermion contributions that should produce the required curvature coupling. It would be interesting to construct explicitly such terms.
When we move to the R–R sector, we observe that in (61) the terms that do not involve derivatives of the $`D0`$ and $`D2`$–brane fields are at order $`𝒪=(\frac{R}{\mathrm{}_s})^2`$, and schematically of the form $`\theta ^T\mathrm{\Gamma }_{ij\mathrm{\Lambda }}\theta X^iX^j`$. For the $`D0`$–brane such terms will likely come from the quadratic fermion pieces that we neglected in the tensor $`T^{}`$, while for the $`D2`$–brane case it is not entirely clear where these terms should come form. The other terms in the R–R action (61) are higher derivative terms in the R–R fields, and so we would only expect to match these terms to higher moments of our couplings. Therefore, the overall comparison of our results with the ones for the one string action is rather schematic and qualitative. But the comparison can still be of some use in predicting some possible new coupling terms coming up in the full curved action.
### 6.2 Matrix Theory in Curved Backgrounds
As we mentioned before, in the Green–Schwarz sigma model there are four fermion couplings to the background Riemann tensor. One could expect that this coupling would correspond to the free string limit of some nonabelian coupling between world–sheet fields and the background curvature. Indeed, given the previous match between abelian and nonabelian actions, one has some clues for the form of the couplings to background curvature. This would amount to terms in the full non–linear matrix string action, and therefore to terms involving the Riemann curvature and other non–linear combinations of the supergravity background fields in the Matrix theory action.
As we have just seen, the term involving the curvature tensor is schematically of the form $`\widehat{R}_{ijkl}\theta ^T\mathrm{\Gamma }^{ij}\theta \theta ^T\mathrm{\Gamma }^{kl}\theta `$, and we have no tensor coupling with this index structure among the zeroeth moment terms. One naturally expects that such tensors would start making their appearance once one performs a higher loop calculation in Matrix theory or in matrix string theory as we would obtain, e.g., quadratic pieces in the metric, $`h`$. For the moment we simply observe that the actions (58) and (61) yield already some information on what one will obtain from such a higher loop calculation by telling us what will be the abelian limit of the tensors one would eventually obtain. Indeed we can predict that there will be a tensor coupling to the quadratic metric piece, of the type,
$$G^{ijkl}=\theta ^T\mathrm{\Gamma }^{ij}\theta \theta ^T\mathrm{\Gamma }^{kl}\theta +\mathrm{Nonabelian}\mathrm{Terms}.$$
(64)
This is of course the required coupling for the curved matrix string theory action to match, in the free abelian limit, the Green–Schwarz action. But because we are dealing in this section with background fields that are non–trivial only in the eight transverse directions, this term actually is lifted to a term in the curved Matrix theory action. We see therefore that in order for a supersymmetric completion of the curved background Matrix action to be done, there will be at least quartic fermion terms required, at zeroeth moment terms.
Because the curvature term in (58) also includes a coupling to the NS–NS $`B`$–field field strength, a similar story will also work for that term. Likewise, due to the higher derivative terms in (61), we shall actually obtain quite a few predictions for higher moments and other tensor couplings, along the lines of the previous discussion for the background Riemann curvature. In summary, a full extension of Matrix theory for the case of curved backgrounds will probably not be as simple as the sigma model proposal in . Its supersymmetric completion however, will have to include a series of non–linear background couplings, as we have just discussed. Such types of couplings, involving the Riemann tensor and four fermion fields, are common in supersymmetric completions of bosonic sigma models and so are quite natural to be expected in here as well.
## 7 Noncommutative Backgrounds
In this section we wish to exemplify the nonabelian nature of our action. Ideally one hopes that coherent states of gravitons can be made out of many fundamental strings (by some sort of fundamental string condensation), and by using infinite dimensional matrices to describe such solutions one could then be able to build fully curved spacetime geometries – somewhat like when one uses infinite matrices to describe non–compact curved membranes in Matrix theory . In practice the situation is not as idealistic, as it is not clear how to go from the description of the coherent state in terms of the strings to their effects on the other strings. This would correspond to a higher loop calculation in Matrix theory or matrix string theory.
Still, an interesting question is whether one can exponentiate the noncommutative vertex operators we have obtained in our linear action, and from that build the full non–linear matrix sigma model. Recall that the results for the $`I`$ tensors can be used (loosely speaking, via multiplication by $`\mathrm{exp}(ipX)`$) to obtain the noncommutative vertex operators of matrix string theory. However, precisely because of this noncommutative nature of the vertex operators, one still lacks an ordering for the exponentiation. Such an ordering would moreover have to produce terms with derivatives of the background fields, as such terms are expected in order to satisfy the geodesic length condition in . It is certainly not clear how to choose the ordering of such an exponential at this stage. There is also the question of whether the background satisfies the equations of motion of supergravity. To clear this issue, one would again need the full matrix sigma model in order to compute noncommutative beta functions and from there derive the noncommutative equations of motion for the background.
For the moment we will aim lower and consider a very simple example involving non–trivial R–R flux. There is a particular interest in examples involving R–R flux, due to its possible connection to noncommutative spacetime geometry. Recently there has been some study in the applications of noncommutative geometry to string theory. This has, however, been mainly studied at the world–volume level where the noncommutativity appears as a result of non–trivial NS–NS flux . But it has also been suggested that in situations involving R–R flux rather than NS–NS flux, the noncommutativity could make its appearance at the background spacetime level due to small distance stringy effects . It would be quite interesting to further study this issue.
### 7.1 R–R Flux and String Condensation
An interesting situation is the one where there exists non–trivial R–R flux. In recent work it was studied an example where a collection of $`D0`$–branes was polarized into a noncommutative 2–sphere configuration by an external R–R field. The question quickly arises of whether a similar situation could exist in this case, where we are dealing with a collection of fundamental strings. If they can indeed be polarized into noncommutative configurations by some external R–R fields, this would then correspond to the creation of some sort of noncommutative stringy object. We shall see that such indeed happens, and so R–R flux can act as a source for fundamental string world–sheet noncommutativity.
Let us consider a situation where there is non–trivial R–R 3–form flux. This case will be quite similar to the one of dielectric branes considered in , the difference being that now we have “dielectric strings”. We will moreover consider a simplified case where we take all fermionic fields to vanish, $`\theta =0`$. As an ansatz, let us also consider $`A=0`$ and $`X^i=0`$. At the background level, we shall set all other fields (except for the membrane current) to zero. All this done, one is left with the flat space matrix string theory action,
$$S_{Flat}=\frac{1}{2\pi }𝑑\sigma 𝑑\tau \mathrm{𝐓𝐫}\left(\frac{1}{2}(\dot{X}^i)^2+\frac{1}{4g_s^2}[X^i,X^j]^2\right),$$
(65)
supplemented by the $`D2`$–brane linear coupling,
$$S_{D2brane}=\frac{1}{2\pi }𝑑\sigma 𝑑\tau C_{\mu \nu \lambda }(X)I_2^{\mu \nu \lambda }.$$
(66)
As to the $`D2`$–brane linear coupling, we shall focus our attention in the lowest order terms in the derivative expansion. In particular, we will not retain terms at order $`𝒪=(\frac{R}{\mathrm{}_s})`$ and above (this corresponds to two operator insertions and is the same order as the flat space matrix string action). In this case, the tensor couplings we need to consider are:
$`I_2^{0ij}`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}\left({\displaystyle \frac{i}{6g_s}}\left({\displaystyle \frac{\mathrm{}_s}{R}}\right)[X^i,X^j]+𝒪({\displaystyle \frac{R}{\mathrm{}_s}})\right),`$
$`I_2^{ijk}`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}\left({\displaystyle \frac{i}{6g_s}}\dot{X}^i[X^j,X^k]{\displaystyle \frac{i}{6g_s}}\dot{X}^j[X^k,X^i]{\displaystyle \frac{i}{6g_s}}\dot{X}^k[X^i,X^j]\right),`$ (67)
so that the $`D2`$–brane linear coupling term becomes,
$`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau C_{\mu \nu \lambda }(X)I_2^{\mu \nu \lambda }}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \left(3C_{0ij}(X)I_2^{0ij}+C_{ijk}(X)I_2^{ijk}\right)}`$ (68)
$`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{i}{2g_s}}\left({\displaystyle \frac{\mathrm{}_s}{R}}\right)C_{0ij}(X)[X^i,X^j]`$
$`{\displaystyle \frac{i}{2g_s}}C_{ijk}(X)\dot{X}^i[X^j,X^k]+𝒪({\displaystyle \frac{R}{\mathrm{}_s}})),`$
where we still have to expand $`C_{0ij}`$ to first order in derivatives, $`C_{0ij}^{D2}(X)=C_{0ij}^{D2}(0)+(\frac{R}{\mathrm{}_s})X^k_kC_{0ij}^{D2}(0)+\mathrm{}`$. Doing this and integrating the time derivative by parts, one obtains,
$$\frac{1}{2\pi }𝑑\sigma 𝑑\tau C_{\mu \nu \lambda }(X)I_2^{\mu \nu \lambda }=\frac{1}{2\pi }𝑑\sigma 𝑑\tau \frac{i}{2g_s}\mathrm{𝐒𝐓𝐫}\left((_0C_{ijk}_kC_{0ij})X^k[X^i,X^j]\right)+\mathrm{},$$
(69)
where we have further specialized for the particular case of time independent solutions, $`\dot{X}^i=0`$. Any solutions we shall obtain will correspond to static backgrounds. Now, the $`D2`$–brane 3–form potential, $`𝐂_3`$, is related to the $`D2`$–brane 4-form field strength, $`𝐅_4`$, by the standard relation $`𝐅_4=d𝐂_3`$, and so one simply has the expected gauge invariant coupling:
$$\frac{1}{2\pi }𝑑\sigma 𝑑\tau C_{\mu \nu \lambda }(X)I_2^{\mu \nu \lambda }=\frac{1}{2\pi }𝑑\sigma 𝑑\tau \frac{i}{6g_s}\mathrm{𝐒𝐓𝐫}\left(F_{0ijk}X^k[X^i,X^j]\right)+\mathrm{}.$$
(70)
If we furthermore choose the $`D2`$–brane field strength as constant, $`F_{0ijk}=2Fϵ_{ijk}`$ for $`i,j,k=1,2,3`$, and zero otherwise, we are led into a situation very similar to the $`D0`$–brane case of , only now with $`F`$–strings. Indeed, the effective potential we have obtained for the static configuration of $`N`$ fundamental strings is,
$$V_{eff}(X)=\frac{1}{4g_s^2}\mathrm{𝐓𝐫}[X^i,X^j]^2\frac{i}{6g_s}F_{0ijk}\mathrm{𝐒𝐓𝐫}[X^i,X^j]X^k.$$
(71)
with the following equation for the extrema,
$$[[X^i,X^j],X^j]\frac{i}{2}g_sF_{0ijk}[X^j,X^k]=0.$$
(72)
The case of commuting matrices, $`[X^i,X^j]=0`$, is a solution with zero potential energy. This corresponds to the free string limit, i.e., it corresponds to a situation describing separated, straight and static free strings. More interesting to us would be a noncommuting solution. Following , we consider the ansatz $`X^i=\phi \sigma ^i`$, $`i=1,2,3`$, where $`\phi `$ is a constant and $`\sigma ^i`$ are some $`N`$–dimensional matrix representation of the $`su(2)`$ algebra,
$$[\sigma ^i,\sigma ^j]=2iϵ^{ijk}\sigma ^k.$$
(73)
Using this ansatz in the equation for the extrema of the string effective potential, one immediately obtains that this is indeed a solution once we set the constant to be $`\phi =\frac{1}{2}g_sF`$. If we moreover consider the $`\sigma ^i`$ as an irreducible representation, one computes the Casimir as $`\mathrm{𝐓𝐫}(\sigma ^i)^2=\frac{1}{3}N(N^21)`$, and so the nonabelian solution has an effective potential of,
$$V_{NC}=\frac{g_s^2F^4}{48}N(N^21),$$
(74)
i.e., the noncommutative solution has an energy lower than the commuting one. This means of course that the configuration corresponding to separated static free strings is unstable, and the strings will actually condense into a noncommutative solution. This solution is the noncommutative sphere, with radius $`R=\frac{1}{2}g_sFN(1\frac{1}{N^2})^{\frac{1}{2}}`$, as can be seen from the algebra (73). So, in conclusion, the presence of an R–R field has condensed the initially free fundamental strings into a static noncommutative spherical configuration. Also, we should point out that this solution corresponds to a string theory derivation of the commutation relations for fundamental strings in the presence of R–R fields, proposed in .
This phenomena leads to an interesting question. In the $`D`$–brane situation, the Myers’ effect tells us that an external R–R field can polarize a collection of $`D`$–branes into a noncommutative configuration. The noncommutativity is present at the world–volume level, such that there is a background commutative spacetime with a noncommutative object made out of many $`D`$–branes inside. In the situation described in this section we are dealing with $`F`$–strings. So, even though the situation seems quite analogous to the one of the brane system, one also needs to take into account the fact that the $`F`$–strings actually describe gravitons, and we are thus led to a situation where the R–R field is building a noncommutative object made out of many gravitons. Now, graviton states generally correspond to curved spacetimes, however describing small fluctuations such as gravitational waves with small amplitudes. If one wishes to describe large fluctuations one needs to consider states with a semiclassical behavior which would correspond to coherent states – where one would run into the mentioned problems concerning exponentiation of nonabelian couplings.
The question still remains on how to interpret a noncommutative object made out of many gravitons, inside a background commutative spacetime. One speculative possibility is that this could actually correspond to a noncommutative background spacetime geometry. Indeed, one could imagine that if we put enough gravitons together, the noncommutative spherical configuration will grow up to a stage where its curvature is actually weak. Then, we would be in a position to expect that this large sphere would take up the role of the background spacetime, i.e., the R–R flux we considered would have created some sort of noncommutative background spacetime geometry. It would be quite interesting to further study more complex examples of such situations.
## 8 Conclusions
We have seen in this paper how to construct an action for matrix string theory in weakly curved background fields. In the process, we have also studied its relation to $`T`$ and $`S`$ dualities. Such an action provides working ground to study multiple interacting strings in both NS–NS and R–R backgrounds. For the particular case of an R–R background, we have seen how fundamental strings can condense into a nonabelian configuration thus building a noncommutative stringy configuration. The action we derived also allows for some discussion on how background non–linear curvature terms will couple to the Matrix theory action in a general curved background. With all this in hand, we believe there are quite a few interesting lines for future research on this subject. We present some of these lines below.
One possible application of the action we have obtained is to describe second quantized superstring theory in general backgrounds with R–R fields turned on. This is quite an interesting venue of work, given that there has been some recent interest in such ideas, e.g. . Along these lines, one should also try to further understand the possible background geometry noncommutativity due to the R–R flux, in particular it would be quite interesting if some connection to the work in could be done. For this, it could be of some interest to complete further examples involving diverse background fields, in order to fully explore the nonabelian aspects of curved stringy spacetimes. This could be a first step towards the more ambitious goal of constructing fully curved noncommutative spacetime geometries from string theory.
The resulting action from our work could also be of use in the study of diverse scattering or absorption problems, involving branes, black holes, or in the context of the AdS/CFT correspondence . It would be quite interesting if some work along the lines in could be accomplished. Probably, from the computation of scattering amplitudes, the role of the noncommutative vertex operators would then become more translucid (there has been some very recent work on vertex operators in ). This would then be of some interest should it yield further insight on how one should exponentiate such operators, in order to obtain the full non–linear action. Indeed, one particularly interesting use of our work would be to further use this action in order to try to infer some new information about Matrix theory in a general curved background. This could perhaps be accomplished via a more detailed and direct comparison with the Type IIA string theory abelian limit. Some work towards such goal has been done in , and one should try to understand any possible relations between those papers and the work presented in here. Probably one particularly important relation to understand is the one with the work in . In order to achieve these goals, some research should be done on understanding how to construct the precise lift of the matrix string theory action from the cylinder to arbitrary Riemann surfaces, and so establish the precise connection to the Green–Schwarz action along the lines of . We hope to address some of these questions in the future.
Acknowledgments: I would like to thank Washington Taylor for many discussions and critical comments as well as suggestions on the draft of the paper. I would also like to thank Lorenzo Cornalba, José Mourão and João Nunes for helpful discussions and/or comments. The author is supported in part by the Fundação para a Ciência e Tecnologia, under the grant Praxis XXI BPD-17225/98 (Portugal).
## Appendix A Supercurrents from Matrix Theory
We reproduce here the Matrix theory forms of the multipole moments of the 11–dimensional supercurrents found in , written in 10–dimensional form as in . Dropping a factor of $`1/R`$ from each expression, the stress tensor $`T^{IJ}`$, $`M2`$–brane current $`J^{IJK}`$ and $`M5`$–brane current $`M^{IJKLMN}`$ have integrated (monopole) components:
$`T^{++}`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}(\text{1}\text{ }\text{1})=N`$
$`T^{+i}`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}(F^{0i})`$
$`T^+`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}(F^{0\mu }F^0{}_{\mu }{}^{}+{\displaystyle \frac{1}{4}}F_{\mu \nu }F^{\mu \nu }+{\displaystyle \frac{i}{2}}\overline{\mathrm{\Theta }}\mathrm{\Gamma }^0D_0\mathrm{\Theta })`$
$`T^{ij}`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}(F^{i\mu }F_\mu {}_{}{}^{j}+{\displaystyle \frac{i}{4}}\overline{\mathrm{\Theta }}\mathrm{\Gamma }^iD_j\mathrm{\Theta }+{\displaystyle \frac{i}{4}}\overline{\mathrm{\Theta }}\mathrm{\Gamma }^jD_i\mathrm{\Theta })`$
$`T^i`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}(F^{0\mu }F_{\mu \nu }F^{\nu i}+{\displaystyle \frac{1}{4}}F^{0i}F_{\mu \nu }F^{\mu \nu }{\displaystyle \frac{i}{8}}F_{\mu \nu }\overline{\mathrm{\Theta }}\mathrm{\Gamma }^i\mathrm{\Gamma }^{\mu \nu }D_0\mathrm{\Theta }`$
$`+{\displaystyle \frac{i}{8}}F_{\mu \nu }\overline{\mathrm{\Theta }}\mathrm{\Gamma }^0\mathrm{\Gamma }^{\mu \nu }D_i\mathrm{\Theta }{\displaystyle \frac{i}{4}}F_{\mu \nu }\overline{\mathrm{\Theta }}\mathrm{\Gamma }^\nu \mathrm{\Gamma }^{0i}D^\mu \mathrm{\Theta }{\displaystyle \frac{1}{8}}\overline{\mathrm{\Theta }}\mathrm{\Gamma }^{0\mu i}\mathrm{\Theta }\overline{\mathrm{\Theta }}\mathrm{\Gamma }_\mu \mathrm{\Theta })`$
$`T^{}`$ $`=`$ $`{\displaystyle \frac{1}{4}}\mathrm{𝐒𝐓𝐫}(F_{\mu \nu }F^{\nu \gamma }F_{\gamma \delta }F^{\delta \mu }{\displaystyle \frac{1}{4}}F_{\mu \nu }F^{\mu \nu }F_{\gamma \delta }F^{\gamma \delta }+iF_{\mu \nu }F_{\gamma \delta }\overline{\mathrm{\Theta }}\mathrm{\Gamma }^\nu \mathrm{\Gamma }^{\gamma \delta }D^\mu \mathrm{\Theta }+𝒪(\mathrm{\Theta }^4))`$
$`J^{+ij}`$ $`=`$ $`{\displaystyle \frac{1}{6}}\mathrm{𝐒𝐓𝐫}(F^{ij})`$
$`J^{+i}`$ $`=`$ $`{\displaystyle \frac{1}{6}}\mathrm{𝐒𝐓𝐫}(F^{0\mu }F_\mu {}_{}{}^{i}+{\displaystyle \frac{i}{4}}\overline{\mathrm{\Theta }}\mathrm{\Gamma }^0D_i\mathrm{\Theta }{\displaystyle \frac{i}{4}}\overline{\mathrm{\Theta }}\mathrm{\Gamma }^iD_0\mathrm{\Theta })`$
$`J^{ijk}`$ $`=`$ $`{\displaystyle \frac{1}{6}}\mathrm{𝐒𝐓𝐫}(F^{0i}F^{jk}+F^{0j}F^{ki}+F^{0k}F^{ij}{\displaystyle \frac{3i}{4}}\overline{\mathrm{\Theta }}\mathrm{\Gamma }^{0[ij}D^{k]}\mathrm{\Theta }+{\displaystyle \frac{i}{4}}\overline{\mathrm{\Theta }}\mathrm{\Gamma }^{ijk}D_0\mathrm{\Theta })`$
$`J^{ij}`$ $`=`$ $`{\displaystyle \frac{1}{6}}\mathrm{𝐒𝐓𝐫}(F^{i\mu }F_{\mu \nu }F^{\nu j}+{\displaystyle \frac{1}{4}}F^{ij}F_{\mu \nu }F^{\mu \nu }{\displaystyle \frac{i}{8}}F_{\mu \nu }\overline{\mathrm{\Theta }}\mathrm{\Gamma }^j\mathrm{\Gamma }^{\mu \nu }D_i\mathrm{\Theta }`$
$`+{\displaystyle \frac{i}{8}}F_{\mu \nu }\overline{\mathrm{\Theta }}\mathrm{\Gamma }^i\mathrm{\Gamma }^{\mu \nu }D_j\mathrm{\Theta }{\displaystyle \frac{i}{4}}F_{\mu \nu }\overline{\mathrm{\Theta }}\mathrm{\Gamma }^\nu \mathrm{\Gamma }^{ij}D^\mu \mathrm{\Theta }+{\displaystyle \frac{1}{8}}\overline{\mathrm{\Theta }}\mathrm{\Gamma }^{\mu ij}\mathrm{\Theta }\overline{\mathrm{\Theta }}\mathrm{\Gamma }_\mu \mathrm{\Theta })`$
$`M^{+ijkl}`$ $`=`$ $`{\displaystyle \frac{1}{12}}\mathrm{𝐒𝐓𝐫}(F^{ij}F^{kl}+F^{ik}F^{lj}+F^{il}F^{jk}i\overline{\mathrm{\Theta }}\mathrm{\Gamma }^{[ijk}D^{l]}\mathrm{\Theta })`$
$`M^{ijklm}`$ $`=`$ $`{\displaystyle \frac{5}{4}}\mathrm{𝐒𝐓𝐫}(F^{0[i}F^{jk}F^{lm]}+{\displaystyle \frac{i}{2}}F^{[0i}\overline{\mathrm{\Theta }}\mathrm{\Gamma }^{jkl}D^{m]}\mathrm{\Theta }).`$
Here, $`\mathrm{𝐒𝐓𝐫}`$ denotes a symmetrized trace in which one takes the average over all possible orderings of the matrices inside the trace, with commutators being treated as a unit block. Time derivatives are taken with respect to Minkowski time $`t`$. Indices $`i,j,\mathrm{}`$, run from 1 through 9, while indices $`a,b,\mathrm{}`$, run from 0 through 9. In these expressions one should use the definitions $`F_{0i}=\dot{X}^i`$ and $`F_{ij}=i[X^i,X^j]`$. A Matrix form for the transverse 5–brane current components $`M^{+ijklm}`$ and $`M^{ijklmn}`$ is not known, and in fact comparison with supergravity suggests that these should be zero for any Matrix theory configuration.
The higher multipole moments of these currents contain one set of terms which are found by including the matrices $`X^{k_1},\mathrm{},X^{k_n}`$ into the symmetrized trace as well as more complicated spin contributions. We may write these as,
$`T^{IJ(i_1\mathrm{}i_k)}`$ $`=`$ $`\mathrm{Sym}(T^{IJ};X^{i_1},\mathrm{},X^{i_k})+T_{\mathrm{Fermion}}^{IJ(i_1\mathrm{}i_k)}`$
$`J^{IJK(i_1\mathrm{}i_k)}`$ $`=`$ $`\mathrm{Sym}(J^{IJK};X^{i_1},\mathrm{},X^{i_k})+J_{\mathrm{Fermion}}^{IJK(i_1\mathrm{}i_k)}`$
$`M^{IJKLMN(i_1\mathrm{}i_k)}`$ $`=`$ $`\mathrm{Sym}(M^{IJKLMN};X^{i_1},\mathrm{},X^{i_k})+M_{\mathrm{Fermion}}^{IJKLMN(i_1\mathrm{}i_k)},`$
where some simple examples of the two–fermion contribution to the first moment terms are,
$`T_{\mathrm{Fermion}}^{+i(j)}`$ $`=`$ $`{\displaystyle \frac{1}{8R}}\mathrm{𝐓𝐫}(\overline{\mathrm{\Theta }}\mathrm{\Gamma }^{[0ij]}\mathrm{\Theta })`$
$`T_{\mathrm{Fermion}}^{+(i)}`$ $`=`$ $`{\displaystyle \frac{1}{16R}}\mathrm{𝐓𝐫}(F_{\mu \nu }\overline{\mathrm{\Theta }}\gamma ^{[\mu \nu i]}\mathrm{\Theta }4\overline{\mathrm{\Theta }}F_{0\mu }\gamma ^{[0\mu i]}\mathrm{\Theta })`$
$`T_{\mathrm{Fermion}}^{ij(l)}`$ $`=`$ $`{\displaystyle \frac{1}{8R}}\mathrm{𝐓𝐫}(F_{j\mu }\overline{\mathrm{\Theta }}\gamma ^{[\mu il]}\mathrm{\Theta }+\overline{\mathrm{\Theta }}F_{i\mu }\gamma ^{[\mu jl]}\mathrm{\Theta })`$
$`J_{\mathrm{Fermion}}^{+ij(k)}`$ $`=`$ $`{\displaystyle \frac{i}{48R}}\mathrm{𝐓𝐫}(\overline{\mathrm{\Theta }}\mathrm{\Gamma }^{[ijk]}\mathrm{\Theta })`$
$`J_{\mathrm{Fermion}}^{+i(j)}`$ $`=`$ $`{\displaystyle \frac{1}{48R}}\mathrm{𝐓𝐫}(F_{0\mu }\overline{\mathrm{\Theta }}\gamma ^{[\mu ij]}\mathrm{\Theta }+\overline{\mathrm{\Theta }}F_{i\mu }\gamma ^{[\mu 0j]}\mathrm{\Theta })`$
$`M_{\mathrm{Fermion}}^{+ijkl(m)}`$ $`=`$ $`{\displaystyle \frac{i}{16R}}\mathrm{𝐒𝐓𝐫}\left(\overline{\mathrm{\Theta }}F^{[jk}\mathrm{\Gamma }^{il]m}\mathrm{\Theta }\right).`$
The remaining two–fermion contributions to the first moments and some four–fermion terms are also determined by the results in . There are also fermionic components of the supercurrent which couple to background fermion fields in the supergravity theory. These couplings have not been discussed in this paper, but the Matrix theory form of the currents is determined in .
Finally, there is also a 6–brane current appearing in Matrix theory related to nontrivial 11–dimensional background metrics. The components of this current as well as its first moments are:
$`S^{+ijklmn}`$ $`=`$ $`{\displaystyle \frac{1}{R}}\mathrm{𝐒𝐓𝐫}\left(F_{[ij}F_{kl}F_{mn]}\right)`$
$`S^{+ijklmn(p)}`$ $`=`$ $`{\displaystyle \frac{1}{R}}\mathrm{𝐒𝐓𝐫}\left(F_{[ij}F_{kl}F_{mn]}X_p\theta F_{[kl}F_{mn}\gamma _{pqr]}\theta \right)`$
$`S^{ijklmnp}`$ $`=`$ $`{\displaystyle \frac{7}{R}}\mathrm{𝐒𝐓𝐫}\left(F_{[ij}F_{kl}F_{mn}\dot{X}_{p]}+𝒪(\theta ^2,\theta ^4)\right)`$
$`S^{ijklmnp(q)}`$ $`=`$ $`{\displaystyle \frac{7}{R}}\mathrm{𝐒𝐓𝐫}\left(F_{[ij}F_{kl}F_{mn}\dot{X}_{p]}X_q\theta \dot{X}_{[j}F_{kl}F_{mn}\gamma _{pqr]}\theta +{\displaystyle \frac{i}{2}}\theta F_{[jk}F_{lm}F_{np}\gamma _{qr]}\theta \right).`$
## Appendix B DVV Reduction of Matrix Theory Tensors
The Matrix stress tensor components are:
$`T^{++}`$ $`=`$ $`{\displaystyle \frac{1}{R}}\mathrm{𝐒𝐓𝐫}\left(\text{1}\text{ }\text{1}\right),`$
$`T^{+i}`$ $`=`$ $`{\displaystyle \frac{1}{R}}\mathrm{𝐒𝐓𝐫}\left(\dot{X_i}\right),`$
$`T^+`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}\left({\displaystyle \frac{1}{2R}}\dot{X_i}\dot{X_i}{\displaystyle \frac{RM_P^6}{8\pi ^2}}{\displaystyle \underset{i<j}{}}[X^i,X^j]^2+{\displaystyle \frac{RM_P^6}{8\pi ^2}}\theta \gamma ^i[X^i,\theta ]\right),`$
$`T^{ij}`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}\left({\displaystyle \frac{1}{R}}\dot{X_i}\dot{X_j}{\displaystyle \frac{RM_P^6}{4\pi ^2}}[X^i,X^k][X^k,X^j]{\displaystyle \frac{RM_P^6}{16\pi ^2}}\theta \gamma ^i[X_j,\theta ]{\displaystyle \frac{RM_P^6}{16\pi ^2}}\theta \gamma ^j[X_i,\theta ]\right),`$
$`T^i`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}\left({\displaystyle \frac{1}{2R}}\dot{X_i}(\dot{X_j})^2{\displaystyle \frac{RM_P^6}{8\pi ^2}}\dot{X_i}{\displaystyle \underset{j<k}{}}[X^j,X^k]^2{\displaystyle \frac{RM_P^6}{4\pi ^2}}[X^i,X^j][X^j,X^k]\dot{X_k}\right)`$
$`\mathrm{𝐒𝐓𝐫}\left({\displaystyle \frac{RM_P^6}{16\pi ^2}}\theta _\alpha \dot{X_k}[X_m,\theta _\beta ]\right)\{\gamma ^k\delta _{im}+\gamma ^i\delta _{mk}2\gamma ^m\delta _{ki}\}_{\alpha \beta }`$
$`\mathrm{𝐒𝐓𝐫}\left({\displaystyle \frac{iR^2M_P^9}{64\pi ^3}}\theta _\alpha [X^k,X^l][X_m,\theta _\beta ]\right)\{\gamma ^{[iklm]}+2\gamma ^{[lm]}\delta _{ki}+4\delta _{ki}\delta _{lm}\}_{\alpha \beta }`$
$`+\mathrm{𝐓𝐫}\left({\displaystyle \frac{iR^2M_P^9}{64\pi ^3}}\theta \gamma ^{[ki]}\theta \theta \gamma ^k\theta \right),`$
$`T^{}`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}({\displaystyle \frac{1}{4R}}(\dot{X^i})^2(\dot{X^j})^2+{\displaystyle \frac{RM_P^6}{4\pi ^2}}\dot{X^i}\dot{X^j}[X^i,X^k][X^k,X^j]+{\displaystyle \frac{RM_P^6}{8\pi ^2}}(\dot{X^i})^2{\displaystyle \underset{j<k}{}}[X^j,X^k]^2`$ (75)
$`+{\displaystyle \frac{R^3M_P^{12}}{64\pi ^4}}[X^i,X^j][X^j,X^k][X^k,X^m][X^m,X^i]{\displaystyle \frac{R^3M_P^{12}}{64\pi ^4}}{\displaystyle \underset{i<j}{}}[X^i,X^j]^2{\displaystyle \underset{k<m}{}}[X^k,X^m]^2`$
$`+𝒪(\theta ^2)+𝒪(\theta ^4)).`$
To these components, one should now perform the $`T`$–duality for the $`911`$ flip, followed by the rescalings of world–sheet coordinates, background fields and coupling constants. The final result to obtain is the explicit form of the previous components of the stress tensor, this time in matrix string theory (with $`i,j9`$):
$`T^{++}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{\mathrm{}_s}{R}}\right)^2{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\text{1}\text{ }\text{1}\right)},`$
$`T^{+i}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{\mathrm{}_s}{R}}\right){\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\dot{X_i}\right)},`$
$`T^{+9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{\mathrm{}_s}{R}}\right){\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(g_s\dot{A}\right)},`$
$`T^+`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{1}{2}}\dot{X_i}^2+{\displaystyle \frac{1}{2}}(DX^i)^2+{\displaystyle \frac{1}{2}}g_s^2\dot{A}^2{\displaystyle \frac{1}{2g_s^2}}{\displaystyle \underset{i<j}{}}[X^i,X^j]^2`$
$`+{\displaystyle \frac{1}{g_s}}\theta \gamma _i[X^i,\theta ]+i\theta \gamma ^9D\theta ),`$
$`T^{ij}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}(\dot{X_i}\dot{X_j}DX^iDX^j{\displaystyle \frac{1}{g_s^2}}[X^i,X^k][X^k,X^j]`$
$`{\displaystyle \frac{1}{2g_s}}\theta \gamma ^i[X_j,\theta ]{\displaystyle \frac{1}{2g_s}}\theta \gamma ^j[X_i,\theta ]),`$
$`T^{i9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}(g_s\dot{X_i}\dot{A}+\frac{i}{g_s}[X^i,X^k]DX^k\frac{i}{2}\theta \gamma ^iD\theta \frac{1}{2g_s}\theta \gamma ^9[X_i,\theta ])},`$
$`T^{99}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(g_s^2\dot{A}^2(DX^k)^2i\theta \gamma ^9D\theta \right)},`$
$`T^i`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{1}{2}}\dot{X_i}(\dot{X_j})^2+{\displaystyle \frac{1}{2}}g_s^2\dot{X_i}\dot{A}^2{\displaystyle \frac{1}{2g_s^2}}\dot{X_i}{\displaystyle \underset{j<k}{}}[X^j,X^k]^2+{\displaystyle \frac{1}{2}}\dot{X_i}(DX^j)^2`$
$`{\displaystyle \frac{1}{g_2^2}}[X^i,X^j][X^j,X^k]\dot{X_k}DX^iDX^k\dot{X_k}+i[X^i,X^j]DX^j\dot{A}`$
$`{\displaystyle \frac{1}{2g_s}}\theta _\alpha \dot{X_k}[X_j,\theta _\beta ]\{\gamma ^k\delta _{ij}+\gamma ^i\delta _{jk}2\gamma ^j\delta _{ki}\}_{\alpha \beta }{\displaystyle \frac{1}{2}}\dot{A}\theta \gamma ^9[X_i,\theta ]+i\dot{X_i}\theta \gamma ^9D\theta `$
$`{\displaystyle \frac{i}{2}}g_s\dot{A}\theta \gamma ^iD\theta {\displaystyle \frac{i}{4g_s^2}}\theta _\alpha [X^k,X^j][X^l,\theta _\beta ]\{\gamma ^{[ikjl]}+2\gamma ^{[jl]}\delta _{ki}+4\delta _{ki}\delta _{jl}\}_{\alpha \beta }`$
$`{\displaystyle \frac{1}{2g_s}}\theta _\alpha DX^k[X^j,\theta _\beta ]\{\gamma ^{[ik9j]}+\gamma ^{[9j]}\delta _{ki}\}_{\alpha \beta }+{\displaystyle \frac{1}{4g_s}}\theta _\alpha [X^k,X^j]D\theta _\beta \{\gamma ^{[ikj9]}+2\gamma ^{[j9]}\delta _{ki}\}_{\alpha \beta }`$
$`iDX^i\theta D\theta +{\displaystyle \frac{i}{2}}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)(\theta \gamma ^{[ki]}\theta \theta \gamma ^k\theta +\theta \gamma ^{[9i]}\theta \theta \gamma ^9\theta )),`$
$`T^9`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{1}{2}}g_s\dot{A}(\dot{X^i})^2+{\displaystyle \frac{1}{2}}g_s^3\dot{A}^3{\displaystyle \frac{1}{2g_s}}\dot{A}{\displaystyle \underset{i<j}{}}[X^i,X^j]^2`$
$`{\displaystyle \frac{i}{g_s}}DX^i[X^i,X^j]\dot{X^j}{\displaystyle \frac{1}{2}}g_s(DX^i)^2\dot{A}{\displaystyle \frac{1}{2g_s}}\dot{X^i}\theta \gamma ^9[X^i,\theta ]+\dot{A}\theta \gamma ^i[X^i,\theta ]`$
$`{\displaystyle \frac{i}{2}}\dot{X^i}\theta \gamma ^iD\theta {\displaystyle \frac{i}{4g_s^2}}[X^i,X^j]\theta \gamma ^{[9ijk]}[X^k,\theta ]+{\displaystyle \frac{1}{2g_s}}\theta _\alpha DX^i[X^j,\theta _\beta ]\{\gamma ^{[ij]}+2\delta _{ij}\}_{\alpha \beta }`$
$`+{\displaystyle \frac{i}{2}}DX^i\theta \gamma ^{[i9]}D\theta +{\displaystyle \frac{i}{2}}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)\theta \gamma ^{[i9]}\theta \theta \gamma ^i\theta ),`$
$`T^{}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)^2{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{1}{4}}(\dot{X^i})^2(\dot{X^j})^2+{\displaystyle \frac{1}{2}}g_s^2\dot{A}^2(\dot{X^i})^2+{\displaystyle \frac{1}{4}}g_s^4\dot{A}^4`$ (76)
$`+{\displaystyle \frac{1}{g_s^2}}\dot{X^i}\dot{X^j}[X^i,X^k][X^k,X^j]2i\dot{X^i}\dot{A}[X^i,X^k]DX^k+{\displaystyle \frac{1}{2}}g_s^2\dot{A}^2(DX^i)^2`$
$`+\dot{X^i}\dot{X^j}DX^iDX^j+{\displaystyle \frac{1}{2g_s^2}}(\dot{X^i})^2{\displaystyle \underset{j<k}{}}[X^j,X^k]^2{\displaystyle \frac{1}{2}}(\dot{X^i})^2(DX^j)^2+{\displaystyle \frac{1}{2}}\dot{A}^2{\displaystyle \underset{i<j}{}}[X^i,X^j]^2`$
$`+{\displaystyle \frac{1}{4g_s^4}}[X^i,X^j][X^j,X^k][X^k,X^m][X^m,X^i]+{\displaystyle \frac{1}{g_s^2}}DX^iDX^j[X^i,X^k][X^k,X^j]`$
$`{\displaystyle \frac{1}{4g_s^4}}{\displaystyle \underset{i<j}{}}[X^i,X^j]^2{\displaystyle \underset{k<m}{}}[X^k,X^m]^2+{\displaystyle \frac{1}{2g_s^2}}(DX^i)^2{\displaystyle \underset{j<k}{}}[X^j,X^k]^2+{\displaystyle \frac{1}{4}}(DX^i)^2(DX^j)^2`$
$`+𝒪(\theta ^2)+𝒪(\theta ^4)).`$
Finally, the free string limit can be taken. The result for the conformal field theory limit of the matrix string stress tensor is:
$`\underset{g_s0}{lim}T^{++}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{\mathrm{}_s}{R}}\right)^2{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\text{1}\text{ }\text{1}\right)},`$
$`\underset{g_s0}{lim}T^{+i}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{\mathrm{}_s}{R}}\right){\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\dot{X_i}\right)},`$
$`\underset{g_s0}{lim}T^{+9}`$ $`=`$ $`0,`$
$`\underset{g_s0}{lim}T^+`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{1}{2}\dot{X_i}^2+\frac{1}{2}(X^i)^2+i\theta \gamma ^9\theta \right)},`$
$`\underset{g_s0}{lim}T^{ij}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\dot{X_i}\dot{X_j}X^iX^j\right)},`$
$`\underset{g_s0}{lim}T^{i9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{i}{2}\theta \gamma ^i\theta \right)},`$
$`\underset{g_s0}{lim}T^{99}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left((X^i)^2i\theta \gamma ^9\theta \right)},`$
$`\underset{g_s0}{lim}T^i`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{1}{2}}\dot{X_i}(\dot{X_j})^2+{\displaystyle \frac{1}{2}}\dot{X_i}(X^j)^2X^iX^k\dot{X_k}`$
$`+i\dot{X_i}\theta \gamma ^9\theta iX^i\theta \theta +𝒪\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)),`$
$`\underset{g_s0}{lim}T^9`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{i}{2}\dot{X^i}\theta \gamma ^i\theta +\frac{i}{2}X^i\theta \gamma ^{[i9]}\theta +𝒪\left(\frac{R}{\mathrm{}_s}\right)\right)},`$
$`\underset{g_s0}{lim}T^{}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)^2{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{1}{4}}(\dot{X^i})^2(\dot{X^j})^2+\dot{X^i}\dot{X^j}X^iX^j{\displaystyle \frac{1}{2}}(\dot{X^i})^2(X^j)^2`$ (77)
$`+{\displaystyle \frac{1}{4}}(X^i)^2(X^j)^2+𝒪(\theta ^2)+𝒪(\theta ^4)).`$
The next terms we look at are the zeroth moments of the components of the Matrix membrane current. These components are:
$`J^{+ij}`$ $`=`$ $`{\displaystyle \frac{iM_P^3}{12\pi }}\mathrm{𝐒𝐓𝐫}\left([X^i,X^j]\right),`$
$`J^{+i}`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}\left({\displaystyle \frac{iM_P^3}{12\pi }}[X^i,X^j]\dot{X^j}{\displaystyle \frac{RM_P^6}{48\pi ^2}}\theta [X^i,\theta ]+{\displaystyle \frac{RM_P^6}{96\pi ^2}}\theta \gamma ^{[ki]}[X^k,\theta ]\right),`$
$`J^{ijk}`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}\left({\displaystyle \frac{iM_P^3}{12\pi }}\dot{X^i}[X^j,X^k]+{\displaystyle \frac{iM_P^3}{12\pi }}\dot{X^j}[X^k,X^i]+{\displaystyle \frac{iM_P^3}{12\pi }}\dot{X^k}[X^i,X^j]{\displaystyle \frac{RM_P^6}{96\pi ^2}}\theta \gamma ^{[ijkl]}[X_l,\theta ]\right),`$
$`J^{ij}`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}({\displaystyle \frac{iM_P^3}{12\pi }}\dot{X^i}\dot{X^k}[X^k,X^j]{\displaystyle \frac{iM_P^3}{12\pi }}\dot{X^j}\dot{X^k}[X^k,X^i]{\displaystyle \frac{iM_P^3}{24\pi }}(\dot{X^k})^2[X^i,X^j]`$ (78)
$`{\displaystyle \frac{iR^2M_P^9}{96\pi ^3}}[X^i,X^j]{\displaystyle \underset{k<l}{}}[X^k,X^l]^2{\displaystyle \frac{iR^2M_P^9}{48\pi ^3}}[X^i,X^k][X^k,X^l][X^l,X^j])`$
$`+{\displaystyle \frac{RM_P^6}{96\pi ^2}}\mathrm{𝐒𝐓𝐫}\left(\theta _\alpha \dot{X^k}[X^m,\theta _\beta ]\right)\{\gamma ^{[kijm]}+\gamma ^{[jm]}\delta _{ki}\gamma ^{[im]}\delta _{kj}+2\delta _{jm}\delta _{ki}2\delta _{im}\delta _{kj}\}_{\alpha \beta }`$
$`+{\displaystyle \frac{iR^2M_P^9}{64\pi ^3}}\mathrm{𝐒𝐓𝐫}\left(\theta _\alpha [X^k,X^l][X^m,\theta _\beta ]\right)\{\gamma ^{[jkl]}\delta _{mi}\gamma ^{[ikl]}\delta _{mj}+2\gamma ^{[lij]}\delta _{km}+2\gamma ^l\delta _{jk}\delta _{im}`$
$`2\gamma ^l\delta _{ik}\delta _{jm}+2\gamma ^j\delta _{il}\delta _{km}2\gamma ^i\delta _{jl}\delta _{km}\}_{\alpha \beta }`$
$`+{\displaystyle \frac{iR^2M_P^9}{384\pi ^3}}\mathrm{𝐒𝐓𝐫}\left(\theta \gamma ^{[kij]}\theta \theta \gamma ^k\theta \theta \gamma ^{[ij]}\theta \theta \theta \right).`$
To these components we now perform the $`T`$–duality for the $`911`$ flip, followed by the rescalings of world–sheet coordinates, background fields and coupling constants. One obtains the explicit form of the previous components of the membrane current, in matrix string theory (with $`i,j,k9`$):
$`J^{+ij}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{\mathrm{}_s}{R}}\right){\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{i}{6g_s}[X^i,X^j]\right)},`$
$`J^{+i9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{\mathrm{}_s}{R}}\right){\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{1}{6}DX^i\right)},`$
$`J^{+i}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{i}{6g_s}}[X^i,X^j]\dot{X^j}+{\displaystyle \frac{1}{6}}g_s\dot{A}DX^i{\displaystyle \frac{1}{6g_s}}\theta [X^i,\theta ]`$
$`+{\displaystyle \frac{1}{12g_s}}\theta \gamma ^{[ki]}[X^k,\theta ]+{\displaystyle \frac{i}{12}}\theta \gamma ^{[9i]}D\theta ),`$
$`J^{+9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{1}{6}\dot{X^i}DX^i\frac{i}{6}\theta D\theta +\frac{1}{12g_s}\theta \gamma ^{[i9]}[X^i,\theta ]\right)},`$
$`J^{ijk}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{i}{6g_s}}\dot{X^i}[X^j,X^k]{\displaystyle \frac{i}{6g_s}}\dot{X^j}[X^k,X^i]{\displaystyle \frac{i}{6g_s}}\dot{X^k}[X^i,X^j]`$
$`+{\displaystyle \frac{1}{12g_s}}\theta \gamma ^{[ijkl]}[X_l,\theta ]+{\displaystyle \frac{i}{12}}\theta \gamma ^{[ijk9]}D\theta ),`$
$`J^{ij9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{1}{6}\dot{X^i}DX^j+\frac{1}{6}\dot{X^j}DX^i\frac{i}{6}\dot{A}[X^i,X^j]+\frac{1}{12g_s}\theta \gamma ^{[ij9l]}[X_l,\theta ]\right)},`$
$`J^{ij}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{i}{6g_s}}\dot{X^i}\dot{X^k}[X^k,X^j]{\displaystyle \frac{i}{6g_s}}\dot{X^j}\dot{X^k}[X^k,X^i]{\displaystyle \frac{g_s}{6}}\dot{A}\dot{X^i}DX^j`$
$`+{\displaystyle \frac{g_s}{6}}\dot{A}\dot{X^j}DX^i{\displaystyle \frac{i}{12g_s}}(\dot{X^k})^2[X^i,X^j]{\displaystyle \frac{i}{12}}g_s\dot{A}^2[X^i,X^j]{\displaystyle \frac{i}{12g_s^3}}[X^i,X^j]{\displaystyle \underset{k<l}{}}[X^k,X^l]^2`$
$`+{\displaystyle \frac{i}{12g_s}}[X^i,X^j](DX^k)^2{\displaystyle \frac{i}{6g_s^3}}[X^i,X^k][X^k,X^l][X^l,X^j]{\displaystyle \frac{i}{6g_s}}DX^iDX^k[X^k,X^j]`$
$`+{\displaystyle \frac{i}{6g_s}}DX^jDX^k[X^k,X^i]+{\displaystyle \frac{1}{12g_s}}\theta _\alpha \dot{X^k}[X^m,\theta _\beta ]\{\gamma ^{[kijm]}+\gamma ^{[jm]}\delta _{ki}\gamma ^{[im]}\delta _{kj}`$
$`+2\delta _{jm}\delta _{ki}2\delta _{im}\delta _{kj}\}_{\alpha \beta }`$
$`+{\displaystyle \frac{1}{12}}\dot{A}\theta \gamma ^{[9ijm]}[X^m,\theta ]+{\displaystyle \frac{i}{12}}\theta _\alpha \dot{X^k}D\theta _\beta \{\gamma ^{[kij9]}+\gamma ^{[j9]}\delta _{ki}\gamma ^{[i9]}\delta _{kj}\}_{\alpha \beta }`$
$`+{\displaystyle \frac{i}{4g_s^2}}\theta _\alpha [X^k,X^l][X^m,\theta _\beta ]\{\gamma ^{[jkl]}\delta _{mi}\gamma ^{[ikl]}\delta _{mj}+2\gamma ^{[lij]}\delta _{km}+2\gamma ^l\delta _{jk}\delta _{im}`$
$`2\gamma ^l\delta _{ik}\delta _{jm}+2\gamma ^j\delta _{il}\delta _{km}2\gamma ^i\delta _{jl}\delta _{km}\}_{\alpha \beta }`$
$`+{\displaystyle \frac{1}{2g_s}}\theta _\alpha DX^k[X^m,\theta _\beta ]\{\gamma ^{[jk9]}\delta _{mi}\gamma ^{[ik9]}\delta _{mj}+\gamma ^{[9ij]}\delta _{km}+\gamma ^9\delta _{jk}\delta _{im}\gamma ^9\delta _{ik}\delta _{jm}\}_{\alpha \beta }`$
$`{\displaystyle \frac{i}{2}}\theta _\alpha DX^lD\theta _\beta \{\gamma ^{[lij]}+\gamma ^j\delta _{il}\gamma ^i\delta _{jl}\}_{\alpha \beta }`$
$`+{\displaystyle \frac{i}{12}}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)(\theta \gamma ^{[kij]}\theta \theta \gamma ^k\theta +\theta \gamma ^{[9ij]}\theta \theta \gamma ^9\theta \theta \gamma ^{[ij]}\theta \theta \theta )),`$
$`J^{i9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{1}{6}}\dot{X^i}\dot{X^k}DX^k{\displaystyle \frac{i}{6}}\dot{A}\dot{X^k}[X^k,X^i]+{\displaystyle \frac{1}{12}}g_s^2\dot{A}^2DX^i`$ (79)
$`{\displaystyle \frac{1}{12}}(\dot{X^k})^2DX^i{\displaystyle \frac{1}{12g_s^2}}DX^i{\displaystyle \underset{k<l}{}}[X^k,X^l]^2{\displaystyle \frac{1}{12}}DX^i(DX^k)^2`$
$`{\displaystyle \frac{1}{6g_s^2}}[X^i,X^k][X^k,X^l]DX^l+{\displaystyle \frac{1}{12g_s}}\theta _\alpha \dot{X^k}[X^m,\theta _\beta ]\{\gamma ^{[ki9m]}+\gamma ^{[9m]}\delta _{ki}\}_{\alpha \beta }`$
$`{\displaystyle \frac{1}{12}}\dot{A}\theta _\alpha [X^m,\theta _\beta ]\{\gamma ^{[im]}+2\delta _{im}\}_{\alpha \beta }+{\displaystyle \frac{i}{6}}\dot{X^i}\theta D\theta {\displaystyle \frac{i}{12}}\dot{A}\theta \gamma ^{[i9]}D\theta `$
$`+{\displaystyle \frac{i}{4g_s^2}}\theta _\alpha [X^k,X^l][X^m,\theta _\beta ]\{\gamma ^{[9kl]}\delta _{mi}+2\gamma ^{[li9]}\delta _{km}+2\gamma ^9\delta _{il}\delta _{km}\}_{\alpha \beta }`$
$`{\displaystyle \frac{1}{2g_s}}DX^k\theta \gamma ^k[X^i,\theta ]{\displaystyle \frac{1}{2g_s}}DX^k\theta \gamma ^i[X^k,\theta ]`$
$`+{\displaystyle \frac{i}{4g_s}}\theta _\alpha [X^k,X^l]D\theta _\beta \{\gamma ^{[ikl]}+2\gamma ^l\delta _{ik}\}_{\alpha \beta }iDX^i\theta \gamma ^9D\theta `$
$`+{\displaystyle \frac{i}{12}}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)(\theta \gamma ^{[ki9]}\theta \theta \gamma ^k\theta \theta \gamma ^{[i9]}\theta \theta \theta )).`$
The free string limit can now be taken. The result for the conformal field theory limit of the matrix string membrane current is:
$`\underset{g_s0}{lim}J^{+ij}`$ $`=`$ $`0,`$
$`\underset{g_s0}{lim}J^{+i9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{\mathrm{}_s}{R}}\right){\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{1}{6}X^i\right)},`$
$`\underset{g_s0}{lim}J^{+i}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{i}{12}\theta \gamma ^{[9i]}\theta \right)},`$
$`\underset{g_s0}{lim}J^{+9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{1}{6}\dot{X^i}X^i\frac{i}{6}\theta \theta \right)},`$
$`\underset{g_s0}{lim}J^{ijk}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{i}{12}\theta \gamma ^{[ijk9]}\theta \right)},`$
$`\underset{g_s0}{lim}J^{ij9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{1}{6}\dot{X^i}X^j+\frac{1}{6}\dot{X^j}X^i\right)},`$
$`\underset{g_s0}{lim}J^{ij}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{i}{12}}\theta _\alpha \dot{X^k}\theta _\beta \{\gamma ^{[kij9]}+\gamma ^{[j9]}\delta _{ki}\gamma ^{[i9]}\delta _{kj}\}_{\alpha \beta }`$
$`{\displaystyle \frac{i}{2}}\theta _\alpha X^l\theta _\beta \{\gamma ^{[lij]}+\gamma ^j\delta _{il}\gamma ^i\delta _{jl}\}_{\alpha \beta }+𝒪\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)),`$
$`\underset{g_s0}{lim}J^{i9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{1}{6}}\dot{X^i}\dot{X^k}X^k{\displaystyle \frac{1}{12}}(\dot{X^k})^2X^i{\displaystyle \frac{1}{12}}X^i(X^k)^2`$ (80)
$`+{\displaystyle \frac{i}{6}}\dot{X^i}\theta \theta iX^i\theta \gamma ^9\theta +𝒪\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)).`$
Next, we look at the zeroth moments of the components of the Matrix 5–brane current. Explicitly, these components are:
$`M^{+ijkl}`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}({\displaystyle \frac{RM_P^6}{48\pi ^2}}[X^i,X^j][X^k,X^l]{\displaystyle \frac{RM_P^6}{48\pi ^2}}[X^i,X^k][X^l,X^j]`$
$`{\displaystyle \frac{RM_P^6}{48\pi ^2}}[X^i,X^l][X^j,X^k]+{\displaystyle \frac{RM_P^6}{48\pi ^2}}\theta \gamma ^{[jkl}[X^{i]},\theta ]),`$
$`M^{ijklm}`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}({\displaystyle \frac{5RM_P^6}{16\pi ^2}}\dot{X}^{[i}[X^j,X^k][X^l,X^{m]}]{\displaystyle \frac{5RM_P^6}{48\pi ^2}}\theta \dot{X}^{[i}\gamma ^{jkl}[X^{m]},\theta ]`$ (81)
$`{\displaystyle \frac{5iR^2M_P^9}{192\pi ^3}}\theta [X^{[i},X^j]\gamma ^{klm]}\gamma ^n[X^n,\theta ]).`$
To these components one now performs the $`T`$–duality for the $`911`$ flip, followed by the rescalings of world–sheet coordinates, background fields and coupling constants. We then obtain the explicit form of the previous components of the matrix string theory 5–brane current (with $`i,j,k,l,m9`$):
$`M^{+ijkl}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{1}{12g_s^2}}[X^i,X^j][X^k,X^l]{\displaystyle \frac{1}{12g_s^2}}[X^i,X^k][X^l,X^j]`$
$`{\displaystyle \frac{1}{12g_s^2}}[X^i,X^l][X^j,X^k]+{\displaystyle \frac{1}{6g_s}}\theta \gamma ^{[jkl}[X^{i]},\theta ]),`$
$`M^{+ijk9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{i}{2g_s}}DX^k[X^i,X^j]{\displaystyle \frac{i}{2g_s}}DX^j[X^i,X^k]+{\displaystyle \frac{i}{2g_s}}DX^i[X^j,X^k]`$
$`+{\displaystyle \frac{i}{6}}\theta \gamma ^{[ikj]}D\theta {\displaystyle \frac{1}{6g_s}}(\theta \gamma ^{9[kj}[X^{i]},\theta ]+\theta \gamma ^{[i|9|j}[X^{k]},\theta ]+\theta \gamma ^{[ik|9|}[X^{j]},\theta ])),`$
$`M^{ijklm}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{5}{4g_s^2}}\dot{X}^{[i}[X^j,X^k][X^l,X^{m]}]{\displaystyle \frac{5}{6g_s}}\theta \dot{X}^{[i}\gamma ^{jkl}[X^{m]},\theta ]`$
$`{\displaystyle \frac{5i}{12g_s^2}}\theta [X^{[i},X^j]\gamma ^{klm]}\gamma ^n[X^n,\theta ]+{\displaystyle \frac{5}{12g_s}}\theta [X^{[i},X^j]\gamma ^{klm]}\gamma ^9D\theta ),`$
$`M^{ijkl9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{5i}{g_s}}\dot{X}^{[i}[X^j,X^k]DX^{l]}+{\displaystyle \frac{5}{4g_s}}\dot{A}[X^{[i},X^j][X^l,X^{k]}]`$ (82)
$`{\displaystyle \frac{5i}{6}}\theta \dot{X}^{[i}\gamma ^{jkl]}D\theta +{\displaystyle \frac{5}{6}}\dot{A}\theta \gamma ^{[ijl}[X^{k]},\theta ]+{\displaystyle \frac{5}{2g_s}}\theta \dot{X}^{[i}\gamma ^{|9|kl}[X^{j]},\theta ]`$
$`{\displaystyle \frac{5i}{4g_s^2}}\theta [X^{[i},X^j]\gamma ^{kl]9}\gamma ^n[X^n,\theta ]+{\displaystyle \frac{5}{4g_s}}\theta [X^{[i},X^j]\gamma ^{kl]9}\gamma ^9D\theta `$
$`{\displaystyle \frac{5}{6g_s}}\theta DX^{[i}\gamma ^{jlk]}\gamma ^n[X^n,\theta ]{\displaystyle \frac{5i}{6}}\theta DX^{[i}\gamma ^{jlk]}\gamma ^9D\theta ).`$
One can now take the free string limit. The result for the conformal field theory limit of the matrix string 5–brane current is:
$`\underset{g_s0}{lim}M^{+ijkl}`$ $`=`$ $`0,`$
$`\underset{g_s0}{lim}M^{+ijk9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{i}{6}\theta \gamma ^{[ikj]}\theta \right)},`$
$`\underset{g_s0}{lim}M^{ijklm}`$ $`=`$ $`0,`$
$`\underset{g_s0}{lim}M^{ijkl9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{5i}{6}\theta \dot{X}^{[i}\gamma ^{jkl]}\theta \frac{5i}{6}\theta X^{[i}\gamma ^{jlk]}\gamma ^9\theta \right)}.`$ (83)
Finally, we look at are the zeroth moments of the components of the Matrix 6–brane current (related to nontrivial $`11`$–dimensional background metrics), which are given explicitly by:
$`S^{+ijklmn}`$ $`=`$ $`{\displaystyle \frac{iR^2M_P^9}{8\pi ^3}}\mathrm{𝐒𝐓𝐫}\left([X^{[i},X^j][X^k,X^l][X^m,X^{n]}]\right),`$
$`S^{ijklmnp}`$ $`=`$ $`\mathrm{𝐒𝐓𝐫}\left({\displaystyle \frac{7iR^2M_P^9}{8\pi ^3}}[X^{[i},X^j][X^k,X^l][X^m,X^n]\dot{X}^{p]}+𝒪(\theta ^2,\theta ^4)\right).`$ (84)
To these components we now perform $`T`$–duality for the $`911`$ flip, followed by the rescalings of world–sheet coordinates, background fields and coupling constants. We obtain the explicit form of the previous components of the matrix string theory 6–brane current (with $`i,j,k,l,m,n9`$):
$`S^{+ijklmn}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{i}{g_s^3}[X^{[i},X^j][X^k,X^l][X^m,X^{n]}]\right)},`$
$`S^{+ijklm9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right){\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{6}{g_s^2}[X^{[i},X^j][X^k,X^l]DX^{m]}\right)},`$
$`S^{ijklmnp}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)^2{\displaystyle 𝑑\sigma 𝑑\tau \mathrm{𝐒𝐓𝐫}\left(\frac{7i}{g_s^3}[X^{[i},X^j][X^k,X^l][X^m,X^n]\dot{X}^{p]}+𝒪(\theta ^2,\theta ^4)\right)},`$
$`S^{ijklmn9}`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}\left({\displaystyle \frac{R}{\mathrm{}_s}}\right)^2{\displaystyle }d\sigma d\tau \mathrm{𝐒𝐓𝐫}({\displaystyle \frac{7i}{g_s^2}}[X^{[i},X^j][X^k,X^l][X^m,X^{n]}]\dot{A}`$ (85)
$`{\displaystyle \frac{42}{g_s^2}}DX^{[j}][X^k,X^l][X^m,X^{n]}]\dot{X}^{i]}+𝒪(\theta ^2,\theta ^4)).`$
One can now take the free string limit. The result for the conformal field theory limit of the matrix string 6–brane current is:
$`\underset{g_s0}{lim}S^{+ijklmn}`$ $`=`$ $`0,`$
$`\underset{g_s0}{lim}S^{+ijklm9}`$ $`=`$ $`0,`$
$`\underset{g_s0}{lim}S^{ijklmnp}`$ $`=`$ $`0,`$
$`\underset{g_s0}{lim}S^{ijklmn9}`$ $`=`$ $`0.`$ (86)
|
warning/0005/quant-ph0005115.html
|
ar5iv
|
text
|
# Three qubits can be entangled in two inequivalent ways
## I Introduction
The understanding of entanglement is at the very heart of Quantum Information Theory (QIT). In recent years, there has been an ongoing effort to characterize qualitatively and quantitatively the entanglement properties of multiparticle systems. A situation of particular interest in QIT consists of several parties that are spatially separated from each other and share a composite system in an entangled state. This setting conditionates the parties —which are typically allowed to communicate through a classical channel— to only act locally on their subsystems. But even restricted to local operations assisted with classical communication (LOCC), the parties can still modify the entanglement properties of the system and in particular they can try to convert an entangled state into another. This possibility leads to natural ways of defining equivalence relations in the set of entangled states —where equivalent states are then said to contain the same kind of entanglement—, also of establishing hierarchies between the resulting classes.
For instance, we could agree in identifying any two states which can be obtained from each other with certainty by means of LOCC. Clearly, this criterion is interesting in QIT because the parties can use these two states indistintively for exactly the same tasks. It is a celebrated result that, when applied to many copies of a state, this criterion leads to identifying all bipartite pure-state entanglement with that of the EPR state $`1/\sqrt{2}(|00+|11)`$. That is, the entanglement of any pure state $`|\psi _{AB}`$ is asymptotically equivalent, under deterministic LOCC, to that of the EPR state, the entropy of entanglement $`E(\psi _{AB})`$ —the entropy of the reduced density matrix of either system $`A`$ or $`B`$— quantifying the amount of EPR entanglement contained asymptotically in $`|\psi _{AB}`$. In contrast, recent contributions have shown that in systems shared by three or more parties there are several inequivalent forms of entanglement under asymptotic LOCC .
This paper is essentially concerned with the entanglement properties of a single copy of a state, and thus asymptotic results do not apply here. For single copies it is known that two pure states $`|\psi `$ and $`|\varphi `$ can be obtained with certainty from each other by means of LOCC if and only if they are related by local unitaries LU . But even in the simplest bipartite systems, $`|\psi `$ and $`|\varphi `$ are typically not related by LU, and continuous parameters are needed to label all equivalence classes . That is, one has to deal with infinitely many kinds of entanglement. In this context an alternative, simpler classification would be advisable.
One such classification is possible if we just demand that the conversion of the states is through stochastic local operations and classical communication (SLOCC) ; that is, through LOCC but without imposing that it has to be achieved with certainty. In that case we can establish an equivalence relation stating that two states $`|\psi `$ and $`|\varphi `$ are equivalent if the parties have a non-vanishing probability of success when trying to convert $`|\psi `$ into $`|\varphi `$, and also $`|\varphi `$ into $`|\psi `$ . This relation has been termed stochastic equivalence in Ref. . Their equivalence under SLOCC indicates that both states are again suited to implement the same tasks of QIT, although this time the probability of a successful performance of the task may differ from $`|\varphi `$ to $`|\psi `$. Notice in addition that since LU are a particular case of SLOCC, states equivalent under LU are also equivalent under SLOCC, the new classification being a coarse graining of the previous one.
The main aim of this work is to identify and characterize all possible kinds of pure-state entanglement of three qubits under SLOCC. Unentangled states, and also those which are product in one party while entangled with respect to the remaining two, appear as expected, trivial cases. More surprising is the fact that there are two different kinds of genuine tripartite entanglement. Indeed, we will show that any (non-trivial) tripartite entangled state can be converted, by means of SLOCC, into one of two standard forms, namely either the GHZ state
$$|GHZ=1/\sqrt{2}(|000+|111),$$
(1)
or else a second state
$$|W=1/\sqrt{3}(|001+|010+|100),$$
(2)
and that this splits the set of genuinely trifold entangled states into two sets which are unrelated under LOCC. That is, we will see that if $`|\psi `$ can be converted into the state $`|GHZ`$ in (1) and $`|\varphi `$ can be converted into the state $`|W`$ in (2), then it is not possible to transform, not even with only a very small probability of success, $`|\psi `$ into $`|\varphi `$ nor the other way round.
The previous result is based on the fact that, unlike the GHZ state, not all entangled states of three qubits can be expressed as a linear combination of only two product states. Remarkably enough, the inequivalence under SLOCC of the states $`|GHZ`$ and $`|W`$ can alternatively be shown from the fact that the 3-tangle (residual tangle), a measure of tripartite correlations introduced by Coffman et. al. , does not increase on average under LOCC, as we will prove here.
We will then move to the second main goal of this work, namely the analysis of the state $`|W`$. It can not be obtained from a state $`|GHZ`$ by means of LOCC and thus one could expect, in principle, that it has some interesting, characteristic properties. Recall that in several aspects the GHZ state can be regarded as the maximally entangled state of three qubits. However, if one of the three qubits is traced out, the remaining state is completely unentangled. Thus, the entanglement properties of the state $`|GHZ`$ are very fragile under particle losses. We will prove that, oppositely, the entanglement of $`|W`$ is maximally robust under disposal of any one of the three qubits, in the sense that the remaining reduced density matrices<sup>*</sup><sup>*</sup>*The reduced density matrix $`\rho _{AB}`$ of a pure tripartite state $`|\psi `$ is defined as $`\rho _{AB}tr_C(|\psi \psi |)`$. $`\rho _{AB}`$, $`\rho _{BC}`$ and $`\rho _{AC}`$ retain, according to several criteria, the greatest possible amount of entanglement, compared to any other state of three qubits, either pure or mixed.
We will finally analyze entanglement under SLOCC in more general multipartite systems. We will show that, for most of these systems, there is typically no chance at all to transform locally a given state into some other if they are chosen randomly, because the space of entangled pure states depends on more parameters than those that can be modified by acting locally on the subsystems.
The paper is organized as follows. In section II we characterize mathematically the equivalence relation established by stochastic conversions under LOCC, and illustrate its performance by applying it to the well-known bipartite case. In section III we move to consider a system of three qubits, for which we prove the existence of 6 classes of states under SLOCC —including the 2 genuinely tripartite ones—. Section IV is devoted to study the endurance of the entanglement of the state $`|W`$ against particle losses. In section V more general multipartite systems are considered. Section VI contains some conclusions. Finally, appendix A to C prove, respectively, some needed results related to SLOCC, the monotonicity of the 3-tangle under LOCC and the fact that $`|W`$ retains optimally bipartite entanglement when one qubit is traced out.
## II Kinds of entanglement under Stochastic LOCC
In this work we define as equivalent the entanglement of two states $`|\psi `$ and $`|\varphi `$ of a multipartite system iff local protocols exist that allow the parties to convert each of the two states into the other one with some a priori probability of success. In this approach, we follow the definition for stochastic equivalence as given in Stochastic transformations under LOCC had been previously analyzed in .. The underlying motivation for this definition is that, if the entanglement of is $`|\psi `$ and $`|\varphi `$ is equivalent, then the two states can be used to perform the same tasks, although the probability of a successful performance of the task may depend on the state that is being used.
### A Invertible local operators
Sensible enough, this classification would remain useless if in practice we would not be able to find out which states are related by SLOCC. Let us recall that, all in all, no practical criterion is known so far that determines whether a generic transformation can be implemented by means of LOCC. However, we can think of any local protocol as a series of rounds of operations, where in each round a given party manipulates locally its subsystem and communicates classically the result of its operation (if it included a measurement) to the rest of parties. Subsequent operations can be made dependent on previous results and the protocol splits into several branches. This picture is useful because for our purposes we need only focus on one of these branches. Suppose that state $`|\psi `$ can be locally converted into state $`|\varphi `$ with non-zero probability. This means that at least one branch of the protocol does the job. Since we are concerned only with pure states we can always characterize mathematically this branch as an operator which factors out as the tensor product of a local operator for each party. For instance, in a three-qubit case we would have that $`|\psi `$ can be locally converted into $`|\varphi `$ with some finite probability iff an operator $`ABC`$ exists such that
$$|\varphi =ABC|\psi ,$$
(3)
where operator $`A`$ contains contributions coming from any round in which party A acted on its subsystem, and similarly for operators $`B`$ and $`C`$ In practice the constraints $`A^{}A,B^{}B,C^{}C1`$ should be fulfilled if the invertible operators $`A,B,C`$ are to come from local POVMs. In this work we do not normalize them in order to avoid introducing unimportant constants to the equations. Instead, both the initial and final states are normalized. . Carrying on with the 3-qubit example, let us now consider for simplicity that both states $`|\psi `$ and $`|\varphi `$ have rank 2 reduced density matrices $`\rho _A`$ tr$`{}_{BC}{}^{}(|\psi \psi |),\rho _B`$ and $`\rho _C`$. Then clearly the rank of operators $`A`$, $`B`$ and $`C`$ need to be 2 (see appendix A). In other words, each of these operators is necessarily invertible, and in particular
$$|\psi =A^1B^1C^1|\varphi .$$
(4)
We see thus that, under the assumption of maximal rank for the reduced density matrices, two-way convertibility implies the existence of invertible operators $`A`$, $`B`$ and $`C`$ as in (3) \[actually, one-way convertibility alone has already implied that an invertible local operator (ILO) $`ABC`$ exists\]. Obviously, the converse also holds, namely that if an ILO $`ABC`$ exists then for each direction of the conversion a local protocol can be build that succeeds with non-zero probability. As explained in appendix A in detail, we can get rid of the previous assumption on the ranks and announce with generality,
Result: States $`|\psi `$ and $`|\varphi `$ are equivalent under stochastic local operations and classical communication —SLOCC— iff an invertible local operator —ILO— relating them \[as in, for instance, equation (3)\] exists.
### B Bipartite entanglement under SLOCC
What does this classification implies in the well-known case of bipartite systems? Since LU are included in SLOCC, we can take the Schmidt decomposition of a pure state $`|\psi \mathrm{I}\mathrm{C}^n\mathrm{I}\mathrm{C}^m`$, $`nm`$, as the starting point for our analysis. Thus,
$$\underset{i=1}{\overset{n_\psi }{}}\sqrt{\lambda _i}|i|i=U_AU_B|\psi ;\lambda _i>0,n_\psi n,$$
(5)
where $`U_A`$ and $`U_B`$ are some proper local unitaries, the coefficients $`\lambda _i`$ decrease with $`i`$, and $`n_\psi `$ is the number of non-vanishing terms in the Schmidt decomposition. Clearly, the ILO
$$\frac{1}{\sqrt{n_\psi }}(\underset{i=1}{\overset{n_\psi }{}}\frac{1}{\sqrt{\lambda _i}}|ii|+\underset{i=n_\psi +1}{\overset{n}{}}|ii|)1_B$$
(6)
transforms (5) into a maximally entangled state
$$\frac{1}{\sqrt{n_\psi }}\underset{i}{\overset{n_\psi }{}}|i|i,$$
(7)
which depends only on the Schmidt number $`n_\psi `$. Since SLOCC cannot modify the rank of the reduced density matrices $`\rho _A`$ and $`\rho _B`$, which is given by $`n_\psi `$, we conclude that in $`\mathrm{I}\mathrm{C}^n\mathrm{I}\mathrm{C}^m`$, $`nm`$, there are $`n`$ different kinds of entangled states, corresponding to $`n`$ different classes under SLOCC. Each of these classes is characterized by a given Schmidt number, and we can choose as their representatives the state (7) with $`n_\psi =1,\mathrm{},n`$. Clearly $`n_\psi =1`$ corresponds to states that are less entangled than the rest (they are, after all, unentangled). This hierarchical relation can be extended to the rest of classes by noting that none-invertible local operators can project out some of the Schmidt terms and thus diminish the Schmidt number of a state. Therefore the state $`|\psi `$ can be locally converted into the state $`|\varphi `$ with some finite probability iff $`n_\psi n_\varphi `$, or in terms of kinds of entanglement, we can say that the entanglement of the class characterized by a given Schmidt number is more powerful than that of a class with a smaller Schmidt number.
For later reference we also note that in a two-qubit system, $`=\mathrm{I}\mathrm{C}^2\mathrm{I}\mathrm{C}^2`$, we can write any state, after using a convenient LU, uniquely as
$$|\psi =c_\delta |0|0+s_\delta |1|1;c_\delta s_\delta 0,$$
(8)
where $`c_\delta ,s_\delta `$ stand for cos$`\delta `$ and sin$`\delta `$. This is either a product (unentangled) state $`|\psi _{AB}=|0|0`$ for $`c_\delta =1`$ or else an entangled state that can be converted into the EPR state,
$$\frac{1}{\sqrt{2}}(|0|0+|1|1),$$
(9)
with probability $`p=E_2(\psi )`$, where $`E_2(\psi )\lambda _2`$ is the entanglement monotone that provides a quantitative description of the non-local resources contained in a single copy of a two-qubit pure state . Any state $`|\psi `$ can be obtained from state (9) with certainty, this contributing to the fact that the EPR state is considered the maximally entangled state of two qubits.
## III Entanglement of pure states of three qubits
In this section we analyze a system of three qubits. We show that SLOCC split the set of pure states into 6 inequivalent classes, which further structure themselves into a three-grade hierarchy when non-invertible local operations are used to relate them. At the top of the hierarchy we find two inequivalent classes of true tripartite entanglement, which we name GHZ-class and W-class after our choice of corresponding representatives. The three possible classes of bipartite entanglement are accessible (with some non-vanishing probability) from any state of the W and GHZ classes by means of a non-invertible local operator. Finally, at the bottom of the hierarchy we find non-entangled states.
The ranks r$`(\rho _A)`$, r$`(\rho _B)`$ and r$`(\rho _C)`$ of the reduced density matrices, together with the range $`R(\rho _{BC})`$ of $`\rho _{BC}`$, will be the main mathematical tools used through the first part of this section. By analysing them we will be able to make an exhaustive classification of three-qubit entanglement. Later on we will rephrase some of these results in terms of well-known measures of entanglement. In particular, we will see that the existence of two inequivalent kinds of true tripartite entanglement under SLOCC is very much related to the fact that the 3-tangle, a measure of tripartite entanglement introduced in , is an entanglement monotone (see appendix B).
At the end of the section also a practical way to identify the class an arbitrary state belongs to will be discussed.
### A Non-entangled states and bipartite entanglement.
If at least one of the local ranks r$`(\rho _A)`$, r$`(\rho _B)`$ or r$`(\rho _C)`$ is 1, then the pure state of the three qubits factors out as the tensor product of two pure states, and this implies that at least one of the qubits is uncorrelated with the other two. SLOCC distinguish states with this feature depending on which qubits are uncorrelated from the rest.
Class A-B-C (product states)
This class corresponds to states with $`r(\rho _A)=r(\rho _B)=r(\rho _C)=1`$. They can be taken, after using some convenient LU, into the form
$$|\psi _{ABC}=|0|0|0,$$
(10)
where we have already relaxed the notation for $`|0|0|0`$.
Classes A-BC, AB-C and C-AB
(bipartite entanglement)
These three classes of states contain only bipartite entanglement between two of the qubits, one of the reduced density matrices having rank 1 and the other two having rank 2. For example, the states in class $`ABC`$ possess entanglement between the systems $`B`$ and $`C`$ ($`r(\rho _B)=r(\rho _C)=2`$) and are product with respect to system $`A`$ ($`r(\rho _A)=1`$). LU allow us to write uniquely states of the class $`ABC`$ as
$$|\psi _{ABC}=|0(c_\delta |0|0+s_\delta |1|1),c_\delta s_\delta >0,$$
(11)
and similary for $`|\psi _{BAC}`$ and $`|\psi _{CAB}`$. We choose the maximally entangled state
$$\frac{1}{\sqrt{2}}|0(|0|0+|1|1)$$
(12)
as representative of the class $`ABC`$. Any other state within this class can be obtained from (12) with certainty by means of LOCC.
The proof that these four marginal classes are inequivalent under SLOCC is very simple. We only need to recall that the local ranks are invariant under ILO (see appendix A). In what follows we will analyze the more interesting case of $`r(\rho _\kappa )=2,\kappa =A,B,C`$. To see that there are two inequivalent classes fulfilling this condition we will have to study possible product decompositions of pure states.
### B True three-qubit entanglement.
There turns out to be a close connection between convertibility under SLOCC and the way entangled states can be expressed minimally as a linear combination of product states. For instance, as we shall prove later on, the GHZ and W states have a different number of terms in their minimal product decompositions (1) and (2), namely 2 and 3 product terms respectively, and this readily implies that there is no way to convert one state into the other by means of an ILO $`ABC`$. Indeed, let us consider, e.g., the most general pure state that can be obtained reversibly from a $`|GHZ`$. It reads
$$ABC|GHZ=\frac{1}{\sqrt{2}}(|A0|B0|C0+|A1|B1|C1),$$
(13)
where $`|A0`$ and $`|A1`$ are linearly independent vectors (since $`A`$ is invertible) and similarly for the other two qubits. That is, the minimal number of terms in a product decomposition for the state (13) is also 2. Actually, we have that also for a general multipartite system,
Observation: The minimal number of product terms for any given state remains unchanged under SLOCC.
This simple observation tells us already that in three qubits there are at least two inequivalent kinds of genuine tripartite entanglement under SLOCC, that of $`|GHZ`$ and that of $`|W`$.
However, we still have to prove that the state $`|W`$ cannot be expressed as a linear combination of just two product vectors. In order to complete our classification we also have to show that any pure state of three qubits with maximal local ranks can be reversibly converted into either the state $`|GHZ`$ or the state $`|W`$. We start with an obvious lemma regarding product decompositions:
Lemma: Let $`_{i=1}^l|e_i|f_i`$ be a product decomposition for the state $`|\eta _E_F`$. Then the set of states $`\{|e_i\}_{i=1}^l`$ span the range of $`\rho _E`$Tr$`{}_{F}{}^{}|\eta \eta |`$.
Proof: We have that $`\rho _E=_{i,j=1}^lf_i|f_j|e_je_i|`$. On the other hand $`|\nu `$ is in the range of $`\rho _E`$ iff a state $`|\mu `$ exists such that $`|\nu =\rho _E|\mu `$, that is $`|\nu =_{i,j=1}^lf_i|f_je_i|\mu |e_j`$. $`\mathrm{}`$
In particular, $`r(\rho _A)=2`$ implies that at least two product terms are needed to expand $`|\psi \mathrm{I}\mathrm{C}^2\mathrm{I}\mathrm{C}^2\mathrm{I}\mathrm{C}^2`$. Let us suppose that a product decomposition with only two terms is possible, namely
$$|\psi =|a_1|b_1|c_1+|a_2|b_2|c_2.$$
(14)
Then, also according to the previous lemma, $`|b_1`$$`|c_1`$ and $`|b_2`$$`|c_2`$ have to span the range of $`\rho _{BC}`$, R$`(\rho _{BC})`$.
But R$`(\rho _{BC})`$ is a two dimensional subspace of $`\mathrm{I}\mathrm{C}^2\mathrm{I}\mathrm{C}^2`$. Therefore it always contains either only one or only two product states \[unless R$`(\rho _{BC})`$ was supported in $`\mathrm{I}\mathrm{C}\mathrm{I}\mathrm{C}^2`$ or $`\mathrm{I}\mathrm{C}^2\mathrm{I}\mathrm{C}`$, but we already excluded this possibility because we are considering r$`(\rho _B)=`$r$`(\rho _C)=2`$\]. Notice that a two-term decomposition (14) requires that R$`(\rho _{BC})`$ contains at least two product vectors. Only one product vector in R$`(\rho _{BC})`$, and thus the impossibility of decomposition (14), is going to be precisely the trait of the states in the W-class.
GHZ-class
Let us suppose first that R$`(\rho _{BC})`$ contains two product vectors, $`|b_1`$$`|c_1`$ and $`|b_2`$$`|c_2`$. Then decomposition (14) is possible, and actually unique, with $`|a_i=\xi _i|\psi `$, $`i=1,2`$, where $`|\xi _i`$ are the two vectors supported in $`R(\rho _{BC})`$ that are biorthonormal to the $`|b_i`$$`|c_i`$. In this case we can use LU in order to take $`|\psi `$ into the useful standard product form (see also )
$$|\psi _{GHZ}=\sqrt{K}(c_\delta |0|0|0+s_\delta e^{i\phi }|\phi _A|\phi _B|\phi _C),$$
(15)
where
$`|\phi _A=c_\alpha |0+s_\alpha |1`$ (16)
$`|\phi _B=c_\beta |0+s_\beta |1`$ (17)
$`|\phi _C=c_\gamma |0+s_\gamma |1`$ (18)
and $`K=(1+2c_\delta s_\delta c_\alpha c_\beta c_\gamma c_\phi )^1(1/2,\mathrm{})`$ is a normalization factor. The ranges for the five parameters are $`\delta (0,\pi /4],\alpha ,\beta ,\gamma (0,\pi /2]`$ and $`\phi [0,2\pi )`$.
All these states are in the same equivalence class as the $`|GHZ`$ (1) under SLOCC. Indeed, the ILO
$`\sqrt{2K}\left(\begin{array}{cc}c_\delta \hfill & s_\delta c_\alpha e^{i\phi }\hfill \\ 0\hfill & s_\delta s_\alpha e^{i\phi }\hfill \end{array}\right)\left(\begin{array}{cc}1\hfill & c_\beta \hfill \\ 0\hfill & s_\beta \hfill \end{array}\right)\left(\begin{array}{cc}1\hfill & c_\gamma \hfill \\ 0\hfill & s_\gamma \hfill \end{array}\right),`$ (25)
applied to $`|GHZ`$ produces the state (15).
The GHZ state is a remarkable representative of this class. It is maximally entangled in several senses . For instance, it maximally violates Bell-type inequalities, the mutual information of measurement outcomes is maximal, it is maximally stable against (white) noise and one can locally obtain from a GHZ state with unit propability an EPR state shared between any two of the three parties. Another relevant feature is that when any one of the three qubits is traced out, the remaining two are in a separable —and therefore unentangled— state.
W-class
Let us move to analyze the case where R$`(\rho _{BC})`$ contains only one product vector. We already argued that decomposition (14) is now not possible. Instead we can (uniquely) write
$$|\psi =|a_1|b_1|c_1+|a_2|\varphi _{BC},$$
(26)
where $`|\varphi _{BC}`$ is the vector of $`R(\rho _{BC})`$ which is orthogonal to $`|b_1`$$`|c_1`$, and $`|a_1`$ and $`|a_2`$ are given by $`b_1|c_1|\psi `$ and $`\varphi _{BC}|\psi `$. By means of LU (26) can be always rewritten as
$`|\psi =(\sqrt{c}|1+\sqrt{d}|0)`$ $`|00`$ (27)
$`+|0`$ $`(\sqrt{a}|01+\sqrt{b}|10).`$ (28)
Indeed, we first take $`|b_1`$$`|c_1`$ into $`|0`$$`|0`$. Then, since $`|\varphi _{BC}`$ has been chosen orthogonal to $`|b_1`$$`|c_1`$, it must become $`x|01+y|10+z|11`$. By requiring that a linear combination of these two vectors has no second product vector we obtain that $`z=0`$ . In addittion the coefficients $`\sqrt{a}x,\sqrt{b}y,\sqrt{c}`$ and $`\sqrt{d}`$ can be made positive by absorbing the three relative phases into the definition of state $`|1`$ of subsystems $`A`$, $`B`$ and $`C`$. Thus case (i) has been taken into the form (28) by just using LU. Before we showed that 2 terms could not suffice in a product decomposition of the state. Now we see that 3 product terms always do the job, for instance $`(\sqrt{c}|1+\sqrt{d}|0)|00`$, $`\sqrt{a}|0|01`$ and $`\sqrt{b}|0|10`$ once we took the original state into the standard, unique form
$$|\psi _W=\sqrt{a}|001+\sqrt{b}|010+\sqrt{c}|100+\sqrt{d}|000,$$
(29)
where $`a,b,c>0`$, and $`d1(a+b+c)0`$.
The parties can locally obtain the state (29) from the state $`|W`$ in (2), which we choose as a representative of the class —and whose study we postpone for later on—, by application of an ILO of the form
$`\left(\begin{array}{cc}\sqrt{a}\hfill & \sqrt{d}\hfill \\ 0\hfill & \sqrt{c}\hfill \end{array}\right)\left(\begin{array}{cc}\sqrt{3}\hfill & 0\hfill \\ 0\hfill & \frac{\sqrt{3b}}{\sqrt{a}}\hfill \end{array}\right)\left(\begin{array}{cc}1\hfill & 0\hfill \\ 0\hfill & 1\hfill \end{array}\right).`$ (36)
Before moving to relate these classes by means of non–invertible local operators, we note that states within the GHZ-class and the W-class depend, respectively, on 5 and 3 parameters that cannot be changed by means of LU. Previous works have shown that a generic state of three qubits depends, up to LU, on 5 parameters. This means that states tipically belong to the GHZ-class, or equivalently, that a generic pure state of three qubits can be locally transformed into a GHZ with finite probability of success (see also ). The W-class is of zero measure compared to the GHZ-class. This does not mean, however, that it is irrelevant. In a similar way as separable mixed states are not of zero measure with respect to entangled states, even though product states are, it is in principle conceivable that mixed states having only W-class entanglement are also not of zero measure in the set of mixed states.
### C Relating SLOCC–classes by means of non–invertible operators
In this subsection, we investigate the hierarchical relation of the 6 SLOCC-equivalence classes under non–invertible operators, i.e. under general LOCC.
A non–invertible local operator transforms $`|\psi `$ into $`|\varphi `$ according to (3), but with at least one of the local operators $`A`$, $`B`$ and $`C`$ having rank 1. This means that the local ranks of the pure states can be diminished. For instance, if the initial state $`|\psi `$ belongs either to the GHZ or W class, then a non-invertible operator will diminish at least one of the local ranks. That is, $`|\varphi `$ belongs necessarily to one of the bipartite classes $`\kappa \mu \nu (\kappa \mu \nu \{A,B,C\})`$ or else is a product state $`ABC`$.
Thus we have that the classes GHZ and W are also inequivalent even under most general LOCC, whereas e.g. a measurement of the projector $`P=|++|`$ with $`|+=1/\sqrt{2}(|0+|1)`$ in party $`A`$ maps states within the classes $`W`$ (29) and $`GHZ`$ (15) to states within the class $`ABC`$. In a similar way, non–invertible local operators (local, standard measurements) can convert states within one of the classes $`\kappa \mu \nu `$ to states within the class $`ABC`$. Note that in all cases described above, the inverse transformations, e.g. from the class $`ABC`$ to one of the classes $`\kappa \mu \nu `$ are impossible as they would imply an increase of the rank of at least one of the reduced density operators $`\rho _A,\rho _B,\rho _C`$. These results are summarized in Fig. 1.
### D Measures of entanglement and classes under SLOCC
Several measures have been introduced so far in the literature in order to quantify entanglement. Although this section is mainly concerned with qualitative aspects of multipartite quantum correlations, we would like to relate some of these measures, namely some bipartite ones and the tripartite 3-tangle (see appendix B), to our classification. Remarkably, the existence of two kinds of genuine tripartite entanglement in a three-qubit system, as well as the inequivalence between bipartite and tripartite entanglement, can be easily understood from the non-increasing character of these measures under LOCC. In addittion, the 3-tangle allows for a systematic and practical identification of which class under SLOCC any pure state belongs to.
For each $`\kappa =A,B`$ and $`C`$ we can regard the three-qubit system as a bipartite system, with qubit $`\kappa `$, say $`A`$ for concreteness, being one part of the system and the remaining two qubits, $`B`$ and $`C`$, being the other. Correspondingly, a state $`|\psi `$ of the three qubits can be viewed as a bipartite state $`|\psi _{A(BC)}`$. For bipartite states several measures are known, which are entanglement monotones ; that is, which cannot be increased, on average, under LOCC. For instance, we already mentioned the entropy of entanglement $`E(\psi )`$ for asymptotic conversions –given by the entropy $`S_A`$ of the eigenvalues of $`\rho _A`$— and the monotone $`E_2(\psi )`$ for the single copy case —which is given by the smallest eigenvalue $`\lambda _2`$ of $`\rho _A`$. They all satisfy that vanish for product states (corresponding to $`\rho _A`$ with rank 1) while having a positive value for any other state (corresponding to $`\rho _A`$ with rank 2). Thus we can interpret the inequivalence under SLOCC of states whose reduced density matrices differ in rank also in terms of the impossibility of creating any of the bipartite measures. For instance, a state in the $`ABC`$ class has $`S_A=0`$, and thus cannot be transformed with any finite probability into a state of the $`ABC`$ class, because this would have $`S_A>0`$. We conclude that the monotonicity of these measures readily split the set of pure states of three qubits into five subsets which are inequivalent under SLOCC, namely unentagled states $`ABC`$, the three classes $`ABC`$, $`ABC`$ and $`CAB`$ containing only bipartite entanglement, and a fifth subset of entangled states with $`S_A,S_B,S_C0`$ (i.e. r$`(\rho _A)=`$r$`(\rho _B)=`$r$`(\rho _C)=2`$). Bipartite measures cannot, however, determine the inequivalence of the GHZ and W classes.
Is there any known measure of tripartite entanglement which can distinguish between these two classes? The 3-tangle does. Indeed, it can be computed from the product decompositions (15) and (29) (see for details), and reads
$$\tau (\psi _{GHZ})=(2Ks_\alpha s_\beta s_\gamma s_\delta c_\delta )^20$$
(37)
for any state in the GHZ class, while it vanishes for any state in the W class. In the appendix B we prove that the 3-tangle is an entanglement monotone, a very desirable property for any quantity aiming at measuring entanglement. Consequently, a state in the W class cannot be transformed by means of LOCC (and in particular SLOCC) to a state in the GHZ class, which is an independent proof of the fact that the two kinds of true tripartite entanglement are indeed inequivalent under SLOCC.
### E Practical identification
Given an arbitrary state $`|\psi `$ of three qubits, expressed in any basis, it may be interesting to know, for instance, whether it can be converted by means of LOCC into a GHZ or a W state, if any, or into a EPR state shared between two of the parties. In our original analysis of the classes we already have provided a constructive method, based on the analysis of r$`(\rho _\kappa )`$ and R$`(\rho _{BC})`$, to determine the class of $`|\psi `$ under SLOCC. Analysing the R$`(\rho _{BC})`$ may, however, not be the most practical way to proceed. Here we suggest to proceed instead according to the following two steps:
* compute $`\rho _\kappa `$, $`\kappa =A,B`$ and $`C`$, and check whether they have a vanishing determinant. \[note that det$`\rho _\kappa =0S_\kappa =0`$r$`(\rho _\kappa )=1`$\]
* If none of the previous determinants vanish \[that is, $`|\psi `$ has true tripartite entanglement\], then compute the $`3`$-tangle using the recipe introduced in .
Then Table I, which sumarizes the relation between classes under SLOCC and measures of entanglement, can be used to catalogue state $`|\psi `$.
## IV State $`|W`$ and residual bipartite entanglement.
As mentioned in the previous section, in several aspects the state $`|GHZ`$ is the maximally entangled state of three qubits. It also has the feature that when one of the qubits is traced out, then the remaining two are completely unentangled. This means, in particular, that if one of the three parties sharing the system decides not to cooperate with the other two, then they can not use at all the entanglement resources of the state. The same happens if for some reason the information about one of the qubits —namely the identity of the corresponding states $`|0`$ and $`|1`$ in (1)— is lost.
Here we would like to investigate the robustness of the entanglement of a three-qubit state $`|\psi `$ against disposal of one of the qubits . The residual, two-qubit states $`\rho _{AB}`$, $`\rho _{AC}`$ and $`\rho _{BC}`$ are in general mixed states. There are several measures of entanglement of mixed states and therefore multiple ways of quantifying how much (mixed-state) bipartite entanglement the state $`|\psi `$ turns into when one of the qubits is traced out. Nevertheless, most of the criteria we have examined coincide in pointing out the state $`|W`$ as the one that maximally retains bipartite entanglement. Note that the reduced density matrix of $`|W`$ is identical for any two subsystems and is e.g. given by
$$\rho _{AB}=\frac{2}{3}|\mathrm{\Psi }^+\mathrm{\Psi }^+|+\frac{1}{3}|0000|,$$
(38)
with $`|\mathrm{\Psi }^+=1/\sqrt{2}(|01+|10)`$ being a maximally entangled state of two qubits. Note that one can obtain from a single copy of (38) a state which is arbitrarily close to the state $`|\mathrm{\Psi }^+`$ by means of a filtering measurement .
### A Average residual entanglement
Let us consider first which is the amount of bipartite entanglement, according to some measure $`(\rho )`$, that the two remaining qubits retain on average when a third one is traced out, that is,
$$\overline{}(\psi )\frac{1}{3}((\rho _{AB})+(\rho _{AC})+(\rho _{BC})).$$
(39)
In general, computing the amount of entanglement $`(\rho )`$ for bipartite mixed states is a difficult problem. However numerical results have shown that $`|W`$ maximizes the average entanglement of formation, that is the choice $`(\rho )=E_f(\rho )`$, where $`E_f(\rho )`$ <sup>§</sup><sup>§</sup>§The entanglement of formation is given by $`E_f(\rho )=h(\frac{1}{2}+\frac{1}{2}\sqrt{1𝒞^2})`$, where $`𝒞`$ is the concurrence and $`h`$ is the binary entropy function $`h(x)=x\mathrm{log}_2x(1x)\mathrm{log}_2(1x)`$. is the minimal amount of bipartite pure-state entanglement \[as quantified by means of the entropy of entanglement\] required to prepare locally one single copy of the state $`\rho `$ .
In addition, we have managed to show analytically (see appendix C) for the particular choice $`(\rho )=𝒞(\rho )^2`$, where $`𝒞(\rho )`$ is the concurrence (for a definition of the concurrence see e.g. ), the state $`|W`$ reaches the maximal average value $`\overline{𝒞^2}(W)=4/9`$, which no other state can match.
### B Least entangled pair
Another way of quantifying how resistent the entanglement of a tripartite state $`|\psi `$ is to dismissal of one part of the system consists in looking at the least entangled of the three possible remaining parts, namely at the function
$$_{\mathrm{min}}(\psi )\mathrm{min}((\rho _{AB}),(\rho _{AC}),(\rho _{BC})).$$
(40)
For this “worst case scenario” we have been able to prove analytically (see appendix C) that the maximal value of $`_{\mathrm{min}}(\psi )`$ is obtained by the state $`|W`$ for any bipartite measure $`(\rho )`$ which is monotonic with the concurrence, $`𝒞(\rho )`$, such as the entanglement of formation $`E_f(\rho )`$ and the monotone $`E_2(\rho )`$ The entanglement monotone $`E_2`$, expressed in terms of the concurrence $`𝒞`$ is given by $`E_2(\rho )=\frac{1}{2}\frac{1}{2}\sqrt{1𝒞^2}`$., which denotes the minimal amount of bipartite pure-state entanglement \[quantified by means of $`E_2(\psi )`$\] required to prepare locally one single copy of the state $`\rho `$.
We conclude that the state $`|W`$ is the state of three-qubits whose entanglement has the highest degree of endurance against loss of one of the three qubits. We conceive this property as important in any situation where one of the three parties sharing the system, say Alice, may suddenly decide not to cooperate with the other two. Notice that even in the case that Alice would decide to try to destroy the entanglement between Bob and Claire, this would not be possible, since any local action on A cannot prevent Bob and Claire from sharing, at least, the entanglement contained in $`\rho _{BC}`$ (for instance, by simply ignoring Alice’s actions). Therefore, although essentially tripartite, the entanglement of the state $`|W`$ is also readily bipartite, in contrast to that of the state $`|GHZ`$, which only after some local manipulation can be brought into a bipartite form.
## V Generalization to $`N`$ parties
In this last section we would like to apply the same techniques to analyze the entanglement of more general multipartite systems. We will learn that the set of entangled states is a rather inaccessible jungle for the local explorer, for two pure states $`|\psi `$ and $`|\varphi `$ are typically not connected by means of LOCC, so that the parties are usually unable to convert states locally. We will also study generalizations to $`N`$ qubits of the state $`|W`$.
### A Local inaccessibility of states in general multipartite systems
Let us consider first $`N`$ parties each possessing a qubit. The Hilbert space of the system is
$$^{(N)}=\underset{N}{\underset{}{\mathrm{I}\mathrm{C}^2\mathrm{I}\mathrm{C}^2\mathrm{}\mathrm{I}\mathrm{C}^2}},$$
(41)
and therefore up to a global, physically irrelevant complex constant, a generic vector depends on $`2(2^N1)`$ real parameters. On the other hand we want to identify vectors which are related by means of a ILO. A general one-party, invertible operator $`A`$ must have non-vanishing determinant, which we can fix to one, det$`A=1`$, because the operator $`kA`$ only differs in that it introduces in the transformed states an extra constant factor $`k\mathrm{I}\mathrm{C}`$, which we have already addressed. That is, $`ASL_2(\mathrm{I}\mathrm{C})`$, and it depends on $`6`$ real parameters. Therefore the set of equivalence classes under SLOCC,
$$\frac{^{(N)}}{\underset{N}{\underset{}{SL_2(\mathrm{I}\mathrm{C})\times SL_2(\mathrm{I}\mathrm{C})\times \mathrm{}\times SL_2(\mathrm{I}\mathrm{C})}}},$$
(42)
depends at least on $`2(2^N1)6N`$ parameters. This lower bound allows for a finite number of classes for $`N=3`$, but shows that for any larger number $`N`$ of qubits there are infinitely many classes, labeled by at least one continuous parameter. The reason is that the number of parameters from a state $`|\psi `$ which the parties can modify by means of a general ILO $`AB\mathrm{}N`$ grows linearly with $`N`$ ($`6N`$ for the multi-qubit case), whereas the number of parameters required to specify $`|\psi `$ grows exponentially with $`N`$.
More generally, if the Hilbert space is given by $`=\mathrm{I}\mathrm{C}^{n_1}\mathrm{}\mathrm{I}\mathrm{C}^{n_N}`$, then the set of equivalence classes under SLOCC,
$$\frac{\mathrm{I}\mathrm{C}^{n_1}\mathrm{}\mathrm{I}\mathrm{C}^{n_N}}{SL_{n_1}(\mathrm{I}\mathrm{C})\times \mathrm{}\times SL_{n_N}(\mathrm{I}\mathrm{C})},$$
(43)
depends at least on $`2(n_1n_2\mathrm{}n_N1)2_{i=1}^N(n_i^21))`$. This shows that only for $`N=3`$ there are still some systems with (potentially) only a finite number of classe under SLOCC, namely those with Hilbert space $`\mathrm{I}\mathrm{C}^2\mathrm{I}\mathrm{C}^{n_2}\mathrm{I}\mathrm{C}^{n_3}`$, that is, having a qubit as at least one of the subsystems. In all other cases, one finds an infinite number of classes.
We notice that even allowing for non-invertible local operations the amount of parameters that can be changed by local manipulations is typically smaller than that the state depends on. That is, the subset of states that can be reached locally from a given state $`|\psi `$ is of zero measure in the set of states of the multipartite system. Recall that in the bipartite scenario, $`=\mathrm{I}\mathrm{C}^n\mathrm{I}\mathrm{C}^m`$, there is always a maximally entangled state from which all the other states can be locally prepared with certainty of success. We see now that, in constrast, there is typically in a multipartite system no state from which all the others can be prepared, not even with some probability of success. Of course, the parties can always resort to, say, using a sufficient amount of EPR states distributed among them to prepare any multipartite state by standard teleportation. This implies, however, using an initial state (that of many EPR states) which belongs to a Hilbert space much larger than the Hilbert space of the state the parties are trying to create, and thus does not change the previous conclusion.
### B State $`|W`$ in multi-qubit systems
Let us have a look at the generalized form $`|W_N`$ of the state $`|W`$ (2). We define the state
$$|W_N1/\sqrt{N}|N1,1,$$
(44)
where $`|N1,1`$ denotes the totally symmetric state including $`N1`$ zeros and $`1`$ ones. For example, we obtain for $`N=4`$
$$|W_4=1/\sqrt{4}(|0001+|0010+|0100+|1000).$$
(45)
One immediately observes that the entanglement of this state is again very robust against particle losses, i.e. the state $`|W_N`$ remains entangled even if any $`N2`$ parties lose the information about their particle. This means that any two out of $`N`$ parties possess an entangled state, independently of whether the remaining $`(N2)`$ parties decide to cooperate with them or not. This can be seen by computing the reduced density operator $`\rho _{AB}`$ of $`|W_N`$, i.e. by tracing out all but the first and the second systems. By symmetry of the state $`|W_N`$, we have that all reduced density operators $`\rho _{\kappa \mu }`$ are identical and we obtain
$$\rho _{\kappa \mu }=\frac{1}{N}(2|\mathrm{\Psi }^+\mathrm{\Psi }^+|+(N2)|0000|).$$
(46)
The concurrence can easily determined to be
$$𝒞_{\kappa \mu }(W_N)=\frac{2}{N},$$
(47)
which shows that $`\rho _{\kappa \mu }`$ is entangled, even distillable. We conjecture that the average value of the square of the concurrence for $`|W_N`$,
$$\frac{2}{N(N1)}\underset{\kappa }{}\underset{\mu \kappa }{}𝒞_{\kappa \mu }^2(W_N)=\frac{4}{N^2},$$
(48)
is again the maximal value achievable for any state of $`N`$ qubits.
## VI Summary and conclusions
In this work, we investigated equivalence classes of multipartite states specified by stochastic local operations and classical communication. We showed that for pure states of three qubits there are 6 different classes of this kind. In particular, we found that there are two inequivalent types of genuine tripartite entanglement, represented by the GHZ state and the state W. We showed that the state W is the state of three qubits that retains a maximal amount of bipartite entanglement when any one of the three qubits is traced out. For multipartite ($`N4`$) and multilevel systems, we showed that there exist infinitely many inequivalent kinds of entanglement (i.e. classes under SLOCC).
## Acknowledgments
This work was supported by the Austrian Science Foundation under the SFB “control and measurement of coherent quantum systems” (Project 11), the European Community under the TMR network ERB–FMRX–CT96–0087, the European Science Foundation and the Institute for Quantum Information GmbH. G.V also acknowledges a Marie Curie Fellowship HPMF-CT-1999-00200 (European Community).
## Appendix A: SLOCC and local ranks
In this appendix we show that states $`|\psi `$ and $`|\varphi `$ belong to the same class under SLOCC iff they are related by means of a invertible local operator (ILO). From this connection it follows easily that the local ranks of a pure state, r($`\rho _\kappa )`$, $`\kappa =A,B,\mathrm{}`$, are invariant under SLOCC, whereas under LOCC they can only decrease.
Lemma: If the bipartite vectors $`|\psi `$ and $`|\varphi \mathrm{I}\mathrm{C}^n\mathrm{I}\mathrm{C}^m`$ fulfill
$$|\varphi =A1_B|\psi ,$$
(49)
then the ranks of the corresponding reduced density matrices satisfy r$`(\rho _A^\psi )`$ r$`(\rho _A^\varphi )`$ and r$`(\rho _B^\psi )`$r$`(\rho _B^\varphi )`$.
Proof: We consider the Schmidt decomposition of $`|\psi `$,
$$|\psi =\underset{i=1}{\overset{n_\psi }{}}\sqrt{\lambda _i^\psi }|i|i,\lambda _i^\psi >0,n_\psi \mathrm{min}(n,m),$$
(50)
and write the operator $`A`$ as
$$A=\underset{i=1}{\overset{n}{}}|\mu _ii|,$$
(51)
where $`|\mu _i\mathrm{I}\mathrm{C}^n`$ do not need to be normalized nor linearly independent. Then we have that $`\rho _A^\psi =_{i=1}^{n_\psi }|ii|`$ and $`\rho _A^\varphi =A\rho _A^\psi A^{}=_{i=1}^{n_\psi }|\mu _i\mu _i|`$, so that r$`(\rho _A^\varphi )n_\psi `$. The second inequality of the Lemma follows from the fact that for any bipartite vector r$`(\rho _A)=`$ r$`(\rho _B)`$. $`\mathrm{}`$
Corollary: If the vectors $`|\psi ,|\varphi _A_B\mathrm{}_N`$ are connected by a local operator as $`|\varphi =AB\mathrm{}N|\psi `$, then the local ranks satisfy r$`(\rho _\kappa ^\psi )`$ r$`(\rho _\kappa ^\varphi )`$, $`\kappa =A,B,\mathrm{},N`$.
Proof: Indeed, for each of the parties, say Alice for concreteness, we can view the operator $`AB\mathrm{}N`$ as the composition of two local operators, $`A1_{B\mathrm{}N}`$ and $`1_A(B\mathrm{}N)`$, and the Hilbert space as $`_A_{B\mathrm{}N}`$. Then, because of the previous lemma, application of the first operator cannot increase r($`\rho _A`$), and the same happens with the second operator, which cannot increase r$`(\rho _{B\mathrm{}N})`$ \[recall that for any pure state r$`(\rho _A)=`$ r$`(\rho _{B\mathrm{}N})`$\]. $`\mathrm{}`$
Theorem: Two pure states of a multipartite system are equivalent under SLOCC iff they are related by a local invertible operator.
Proof: If
$$|\varphi =AB\mathrm{}N|\psi ,$$
(52)
then a local protocol exists for the parties to transform $`|\psi `$ into $`|\varphi `$ with a finite probability of success. Indeed, each party needs simply perform a local POVM including a normalized version of the corresponding local operator in (52). For instance, Alice has to apply a POVM defined by operators $`\sqrt{p_A}A`$ and $`\sqrt{1_Ap_AA^{}A}`$, where $`p_A1`$ is a positive weight such that $`p_AA^{}A1_A`$, and similarly for the rest of the parties. Then such a local protocol converts $`|\psi `$ succesfully into $`|\varphi `$ with probability $`p_Ap_B\mathrm{}p_N`$. If, in addition, $`A,B,\mathrm{},N`$ are invertible operators, then obviously
$$|\psi =A^1B^1\mathrm{}N^1|\varphi $$
(53)
and the conversion can be reversed locally. Let us then move to prove the converse. We already argued (section II.A) that if $`|\psi `$ can be converted into $`|\varphi `$ by LOCC, then a local operator relate them. We want to prove now that equivalence of $`|\psi `$ and $`|\varphi `$ under SLOCC implies that this operator can always be chosen to be invertible. For simplicity, we will assume that $`|\psi `$ and $`|\varphi `$ are related by a local operator acting non-trivially only in Alice’s part,
$$|\varphi =A1_{B\mathrm{}N}|\psi .$$
(54)
\[The general case would correspond to composing operator $`A1_{B\mathrm{}N}`$ with operator $`1_AB1_{C\mathrm{}N}`$, and similarly for the rest of the parties. The following argumentation should then be applied sequentially to each party individually.\] We can then consider the Schmidt decomposition of the states with respect to part $`A`$ and part $`B\mathrm{}N`$
$`|\psi ={\displaystyle \underset{i=1}{\overset{n_\psi }{}}}\sqrt{\lambda _i^\psi }|i|\tau _i,\lambda _i^\psi >0`$ (55)
$`|\varphi ={\displaystyle \underset{i=1}{\overset{n_\varphi }{}}}\sqrt{\lambda _i^\varphi }(U_A|i)|\tau _i,\lambda _i^\varphi >0`$ (56)
where the local unitary $`U_A`$ relate the two local Schmidt basis in Alice’s part, $`\{|i\}_{i=1}^n_A=\mathrm{I}\mathrm{C}^n`$, $`|\tau _i_B\mathrm{}_N`$, and $`n_\psi =n_\varphi `$ because of the previous corollary. Now, operator $`A`$ in equation (54) must be of the form (up to some irrelevant permutations in the Schmidt basis)
$`A=U_A(A_1+A_2)`$ (57)
$`A_1{\displaystyle \underset{i=1}{\overset{n_\psi }{}}}\sqrt{{\displaystyle \frac{\lambda _i^\varphi }{\lambda _i^\psi }}}|ii|,`$ (58)
$`A_2{\displaystyle \underset{i=n_\psi +1}{\overset{n}{}}}|\mu _ii|`$ (59)
where $`|\mu _i`$ are arbitrary unnormalized vectors. Notice that vectors $`|\mu _i`$ play no role in equation (54) since $`A_21_{B\mathrm{}N}|\psi =0`$. Therefore we can redefine
$$A_2\underset{i=n_\psi +1}{\overset{n}{}}|ii|,$$
(60)
which implies that $`A`$ is an invertible operator.$`\mathrm{}`$
## Appendix B: $`\tau `$ is an entanglement monotone
In this appendix, we show that the 3-tangle $`\tau `$ is an entanglement monotone, i.e. decreasing on average under LOCC in all the three parties. We first note that any local protocol can be decomposed into POVM’s such that only one party performs operations on the system. This, together with the invariance of the 3-tangle $`\tau `$ under permutations of the parties, ensures that it is sufficient to consider a local POVM in $`A`$ only. Furthermore, we can restrict ourselves to two–outcome POVM’s due to the fact that a genarlized (local) POVM can be implemented by a sequence of two outcome POVM’s. Let $`A_1,A_2`$ be the two POVM elements such that $`A_1^{}A_1+A_2^{}A_2=1𝐥`$. We can write $`A_i=U_iD_iV`$, where $`U_i`$, $`V`$ are unitary matrices and $`D_i`$ are diagonal matrices with entries $`(a,b)`$ $`[((1a^2)^{\frac{1}{2}},(1b^2)^{\frac{1}{2}}`$)\] respectively. Note that we used the singular value decomposition for $`A_i`$, and we have that the restriction that $`A_1,A_2`$ constitute a POVM immediately implies that the unitary operation $`V`$ can be chosen to be the same in both cases. We consider an initial state $`|\psi `$ with 3-tangle $`\tau (\psi )`$. Let $`|\stackrel{~}{\varphi }_i=A_i|\psi `$ be the (unnormalized) states after the application of the POVM. Normalizing them, we obtain $`|\varphi _i=|\stackrel{~}{\varphi }_i/\sqrt{p_i}`$ with $`p_i=\stackrel{~}{\varphi }_i|\stackrel{~}{\varphi }_i`$ and $`p_1+p_2=1`$. We want to show that $`\tau ^\eta `$, $`0<\eta 1`$ is, on average, always decreasing and thus an entanglement monotone, i.e for
$$<\tau ^\eta >=p_1\tau ^\eta (\varphi _1)+p_2\tau ^\eta (\varphi _2)$$
(61)
we have that
$$<\tau ^\eta >\tau ^\eta (\psi )$$
(62)
is fulfilled for all possible choices of the POVM $`\{A_1,A_2\}`$. Using that $`\tau `$ is invariant under local unitaries, we do not have to consider the unitary operations $`U_i`$ in our calculations, i.e. $`\tau (U_iD_iV\psi )=\tau (D_iV\psi )`$. Taking this simplification into account, a straightforward calculation shows that
$$\tau (\varphi _1)=\frac{a^2b^2}{p_1^2}\tau (\psi ),\text{ }\tau (\varphi _2)=\frac{(1a^2)(1b^2)}{p_2^2}\tau (\psi ),$$
(63)
where we used that $`\tau (ϵ\stackrel{~}{\varphi _i})=ϵ^4\tau (\stackrel{~}{\varphi _i})`$, which can be checked by noting that $`\tau `$ is a quartic function with respect to its coefficients in the standard basis. Note that the dependence of $`\tau (\varphi _i)`$ on the unitary operation $`V`$ is hidden in $`p_i`$. For $`\eta =1/2`$, one obtains for example $`\tau ^{\frac{1}{2}}(\varphi _1)=ab/p_1\tau ^{\frac{1}{2}}(\psi )`$. Substituting in (61), we find
$$<\tau ^{\frac{1}{2}}>=(ab+\sqrt{(1a^2)(1b^2)})\tau ^{\frac{1}{2}}(\psi ).$$
(64)
In this case, one can easily check that ($`\text{64})\tau ^{\frac{1}{2}}`$ by noting that (64) is maximized for $`a=b`$. We thus have that $`\tau ^{\frac{1}{2}}`$ is, on average, always decreasing and thus an entanglement monotone. In a similar way, one can check for $`0<\eta 1`$ that $`\tau ^\eta `$ is an entanglement monotone. However, for $`\eta 1/2`$, the derivation is a bit more involved due to the fact that in this case the propabilities $`p_i`$ in the expression for $`<\tau ^\eta >`$ do no longer cancel and have to be calculated explicitly.
## Appendix C: $`|W`$ maximizes residual bipartite entanglement
Here we show that for all tripartite pure states, except the state $`|W`$ the following inequality holds
$$E_\tau 𝒞_{AB}^2+𝒞_{AC}^2+𝒞_{BC}^2<\frac{4}{3},$$
(65)
while the state $`|W`$ reaches the value $`E_\tau =4/3`$. Note that we used the shorthand notation $`𝒞_{AB}`$ for the concurrence of the reduced density operator $`\rho _{AB},𝒞(\rho _{AB})`$, and similary for $`𝒞_{AC}`$,$`𝒞_{BC}`$.
Inequality (65) already implies that the state $`|W`$ reaches the maximum average value $`\overline{}(\psi )`$ of Equ. (39) for the choice of $`(\rho )=𝒞(\rho )^2`$, namely $`\overline{}(W)=4/9`$.
At the same time, inequality (65) also shows that the state $`|W`$ maximizes the function $`_{\mathrm{min}}(\psi )`$ (40) for the choice of $`(\rho )=𝒞(\rho )^2`$, since (65) implies that
$$𝒞_{\mathrm{min}}^2(\psi )\mathrm{min}(𝒞_{AB}^2,𝒞_{AC}^2,𝒞_{BC}^2)<4/9$$
(66)
for all states except the state $`|W`$, for which the value $`4/9`$ is reached. From (66) follows that for any bipartite measure of entanglement $`(\rho )`$ which is monotonically increasing with the square of the concurrence (and hence with the concurrence itself), the state $`|W`$ maximizes the function $`_{\mathrm{min}}(\psi )`$ (40), i.e.
$$_{\mathrm{min}}(\psi )<_{\mathrm{min}}(W)=(𝒞^2=4/9).$$
(67)
Assume that this is not the case, i.e. there exist a state $`\psi `$ for which $`_{\mathrm{min}}(\psi )>_{\mathrm{min}}(W)`$. Since by assumption $``$ is monotonically increasing with the concurrence, this would imply that $`𝒞_{\mathrm{min}}^2(\psi )>4/9`$, which contradicts Equ. (66) and is hence impossible.
Note in addition that any good measure of entanglement should be a convex function , as $`𝒞(\rho ),E_f(\rho )`$ and $`E_2(\rho )`$ are. This implies, when applied to (39) and (40) that the optimal values for $`\overline{}`$ and $`_{\mathrm{min}}`$ are achieved for pure states.
Ther remainder of this appendix is devoted to prove inequality (65). Using the definition of the 3-tangle, $`\tau \tau _{ABC}=𝒞_{A(BC)}^2𝒞_{AB}^2𝒞_{AC}^2`$ and the invariance of the 3-tangle under permutations of the parties, we can rewrite $`E_\tau `$ as $`1/2(𝒞_{A(BC)}^2+𝒞_{B(AC)}^2+𝒞_{C(AB)}^23\tau )`$. Using that $`𝒞_{\kappa (\mu \nu )}^2=4\mathrm{d}\mathrm{e}\mathrm{t}\rho _\kappa `$, we can evaluate $`E_\tau `$ for the different classes.
Starting with the class $`ABC`$, we immeadetly obtain that $`E_\tau (\mathrm{\Psi }_{ABC})=0`$. For the class $`ABC`$, we have that $`\tau =0`$ and $`𝒞_{A(BC)}^2=0`$. Since $`𝒞_{B(AC)}^2,𝒞_{C(AB)}^21`$, we have that $`E_\tau (\mathrm{\Psi }_{ABC})1`$ in this case (and similary for the classes $`BAC,CAB`$).
Now we consider the class $`W`$, specified by equ. (29). Again, we have that $`\tau =0`$. We find that $`E_\tau (\mathrm{\Psi }_W)=4(ab+ac+bc)`$ (which does not depend on $`d`$). Notice that $`E_\tau `$ is maximized for $`a=b=c=1/3`$ \- which corresponds to the state $`|W`$ \- and leads to $`E_\tau =4/3`$. For all other values of $`a,b,c,d`$, we have that $`E_\tau <4/3`$.
Let us now turn to the class GHZ, specified in eq. (15). Using that $`\tau (\mathrm{\Psi }_{GHZ})`$ is given in eq. (37) and $`det\rho _A=K^2c_\delta ^2s_\delta ^2s_\alpha ^2(1c_\beta ^2c_\gamma ^2)`$ (and similary for $`det\rho _{B,C}`$), we obtain
$$E_\tau =\frac{4c_\delta ^2s_\delta ^2[(s_\alpha ^2s_\beta ^2+s_\alpha ^2s_\gamma ^2+s_\beta ^2s_\gamma ^2)3s_\alpha ^2s_\beta ^2s_\gamma ^2]}{(1+2c_\delta s_\delta c_\alpha c_\beta c_\gamma c_\phi )^2}$$
(68)
One readily checks that (68) is maximized for $`\delta =\pi /4`$ and $`\phi =\pi `$ (which corresponds to $`c_\delta =s_\delta =1/\sqrt{2}`$ and $`c_\phi =1`$), independent of the values of $`\alpha ,\beta ,\gamma (0,\pi /2]`$. Thus we have that $`E_\tau E_\tau (\delta =\pi /4,\phi =\pi )`$ and after some algebra we obtain
$$E_\tau \frac{(c_\alpha ^2+c_\beta ^2+c_\gamma ^2)2(c_\alpha ^2c_\beta ^2+c_\alpha ^2c_\gamma ^2+c_\beta ^2c_\gamma ^2)+3c_\alpha ^2c_\beta ^2c_\gamma ^2}{(1+c_\alpha c_\beta c_\gamma )^2}$$
(69)
We want to show that the (rhs) of eq. 69 $`<4/3`$. Let us call $`xc_\alpha ,yc_\beta ,zc_\gamma `$ with $`0x,y,z<1`$. We thus have to show that
$`f(x,y,z)`$ $`3(x^2+y^2+z^2)6(x^2y^2+x^2z^2+y^2z^2)`$ (70)
$`+`$ $`5(x^2y^2z^2)4+8xyz<0`$ (71)
Let us calculate the maximum of $`f(x,y,z)`$. We therefore take the derivatives of $`f(x,y,z)`$ with respect to $`x,y,z`$ respectively (which we denote by $`f_x,f_y,f_z`$) and set them to zero. One immeadetly observes (by considering linear combination of the resulting equations, e.g. $`xf_xyf_y`$, where one e.g. obtains $`(x^2y^2)(12z^2)=0`$), that for a maximum we must have $`x=y=z`$. The possible solutions of the resulting polynomial of degree 5 can be checked to lie outside the intervall $`[0,1)`$, i.e. outside the range of $`x,y,z`$ except for $`x=y=z=0`$. It can however be easily verified that this solution give rise to a minimum of $`f(x,y,z)`$, namely $`f(0,0,0)=4`$. Thus the maximum of $`f(x,y,z)`$ is obtained at the border of the range for $`x,y,z`$, which corresponds to the surfaces of a cube. Due to the fact that $`f(x,y,z)`$ is invariant under permutations of the variables, we only have to check two of the surfaces, e.g. the surfaces specified by $`x=0`$ and $`x=1`$ (actually $`x=1ϵ`$, where $`ϵ`$ is an infinitesimally small positive number) and we find (i) $`f(0,x,y)=3(y^2+z^2)6y^2z^241`$ (the maximum in this case is e.g. obtained for $`y=0,z=1ϵ`$)) and (ii) $`f(1,y,z)=8yz3(y^2+z^2)y^2z^21<0`$. In (ii), it can be checked that a necessary condition for a maximum is $`y=z`$ and that $`f(1,y,y)`$ is monotonically increasing in $`[0,1)`$ and is thus maximized for $`y=z=(1ϵ)`$. One obtains $`f(x,y,z)f(1,1ϵ,1ϵ)<0`$ as desired.
So we managed to show that the state $`|W`$ is the only state which fulfills $`E_\tau =4/3`$, and for all other tripartite pure states we have that $`E_\tau <4/3`$.
|
warning/0005/hep-ph0005238.html
|
ar5iv
|
text
|
# May 2000 UM-P-00/021 RCHEP-00/003 Discovering mirror particles at the Large Hadron Collider and the implied cold universe
## I Introduction
The Mirror Matter or Exact Parity Model (EPM) sees every standard particle paired with a parity partner. This idea was first mentioned by Lee and Yang in their seminal paper on parity violation as a way to retain the full Poincaré Group as a symmetry of nature despite the $`VA`$ character of weak interactions. Some follow up work was performed in the ensuing decades on some aspects of the mirror matter hypothesis . In 1991, the idea was independently rediscovered and the full gauge theory constructed for the first time . Shortly thereafter the EPM was extended to include nonzero neutrino masses and mixings and applied to the solar, atmospheric and LSND anomalies . The EPM can also alter standard Big Bang cosmology in interesting ways, through the possible identification of some dark matter with mirror matter, and through modifications of Big Bang Nucleosynthesis (BBN) .
The ordinary and mirror particle sectors can interact in a number of ways. The first is through gravitation, with immediate consequences for the dark matter problem and astrophysics. Non-gravitational interactions can be induced through the mixing of colourless and neutral particles with their mirror counterparts. Neutrinos, the photon, the $`Z`$ boson and the physical neutral Higgs boson can mix with the corresponding mirror states. Coloured and/or electrically charged particles are prevented from mixing with their mirror analogues by colour and electric charge conservation laws.
The purpose of this paper is to study the Higgs boson sector of the EPM. It has been previously noted that the mass eigenstate physical Higgs bosons must be maximal mixtures of the underlying ordinary and mirror states because of the unbroken parity symmetry . Each mass eigenstate will therefore decay $`50\%`$ of the time into invisible mirror particles, providing a striking experimental signature in principle. The production cross-section for such Higgs bosons would be 1/2 of that of the standard Higgs boson of the same mass. This is a very simple and important observation, because it provides a clear way to experimentally establish the existence of the mirror world <sup>*</sup><sup>*</sup>*From the recent results of the L3 Collaboration we can establish a lower bound of about 65 GeV for a Higgs boson with these properties.. In recent years there has been a strong focus on using the neutrino anomalies as a way to discover mirror matter . Neutrino oscillation physics certainly does provide a very interesting way to garner experimental evidence for mirror matter, or to at least constrain the model (if one is being pessimistic). However, the terrestrial neutrino phenomenology of the EPM is similar to that of pseudo-Dirac neutrinos , so complementary information would be useful. The Higgs boson sector is one potentially important way to obtain this information. (The mixing of ortho-positronium with mirror-ortho-positronium is another .)
The strength of Higgs boson mixing with its mirror partner is controlled by an a priori independent dimensionless parameter $`\lambda _{HH^{}}`$. The mass splitting between the mass eigenstate Higgs bosons is proportional to this same parameter. Standard cosmology, through BBN, can be used to constrain $`\lambda _{HH^{}}`$ and hence the Higgs boson mass splitting also . In this paper, we will demonstrate that the temperature of the radiation dominated (RD) phase of the universe should never have exceeded a few tens of GeV if the mass splitting is to be substantial (of order 1 GeV). Chaotic inflation with very inefficient reheating is an example of how such a cold cosmology could arise. Remarkably, the Large Hadron Collider (LHC) could thus discover the mirror world as a byproduct of its Higgs boson search programme, and simultaneously establish an upper bound on the temperature of the RD phase of the universe.
## II The Higgs boson sector and cosmological constraints
Consider a minimal Higgs boson sector for the EPM. It contains the standard Higgs doublet $`\varphi `$, transforming as a $`\mathrm{𝟐}(1)`$ representation under the electroweak gauge group SU(2)$``$U(1)<sub>Y</sub>. It also contains a mirror Higgs doublet $`\varphi ^{}`$ which transforms as a $`\mathrm{𝟐}(1)`$ representation under the mirror electroweak gauge group SU(2)$`{}_{}{}^{}`$U(1)$`{}_{Y}{}^{}{}_{}{}^{}`$. The standard doublet $`\varphi `$ is a singlet under the mirror gauge group, while $`\varphi ^{}`$ is a singlet under the ordinary gauge group. Under the discrete parity symemtry, $`\varphi \varphi ^{}`$.
We focus on the Higgs potential in this paper. It is very simply given by
$$V=\lambda _+(\varphi ^{}\varphi +\varphi ^{}\varphi ^{}2v^2)^2+\lambda _{}(\varphi ^{}\varphi \varphi ^{}\varphi ^{})^2.$$
(1)
In the $`\lambda _\pm >0`$ region of parameter space, the vacuum is clearly given by
$$\varphi =\varphi ^{}=\left(\begin{array}{c}0\\ v\end{array}\right).$$
(2)
In this region of parameter space, the parity or mirror symmetry is respected by the vacuum because of the equality between the vacuum expectation values (VEVs) of the ordinary and mirror Higgs doublets.A parity breaking global minimum of this Higgs potential can be found in another region of parameter space . Going to unitary gauge and shifting the neutral components as per $`\varphi ^0=v+H/\sqrt{2}`$ and $`\varphi ^0=v+H^{}/\sqrt{2}`$ we see from Eq.(1) that the mass eigenstates are
$$H_\pm =\frac{H\pm H^{}}{\sqrt{2}}$$
(3)
with masses given by
$$m_+^2=8\lambda _+v^2\text{and}m_{}^2=8\lambda _{}v^2$$
(4)
respectively. We therefore see that the mass splitting
$$\mathrm{\Delta }mm_+m_{}=(\lambda _+\lambda _{})\frac{8v^2}{m_++m_{}}$$
(5)
is controlled by the parameter
$$\lambda _{HH^{}}\lambda _+\lambda _{}.$$
(6)
From Eq.(1) we also see that the coefficient of the $`HH^{}`$ mixing term, $`4\lambda _{HH^{}}v^2`$, is proportional to the same parameter. In addition, the coefficient of the quartic coupling term $`\varphi ^{}\varphi \varphi ^{}\varphi ^{}`$ is $`2\lambda _{HH^{}}`$.
It is clear that each mass eigenstate physical neutral Higgs boson $`H_\pm `$ decays $`50\%`$ of the time into ordinary particles and $`50\%`$ of the time into mirror, and hence invisible, particles. The total decay rate of $`H_+`$ or $`H_{}`$ is the same as that for a SM physical neutral Higgs boson of the same mass. Note also that each mass eigenstate couples to ordinary particles with strength reduced by $`1/\sqrt{2}`$ compared to the coupling of the standard Higgs boson to those same particles .
We now turn to cosmological constraints from BBN on $`\lambda _{HH^{}}`$, or equivalently, $`\mathrm{\Delta }m`$. BBN does not allow the mirror plasma to be in thermal equilibrium with the ordinary plasma during the relevant epoch. The parameters controlling the mixing of colourless and neutral particles with their mirror partners must therefore obey upper bounds, assuming that the standard theory of BBN is correct. The derivations of these bounds for the photon and neutrino systems have been described elsewhere .
The Higgs system situation was briefly discussed in Ref. and will be fully explored here. There are two different epochs to consider: (i) temperatures $`T\stackrel{>}{}100`$’s of GeV, where the Higgs bosons exist as real particles in the plasma, and (ii) the opposite limit where they do not.
Epoch (i) was considered in Ref.. The physics is very simple. Suppose that no mirror particles exist in the plasma to begin with. We then have to ensure that processes driven by $`\lambda _{HH^{}}`$ do not bring the mirror Higgs bosons, and hence all other mirror particles, into thermal equilibrium. During epoch (i), the electroweak symmetry is presumably restored, so the relevant term is $`2\lambda _{HH^{}}\varphi ^{}\varphi \varphi ^{}\varphi ^{}`$ from the Higgs potential. By dimensional analysis, the rate for $`\varphi \varphi \varphi ^{}\varphi ^{}`$ scattering will be approximately given by $`(\lambda _{HH^{}})^2T`$. Requiring that this be less than the expansion rate of the universe $`\sqrt{g}T^2/m_P`$, where $`g100`$ is the effective number of massless degrees of freedom and $`m_P`$ is the Planck mass, we find the bound
$$\lambda _{HH^{}}\stackrel{<}{}10^8\sqrt{\frac{m_\varphi }{100\text{GeV}}}.$$
(7)
This bound is obtained by setting the temperature $`T`$ to be about $`m_\varphi `$ in order to get the most restrictive condition, where $`m_\varphi 100`$’s of GeV is the Higgs mass in the symmetric phase. This is clearly a severe bound. If the assumptions behind its derivation were unassailable, then the EPM would have a significant fine-tuning problem: why is $`\lambda _{HH^{}}`$ so small? It has been observed that the supersymmetric extension of the EPM yields $`\lambda _{HH^{}}=0`$ . While this is of interest, we will look for an alternative solution, because $`\lambda _{HH^{}}=0`$ eliminates the chance for the LHC to discover the mirror world. For a sufficiently small $`\mathrm{\Delta }m`$, the ordinary particles from which the Higgs bosons are produced yield a coherent superposition of $`H_+`$ and $`H_{}`$ which is precisely the ordinary Higgs boson $`H`$. The Higgs physics of the EPM is then indistinguishable from that of the Standard Model. According to Ref., the LHC is expected to measure the Higgs boson mass with an accuracy of roughly $`1\%`$. This means that $`\lambda _{HH^{}}`$ must be larger than about $`0.01`$ for the mass difference between $`H_+`$ and $`H_{}`$ to be observable.
So, let us suppose that the radiation dominated phase of the universe was never hot enough for Higgs bosons to exist as real particles in the plasma! The bound of Eq.(7) is then irrelevant. Such a “cold universe” can be produced, for example, by inefficient reheating after inflation. We will discuss this further in the next section.
For $`Tm_\varphi `$, mirror particles can be brought into thermal equilibrium via the $`f\overline{f}f^{}\overline{f^{}}`$ process mediated by virtual Higgs boson exchange as depicted in the Figure. We will also take $`T`$ to be less than the temperature for the electroweak phase transition (about 100 GeV), so we are in the broken phase. The rate is given roughly by
$$\mathrm{\Gamma }h_f^4\lambda _{HH^{}}^2\frac{v^4T^5}{m_\varphi ^8},$$
(8)
where $`h_f=m_f/v`$ is the Yukawa coupling constant for the fermions and mirror fermions in the initial and final states. For $`T100`$ GeV, top quarks are not a significant component of the plasma, so the bottom quark $`b\overline{b}b^{}\overline{b^{}}`$ process will dominate all others. We therefore set $`h_f=h_b=m_b/v`$. The condition that $`\mathrm{\Gamma }`$ is always less than the expansion rate then implies that
$$\lambda _{HH^{}}\stackrel{<}{}\frac{m_\varphi ^4g^{1/4}}{m_b^2\sqrt{m_PT^3}}0.1\left(\frac{T_{\text{max}}}{\text{GeV}}\right)^{\frac{3}{2}}\left(\frac{m_\varphi }{200\text{GeV}}\right)^4.$$
(9)
which is most restrictive for the highest temperature $`T_{\text{max}}`$ we hypothesise the radiation dominated phase of the universe to reach. If we require that the Higgs boson mass difference be observable at the LHC ($`\lambda _{HH^{}}\stackrel{>}{}0.01`$) then $`T_{\text{max}}`$ cannot be higher than a few tens of GeV. Furthermore, if $`T_{\text{max}}`$ does not exceed a few GeV then $`\lambda _{HH^{}}1`$ is allowed If $`T_{\text{max}}`$ is below $`m_b4.4`$ GeV, then charmed quarks and tau leptons should be used instead of bottom quarks..
The cold universe hypothesis simultaneously remedies the $`\lambda _{HH^{}}`$ fine-tuning problem, and allows a large mass splitting between $`H_+`$ and $`H_{}`$. Remarkably, the LHC could simultaneously discover the mirror world and produce strong evidence that the RD phase of the universe was never hotter than a few tens of GeV!
## III Cold cosmology?
We will now make a few remarks about how such a cold cosmology could be constructed. As well as providing a low $`T_{\text{max}}`$, the cosmological model would also have to explain why the early universe was predominantly composed of ordinary matter in the first place. A universe with an ordinary and mirror plasma in thermal equilibrium with each other is ruled out by BBN. Phrased another way, we must require that the temperature $`T^{}`$ of the mirror plasma be less than about half of the temperature of the ordinary plasma during the BBN epoch in order for the expansion rate of the universe to not be too high.
The $`T^{}`$ issue has already been addressed in the literature through inflationary models . As an example, Ref. introduces an inflaton $`\sigma `$ and a mirror inflaton $`\sigma ^{}`$ with the potential
$$U=\frac{1}{2}m_\sigma ^2(\sigma ^2+\sigma ^2)$$
(10)
in the context of the chaotic inflationary paradigm (see, e.g. ). They then suppose that the chaotic initial conditions set up $`\sigma ^{}\sigma `$ by chance.<sup>§</sup><sup>§</sup>§This important idea illustrates how a cosmology which is asymmetric between the ordinary and mirror sectors can arise despite the identical microphysics: exploit fluctuations. The equations of motion derived from $`U`$ then show that $`\sigma ^{}/\sigma `$ remains constant during the inflationary phase. This means that the $`\sigma ^{}`$ field will begin oscillating about $`\sigma ^{}=0`$ while $`\sigma `$ is still driving inflation. Assuming further that $`\sigma ^{}`$ $`(\sigma )`$ couples only to mirror (ordinary) particles, the reheated mirror plasma created by the decays of $`\sigma ^{}`$ gets diluted by the inflationary expansion that is still occurring. When $`\sigma `$ subsequently ceases driving inflation, it then produces a reheated ordinary plasma that has a much higher temperature than the diluted mirror plasma. All we need to further postulate is very inefficient reheating to produce the required cold universe. One will ultimately also need a baryogenesis mechanism that can work at such low temperatures. Some proposals already exist in the literature, including the exploitation of the pre-heating process for this purpose .
The above is but an example of how a cold cosmology with asymmetric temperatures for the ordinary and mirror plasmas might arise. Many other issues need to be addressed, including the origin of the inflaton potential, the precise mechanism of reheating, and whether a substantial (but still subdominant) amount of mirror matter can be produced in addition to the ordinary matter during re- or pre-heating (as would be needed for mirror dark matter purposes). For the moment, our focus should be on the interesting and simple Higgs physics of the EPM. If the LHC discovers a large mass splitting between $`H_+`$ and $`H_{}`$, then this terrestrially obtained data will provide good motivation for further work in cosmological model building.
## IV Conclusions
The Exact Parity or Mirror Model predicts some simple and interesting Higgs physics. There will be two physical neutral Higgs boson mass eigenstates, each with a $`50\%`$ invisible width. This would be a remarkable way to discover mirror particles. A detectable mass splitting between the two eigenstates would strongly suggest that the radiation dominated phase of the universe was never hotter than, say, a few tens of GeV. This would in turn be interesting information for cosmological model builders.
###### Acknowledgements.
We would like to thank Nicole Bell and Sergei Gninenko for discussions and Robert Foot for comments on a draft manuscript. A.I. is grateful to D.Grigoriev and M.Shaposhnikov for helpful discussions. This work was supported by the Australian Research Council.
FIGURE
Diagram of the process $`f\overline{f}f^{}\overline{f^{}}`$ mediated by Higgs–mirror-Higgs boson mixing.
|
warning/0005/cond-mat0005501.html
|
ar5iv
|
text
|
# A new method for a local study of nonlinear microwave properties of superconductors
## Abstract
We report a set of experimental data on local third-harmonic generation at microwave frequencies (0.5 GHz) in $`YBa_2Cu_3O_7`$ and Nb films. For local investigations of the nonlinear response a probe with inductive coupling was elaborated. The map of the nonlinear microwave response of a $`YBa_2Cu_3O_7`$ thin film is plotted below $`T_c`$ with high resolution. The third-harmonic power is measured as a function of temperature, input power and dc magnetic field at some areas of the film. The correlation between the depinning current density $`J_p`$ and the nonlinear microwave response is also demonstrated.
HTS films find extensive application in passive microwave devices such as transmission lines, antennas and filters , owing to a low surface resistance $`R_s`$ of HTS films, which ensures low losses at low power levels. At higher power surface resistance $`R_s`$ increases, due to nonlinear properties of these films which leads to higher losses and a shift of the resonance frequency, thus limiting the applicability range for these films. The nonlinearity of surface resistance $`R_s`$ is generally associated with the Ginzburg-Landau nonlinearity , thermal nonlinearity , hysteretic losses , and the Josephson nonlinearity. Yet, despite the numerous experiments carried out in this field, the origin of a nonlinear microwave response of superconductor has not been fully understood thus far. Therefore, investigation of nonlinearity is equally important in terms of gaining an insight into the fundamental properties of HTS films and towards applications in superconducting electronics.
In this work we propose a new technique for measuring a nonlinear local microwave response of a superconducting film basing on developed a near-field inductive coupling probe. Using this method, we have plotted a map of the nonlinear local response of a HTS film at a temperature below $`T_c`$, and also measured the third-harmonic power as a function of input power, temperature, and applied dc magnetic field for HTS and Nb films.
A wide use in measurements of the nonlinear properties of superconducting films lately has been made of a resonator technique which allows to produce rf magnetic fields that are close in value to the characteristic rf magnetic fields of nonlinearity. This is achieved by using a stripline resonator or a cavity resonator with a sample placed inside. Characteristic values for the current density or the magnetic field of film nonlinearity are determined either from the power dependence of the third-harmonic power or from measurements of surface impedance which is related to microwave losses . Note that the averaged nonlinear characteristics of a microwave device are measured in this case. Local measurements of surface resistance $`R_s`$ in the microwave range are aided by near-field microscopes in wide use currently. Essentially, the idea of near-field microscopy is in localization of a magnetic or an electric field near probe on scales much less than a wavelength. A variety of probe designs conventionally used for local investigations includes a circular aperture, open-ended coaxial cable, small loop, etc.
Here we present an original near-field probe used for measuring a local nonlinear response of superconducting films, that has been designed with due regard for the earlier developed methods and approaches. A block diagram of the probe is shown in Fig.1. The probe is essentially a 2 mm long 50 $`\mu `$m diameter wire connecting the outer and the inner conductors of a coaxial cable. Reflection of a microwave signal from such a probe gives rise to a high current flow in the wire because the probe impedance is much less than the wave impedance of the coaxial cable. The current induces a fairly strong quasistatic magnetic field localized on a scale of order of the probe diameter. The nonlinear properties of a superconducting film are responsible for generation of higher harmonics which are picked up by the same probe. The incident wave frequency is 472 MHz, and the nonlinear response is measured at the third harmonic frequency of 1.42 GHz. To avoid contact effects preventing observation of a superconducting film nonlinearity, a 10 $`\mu `$m thick teflon film is placed between the probe and the sample. Note that for the dimensions and the geometry of the probe studied here the maximum of power of about 100 mW produces a maximum current density $`J_{rf}`$ in the 100 nm film of about $`10^6A/cm^2`$.
In this work we did an experimental study of 30-100 nm thick $`YBa_2Cu_3O_7`$ films magnetron sputtered on a $`GaNdO_3`$ substrate. The films quality was quite high (critical current density of $`10^6A/cm^2`$ ). We also investigated Nb films of 30 nm thickness.
Fig.2 shows a temperature dependence of the third-harmonic signal at different levels of input power, which features a nonlinearity peak below $`T_c`$. It should be noted that nonlinearity maxima near $`T_c`$ were observed in a number of works . In the nonlinearity peak was shown to appear by penetration of a magnetic flux through the film edges. In our case the probe was placed in a film center, but a sharpest peak remained. For a qualitative analysis of the temperature dependence of nonlinear response, we used the measurements of the temperature dependences of the depinning current density $`J_p`$, found from measurements of the residual magnetization produced in a film by an external uniform magnetic field, the current density of vortex penetration $`J_c`$ (which corresponds to the Ginzburg-Landau pair-breaking current density $`J_{GL}`$ for perfect superconductor ) and the resistivity $`\rho `$, kindly provided by the authors of . By comparing these dependences we found out that the temperature of the peak $`T_{max}`$, correlates with that at which the depinning current density $`J_p`$ disappears, and the temperature of nonlinearity vanishing corresponds to the off-set temperature $`\rho (T)`$. The correlation between the nonlinear microwave properties and the depinning current density indicates that the nonlinearity observed at temperatures close to $`T_c`$ is of a vortex origin.
In Fig.3 the third-harmonic power $`P_{3\omega }(P_\omega `$) is shown as a function of input power on a log-log scale for $`YBa_2Cu_3O_7`$ and Nb films. The data are readily approximated by the power law $`P_{3\omega }AP_{\omega }^{}{}_{}{}^{n}`$ . At temperatures close to $`T_c`$ the HTS films exhibit a deviation from the exponent n = 3 (which is characteristic of an ordinary cubic nonlinearity described by the Ginzburg-Landau equations), while Nb films feature a marked power threshold. The exponent deviation from n = 3 for HTS films occurs through saturation of the power dependence of the third-harmonic signal at high input powers or at temperatures close to $`T_c`$ .
The third-harmonic power $`P_{3\omega }(H_{dc})`$ as a function of a dc magnetic field $`H_{dc}`$ at a temperature near the nonlinearity peak and at T = 77 K is shown in Fig.4. The behavior of $`P_{3\omega }(H_{dc})`$ differs qualitatively at liquid nitrogen temperature and temperatures close to $`T_c`$. At T = 77 K there is a rise in the third-harmonic value and a fairly strong hysteresis is observed, whereas in the vicinity of $`T_c`$ irreversibility disappears and an increase in the field causes suppression of nonlinearity. A strong dependence on an external magnetic field is also evidence of the vortex origin of the observed nonlinearity; it may be connected with a decrease in the depinning current density $`J_p`$ at higher temperatures (Fig.2).
The method developed was used to map a nonlinear local microwave response for $`YBa_2Cu_3O_7`$ at liquid nitrogen temperature, as shown in Fig.5. We have chosen a positioning system such that it would allow the probe to be moved at a 125 $`\mu `$m step in the direction of the x and y axes. Fig.5 demonstrates a nonuniform distribution of the nonlinear response across the film surface, which depends on the inhomogeneity of the critical current density in the sample. Note also that nonlinearity increases in areas lying closer to the sample edge, when the probe is parallel to the film boundary. This effect can be explained by an increasing density of the current excited in the film.
Although a complete quantitative analysis of the experimental data is impossible currently, some qualitative considerations seem plausible enough. Estimated value of the highest current density $`J_{rf}10^510^6A/cm^2`$ in the film is higher or of the order of the current density of vortex penetration $`J_c`$ and it is naturally to consider that the nonlinear response is due to creation of vortices by microwave field. At the same time the relation between the temperature dependences of the nonlinearity and the depinning current density $`J_p`$ (Fig.2), and disappearance of irreversibility in the dc magnetic field dependence of the third-harmonic power (Fig.4) demonstrate the substantial role of thermal fluctuations in the vortex response. The nonlinear response of low-temperature superconductors, unlike in the HTS case, demonstrates a power threshold likely to be related to the onset of vortex creation by the microwave field.
In summary, a new method for local investigation of the nonlinear microwave properties of superconducting films has been developed and used for mapping of a nonlinear response from a HTS film at liquid nitrogen temperature. The third-harmonic power was measured as a function of temperature, input power and external dc magnetic field for superconducting films. It argues that the origin of the observed nonlinear response is likely due to creation of vortices in the film by the microwave field or an external dc magnetic field.
The authors are thankful to A. Vorob’ev for HTS films preparation and to A.A. Andronov for fruitful discussions and critical comments on the manuscript. This work was supported by the Russian Foundation for Basic Research, grant No. 00-02-16158 and partly by grant No. 00-02-16528.
|
warning/0005/cond-mat0005091.html
|
ar5iv
|
text
|
# LOCAL MAGNETISM OF ISOLATED Mo ATOMS AT SUBSTITUTIONAL AND INTERSTITIAL SITES IN Yb METAL: EXPERIMENT AND THEORY
## Abstract
Using TDPAD experiment and local spin density calculations, we have observed large 4d moments on isolated Mo atoms at substitutional and octahedral interstitial lattice sites in Yb metal, showing Curie-Weiss local susceptibility and Korringa like spin relaxation rate. As a surprising feature, despite strong hybridization with the Yb neighbours, interstitial Mo atoms show high moment stability with small Kondo temperature. While, magnetism of Mo, at substitutional site is consistent with Kondo type antiferromagnetic d-sp exchange interaction, we suggest that moment stability at the interstitial site is strongly influenced by ferromagnetic polarization of Yb-4f5d band electrons.
The formation and stability of local magnetic moments on d-impurities in metallic hosts has been a topic of intense experimental and theoretical investigations over the past several years. While extensive studies have been made for several 3d impurities , much less information is available on the magnetism of 4d atoms in metals. In general, the d-electrons in 4d metals are regarded as being itinerant and do not show tendency towards local moment formation. Recently, applying time differential perturbed angular distribution/correlation (TDPAD/TDPAC) methods, strong magnetic behaviour has been observed for some substitutional 4d impurities in few metals and alloys . The results revealed that, similar to the behaviour observed for 3d impurities, magnetism of 4d ions in metals strongly depends on the type of conduction electrons in the host. Furthermore, it has been argued that moment stability measured by the Kondo temperature, T<sub>K</sub> is strongly influenced by induced spin polarization of host band electrons.
In view of these aspects of magnetism for substitutional 3d and 4d impurities in metals, one can ask : can local moment occur on 3d/4d atoms at interstitial lattice sites also? If so, how stable are the moments? What is the magnitude and sign of host spin polarization and how does it influence the impurity magnetism? Intuitively, due to reduced interatomic distances and the consequent increase in the hybridization strength, one would expect suppressed magnetism for d atoms at interstitial lattice sites. However, large 3d moment has recently been observed for Fe at interstitial sites in fcc Yb metal . Hitherto, magnetism of 4d atoms at interstitial lattice sites in a metal has not been investigated. In this letter we report experimental and theoretical studies on the magnetism of isolated Mo impurity atoms at substitutional and interstitial lattice sites in Yb metal. The results, obtained from TDPAD measurements and local spin density (LSD) calculations show that Mo atoms occupying substitutional and octahedral interstitial sites posses large stable 4d moments with rather small T<sub>K</sub> values. We show that lattice site dependent magnetism of Mo, especially the moment stability, is strongly influenced by the sign and strength of host conduction electron spin polarization.
TDPAD experiments were carried out at the TIFR/BARC Pelletron accelerator facility, Mumbai. We used the 8<sup>+</sup> isomer in <sup>94</sup>Mo (T<sub>1/2</sub> = 98ns, g<sub>N</sub> = 1.31) as a nuclear probe for the detection of magnetic response. Heavy ion reaction <sup>82</sup>Se(<sup>16</sup>O,4n)<sup>94</sup>Mo with pulsed <sup>16</sup>O beam of energy 62 MeV was used to produce and recoil implant the probes in Yb hosts. The estimated concentration of Mo in Yb was less than 1 ppm. Spin rotation spectra, R(t) were recorded in the temperature range 10 to 300 K by applying magnetic field of 2 T. Further details on the experimental method can be found in Ref. . Measurements were performed in fcc and hcp Yb. Following Ref., fcc Yb could be obtained by rolling a piece of pure (99.9$`\%`$) metal to a thickness of about 5 to 10 mg/cm<sup>2</sup>. To get the hcp phase, a thin disk of Yb metal was annealed at 400C for 12 hours followed by repeated cycles of dipping in liquid nitrogen and warming to 300 K. The crystal structures of the samples were verified by X-ray diffraction measurements.
Figure 1 shows some examples of spin rotation spectra, R(t) for <sup>94</sup>Mo in fcc and hcp Yb along with their Fourier transforms. The spectra show superposition of three frequency components arising from Mo atoms at different lattice sites. They were fitted by the function : R(t) =$`_iA_i\mathrm{exp}(t/\tau _{N_i})sin[2(\omega _{L_i}t\theta )]`$ to extract the amplitude A, Larmor frequency $`\omega _L`$ and the nuclear spin relaxation time $`\tau _N`$ of each component. Fig.2 shows the local susceptibilities, $`\chi _{loc}(T)=\beta (T)`$-1 of Mo deduced from the relation $`\omega _L(T)=\mathrm{}^1g_N\mu _NB_{ext}\times \beta (T)`$. It can seen that $`\beta `$ for to two of the components in fcc as well as hcp Yb, strongly vary with temperature which could be fitted to a Curie-Weiss law: $`\beta (T)1=C/(T+T_K)`$. The Curie constants C for the two components, in both phases of Yb, were found to be +15(2) K and -32(2) K. Furthermore, the corresponding nuclear relaxation times $`\tau _N`$, shown in Fig. 3, exhibit Korringa like behaviour ($`\tau _N`$T). Both these features reflect strong local magnetism of Mo. From the amplitudes (A<sub>i</sub>), it turns out that nearly 50$`\%`$ of the implanted Mo atoms in fcc as well as hcp Yb show positive Curie constant while a fraction of $`25\%`$ exhibit negative C value. The remaining $`25\%`$ show temperature independent $`\chi _{loc}`$ with $`\beta (T)=1.00\pm 0.02`$. Tentatively, we assign the components characterized by $`\beta (T)<`$1 and $`\beta (T)>`$1, respectively to substitutional and octahedral interstitial site. The third fraction ($`\beta (T)`$ = 1) presumably corresponds to a tetrahedral interstitial site. Since fcc and hcp Yb has identical near neighbour environments, for our later discussion of Mo magnetism, we mainly consider the components ascribed to substitutional and octahedral interstitial sites in fcc Yb.
To get theoretical understanding on the lattice site dependent magnetism of Mo in Yb host, we have performed first-principles spin polarized, semi-relativistic supercell electronic structure calculations within local spin density approximation (LSDA), employing tight binding linear muffin-tin orbital method in atomic sphere approximation (TB-LMTO-ASA) and Von-Barth-Hedin parameterization for the exchange correlation potential. Calculations were carried out for a single Mo impurity at substitutional as well as octahedral interstitial site, using cubic supercells (space group Pm3m) of dimension twice the lattice constant of fcc Yb, with 32 (33 for interstitial case) atoms and treating Yb-4f as band electrons. In order to accommodate the large Mo impurity in the interstitial site without violating the overlap criterion prescribed by ASA, the nearest neighbour (nn) Yb atoms had to be relaxed outwards by at least 5$`\%`$ of the nn distance. The next nearest neighbour atoms were left unrelaxed thus keeping the unit cell volume intact. No relaxation was necessary for treating the substitutional case.
Figure 4 show the up and down spin local density of states (LDOS) of Mo-d electron. The results showing large splitting of the two spin sub bands clearly indicate the presence of high local 4d moments on Mo. The calculated moments, summarized in Table I, turn out to be 3.56$`\mu _B`$ for substitutional and 1.12$`\mu _B`$ for the interstitial Mo atoms. For further confirmation of the rather high moment of Mo, especially at the interstitial site, we performed additional calculations using spin polarized full-potential linearized augmented plane wave (FLAPW) method as implemented in WIEN code . The resulting Mo moments 3.21$`\mu _B`$ for substitutional and 1.29$`\mu _B`$ for the interstitial sites closely agree with the LMTO-ASA results. For the interstitial case, calculations with higher lattice relaxation, 17$`\%`$ taken from assumption of hard sphere atomic radii, yielded larger moment $``$1.8$`\mu _B`$ (2.1$`\mu _B`$ from FLAPW method).
Coming back to the experimental results, the Mo magnetic moments can be estimated from the Curie constants using C = g(S+1)$`\mu _B`$B(0)/3k<sub>B</sub> where B(0) is the hyperfine field at 0K. For half filled 4d-shell, neglecting orbital contribution, the measured B(0) consists of a negative core polarization field, B<sub>CP</sub> and a positive term (B<sub>val</sub>) arising from the valence electrons. The net spin contact hyperfine fields for many 4d impurities have been found to lie between -200 kG and -280 kG . Assuming B(0) = -240 kG, the Mo moment for the minority fraction ($`\beta (T)<`$1) turned out to be 3.96 $`\mu _B`$ which is close to the value calculated for substitutional site. This supports our earlier attribution of the $`\beta (T)<`$1 component to Mo atoms at substitutional lattice site. For the majority fraction, the observed $`\beta (T)>`$1 behaviour implies B(0) to be positive which can arise from a larger contribution from the valence 5s electrons and much reduced value of B<sub>CP</sub>. The exact value of B(0) though difficult to calculate, one can get a reasonable estimate from the $`\tau _N`$ data which came out to be $``$ +19$`\pm `$(2) T. With this B(0), the Mo moment for majority fraction was found to be 1.5$`\pm 0.3\mu _B`$, in accordance with the values calculated for octahedral interstitial site. This gives credence to our assumption that the component with $`\beta (T)>`$1 arises from Mo atoms at octahedral interstitial site. The above site assignment is also supported from results reported for Fe in Yb . Here, we like to emphasize that any uncertainty in B(0) and the consequent spread in the Mo moments does not influence the main conclusions of this work.
We now examine the stability of Mo magnetic moments in Yb which can be scaled with the Kondo temperature, T<sub>K</sub>. The later can be derived from the Curie-Weiss fit of $`\beta (T)`$ data. The measured $`\beta (T)`$ corresponding to substitutional as well as interstitial sites in fcc Yb yielded T<sub>K</sub> = 60$`\pm `$10 K. For Mo in hcp Yb the T<sub>K</sub> values for the two sites were found to be 35$`\pm `$5 K. An estimation of T<sub>K</sub> could also be obtained from the spin relaxation rates $`\tau _J`$ extracted from the $`\tau _N`$ data . The T<sub>K</sub> derived from $`\tau _N`$ data again turn out to be similar for both substitutional and interstitial sites, being $``$ 55 K for fcc Yb and $``$ 20 K for hcp Yb. The $`\beta (T)`$ and $`\tau _N`$ results indicate rather high moment stability for Mo at substitutional as well as interstitial lattice sites.
Finally, by analyzing the Kondo temperatures of site specific Mo moments in Yb we show that magnetism, particularly moment stability, is strongly influenced by the sign and strength of spin polarization of host conduction band electrons. Starting from Kondo model, instability of a magnetic moment is caused by antiferromagnetic exchange interaction between impurity-d and host conduction electrons. The degree of instability proportional to T<sub>K</sub> is governed mainly by the Kondo resonance width $`\mathrm{\Gamma }=\pi N(E_F)V_{kd}^2`$, where V<sub>kd</sub> is the hybridization strength. For divalent Yb with dominantly sp type conduction electrons, the hybridization strength can be roughly estimated using the procedure given in Ref and are listed in Table I. Using these V<sub>kd</sub> values the T$`{}_{K}{}^{}(=\mathrm{\Gamma }/k_B)`$ of substitutional Mo impurity in fcc Yb was found to be $``$75 K in close agreement with the value 60 K measured experimentally. Further more, the magnetism of substitutional Mo atom in Yb is consistent with the trend observed in alkali and alkaline earth metals where by the reduction in moment correlates with the hybridization strength. The above analysis show that magnetism of substitutional Mo in Yb can be well understood within Kondo model. Extrapolating the same physical picture to Mo impurity at interstitial site, due to stronger hybridization strength (see Table I) one would expect the moment to be highly unstable with T$`{}_{K}{}^{}>400K`$ leading close to nonmagnetic behaviour with $`\beta (T)`$ = 1. Instead, interstitial Mo atoms in Yb show rather stable moment with Curie-Weiss type $`\beta (T)`$ and a low T$`{}_{K}{}^{}`$ 60 K.
What is the physical reason for the high moment stability of interstitial Mo atom in Yb? To understand this, we look into the host spin polarization by examining the induced moments at the Yb sites. The results listed in Table I clearly reveal that sign and strength of host polarization for the interstitial site qualitatively differ from the features seen for substitutional case. For substitutional Mo, the small negative moment at Yb site, mainly arising from sp-band electrons, implies an antiferromagnetic polarization of host band electrons. In contrast, Yb atoms surrounding the interstitial Mo impurity show substantial positive moment largely due to ferromagnetic polarization of 4f5d band electrons of Yb. From the results presented above, we believe that this induced ferromagnetic polarization causing strong interatomic interaction between Mo-4d and host conduction electrons is mainly responsible for the high moment stability of interstitial Mo atoms in Yb. The features of host polarization found for interstitial Mo in Yb and its influence on T<sub>K</sub> show striking similarity with the results observed for 3d, 4d impurities in some d band metals viz. Pd and PdFe alloys . As a plausible mechanism, we suggest that interatomic ferromagnetic interaction between Mo-4d and host conduction band electrons can compete and successfully suppress T<sub>K</sub> arising from antiferromagnetic d-sp exchange interaction and thereby stabilize the magnetic moment of Mo at the interstitial lattice site. The above physical picture is consistent with prediction of recent theoretical calculations where a ferromagnetic interaction between the impurity and host conduction electrons has been shown to suppress the Kondo resonance at Fermi energy .
To conclude, combining TDPAD experiments with local spin density calculations, large 4d local moments have been observed for Mo atoms at substitutional and octahedral interstitial lattice sites of fcc and hcp Yb metal. While, the magnetism of Mo, for the substitutional site is consistent with Kondo type antiferromagnetic d-sp exchange interaction, we find that magnetism and Kondo temperature of interstitial Mo atom is strongly influenced by ferromagnetic polarization of host Yb-4f5d band electrons. The results and interpretations presented in this letter provide an important basis for understanding local magnetism of interstitial d-impurities in a metallic host. They also yield insight on the key role of host polarization on the occurrence and stability of local moments in general.
We thank our Pelletron staff for their excellent cooperation during the experiments.
|
warning/0005/nucl-ex0005004.html
|
ar5iv
|
text
|
# The Emission Pattern of High-Energy Pions: A New Probe for the Early Phase of Heavy Ion Collisions
## Abstract
The emission pattern of charged pions has been measured in Au+Au collisions at 1 GeV/nucleon incident energy. In peripheral collisions and at target rapidities, high-energy pions are emitted preferentially towards the target spectator matter. In contrast, low-energy pions are emitted predominantly in the opposite direction. The corresponding azimuthal anisotropy is explained by the interaction of pions with projectile and target spectator matter. This interaction with the spectator matter causes an effective shadowing which varies with time during the reaction. Our observations show that high-energy pions stem from the early stage of the collision whereas low-energy pions freeze out later.
Heavy-ion collisions at relativistic energies provide a unique possibility to study nuclear matter at high densities and at high temperatures in the laboratory. These reactions last only several $`10^{23}`$ s and within this time interval the baryonic density varies between about three times normal nuclear matter density ($`\rho _0`$ = 0.17 fm<sup>-3</sup>) in the early phase and about $`0.2\times \rho _0`$ at freeze out when the particles cease to interact . A prerequisite for the study of the properties (e.g. the nuclear equation of state) of the dense “fireball” is to obtain information on the space-time evolution of the nuclear matter distribution in the course of the collision.
Pions are considered to be a sensitive probe of the reaction dynamics. They are produced abundantly and due to the large $`\pi N`$ cross section pions are continuously “trapped” by forming baryonic resonances (e.g. $`\pi N\mathrm{\Delta }`$) which then can decay by pion emission. Therefore, pions - especially those with momenta between 0.2 GeV/c and 0.5 GeV/c - are expected to freeze out predominantly in the late and dilute stage of the collision. High-energy pions, however, interact less strongly with the nucleons and hence have a chance to decouple already in an early phase. Therefore, a detailed study of high-energy pions may shed light on the hot and dense stage of the collision as suggested in. First evidence for different freeze-out conditions was extracted from the $`\pi ^{}`$/$`\pi ^+`$ ratio measured in central Au+Au collisions at 1 GeV/nucleon. The $`\pi ^{}`$/$`\pi ^+`$ ratio as a function of transverse momentum was analyzed in terms of Coulomb interaction between pions and the nuclear fireball. It was found that high-energy pions are emitted from a more compact source than low-energy pions .
Our experimental approach to investigate the space-time evolution of the pion source is to exploit the absorption or rescattering of pions when interacting with the spectator fragments. The shadow cast by the spectator matter leads to a depletion of pions according to the emission time of the pions and the motion of the spectator matter. Preferential emission of pions in the reaction plane was found in asymmetric collisions and has been interpreted as an effect of shadowing by a large target nucleus. In symmetric collision systems, a preferential emission perpendicular to the reaction plane has been observed both for charged and neutral pions and has been interpreted as absorption or rescattering effects in the spectator matter. Recently, an enhanced in-plane emission of pions was observed in Au+Au collisions with more (positive) pions being emitted opposite to the target spectator fragments. Since in a hydro-dynamical interpretation the preferential motion of nucleons and composite particles towards the spectator fragments is described as “flow”, the complementary effect observed for pions was called “antiflow”. The “antiflow” of pions is found to be pronounced only in peripheral Au+Au collisions and vanishes for central collisions.
In this Letter we present data on pion production in Au+Au collisions at a bombarding energy of 1 GeV/nucleon as a function of the pion azimuthal emission angle, the rapidity, the transverse momentum, and the collision centrality. With this detailed information one can monitor the effect of spectator shadowing on the pion emission pattern at subsequent stages of the collision. The fast moving spectator matter represents an obstacle for the pions emitted from the fireball. This introduces a time scale which allows to follow the evolution of the pion source and to study correlations of pion energy and freeze-out time. Emphasis is put on the investigation of high-energy pions which are measured with high statistics.
The experiments were performed with the Kaon Spectrometer at the heavy-ion synchrotron SIS at GSI (Darmstadt) which delivered a beam of <sup>197</sup>Au<sup>65+</sup> impinging onto a 1.93 g/cm<sup>2</sup> Au target. The spectrometer covers a momentum-dependent solid angle of $`\mathrm{\Omega }=1535`$msr and a momentum bite of $`p_{max}/p_{min}2`$ for a given magnetic field setting. The measured laboratory momenta vary between 0.156 GeV/$`c`$ and 1.5 GeV/$`c`$ with data taken in four different magnetic field settings. The particle trajectories and momenta are reconstructed using three multi-wire proportional chambers. The particle velocities are determined with two time-of-flight arrays. Both measurements allow to identify pions up to 1.5 GeV/$`c`$. The collision centrality is determined by means of the charged-particle multiplicity measured in a polar angle range between 12 and 48 degrees using a 84-fold segmented plastic-scintillator detector. For our study we select peripheral collisions ($`65\pm 5`$% of the reaction cross section) and the $`14\pm 4`$% most central collisions. The reaction cross section (5.9$`\pm `$0.4 barn) was measured with a minimum bias trigger which required a charged particle multiplicity of more than two in the polar angle range given above.
The determination of the reaction plane in every collision is based on the measurement of charged projectile spectator fragments detected between $`0.5^o\theta _{lab}5^o`$ using a plastic scintillator wall of 380 modules positioned 7 m downstream from the target. The orientation of the event plane is determined by the sum of transverse momenta of the charged projectile spectator particles . The dispersion of the reaction plane amounts to $`45^o`$ for peripheral collisions.
Figure 1 depicts the nuclear matter distribution for a Au+Au collision at a beam energy of 1 GeV/nucleon at 6.5, 12.5 and 16.5 fm/$`c`$ after time zero (which is the time instant when both nuclei have a distance projected to the beam axis of two times the nuclear radius).
These pictures are the result of a transport calculation for an impact parameter of $`b`$ = 7 fm . The snapshots sketch the effect of pion shadowing by spectator matter at different stages of the collision. Those pions which are emitted in the early phase of the collision and are detected around target rapidity (i.e. at backward angles as indicated by the arrows in Fig. 1) will be shadowed by the projectile spectator on one side and therefore exhibit “flow” to the other side. In contrast, if pions freeze out at a late stage of the collision they will be shadowed (at target rapidity) by the target spectator which results in an “antiflow”-like configuration. Experimentally, we compare the number of pions emitted in the reaction plane to the side of the projectile spectator $`N(\varphi =0^{})`$ with the number of those pions emitted into the opposite direction $`N(\varphi =180^{})`$. The azimuthal angle $`\varphi =0^{}`$ is defined by the projectile. $`N_\pi (0^{})`$ refers to the angular range of $`45^{}<\varphi <45^{}`$, and $`N_\pi (180^{})`$ to the angular range of $`135^{}<\varphi <225^{}`$.
Figure 2 shows the ratio of these numbers for $`\pi ^+`$ and $`\pi ^{}`$ mesons as a function of transverse momentum $`p_T`$ in two different rapidity regions and both for near-central and peripheral collisions. In all four cases both $`\pi ^+`$ and $`\pi ^{}`$ (open and full symbols) show a similar behavior ruling out Coulomb effects as the origin of the observed effect. At mid rapidity the measured ratios are close to one both for near-central and peripheral collisions, as expected for symmetric systems. The deviations of about 10% (for near-central collisions) reflect the systematic error of the measurement which is attributed mainly to the uncertainty in the determination of the reaction plane.
In near-central collisions (right panels in Fig. 2) the spectator fragments are small and hence shadowing effects are strongly reduced. Nevertheless, at target rapidities we observe that all pions - independent of their momentum - are emitted preferentially to the side of the target spectator (upper right panel of Fig. 2). The same asymmetry - which corresponds to pion “flow” - was observed by the EOS collaboration at slightly higher incident energies . Transport calculations have predicted a transition from pion “antiflow” to “flow” with decreasing impact parameter. According to these calculations, pion “flow” in near-central collisions is a remainder of the flow of $`\mathrm{\Delta }`$-resonances which decay into protons and pions .
For peripheral collisions and at target rapidity (upper left panel of Fig. 2) the ratio $`N_\pi (0^{})/N(180^{})`$ decreases from about 1.2 at low p<sub>T</sub> values to about 0.5 at high p<sub>T</sub>. This behavior corresponds to a transition from pion number “antiflow” to “flow” with increasing transverse momentum. In earlier measurements it has been integrated over pion momentum and hence, pion “antiflow” has been found since the pion yield is dominated by low-momentum pions. According to transport calculations, “antiflow” is caused by rescattering of pions at the spectator matter in the late stage of the collision . The transition from “antiflow” to “flow” as a function of transverse momentum as shown in the upper left panel of Fig. 2 is a new observation which will be discussed in more details along with Fig. 3.
A more detailed picture of pion emission is obtained by comparing the yield of pions emitted in plane to the yield of pions emitted perpendicular to the reaction plane. The latter pions are expected to be much less affected by shadowing or rescattering by spectator matter and hence provide a nearly undisturbed view onto the pion source. In order to visualize the effect of shadowing for different pion momenta we normalize the in-plane pion spectra ($`N_\pi (0^{})`$ and $`N_\pi (180^{})`$) to the out-of-plane spectra (N$`{}_{\pi }{}^{}(perp)`$ = $`(N_\pi (90^{})+N_\pi (270^{}))/2`$).
Figure 3 shows the ratios $`R_0=N_\pi (0^{})/N_\pi (perp)`$ (“projectile side”, upper panel) and $`R_{180}=N_\pi (180^{})/N_\pi (perp)`$ “target side”, lower panel) as a function of transverse momentum for peripheral collisions and target rapidities. The ratios $`R_{0,180}`$ in Fig. 3 do not exceed unity. This indicates that the azimuthal asymmetry as shown in Fig. 2 is not caused by an enhanced pion emission but rather by losses due to absorption or rescattering (which result in ratios R inferior to unity). Figure 3 allows to extract detailed information on the emission time of pions as a function of their momentum. At pion momenta around 0.4 GeV/$`c`$, the upper left panel of Fig. 2 exhibits no asymmetry ($`N_\pi (0^{})/N_\pi (180^{})1`$) whereas Fig. 3 clearly shows that pion emission into the reaction plane is depleted ($`R_{0,180}<1`$ both at the projectile and target side). This effect is expected if pions are emitted at about 13 fm/$`c`$ when they are shadowed by both the target and projectile spectator (see Fig. 1).
Above momenta of 0.4 GeV/$`c`$, the pion loss increases with increasing momentum for pions emitted towards the projectile side (upper panel of Fig. 3) whereas the opposite trend is observed for pions emitted towards the target side (lower panel of Fig. 3). This finding shows that high-momentum pions are correlated with early emission times (which are even shorter than 13 fm/$`c`$). In contrast, low-energy pions predominantly freeze out at a later stage of the collision. This information is based on the observation that pions with momenta below 0.3 GeV/$`c`$ suffer from absorption or rescattering when emitted towards the target side but remain undisturbed when emitted to the projectile side. The depletion for low-energy pions (as shown in the lower panel Fig. 3) is less pronounced than for high-energy pions (upper panel Fig. 3). This effect indicates that low-energy pions freeze out over an extended time span. Again, $`\pi ^{}`$ and $`\pi ^+`$ mesons behave very similarly which demonstrates that the observed effects are not caused by Coulomb interaction.
Before drawing conclusions on the time evolution of pion emission we investigate another effect. An anisotropy of the ratio $`N_\pi (0^{})/N_\pi (180^{})`$ as function of the pion transverse momentum may be caused by the momentum dependence of the pion-nucleon cross section. The same value of transverse momentum for pions emitted towards the target and projectile remnants corresponds to different relative momenta between pions and nucleons in the remnants. We have performed calculations with a shadowing model using measured pion-nucleon scattering cross sections as functions of the relative pion-nucleon momentum. The model allows to vary the correlation between pion energy and emission time of the pion. It turns out that the momentum-dependent cross section causes the ratio $`N_\pi (0^{})/N_\pi (180^{})`$ to decrease with increasing transverse momentum, qualitatively similar to the observation. However, the calculated ratio $`N_\pi (0^{})/N_\pi (180^{})`$ remains well above unity for all transverse momenta up to 0.8 GeV/$`c`$. The observed reduction of the ratio $`N_\pi (0^{})/N_\pi (180^{})`$ to values well below 1 for large transverse momenta can only be reproduced if hard pions are emitted prior to the instant of closest approach at 13 fm/$`c`$.
Our data show that in Au + Au collisions at 1 GeV/nucleon most of the high-energy pions freeze out within 13 fm/$`c`$ after time zero. Transport calculations predict a similar time scale for the emission of high-energy pions . According to calculations, the nuclear density exceeds twice the saturation value in central Au+Au collisions at 1 GeV/nucleon within the first 15 fm/$`c`$ . Therefore, the investigation of high-energy pions may open a new way to study the nuclear matter equation of state at high baryonic densities.
In summary, we have studied pion production in peripheral and near-central Au + Au collisions at 1 GeV/nucleon as a function of pion transverse momentum, the azimuthal emission angle, and at different rapidities. In peripheral collisions at target rapidity, a reduced yield of high-energy pions is observed at the projectile side. This finding indicates that high-energy pions are shadowed by the incoming projectile spectator and, therefore, are emitted within the first 13 fm/$`c`$ of the collision. In contrast, low-energy pions observed at backwards angles are shadowed by the target spectator which means that they predominantly freeze out in the late phase of the collision.
This work is supported by the Bundesministerium für Bildung und Wissenschaft, Forschung und Technologie under contract 06 DA 819 and by the Gesellschaft für Schwerionenforschung under contract DA OESK and by the Polish Committee of Scientific Research under contract No. 2P3B11515.
|
warning/0005/hep-th0005016.html
|
ar5iv
|
text
|
# 1 The one-loop diagram generating the Ricci scalar for 4D graviton. Wave lines denote gravitons, solid lines denote massive scalars/fermions. Vertical short lines on scalar/fermion propagators indicate that they are massive.
NYU-TH/00/04/01
April 25, 2000
4D Gravity on a Brane in 5D Minkowski Space
Gia Dvali, Gregory Gabadadze, Massimo Porrati
Department of Physics, New York University, New York, NY 10003
Abstract
We suggest a mechanism by which four-dimensional Newtonian gravity emerges on a 3-brane in 5D Minkowski space with an infinite size extra dimension. The worldvolume theory gives rise to the correct 4D potential at short distances whereas at large distances the potential is that of a 5D theory. We discuss some phenomenological issues in this framework.
1. Introduction
The observed weakness of gravity may be due to the fact that we live on a brane embedded in space with large extra dimensions . The correct 4D gravity can be reproduced at large distances due to the finite volume of extra space. This is usually achieved by compactifying extra space. Alternatively, this can be obtained by keeping extra space uncompactified but warped as in the scenario of , where the size $`L`$ of extra space is still finite ($`L=2_0^{\mathrm{}}\sqrt{g}𝑑y<\mathrm{}`$) .
In this work we shall discuss the fate of 4D gravity in theories with infinite size flat extra dimensions. These models may shed new light on supersymmetry breaking and the cosmological constant problem , since they make compatible unbroken bulk supersymmetry with fermi-bose non-degeneracy on a brane (for related discussions, see ). A first example with an infinite size extra dimension and the correct 4D Newtonian potential was proposed in as a generalization of the scenario of . However, the model of has ghosts <sup>1</sup><sup>1</sup>1 The 4D Newtonian force in is mediated by a resonance graviton which has extra polarization degrees of freedom . The brane construction of violates the null-energy condition , and, as a result, it has a state with negative norm, a ghost . A brane bending term which emerges in the traceless-transverse gauge, is a manifestation of a ghost . In the harmonic gauge the ghost can be identified with the radion field which has a negative kinetic term . This ghost can cancel the unwanted polarizations of a resonance graviton . However, the presence of a propagating ghost signals the inconsistency of the model of Ref. ..
In the present paper we suggest a general mechanism by which 4D Newtonian gravity may be generated on any static 3-brane embedded in 5D Minkowski space. We show that the 4D scalar curvature term in the worldvolume brane action
$`M_P^2{\displaystyle d^4x\sqrt{|g|}R},`$
can be responsible for the correct 4D Newtonian interaction on a brane, despite of the fact that gravity propagates in 5D Minkowski space. Such a term is compatible with all the symmetries left unbroken on a brane and can be included in the theory. Moreover, if it is absent in a classical theory, it may be generated on a brane by quantum corrections. It turns out that inclusion of such a term automatically generates the $`1/r`$ gravitational potential at short distances for the sources localized on a brane. As a result, an observer on a brane will see correct Newtonian gravity despite of the fact that gravity propagates in extra space which is flat and has the infinite extent. We should note here that the modification of gravity on a brane due to this effect in weakly coupled string theory would take place at very short, phenomenologically unacceptable distances. Nevertheless, it is interesting to study this mechanism in a low-energy field theory framework.
The paper is organized as follows. In section 2 we present a 5D model with an infinite size extra dimension and discuss mechanisms by which 4D Newtonian gravity on a brane can be obtained from higher dimensional theory. In section 3 we study a prototype model which involves only a scalar field and calculate the potential which is produced by mediation of this scalar. At short distances the potential has 4D nature, while it has 5D behavior at large distances. In section 4 we study the 4D graviton propagator in the infinite extra dimensional model. As in the case with a scalar, gravitons mediate the 4D Newton potential at small distances with a crossover to the 5D potential at large distances. Discussions of some phenomenological issues are given in section 5.
2. A Model of 4D Gravity in 5D Minkowski Space
We start with a $`D=(4+1)`$ dimensional theory. Let us suppose there is a 3-brane embedded in $`5`$-dimensional space-time. We assume that this is a zero-tension brane in $`5`$-dimensions. The four coordinates of our world are $`x_\mu ,\mu =0,1,2,3`$; the extra coordinate will be denoted by $`y`$. Capital letters and subscripts will be used for 5D quantities ($`A,B,C=0,1,2,3,5`$); the metric convention is mostly negative.
Let us consider the action:
$`S=M^3{\displaystyle d^5X\sqrt{G}_{(5)}}+M_P^2{\displaystyle d^4x\sqrt{|g|}R},`$ (1)
where $`M`$ stands for the 5D Planck mass, and $`M_P`$ is the 4D Planck mass; as they stand in (1) $`M`$ and $`M_P`$ are independent (in general they can be dependent). $`G_{AB}(X)G_{AB}(x,y)`$ denotes a 5D metric for which the 5D Ricci scalar is $`_{(5)}`$. The brane is located at $`y=0`$. The induced metric on the brane is denoted by
$`g_{\mu \nu }(x)G_{\mu \nu }(x,y=0).`$ (2)
The 4D Ricci scalar for $`g_{\mu \nu }(x)`$ is $`R=R(x)`$. Possible additional terms of the corresponding SUGRA and/or matter fields confined to a brane are omitted in Eq. (1) for simplicity.
In the limit $`M0`$ with finite $`M_P`$ the action (1) describes 4D gravity on a brane. On the other hand, in the limit $`M_P0`$ with finite $`M`$ it describes 5D bulk gravity. In what follows we study 4D gravitational interactions on a brane when both $`M`$ and $`M_P`$ are finite.
Before we turn to these discussions let us try to understand a possible origin of the action (1).
The first possibility is related to the fact that in certain cases (to be specified below) the localized matter on a brane can induce via loop corrections a 4D kinetic term for gravitons. To demonstrate this, suppose there are matter fields confined to a brane. Thus, the matter energy-momentum tensor can be written as follows:
$`T_{AB}=\left(\begin{array}{cc}T_{\mu \nu }(x)\delta (y)& \text{ 0}\\ \text{0}& \text{ 0}\end{array}\right).`$ (5)
As a result, the interaction Lagrangian of localized matter with 5D metric fluctuations $`h_{AB}(x,y)G_{AB}(x,y)\eta _{AB}`$, reduces to the following expression:
$`_{\mathrm{int}}={\displaystyle 𝑑yh^{\mu \nu }(x,y)T_{\mu \nu }(x)\delta (y)}=h^{\mu \nu }(x,0)T_{\mu \nu }(x),`$ (6)
where the 4D induced metric $`g_{\mu \nu }(x)=\eta _{\mu \nu }+h_{\mu \nu }`$ is defined as in (2). Due to this interaction, a 4D kinetic term can be generated for $`g_{\mu \nu }(x)`$ in the full quantum theory. For instance, the diagram of Fig. 1 with massive scalars , or fermions running in the loop would induce the following 4D term in the low-energy action:
$`{\displaystyle d^4x𝑑y\delta (y)\sqrt{|g|}R}.`$ (7)
The corresponding induced gravitational constant will be determined by a correlation function of the world-volume matter theory<sup>2</sup><sup>2</sup>2For instance, in the 4D framework of Refs. :
$$M_P^2=\frac{i}{96}d^4xx^2\left\{TS(x)S(0)S^2\right\},$$
where $`S(x)T_\mu ^\mu `$ is the trace of the energy-momentum tensor of 4D states running in the loop.. The magnitude of this constant depends on a worldvolume theory at hand and is vanishing in conformaly invariant models, or nonzero if conformal invariance is broken (for detailed discussions see ). We will not attempt to discuss these model dependent features here, we rather assume that the worldvolume theory is such that the second term in (1) is generated with a proper sing and magnitude. We also neglect the induced 4D cosmological constant (which renormalizes the brane tension), $`\mathrm{\Lambda }=0|T_\mu ^\mu |0`$, as well as higher derivative terms $`𝒪(R^2)`$ which can be generated in this case as well. The induced 4D cosmological constant, in general, should be canceled by the brane tension and/or by the vacuum energy due to light matter fields on a brane. Without SUSY this is a usual fine tuning. However, if the bulk is supersymmetric (even though SUSY is broken on a brane) it might require the exact cancellation of the total 4D cosmological constant in a model which at large distances (low energies) becomes effectively 5-dimensional. This certainly needs further separate investigation. Here we assume that the total 4D cosmological constant is zero.
The second way to think of the origin of the 4D term in (1) is to imagine that 5D gravity is coupled to a certain 5D scalar field $`\varphi `$. Suppose that in addition to usual terms there are terms in which $`\varphi `$ couples to the five-dimensional Ricci scalar. Furthermore, suppose that the scalar field potential is such that there is a kink type solution to its equations of motion: $`\varphi _{\mathrm{classical}}=v\mathrm{tanh}(vy)`$. This background, due to the presence of higher derivative interactions might produce the 4D Ricci scalar which would be peaked around the point $`y=0`$. In the limit of a zero width for the kink this term can be approximated by (7). A prototype example with scalars will be presented in the next section. Then we turn to the consideration of gravitons.
3. A Simple Example with Scalars
In this section we study an instructive example which involves a single 5D scalar field. We write the action of the model as follows:
$`S=M^3{\displaystyle d^4x𝑑y_A\mathrm{\Phi }(x,y)^A\mathrm{\Phi }(x,y)}+M_P^2{\displaystyle d^4x𝑑y\delta (y)_\mu \mathrm{\Phi }(x,0)^\mu \mathrm{\Phi }(x,0)}.`$ (8)
For the purpose of comparison with gravity we choose to work with the unconventionally normalized dimensionless scalar field $`\mathrm{\Phi }`$. Let us notice that this kind of action could arise if the 5D field $`\mathrm{\Phi }`$ is placed in the background $`\chi =v\mathrm{tanh}(vy)`$, and the interactions between these fields are taken as: $`(_A\mathrm{\Phi })^2+(_A\chi )^2(_B\mathrm{\Phi })^2(_A\mathrm{\Phi }^A\chi )^2+\mathrm{}`$. This leads to the equation of motion for $`\mathrm{\Phi }`$: $`2_A^2\mathrm{\Phi }+(\chi ^{})^2_\mu ^2\mathrm{\Phi }=0`$. Here the prime denotes derivative with respect to $`y`$. $`(\chi ^{})`$ is a peaked function at $`y=0`$ point. Therefore, this equation can be modeled by the action (8).
Our goal is to determine the distance dependence of interactions which are mediated by this scalar in a 4D worldvolume theory. For this we should find the corresponding retarded Green function and calculate the potential. The classical equation for the Green function looks as follows:
$`\left(M^3_A^A+M_P^2\delta (y)_\mu ^\mu \right)G_R(x,y;0,0)=\delta ^{(4)}(x)\delta (y),`$ (9)
where $`G_R(x,y;0,0)=0`$ for $`x_0<0`$.
The potential mediated by the scalar $`\mathrm{\Phi }`$ on the 4D worldvolume of the brane is determined as:
$`V(r)={\displaystyle }G_R(t,\stackrel{}{x},y=0;0,0,0)dt,`$ (10)
where $`r\sqrt{x_1^2+x_2^2+x_3^2}`$. To find a solution of (9) let us turn to Fourier-transformed quantities with respect to the worldvolume four-coordinates $`x_\mu `$:
$`G_R(x,y;0,0){\displaystyle \frac{d^4p}{(2\pi )^4}e^{ipx}\stackrel{~}{G}_R(p,y)}.`$ (11)
Turning to Euclidean space the equation (9) takes the form:
$`\left(M^3(p^2_y^2)+M_P^2p^2\delta (y)\right)\stackrel{~}{G}_R(p,y)=\delta (y).`$ (12)
Here $`p^2`$ denotes the square of an Euclidean four-momentum. The solution with appropriate boundary conditions takes the form:
$`\stackrel{~}{G}_R(p,y)={\displaystyle \frac{1}{M_P^2p^2+2M^3p}}\mathrm{exp}(p|y|),`$ (13)
where $`p\sqrt{p^2}=\sqrt{p_4^2+p_1^2+p_2^2+p_3^2}`$. Using this expression and Eq. (10) one finds the following (properly normalized) formula for the potential mediated by a scalar in 4D brane worldvolume:
$`V(r)={\displaystyle \frac{1}{8\pi ^2M_P^2}}{\displaystyle \frac{1}{r}}\left\{\mathrm{sin}\left({\displaystyle \frac{r}{r_0}}\right)\mathrm{Ci}\left({\displaystyle \frac{r}{r_0}}\right)+{\displaystyle \frac{1}{2}}\mathrm{cos}\left({\displaystyle \frac{r}{r_0}}\right)\left[\pi 2\mathrm{Si}\left({\displaystyle \frac{r}{r_0}}\right)\right]\right\},`$ (14)
where $`\mathrm{Ci}(z)\gamma +\mathrm{ln}(z)+_0^z(\mathrm{cos}(t)1)𝑑t/t`$, $`\mathrm{Si}(z)_0^z\mathrm{sin}(t)𝑑t/t`$, $`\gamma 0.577`$ is the Euler-Masceroni constant, and the distance scale $`r_0`$ is defined as follows:
$`r_0{\displaystyle \frac{M_P^2}{2M^3}}.`$ (15)
It is useful to study the short distance and long distance behavior of this expression.
At short distances when $`r<<r_0`$ we find:
$`V(r){\displaystyle \frac{1}{8\pi ^2M_P^2}}{\displaystyle \frac{1}{r}}\left\{{\displaystyle \frac{\pi }{2}}+\left[1+\gamma +\mathrm{ln}\left({\displaystyle \frac{r}{r_0}}\right)\right]\left({\displaystyle \frac{r}{r_0}}\right)+𝒪(r^2)\right\}.`$ (16)
As we expected, at short distances the potential has the correct 4D Newtonian $`1/r`$ scaling. This is subsequently modified by the logarithmic repulsion term in (16).
Let us turn now to the large distance behavior. Using (14) we obtain for $`r>>r_0`$:
$`V(r){\displaystyle \frac{1}{8\pi ^2M_P^2}}{\displaystyle \frac{1}{r}}\left\{{\displaystyle \frac{r_0}{r}}+𝒪\left({\displaystyle \frac{1}{r^2}}\right)\right\}.`$ (17)
Thus, the long distance potential scales as $`1/r^2`$ in accordance with laws of 5D theory. These properties are similar to those obtained in the model of Ref. .
In the next section we consider the system defined in (1). We will show that the short and long distance behavior of the Newtonian potential for (1) is determined by the scalar example studied in this section.
4. Gravitational potential
Based on the scalar field example discussed in the previous section we expect that the system (1) will produce the $`1/r`$ gravitational potential at short distances and $`1/r^2`$ potential at large scales. Therefore, if $`r_0`$ in (15) is big enough there will be no contradictions with Newtonian predictions. However, there is a subtlety when it comes to relativistic effects. This is related to the structure of the graviton propagator. The tensor structure of a 4D massless graviton propagator looks as follows (we omit momentum-dependent parts):
$`{\displaystyle \frac{1}{2}}\eta ^{\mu \alpha }\eta ^{\nu \beta }+{\displaystyle \frac{1}{2}}\eta ^{\mu \beta }\eta ^{\nu \alpha }{\displaystyle \frac{1}{2}}\eta ^{\mu \nu }\eta ^{\alpha \beta },`$ (18)
while for a 4D massive case, or equivalently for a 5D massless case, it takes the following form:
$`{\displaystyle \frac{1}{2}}\eta ^{\mu \alpha }\eta ^{\nu \beta }+{\displaystyle \frac{1}{2}}\eta ^{\mu \beta }\eta ^{\nu \alpha }{\displaystyle \frac{1}{3}}\eta ^{\mu \nu }\eta ^{\alpha \beta }.`$ (19)
The difference in the last coefficient ( $`1/2`$ versus $`1/3`$) is vital for description of observations. It was shown in Refs. that no matter how small the graviton mass is in 4D, due to the difference in the tensor structures in (18) and (19), predictions for bending of light are substantially different in the two cases. Moreover, the structure (19) gives contradictions with observations. Any theory with massive gravitons, no matter how light they are, will have to face this problem since there is a discontinuity in the limit when the graviton mass is taken to zero. This can easily be understood in terms of degrees of freedom: a massive graviton has 5 degrees of freedom 3 of which couple to a conserved energy-momentum tensor. Thus, having the propagator as in (19) is equivalent of having a tensor-scalar gravity from 4D point of view. This extra scalar polarization degree of freedom yields additional attractive force. We will continue this discussion in the next section after we find out what is the correct tensor structure of the graviton propagator in the model (1).
To find the propagator let us introduce the metric fluctuations:
$`G_{AB}=\eta _{AB}+h_{AB}.`$ (20)
We choose the harmonic gauge in the bulk:
$`^Ah_{AB}={\displaystyle \frac{1}{2}}_Bh_C^C.`$ (21)
It can be checked that the choice
$`h_{\mu 5}=0,`$ (22)
is consistent with the equations of motions for (1). Thus, the surviving components of $`h_{AB}`$ are $`h_{\mu \nu }`$ and $`h_{55}`$. In this gauge the (55) component of Einstein’s equations yields:
$`_\mu ^\mu h_\nu ^\nu =_\mu ^\mu h_5^5,`$ (23)
which combined with the gauge fixing conditions (21, 22) implies:
$`_A^Ah_\nu ^\nu =_B^Bh_5^5.`$ (24)
Subscripts and superscripts in all these equations are raised and lowered by a flat space metric tensor. Finally, we come to the ($`\mu \nu `$) components of the Einstein equation for (1). After some rearrangements it takes the form:
$`\left(M^3_A^A+M_P^2\delta (y)_\mu ^\mu \right)h_{\mu \nu }(x,y)=\left\{T_{\mu \nu }{\displaystyle \frac{1}{3}}\eta _{\mu \nu }T_\alpha ^\alpha \right\}\delta (y)`$
$`+M_P^2\delta (y)_\mu _\nu h_5^5.`$ (25)
This has a structure of a massive 4D graviton or, equivalently that of a massless 5D graviton, indicating that the tensor structure of the propagator looks as in (19). In this respect, it is instructive to rewrite the expression (25) in the following form:
$`\left(M^3_A^A+M_P^2\delta (y)_\mu ^\mu \right)h_{\mu \nu }(x,y)=\left\{T_{\mu \nu }{\displaystyle \frac{1}{2}}\eta _{\mu \nu }T_\alpha ^\alpha \right\}\delta (y)`$
$`{\displaystyle \frac{1}{2}}M^3\eta _{\mu \nu }_A^Ah_\alpha ^\alpha +M_P^2\delta (y)_\mu _\nu h_5^5.`$ (26)
Here the tensor structure on the r.h.s. is that of a 4D massless graviton (18). However, there is an additional contribution due to the trace part $`h_\mu ^\mu `$ which is nonzero. Therefore, one is left with the theory of gravity which from the 4D point of view is mediated by a graviton plus a scalar<sup>3</sup><sup>3</sup>3The same result could be obtained by using the traceless-transverse gauge and taking into account the “brane bending” effects . We would like to emphasize here that the change of the coefficient $`1/3`$ to the coefficient $`1/2`$ in is due to the possibility to perform gauge transformations of $`h_{\mu \nu }`$ which generate terms proportional to $`\eta _{\mu \nu }`$. This term appears due to the nontrivial warp factor. In the present case the warp-factor is absent, thus the terms proportional to $`\eta _{\mu \nu }`$ are absent in gauge transformations and the change of the coefficient $`1/3`$ into $`1/2`$ does not take place..
Let us now present the exact form for the graviton propagator of (1). Turning to the Fourier images in the Euclidean space as in the previous section we find:
$`\stackrel{~}{h}_{\mu \nu }(p,y=0)\stackrel{~}{T}^{\mu \nu }(p)={\displaystyle \frac{\stackrel{~}{T}^{\mu \nu }\stackrel{~}{T}_{\mu \nu }\frac{1}{3}\stackrel{~}{T}_\mu ^\mu \stackrel{~}{T}_\nu ^\nu }{M_P^2p^2+2M^3p}}.`$ (27)
Here the tilde sign denotes the Fourier-transformed quantities. Thus, the tensor structure of the graviton propagator in 4D worldvolume theory looks as follows:
$`D^{\mu \nu \alpha \beta }={\displaystyle \frac{1}{2}}\eta ^{\mu \alpha }\eta ^{\nu \beta }+{\displaystyle \frac{1}{2}}\eta ^{\mu \beta }\eta ^{\nu \alpha }{\displaystyle \frac{1}{3}}\eta ^{\mu \nu }\eta ^{\alpha \beta }+𝒪(p).`$ (28)
At short distances the potential scales as $`1/r`$ with the logarithmic corrections defined in (16). On the other hand, at large distances the $`1/r^2`$ behavior is recovered in (17). The tensor structure of the propagator is that of 4D tensor-scalar gravity.
5. Discussions and Outlook
So far we worried about the theoretical consistency of the model with an infinite size extra dimension (1). In this section we address the issue whether this model could be phenomenologically viable. There are two outstanding questions in this regard. The first one deals with the extra degree of freedom which shows up in the propagator (28): one should wonder if it is possible to cancel it.
The second question concerns the modification of the Newton law by logarithmic corrections (16): what is the distance at which these modifications will be harmless?
We will discuss these two issues in turn. Let us start with the extra degree of freedom. As we said above, there is additional attraction in the theory due to the extra scalar mode. We should look for some new states which could compensate for this extra attraction. This can certainly be done by a ghost field which gives rise to repulsive force. However, it is hard to make sense of a theory with a manifest ghost. The ghost could be produced as a final state in various processes and this would ruin the consistency of the model at hand. Thus, we should exclude the possibility of using a ghost and try to utilize some other means.
A possibility to compensate for additional attraction would be to use an exchange of a vector particle. One could think of a scenario where all matter fields are given additional $`U(1)`$ charges so that the corresponding gauge fields lead to repulsive interactions. Then one should tune the parameters of the model so that this repulsion cancels (at least partially) the additional attraction in (28). Whether this scenario suffices a single extra $`U(1)`$ or needs a few of them (so that the possible equivalence principle violation is not observable) should be determined by separate phenomenological studies. It is also possible that there exists a more elegant solution to this problem.
Let us now turn to the second issue, what is the crossover scale where gravity in (1) changes its behavior? As we discussed in section 3 this scale is determined by the ratio $`r_0=M_P^2/2M^3`$. Taking the value of the four-dimensional Planck scale $`M_P10^{19}\mathrm{GeV}`$ and assuming that the five dimensional Planck scale is in the $`\mathrm{TeV}`$ region, we find that $`r_010^{15}\mathrm{cm}`$. This is precisely the size of the solar system. However, the crossover scale should be much bigger than the solar system size. For instance, requiring that the logarithmic corrections in (16) give the effects which are small compared to the General Relativity corrections to the precession of the Mercury perihelion, one obtains that the crossover scale should be about $`10^8`$ times bigger then the solar system size. This means that $`M`$ should be about $`10^210^3`$ times smaller then the $`\mathrm{TeV}`$ scale<sup>4</sup><sup>4</sup>4As we mentioned before, in non-supersymmetric D-brane constructions within weakly coupled string theory induced 4D terms will modify gravity at distances less than the string scale.. It is not obvious at present whether such a low scale can be obtained from higher dimensional theories.
In conclusion, we suggested a class of models in which 4D Newtonian gravity can emerge on a brane in 5D flat space. A crucial feature of these models is that the extra dimension is neither compact nor warped and its size is truly infinite. The modification of the Newton law takes place at large distances in this framework. In the minimal setup there are phenomenological subtleties. We outlined some possible ways to circumvent them by adding extra fields. More detailed studies are needed in order determine whether models with infinite size extra dimensions can be given a complete, phenomenologically viable form.
We would like to emphasize that the similar mechanism can be used for “localization” of scalar (discussed in section 3) and vector fields on a brane in flat higher dimensional space. In these cases, the phenomenological subtleties which were present for gravitons are gone. As a result, one is left with theoretically, as well as phenomenologically consistent models of localization for spin-0 and spin-1 fields.
Acknowledgments
The work of G.D. is supported in part by a David and Lucile Packard Foundation Fellowship for Science and Engineering. G.G. is supported by NSF grant PHY-94-23002. M.P. is supported in part by NSF grant PHY-9722083.
|
warning/0005/physics0005008.html
|
ar5iv
|
text
|
# Parameter Sets for 10 TeV and 100 TeV Muon Colliders, and their Study at the HEMC’99 Workshop To appear in Proc. HEMC’99 Workshop – Studies on Colliders and Collider Physics at the Highest Energies: Muon Colliders at 10 TeV to 100 TeV; Montauk, NY, September 27-October 1, 1999, web page http://pubweb.bnl.gov/people/bking/heshop. This work was performed under the auspices of the U.S. Department of Energy under contract no. DE-AC02-98CH10886.
## I Introduction
Self-consistent example parameter sets for the acceleration and collider ring parameters of many-TeV muon colliders were an important focal point for the discussions at the HEMC’99 Workshop – “Studies on Colliders and Collider Physics at the Highest Energies: Muon Colliders at 10 TeV to 100 TeV”, held at Montauk, NY from September 27-October 1, 1999. They served as straw-man examples to be criticized, fleshed-out and improved upon by the accelerator experts attending the workshop, and the physics-related parameters helped the experimental and theoretical physicists at the workshop in their evaluations and comments on the physics potential of such colliders.
Three acceleration and collider parameter sets were used at HEMC’99: one at a center-of-mass energy of 10 TeV (set A) and two at 100 TeV (sets B and C). The collider ring and accelerator parameters are presented in tables 1 and 2, respectively. For comparison, table 1 also includes the parameter ranges for the lower energy muon colliders that have been studied by the Muon Collider Collaboration (MCC). This paper describes the methods used to generate the parameter sets, details the motivations and assumptions for the specific choices of parameters and summarizes the evaluations, conclusions and suggestions for the parameter sets that were given by the workshop participants.
In more detail, the collider ring parameters are presented first, in section II, since they were considered the more critical of the two for assessing the feasibility of many-TeV muon colliders. They also determine the initial assumptions used for the acceleration parameters, which are then discussed in section III. The level of understanding advanced substantially during the workshop, and section IV goes over the issues and viewpoints raised during the workshop as well as referencing the more detailed studies that are included elsewhere in these proceedings and discussing their impact on our assessment of the parameter sets. Finally, the Outlook and Conclusions section, section V, summarizes the results discussed in the preceding section in the more general context of what they imply for the feasibility of many-TeV muon colliders. This concluding section also discusses the outlook for iterations and refinements on the parameter sets and, more generally, previews some plans for further studies on many-TeV muon colliders.
## II Straw-man Muon Collider Ring Parameter Sets at 10 TeV and 100 TeV
### II.1 Generation of the Parameter Sets
The parameter sets in table 1 were generated through iterative runs of a stand-alone computer program, as has been described previously epac98 ; pac99he .
The most important physics parameter for a specified collider energy is the luminosity, $``$. This is derived in terms of several input parameters according to the formula epac98 :
$`[\mathrm{cm}^2.\mathrm{s}^1]`$ $`=`$ $`2.11\times 10^{33}\times \mathrm{H}_\mathrm{B}\times (1\mathrm{e}^{2\mathrm{t}_\mathrm{D}[\gamma \tau _\mu ]})`$ (1)
$`\times {\displaystyle \frac{\mathrm{f}_\mathrm{b}[\mathrm{s}^1](\mathrm{N}_0[10^{12}])^2(\mathrm{E}_{\mathrm{CoM}}[\mathrm{TeV}])^3}{\mathrm{C}[\mathrm{km}]}}`$
$`\times \left({\displaystyle \frac{\sigma _\theta [\mathrm{mr}].\delta [10^3]}{ϵ_{6\mathrm{N}}[10^{12}]}}\right)^{2/3},`$
where the input variables are the CoM energy ($`\mathrm{E}_{\mathrm{CoM}}`$), the collider ring circumference (C), the beams’ fractional momentum spread ($`\delta `$) and 6-dimensional invariant emittance ($`ϵ_{6\mathrm{N}}`$), the time until the beams are dumped ($`\mathrm{t}_\mathrm{D}`$), the bunch repetition frequency ($`\mathrm{f}_\mathrm{b}`$), the initial number of muons per bunch ($`N_0`$), and the beam divergence at the interaction point ($`\sigma _\theta `$). Units in equations throughout this paper are given in square brackets. (The time-to-dump, $`t_D`$, is given in units of the boosted muon lifetime, $`\gamma \tau _\mu `$.) This formula uses the standard assumption from the Muon Collider Collaboration that the ratio of transverse to longitudinal emittances can be manipulated freely in the muon cooling channel to maximize the luminosity for a given $`ϵ_{6\mathrm{N}}`$. The pinch enhancement factor, $`\mathrm{H}_\mathrm{B}`$, is very close to unity (see table 1), and the numerical coefficient in equation 1 includes a geometric correction factor of 0.76 for the non-zero bunch length, $`\sigma _z=\beta ^{}`$ (the “hourglass effect”) .
In practice, the muon beam power and current are limiting parameters for energy frontier muon colliders, so the parameters are actually chosen to optimize the “specific luminosity”:
$$l\frac{}{f_b\times N_0}.$$
(2)
The luminosity is then determined from the choice of beam current that corresponds to the highest plausible beam powers.
Several further parameters in table 1 have been derived from the input parameters that determine the luminosity. These include, for example, the beam-beam tune disruption parameter, $`\mathrm{\Delta }\nu `$. Other output parameters require additional modeling assumptions and/or further input parameters epac98 ; pac99he . Examples include some of the output parameters for the final focus; these require both the input of a reference pole-tip magnetic field for the final focus quadrupoles ($`\mathrm{B}_{5\sigma }`$) and a much simplified model for the final focus magnet lattice that is a linearized extrapolation from existing final focus lattice designs for lower energy muon colliders.
The physics parameters in table 1 include two examples of event sample sizes. As is discussed in references pr98 ; hemc99intro these give an indication of the physics potential corresponding to the specified luminosity and energy. Briefly, the number of $`\mu \mu \mathrm{ee}`$ events gives a benchmark estimate of the discovery potential for elementary particles at the full CoM energy of the collider, while the production of hypothesized 100 GeV Higgs particles indicates roughly how the colliders might perform in studying physics at a lower energy scale.
### II.2 Optimization of the 10 TeV and 100 TeV Parameter Sets
#### The Initial Choice of Energies
The two energies for the parameter sets, $`\mathrm{E}_{\mathrm{CoM}}=10`$ TeV and 100 TeV, were chosen because they bracket that energy decade. The 10 TeV lower limit was chosen to be well above the highest energy that had been studied in detail, namely, $`\mathrm{E}_{\mathrm{CoM}}=4`$ TeV for the Snowmass’96 workshop Snowmass . Further, the neutrino radiation for very high luminosity $`\mu ^+\mu ^{}`$ colliders at 10 TeV and above is high enough to rule out siting them at an existing laboratory, as is covered elsewhere in these proceedings hemc99nurad . This necessitates a fresh outlook for the design optimization of the $`\mu ^+\mu ^{}`$ colliders that is free from site-specific preconceptions involving existing laboratories, which was considered a good thing.
The choice of the upper energy limit was more technically constrained. For the 100 TeV parameter sets, the synchrotron radiation power had risen to become almost identical to the beam power, signaling a clear upper bound for the feasibility of circular $`\mu ^+\mu ^{}`$ colliders.
To preview later discussion, it is noted that our understanding of the constraints on high energy muon colliders advanced during the workshop, as will be covered in section IV. An additional constraint on the maximum possible energies for circular muon colliders was discovered Telnovsynch , due to beam heating arising from the quantum mechanical nature of the synchrotron radiation. On the other hand, the future prospects of many-TeV muon colliders were given a boost when the possible potential for linear colliders at even higher energies was uncovered Zimmermann .
#### Balancing Luminosity against Technical Difficulty
After deciding on the collision energies, it was then decided that the 10 TeV (set A) and the first of the 100 TeV parameter sets (set B) should assume only evolutionary changes in technology from the base-line parameters that have been previously posited for lower energy colliders status . For example, the assumed 6-dimensional emittances are factors of 3.5 (10 TeV) or 50 (100 TeV) smaller than the value $`170\times 10^{12}\mathrm{m}^3`$ that is normally used in Muon Collider Collaboration scenarios for first generation muon colliders. The smaller emittances assume that the performance of the muon cooling channel will be progressively improved through further design optimization, stronger magnets, higher gradient rf cavities and other technological advancements and innovations.
The second parameter set at 100 TeV (i.e., set C) encouraged study on some of the possibilities for using exotic technologies to improve the potential performance of future many-TeV $`\mu ^+\mu ^{}`$ colliders. The additional assumed advances increased the luminosity by two orders of magnitude over the evolutionary parameter set at 100 TeV, to what would be a very impressive $`1\times 10^{38}\mathrm{cm}^2.\mathrm{s}^1`$. (The luminosity should ideally rise as $`\mathrm{E}_{\mathrm{CoM}}^{}{}_{}{}^{2}`$, as is explained in reference hemc99intro .) The hypothesized technical advances included:
1. exotic cooling, to obtain a phase space density that is a further 3 orders of magnitude larger than the assumption for the evolutionary parameter set at 100 TeV
2. charge compensation at the interaction point (ip), to reduce the effective charge by a factor of 10. This assumption led rather directly to a corresponding increase in the luminosity by about a factor of 10.
3. more aggressive final focus parameters were included to allow for potential improvements in the final focus design, perhaps using exotic focusing technologies
4. the beam power was almost doubled from the evolutionary parameter set (B), to “top up” the luminosity to $`1\times 10^{38}\mathrm{cm}^2.\mathrm{s}^1`$ .
#### Final Focus Constraints
The final focus design may well present the most difficult design challenges that are relatively specific to high energy muon colliders. (This is to be contrasted with the muon cooling channel, which is a formidable challenge for all muon colliders.) References epac98 and pac99he have previously addressed the general design constraints and issues for final focus designs at many-TeV muon colliders.
To re-cap the discussion of references epac98 and pac99he , higher energies demand progressively stronger focusing to generate the smaller spot sizes needed to increase the luminosity. Two simply defined parameters were used as benchmarks to obtain final focus specifications that might provide plausible starting assumptions for first attempts at magnet lattice designs. Firstly, an overall beam demagnification parameter is defined epac98 in terms of one of the Courant-Snyder lattice parameters, $`\beta `$, as
$$M\sqrt{\frac{\beta _{\mathrm{max}}}{\beta ^{}}}.$$
(3)
This is a dimensionless parameter that gauges the strength of the focusing. The size of $`M`$ should be closely correlated with fractional tolerances in magnet uniformity, residual chromaticity, etc., where the chromaticity is a measure of the change in response of the final focus to off-momentum particles. Secondly, a high residual chromaticity can be compensated for by decreasing the fractional momentum spread of the beams, $`\delta `$. This suggests that another measure of the final focus difficulty might come from the product of the demagnification and momentum spread,
$$qM\delta ,$$
(4)
where $`q`$ has been referred to epac98 as the “chromaticity quality factor”.
In generating the parameter sets, the values of $`M`$ and $`q`$ were compared to those for existing $`\mathrm{e}^+\mathrm{e}^{}`$ and $`\mu ^+\mu ^{}`$ final focus designs, as was discussed in references epac98 ; pac99he . In practice, slightly more attention was paid to $`q`$ than to $`M`$ in obtaining the final parameters. It can be seen from table 1 that the two “evolutionary” parameter sets, A and B, were constrained to the value $`q=10`$, which is very similar to the calculated value, $`q=11`$, for the final focus lattice design of the 3 TeV $`\mu ^+\mu ^{}`$ collider in reference status . A more aggressive value, $`q=45`$, was allowed for in the second parameter set at 100 TeV.
It is noted that the parameter sets at these high energies are always limited by $`\mathrm{\Delta }\nu `$ and it is useful and straightforward to rewrite equations 1 and 2 in the form:
$$l\frac{\mathrm{\Delta }\nu }{\beta ^{}},$$
(5)
which has no explicit dependence on emittance or bunch size for a given energy. The experience with optimizing the parameter sets was that this independence is true as an approximation only pac99he ; residual dependences on limiting magnet apertures etc. meant that, in practice, it was almost always possible to slightly improve the specific luminosity by re-optimizing to parameter sets with smaller assumed emittances.
A value of $`\mathrm{\Delta }\nu =0.10`$ was assumed for all parameter sets. This was estimated by interpolating the results from a beam tracking study described in reference Snowmass . Equation 5 and the discussion that follows indicate that the luminosity will scale approximately linearly with different assumed values for $`\mathrm{\Delta }\nu `$.
#### Constraints on Energy Spread from Beamstrahlung
The “chromaticity quality factor” figure of merit, equation 4, favors decreasing the fractional momentum spread, $`\delta `$, in order to ease the difficulty of the final focus, and this strategy was found to be effective in optimizing the luminosity for all three parameter sets. By the $`\mathrm{E}_{\mathrm{CoM}}=100`$ TeV energy scale, however, the value of $`\delta `$ was found pac99he to be limited from below by the rapidly rising beamstrahlung at collision. This occurred even though the fractional beamstrahlung energy loss, $`(\mathrm{\Delta }E)_{brem}`$, remained at the level of parts-per-million per beam crossing, i.e., much less than the percent level expected at TeV-scale linear $`\mathrm{e}^+\mathrm{e}^{}`$ colliders. The difference is the need for multiple passes at $`\mu ^+\mu ^{}`$ colliders, which compounds the sensitivity to beamstrahlung losses.
The average beamstrahlung energy losses can be replaced by rf acceleration, of course. However, the particle-by-particle variations will contribute to the spread in the beam momentum, and any such contributions from beamstrahlung must be limited to somewhat below the original momentum spread of the beam. The residual contributions to the beam energy spread should rise as the square root of the number of passes, since they will be statistically independent from turn to turn. Therefore, an appropriate criterion that was chosen to set lower limits on $`\delta `$ is:
$$\frac{(\mathrm{\Delta }E)_{brem}\times \sqrt{n_{turn}^{eff}}}{\delta }\stackrel{<}{}1,$$
(6)
where the effective (i.e. luminosity-weighted) number of turns, $`n_{turn}^{eff}`$, has values in the range $`n_{turn}^{eff}1000`$. The evolutionary (B) and ultra-cool (C) parameter sets at $`\mathrm{E}_{\mathrm{CoM}}=100`$ TeV had chosen values of 0.49 and 0.33 for the left hand side of equation 6, respectively.
As an aside, it is noted that reference pac99he had suggested following the lead of proposed TeV-scale $`\mathrm{e}^+\mathrm{e}^{}`$ colliders by considering the option of using flat, rather than round, beam spots at the ip in order to reduce the beamstrahlung. This was tried, but all attempts led to disappointing luminosities and so round beam spots were retained for the parameter sets.
## III Straw-man Acceleration Parameters
### III.1 Introduction
Table 2 gives straw-man acceleration scenarios that reproduce the final energy and bunch charge for each of the three straw-man muon collider ring scenarios given in table 1, labeled as A) 10 TeV with $`10^{36}`$ luminosity, B) 100 TeV with $`10^{36}`$ luminosity and C) 100 TeV with $`10^{38}`$ luminosity. The layout of each of the two recirculating complexes for table 2 is sketched schematically in figure 1.
The acceleration scenarios of table 2 and figure 1 will be described in subsection III.4. For now, we note that the table contains only a minimal amount of information – much less than was provided for the collider ring – and, in practice, the acceleration parameters were much less critical than the collider ring parameters for determining the technical feasibility or otherwise of the collider scenarios. This viewpoint is supported by a much more detailed and knowledgeable acceleration scenario that is presented elsewhere in these proceedings Berg .
Aside from the technical considerations, the acceleration is expected to dominate the cost of the colliders so its cost optimization will be very important and this was the main design criterion for the straw-man scenarios presented in table 2. To minimize the cost, the scenarios use configurations of recirculating linacs with “fixed field alternating gradient” (FFAG) magnet lattices.
The rest of this section is organized as follows. A very simplistic and non-technical introduction to FFAGs will be given in the next subsection. Some preliminaries on calculating decay losses during acceleration occupy the subsection after that before, in subsection III.4, returning to describe the motivation for the parameter choices in table 2.
### III.2 FFAG Recirculating Arcs
The amount of expensive rf acceleration can be reduced many-fold relative to linear accelerators by bending the muons around for many passes through the same length of linac. The onus then shifts to minimizing the cost of the magnets in the recirculating arcs. In turn, it is then desirable that each of the arcs be able to accept a wide range of momenta so it can be reusable for many traverses. The most promising option for doing this appears to lie in a class of either quadrupole-loaded or combined function magnet lattices that are referred to as “fixed focus, alternating gradient” or “FFAG” lattices. Fast-ramping synchrotrons may also be considered Snowmass ; status but steady-state operation of FFAGs appears likely to be cheaper and might well be more reliable.
Figure 2 gives a very conceptual illustration of the basic idea of FFAGs. It can be seen that the alternating sign of the bending field results in net bending in the direction of the stronger dipoles a provides the scalloped beam trajectories that are characteristic of FFAGs. Further out trajectories see a progressively stronger average bending field and so are appropriate for transporting larger momenta in proportion to the average magnetic field strength.
FFAGs were first considered back in the 1950’s FFAGhistory but, presumably, were not developed further at that time because the simpler alternative of slowly ramping synchrotrons was adequate for the acceleration of stable particles. Impressively, FFAG lattices have now been designed that transport as much as factors of 5 to 10 in muon momentum, although such extreme designs require very large apertures and the peak magnetic fields are several times the average bending field. Some initial design studies for more practical FFAG lattices are presented elsewhere in these proceedings Garren ; Dejan .
Perhaps the biggest technical problem with all FFAG scenarios is the difficulty in maintaining turn-by-turn an appropriate phase relationship with the rf acceleration, since the path lengths of the muon orbits within the FFAG lattice get progressively larger with increasing energy – as is conceptually illustrated in figure 2. It is a nice feature of many-TeV colliders that these problems become progressively less at higher energies because the increasing revolution period through the arcs gives more time for adjustments between passes through the linac.
### III.3 Rf Acceleration and Decay Losses
The amount of radio-frequency (rf) acceleration per turn will be determined by a trade-off between minimizing the expense and the tunnel length occupied by rf (favors less rf) and minimizing the number of turns and the decay losses (favors more rf).
A formula relating the decay losses to the rf and recirculator parameters can be derived directly from the decay equation for the change in the number of muons, $`N`$, with distance, $`x`$:
$$\frac{1}{N}\frac{dN}{dx}=\frac{1}{\beta \gamma c\tau },$$
(7)
where $`c`$ is the speed of light, the scaled muon velocity is essentially unity for the muon energies under consideration, $`\beta =1`$, $`\gamma \frac{E_\mu }{m_\mu c^2}`$ is the conventional relativistic gamma factor and the muon mass and its lifetime, $`\tau `$, are such that $`\frac{mc^2}{c\tau }=0.1604\mathrm{GeV}.\mathrm{km}^1`$.
It follows easily that muon decay losses lead to ratios of initial to final bunch populations, $`\frac{N_i}{N_f}`$, that are related to the recirculator tunnel lengths in units of kilometers, $`L^j[km]`$, the number of GeV per turn of rf acceleration, $`E_{rf}^j[GeV]`$, and the ratio of final to initial energies in the recirculator, $`\frac{E_i^j}{E_f^j}`$, through
$$\mathrm{ln}\left(\frac{N_f}{N_i}\right)=0.1604\underset{j=1,N}{}\frac{L^j[km]}{E_{rf}^j[GeV]}\mathrm{ln}\left(\frac{E_i^j}{E_f^j}\right),$$
(8)
where $`j=1,N`$ is the index for the $`j^{th}`$ of N recirculators. Equation 8 has made the approximation of averaging the acceleration to an assumed constant gradient over the length of the recirculator rather than the real situation where it will be concentrated in one or more rf linacs placed around the recirculator. This should introduce only small fractional errors in the calculated particle losses for the parameters given in table 2.
### III.4 Optimization of the Straw-man Acceleration Scenario
The straw-man acceleration scenario presented in table 2 starts at 500 GeV, working on the assumption that the acceleration to this energy range has already been developed and used for a previous TeV-scale $`\mu ^+\mu ^{}`$ collider. The acceleration scenario for the $`\mathrm{E}_{\mathrm{CoM}}=10`$ TeV collider (set A) then needs to provide exactly one decade of energy gain, accelerating the beams from 0.5 TeV up to their collision energy of 5 TeV. It economizes on expensive rf acceleration by utilizing only a relatively modest 50 GV of rf cavities.
The $`\mathrm{E}_{\mathrm{CoM}}=100`$ TeV collider scenarios start with the $`\mathrm{E}_{\mathrm{CoM}}=10`$ TeV acceleration scenario and add a further decade of acceleration to raise the beam energies to 50 TeV. A further 250 GV of rf cavities are utilized, which is, for example, much less rf than is required for the next generation of $`\mathrm{e}^+\mathrm{e}^{}`$ colliders and so should be easily compatible with the budget constraints on a 100 TeV collider.
In more detail, it can be seen from table 2 that the recirculating accelerator to 50 TeV is essentially a scaled copy of that to 5 TeV. Both recirculators use 5 rings of FFAG arcs and the fractional momentum increment in each of the 5 rings is the same between the first and second recirculator. As one difference, the average bending fields, $`\mathrm{B}_{\mathrm{ave}}`$, in the second recirculator are assumed to be a factor of 1.5 times higher than in the first.
Figure 1 is a schematic diagram for a possible layout of either of the two recirculators. As a specific suggestion of this layout, it assumes the 5 FFAG rings to be housed in the same tunnel. Further, this tunnel is assumed to be the collider tunnel, i.e., the 5+5 TeV collider for the first recirculator and the 50+50 TeV collider for the second.
The obvious motivation for the layout of figure 1 is to minimize the complexity and tunnel expense of the scenario. However, requiring all 5 FFAG rings in a recirculator to have the same radius as the collider ring has the obvious consequences of fixing the FFAG ring radii and of constraining the average bending magnetic fields in each ring according to the ranges of transported momenta in that ring. Table 2 gives a specific scenario for doing this.
The design of the later FFAG rings in each recirculator is clearly more constrained than those for the earlier rings because the average bending field must be closer to the (assumed high) average bending field of the collider ring. This is dealt with in table 2 by constraining the momentum swing to become progressively smaller for the later rings in the recirculators. Assumed energy ranges covered by the arcs range from a factor of 2.5 increase – for the lowest energy arcs in each recirculator – down to 10% energy gain for the highest energy arcs in each recirculator. These are really no more than guesses since, for example, the magnet apertures and ratios of peak-to-average magnetic fields required for this scenario are unknown.
Assumed average gradients for superconducting rf of 25 MV/m, as is assumed for the proposed TESLA $`\mathrm{e}^+\mathrm{e}^{}`$ collider, correspond to total rf lengths of 2 km (10 km) for the 10 TeV (100 TeV) colliders, which is 13.3% (10.0%) of the collider ring circumference. The example schematic layout of figure 1 shows the rf to be split equally between the two straight sections of tunnel on the opposing sides of the “race-track” collider ring, although this choice was somewhat arbitrary.
Decay losses were calculated according to equation 8. Non-decay losses were neglected. Synchrotron radiation energy losses – which range up to about 10% per turn at 50 TeV – have been included in a simple approximate manner. Table 2 shows the overall decay losses to be acceptably low, at 10.5% and 13.9% respectively, for each of the two decades of energy gain.
Having detailed the scenario, it should again be emphasized that the overall scenario, together with its specific choices and assumptions, was intended to do no more than provide the seed for more credible design studies from the accelerator physicists attending this workshop. Nothing but the qualitative assumptions of the scenario should be considered at all, and even these only at the reader’s discretion. Of course, none of the specific numerical assumptions should be taken at all seriously, beyond perhaps obtaining a rough qualitative feel for such parameters as the amount of rf acceleration required and the magnitudes for the fractional decay losses.
Bearing the preceding paragraph in mind, we conclude this section by again referring the reader to the vastly more competent and detailed acceleration studies that emerged from the workshop: the overall acceleration scenarios of reference Berg and the FFAG design studies in references Garren and Dejan .
## IV Assessment of the Muon Collider Parameter Sets at HEMC’99
This section reviews the studies and assessments at HEMC’99 of the collider ring and acceleration parameter sets of tables 1 and 2. It will concentrate on the muon collider design issues arising out of the parameter sets. The reader is also referred to the summary paper by Willis Willis for a more general overview of the findings of the workshop.
Table 3 summarizes the status of the acceleration and collider parameter sets after review at the workshop. As an important piece of contextual information, the assessment of parameter set A (10 TeV), assumes that a TeV-scale muon collider has already been built and successfully operated and the parameter set in each successive column assumes that the collider of the preceding column has already been built.
The following subsections have been grouped according to subject areas that follow fairly closely, but not exactly, the rows of table 3: on luminosity, acceleration, detectors, cooling, synchrotron radiation, final focus design and beam instabilities. A more general outlook and list of conclusions based on these observations will be deferred to the final section, section V.
### IV.1 Assessment of Luminosities for Physics
The luminosity requirements for $`\mu ^+\mu ^{}`$ colliders are discussed in some detail elsewhere in these proceedings hemc99intro . Ideally, collider luminosities should rise as $`\mathrm{E}_{\mathrm{CoM}}^{}{}_{}{}^{2}`$ and it is seen that, indeed, the luminosities for all three parameter sets in table 1 are higher than for any existing or (to the author’s knowledge) other proposed collider.
Both parameter sets A and C have excellent luminosities, even considering their high energies, while the luminosity of parameter set B was still considered to be “fair” for a 100 TeV lepton collider. (See reference hemc99intro for further discussion.)
### IV.2 Assessment of the Acceleration Scenario
Studies at HEMC’99 focused more on the collider ring parameter sets of table 1 than on the acceleration scenario of table 2. The acceleration scenario was considered critical mostly to the extent that it would be expected to be the biggest single component of the overall cost of the collider. Unfortunately, the cost of the FFAG magnets was not able to be explicitly addressed in any detail due to the newness and developing nature of FFAG scenarios Garren ; Dejan for muon colliders. A rather indirect source for some optimism on the acceleration costs could come from any assumed correlation with some relatively favorable cost estimates for the collider ring magnets, by Harrison Harrison , who roughly assessed the cost for the collider magnets for the 10 TeV scenario (set A) to be perhaps of order 400 million dollars.
Technically, muon acceleration tends to get easier at higher energies due to the increasing muon lifetime, smaller beam sizes and lower circulation frequencies in recirculating linacs. Hence, the technical feasibility of acceleration up to the energies in the table is automatically established to a large extent by the assumed previous success of the acceleration at a TeV-scale $`\mu ^+\mu ^{}`$ collider. As a minor caveat to this, Harrison pointed out the increased load due to synchrotron radiation in the FFAGs. However, the collider ring magnets will need to handle the synchrotron radiation load for many times more turns than the FFAG arcs, so even this technical difficulty is concentrated more in the collider ring than the accelerating lattice.
Berg Berg pointed out that slightly increased technical difficulties might instead be expected for the low energy end of the acceleration for parameter sets A and, especially, B. This could result from the higher specified values for the longitudinal emittance in the many-TeV parameter sets: table 1 shows the longitudinal emittances for these parameter sets to be, respectively, similar to, and about twice as large as, the longitudinal emittance for the 3 TeV parameter set of reference status .
### IV.3 Detector Backgrounds
All muon collider detectors face challenging backgrounds resulting from the electron daughters of decaying muons near the interaction point. However, the amount of electromagnetic “junk” entering the detector is relatively independent of the collider energy since the power density of deposited electromagnetic energy depends primarily on the beam current rather than the beam energy. (For confirmation of this statement, see the values in the “power density into magnet liner” row of table 1.) Hence, such backgrounds are expected to be manageable for these many-TeV parameter sets under the stated assumption that the problem has already been solved at TeV-scale collider detectors. (A specific strategy for handling these backgrounds that was developed at the workshop is described in reference Rehak of these proceedings.)
Muons entering the side of the detector, either from beam halo or Bethe-Heitler $`\mu ^+\mu ^{}`$ pair production, are the one background that is expected to evolve markedly with energy. As muons become more relativistic they become less and less like minimum-ionizing particles and deposit larger amounts of energy “catastrophically” in, mainly, electromagnetic showers. This issue was not addressed at the workshop and it deserves further study.
### IV.4 Beam Cooling
Parameter sets A and B assume only evolutionary improvements in the ionization cooling performance over that assumed (but far from demonstrated status !) for TeV-scale colliders so, by definition, the beam cooling should probably be OK if following on from the TeV-scale collider. Parameter set C is very different, assuming that some form of exotic cooling will be able to increase the phase space density of the muon beams by three orders of magnitude from that assumed for parameter set B.
Such ultra-cold muon beams are still looking plausible but have not yet progressed beyond that. The most promising of the exotic cooling methods is optical stochastic cooling Zholents . This method clearly has formidable technical challenges but no obvious show-stoppers. Other, very low energy, cooling methods were also presented at the workshop Lebrun ; Nagamine . There is some concern that any cooling method using non-relativistic muons (i.e. with scaled velocity $`\beta 1`$) may well not be feasible for preparing the high-charge muon bunches needed for colliders, due to space charge limitations.
It is noted that parameter set C provides a specific example of a general feature for ultra-cold muon beams. Since the collisions at many-TeV colliders would normally be tune-shift limited anyway, it is likely that improved cooling would also require ip compensation to substantially benefit the luminosity. We now discuss yet another barrier to the use of ultra-cold beams, at least at very high energies, from synchrotron radiation.
### IV.5 Synchrotron Radiation
It has already been noted that the synchrotron radiation power in the 100 TeV colliders is already comparable to the beam power. During the workshop, Telnov Telnovsynch raised what might possibly be a stronger constraint from synchrotron radiation on the energy reach of circular muon colliders, namely, the quantum nature of synchrotron radiation may lead to heating, rather than damping, of the horizontal beam emittance if the beam energy is high enough and the emittance is already very small.
Telnov’s observation clearly spells the end of parameter set C, with its ultra-cold beam at $`\mathrm{E}_{\mathrm{CoM}}=100`$ TeV. The other parameter set at 100 TeV (set B) is also borderline, with an initial horizontal emittance that is larger by a factor of five Telnovsynch than the equilibrium emittance due to this effect, as calculated by Telnov using a simple approximate model.
The most likely possible loop-hole for parameter set B is that the heating effect is reduced for a very strongly focusing collider lattice. More specifically, Telnov’s equation 2 shows the equilibrium emittance to be proportional to the average of the “H-function” around the collider ring, where
$$H\frac{\beta ^3}{\rho ^2},$$
(9)
for $`\beta `$ the standard Courant-Snyder parameter and $`\rho `$ the collider ring’s radius. (Stronger focusing corresponds to smaller $`\beta `$ values around the ring.)
To consider adjustments to parameter set B, equation 9 suggests that 100 TeV colliders with the emittances expected from ionization cooling still look to be feasible by increasing the ring radius, $`\rho `$, by, for example, a factor of two. This would lower both the equilibrium emittance by a factor of four and the radiated energy per turn by a factor of two, which is substantial compensation for halving the number of collisions per bunch.
A much more dramatic approach to beating the energy limits from synchrotron radiation has come from Zimmermann Zimmermann , in the form of single pass linear $`\mu ^+\mu ^{}`$ colliders. Example parameter sets are included in Zimmermann’s paper, and are commented on in more detail elsewhere in these proceedings hemc99intro .
### IV.6 Final Focus Design
The final focus design extrapolations discussed in section II seemed to work well for the 10 TeV parameter set A. A magnet layout for the final focus from Johnstone Carol closely reproduced the predicted $`\beta _{max}`$ in table 1. Further, the lattice design experts at the workshop seemed to appreciate the extremely challenging nature of the 10 TeV final focus parameters without everybody actually condemning them as being clearly unrealistic, i.e., an appropriate level of difficulty for a workshop of this nature! See reference Zimmermann for more detailed studies and comments.
The 100 TeV parameter sets were less fortunate. Even the “evolutionary” parameter set B was immediately dismissed by the lattice experts as being incompatible with any final focus lattice designs using conventional magnets. It will be very useful to get further feedback on what exactly broke down in the simplistic energy extrapolation that was described in section II.1. Hopefully, such feedback can then be used to obtain a better parameterization of the energy evolution in the final focus parameters. A more realistic and better established parameterization could then be used to predict the luminosity scaling with energy that might be expected using conventional final focus technologies.
Finally, two exotic final focus options were discussed that might go beyond conventional magnet designs: “dynamic focusing” (using auxiliary beams to focus the colliding beams) and plasma focusing. Discouragingly, both options looked much less plausible than when considered for single pass $`\mathrm{e}^+\mathrm{e}^{}`$ colliders, due to both the need for multiple passes and the larger bunch currents assumed for $`\mu ^+\mu ^{}`$ collider parameters. Also disappointing are the obstacles to beam compensation at collision (as was assumed in parameter set C), which call into question the possibility of being able to do this – see reference Telnovcomp for discussion on this topic.
### IV.7 Beam Instabilities in the Collider Ring
Papers by Keil Keil and Zimmermann Zimmermann provide studies on beam instabilities. Keil provides a systematic assessment of the classes of instabilities, including parameter comparisons with the LHC collider ring. Zimmermann’s tracking studies demonstrated that even circulating the beams for a single turn should not be taken for granted, let alone for of order 1000 turns over the lifetime of the muons.
As a connection to the physics capabilities of the collider ring that needs to be borne in mind, the common assumption status of collider rings that are isochronous is disfavored for retaining the beam polarization. (See also reference Heusch for a discussion on the importance of polarization.) As a rough hand-waving explanation, the rate of polarization precession while circulating in the collider ring is proportional to the muon’s energy. It is intuitively clear that the polarization will decay away more slowly if the energies of all the particles are allowed to slosh around the beam average energy – sometimes gaining in polarization precession (higher energy) over the bunch average precession and sometimes losing (lower energy than the bunch average). This is what happens in a collider ring with longitudinal focusing as opposed to isochronous rings. The same argument also favors small beam energy spreads.
## V Outlook and Conclusions
The preceding section has reviewed the insights from HEMC’99 on the parameter sets of tables 1 and 2. More generally than this, HEMC’99 has provided the first speculative insights into (i) the ultimate physics potential for future colliders at the high energy frontier and (ii) the potential challenges to reaching very high energies with muon colliders. A personal interpretation of the workshop’s findings through the energy decades is:
* muon colliders to the TeV scale: (added for completeness – these energies were not discussed in detail at the workshop) beam cooling is the dominant technical challenge. Other major challenges are the final focus region, backgrounds in the detector, cost-efficient acceleration and beam stability throughout the cooling, acceleration and storage in the collider ring. Neutrino radiation will impose significant design constraints and the beam currents may be well below those for the straw-man parameters in reference status (i.e. $`6\times 10^{20}`$ muons/sign/year in collision).
* to advance to the 10 TeV scale: neutrino radiation will probably dictate a new site. The final focus region of the collider and magnet cost reduction for acceleration may be the other major technical design issues.
* to advance to the 100 TeV scale: major breakthroughs are needed in magnet costs and in the final focus region.
* to advance to the 1 PeV scale and beyond: this is not absolutely ruled out in the far distant future using a linac and many technological breakthroughs, as illustrated by the parameter set in reference Zimmermann and discussed further in reference hemc99intro .
It would certainly be very valuable to follow up on the understandings gained at this workshop. As a small first step, modified parameter sets for many-TeV muon colliders are being generated epac2000 that take into account the insights gained at HEMC’99. As a refinement to make interpolations easier, a parameter set at the intermediate center-of-mass energy of 30 TeV will be included.
More substantially, there is need for a new study and workshop. Preferably, this should include all three of the main accelerator technologies – pp, $`\mathrm{e}^+\mathrm{e}^{}`$ and $`\mu ^+\mu ^{}`$ colliders. This is motivated hemc99intro both for a more coherent understanding of the future of experimental high energy physics and in recognition that the three accelerator technologies are deeply intertwined. Planning is underway for such a study to take place in the Summer and Fall of 2001.
|
warning/0005/astro-ph0005428.html
|
ar5iv
|
text
|
# Exhaustive Study of Cosmic Microwave Background Anisotropies in Quintessential Scenarios
## I Introduction
Recent measurements of the luminosity distance versus redshift relation for type Ia supernovae , if confirmed, are compatible with an expanding (accelerating) universe driven by a new type of matter whose equation of state $`p=\omega \rho `$ is characterized by a negative $`\omega `$. One of the possible explanations is the existence of a non-zero vacuum energy, i.e. a “cosmological constant”. Another pragmatic possibility which has been proposed is to assume the existence of a yet unknown mechanism guarenteeing that the true cosmological constant vanishes, the remaining energy density being then due to the presence of a scalar field, the quintessence field, almost decoupled from ordinary matter . The main difference between a quintessence fluid and a cosmological constant comes from their equation of state where $`\omega _\mathrm{\Lambda }=1`$ for a cosmological constant and $`1\omega _\mathrm{Q}0`$ for the quintessence fluid.
One of the puzzles in the interpretation of these data is the extremely small value of the energy density due to the new form of matter. From the point of view of particle physics a vanishing value for the cosmological constant is one of the major challenges . At present there is no known mechanism which prevents the vacuum energy from picking large values due to radiative corrections and one expects typically a contribution equal to $`(\mathrm{}c/2)d𝐤k/(2\pi )^3\mathrm{}ck_{\mathrm{m}ax}^4/(16\pi ^2)`$, where $`k_{\mathrm{m}ax}`$ is a cut-off which can naturally be taken as the Planck wavenumber. This gives a contribution which is $`120`$ orders of magnitude above the observed one. One possibility which is often advocated is the presence of some global supersymmetry (SUSY) which would guarantee that the energy of the vacuum is zero. Unfortunately SUSY has to be broken to take into account the absence of experimental evidence in favour of particle superpartners leading to a natural contribution to the vacuum energy of order $`M_{\mathrm{S}USY}^4`$ where $`M_{\mathrm{S}USY}`$ is the SUSY breaking scale estimated around $`1\mathrm{TeV}`$ . The measurement of a vacuum energy some $`60`$ orders of magnitude below this expected value indicates that some new physics must be at play here.
In the quintessence hypothesis, the small vacuum energy density is due to the rolling down of the quintessence field $`Q`$ along a decreasing potential. A typical potential is the Ratra-Peebles potential $`V(Q)=\mathrm{\Lambda }^{4+\alpha }/Q^\alpha `$ . From the particle physics point of view one would like to justify the existence of the quintessence field. Several natural candidates have been ruled out such as the axion-dilaton field , the moduli fields of toroidal compactifications in string theory and finally the meson fields of supersymmetric gauge theories . Nevertheless, it seems reasonnable to expect that SUSY will play a role in the solution. Within this framework it is a matter of fact that the quintessence field must be part of supergravity (SUGRA) models . This comes from the large value $`Qm_{\mathrm{P}l}`$ of the field at small redshift which implies that SUGRA corrections cannot be neglected. In an effective theory approach has been used to deduce the general form of quintessence SUGRA potentials, they are of the type
$$V(Q)=\frac{\mathrm{\Lambda }^{4+\alpha }}{Q^\alpha }e^{\kappa Q^2/2},$$
(1)
where $`\kappa 8\pi G`$, $`G`$ being the Newton constant, and where the exponential factor comprises the SUGRA corrections. $`\mathrm{\Lambda }`$ and $`\alpha `$ are free parameters. The fine-tuning is not too severe as for typical values $`\alpha =6`$ the scale $`\mathrm{\Lambda }10^6\mathrm{GeV}`$ is compatible with high energy scales. Notice that the SUGRA corrections become relevant towards the end of the evolution and decouple at small $`Qm_{\mathrm{P}l}`$. Different types of potentials can be distinguished because they lead to different values of the equation of state parameter. For example, for $`\alpha =11`$, the Ratra-Peebles potential is such that $`\omega _\mathrm{Q}0.29`$ whereas the SUGRA potential gives $`\omega _\mathrm{Q}0.82`$ (for $`\mathrm{\Omega }_\mathrm{Q}=0.7`$).
It is also worth noticing that there exists quintessence models where the field is non-minimally coupled with the metric. Such models induce time-variation of the Newton constant and are therefore already constrained, for example by observations in the solar system or by pulsar timing measurements . They lead to the same tracking behaviour, as stressed in Refs. , as soon as the coupling term is proportional to a power of the potential. However, some important differences occur when the field starts dominating; for example its effective equation of state can reach extreme values such that $`\omega 3`$ . Also, these models can lead (especially in the context of quintessential inflation ) to clear observable features in the gravitational waves spectrum .
In view of the numerous phenomenological successes of quintessence, it is relevant to deduce its consequences for Cosmic Microwave Background (CMB) anisotropies and structure formation. The aim is two–fold. First, we have to study whether quintessence leads to acceptable scenarios and, second, we have to learn how we could use high-precision measurements recently obtained by the BOOMERanG and MAXIMA-1 experiments or to be performed in the near future by NASA’s Microwave Anisotropy Probe (MAP) satellite , ESA’s Planck satellite or the Sloan Digital Sky Survey (SDSS) to put constraints on the quantities characterizing quintessence like $`\mathrm{\Omega }_\mathrm{Q}`$ or $`\omega _\mathrm{Q}`$. The second possibility has of course already been investigated for the cosmological constant case. For example, the fraction $`\mathrm{\Omega }_\mathrm{\Lambda }`$ of the critical density is not determined entirely from the supernovae data. Indeed, the data from the supernovae observations are degenerate in the plane $`(\mathrm{\Omega }_\mathrm{m},\mathrm{\Omega }_\mathrm{\Lambda })`$, where $`\mathrm{\Omega }_\mathrm{m}`$ is the matter (i.e. cold dark matter plus baryons) component preventing a clear cut determination of the fraction $`\mathrm{\Omega }_\mathrm{\Lambda }`$. The situation changes drastically if one includes the measurements of the CMB anisotropies (even without the BOOMERanG or MAXIMA-1 data). In that case, the degeneracy is removed leading to a probable $`70\%`$ of the total energy density of the universe carried by the negative pressure fluid while the remaining $`30\%`$ are the matter components ensuring that $`\mathrm{\Omega }_0=1`$ in agreement with a spatially flat universe. This conclusion can be drawn from the measurements of the location of the first Doppler peak. This result has been confirmed by other measurements . Another use of combined data has been to put constraints on the equation of state parameter. However, this has been done only for constant or for very simple time-dependent $`\omega _\mathrm{Q}`$ .
CMB anisotropies and the power spectrum are calculated with the help of the theory of cosmological perturbations. Cosmological perturbations in the presence of quintessence have been studied by Ratra and Peebles but only in the tracking regime . CMB multipoles moments and/or the power spectrum have already been calculated for the Ratra-Peebles potential in Ref. and for other models of quintessence in Refs. . One important issue is to understand whether the final evolution of the various perturbed quantities depend on the initial conditions imposed at reheating (of the inflationary type or not). Another way to put the same problem is the following: do the multipole moments depend on the value of $`\delta Q(\eta _\mathrm{i})`$ and $`\delta Q^{}(\eta _\mathrm{i})`$ at initial time? In Ref. , it was noticed that the answer to this question is no but no explanations were provided. Here, we confirm the remark of Ref. and show that this is due to the fact that the perturbed Einstein equations also possess an attractor which renders the multipole moments insensitive to the initial conditions.
One of the main purpose of this article is the study of the general properties of the multipoles moments of the CMB anisotropies in presence of the quintessence field. We present the CMB multipole moments for the Ratra-Peebles potential and, for the first time, for the SUGRA tracking potential. In addition, we also display the matter power spectrum for these two models. Recently, it has been shown by Kamionkowski and Buchalter that the location of the second peak in the CMB power spectrum is an efficient way of revealing some features of the dark energy sector. Therefore, we pay special attention to this question. In particular, in Ref. , only the cosmological constant case was studied and it was argued that the quintessence case (the authors refer to the Ratra-Peebles potential) must not differ significantly from the cosmological constant case. In the present article, we demonstrate that this is not the case and that, as a matter of fact, quintessence leads to a different location (denoted, in the following, $`l_2`$) of the second peak. In addition, we show that the location of the second peak in the quintessence case and in the cosmological constant case can be easily distinguished. Following Ref. , we display the contour plots of $`l_2`$ in the plane $`(\mathrm{\Omega }_\mathrm{m},h)`$ for the Ratra-Peebles and SUGRA tracking potentials.
The article is organized as follows. In section II, we give a description of the background evolution in terms of two physical quantities: the equation of state parameter $`\omega _\mathrm{Q}`$ and the sound velocity $`c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}`$. In section III, we study the cosmological perturbations for the quintessence field. In section IV, we present the results of full numerical calculations with the help of a Boltzmann code developped by one of us (A.R.) for the CMB anisotropies and power spectra in the case of the Ratra-Peebles and SUGRA potentials. Then, detailed comparisons with the recent BOOMERanG and MAXIMA-1 data are performed. We end with our main conclusions in section V.
## II The background evolution
We suppose that the Universe can be described by a Friedman-Lemaître-Robertson-Walker metric the spacelike sections of which are flat
$$\mathrm{d}s^2=a^2(\eta )(\mathrm{d}\eta ^2+\delta _{ij}\mathrm{d}x^i\mathrm{d}x^j).$$
(2)
In this equation, $`\eta `$ is the conformal time related to the cosmic time by $`a(\eta )\mathrm{d}\eta \mathrm{d}t`$. The matter content is as follows. The Universe is filled with a mixture of five fluids: photons ($`\gamma `$), neutrinos ($`\nu `$), baryons ($`\mathrm{b}`$), cold dark matter ($`\mathrm{c}dm`$) and a scalar field $`Q`$ named quintessence. The stress energy tensor of each of these species is the one of a perfect fluid, $`T_{\mu \nu }=(p+\rho )u_\mu u_\nu +pg_{\mu \nu }`$, where $`u_\mu `$ is the 4-velocity of the fluid. The energy density and the pressure of the scalar field are given by $`\rho _\mathrm{Q}=\frac{1}{2}(Q^{}/a)^2+V(Q)`$ and $`p_\mathrm{Q}=\frac{1}{2}(Q^{}/a)^2V(Q)`$, where $`V(Q)`$ is the potential of quintessence whose shape will be very important in what follows. Each fluid is also characterized by its equation of state $`p_\mathrm{i}\omega _\mathrm{i}\rho _\mathrm{i}`$ where $`\mathrm{i}=\gamma ,\nu ,\mathrm{b},\mathrm{c}dm`$ or $`\mathrm{Q}`$. We have $`\omega _\gamma =\omega _\nu =1/3`$ and $`\omega _\mathrm{b}=\omega _{\mathrm{c}dm}=0`$. The case of $`\omega _\mathrm{Q}`$ is more complicated since this is a time-dependent function such that $`1\omega _\mathrm{Q}+1`$. Its expression reads $`\omega _\mathrm{Q}=12V(Q)/\rho _\mathrm{Q}`$. The fact that $`\omega _\mathrm{Q}`$ is a time-dependent function directly comes from the fact that, for a scalar field, the sound velocity defined as
$$c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}\frac{p_\mathrm{Q}^{}}{\rho _\mathrm{Q}^{}}=1+\frac{4a^2}{3Q^{}}\frac{\mathrm{d}V(Q)}{\mathrm{d}Q}=\frac{1}{3}\left(2\frac{Q^{\prime \prime }}{Q^{}}+1\right),$$
(3)
is not equal to the equation of state parameter $`\omega _\mathrm{Q}`$. As a consequence $`\omega _\mathrm{Q}`$ has to change in time as revealed by the following equation
$$\omega _\mathrm{Q}^{}=3(1+\omega _\mathrm{Q})(c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}\omega _\mathrm{Q}),$$
(4)
unless $`\omega _\mathrm{Q}=1`$.
The evolution of the Universe can be calculated with the help of the Friedman and conservation equations
$`{\displaystyle \frac{1}{a^2}}^2`$ $`=`$ $`{\displaystyle \frac{8\pi }{m_{\mathrm{P}l}^2}}{\displaystyle \underset{i}{}}\rho _\mathrm{i},`$ (5)
$`\rho _\mathrm{i}^{}`$ $`=`$ $`3(1+\omega _\mathrm{i})\rho _\mathrm{i},\mathrm{i}=\gamma ,\nu ,\mathrm{b},\mathrm{c}dm\text{ or }\mathrm{Q},`$ (6)
where $`m_{\mathrm{P}l}`$ is the Planck mass and $`a^{}/a`$ is related to the Hubble constant by the equation $`H=/a`$. The equations of conservation simply express the fact that the energy is conserved for each species which do not interact. The equation of conservation of the quintessence field can also be written as the Klein-Gordon equation
$$Q^{\prime \prime }+2Q^{}+a^2\frac{\mathrm{d}V}{\mathrm{d}Q}=0.$$
(7)
We now need to give the last piece of information necessary to have a complete description of the system, i.e. the shape of the potential $`V(Q)`$. In order to be an interesting theory and to represent an improvement over the current situation, quintessence has to address the following four problems: the fine-tuning problem, the coincidence problem, the equation of state problem and the model building problem. The fine-tuning problem amounts to understanding whether one can have $`\mathrm{\Omega }_\mathrm{Q}0.7`$ with the free parameters of the potential taking “natural” values, i.e. close to the energy scale of the theory under consideration. The coincidence problem is the question of the initial conditions: does the final value of $`\rho _\mathrm{Q}`$ strongly depend on the chosen initial values of $`Q`$ and $`Q^{}`$? The equation of state problem is the question of the value of $`\omega _\mathrm{Q}`$. In order to be compatible with observational data, it should be such that $`1<\omega _\mathrm{Q}<0`$. According to recent papers, even more stringent restrictions can be put, namely $`1<\omega _\mathrm{Q}<0.6`$ or even $`1<\omega _\mathrm{Q}<0.8`$ . In particular, this already rules out a network of cosmic strings since the corresponding fluid has an equation of state parameter equal to $`1/3`$. Finally, the model building problem consists in justifying the shape of the potential from the high energy physics point of view. Many different shapes of potential which allow, at least partially, to solve these problems have been investigated in the litterature and Table I summarizes these proposals.
In particular, the first possibility has been studied thoroughly in the past years. In this article, we will mainly concentrate on the Ratra-Peebles potential and the SUGRA tracking potential .
Let us briefly see how the four questions evoked previously can be addressed with these potentials.
### A The fine-tuning problem
Let us start with the fine-tuning problem which is clearly a delicate question. This problem is crucial for the cosmological constant. Indeed, from very simple high energy physics considerations, one typically expects $`\rho _\mathrm{\Lambda }m_{\mathrm{P}l}^410^{76}\mathrm{GeV}^4`$ whereas one measures $`\rho _\mathrm{\Lambda }\mathrm{\Omega }_\mathrm{\Lambda }\rho _\mathrm{c}10^{47}\mathrm{GeV}^4`$ since the critical energy density is $`\rho _\mathrm{c}8.1h^2\times 10^{47}\mathrm{GeV}^4`$. Do we gain something in the case of quintessence? This question is controversial. For example in Ref. , the authors clearly answer no and write “Two proposals to explain these observations are a non-vanishing cosmological constant or a very slowly rolling scalar field, often dubbed quintessence. Both proposals, however, are plagued with formidable fine tuning problems.” However, one should look more carefully at this point. To illustrate this issue, let us consider the general argument given against quintessence. If we consider the potential $`V(Q)=(m^2/2)Q^2`$ then the mass of such a field, which is also the only free parameter of the potential, should be $`m=\sqrt{2\mathrm{\Omega }_\mathrm{Q}\rho _\mathrm{c}}/m_{\mathrm{P}l}10^{33}\mathrm{eV}`$, a very tiny mass indeed. Justifying such a value for the free parameter $`m`$ is probably the same problem as justifying a very low value for $`\rho _\mathrm{\Lambda }`$. However, such a model has never been advocated for the quintessence field. As already mentioned above, one typically considers models such that $`V(Q)=\mathrm{\Lambda }^{4+\alpha }/Q^\alpha `$. This changes the argument. Now, the free parameter of the theory is $`\mathrm{\Lambda }`$. In order to have $`\rho _\mathrm{Q}=\mathrm{\Omega }_\mathrm{Q}\rho _\mathrm{c}`$ today, one has $`\mathrm{\Lambda }10^{11}\mathrm{GeV}`$, for $`\alpha =11`$. This time, the free parameter of the theory has a value comparable to the natural scales of high energy physics. Therefore, something has been gained and it seems unfair not to emphasize this point. On the other hand the mass of the field is given by $`m=\alpha (\alpha +1)\mathrm{\Omega }_\mathrm{Q}\rho _\mathrm{c}/m_{\mathrm{P}l}^210^{33}\mathrm{eV}`$ but this number should be interpreted completely differently. Here the mass $`m`$ is just a “by-product” and its value is naturally very small without any artificial fine-tuning of $`\mathrm{\Lambda }`$. Of course the very small value of the mass implies that the quintessence field is almost completely decoupled from the other matter fields. This renders the model building issue even more acute.
### B The coincidence problem
The coincidence problem as formulated in the introduction, i.e. the dependence upon the initial conditions, is solved because the Klein-Gordon equation possesses an attractor. In order to prove this property, we have to rely either on numerical calculations or on approximate methods. All the plots and numerical estimates displayed in this article will be made with the help of numerical calculations. However, it is always useful to understand the tracking property by means of analytical methods and we now turn to this question. It is convenient, for analytical calculations, to consider that there is in fact only one “background” fluid with a time dependent equation of state such that $`\omega _\mathrm{B}=1/3`$ during the radiation dominated epoch and $`\omega _\mathrm{B}=0`$ during the matter dominated era. In addition to the background fluid, we assume that there also exists the quintessence scalar field field. Following the treatment of Ratra and Peebles , it will be considered that this scalar field is a test field. This is a good approximation since this field must be sub-dominant in particular during Big Bang Nucleosynthesis (BBN) in order not to modify the behaviour of the scale factor and, as a consequence, not to spoil the success of BBN. This means that the behaviour of the scale factor is essentially determined by the background fluid and that $`_i\rho _\mathrm{i}\rho _\mathrm{B}`$. This hypothesis breaks down at very small redshift when quintessence starts dominating the matter content of the Universe. Since quintessence is only a test field which does not interact with the background fluid, the scale factor and the quantity $``$ can be written as
$$a(\eta )\eta ^{2/(1+3\omega _\mathrm{B})},(\eta )=\frac{2}{(1+3\omega _\mathrm{B})\eta }.$$
(8)
For the sake of illustration, let us now consider the radiation dominated era where $`\omega _\mathrm{B}=1/3`$. Under the previous assumptions, the Klein-Gordon equation has a particular solution given by
$$Q_\mathrm{p}=Q_0\eta ^{4/(\alpha +2)},$$
(9)
where $`Q_0`$ is a constant which depends on the free parameters of the potential, i.e. $`\mathrm{\Lambda }`$ and $`\alpha `$. The tracking behaviour is revealed by the behaviour of small perturbations around $`Q_\mathrm{p}`$. Let us introduce the new time $`\tau `$ defined by $`\eta e^\tau `$ and define $`u`$ and $`p`$ by $`Q=Q_\mathrm{p}u`$ and $`p=\mathrm{d}u/\mathrm{d}\tau `$. The Klein-Gordon equation, viewed as a dynamical system in the plane $`(p,u)`$, possesses a critical point $`(0,1)`$ and small perturbations around this point $`\delta u,\delta p`$ obey the following equation
$`{\displaystyle \frac{\mathrm{d}}{\mathrm{d}\tau }}\left(\begin{array}{c}\delta p\\ \delta u\end{array}\right)=\left(\begin{array}{cc}\frac{\alpha +10}{\alpha +2}& \frac{4(\alpha +6)}{\alpha +2}\\ 1& 0\end{array}\right)\left(\begin{array}{c}\delta p\\ \delta u\end{array}\right).`$ (10)
Solutions to the equation $`\text{det}(M\lambda I)=0`$, where $`M`$ is the matrix defined above, are given by
$$\lambda _\pm =\frac{\alpha +10}{2(\alpha +2)}\pm \frac{i}{2(\alpha +2)}\sqrt{15\alpha ^2+108\alpha +92}.$$
(11)
The real part of $`\lambda _\pm `$ is always negative and the critical point is a spiral point. Therefore, every solution will tend to $`Q=Q_\mathrm{p}`$ after an intermediate regime: $`Q=Q_\mathrm{p}`$ is an attractor and no fine-tuning of the initial conditions is required.
Before reaching the attractor, the quintessence field undergoes different regimes that we are now going to describe. These regimes are in fact characterized by two physical quantities already introduced previously: the equation of state parameter $`\omega _\mathrm{Q}`$ and the sound velocity $`c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}`$. We study the case of an “overshoot”, in the terminology of Ref. , since this corresponds to initial conditions that are physically more relevant (in particular this includes the case of equipartition, i.e. $`\rho _\mathrm{Q}10^4\rho _\mathrm{B}`$ initially). We also assume that the background is radiation dominated $`\omega _\mathrm{B}=1/3`$.
Initially, the kinetic energy dominates the potential energy, i.e. $`Q^2/(2a^2)V(Q)`$. This means that the energy density redshifts as $`\rho _\mathrm{Q}1/a^6`$ and that the equation of state parameter is $`\omega _\mathrm{Q}=1`$. As a consequence, due to the constancy of $`\omega _\mathrm{Q}`$ and Eq. (4) (and also $`\omega _\mathrm{Q}1`$), we have $`c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}=1`$ as well. The scalar field itself evolves like
$$Q=Q_\mathrm{f}\frac{A}{a},$$
(12)
where $`Q_\mathrm{f}`$ and $`A`$ are constant. These constants are such that the term $`A/a`$ becomes rapidly small in comparison with the frozen value $`Q_\mathrm{f}`$ and we have the amusing situation that the field can be (almost) considered as frozen even if the kinetic energy still dominates. This is illustrated in Fig. 1.
As a consequence, during this regime the potential energy is also almost constant except at the very beginning. Using the definition of $`\omega _\mathrm{Q}`$ and $`c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}`$, see Eq. (3), we deduce that, during the kinetic regime, we have
$$\omega _\mathrm{Q}1a^6,c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}1a^5.$$
(13)
The fact that, in the parametrization adopted here, the scale factor is very small during the kinetic regime explains that there is no contradiction between these equations and the values of $`\omega _\mathrm{Q}`$ and $`c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}`$ deduced above.
Since the kinetic energy decreases while the potential energy is almost constant, the kinetic regime cannot last forever. When the potential energy becomes larger than the kinetic one, the equation of state parameter suddenly jumps from $`+1`$ to $`1`$ while the sound velocity still remains equal to $`+1`$ since Eq. (4) does not imply a change of this quantity in the case $`\omega _\mathrm{Q}=1`$. The fact that the equation of state parameter changes before the sound velocity is explained by Eq. (13). We call this regime the transition regime. During this regime, the kinetic energy still redshifts as $`1/a^6`$ and $`V(Q)`$ is approximately constant but of course now $`\rho _\mathrm{Q}V(Q)`$.
Due to the second of Eq. (13), the sound velocity has also to change at some later time. This implies that the quintessence field can no longer behave according to Eq. (12). This is the starting point of the potential regime. In order to study the behaviour of the system in this regime, we need to find an expression for the second derivative of the potential. Differentiating once the definition of the sound velocity, Eq. (3), we arrive at
$`{\displaystyle \frac{\mathrm{d}^2V(Q)}{\mathrm{d}Q^2}}`$ $`=`$ $`{\displaystyle \frac{3}{2}}H^2\{{\displaystyle \frac{1}{}}c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}`$ (14)
$`+`$ $`(c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}1)[{\displaystyle \frac{^{}}{^2}}{\displaystyle \frac{1}{2}}(3c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}+5)]\}.`$ (15)
No approximation has been made in the derivation of this relation. This formula generalizes Eq. (3) of Ref. . This formula will turn out to be very useful when we study the perturbations in the next section. With the scale factor given by Eqns. (8), this relation can be re-written as
$$\frac{2}{3H^2}\frac{\mathrm{d}^2V(Q)}{\mathrm{d}Q^2}=\frac{1}{}c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}3(c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}1)(\omega _\mathrm{B}+c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}+2).$$
(16)
In the regime we are interested in, the r.h.s of the previous formula is small. The only way to satisfy this relation is to ensure that the sound velocity changes to the constant $`c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}=2\omega _\mathrm{B}`$. This gives $`c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}=7/3`$ for the radiation dominated era. This evolution is displayed in Fig. 2.
The fact that the sound velocity is a constant implies that the factor $`(4a^2)/(3Q^{})\mathrm{d}V(Q)/\mathrm{d}Q`$ is also a constant. Therefore, the behaviour of the quintessence field is now given by
$$Q=Q_\mathrm{f}+Ba^4,$$
(17)
which implies that the kinetic energy redshifts as $`a^4`$.
Again this regime cannot last forever since the kinetic energy increases while the potential energy still remains constant. At some later time, both contributions become equal and $`\omega _\mathrm{Q}`$ and $`c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}`$ have to change once more. This is the end of the potential regime and the beginning of the tracking regime which has already been described above. The quantities $`p_\mathrm{Q}`$, $`\rho _\mathrm{Q}`$, $`V`$ and the kinetic energy reach a fixed ratio such that
$$\omega _\mathrm{Q}=c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}=\frac{2\alpha \omega _B}{2+\alpha }.$$
(18)
The definitions of the different regimes and the corresponding evolutions of the physically relevant quantities are summed up in Table II.
### C The equation of state parameter problem
The third question evoked previously was the question of the value of the parameter $`\omega _\mathrm{Q}`$ today. As already mentioned, this is an important issue since constraints on this quantity are already available. This problem is also solved by quintessence in the sense that we always have $`1<\omega _\mathrm{Q}<0`$. Here, however, it is relevant to distinguish between the Ratra-Peebles potential and the SUGRA potential. With the first potential, it seems difficult to reach sufficiently small value of $`\omega _\mathrm{Q}`$. On the other hand, this is automatically achieved in the second case. The reason for this is the presence of the factor $`\mathrm{exp}(\kappa Q^2/2)`$ in the potential, a generic feature of SUGRA-based potentials, which drives $`\omega _\mathrm{Q}`$ towards $`1`$. For $`\alpha =11`$ and $`\mathrm{\Omega }_{\mathrm{c}dm}0.3`$, the prediction is $`\omega _\mathrm{Q}0.82`$ a value in agreement with the current data .
### D The model building problem
From the particle physics point of view, one would like to justify the existence of the quintessence field and the shapes of the (so far) phenomenological potentials. Several attempts have already been made in the framework of supersymmetric field theory. In particular, it was shown by Binétruy that the Ratra-Peebles potential can be recovered in the context of global SUSY. However, as already mentioned, SUGRA corrections must be taken into account and this implies that the corresponding potential can be of the type of the SUGRA tracking potential displayed in Eq. (1) which leads to a better agreement with the available data.
Nevertheless, it should be clear that considerable problems remain to be addressed in order to reach a satisfactory situation . Maybe the most crucial question is the problem of SUSY breaking. SUSY must certainly be broken but the models evoked previously do not take into account this basic fact. This could have dramatic consequences and modify the shape of the potential which is so important in order to solve the three previous problems.
## III The cosmological perturbations
We now turn to the study of the cosmological perturbations. A detailed study has already been performed by Ratra and Peebles in Ref. but only for the tracking regime. Cosmological perturbations in a fluid with a constant negative equation of state parameter have been investigated in Ref. . In this article, we study the cosmological perturbations (in the long wavelength approximation) in all the regimes previously described and point out some additional properties. The evolution of the cosmological perturbations mainly depends on the equation of state parameter and the sound velocity. We have shown in the previous section that they can be considered as constant in each regime. This will simplify the analysis a lot.
The fate of the perturbations depends on the initial conditions. It has been noticed for the first time in Ref. that “the observable fluctuation spectrum is insensitive to a broad range of initial conditions, including the case in which the amplitudes of $`\delta Q`$, $`\delta Q^{}`$ are set by inflation”. In that paper, the authors choose $`\delta Q=\delta Q^{}=0`$ initially (in the synchronous gauge). We demonstrate, in this section, that the insensitivity of the spectrum described in Ref. has an origin similar to the insensitivity of the background properties with respect to the initial conditions $`Q`$ and $`Q^{}`$, namely the presence of an attractor for the perturbed quantities. We prove that during all the four regimes undergone by the quintessence field, the attractor is characterized by a “spiral fixed point” as it is the case for the background.
### A General framework
Without loss of generality, the perturbed line element can be written in the synchronous gauge. In this class of coordinates systems, scalar perturbations are completely described by two arbitrary functions. The spatial dependence of the perturbations is given by $`X(x^i)`$ which is the eigenfunction of the Laplace operator on the flat spacelike hypersurfaces. There exists two ways to construct a two rank tensor from a scalar function : either by multiplying it by the spatial background flat metric $`\delta _{ij}`$ or by differentiating it twice. The two arbitrary functions mentioned above are simply the coefficients of these two terms in a Fourier expansion. Therefore, the perturbed metric can be expressed as
$`\mathrm{d}s^2=a^2(\eta )\{`$ $``$ $`\mathrm{d}\eta ^2+[(1+h(\eta )X)\delta _{ij}`$ (19)
$`+`$ $`h_l(\eta ){\displaystyle \frac{1}{k^2}}X_{,i,j}]\mathrm{d}x^i\mathrm{d}x^j\}.`$ (20)
In this equation, the dimensionless quantity $`𝐤`$ is the comoving wavevector related to the physical wavevector $`𝐤^{\mathrm{phys}}`$ through the relation $`𝐤^{\mathrm{phys}}𝐤/a(\eta )`$. As a consequence of Einstein equations, perturbations in the metric are coupled to perturbations in the different matter components. We choose to write the perturbed stress-energy tensor according to
$`T^0_0`$ $`=`$ $`{\displaystyle \frac{ϵ_1}{a^2}}X,T^0{}_{i}{}^{}={\displaystyle \frac{\xi ^{}}{a^2}}X_{,i},T^i{}_{0}{}^{}={\displaystyle \frac{\xi ^{}}{a^2}}X^{,i},`$ (21)
$`T^i_j`$ $`=`$ $`{\displaystyle \frac{p_1}{a^2}}X\delta ^i{}_{j}{}^{},`$ (22)
where we have assumed that the longitudinal pressure $`p_l`$ vanishes for each component. As for the background, one considers that the Universe is filled with two fluids: the background fluid, an hydrodynamical perfect fluid which is either radiation or dust (again, the corresponding quantities will carry the index $`\mathrm{B}`$) and a scalar field $`Q`$ describing the quintessence field (in this case the corresponding quantities will carry the index $`\mathrm{Q}`$). The perturbed Einstein equations which govern the evolution of the quantities $`h`$ and $`h_l`$ are given by:
$`3h^{}+k^2hh_l^{}`$ $`=`$ $`\kappa ϵ_{1\mathrm{B}}+\kappa ϵ_{1\mathrm{Q}},`$ (23)
$`h^{}`$ $`=`$ $`\kappa \xi _\mathrm{B}^{}+\kappa \xi _\mathrm{Q}^{},`$ (24)
$`h^{\prime \prime }2h^{}`$ $`=`$ $`\kappa p_{1\mathrm{B}}+\kappa p_{1\mathrm{Q}},`$ (25)
$`h_l^{}+2h_l^{}k^2h`$ $`=`$ $`0.`$ (26)
Finally, it turns out to be more convenient to work with the density contrast $`\delta `$ and the velocity divergence $`\theta `$ defined by the equations:
$$\delta \frac{ϵ_1}{a^2ϵ_0},\xi ^{}\frac{a^3ϵ_0}{k^2}(1+\omega )\theta .$$
(27)
In the following, we study analytically the time evolution of the density contrast for the background fluid and for quintessence in the long wavelength limit.
### B The background fluid
The equations satisfied by the background density contrast and divergence can be obtained either from combinations of the Einstein equations (23-26) or, more directly, from the conservation of the perturbed background fluid stress-energy tensor (since the background fluid and quintessence only interact gravitationally). They read:
$`\delta _\mathrm{B}^{}+a(1+\omega _\mathrm{B})\theta _\mathrm{B}+{\displaystyle \frac{1+\omega _\mathrm{B}}{2}}(3h^{}h_l^{})`$ $`=`$ $`0,`$ (28)
$`\theta _\mathrm{B}^{}+(23\omega _\mathrm{B})\theta _\mathrm{B}{\displaystyle \frac{k^2c_{\mathrm{s}}^{2}{}_{B}{}^{}}{(1+\omega _\mathrm{B})a}}\delta _\mathrm{B}`$ $`=`$ $`0.`$ (29)
These two equations are equivalent to Eqns. (7.15) and (7.16) of Ref. . From them, we can derive the relation
$$3h^{\prime \prime }h_l^{\prime \prime }=\frac{2}{1+\omega _\mathrm{B}}\delta _\mathrm{B}^{\prime \prime }+2(13\omega _\mathrm{B})a^{}\theta _\mathrm{B}\frac{2k^2c_{\mathrm{s}}^{2}{}_{B}{}^{}}{1+\omega _\mathrm{B}}\delta _\mathrm{B},$$
(30)
where we have assumed that $`\omega _\mathrm{B}`$ is a constant. On the other hand, from the Einstein equations we get
$`(3h^{\prime \prime }h_l^{\prime \prime })(3h^{}h_l^{})`$ (31)
$`=3^2\left[(1+3c_\mathrm{}\mathrm{B}^2)\mathrm{\Omega }_\mathrm{B}\delta _\mathrm{B}+(1+3c_\mathrm{}\mathrm{Q}^2)\mathrm{\Omega }_\mathrm{Q}\delta _\mathrm{Q}\right],`$ (32)
where $`c_\mathrm{}\mathrm{Q}^2p_{1\mathrm{Q}}/ϵ_{1\mathrm{Q}}`$ which needs not to coincide with the definition of $`c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}`$. In order to derive the formula satisfied by the density contrast of the background fluid in the long wavelength limit, we neglect the term proportional to $`k^2`$ in Eq. (30) and we use the fact that $`\mathrm{\Omega }_\mathrm{Q}\mathrm{\Omega }_\mathrm{B}`$. Then, straightforward manipulations lead to
$`\delta _\mathrm{B}^{\prime \prime }+\delta _\mathrm{B}^{}`$ $``$ $`{\displaystyle \frac{3}{2}}^2(1+3\omega _\mathrm{B})(1+\omega _\mathrm{B})\delta _\mathrm{B}`$ (33)
$`=`$ $`3\omega _\mathrm{B}(1+\omega _\mathrm{B})a\theta _\mathrm{B},`$ (34)
where we used the fact that $`c_\mathrm{}\mathrm{B}^2=\omega _\mathrm{B}`$ for an hydrodynamical fluid. This equation shows that the evolution of the background density contrast is essentially unaffected by the presence of quintessence. This is of course an expected result since we have assumed $`\mathrm{\Omega }_\mathrm{Q}\mathrm{\Omega }_\mathrm{B}`$. The general solution to Eq. (33) can be easily found and reads
$`\delta _\mathrm{B}(\eta )=A_1\left({\displaystyle \frac{a}{a_0}}\right)^{x_+}+A_2\left({\displaystyle \frac{a}{a_0}}\right)^x_{}`$ (35)
$`+`$ $`{\displaystyle \frac{\omega _\mathrm{B}(1+\omega _\mathrm{B})(1+3\omega _\mathrm{B})a_0\theta _{\mathrm{B0}}\eta _0}{(1\omega _\mathrm{B})(1+6\omega _\mathrm{B})}}\left({\displaystyle \frac{a}{a_0}}\right)^{(9\omega _\mathrm{B}1)/2},`$ (36)
where we have defined
$`x_\pm `$ $``$ $`{\displaystyle \frac{(13\omega _\mathrm{B})}{4}}`$ (38)
$`\pm {\displaystyle \frac{1}{4}}\sqrt{(13\omega _\mathrm{B})^2+24(1+\omega _\mathrm{B})(13\omega _\mathrm{B})}.`$
The results for the radiation dominated and matter dominated epochs are summarized in Table III.
These results are consistent with those obtained in Ref. . In particular, it can be shown that the branch $`\delta _\mathrm{B}a^x_{}`$ corresponds in fact to a residual gauge mode, i.e. there exists a synchronous system of coordinates such that this mode can be removed and therefore must not be considered as a physical mode.
### C Quintessential perturbations
We now describe how the long wavelength quintessential perturbations evolve with time. A similar study has already been performed by Ratra and Peebles but only on the tracking solution. We give here a complete description of the evolution of the quintessence density contrast in the four regimes defined in the previous section. In addition, we prove that there exists an attractor for the perturbations as it is the case for the background solution. As a consequence, the final value of the density contrast is always the same whatever the initial conditions are.
In order to obtain the fundamental equations to be solved, we can proceed as for the background fluid. However, it is important to notice that the link between the perturbed energy density and the perturbed pression, which is just a constant for the background fluid, is more complicated in the case of quintessence. In general, we can write $`p_{1\mathrm{Q}}=c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}ϵ_{1\mathrm{Q}}+a^2\tau \delta S`$ where the second term represents entropy perturbations. In the synchronous gauge, we obtain
$$p_{1\mathrm{Q}}=c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}ϵ_{1\mathrm{Q}}+(1c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{})\frac{1}{\kappa }(h_l^{\prime \prime }+h_l^{}).$$
(39)
We can now establish the equations satisfied by the quintessence density contrast and divergence. The conservation of the perturbed stress-energy tensor leads to
$`\delta _\mathrm{Q}^{}+3\alpha (c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}\omega _\mathrm{Q})\delta _\mathrm{Q}+a(1+\omega _\mathrm{Q})\theta _\mathrm{Q}`$ (40)
$`+`$ $`{\displaystyle \frac{1}{2}}(1+\omega _\mathrm{Q})(3h^{}h_l^{})={\displaystyle \frac{^1}{\mathrm{\Omega }_\mathrm{Q}}}(c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}1)(h_l^{\prime \prime }+h_l^{}),`$ (42)
$`\theta _\mathrm{Q}^{}+(23\omega _\mathrm{Q})\alpha \theta _\mathrm{Q}{\displaystyle \frac{k^2c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}}{(1+\omega _\mathrm{Q})a}}\delta _\mathrm{Q}`$
$`=`$ $`{\displaystyle \frac{\omega _\mathrm{Q}^{}}{1+\omega _\mathrm{Q}}}\theta _\mathrm{Q}+{\displaystyle \frac{(1c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{})k^2}{3a^2(1+\omega _\mathrm{Q})\mathrm{\Omega }_\mathrm{Q}}}(h_l^{\prime \prime }+h_l^{}).`$ (43)
In these two equations, no approximation have been made: they are valid for any wavenumber, any equation of state parameter and sound velocity. In practice, it turns out to be more convenient to use the perturbed Klein-Gordon equation to analyse the problem. This can be obtained directly from the first of the two previous equations if one notices that the quantities describing the perturbed scalar field stress-energy tensor can be expressed in terms of the perturbed scalar field $`\delta Q(\eta ,𝐱)`$ according to
$`ϵ_{1\mathrm{Q}}`$ $`=`$ $`Q^{}\delta Q^{}+a^2\delta Q{\displaystyle \frac{\mathrm{d}V(Q)}{\mathrm{d}Q}},`$ (44)
$`\xi _\mathrm{Q}^{}`$ $`=`$ $`Q^{}\delta Q,`$ (45)
$`p_{1\mathrm{Q}}`$ $`=`$ $`Q^{}\delta Q^{}a^2\delta Q{\displaystyle \frac{\mathrm{d}V(Q)}{\mathrm{d}Q}}.`$ (46)
Inserting the corresponding expression for the density contrast and the divergence in Eq. (40), we get
$$\delta Q^{\prime \prime }+2\delta Q^{}+\left[k^2+a^2\frac{\mathrm{d}^2V(Q)}{\mathrm{d}Q^2}\right]\delta Q+\frac{Q^{}}{2}(3h^{}h_l^{})=0.$$
(47)
This is similar to Eq. (7.20) of Ref. . One can check that Eq. (42) is automatically verified since it is equivalent to the unperturbed Klein-Gordon equation (times an unimportant factor). Using Eq. (28) to express the factor $`3h^{}h_l^{}`$ and neglecting the $`k^2`$ term, we arrive at
$$\delta Q^{\prime \prime }+2\delta Q^{}+a^2\frac{\mathrm{d}^2V(Q)}{\mathrm{d}Q^2}\delta Q=Q^{}a\theta _\mathrm{B}+\frac{Q^{}}{1+\omega _\mathrm{B}}\delta _\mathrm{B}^{}.$$
(48)
We are now going to analyse this equation in detail. We now need to utilize the general expression for the second derivative of the potential, Eq. (14). On the tracking solution, we have $`\omega _\mathrm{Q}=c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}`$ and $`\omega _\mathrm{Q}=(2+\alpha \omega _\mathrm{B})/(2+\alpha )`$ and this equation reduces to
$$\frac{\mathrm{d}^2V(Q)}{\mathrm{d}Q^2}=\frac{9}{2}H^2\frac{\alpha +1}{\alpha }(1\omega _\mathrm{Q}^2).$$
(49)
For our purpose, as proven in the previous section, it is sufficient to consider a regime where $`c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}`$ is constant and where the scalar field is a test field. Under these conditions, we obtain
$$\frac{\mathrm{d}^2V(Q)}{\mathrm{d}Q^2}=\frac{3}{4}H^2(1c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{})(6+3\omega _\mathrm{B}+3c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}).$$
(50)
Let us now concentrate on the homogeneous part of Eq. (48). Using the previous equation, it can be expressed as
$`\delta Q^{\prime \prime }+{\displaystyle \frac{4}{1+3\omega _\mathrm{B}}}{\displaystyle \frac{1}{\eta }}\delta Q^{}`$ (51)
$`+`$ $`{\displaystyle \frac{3}{(1+3\omega _\mathrm{B})^2}}{\displaystyle \frac{1}{\eta ^2}}(1c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{})(6+3\omega _\mathrm{B}+3c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{})\delta Q=0.`$ (52)
This linear equation can easily be solved: its solutions are just power law of the conformal time. However, in order to show explicitly the complete analogy with the background attractor, we choose to analyse it in a rather roundabout way. Let us proceed exactly as for the unperturbed Klein-Gordon equation \[see the discussion around Eq. (10)\]. We define the time $`\tau `$ by $`\eta e^\tau `$ and introduce the quantity $`\delta u`$ and $`\delta p`$ defined by $`\delta u\delta Q`$ and $`\delta p\mathrm{d}(\delta Q)/\mathrm{d}\tau `$. Then, Eq. (51) can re-expressed as
$`{\displaystyle \frac{\mathrm{d}}{\mathrm{d}\tau }}\left(\begin{array}{c}\delta p\\ \delta u\end{array}\right)`$ (53)
$`=\left(\begin{array}{cc}\frac{3(\omega _\mathrm{B}1)}{1+3\omega _\mathrm{B}}& \frac{9(c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}1)}{(1+3\omega _\mathrm{B})^2}(2+\omega _\mathrm{B}+c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{})\\ 1& 0\end{array}\right)\left(\begin{array}{c}\delta p\\ \delta u\end{array}\right).`$ (54)
The form of this equation clearly shows the complete analogy with Eq. (10). The eigenvalues of the system are found by solving the equation $`\text{det}(M\lambda I)=0`$, where $`M`$ is the matrix defined above and $`I`$ the identity matrix. Straigthforwards calculations show that the solutions are given by
$$\lambda _\pm =\frac{3}{2}\frac{\omega _\mathrm{B}1}{1+3\omega _\mathrm{B}}\left[1\pm \sqrt{1+4\frac{c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}1}{(\omega _\mathrm{B}1)^2}(2+\omega _\mathrm{B}+c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{})}\right].$$
(55)
Of course, this is just a simple rephrasing of the fact that the solution of Eq. (51) is $`\delta QA_+\eta ^{\lambda _+}+A_{}\eta ^\lambda _{}`$. The presence of an attractor is linked to the negative sign of the real part of $`\lambda _\pm `$. It is easy to see that the real part is always negative in all four regimes, in particular this is true for any value of $`\alpha `$. This is displayed in Fig. 3 in the plane $`(\omega _\mathrm{B},c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{})`$. The green and purple regions are the regions where these real parts are negative. The green region is the region where the argument of the square root is negative, i.e. where the square root is an imaginary number. The exact “trajectories” of the system for the usual tracking potential (short line) and for the SUGRA tracking potential (long line) are also shown for the case $`\alpha =11`$. They have been obtained by full numerical integration. The remarkable property is that these trajectories are always in the stable region. This means that, in each region, the system tends to an attractor which is given by the inhomogeneous part of the perturbed Klein-Gordon equation. The system starts at $`\omega _\mathrm{B}=1/3`$ and goes from $`c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}=1`$ to $`c_{\mathrm{s}}^{2}{}_{\mathrm{Q}}{}^{}=7/3`$. Then, the system approaches the transition to the matter dominated era and leaves the vertical line. Finally, it stops when the redshift vanishes at $`\omega _\mathrm{B}0.29`$ for the tracking potential and at $`\omega _\mathrm{B}0.82`$ for the SUGRA tracking potential. The two lines separate when the exponential factor becomes important in the SUGRA tracking potential.
The conclusion is that the final value of the quintessence perturbations is insensitive to the initial conditions, a property completely similar to what has been shown in Ref. for the background. Strictly speaking, this property has been demonstrated for long wavelength modes only. However, we have checked by numerical calculations that this is also true for shorter wavelength modes. Having proven that the final result does not depend on the initial conditions of the quintessence perturbations, we can now proceed further and embark in a rather detailed study of the CMB anisotropies predictions in the presence of quintessence.
## IV Predictions for the power spectrum and the multipole moments
The presence of cosmological perturbations induces directional variations in the CMB photon redshift. This is the so-called Sachs-Wolfe effect . Since these variations are the same regardless of the wavelength of the photons, they translate into variations in the temperature of the black body on the celestial sphere. Their amplitude has been measured by the COBE satellite and is of the order of magnitude $`\delta T/T_010^5`$ . The detailed angular structure of the CMB anisotropies is usually characterized by the two-point correlation function which can be expanded according to
$$\frac{\delta T}{T}(𝐞_1)\frac{\delta T}{T}(𝐞_2)=\frac{1}{4\pi }\underset{l}{}(2l+1)C_lP_l(\mathrm{cos}\gamma ),$$
(56)
where $`\gamma `$ is the angle between the directions $`𝐞_1`$ and $`𝐞_2`$ and $`P_l`$ is a Legendre polynomial. The coefficients $`C_l`$ are the multipole moments. In what follows, we will be mainly interested in the so-called band power $`\delta T_l`$ defined by the following expression
$$\delta T_lT_0\sqrt{l(l+1)\frac{C_l}{2\pi }},$$
(57)
where $`T_02.7\mathrm{K}`$. The band power has now been measured on a wide range of angular scales from $`10^{}`$ to $`90^{}`$ corresponding roughly to $`l[2,700]`$. Almost $`80`$ data points have been measured. Recently new data obtained by the balloon-borne experiments BOOMERanG and MAXIMA-1 have been published. They clearly show a detection of the first Doppler peak at the expected angular scale $`1^{}`$ corresponding to the size of the Hubble radius at recombination.
On the theoretical side, the multipoles moments depend on the initial spectra for scalar and tensor modes and on how the perturbations evolve from the initial time (after inflation) until now. This evolution is determined by the values of the cosmological parameters, i.e. by the value of the Hubble constant ($`h`$), of the total amount of matter present in our Universe ($`\mathrm{\Omega }_0`$), of the cosmological constant ($`\mathrm{\Omega }_\mathrm{\Lambda }`$), of the baryons density parameter ($`\mathrm{\Omega }_\mathrm{b}`$) and of the cold dark matter density parameter ($`\mathrm{\Omega }_{\mathrm{c}dm}`$). Constraints already exist on some of these parameters. In particular, as already mentioned above, $`\mathrm{\Omega }_\mathrm{\Lambda }0.7`$ according to the SNIa measurements and $`h^2\mathrm{\Omega }_\mathrm{b}0.019\pm 0.002`$ according to BBN . We also assume $`\mathrm{\Omega }_0=1`$ in agreement with the inflation paradigm which has been confirmed by the recent CMB anisotropy measurements. For the initial spectra, it is traditional to assume that they are of the power-law form
$$k^3P_\mathrm{\Phi }(k)=A_\mathrm{S}k^{n_\mathrm{S}1},k^3P_\mathrm{h}(k)=A_\mathrm{T}k^{n_\mathrm{T}},$$
(58)
where the scalar and tensor spectral indices $`n_\mathrm{S}`$ and $`n_\mathrm{T}`$ are related by $`n_\mathrm{S}1=n_\mathrm{T}`$. This last equation is also valid for zeroth order slow-roll inflation. It should be noticed that, a priori, this choice is not the most relevant one since slow-roll inflation is certainly more physically motivated. For spectral indices close to $`n_\mathrm{S}=1`$, we expect a small difference. This is no longer true for larger tilts. Inflation predicts the presence of gravitational perturbations and the tensor to scalar amplitude ratio is given by
$$\frac{A_\mathrm{T}}{A_\mathrm{S}}\frac{200}{9}n_\mathrm{T}.$$
(59)
This equation is valid for power-law inflation with $`n_\mathrm{T}`$ not too large<sup>*</sup><sup>*</sup>*For power-law inflation, the exact expression is given by $`A_\mathrm{T}/A_\mathrm{S}=(200/9)n_\mathrm{T}/(1n_\mathrm{T}/2)`$. or for zeroth order slow-roll inflation. A last remark is in order at this point. All the plots displayed in this article are COBE normalized in the following way: the position of the Sachs-Wolfe plateau is tuned such that it best fits the COBE data points. In practice, this almost amounts to normalize the spectrum to $`C_{10}`$.
In this section, we first study the general properties of the multipoles moments in the quintessence cold dark matter model (QCDM) and point out the main differences with the standard cold dark matter (sCDM) and the cosmic concordance model ($`\mathrm{\Lambda }`$CDM). We also display the corresponding baryonic matter power spectra, given by
$$|\delta (k)|^2\left|\frac{\delta \rho _\mathrm{b}}{\rho _\mathrm{b}}\right|^2,$$
(60)
which is the square of the Fourier transform of the baryonic density contrast. Then, we compare the predictions of the QCDM model for the Ratra-Peebles and SUGRA tracking potentials with the COBE , BOOMERanG , MAXIMA-1 and Saskatoon data. We do not attempt to perform a detailed statistical analysis but we rather indicate roughly how the different models can fit the observational data.
We now turn to simple considerations about the shape of the CMB spectrum. The corresponding band power for the Ratra-Peebles and SUGRA potentials are displayed in Figs. 4 and 5 for $`h=0.5`$, $`\mathrm{\Omega }_\mathrm{b}=0.05`$, $`\mathrm{\Omega }_\mathrm{Q}=0.7`$, $`\mathrm{\Omega }_{\mathrm{c}dm}=1\mathrm{\Omega }_\mathrm{b}\mathrm{\Omega }_\mathrm{Q}`$, $`n_\mathrm{S}=0.99`$ and the tensor contribution neglected. The former set of cosmological parameters has been chosen just for the sake of illustration and discussion.
For simplicity, we start with a comparison of the quintessence multipole moments with those obtained in the $`\mathrm{\Lambda }`$CDM model with similar cosmological parameters. Firstly, since $`\mathrm{\Omega }_\mathrm{m}\mathrm{\Omega }_{\mathrm{c}dm}+\mathrm{\Omega }_\mathrm{b}`$ is the same in the two models, the redshift of equivalence between matter and radiation $`z_{\mathrm{eq}}\mathrm{\Omega }_\mathrm{m}/\mathrm{\Omega }_\mathrm{r}`$, where $`\mathrm{\Omega }_\mathrm{r}\mathrm{\Omega }_\gamma +\mathrm{\Omega }_\nu `$, is also the same in both cases. Therefore, the first peak is boosted in the same way by the early integrated Sachs-Wolfe effect (due to the time variation of the two Bardeen potentials during recombination, see ) and, a priori, one expects the same first peak height. Secondly, the dark energy component (cosmological constant or quintessence) has a negligible contribution before recombination and, as a consequence, the evolution of the perturbations before the last scattering surface is the same in the two models (see the previous section). Thus, one expects again identical acoustic peak patterns. However, despite the previous considerations, the position of the peaks differs because the angular distance-redshift relation is modified at small redshift since the equation of state of the cosmological constant and of quintessence is not the same. The closest to $`1`$ the equation of state parameter is, the largest the shift of the peaks to small angular scales is. As a consequence, the peaks in the $`\mathrm{\Lambda }`$CDM model are more shifted to the right than in the QCDM model. Another feature is that the height of the first peak is not the same in the two types of scenarios. Indeed, at small redshift, the gravitational potential does not behave exactly in the same way in the two models especially because there are scalar field perturbations in the QCDM scenario. This results in a different contribution of the late integrated Sach-Wolfe effect which affects the overall normalization of the spectrum. As a consequence, the height of the first peak is lower in the model which produces a strong late integrated Sachs-Wolfe effect, i.e. in the QCDM model.
The exact shape of the quintessence potential also matters and different potentials lead to different CMB anisotropies. The SUGRA potential and the cosmological constant lead to very similar CMB anisotropy spectra, whereas the difference is stronger in the case of the Ratra-Peebles potential. This is mainly due to the fact that the equation of state parameter is generically closer to $`1`$ in the first case than in the second one. Another difference is that the Ratra-Peebles potential produces a larger late integrated Sachs-Wolfe contribution than the SUGRA potential. This results in a different normalization for both models (note that the normalisation depends on $`\alpha `$) which has for consequence different height of the first Doppler peak. Of course, this difference is also visible in the power spectrum at large scales. Maybe the most interesting property is the following one. The cosmic equation of state (almost) does not depend on $`\alpha `$ in the case of the SUGRA potential. Then, in the same manner, the CMB anisotropies do not depend on $`\alpha `$ contrary to the case of the Ratra-Peebles potential. This means that the multipole moments displayed in Fig. 5 are a generic predictions of the SUGRA QCDM model.
For the sake of completness, let us now describe the corresponding matter power spectra. They are displayed in Fig. 6 and 7.
The matter power spectrum also depends on the nature of the dark energy component (cosmological constant or quintessence) but the difference between the cosmological constant scenario and a quintessence scenario is less important. The matter power spectrum shows a peak the location of which is given by the Hubble radius at equivalence. In the $`\mathrm{\Lambda }`$CDM and QCDM scenarios, the peak is at the same location contrary to the sCDM case for which the peak is located at smaller scales. Also, in models with low matter content, the ratio $`\mathrm{\Omega }_\mathrm{b}/\mathrm{\Omega }_{\mathrm{c}dm}`$ is higher which results in the presence of smooth oscillations at small scales. As for the CMB anisotropy spectrum, the small scales are similar in the $`\mathrm{\Lambda }`$CDM and QCDM scenarios and important differences only occur on larger scales which are more affected by the change in the cosmic equation of state.
Let us now study in more details and for more realistic values of the cosmological parameters, the position and the height of the first Doppler peak. We start with the location of the first peak (denoted in what follows by $`l_1`$) and we study it in the plane $`(\mathrm{\Omega }_\mathrm{m},h)`$ with the following values of the other cosmological parameters: $`h^2\mathrm{\Omega }_\mathrm{b}=0.019`$ (the value predicted by standard BBN), $`\mathrm{\Omega }_{\mathrm{\Lambda },\mathrm{Q}}=0.7`$, $`A_\mathrm{T}=0`$ and $`n_\mathrm{S}=0.99`$. The case of the cosmological constant is displayed in Fig. 8, the case of the Ratra-Peebles QCDM model in Fig. 9 and the case of the SUGRA QCDM model in Fig. 10.
These plots confirm the qualitative predictions made above and in particular the fact that, in general, $`l_1^\mathrm{\Lambda }>l_1^{\mathrm{S}UGRA}>l_1^{\mathrm{R}P}`$. If one assumes that $`\mathrm{\Omega }_\mathrm{m}0.3`$ (since we have assumed $`\mathrm{\Omega }_\mathrm{\Lambda }0.7`$) and $`h0.62`$, this last value being consistent with the Hubble Space Telescope (HST) and SNIa measurements, then we obtain $`l_1^\mathrm{\Lambda }225`$, $`l_1^{\mathrm{S}UGRA}220`$ and $`l_1^{\mathrm{R}P}200`$. It is interesting to compare these values with the recent measurements of the first peak performed by BOOMERanG and MAXIMA-1. The BOOMERanG data indicate that $`l_1=197\pm 6`$ which is compatible with the Ratra-Peebles potential and a spatially flat Universe. On the other hand, the MAXIMA-1 data are consistent with a first peak located at $`l_1220`$ which is, this time, in agreement with a cosmological constant or the SUGRA QCDM model.
Let us now study the height of the first Doppler peak. We study its variation in the plane $`(\mathrm{\Omega }_\mathrm{b},n_\mathrm{S})`$ for the following values of the cosmological parameters: $`h=0.62`$, $`\mathrm{\Omega }_{\mathrm{\Lambda },\mathrm{Q}}=0.7`$. The case of the $`\mathrm{\Lambda }`$CDM model is displayed in Fig. 11 whereas the cases of the Ratra-Peebles QCDM and SUGRA QCDM are presented in Figs. 12 and 13, respectively.
We would like to emphasize that the importance of gravitational waves is crucial in this case. Indeed, as already mentioned, the presence of gravitational waves modifies the normalization and, as a consequence, the height of the peaks. The BOOMERanG data indicate that $`\delta T_{200}69\pm 8\mu \mathrm{K}`$ whereas the MAXIMA-1 ones give $`\delta T_{220}78\pm 6\mu \mathrm{K}`$ , this discrepancy being possibly explained by problems in the calibration of these experiments. If we adopt the value $`\mathrm{\Omega }_\mathrm{b}0.0595`$, compatible with BBN, we see that, in the Ratra-Peebles QCDM model, a height of the first peak compatible with the BOOMERanG and MAXIMA-1 data leads to a value of the scalar spectral index such that $`n_\mathrm{S}>1`$. This is not compatible with standard inflation and cannot be realized with one scalar field. We interpret this as a new evidence that the Ratra-Peebles QCDM model (at least with this value of $`\alpha `$) is excluded by the observations. For the cases of $`\mathrm{\Lambda }`$CDM and SUGRA QCDM, we learn from the previous plots that the spectral index must be very close to one.
We now turn to the study of the second Doppler peak. First of all, we should say something about the observational situation. With regards to the detection of a second peak, it is difficult to deduce something from the BOOMERanG data. The error bars are still large and the data are, for the moment, compatible with a second peak (with a height maybe smaller than predicted by standard inflation) but also with no peak at all, even if one can see a small rise of the signal at $`l_2550`$ . Only $`5\%`$ of the data of this experimement have been analysed so far and one should wait for the rest of the data analysis to be completed. On the other hand, the MAXIMA-1 show “a suggestion of a peak at $`l_2525`$” the height of which would be $`\delta T_{525}48\mu \mathrm{K}`$. One could even argue that the beginning of a third peak has been observed. In fact, considering all the uncertainties in such measurements, we are of the opinion that a reasonable attitute is simply to wait for more data. On the theoretical side, it was argued by Kamionkowski and Buchalter that the location of the second peak can probe the dark energy density. The main idea is to study the contour plots of $`l_2`$ in the plane $`(\mathrm{\Omega }_\mathrm{m},h)`$. Then, a measurement of $`l_2`$, knowing $`h`$ by other means , immediately determines the value of $`\mathrm{\Omega }_\mathrm{m}`$. It was claimed in Ref. that this strategy does not depend on whether the dark energy is a cosmological constant or a quintessence field. We show that this claim is not correct and that the nature of the dark energy matters. The contour plots of $`l_2`$ in the case of a cosmological constant are displayed in Fig. 14 for the cosmological parameters given by $`h=0.62`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.7`$, $`h^2\mathrm{\Omega }_\mathrm{b}=0.019`$, $`n_\mathrm{S}=0.99`$. These plots are in agreement with the results found in Ref. .
The corresponding contour plots for the Ratra-Peebles and SUGRA CDM models are presented in Figs. 15 and 16. In addition, in order to show that there is indeed an important difference, we also display the contour plots for a cosmological constant which, for a given value of $`l_2`$, is always above the QCDM curve. The fact that there is a difference does not totally invalidate the idea of Ref. . But it means that, in order to use it, we should first identify the physical nature of the dark energy, for example with a measurement of its equation of state parameter.
As for the first peak, we have $`l_2^\mathrm{\Lambda }>l_2^{\mathrm{S}UGRA}>l_2^{\mathrm{R}P}`$. Roughly speaking, for $`h0.62`$, $`\mathrm{\Omega }_\mathrm{m}0.3`$, we have $`l_2^\mathrm{\Lambda }550`$, $`l_2^{\mathrm{S}UGRA}525`$ and $`l_2^{\mathrm{R}P}500`$. Interestingly enough, the SUGRA QCDM model seems to predict the correct location of the “suggested second peak” , just in between the location predicted by the $`\mathrm{\Lambda }`$CDM model and the Ratra-Peebles QCDM model. Of course, it is premature to conclude and only more data could allow to know whether this is indeed the case or whether this is just a coincidence.
Finally, we display the multipole moments for the $`\mathrm{\Lambda }`$CDM model, the Ratra-Peebles QCDM model and the SUGRA QCDM model in Figs. 1718 and 19, respectively, for the following cosmological parameters (deduced from the previous considerations): $`h=0.62`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.7`$, $`\mathrm{\Omega }_\mathrm{b}=0.0595`$ and $`n_\mathrm{S}=0.99`$. The data points of COBE, BOOMERanG, MAXIMA-1 and Saskatoon have been added to the plots for comparison. These curves represent the predictions of each model and special attention must be paid to third peak which is certainly one of the next important experimental challenge. In Fig. 20, we present the three curves together in order to make the comparison easier. It should be emphasized again that the multipole moments predicted by the SUGRA QCDM model are unique in the sense that they do not depend on the free parameter in the potential. From these plots, we see that the SUGRA QCDM model is, among the three models studied here, the best fit of the MAXIMA-1 data. It is the only model for which the theoretical curve $`\delta T`$ versus $`l`$ goes through all the $`1\sigma `$ error bars of this experiment. However, we should be careful not to overestimate the relevance of this result since uncertainties are still large, for instance because the comparison of the calibrations of different experiments is always a difficult task. We should also keep in mind that $`2\sigma `$ deviations are always possible. Thus, we are waiting eagerly for the new data to see whether quintessence, and especially SUGRA quintessence, can confirm the hints of this article and fits the data better than the other QCDM models.
## V conclusion
The quintessence scenario provides a general framework within which the issue of the energy density of the Universe can be tackled. In particular long-standing issues such as coincidence problem (and maybe the fine-tuning problem) receive reasonnable answers for a class of models possessing the property of tracking fields, i.e. the evolution of the quintessence field is driven at small redshift towards an attractor independently of the initial conditions. In the same spirit it seems very enticing to draw consequences of the quintessence hypothesis on other cosmological observables, the most prominent ones being the cosmological anisotropies. Recent measurement of the CMB anisotropies by the BOOMERanG and MAXIMA-1 experiments give a first indication on the location of the peaks in the CMB multipoles. It seems therefore topical to understand the consequence of the quintessence hypothesis on the CMB anisotropies.
In this paper we have confronted analytical methods with numerical results. Using the former we establish that the quintessence perturbation are independent of the initial conditions. This is confirmed by a full numerical computation. This allows us to study the CMB anisotropies. In particular we have paid particular attention to the comparison between three possible models: the cosmological constant model, the Ratra-Peebles and SUGRA quintessence models. We have also compared these three models with the existing data from the BOOMERanG and MAXIMA-1 experiments. As a rule the location of the first peak is shifted to the right for models having an equation of state $`\omega `$ closer to $`1`$. This entails that the location of the first peak for the first peak of the MAXIMA-1 data is fitted by the SUGRA model. Similarly the location of the second peak around $`l_2525`$ as suggested by MAXIMA-1 seem to indicate that the SUGRA model comes closer to be the best of these three models. One of the foreseeable challenges will be to carry out a thorough analysis of the forthcoming data in order to distinguish these three models even more clearly.
From the particle physics point of view most of the quintessence models discussed so far have neglected the crucial effects of SUSY breaking. It may well be that the effects of SUSY breaking, on top of necessitating a severe fine-tuning of the cosmological constant, will induce drastic modifications in the functional form of the quintessence potential. It is certainly a tantalizing challenge to include the effects of SUSY breaking within the supergravity models of quintessence . On the other hand there exists the possibility that the cosmological constant problem will be resolved using ideas stemming from extra-dimension scenarios involving an effective supersymmetry in four dimensions . The investigation of such models might well shed new light on the origin of the quintessence field.
As must be clear by now the issues raised by the cosmological constant problem, the quintessence scenario and its proper understanding within particle physics are many-fold. The experimental results which will be available in the near future might help in disentangling some of these very conspicuous matters.
|
warning/0005/hep-th0005041.html
|
ar5iv
|
text
|
# Strings and branes with a modified measure
## 1 Introduction
String and brane theory have appeared as candidates for unifying all interactions of nature. One aspect of string and brane theories seems to many not quite appealing however: this is the introduction from the begining of a fundamental scale, the string or brane tension. The idea that the fundamental theory of nature, whatever that may be, should not contain any fundamental scale has attracted a lot of attention. According to this point of view, whatever scale appears in nature, must not appear in the fundamental lagrangian of physics. Rather, the appearence of these scales must be spontaneous, for example due to boundary conditions in a classical context or a process of dimensional transmutation to give an example of such effect in the context of quantum field theory.
Also, in the context of gravitational theory, the idea that Newton’s constant may originate from a phenomenon of spontaneous symmetry breaking has inspired Zee and others to build models along these lines. Furthermore, it has been shown that Zee’s induced gravity can in turn be obtained from a theory without any fundamental scale and manifest global scale invariance . In such an approach, when integrating the equations of motion, we introduce an integration constant which is responsible for the ssb of scale invariance, as described in the model studied in Refs. 4,5 6, which was shown to be connected, after the ssb, to the Zee model in Ref.3.
The model of Refs.3,4,5,6 is based on the possibility of replacing the measure of integration $`\sqrt{\gamma }d^Dx`$, by another one, $`\mathrm{\Phi }d^Dx`$, where $`\mathrm{\Phi }`$ is a density built out of degrees of freedom independent of the metric $`\gamma _{ab}`$ . Such possibility was studied in a general context , not related to scale invariance also.
Here we want to see what are the consequences of doing something similar in the context of string theory. As we will see, string theories or more generally brane theories without a fundamental scale are possible if the extended objects do not have boundaries (i.e., they are closed).
In the context of the formalism for extended object proposed here, if there are boundaries, we require the coupling of a gauge field that lives in the brane to a lower dimensional object (a point particle in the case of a string) that defines the boundary, since then the equations of motion allow us to end the extended object at this boundary. The coupling constant of the gauge field to the lower dimensional object defines in this case a fundamental scale of the theory. The scales of the theory appear therefore as integration constants in the case of closed extended objects or through the physics of the boundaries of these extended objects.
## 2 String theories with a modified measure
The Polyakov action for the bosonic string is
$$S_P[X,\gamma _{ab}]=T𝑑\tau 𝑑\sigma \sqrt{\gamma }\gamma ^{ab}_aX^\mu _bX^\nu g_{\mu \nu }$$
(1)
Here $`\gamma _{ab}`$ is the metric defined in the $`1+1`$ world sheet of the string and $`\gamma =det(\gamma _{ab})`$. $`g_{\mu \nu }`$ is the metric of the embedding space. $`T`$ is here the string tension, a dimensionfull quantity introduced into the theory, which defines a scale .
We recognize the measure of integration $`d\tau d\sigma \sqrt{\gamma }`$ and as we anticipated before, we want to replace this measure of integration by another one which does not depend on $`\gamma _{ab}`$ .
If we introduce two scalars (both from the point of view of the $`1+1`$ world sheet of the string and from the embedding $`D`$-dimensional universe) $`\phi _i`$, $`i=1,2`$, we can contruct the world sheet density
$$\mathrm{\Phi }=\epsilon ^{ab}\epsilon _{ij}_a\phi _i_b\phi _j$$
(2)
where $`\epsilon ^{ab}`$ is given by $`\epsilon ^{01}=\epsilon ^{10}=1`$, $`\epsilon ^{00}=\epsilon ^{11}=0`$ and $`\epsilon _{ij}`$ is defined by $`\epsilon _{12}=\epsilon _{21}=1`$, $`\epsilon _{11}=\epsilon _{22}=0`$.
It is interesting to notice that $`d\tau d\sigma \mathrm{\Phi }=2d\phi _1d\phi _2`$, that is the measure of integration $`d\tau d\sigma \mathrm{\Phi }`$ corresponds to integrating in the space of the scalar fields $`\phi _1,\phi _2`$.
We proceed now with the construction of an action that uses $`d\tau d\sigma \mathrm{\Phi }`$ instead of $`d\tau d\sigma \sqrt{\gamma }`$. When considering the types of actions we can have under these circumtances, the first one that comes to mind ( a straightforward generalization of the Polyakov action) is
$$S_1=𝑑\tau 𝑑\sigma \mathrm{\Phi }\gamma ^{ab}_aX^\mu _bX^\nu g_{\mu \nu }$$
(3)
Notice that multiplying $`S_1`$ by a constant, before boundary or initial conditions are specified is a meaningless operation, since such a constant can be absorbed in a redefinition of the measure fields $`\phi _1,\phi _2`$ that appear in $`\mathrm{\Phi }`$.
The form (3) is however not a satisfactory action, because the variation of $`S_1`$ with respect to $`\gamma ^{ab}`$ leads to the rather strong condition
$$\mathrm{\Phi }_aX^\mu _bX^\nu g_{\mu \nu }=0$$
(4)
If $`\mathrm{\Phi }0`$, it means that $`_aX^\mu _bX^\nu g_{\mu \nu }=0`$, which means that the metric induced on the string vanishes, clearly not an acceptable dynamics. Alternatively, if $`\mathrm{\Phi }=0`$, no further information is available, also a not desirable situation.
To make further progress, it is important to notice that terms that when considered as contributions to $`L`$ in
$$S=𝑑\tau 𝑑\sigma \sqrt{\gamma }L$$
(5)
which do not contribute to the equations of motion, i.e., such that $`\sqrt{\gamma }L`$ is a total derivative, may contribute when we consider the same $`L`$, but in a contribution to the action of the form
$$S=𝑑\tau 𝑑\sigma \mathrm{\Phi }L$$
(6)
This is so because if $`\sqrt{\gamma }L`$ is a total divergence, $`\mathrm{\Phi }L`$ in general is not.
This fact is indeed crucial and if we consider an abelian gauge field $`A_a`$ defined in the world sheet of the string, in addition to the measure fields $`\phi _1,\phi _2`$ that appear in $`\mathrm{\Phi }`$, the metric $`\gamma ^{ab}`$ and the string coordinates $`X^\mu `$, we can then construct the non trivial contribution to the action of the form
$$S_{gauge}=𝑑\tau 𝑑\sigma \mathrm{\Phi }\frac{\epsilon ^{ab}}{\sqrt{\gamma }}F_{ab}$$
(7)
where
$$F_{ab}=_aA_b_bA_a$$
(8)
So that the total action to be considered is now
$$S=S_1+S_{gauge}$$
(9)
with $`S_1`$ defined as in eq. 3 and $`S_{gauge}`$ defined by eqs.7 and 8.
The action (9) is invariant under a set of diffeomorphisms in the space of the measure fields combined with a conformal transformation of the metric $`\gamma _{ab}`$,
$$\phi _i\phi _i^{^{}}=\phi _i^{^{}}(\phi _j)$$
(10)
So that,
$$\mathrm{\Phi }\mathrm{\Phi }^{^{}}=J\mathrm{\Phi }$$
(11)
where J is the jacobian of the transformation (10) and
$$\gamma _{ab}\gamma _{ab}^{^{}}=J\gamma _{ab}$$
(12)
The combination $`\frac{\epsilon ^{ab}}{\sqrt{\gamma }}F_{ab}`$ is a genuine scalar. In two dimensions is proportional to $`\sqrt{F_{ab}F^{ab}}`$.
Working with (9), we get the following equations of motion: From the variation of the action with respect to $`\phi _j`$
$$\epsilon ^{ab}_b\phi _j_a(\gamma ^{cd}_cX^\mu _dX^\nu g_{\mu \nu }+\frac{\epsilon ^{cd}}{\sqrt{\gamma }}F_{cd})=0$$
(13)
If $`det(\epsilon ^{ab}_b\phi _j)0`$, which means $`\mathrm{\Phi }0`$, then we must have that all the derivatives of the quantity inside the parenthesis in eq.13 must vanish, that is, such a quantity must equal a constant which will be determined later, but which we will call $`M`$ in the mean time,
$$\gamma ^{cd}_cX^\mu _dX^\nu g_{\mu \nu }+\frac{\epsilon ^{cd}}{\sqrt{\gamma }}F_{cd}=M$$
(14)
The equation of motion of the gauge field $`A_a`$, tells us about how the string tension appears as an integration constant. Indeed this equation is
$$\epsilon ^{ab}_b(\frac{\mathrm{\Phi }}{\sqrt{\gamma }})=0$$
(15)
which can be integrated to give
$$\mathrm{\Phi }=c\sqrt{\gamma }$$
(16)
Notice that (16) is perfectly consistent with the conformal symmetry (10), (11) and (12). Equation 14 on the other hand is consistent with such a symmetry only if $`M=0`$. Indeed, we will check that the equations of motion indeed imply that $`M=0`$. In the case of higher dimensional branes, the equations of motion require also a very specific value of $`M`$, but in that case, it will be a non vanishing value.
By calculating the Hamiltonian, after dropping boundary terms (this is totally justified in the case of closed strings) and (only at the end of the process) using eq.16, we find that $`c`$ equals the string tension.
Furthermore, if we couple the gauge field $`A_a`$ to point particles living in the string, we find that the right hand side of eq.15 is not zero anymore, but rather a delta function with non vanishing support at the location of the particle. The solution of the equation will be $`\mathrm{\Phi }=c_1\sqrt{\gamma }`$ to the right of the point particle and $`\mathrm{\Phi }=c_2\sqrt{\gamma }`$ to the left of the point particle. $`c_2c_1`$ will be then the charge of the point particle. We obtain the a picture of a string where the tension has changed from one region to the other according to the charge that we have inserted.
There is the possibility that either $`c_1`$ or $`c_2`$ equal zero. In that case the string itself starts at the location of the elementary charge. This picture could be of use in for example a string confinement model of charged particles. In this case a fundamental scale is introduced, not through the straightforward introduction of a string tension, but by the introduction of the elementary charge of a point particle living in the string, i.e. through the boundary physics of the string.
It is very important to notice that in this formulation the string can finish at a certain definite boundary, in a way that is dictated by the equations of motion, due to the introduction of point like charges at the boundaries of the string. This allows the measure to just vanish when we go beyond the point like charge.
Now let us turn our attention to the equation of motion derived from the variation of (9) with respect to $`\gamma ^{ab}`$. We get then,
$$\mathrm{\Phi }(_aX^\mu _bX^\nu g_{\mu \nu }\frac{1}{2}\gamma _{ab}\frac{\epsilon ^{cd}}{\sqrt{\gamma }}F_{cd})=0$$
(17)
From the constraint (14), we can solve $`\frac{\epsilon ^{cd}}{\sqrt{\gamma }}F_{cd}`$ and insert back into (17), obtaining then (if $`\mathrm{\Phi }0`$)
$$_aX^\mu _bX^\nu g_{\mu \nu }\frac{1}{2}\gamma _{ab}\gamma ^{cd}_cX^\mu _dX^\nu g_{\mu \nu }\frac{1}{2}\gamma _{ab}M=0$$
(18)
Multiplying the above equation by $`\gamma ^{ab}`$ and summing over $`a,b`$, we get that $`M=0`$, that is the equations are exactly those of the Polyakov action. After eq.16 is used, the eq. obtained from the variation of $`X^\mu `$ is seen to be exactly the same as the obtained from the Polyakov action as well.
## 3 Higher Dimensional Extended Objects
Let us now consider a $`d+1`$ extended object, described (generalizing the action (9)),
$$S=S_d+S_{dgauge}$$
(19)
where
$$S_d=d^{d+1}x\mathrm{\Phi }\gamma ^{ab}_aX^\mu _bX^\nu g_{\mu \nu }$$
(20)
and
$$S_{dgauge}=d^{d+1}x\mathrm{\Phi }\frac{\epsilon ^{a_1a_2\mathrm{}a_{d+1}}}{\sqrt{\gamma }}_{[a_1}A_{a_2\mathrm{}a_{d+1}]}$$
(21)
and
$$\mathrm{\Phi }=\epsilon ^{a_1a_2\mathrm{}a_{d+1}}\epsilon _{j_1j_2\mathrm{}j_{d+1}}_{a_1}\phi _{j_1}\mathrm{}._{a_{d+1}}\phi _{j_{d+1}}$$
(22)
This model does not have a symmetry which involves an arbitrary diffeomorphism in the space of the measure fields coupled with a conformal transformation of the metric, except if $`d=1`$ (eqs. (10), (11), (12)). For $`d1`$, there is still a global scaling symmetry where the metric transforms as ($`\theta `$ being a constant),
$$\gamma _{ab}e^\theta \gamma _{ab}$$
(23)
the $`\phi _j`$ are transformed according to
$$\phi _j\lambda _j\phi _j$$
(24)
(no sum on $`j`$) which means $`\mathrm{\Phi }\left(_j\lambda _j\right)\mathrm{\Phi }\lambda \mathrm{\Phi }`$
Finally, we must demand that $`\lambda =e^\theta `$ and that the transformation of $`A_{a_2\mathrm{}a_{d+1}}`$ be defined as
$$A_{a_2\mathrm{}a_{d+1}}\lambda ^{\frac{d1}{2}}A_{a_2\mathrm{}a_{d+1}}$$
(25)
Then we have a symmetry. Also no scale is introduced into the theory from the beginning. This is apparent from the fact that any constants multiplying the separate contributions to the action (20) or (21) is meaningless if no boundary or initial conditions are specified, because then such factors can be absorbed by a redefinition of the measure fields and of the gauge fields. Notice that the existence of a symmetry alone is not enough to guarantee that no fundamental scale appears in the action. For example string theory, as usually formulated has conformal symmetry, but the string tension is still a fundamental scale in the theory.
Another interesting symmetry of the action (up to the integral of a total divergence) consists of the infinite dimensional set of transformations $`\phi _j\phi _j+f_j(L)`$, where $`f_j(L)`$ are arbitrary functions of
$$L=\gamma ^{cd}_cX^\mu _dX^\nu g_{\mu \nu }+\frac{\epsilon ^{a_1a_2\mathrm{}a_{d+1}}}{\sqrt{\gamma }}_{[a_1}A_{a_2\mathrm{}a_{d+1}]}$$
(26)
This symmetry does depend on the explicit form of the lagrangian density $`L`$ , but only the fact that $`L`$ is $`\phi _a`$ independent.
Now we go through the same steps we went through in the case of the string. The variation with respect to the measure field $`\phi _j`$ gives
$$K_j^a_a(\gamma ^{cd}_cX^\mu _dX^\nu g_{\mu \nu }+\frac{\epsilon ^{a_1a_2\mathrm{}a_{d+1}}}{\sqrt{\gamma }}_{[a_1}A_{a_2\mathrm{}a_{d+1}]})=0$$
(27)
where
$$K_j^a=\epsilon ^{aa_2\mathrm{}a_{d+1}}\epsilon _{jj_2\mathrm{}j_{d+1}}_{a_2}\phi _{j_2}\mathrm{}._{a_{d+1}}\phi _{j_{d+1}}$$
(28)
Since $`det(K_j^a)=\frac{(d+1)^{(d+1)}}{(d+1)!}\mathrm{\Phi }^d`$, it therefore follows that for $`\mathrm{\Phi }0`$, $`det(K_j^a)0`$ and
$$\gamma ^{cd}_cX^\mu _dX^\nu g_{\mu \nu }+\frac{\epsilon ^{a_1a_2\mathrm{}a_{d+1}}}{\sqrt{\gamma }}_{[a_1}A_{a_2\mathrm{}a_{d+1}]}=M$$
(29)
where $`M`$ is some constant of integration. If $`d1`$ then $`M0`$ as we will see. Furthermore, under a scale transformation (23), (24), (25), $`M`$ does change from one constant value to another.
The variation with respect to the gauge field $`A_{a_2\mathrm{}a_{d+1}}`$ leads to the equation
$$\epsilon ^{a_1a_2\mathrm{}a_{d+1}}_{a_1}\frac{\mathrm{\Phi }}{\sqrt{\gamma }}=0$$
(30)
which means
$$\mathrm{\Phi }=c\sqrt{\gamma }$$
(31)
once again. As in the case of the string the brane tension has been generated spontaneously instead of appearing as a parameter of the fundamental lagrangian. Again a simple calculation of the Hamiltonian and using after this the above equation, we obtain that $`c`$ equals the brane tension.
The variation of the action with respect to $`\gamma ^{ab}`$ leads to
$$\mathrm{\Phi }(_aX^\mu _bX^\nu g_{\mu \nu }\frac{1}{2}\gamma _{ab}\frac{\epsilon ^{a_1a_2\mathrm{}a_{d+1}}}{\sqrt{\gamma }}_{[a_1}A_{a_2\mathrm{}a_{d+1}]})=0$$
(32)
We can now solve for $`\frac{\epsilon ^{a_1a_2\mathrm{}a_{d+1}}}{\sqrt{\gamma }}_{[a_1}A_{a_2\mathrm{}a_{d+1}]}`$ from equation (29) and then reinsert in the above equation, obtaining then,
$$_aX^\mu _bX^\nu g_{\mu \nu }=\frac{1}{2}\gamma _{ab}(\gamma ^{cd}_cX^\mu _dX^\nu g_{\mu \nu }+M)$$
(33)
This is the same equation that we would have obtained from a Polyakov type action augmented by a cosmological term.
As in the case of the string, $`M`$ can be found by contracting both sides of the equation. For $`d1`$, $`M`$ is non zero and equal to
$$M=\frac{\gamma ^{cd}_cX^\mu _dX^\nu g_{\mu \nu }(1d)}{1+d}$$
(34)
We can also solve for $`\gamma ^{cd}_cX^\mu _dX^\nu g_{\mu \nu }`$ in terms of $`M`$ from (34) and insert in the right hand side of (33), obtaining,
$$\gamma _{ab}=\frac{1d}{M}_aX^\mu _bX^\nu g_{\mu \nu }$$
(35)
Which means that $`\gamma _{ab}`$ is up to the constant factor $`\frac{1d}{M}`$ equal to the induced metric. Since there is the scale invariance (23), (24), (25), an overall constant factor in the evolution of $`\gamma _{ab}`$ cannot be determined. The same scale invariance means however that there is a field redefinition which does not affect any parameter of the lagrangian and which allows us to set $`\gamma _{ab}`$ equal to the induced metric (at least if we start from any negative value of $`M`$), that is,
$$\gamma _{ab}=_aX^\mu _bX^\nu g_{\mu \nu }$$
(36)
In such case $`M`$ is consistently given (inserting (36) into (35) or (34)),
$$M=1d$$
(37)
Notice that in contrast with the standard approach for Polyakov type actions in the case of higher dimensional branes , here we do not have to fine tune a parameter of the lagrangian the brane ”cosmological constant”, so as to force that (36) be satisfied. Rather, it is an integration constant, that appears from an action without an original cosmological term, which can be set to the value given by eq. (37) by means of a scale transformation. Such choice ensures then that (37) is satisfied (and therefore (36)). Furthermore, it appears that this treatment is more appealing if one thinks of all branes on similar footing, since in the approach of this paper they can all be described by a similar looking lagrangian, unlike in the usual aproach which discriminates in a radical way between strings, these having no cosmological constant associated to them, and the higher dimensional branes, which require a fine tuned cosmological constant.
As in the case of the string, the constant $`c`$ provides a spontaneously generated brane tension. In a way similar to that of the string, we can generate a discontinuity in such brane tension by coupling minimally the gauge field $`A_{a_1\mathrm{}.a_d}`$ defined in the brane to a current defined in the boundary of such a brane, which is a lower dimensional brane.
As in the case of the string, the brane can finish at a certain definite boundary, in a way that is dictated by the equations of motion, due to the introduction charges at the boundaries, which are lower dimensional branes. This allows the measure to just vanish when we go beyond the boundaries defined by the lower dimensional brane.
## 4 Discussion and Conclusions
A different approach to the theory of extended objects has been developed by allowing the integration measure in the action to be independent of the metric.
In this approach, in the case of closed objects, no scales appear in the fundamental lagrangian, the brane tension appears as an integration constant. If coupling to lower dimensional branes are allowed, this coupling introduces a a fundamental scale. That it, scales are introduced only as the result of initial conditions or as the result of the physics of the boundaries of the extended object.
Further generalizations and extensions to incorporate supersymmetry should be studied in order to build a relistic model. The fact that both strings and branes can be studied with a fundamental action which does not contain an explicit cosmological term, in contrast with the usual treatment, which requires a different cosmological term for every type of brane, should be of use when trying to achieve a unified treatment of all these branes.
One should notice that other authors have also constructed actions for branes that do not contain a brane-cosmological term . Such formulations depend, unlike what has been developed here, on the dimensionality, in particular whether this is even or odd, so that it is clear that those formulations do not have much relation with what has been done here. Yet other approaches to an action without a brane cosmological involve lagrangians with non linear dependence on the invariant $`\gamma ^{cd}_cX^\alpha _dX^\beta g_{\alpha \beta }`$, also a rather different path to the one followed here. For an interesting analysis of different possible Lagrangians for extendons see .
An approach that has some common features to the one developed here is that of Ref. , where also the tension of the brane is found as an integration constant. Here also gauge fields are introduced, but they appear in a quadratic form rather than in a linear form. Also the scale invariance discussed there is a target space scale invariance since no metric defined in the brane is studied there, i.e. no connection to a Polyakov type action, which is known to be more useful in the quantum theory, is made.
Finally, it will be of use to develop theories along the lines developed here not only as candidate unified models for all fundamental interactions, but also as useful phenomenological tools for the study of confinement of quarks.
## 5 Acknowledgements
I want to thank C.Castro in particular for pointing to me the related work in ref. . I also want to thank A.Davidson, A.Kaganovich, J.Portnoy and L.C.R. Wijewardhana for discussions.
|
warning/0005/hep-ph0005210.html
|
ar5iv
|
text
|
# 1 *
CPT-2000/P.4004
USM-TH-92
The Quark-Antiquark Asymmetry of the Nucleon Sea
from $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ Fragmentation
Bo-Qiang Ma<sup>*</sup><sup>*</sup>*e-mail: mabq@phy.pku.edu.cn<sup>a</sup>, Ivan Schmidte-mail: ischmidt@fis.utfsm.cl <sup>b</sup>, Jacques Soffere-mail: Jacques.Soffer@cpt.univ-mrs.fr<sup>c</sup>, Jian-Jun Yang<sup>§</sup><sup>§</sup>§e-mail: jjyang@fis.utfsm.cl<sup>b,d</sup>
<sup>a</sup>Department of Physics, Peking University, Beijing 100871, China,Mailing address
CCAST (World Laboratory), P.O. Box 8730, Beijing 100080, China,
and Institute of Theoretical Physics, Academia Sinica, Beijing 100080, China
<sup>b</sup>Departamento de Física, Universidad Técnica Federico Santa María,
Casilla 110-V, Valparaíso, Chile
<sup>c</sup>Centre de Physique Th$`\stackrel{´}{\mathrm{e}}`$orique, CNRS, Luminy Case 907, F-13288 Marseille Cedex 9, France
<sup>d</sup>Department of Physics, Nanjing Normal University,
Nanjing 210097, China
Abstract
We present a general analysis of the spin transfer for $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ production in deep-inelastic scattering of polarized charged leptons on the nucleon, and find that the pattern of different behaviors of $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ production observed by the E665 Collaboration suggests the possibility of quark-antiquark asymmetries either in the quark to $`\mathrm{\Lambda }`$ fragmentation functions and/or in the quark and antiquark distributions of the target proton. We also point out that the strange-antistrange asymmetry of the nucleon sea may produce an observable contribution to the different behaviors of $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ production. We find that a softer $`\overline{s}(x)`$ than $`s(x)`$ as predicted by the light-cone baryon-meson fluctuation model of intrinsic quark-antiquark pairs of the nucleon sea might lead to a reasonable picture. However, the magnitude is still too small to explain the E665 data and the conclusion has also strong model-dependence. This may suggest the importance of quark-antiquark asymmetry in the quark to $`\mathrm{\Lambda }`$ fragmentation functions, provided that the E665 data are confirmed.
PACS numbers: 14.20.Jn, 12.38.Bx, 13.87.Fh, 13.88.+e
It is well known that the production of $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ in deep-inelastic scattering (DIS) of lepton on the nucleon may provide information on the quark content of the target nucleon , as well as on the quark to $`\mathrm{\Lambda }`$ fragmentation functions . The idea that the fragmentation of the $`\mathrm{\Lambda }`$ hyperon in DIS of a charged lepton on a nucleon target can supply information concerning the strange content of the nucleon was originally proposed in Refs. . There are four different combinations of the polarizations of the charged lepton beam and the nucleon target:
i) both the lepton beam and the nucleon target are unpolarized;
ii) the nucleon target is polarized while the lepton beam is unpolarized ;
iii) both the lepton beam and the nucleon target are polarized ;
iv) the lepton beam is polarized while the nucleon target is unpolarized .
These different combinations provide different information concerning the quark distributions and quark to $`\mathrm{\Lambda }`$ fragmentation functions. It is suggested in Ref. that there are still large uncertainties in the quark to $`\mathrm{\Lambda }`$ fragmentation function, and it is practically more urgent to measure the $`\mathrm{\Lambda }`$ fragmentation functions before using the $`\mathrm{\Lambda }`$ fragmentation to probe the quark content of the nucleon. Indeed, some symplifying assumptions about the quark to $`\mathrm{\Lambda }`$ fragmentation functions were found to be of little predictive power when applied to $`\mathrm{\Lambda }`$ production in $`e^+e^{}`$ annihilation process at the $`Z`$ resonance , and to semi-inclusive $`\mathrm{\Lambda }`$ production of polarized charged lepton DIS process on the nucleon target .
However, there have been recent progress in order to understand the quark to $`\mathrm{\Lambda }`$ fragmentation functions by connecting them with the quark distributions inside the $`\mathrm{\Lambda }`$ by the Gribov-Lipatov relation (GLR) :
$$D_q^h(z)q_h(x),$$
(1)
where $`D_q^h(z)`$ is the fragmentation function for a quark $`q`$ splitting into a hadron $`h`$ with longitudinal momentum fraction $`z`$, and $`q_h(x)`$ is the quark distribution for finding the quark $`q`$ inside the hadron $`h`$ carrying a momentum fraction $`x`$. $`D_q^h`$ and $`q_h`$ depend also on the energy scale $`Q^2`$, and this relation holds, in principle, in a certain $`Q^2`$ range and in leading order approximation. It is shown recently that the Gribov-Lipatov relation is also verified to hold in leading order for the space- and time-like splitting functions of QCD. Moreover, although Eq. (1) is only valid at $`x1`$ and $`z1`$, it provides a reasonable guidance for a phenomenological parametrization of the various quark to $`\mathrm{\Lambda }`$ fragmentation functions. We are encouraged to find that the predictions of the quark to $`\mathrm{\Lambda }`$ fragmentation functions in an SU(6) quark-diquark model and in a pQCD based model are in good agreement with the experimental data on $`\mathrm{\Lambda }`$ production in both the $`e^+e^{}`$ annihilation process at the $`Z`$ resonance and in polarized positron beam DIS on a nucleon target . Thus we have at least some reasonable parametrizations of quark to $`\mathrm{\Lambda }`$ fragmentation functions, though there are still large uncertainties in the flavor and spin decompositions of these fragmentation functions.
For a longitudinally polarized charged lepton beam and an unpolarized nucleon target, the longitudinal spin transfer to the $`\mathrm{\Lambda }`$ is given in the quark parton model by
$$A^\mathrm{\Lambda }(x,z)=\frac{\underset{q}{}e_q^2[q^N(x,Q^2)\mathrm{\Delta }D_q^\mathrm{\Lambda }(z,Q^2)+(q\overline{q})]}{\underset{q}{}e_q^2[q^N(x,Q^2)D_q^\mathrm{\Lambda }(z,Q^2)+(q\overline{q})]}.$$
(2)
Here $`y=\nu /E`$, $`x=Q^2/2M_N\nu `$, and $`z=E_\mathrm{\Lambda }/\nu `$, where $`q^2=Q^2`$ is the squared four-momentum transfer of the virtual photon, $`M_N`$ is the proton mass, and $`\nu `$, $`E`$, and $`E_\mathrm{\Lambda }`$ are the energies of the virtual photon, the target nucleon, and the produced $`\mathrm{\Lambda }`$ respectively, in the target rest frame; $`q^N(x,Q^2)`$ is the quark distribution for the quark $`q`$ in the nucleon, $`D_q^\mathrm{\Lambda }(z,Q^2)`$ is the fragmentation function for $`\mathrm{\Lambda }`$ production from quark $`q`$, $`\mathrm{\Delta }D_q^\mathrm{\Lambda }(z,Q^2)`$ is the corresponding longitudinal spin-dependent fragmentation function, and $`e_q`$ is the quark charge in units of the elementary charge $`e`$. In a region where $`x`$ is large enough, e.g. $`x>0.2`$, one can neglect the antiquark contributions in Eq. (2), and probe only the valence quarks of the target nucleon. On the contrary, if $`x`$ is much smaller, one is probing the sea quarks and therefore the antiquarks must be considered as well. For $`\overline{\mathrm{\Lambda }}`$ production the spin transfer $`A^{\overline{\mathrm{\Lambda }}}(x,z)`$ is obtained from Eq. (2) by replacing $`\mathrm{\Lambda }`$ by $`\overline{\mathrm{\Lambda }}`$. The $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ fragmentation functions are related since we can safely assume matter-antimatter symmetry, i.e. $`D_{q,\overline{q}}^\mathrm{\Lambda }(z)=D_{\overline{q},q}^{\overline{\mathrm{\Lambda }}}(z)`$ and similarly for $`\mathrm{\Delta }D_{q,\overline{q}}^\mathrm{\Lambda }(z)`$.
Recently, the HERMES Collaboration at DESY reported the result of the longitudinal spin transfer to the $`\mathrm{\Lambda }`$ in polarized positron DIS on the proton . Also the E665 Collaboration at FNAL measured the $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ spin transfers from muon DIS , and they observed very different behaviour for $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ polarizations. The E665 data for the spin transfer are presented as function of the Feynman variable $`x_F`$, although $`x_Fz`$ is a good approximation in the kinematic range of the E665 experiment . Strictly speaking, the magnitude of the measured spin transfer Eq.(2) should be less than unity; thus the E665 data, whose range of magnitude for the measured spin transfer is larger than unity, are of poor precision. But the different behaviors of the $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ spin transfer might still be a realistic effect. Both the HERMES data and the E665 data are measured for $`x_F>0`$, which corresponds to the current fragmentation region. Thus it is natural to try to understand the data from the viewpoint of current fragmentation, rather than target fragmentation as suggested by E665. We will focus our attention on the different behavior of the $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ spin transfer in the E665 data.
It is interesting to notice that the fragmentation functions from the quark-diquark model can give very good descriptions of both the data of $`\mathrm{\Lambda }`$ fragmentations in $`e^+e^{}`$ annihilation at the $`Z`$ resonance , and in polarized positron DIS on the unpolarized proton by HERMES , with only naive parameters without any adjustment. Although the Gribov-Lipatov relation should be of poor validity at small $`x`$, the fragmentation functions obtained by using it in the quark-diquark model seem to give a reasonable relation between different quark to $`\mathrm{\Lambda }`$ fragmentation functions. We would like to mention that the fragmentation functions derived in a quark-diquark picture and in an MIT model framework arrived at similar qualitative results as in Ref. , although the explicit shapes are not the same. Therefore we first use the fragmentation functions from the quark-diquark model as input in order to calculate the spin transfer, Eq. (2), for the $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ production.
We use the recent CTEQ5 parametrizations as input for the quark distributions of the nucleon . In Fig. 1(a) we present the calculated results for the spin transfer of $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$, and compare the results with the the HERMES and E665 data. We notice that the calculations show a trend of increasing positive polarization with increasing $`z`$ that seems to be suggested by the data. This supports the prediction of positive polarized $`u`$ and $`d`$ quarks inside $`\mathrm{\Lambda }`$ at large $`x`$ . However, the calculations cannot produce a difference of the spin transfers for $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ as observed by E665. Our results are also in qualitative agreement with a Monte Carlo simulation based on the naive quark model and a model with SU(3) symmetry , where the different behavior of spin transfer for $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ are not predicted. Although the effect due to target fragmentation has been suggested as a possible mechanism for the E665 different behavior of $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ production , the kinematic region for $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ production corresponds to $`x_F>0`$, which is the current fragmentation region . Thus we need to find a new mechanism for the different behavior of $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ production. The possibility of a quark-antiquark asymmetry in the fragmentation functions has been investigated in Ref. , and the purpose of this paper is to investigate possible asymmetries in the quark distributions of the nucleon target.
Let us consider the spin transfer for $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ production in a region where the sea dominates, namely where the Bjorken $`x`$ is rather small, like in the E665 experiment, which has $`x_B`$=0.005. The E665 data indicates that in this region and for $`z`$ between 0.1 and 0.5, one has
$$A^{\overline{\mathrm{\Lambda }}}(x,z)>>A^\mathrm{\Lambda }(x,z).$$
(3)
Let us consider several possible situations:
1) The sea is fully symmetric, namely $`q(x)=\overline{q}(x)`$, for all flavors $`u,d,s`$. This implies clearly $`A^{\overline{\mathrm{\Lambda }}}(x,z)=A^\mathrm{\Lambda }(x,z)`$, which contradicts the data.
2) The sea is not fully symmetric, and in this case one can consider several scenarios:
(a) One flavor dominates, for example $`u`$-quark. Let us define
$$\mathrm{\Delta }\overline{Q}=\overline{q}(x)[\mathrm{\Delta }D_q^\mathrm{\Lambda }(z)+\mathrm{\Delta }D_{\overline{q}}^\mathrm{\Lambda }(z)],\overline{Q}=\overline{q}(x)[D_q^\mathrm{\Lambda }(z)+D_{\overline{q}}^\mathrm{\Lambda }(z)].$$
(4)
In this case, if the sea is symmetric we are back to case 1) above and Eq. (3) cannot be satisfied. If the sea is not symmetric, namely $`u=\overline{u}+ϵ`$, we have
$$A^\mathrm{\Lambda }=\frac{ϵ\mathrm{\Delta }D_u^\mathrm{\Lambda }+2\mathrm{\Delta }\overline{U}}{ϵD_u^\mathrm{\Lambda }+2\overline{U}},$$
(5)
and
$$A^{\overline{\mathrm{\Lambda }}}=\frac{ϵ\mathrm{\Delta }D_{\overline{u}}^\mathrm{\Lambda }+2\mathrm{\Delta }\overline{U}}{ϵD_{\overline{u}}^\mathrm{\Lambda }+2\overline{U}}.$$
(6)
Let us try to see what conditions one must have in order to fulfill Eq. (3). It seems clear that $`D_u^\mathrm{\Lambda }>>D_{\overline{u}}^\mathrm{\Lambda }`$, therefore if $`ϵ>0`$ (which is the case for $`x=0.005`$ in present parametrizations of quark distributions), Eq. (3) will be satisfied provided
$$\mathrm{\Delta }D_{\overline{u}}^\mathrm{\Lambda }>>\mathrm{\Delta }D_u^\mathrm{\Lambda },$$
(7)
a condition which is assumed and discussed in as a possibility to explain the different behaviors of $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ productions in the E665 data.
(b) All three flavors contribute but only one is asymmetric, say $`u`$, whereas $`d=\overline{d}`$ and $`s=\overline{s}`$. This case is similar to the previous one, since we have
$$A^\mathrm{\Lambda }=\frac{4ϵ\mathrm{\Delta }D_u^\mathrm{\Lambda }+X}{4ϵD_u^\mathrm{\Lambda }+Y},$$
(8)
and
$$A^{\overline{\mathrm{\Lambda }}}=\frac{4ϵ\mathrm{\Delta }D_{\overline{u}}^\mathrm{\Lambda }+X}{4ϵD_{\overline{u}}^\mathrm{\Lambda }+Y},$$
(9)
where $`X=8\mathrm{\Delta }\overline{U}+2\mathrm{\Delta }\overline{D}+2\mathrm{\Delta }\overline{S}`$ and $`Y=8\overline{U}+2\overline{D}+2\overline{S}`$. We reach the same conclusion as above, and since the asymmetric flavor can be either $`d`$ or $`s`$, it seems that one should have more generally
$$\mathrm{\Delta }D_{\overline{q}}^\mathrm{\Lambda }>>\mathrm{\Delta }D_q^\mathrm{\Lambda },$$
(10)
a condition which has been discussed and considered in . Remember that positivity implies $`D_{\overline{q}}^\mathrm{\Lambda }\mathrm{\Delta }D_{\overline{q}}^\mathrm{\Lambda }`$, so it means that one should have a strong bound on $`\mathrm{\Delta }D_q^\mathrm{\Lambda }`$, namely
$$D_{\overline{q}}^\mathrm{\Lambda }>>\mathrm{\Delta }D_q^\mathrm{\Lambda }.$$
(11)
However, the situation will be different in case we have $`ϵ<0`$, which means that $`\overline{q}(x)>q(x)`$. The strange quark-antiquark asymmetry predicted by the light-cone baryon-meson fluctuation model introduces such a behavior for the strange quarks and antiquarks. Thus the pattern of the difference in the $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ productions in the E665 data could suggest a possibility of $`\overline{q}(x)>q(x)`$ in the target proton.
The CTEQ parametrizations of quark distributions are based on data of various structure functions from different DIS processes obtained in the last three decades. The light-flavor $`u`$ and $`d`$ content of the nucleon is well constrained and the uncertainties are not big, though there are still a number of phenomenological anomalies related to the spin and flavor content of the nucleon sea . However, the strange content of the nucleon is less known than the light-flavor $`u`$ and $`d`$ quarks. In the CTEQ parametrizations, identical strange and antistrange quark distributions are assumed. However, it is pointed out in Ref. that within the allowed errors, the CCFR data of $`s(x)/\overline{s}(x)`$ does not rule out a strange-antistrange asymmetry, as suggested by the light-cone baryon-meson fluctuation model . Moreover, this light-cone baryon-meson fluctuation model of intrinsic quark-antiquark ($`q\overline{q}`$) pairs in the nucleon sea suggests a soft $`\overline{s}(x)`$ compared to $`s(x)`$ (i.e., $`\overline{s}(x)>s(x)`$ at small $`x`$ and vice versa at large $`x`$). Remember that a softer $`\overline{s}(x)`$ than $`s(x)`$ was predicted by Burkardt and Warr from the chiral Gross-Neveu model at large $`N_c`$ in the light-cone formalism. It is also pointed out in Ref. that the conflict between two different determinations of the strange quark distributions could be a phenomenological support for $`s(x)\overline{s}(x)`$, or more explicitly, a softer $`\overline{s}(x)`$ compared to $`s(x)`$. Another phenomenological support for a softer $`\overline{s}(x)`$ is also suggested by Barone, Pascaud, and Zomer from a global QCD analysis of structure functions, including neutrino DIS data. More recently, Buccella, Pisanti, and Rosa found, from their analysis of the new CCFR data on structure functions at small $`x`$, an alternative independent support for a softer $`\overline{s}`$, in agreement with the prediction of Ref. . Therefore we can check the possibility that the different behavior of spin transfer for $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ come from the strange-antistrange asymmetry of the nucleon sea. Indeed, the E665 data are measured corresponding to the quark distributions of the nucleon in the Bjorken variable range $`0.0001<x<0.1`$ with $`x=0.005`$, where the antiquark distributions are of the same order as those of the quark distributions. From the light-cone baryon-meson fluctuation model we know that the antistrange quark distribution could be as big as more than two times that of the strange quark distribution at small $`x`$. Therefore we can modify the strange and antistrange quark distributions of the CTEQ parametrization and check the role played by the strange-antistrange asymmetry for the $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ productions.
We choose the values of the quark distributions with quark-antiquark asymmetry of the nucleon sea at $`x=0.005`$ as
$$\begin{array}{ccccc}u=u_0+\delta u=50.32;& & & & \\ d=d_0+\delta d=45.41;& & & & \\ s=s_0+\delta s=17.118;& & & & \\ \overline{u}=\overline{u}_0+\delta \overline{u}=33.55;& & & & \\ \overline{d}=\overline{d}_0+\delta \overline{d}=35.29;& & & & \\ \overline{s}=\overline{s}_0+\delta \overline{s}=17.12+8;& & & & \end{array}$$
(12)
where $`q_0`$ and $`\delta q`$ ($`q=u,d,s,\overline{u},\overline{d},\overline{s}`$) are the quark distributions of CTEQ parametrization at $`x=0.005`$ and the corresponding modifications, respectively. In principle we can also introduce the nucleon sea quark-antiquark asymmetry in the light flavor $`u`$ and $`d`$ quarks, but we still have no phenomenological evidence for doing this. In comparison, there are large uncertainties concerning the strange and antistrange content of the nucleon sea. Therefore we first check the role played by the strange-antistrange asymmetry in the nucleon target for the difference between $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ productions in polarized charged lepton scattering on the nucleon. We present in Fig. 1(b) of the calculated results with strange-antistrange asymmetry. It is interesting to find that the strange-antistrange asymmetry of the nucleon sea predicted by the light-cone baryon-meson fluctuation model can indeed produce a trend for the different behaviors of the spin transfers for $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ production observed by E665, though the magnitude is still not enough to explain the data.
From Eq. (2), we find that the different strange and antistrange quark distributions of the nucleon sea are the reason for the different $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ production. The $`u`$ and $`d`$ ($`\overline{u}`$ and $`\overline{d}`$) quarks mainly contribute to the background of the $`\mathrm{\Lambda }`$ ($`\overline{\mathrm{\Lambda }}`$) production. In the quark-diquark model of the quark to $`\mathrm{\Lambda }`$ fragmentation functions , the quark helicities of $`u`$ and $`d`$ quarks have almost zero net contribution in the whole $`x`$ range $`01`$. But this does not seem to be true from the SU(3) symmetry argument that the $`u`$ and $`d`$ quarks may have net helicilities of the order of -0.2 . Therefore the absolute values of the spin transfers at small $`z`$ might not be correctly predicted by the quark-diquark model parametrization of quark to $`\mathrm{\Lambda }`$ fragmentation functions . The interesting aspect is the difference of the $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ productions from the strange-antistrange asymmetry of the nucleon sea. From Fig. 1(b) we notice that the magnitude of the difference can be the order of $`0.25`$, which should be large enough to cause an observed difference in the measurements of $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ productions in polarizied charged lepton DIS process on the nucleon. This shows that the strange quark content of the nucleon could be probed after we carefully consider the effect of $`u`$ and $`d`$ quarks and antiquarks of the nucleon, and of various quark to $`\mathrm{\Lambda }`$ fragmentation functions.
We would like to mention that the above conclusion depends on the specific forms of the quark to $`\mathrm{\Lambda }`$ fragmentation functions used as input for the spin transfer. We also present in Fig. 2(a) and (b) the calculated results with and without strange-antistrange asymmetry of the nucleon sea, but with the fragmentation functions from a pQCD based model which is also good in describing the data of $`\mathrm{\Lambda }`$ production in $`e^+e^{}`$ annihilation at the $`Z`$ resonance and in the polarized positron DIS on the proton by HERMES . We notice that the difference between the spin transfer for $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ is small in this situation. However, the sea quark-antiquark asymmetry in the quark and antiquark fragmentations to the $`\mathrm{\Lambda }`$ has been found to be an alternative possibility for the different behaviors of the spin tranfer for $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ production .
The above discussion helps us to understand why the strange quark-antiquark asymmetry can provide some contribution to the different behavior of $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ productions. This possibility can only manifest itself in the specific situation when the strange quarks and antiquarks are important and $`\overline{s}(x)>s(x)`$. Strictly speaking, there have been many experimental data related to the $`u`$ and $`d`$ quark and antiquark distributions so that there should be less freedom to introduce $`ϵ<0`$ for the $`u`$ and $`d`$ quarks. However, we notice that a possibility of $`\overline{d}(x)>d(x)`$ is not completely forbidden in the baryon-meson fluctuation picture to understand the Gottfried sum rule violation . Therefore we consider another case with an additional contribution of $`\overline{d}(x)>d(x)`$ in the target proton:
$$\begin{array}{ccccc}u=u_0+\delta u=50.32+8;& & & & \\ d=d_0+\delta d=45.418;& & & & \\ s=s_0+\delta s=17.118;& & & & \\ \overline{u}=\overline{u}_0+\delta \overline{u}=33.558;& & & & \\ \overline{d}=\overline{d}_0+\delta \overline{d}=35.29+8;& & & & \\ \overline{s}=\overline{s}_0+\delta \overline{s}=17.12+8.& & & & \end{array}$$
(13)
The calculated spin transfer for $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ production are presented in Figs. 1(c) and 2(c). We find that $`\overline{d}(x)>d(x)`$ could only provide a very small contribution to the different behaviors of $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ productions with fragmentation functions from both the quark-diquark model and the pQCD based model. This is due to the $`u`$ quark dominance and it also suggests that the $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ production at small $`x`$ is not sensitive to the $`d`$ and $`\overline{d}`$ quark distributions, but might be sensitive to the $`s`$ and $`\overline{s}`$ quark distributions, although there is strong model-dependence in this conclusion.
From the above general analysis we find that the different behaviors of $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ productions could be due to either the quark-antiquark asymmetries in the quark fragmentations and/or in the nucleon sea. The strange-antistrange asymmetry of the nucleon could provide a contribution to the observed difference of $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ productions, but the magnitude is still too small to explain the data. It thus forces us to consider the importance of the quark-antiquark asymmetry in the quark fragmentations if the E665 data are confirmed.
However, due to large uncertainties in the data and in the various quark to $`\mathrm{\Lambda }`$ fragmentation functions, it is still too early for us to arrive at some definite conclusion other than to suggest some interesting possibilities for further study. Thus we still need further efforts in order to reduce the uncertainties in the spin and flavor structure of various quark to $`\mathrm{\Lambda }`$ fragmentation functions. We know that the $`\mathrm{\Lambda }`$ and $`\overline{\mathrm{\Lambda }}`$ fragmentation in neutrino (antineutrino) DIS processes , and the different combinations of beam and target polarizations in the charged lepton DIS on the nucleon, can provide further insight on this issue, in addition to the $`\mathrm{\Lambda }`$ ($`\overline{\mathrm{\Lambda }}`$) fragmentation in the $`e^+e^{}`$ annihilation near the $`Z`$ resonance . We expect further theoretical and experimental work to push forward progress in this direction.
Acknowledgments: This work is partially supported by Fondecyt (Chile) postdoctoral fellowship 3990048, by the cooperation programmes Ecos-Conicyt and CNRS- Conicyt between France and Chile, by Fondecyt (Chile) grant 1990806 and by a Cátedra Presidencial (Chile), and by National Natural Science Foundation of China under Grant Numbers 19605006, 19875024, 19775051, and 19975052.
|
warning/0005/cond-mat0005132.html
|
ar5iv
|
text
|
# Exact sampling from non-attractive distributions using summary states
## I Introduction
In many statistical problems, physical and otherwise, it is useful to be able draw samples from a complex distribution. For example, in statistical physics one is interested in the Boltzmann distribution
$$P(\sigma )=\frac{e^{\beta E(\sigma )}}{Z},$$
(1)
where $`E(\sigma )`$ describes the energy of a system in configuration $`\sigma `$, $`\beta `$ is the inverse temperature (we set $`k_B=1`$), and $`Z`$ is a normalizing constant (the partition function). In general, $`E(\sigma )`$ may be easy to evaluate for a particular configuration, but the number of possible configurations makes it impractical to draw directly from the distribution. Yet some efficient method of sampling is desirable, as this would allow one to calculate properties of the system that might not be easily computed by analytical means.
In traditional Monte Carlo sampling methods, such as the Metropolis-Hastings method and Gibbs sampling (also known as the heat bath algorithm), one constructs an ergodic Markov chain whose stationary distribution is the desired distribution. By starting in some state and evolving the chain for a sufficiently long time, one can approximate a sample from the desired distribution. Unfortunately, such a sample is exact only in the limit of infinite time. In practice, it is often difficult to determine how long to wait to achieve sufficiently good samples, and one inevitably either produces poor samples or wastes time by running the Markov chain for longer than necessary.
However, in 1996, Propp and Wilson demonstrated the possibility of exact sampling by the method of coupling from the past, allowing one to produce perfect samples in a finite number of steps . In the most general case, their method requires the infeasible task of running a Markov chain for every possible initial state of the system. But for certain distributions, termed attractive (such as a ferromagnetic Ising model), Propp and Wilson showed that the task may be greatly simplified by tracking only extremal states, permitting the practical calculation of exact samples. This method was generalized to anti-attractive distributions by Häggström and Nelander . More recently, Huber showed that one may instead track only a single state that summarizes one’s knowledge of the system . Because it does not require that the states be partially ordered, this last method is applicable to non-attractive distributions.
Using the summary state method, we have drawn exact samples from the antiferromagnetic triangular Ising model and from the Potts model. Fig. 1 shows one such sample. In §II, we describe the methods that make this possible. In §III, we briefly discuss the Ising and Potts models. We present results from the exact sampling of these models in §IV. Finally, we discuss the convergence properties of the summary state method, and we suggest practical generalizations.
## II Coupling from the past and the summary state method
Propp and Wilson’s method of coupling from the past is based on the observation that, for a fixed choice of the random numbers used to propagate a Markov chain, its possible paths in state space may ultimately coalesce into a single trajectory. Once two initial states lead to the same state, they will remain in lock step thereafter.
Consider simulating a Markov chain from every possible initial state at some fixed time $`t=T`$, with the goal of taking a sample at $`t=0`$. If all the chains coalesce before $`t=0`$, then this finite procedure yields the same results as a Monte Carlo simulation started at an infinite time in the past, so the result is an exact sample. If the chains fail to coalesce, one can simply double the starting time to $`2T`$, reusing the random numbers for the interval $`[t,0]`$ (i.e., treating the random numbers as a function of simulation time), and repeat until coalescence is achieved.
Having to follow every possible state would make this method exponentially intractable. But for problems that admit a partial ordering of the states and which are “attractive” — that is, which preserve the ordering under evolution of the Markov chain — the computation can be vastly simplified by tracking only the extremal states. An example of an attractive system is the ferromagnetic Ising model, in which it is energetically favorable for spins to align with each other.
Huber and Harvey and Neal have shown that the method of Propp and Wilson may be generalized using a single summary state instead of a pair of extremal states. This single state summarizes one’s knowledge of the possible states of the system, allowing the state of some subsystems to be uncertain.
For example, suppose the system is a collection of variables $`\sigma _i`$ taking on the values $`\{\pm 1\}`$. Conventional Gibbs updating sets
$$\sigma _i\{\begin{array}{cc}+1& \mathrm{if}uP(\sigma _i=+1|\overline{\sigma }_i)\\ 1& \mathrm{if}u>P(\sigma _i=1|\overline{\sigma }_i)\end{array},$$
(2)
where $`u`$ is uniformly distributed on $`[0,1]`$ and $`\overline{\sigma }_i`$ denotes the set of all variables but the $`i`$th. To implement summary states, we allow each variable to take on the additional value ? which indicates uncertainty. We then run a modified Markov chain on this system: $`\sigma _i`$ is updated according to Eq. (2) if the result is the same for any possible assignment of $`\pm 1`$ to the ?’s in $`\overline{\sigma }_i`$; otherwise, $`\sigma _i\text{?}`$. As in the Propp and Wilson method, we run the chain from successively longer times in the past with random numbers as a function of simulation time. When no variables remain in ? states, the algorithm has converged, and we may take a sample at $`t=0`$.
For the case of attractive distributions, this procedure is exactly equivalent to the Propp and Wilson scheme. The value ? denotes variables that differ between the maximal and minimal states, and removal of all ? states corresponds to coalescence of the bounding chains. However, using a single summary state, there is no requirement that the states be ordered in any way. Thus the summary state method can also be applied to non-attractive distributions — for example, the antiferromagnetic Ising model.
Although the samples returned by this method are exact, the algorithm does not necessarily converge after a reasonable amount of time. Huber has shown that for antiferromagnetic spin systems at sufficiently high temperature, the expected running time of the algorithm is polynomial in the number of spins . However, for systems with a phase transition, the convergence time diverges as a power law at the critical temperature, a phenomenon known as critical slowing down .
## III The Ising and Potts models
Consider the Hamiltonian
$$E(\sigma )=\frac{1}{2}\underset{m,n}{}J_{mn}\sigma _m\sigma _n\underset{m}{}H_m\sigma _m,$$
(3)
where $`J_{mn}`$ is the coupling between spins $`m`$ and $`n`$ and $`H_m`$ is the value of an external magnetic field at the location of spin $`m`$. The appropriate Markov chain update rule is Eq. (2) with
$$P(\sigma _i=\pm 1|\overline{\sigma }_i)=\frac{e^{\beta E(\sigma _i=\pm 1)}}{e^{\beta E(\sigma _i=+1)}+e^{\beta E(\sigma _i=1)}}.$$
(4)
In the Ising model , $`J_{mn}`$ is taken to be zero unless spins $`m`$ and $`n`$ are adjacent, in which case it is some constant $`J`$. Cases of particular interest are the square lattice, in which each spin has four neighbors, and the triangular lattice, with six neighbors per spin. In general, the behavior of Ising systems can vary with their spin connectivity. For both kinds of lattices, we use periodic boundary conditions. Because we may simultaneously update the states of spins whose conditional distributions are independent, one iteration of the Markov chain consists of two sweeps for the square lattice and three for the triangular lattice.
In this paper, we use the normalization $`J=\pm 1`$. $`J=+1`$ corresponds to the ferromagnetic case, in which spins prefer to point in the same direction; $`J=1`$ corresponds to the antiferromagnet. As mentioned previously, the ferromagnetic case is attractive. The antiferromagnet on a square lattice is a special case, because its properties are isomorphic to those of a square ferromagnet. However, for a triangular lattice, there is no such isomorphism. With six neighbors per spin, there is no way to minimize the energy locally at all sites: we say the system is frustrated.
It is well known that a two-dimensional ferromagnetic Ising model exhibits a phase transition . Below a critical temperature $`\beta _c^1`$, there is spontaneous symmetry breaking, and the system develops a preferred spin orientation in the absence of any magnetic field. For a square lattice, $`\beta _c^1=2.27`$. At this temperature, the relaxation time of the dynamic system diverges, a phenomenon known as critical slowing down . Correspondingly, there is a divergence in the convergence time for some Markov chain Monte Carlo algorithms, such as coupling from the past, and exact samples cannot be generated for lower temperatures. Note that there is no phase transition in the case of a triangular antiferromagnet , so there cannot be a critical slowing down in the traditional sense.
To circumvent the problem of nonconvergence below the critical temperature, Propp and Wilson actually used a related system, the random cluster model, to generate Ising samples . Unfortunately, this model has no obvious analog in the antiferromagnetic case.
The Potts model is a generalization of the Ising model wherein spins may take on $`q`$ different values $`\{0,1,\mathrm{},q1\}`$ . Spins interact only with others of the same type. The Hamiltonian is
$$E(\sigma )=\frac{1}{2}\underset{m,n}{}J_{mn}\delta _{\sigma _m,\sigma _n}\underset{m,k}{}H_m^k\delta _{\sigma _m,k}.$$
(5)
Specifically, we consider the antiferromagnetic Potts model with $`q=3`$ on a square lattice with zero magnetic field. This model has a critical point only at $`\beta ^1=0`$, so there is no phase transition .
## IV Results
### A Ising model
By implementing the summary state method, we have produced exact samples from the Ising and Potts models. For example, Fig. 1 shows a sample from a triangular Ising antiferromagnet consisting of $`120^2=14,400`$ spins at $`\beta ^1=4.9`$ with zero applied magnetic field.
We find that the number of iterations required for the algorithm to converge diverges at a threshold temperature. We have studied this divergence using a lattice of $`N=63^2=3969`$ spins. Simulations using larger $`N`$ (e.g., $`N=99^2`$) suggest that the outcome is not significantly affected by choosing a larger grid size. Fig. 2 shows the divergence, to which we have fitted a power law of the form
$$t=\frac{a}{(\beta ^1\beta _t^1)^b}+c.$$
(6)
We find that the time diverges with an exponent $`b=1.03\pm 0.01`$ at the threshold temperature $`\beta _t^1=4.839\pm 0.005`$.
This divergence is an important feature of the summary state method. It is qualitatively similar to critical slowing down, but note that no physical phase transition is involved. In divergent situations the augmented Markov chain has a metastable set of distributions with many ?’s, such that it is very unlikely for it to enter a state with no ?’s.
To draw an exact sample using the summary state method, the system must go from a completely uncertain state to a completely certain state. Thus, it must pass through a state with only a few scattered ?’s. For temperatures sufficientlty near the threshold, where we know that such a sparse configuration can be reached, we might expect that the limiting factor is the probability that an isolated ? can cause divergence.
Therefore, as a very rough estimate, we might suppose that the divergence occurs when the probability of a single ? turning one of its six neighbors into a ? rises above $`\frac{1}{6}`$. We expect that the neighbors of any given spin $`\sigma _i`$ should be (on average) half up and half down. Replacing one of these neighbors by a ?, we may assume the configuration $`(\text{?})`$ without loss of generality. Then the threshold temperature is determined by
$$1\frac{1}{1+e^{4\beta }}\frac{1}{2}=\frac{1}{6},$$
(7)
which has the solution $`\beta ^1=4/\mathrm{ln}25.8`$.
To examine the validity of a threshold temperature analysis based on the persistence of single ?’s, we compiled statistics on the stability of an equilibrium system with a single ? added. Because we cannot create exact samples for much of the temperature range of interest, we generated approximate samples by simulating for fixed time (100 iterations) a random initial state. We then set one spin to ? and simulated the system forward. If any uncertainty remained after 500 iterations, we said the system diverged. Fig. 3 shows the fraction of divergent trials for various temperatures. As one would expect, this fraction goes to zero very near the threshold temperature.
It is also interesting to consider how the algorithm behaves when a uniform nonzero magnetic field $`H`$ is applied. Biasing the spins makes it easier for them to choose a particular orientation, so we would expect convergence to be easier. Fig. 4 shows the region of convergence in the $`(\beta ^1,H)`$ plane.
### B Potts model
In addition, we have implemented exact sampling of the Potts model for arbitrary $`q`$. Fig. 5 shows an exact sample with $`q=3`$ for a square antiferromagnetic lattice of $`100^2=10,000`$ spins.
A naïve implementation of the summary state method would augment the possible spin values with a single ?. We refer to this method as algorithm $`A`$. However, it is possible to retain more information about uncertain spins: for each spin, we store a binary $`q`$-bit vector $`(b_1,b_2,\mathrm{},b_q)`$, $`b_i\{0,1\}`$. Bit $`b_i`$ is set to one if it is possible for the spin to take on the value $`i`$: thus the initial state of each spin is $`b=(1,1,\mathrm{},1)`$. In updating the state of the system, we set $`b_i=0`$ only when the spin cannot take on the value $`i`$ for any allowed configuration of its neighbors. We refer to the latter method as algorithm $`B`$.
To demonstrate the advantage of retaining more information in the summary state, we have studied the convergence properties of both algorithms. This comparison is shown in Fig. 6, based on data for a square $`64^2=4096`$ spin lattice. As in the Ising study, both algorithms lead to a power law divergence with an exponent of one ($`b_A=1.04\pm 0.03`$, $`b_B=0.99\pm 0.02`$). However, the threshold temperatures for the two algorithms are quite different: $`\beta _{t,A}^1=2.293\pm 0.005`$, whereas $`\beta _{t,B}^1=1.157\pm 0.004`$. As in the Ising example above, neither of the divergences corresponds to a physical phase transition.
## V Conclusions
We have demonstrated the usefulness of the summary state method for exact sampling from non-attractive distributions. In both the antiferromagnetic Ising and Potts models, the method works above a certain threshold temperature, with a power law divergence in the coalescence time at the threshold. Although similar to the phenomenon of critical slowing down, this divergence does not occur at a physical phase transition. As the Potts example shows, the location of the divergence is a feature of the specific implementation of the summary states, not of the underlying distribution. We have shown that retaining more information in the summary state will allow convergence at lower temperatures.
Based on this result, we may propose an improved algorithm for the triangular antiferromagnetic Ising model. At lower temperatures, the system should be increasingly ordered, and tracking this order might make it easier to gain incremental knowledge of the state of the system. One idea is to keep track of correlations between spins by grouping them into hexagonal clumps of seven, which can be used to tile the triangular lattice. Each tile has $`2^7=128`$ possible states. In analogy to the Potts method presented earlier (in which a $`q`$-bit vector represents the uncertainty about a spin), representing each tile with a 128-bit vector would allow individually tracking the possible arrangements of those seven spins. Within a tile, the summary state can track anticorrelation, which we expect to arise at low temperatures. Each edge of a tile can be easily summarized in a 4-bit vector for comparison with its neighboring tiles. Each tile can then update its summary state by considering the possible states of the neighboring edges.
Of course, summary state sampling remains a valuable tool even if it cannot be done below some threshold. Where it does work, it is exact. The method also has the advantage that we need not even know if the distribution in question is attractive — it can be applied to any system.
## VI Acknowledgments
We wish to thank Radford Neal and David Wilson for several helpful discussions. AMC and RBP acknowledge the support of the Caltech Cambridge Scholars Program. DJCM’s group is supported by the Gatsby Charitable Foundation and by a Partnership Award from IBM Zürich.
|
warning/0005/cond-mat0005005.html
|
ar5iv
|
text
|
# On the electron scattering and dephasing by the nuclear spins.
## Abstract
We show that scattering of the conduction electrons by nuclear spins via the hyperfine interaction may lead the upper limit on the mean free path in clean metals. Nuclear spins with s $`>`$1/2 may cause a strong dephasing in dirty limit due to the quadrupole coupling to the random potential fluctuations caused by static impurities and lattice imperfections.
.
Due to the sharply growing interest to the quantum information processing the study of the electron charge and spin transport in solids has been refocused to the problems of intrinsic and extrinsic sources of decoherence -. While at low temperatures the phonon scattering is eliminated, the impurities and electron interactions remain the main scattering mechanisms . The magnetic impurities, which can strongly influence the electron transport, resulting in e.g. Kondo effect, could be eliminated either by cleaning of the material or by freezing out by a strong magnetic field. In most of intensively studied conductors there exist, however, an intrinsic bath of magnetic scatterers: the nuclear spins. A weak influence of the nuclear spins on the resistivity in strongly doped bulk semiconductors was reported in . A striking example of their influence on the electron transport in low dimensional semiconductors is the observation of sharp spikes in magnetoresistance under the quantum Hall effect conditions.
Much less attention was paid to the magnetic scattering of conduction electrons by nuclear spins in metals. There is however strong evidence that conventional scattering can not explain anomalies in residual resistivity at low temperature . While the normal metals have a quite similar electronic structure, the experimentally observed temperature dependence of the dephasing time $`\tau _\phi `$ may be quite different. This was shown in very recently in , where the value of $`\tau _\phi `$ was defined by the magnetoresistance measurements of long metallic wires $`Cu,Au,Ag`$ in a wide temperature interval $`10^2<T<10^^o`$K. In $`Cu`$ and $`Au`$ wires $`\tau _\phi `$ saturates at low temperatures which contradicts the standard theory . Strangely enough the Ag wires do not show the saturation to the lowest temperatures, in accordance with . The possibility of the hyperfine origin of this discrepancy will be discussed later in this paper.
Here we study the contribution of the hyperfine contact (Fermi) interaction between the conduction electrons and nuclear spins to the temperature and magnetic field dependence of resistivity $`\rho (T,H)`$ . We show that as a result of electron-nuclear interaction the residual resistivity in isotopically clean metals is not vanishing even when the impurity concentration $`C_o0`$ (the universal residual resistivity, URR). The space periodicity of nuclei is of no importance, as long as the nuclear spins are disordered and acts as magnetic impurities with the concentration $`C_n1`$ . It follows that in this temperature interval URR reflects the existence of an upper limit for the mean free path of conduction electrons. This scattering is not operative at extremely low temperatures ($`T`$ $`10^7`$ $`{}_{}{}^{o}K`$in Cu, for example ) when the nuclear spins are ordered.
The residual ”nuclear ”resistivity is due to the Fermi (contact) hyperfine interaction between the nuclear and the conduction electron spins:
$$V_{en}=\frac{8\pi }{3}\mu _e\mu _h\mathrm{\Psi }_e^2(0)\mu _nH_e$$
(1)
here $`\mu _e`$ and $`\mu _h`$ are the operators of the electron and nuclear magnetic moments, $`\mathrm{\Psi }_e^2(0)`$ $`Z`$ is the value of the conduction electron wave function on the nuclei with the nuclear charge $`Z`$ and $`H_e`$ is the magnetic field induced on nuclei by the electrons.
Let us estimate $`V_{en}`$ . In atomic units: $`\mathrm{}=m_e=e=1`$
$$V_{en}Z\alpha ^2\frac{m_e}{m_n}Ry$$
(2)
where $`m_e,m_h`$ are the electron and the nucleon masses, respectively; $`Ry=27`$ ev and $`\alpha =\frac{1}{137}`$ is the fine structure constant.
In metals the effective electron-nuclear interaction constant is
$$g_n\frac{V_{ne}}{ϵ_F}10^7Z\frac{Ry}{ϵ_F}$$
(3)
where the Fermi energy $`ϵ_F`$ varies in wide interval $`\left(0.01÷1\right)Ry`$.The interaction constant $`g_n`$ varies from is $`10^6`$ for $`Li`$ to $`10^1`$ in doped semiconductors with low $`ϵ_F`$ . This estimate of $`g_n`$ is in a good agreement with the experimentally observed values of $`H_e`$ on the nuclei .
The total residual resistivity is therefore a sum of the impurity $`\rho _o(T0)C_o`$ and the nuclear spin $`\rho _n(T0)g_n^2`$ contributions:
$$\rho _o^+(0^+)\rho _{oo}(C_o+g_n^2)$$
which follows also from calculations, based on the magnetic impurity scattering technics introduced in . Here $`0^+`$ is the limit $`T0`$ , while $`TT_c`$,where $`T_c`$ is the temperature of the nuclear ordering, and $`\rho _{oo}1`$ in atomic units:$`\rho _{oo}10^{17}`$ sec . The nuclear contribution to resistivity starts to be operative when the impurity concentration is $`C_og_n^2`$ .
In the limit of an ideally pure ($`C_o=0`$) metal the universal residual resistivity $`\rho _{URR}`$is, therefore $`\rho _{URR}\rho _{oo}g_n^2`$ and the mean free path is limited by $`\frac{10^8}{g_n^2}`$ cm. This yields $`10^4`$ cm in $`Li`$ and $`10^2`$ cm for the rear earth metals. It is interesting to note that in materials with even-even nuclei (zero spin) , like in $`Ca`$ ,$`Ni`$ , $`Fe`$, $`Ce`$ and isotopically clean graphite $`C`$ , where the electron-nuclear scattering is absent, the URR would not be observed.
Consider now the contribution to the temperature and magnetic field dependence of the residual resistivity caused by the hyperfine interaction between the conduction electrons and the nuclear spins. The temperature and the magnetic field dependence of residual resistivity due to nonmagnetic impurities is due mostly to the mesoscopic effects, and is vanishing in the limit $`C_o0`$ . In a magnetic field such that $`\mu _eHT`$ the magnetic impurities freeze out and the Kondo effect is quenched. In order to freeze out the nuclear spins however one should apply much higher magnetic fields, $`\mu _nHT`$ . Therefore in the temperature interval $`\mu _eHT`$ $`\mu _nH`$ the nuclear spin contribution may prevail in metals with magnetic impurities.
The temperature and the magnetic field dependence of the electron-nuclear scattering contribution to resistivity can be written as
$$\rho _n(T)=\rho _n(\mathrm{})f_n(x)$$
(4)
where $`x=\frac{\mu _nH}{T}`$ and the asymptotic of the function $`f_n(x)`$ is given in . Nuclei with spin $`\frac{1}{2}`$ in a magnetic field are equivalent to a two-level system and the function $`f_n(x)`$ can be defined analytically by methods developed in to be:
$$f_n(x)=\frac{2x}{sh2x}$$
(5)
In the limit $`T\mu _nH`$ the temperature dependent part of $`\rho `$ is
$$\rho (T)\rho (0^+)\rho _{00}\left(\frac{T^2}{\epsilon _F^2}g_n^2\frac{(\mu _nH)^2}{T^2}\right)$$
(6)
Since the recent experimental data are plotted as $`\frac{\rho }{T}`$ versus $`T`$ we note that the derivative $`\frac{\rho }{T}`$ experiences a minimum at $`T\sqrt{g\epsilon _F\mu _nH}.`$
In metals like $`Li,Na,K,Rb,Cs,Au,Cu,Al,In`$ the nuclear magnetic moments $`I\frac{1}{2}`$ and even without external magnetic field their $`2I+1`$ degeneracy is lifted partially by the quadrupole effects (in the case of cubic crystal symmetry the quadrupole splitting of the nuclear levels may happen due to the defects ,,) . The hyperfine nuclear contribution to $`\rho _n(T)`$ in this case will have the temperature dependence as in Eq. LABEL:fn1 , where $`\mu _nH`$ should be replaced by the characteristic quadrupole splitting of energy levels.
The influence of the quadrupole nuclear spin splitting on the phase coherence time $`\tau _\phi `$ can be the clue to the puzzling difference between the low temperature dependence of $`\tau _\phi `$ in $`Cu,Au`$ and $`Ag`$ wires , observed in . Indeed, the nuclear spins of both $`Cu`$ and $`Au`$ have a strong quadrupole moment ($`s=3/2`$) and may act as inelastic two-level scatterers ,, once their degeneracy is lifted by the static impurities and other imperfections. The quadrupole splitting in these materials is known to be of the order of $`\mathrm{\Delta }_Q10^3÷10^2`$ K .
This is not the case for $`Ag`$ nuclei since their spin is $`s=\frac{1}{2}`$ .In the absence of magnetic Zeeman splitting present, for an electron spin, just a set of elastic scatterers, and the temperature dependence of $`\tau _\phi `$ should obey the s$`\mathrm{tan}`$dard theory , which indeed the case in the experiments .
It is interesting to continue the measurements of $`\tau _\phi `$ on other materials with (as $`Al:,s=5/2`$) and without ($`Pt`$ and $`Sn,s=1/2`$) the quadrupole nuclear spin splitting.
The Kondo effect appears in $`\rho _n(T)`$ in higher orders of $`g`$ Fig. 1. In analogy with the magnetic impurities the first temperature correction to $`\rho _n(T)`$ is
$$\delta \rho _1\rho _{oo}g_n^3\mathrm{ln}\frac{\epsilon _F}{T}$$
(7)
For positive magnetic nuclear moments one has $`g_n>0`$ and the interaction between the electron and the nuclear spins favors the antiferromagnetic ordering of moments $`\mu _e`$ and $`\mu _n`$. This gives the usual Kondo effect on nuclei, i.e. $`\rho (T)`$ has a minimum at $`T_og^{\frac{3}{2}}\epsilon _F`$ . For metals with large $`Z`$: $`T_o10^3÷10^2`$ $`{}_{}{}^{o}K`$ . Note, that for the nuclear contribution to $`\rho `$ there is no need to summarize the powers of $`\mathrm{ln}\frac{\epsilon _F}{T}`$ , since the nuclear Kondo temperature $`T_k=\epsilon _Fe^{\frac{1}{g_n}}`$ is so low that the nuclear spin interaction start to play the main role. These are the direct dipole-dipole interaction between nuclei and their interaction via conduction electrons. The interaction constant $`J^+`$ is of the order of $`T_c`$ , the temperature of the nuclear ordering. In the case when the interaction via conduction electrons is stronger than the direct one , $`J^+`$ $`g^2\epsilon _F`$ . It can be shown that the contribution to $`\rho `$ from the nuclei-nuclei interactions
$$\delta \rho _{}\rho _{oo}g_n^2\frac{J^+}{T}$$
(8)
are comparable to these of the electron contribution $`\rho _{oo}\left(\frac{T}{\epsilon _F}\right)^2`$ at $`T_1\left(J^+\epsilon _F^2\right)^{\frac{1}{3}}g^{\frac{2}{3}}\epsilon _Fg^{\frac{4}{3}}`$.
By comparing $`T_o`$ and $`T_1`$ one concludes that while considering the Kondo effect on nuclear spins the nuclear spin interaction should not be neglected, since the nuclear spin concentration is always of order of unity. This to be compared with the usual Kondo effect in the case of low concentration of magnetic impurities $`C_m`$ , where the interaction between the localized moments are of the second order with respect to $`C_m`$. The correction to $`\rho `$ , Eq. LABEL:fn1 is analogous to the correction $`J^+`$ to the nuclear susceptibility
$$\chi _n(T)\frac{1}{T}\left(1+\frac{J^+}{T}\right)$$
By measuring the URR in a metal one can therefore establish the sign of $`J^+`$ and to predict the type of the nuclear order at very low temperature.
The high temperature $`TJ^+T_c`$expansion of the residual resistivity of a nonmagnetic metal is
$$\frac{\rho (T)}{\rho _{oo}}C_o+\left(\frac{T}{\epsilon _F}\right)^2+g_n^2\left(1+g_n\mathrm{ln}\frac{\epsilon _F}{T}+\frac{J^+}{T}\right)$$
(9)
It follows, from Eq. 9 that in metals with large $`Z`$ the nuclear effects will contribute to resistivity already at temperatures of the order of $`0.1^oK.`$ Note that the nuclear contribution is nonanalytical in $`T`$ and should be compared with the vanishing, at low temperatures, $`T^2`$ contribution rather than with $`\rho _{oo}C_o`$ (see for more details ).
In conclusion we have suggested that the hyperfine interaction between the conduction electron spins and nuclear spins may result in universal residual resistivity in clean metals at low temperatures. Apart of the fundamental nature of this problem, the natural limitations on the mean free path are decisive in the semiconductor based high speed electronic devices, like heterojunctions and quantum wells. We outline, that the nuclear spin quadrupole splitting due to the static imperfections may be partly responsible for the low temperature behavior of resistivity in such metals as $`Au`$ and $`Cu`$. This mechanism should not be operative in $`Ag`$ where the nuclear spin is $`s=1/2`$ , in agreement with the recent experimental observations . We note also that the influence of the nuclear spins on resistivity should disappear at very low temperatures, where the nuclear spins magnetically order (see e.g. ,).
We acknowledge illuminative discussions with B. Altshuler, D. Esteve, T. Hermannsdorfer,V. Kravtsov, T.Maniv, B. Spivak, P.C.E. Stamp and R. Webb.
|
warning/0005/cond-mat0005192.html
|
ar5iv
|
text
|
# Chain stiffness intensifies the reptation characteristics of polymer dynamics in the melt
## Acknowledgements
We thank B. Dünweg, R. Everaers, A. Heuer, K. Kremer, and M. Pütz for fruitful discussions. Financial support from the German ministry of research (BMBF) is gratefully acknowledged.
|
warning/0005/math-ph0005014.html
|
ar5iv
|
text
|
# The electromagnetic self-force on a charged spherical body slowly undergoing a small, temporary displacement from a position of rest
## Abstract
The self-force of classical electrodynamics on a charged “rigid” body of radius $`R`$ is evaluated analytically for the body undergoing a slow (i.e., with a speed $`vc`$), slight (i.e., small compared to $`R`$), and temporary displacement from an initial position of rest. The results are relevant to the Bohr–Rosenfeld analysis of the measurability of the electromagnetic field, which has been the subject of a recent controversy.
1. Introduction
The problem of the classical electromagnetic self-force on an extended charged body moving along a given trajectory is interesting in its own right; its analysis under some greatly simplifying conditions is a key ingredient of the well known paper on the measurability of the electromagnetic field of Bohr and Rosenfeld (BR) . BR derived an eight-dimensional-integral expression for a time average $`\overline{F}_{\mathrm{BR}x}`$ of the self-force plus the electrostatic force due to a stationary neutralizing body of opposite charge, acting on a charged “rigid” body<sup>*</sup><sup>*</sup>* Of course, absolute rigidity is not allowed in a relativistic theory, and BR went to great lengths to justify an assumption that the body is only rigid to the degree that all its parts participate sufficiently uniformly in the body’s assumed motion. that is undergoing an $`x`$ direction displacement whose time dependence $`D_x(t_1)`$ approaches sufficiently closely a steplike trajectory $`D_x\mathrm{\Theta }(Tt_1)\mathrm{\Theta }(t_1)`$, $`D_x=\mathrm{const}`$ \[$`\mathrm{\Theta }(x)`$ is the Heaviside step function\]. The expression of BR is written as
$$\overline{F}_{\mathrm{BR}x}=\rho _c^2V^2TD_x\overline{A}_{xx}^{(\mathrm{I},\mathrm{I})}$$
(1)
where $`V`$ and $`\rho _c`$ are the displaced body’s volume and constant charge density, respectively, and the quantity $`\overline{A}_{xx}^{(\mathrm{I},\mathrm{I})}`$ is the BR geometric factor for two fully coinciding space-time regions I of volume $`V`$ and duration $`T`$ :
$$\overline{A}_{xx}^{(\mathrm{I},\mathrm{I})}=\frac{1}{V^2T^2}_Tdt_1_Tdt_2_Vd𝒓_1_Vd𝒓_2\left(\frac{^2}{x_1x_2}\frac{^2}{t_1t_2}\right)\frac{\delta (tr)}{r}.$$
(2)
Here and henceforth, $`t=t_2t_1`$, $`r=|𝒓_2𝒓_1|`$, and units such that the speed of light $`c=1`$ are used. This result is valid only when the displacement $`|D_x|a`$, where $`a`$ characterizes the linear dimensions of the body, and the speedWe shall display explicitly the factor $`c`$ in some inequalities. $`|\dot{D}_x(t_1)|c`$—which implies a further condition that $`|D_x|c\mathrm{\Delta }t`$, where $`\mathrm{\Delta }tT`$ is the duration of the time intervals during which the displacement $`D_x(t_1)`$ goes smoothly from zero to the constant value $`D_x`$ and from $`D_x`$ back to zero at the beginning and end, respectively, of the given time period $`0t_1T`$.
Recently, Compagno and Persico (CP) have questioned the use of a steplike trajectory in the BR calculation of the self-force, and have drawn the conclusion that the BR result that a single space-time-averaged component of the electromagnetic field can be measured to arbitrary accuracy only using a compensating spring is incorrect since it is based on the expression (1) that assumes an unphysical steplike trajectory. The paper of CP is criticized in a Comment , where it is shown by an explicit calculation that the limiting BR time-averaged self-force (1) approximates correctly the self-force obtained with a “physical” trajectory of $`|\dot{D}_x(t_1)|c`$ but with a sufficiently short duration $`\mathrm{\Delta }t`$ of the initial and final trajectory segments outside which the body is essentially at rest. In the Reply of CP , this criticism is rejected, claiming that the calculation in is incorrect.
In the present paper, we obtain analytical expressions for the time dependence as well as a time average of the self-force on a spherical charged “rigid”We employ here the same concept of rigidity as BR (see the first footnote). body of radius $`R`$ moving on a trajectory that is subject to the special BR conditions but is not necessarily of a steplike character. Exploiting the spherical symmetry of the problem, we perform the requisite integrations directly with no recourse to the Fourier transform methods used in , but in full agreement with the results obtained using the Fourier transform method in and rejected by CP as incorrect. The expressions obtained are relatively simple, and it is surprising that such or similar results do not seem to have appeared in the literature before (with partial exception of papers )—which perhaps is a factor behind the recently expressed reluctance to accept them, and their implications, as correct.
2. The time average of the self-force
First, we outline the derivation of a multidimensional-integral expression for the self-force in terms of the body’s trajectory $`D_x(t_1)`$ that, while conforming to the BR conditions $`|D_x(t_1)|R`$ and $`|\dot{D}_x(t_1)|c`$, is not necessarily of a steplike character—apart from satisfying the condition that $`D_x(t_1)=0`$ for $`t_1<0`$ and $`t_1>T`$. A detailed derivation of such an expression has been given by CP —but under the complicating conditions of a temporary removal of the neutralizing body, which we shall consider simply as only permanently absent or present. The displaced body’s time-dependent charge density is approximated to first order in a displacement $`𝑫(t_1)`$ as
$$\rho (𝒓_1,t_1)=\rho [𝒓_1𝑫(t_1)][1𝑫(t_1)\mathbf{}\mathbf{}_1]\rho (r_1)$$
(3)
where $`\rho (r_1)`$ is the body’s spherically symmetric charge density before its displacement. Using this approximation and placing the differential operator $`𝑫(t_1)\mathbf{}\mathbf{}_1`$ suitably using integration by parts, the retarded potentials of the electromagnetic self-field of the body can be expressed to first order in the displacement and neglecting also terms of order $`\dot{D}D`$ as
$`\varphi (𝒓_2,t_2)`$ $`=`$ $`{\displaystyle d𝒓_1_{\mathrm{}}^{\mathrm{}}dt_1\frac{\rho (𝒓_1,t_1)}{r}\delta (tr)}`$ (4)
$`=`$ $`{\displaystyle d𝒓_1\frac{\rho (r_1)}{r}}+{\displaystyle d𝒓_1\rho (r_1)_{\mathrm{}}^{\mathrm{}}dt_1𝑫(t_1)\mathbf{}\mathbf{}_1\frac{\delta (tr)}{r}}`$ (5)
$`𝑨(𝒓_2,t_2)`$ $`=`$ $`{\displaystyle d𝒓_1_{\mathrm{}}^{\mathrm{}}dt_1\frac{\rho (𝒓_1,t_1)\dot{𝑫}(t_1)}{r}\delta (tr)}`$ (6)
$`=`$ $`{\displaystyle d𝒓_1\rho (r_1)_{\mathrm{}}^{\mathrm{}}dt_1\dot{𝑫}(t_1)\frac{\delta (tr)}{r}}.`$ (7)
Assuming now that the displacement is along the $`x`$ direction, the $`x`$ component of the body’s electric self-field can be written as
$`E_x(𝒓_2,t_2)`$ $`=`$ $`{\displaystyle \frac{\varphi (𝒓_2,t_2)}{x_2}}{\displaystyle \frac{A_x(𝒓_2,t_2)}{t_2}}`$ (8)
$`=`$ $`{\displaystyle d𝒓_1\frac{}{x_2}\frac{\rho (r_1)}{r}}`$ (10)
$`{\displaystyle d𝒓_1\rho (r_1)_{\mathrm{}}^{\mathrm{}}dt_1D_x(t_1)\left(\frac{^2}{x_1x_2}\frac{^2}{t_1t_2}\right)\frac{\delta (tr)}{r}}.`$
(The magnetic field is neglected in view of the assumption that the body’s speed $`|\dot{D}_x(t_1)|c`$.) This results in a self-force $`_x(t_2)`$ on the displaced body given to first order in $`D_x`$ by
$$_x(t_2)=d𝒓_2\left\{\left[1D_x(t_2)\frac{}{x_2}\right]\rho (r_2)\right\}E_x(𝒓_2,t_2)=F_{0x}(t_2)+F_x(t_2)$$
(11)
where
$$F_{0x}(t_2)=D_x(t_2)\rho _c^2_{|𝒓_1|<R}d𝒓_1_{|𝒓_2|<R}d𝒓_2\frac{^2}{x_2^2}\frac{1}{r}$$
(12)
which equals<sup>§</sup><sup>§</sup>§This follows on the replacement of $`^2/x_2^2`$ in (12) by $`\frac{1}{3}_2^2`$, which is allowed by the spherical symmetry of the problem, and the use of $`_2^2(1/r)=4\pi \delta ^{(3)}(𝒓_2𝒓_1)`$. $`\rho _c^2V^2D_x(t_2)/R^3`$ and is, for $`|D_x|R`$, the electrostatic repulsive force that would be due to an identical body placed at the undisplaced position; and
$$F_x(t_2)=\rho _c^2_{|𝒓_1|<R}d𝒓_1_{|𝒓_2|<R}d𝒓_2_0^Tdt_1D_x(t_1)\left(\frac{^2}{x_1x_2}\frac{^2}{t_1t_2}\right)\frac{\delta (tr)}{r}$$
(13)
which would be the net force in the presence of an oppositely charged neutralizing body occupying permanently the space region of the undisplaced body The bodies may be assumed to have a fine tubular structure that enables them to move without hindrance through each other along a given direction (see )., as the electrostatic force of attraction to the neutralizing body would cancel the force $`F_{0x}(t_2)`$ of equation (12). We assumed in equations (12) and (13) that the body has a constant charge density $`\rho _c`$, and replaced the infinite region of the time integration with the time interval $`(0,T)`$ in view of the fact that the displacement $`D_x(t_1)0`$ outside this interval. We shall be calling, following CP, the force $`F_x(t_2)`$ of equation (13) also a “self-force.”
We now evaluate the time average of the self-force as a one-dimensional quadrature involving the body’s trajectory $`D_x(t_1)`$. A time-averaged self-force $`\overline{F}_x`$ is obtained by averaging the expression (13) with respect to time $`t_2`$,
$$\overline{F}_x=\frac{1}{T}_0^Tdt_2F_x(t_2)$$
(14)
and we note that when the trajectory $`D_x(t_1)`$ in (13) is replaced formally by a steplike trajectory $`D_x\mathrm{\Theta }(Tt_1)\mathrm{\Theta }(t_1)`$, the averaging results in the limiting BR time-averaged self-force $`\overline{F}_{\mathrm{BR}x}`$ of equation (1). The time-averaged self-force (14) can be written as
$$\overline{F}_x=\frac{\rho _c^2V^2}{T}_0^Tdt_1D_x(t_1)f(t_1)V=\frac{4}{3}\pi R^3$$
(15)
where the function $`f(t_1)`$ is defined by
$$f(t_1)=\frac{1}{V^2}_{|𝒓_1|<R}d𝒓_1_{|𝒓_2|<R}d𝒓_2_0^Tdt_2\left(\frac{^2}{x_1x_2}\frac{^2}{t_1t_2}\right)\frac{\delta (tr)}{r}.$$
(16)
We note with CP that the function $`f(t_1)`$ can be written as
$$f(t_1)=\frac{1}{V^2}_{|𝒓_1|<R}d𝒓_1_{|𝒓_2|<R}d𝒓_2_0^Tdt_2\left[\frac{1}{3}\left(_2^2\frac{^2}{t_2^2}\right)+\frac{2}{3}\frac{^2}{t_1t_2}\right]\frac{\delta (tr)}{r}$$
(17)
because $`/x_1=/x_2`$ and $`/t_1=/t_2`$ when operating on $`\delta (tr)/r`$, and $`^2/x_2^2`$ can be replaced by $`\frac{1}{3}_2^2`$ in view of the spherical symmetry of the problem. Using the well-known equation for the retarded Green’s function $`\delta (tr)/r`$,
$$\left(_2^2\frac{^2}{t_2^2}\right)\frac{\delta (tr)}{r}=4\pi \delta ^{(3)}(𝒓_2𝒓_1)\delta (t_2t_1)$$
(18)
and performing also the integration with respect to $`t_2`$ in the second term of (17), $`f(t_1)`$ is obtained as
$`f(t_1)=`$ $`{\displaystyle \frac{1}{R^3}}[\mathrm{\Theta }(Tt_1)\mathrm{\Theta }(t_1)]`$ (20)
$`{\displaystyle \frac{2}{3V^2}}{\displaystyle _{|𝒓_1|<R}}d𝒓_1{\displaystyle _{|𝒓_2|<R}}d𝒓_2{\displaystyle \frac{1}{r}}[\delta ^{}(Tt_1r)\delta ^{}(t_1r)].`$
Defining an integral
$$I(s)=\frac{1}{V^2}_{|𝒓_1|<R}d𝒓_1_{|𝒓_2|<R}d𝒓_2\frac{\delta ^{}(sr)}{r}V=\frac{4}{3}\pi R^3r=|𝒓_2𝒓_1|$$
(21)
the function $`f(t_1)`$ can now be written as
$$f(t_1)=\frac{1}{R^3}[\mathrm{\Theta }(Tt_1)\mathrm{\Theta }(t_1)]\frac{2}{3}[I(Tt_1)I(t_1)].$$
(22)
Due to the symmetry of the problem, the integral (21) reduces to a three-dimensional quadrature:
$$I(s)=\frac{9}{2R^3}_0^1d\zeta _1\zeta _1^2_0^1d\zeta _2\zeta _2^2_1^1dx\frac{\delta ^{}(\xi \sqrt{\zeta _1^2+\zeta _2^22\zeta _1\zeta _2x})}{\sqrt{\zeta _1^2+\zeta _2^22\zeta _1\zeta _2x}}\xi =\frac{s}{R}.$$
(23)
Let us do the integration with respect to $`x`$ first. On the substitution $`\xi (\zeta _1^2+\zeta _2^22\zeta _1\zeta _2x)^{1/2}=y`$, this yields
$$_1^1dx\frac{\delta ^{}(\xi \sqrt{\zeta _1^2+\zeta _2^22\zeta _1\zeta _2x})}{\sqrt{\zeta _1^2+\zeta _2^22\zeta _1\zeta _2x}}=\frac{\delta [\sqrt{(\zeta _1\zeta _2)^2}\xi ]\delta [\sqrt{(\zeta _1+\zeta _2)^2}\xi ]}{\zeta _1\zeta _2}.$$
(24)
The integration with respect to a radial variable, say $`\zeta _1`$, can be done using the rule $`\delta [f(x)]=_i\delta (xx_i)/|f^{}(x_i)|`$, where $`x_i`$ are the roots of $`f(x)=0`$; $`\delta [f(x)]=0`$ when there are no real roots of $`f(x)=0`$. When $`\xi 0`$, the roots of $`f_+(\zeta _1)[(\zeta _1+\zeta _2)^2]^{1/2}\xi `$ are $`\zeta _{1i}^{(+)}=\zeta _2\pm \xi `$, and the roots of $`f_{}(\zeta _1)[(\zeta _1\zeta _2)^2]^{1/2}\xi `$ are $`\zeta _{1i}^{()}=\zeta _2\pm \xi `$, with $`|f_\pm ^{}(\zeta _{1i}^{(\pm )})|=1`$ for all these roots; when $`\xi <0`$, there are no real roots of $`f_\pm (\zeta _1)=0`$. We thus obtain
$`K(\zeta _2,\xi )`$ $``$ $`{\displaystyle _0^1}d\zeta _1\zeta _1^2{\displaystyle \frac{\delta [\sqrt{(\zeta _1\zeta _2)^2}\xi ]\delta [\sqrt{(\zeta _1+\zeta _2)^2}\xi ]}{\zeta _1\zeta _2}}`$ (25)
$`=`$ $`\mathrm{\Theta }(\xi ){\displaystyle _0^1}d\zeta _1{\displaystyle \frac{\zeta _1}{\zeta _2}}[\delta (\zeta _1\zeta _2+\xi )+\delta (\zeta _1\zeta _2\xi )\delta (\zeta _1+\zeta _2+\xi )\delta (\zeta _1+\zeta _2\xi )]`$ (26)
$`=`$ $`{\displaystyle \frac{\mathrm{\Theta }(\xi )}{\zeta _2}}[(\zeta _2+\xi )\mathrm{\Theta }(1\zeta _2\xi )+(\zeta _2\xi )\mathrm{\Theta }(1+\zeta _2\xi )]\mathrm{\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}0}<\zeta _2<1.`$ (27)
Finally,
$$I(s)=\frac{9}{2R^3}_0^1d\zeta _2\zeta _2^2K(\zeta _2,\xi )=\frac{3}{4R^3}(2\xi )(22\xi \xi ^2)\mathrm{\Theta }(\xi )\mathrm{\Theta }(2\xi )\xi =\frac{s}{R}$$
(28)
and using this in equation (22), the function $`f(t_1)`$ is evaluated for $`0<t_1<T`$ in closed form as
$$f(t_1)=\frac{1}{R^3}\frac{1}{2R^3}(2\chi )(22\chi \chi ^2)\mathrm{\Theta }(2\chi )\chi =\frac{Tt_1}{R}\mathrm{\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}0}<t_1<T$$
(29)
which completes the evaluation of the time-averaged self-force (15) as a one-dimensional quadrature involving the test body’s displacement $`D_x(t_1)`$. The BR geometric factor (2) for coinciding spherical space-time regions can now be obtained in closed form as
$$\overline{A}_{xx}^{(\mathrm{I},\mathrm{I})}=\frac{1}{T^2}_0^Tdt_1f(t_1)=\frac{1}{R^4\kappa }\frac{1}{8R^4\kappa }(4+\kappa )(2\kappa )^2\mathrm{\Theta }(2\kappa )\kappa =\frac{T}{R}$$
(30)
in agreement with equation (100) of . The terms $`1/R^3`$ and $`1/R^4\kappa `$ in expressions (29) and (30), respectively, are due to the electrostatic force of attraction to the neutralizing body—when the latter is absent, or one is interested only in the proper self-force itself, these terms must be subtracted from the above expressions. According to equation (30), the BR geometric factor $`\overline{A}_{xx}^{(\mathrm{I},\mathrm{I})}`$ equals the electrostatic term $`1/R^3T`$ only when the duration of the displacement $`T2R`$; the result of CP that $`\overline{A}_{xx}^{(\mathrm{I},\mathrm{I})}=1/R^3T`$ for all values of $`T`$ (see \[5, equation (11)\]) was obtained by an incorrect use of a Taylor expansion of the derivative of the delta function in an integration with finite limits.
We assume that the trajectory $`D_x(t_1)`$ can be expanded about the point $`t_1=0`$ as a Taylor series, valid for $`0<t_1<T`$:
$$D_x(t_1)=\underset{n=0}{\overset{\mathrm{}}{}}\frac{D_x^{(n)}(0^+)}{n!}t_1^n.$$
(31)
This enables us to evaluate analytically the time-averaged self-force (15) in terms of the time derivatives $`D_x^{(n)}(0^+)lim_{t_10^+}\mathrm{d}^nD_x(t_1)/\mathrm{d}t_1^n`$ using the closed-form expression (29) for the function $`f(t_1)`$:
$`\overline{F}_x`$ $`=`$ $`{\displaystyle \frac{\rho _c^2V^2}{T}}{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{D_x^{(n)}(0^+)}{n!}}{\displaystyle _0^T}dt_1t_1^nf(t_1)`$ (32)
$`=`$ $`{\displaystyle \frac{3\rho _c^2V^2}{TR^2}}{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{R^nD_x^{(n)}(0^+)}{(n+4)!}}\{[\kappa ^2+2(n+2)\kappa +n(n+3)](\kappa 2)^{n+2}\mathrm{\Theta }(\kappa 2)`$ (34)
$`\kappa ^{n+4}+(n+3)(n+4)\kappa ^{n+2}(n+2)(n+3)(n+4)\kappa ^{n+1}\}\kappa ={\displaystyle \frac{T}{R}}.`$
The electrostatic term $`1/R^3`$ in (29) contributes here one-third of the $`\kappa ^{n+1}`$ term, and so the time-averaged proper self-force itself (or the time-averaged “radiation-reaction” component of the “self-force” ) is obtained by replacing $`\kappa ^{n+1}`$ in (34) with $`\frac{2}{3}\kappa ^{n+1}`$.
Figure 1 exhibits the dependence on the displacement duration $`T`$ of the time-averaged self-force $`\overline{F}_x`$ for a trajectory $`D_x(t_1)=D_x[1\mathrm{cos}(2\pi t_1/T)]\mathrm{\Theta }(Tt_1)\mathrm{\Theta }(t_1)`$, as calculated according to equation (34), together with that of the limiting time-averaged self-force $`\overline{F}_{\mathrm{BR}x}`$, as given by equations (1) and (30).
3. The time dependence of the self-force
Using the results obtained in the course of calculating the time average of the self-force, we can also evaluate analytically the time dependence $`F_x(t_2)`$ of the self-force in terms of the derivatives $`D_x^{(n)}(0^+)`$ of the body’s trajectory $`D_x(t_1)`$. The self-force (13) can be written as
$$F_x(t_2)=\rho _c^2V^2_0^Tdt_1D_x(t_1)g(t_2t_1)$$
(35)
where the function $`g(t)`$ is defined by
$$g(t)=\frac{1}{V^2}_{|𝒓_1|<R}d𝒓_1_{|𝒓_2|<R}d𝒓_2\left(\frac{^2}{x_1x_2}\frac{^2}{t_1t_2}\right)\frac{\delta (tr)}{r}$$
(36)
which differs from the definition (16) of the function $`f(t_1)`$ only by the absence of the integration with respect to time $`t_2`$. Thus, using equation (22), the function $`g(t)`$ can be expressed in terms of the derivative of the integral $`I(s)`$, and using the expression (28) for $`I(s)`$, we get
$$g(t)=\frac{1}{R^3}\delta (t)\frac{2}{3}\frac{\mathrm{d}I(t)}{\mathrm{d}t}=\frac{3}{R^3}\delta (t)+\frac{3}{2R^4}(2\xi ^2)\mathrm{\Theta }(\xi )\mathrm{\Theta }(2\xi )\xi =\frac{t}{R}.$$
(37)
Only one-third of the $`\delta (t)`$ term on the right-hand side arises from the electrostatic term as the derivative of the step function $`\mathrm{\Theta }(\xi )`$ in $`I(s)`$ also contributest. Using in equation (35) the closed-form expression (37) for $`g(t)`$ and the Taylor expansion (31) for $`D_x(t_1)`$ in the non-delta-function term, we obtain the following analytical expression for the time dependence $`F_x(t_2)`$ of the self-force:
$`F_x(t_2)`$ $`=`$ $`{\displaystyle \frac{3\rho _c^2V^2}{R^3}}D_x(t_2)+{\displaystyle \frac{3\rho ^2V^2}{R^3}}\mathrm{\Theta }(t_2){\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{R^nD_x^{(n)}(0^+)}{(n+3)!}}`$ (39)
$`\times \left[a_n(\kappa )\mathrm{\Theta }(\kappa 2)\mathrm{\Theta }(2\xi )+b_n(\kappa )\kappa ^{n+1}\mathrm{\Theta }(\xi )+c_n(\kappa )\mathrm{\Theta }(\xi )\mathrm{\Theta }(2\xi )\right]`$
$`a_n(\kappa )`$ $`=`$ $`[\kappa ^2+2(n+1)\kappa +n^2+n2](\kappa 2)^{n+1}b_n(\kappa )=\kappa ^2+n^2+5n+6`$ (40)
$`c_n(\kappa )`$ $`=`$ $`{\displaystyle \frac{T^{n+1}}{R^{n+1}}}[b_n(\kappa )(n+1)\xi \kappa {\displaystyle \frac{1}{2}}(n^2+3n+2)\xi ^2]\kappa ={\displaystyle \frac{t_2}{R}}\xi =\kappa {\displaystyle \frac{T}{R}}.`$ (41)
We note that, interestingly, the electrostatic force $`\rho _c^2V^2D_x(t_2)/R^3`$ of attraction to the neutralizing body contributes here only one third of the term that is directly proportional to the instantaneous distance $`|D_x(t_2)|R`$ from the neutralizing body. The averaging of expression (41) according to equation (14) confirms equation (34) for the time-averaged self-force $`\overline{F}_x`$; as expected, the self-force $`F_x(t_2)`$ vanishes when the variable $`\xi 2`$ (i.e., when $`t_2T+2R`$). The limiting BR self-force $`F_{\mathrm{BR}x}(t_2)`$ is obtained with a steplike trajectory $`D_x(t_2)=D_x\mathrm{\Theta }(Tt_2)\mathrm{\Theta }(t_2)`$, for which only the $`n=0`$ term \[with $`D_x^{(0)}(0^+)=D_x`$\] in the series in equation (41) is nonzero.
Figures 2 and 3 show the time dependence of the self-force $`F_x(t_2)`$, calculated using equation (41) for the trajectory $`D_x(t_2)=D_x[1\mathrm{cos}(2\pi t_2/T)]\mathrm{\Theta }(Tt_2)\mathrm{\Theta }(t_2)`$ and for the limiting steplike trajectory $`D_x\mathrm{\Theta }(Tt_2)\mathrm{\Theta }(t_2)`$.
An alternative expression for the self-force $`F_x(t_2)`$ in terms of the derivatives $`D_x^{(n)}(t_2)`$ at a current time $`t_2`$ along the trajectory should be instructive since the radiation-reaction force is usually expressed in terms of such derivatives. This can be done easily by suitably changing the integration variable in equation (35) before expanding the trajectory in a Taylor series:
$$F_x(t_2)=\rho _c^2V^2\underset{n=0}{\overset{\mathrm{}}{}}\frac{(1)^n}{n!}D_x^{(n)}(t_2)_{t_2T}^{t_2}dtt^ng(t).$$
(42)
However, this is valid only for $`t_2<T`$, as the function $`D_x(t_2)0`$ for $`t_2>T`$ and as such cannot be expanded about a point $`t_2>T`$ for use in the interval $`(0,T)`$. The integration in equation (42) with the closed-form expression (37) for the function $`g(t)`$ leads to the following result:
$`F_x(t_2)`$ $`=`$ $`{\displaystyle \frac{3\rho _c^2V^2}{R^3}}D_x(t_2){\displaystyle \frac{3\rho _c^2V^2}{2R^3}}\mathrm{\Theta }(t_2){\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{(1)^nR^nD_x^{(n)}(t_2)}{(n+3)(n+1)!}}`$ (44)
$`\times \{[2^{n+2}(n1)d_n(\kappa )]\mathrm{\Theta }(\kappa 2)+d_n(\kappa )\}`$
$`d_n(\kappa )`$ $`=`$ $`[(n+1)\kappa ^22n6]\kappa ^{n+1}\kappa ={\displaystyle \frac{t_2}{R}}<{\displaystyle \frac{T}{R}}.`$ (45)
For times $`t_2>2R`$ (i.e., for $`\kappa >2`$), equation (45) gives
$$F_x(t_2)=\frac{\rho _c^2V^2}{R^3}D_x(t_2)\frac{24\rho _c^2V^2}{R^3}\underset{n=0}{\overset{\mathrm{}}{}}\frac{(2)^nR^{n+2}D_x^{(n+2)}(t_2)}{(n+5)(n+3)(n+2)n!}\mathrm{\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}\hspace{0.33em}2}R<t_2<T.$$
(46)
Here, the original $`n=0`$ term reduced the first term of equation (45) to the electrostatic force of attraction to the neutralizing body, the original $`n=1`$ term vanished, and the summation of the series was relabeled so that it now begins with the $`D_x^{(2)}(t_2)`$ term. The series in equation (46) agrees with the expression given by Jackson for the electromagnetic self-force Note that Jackson’s self-force is defined so that its sign is opposite to ours. on a body carrying a spherically symmetric charge distribution. This can be seen on noting that, in the case of a uniform spherically symmetric charge density, the integral appearing in that expression has the following value:
$$_{|𝒓_1|<R}d𝒓_1_{|𝒓_2|<R}d𝒓_2r^{n1}=\frac{9V^22^{n+2}R^{n1}}{(n+5)(n+3)(n+2)}.$$
(47)
This integral was evaluated by reducing it to a three-dimensional quadrature in the same way as that of the reduction of integral (21) to integral (23) and performing the resulting three-dimensional integral analytically.
In conclusion, we remark that the fact that the time-averaged self-force (15) is proportional to the displacement $`D_x`$ even in the absence of the neutralizing body—for a displacement duration $`T<2R`$ and in the limit of a steplike trajectory—does not contradict the translational invariance of the Lagrangian of the system consisting of the displaced body and the electromagnetic field. Such invariance is irrelevant to the case under the consideration because the body is assumed to be displaced by an external force whose origin is outside this system.
|
warning/0005/astro-ph0005149.html
|
ar5iv
|
text
|
# Radio Properties of NLS1s
## 1 Introduction
It is well established that the optical and X-ray characteristics of narrow-line Seyfert 1 galaxies distinguish them from all other types of active galactic nuclei (AGNs). Unfortunately, the radio properties of NLS1s have been less well explored. In the only study dedicated to the subject, Ulvestad, Antonucci, & Goodrich (1995; hereafter UAG) found that NLS1s do not differ noticeably from nearby classical type 1 and type 2 Seyfert galaxies at centimeter wavelengths, in contrast to the results obtained in the optical and X-ray bands. This conclusion was based on the modest radio powers ($`10^{20}`$$`10^{23}`$ W Hz<sup>-1</sup>) and small radio source sizes ($`<300`$ pc) of the objects they examined. But as UAG candidly noted, their sample of NLS1s was not defined using a uniform set of criteria, and only a fraction of the galaxies in it (9/15) was detected. We have investigated the radio emission of a larger, uniformly selected sample of NLS1s in order to gain further insight into the radio nature of these objects and their relation to other classes of AGNs.
Our sample of 24 NLS1s is drawn from the catalog of IRAS sources detected in the ROSAT All-Sky Survey (Boller et al. 1992; Moran et al. 1996). Full details regarding the sample definition are provided in Moran et al. (2000). We have obtained simultaneous high-resolution A-array VLA observations at 20 cm and 3.6 cm of most of the IRAS- and ROSAT-Observed NLS1 (“IRON”) galaxies. In addition, nearly all of the objects have been imaged at 20 cm in the moderate-resolution B and C arrays by Condon et al. (1998a) and in the low-resolution D array as part of the NRAO VLA Sky Survey (Condon et al. 1998b). All but one of the IRON galaxies are detected at 20 cm; 22 have three or more flux density measurements at that wavelength.
## 2 Population Statistics
Radio Power Distribution. As Figure 1$`a`$ illustrates, the majority of the IRON objects have radio powers in excess of $`10^{23}`$ W Hz<sup>-1</sup>, and seven are more luminous than $`10^{24}`$ W Hz<sup>-1</sup>—in stark contrast to the 20 cm radio power distribution for nearby classical Seyfert galaxies (Ulvestad & Wilson 1989). Thus, it would appear that NLS1s are frequently more luminous than nearby Seyferts in the radio band. We have also determined the 20 cm luminosity distribution for 77 classical Seyfert galaxies in the IRAS-ROSAT catalog from which the IRON sample was drawn, based on VLA observations by Condon et al. (1998a). As indicated in Figure 1$`b`$, the radio luminosities of the IR/X-ray–selected Seyferts tend to be higher than those of the nearest classical Seyfert galaxies, but they do not extend to the very high luminosities displayed by NLS1s selected the same way. Interestingly, the radio–to–infrared and radio–to–X-ray flux ratio distributions of the IRON galaxies and the IRAS-ROSAT Seyferts do not differ significantly, suggesting that the IRON galaxies have higher 20 cm radio powers because they are more luminous sources at several wavelengths, not because their radio emission is enhanced in some way.
Radio Source Sizes. Most of the IRON galaxies are unresolved at $`1^{\prime \prime }`$ resolution, confirming the conclusions of UAG that the nuclear radio sources in NLS1s are compact. However, in our 3.6 cm observations (0$`.^{\prime \prime }`$25 resolution), two sources (IRAS 06269$``$0543 and Ark 564) exhibit an unresolved core and what appears to be a small-scale ($`1^{\prime \prime }`$) jet (Fig. 2).
Spectral Index Distribution. In Figure 3 we have plotted the radio spectral index distribution for the IRON galaxies; also shown is the distribution of spectral slopes for 59 of the classical Seyfert galaxies in the distance-limited sample of Morganti et al. (1999). Clearly, the IRON galaxies tend to have significantly steeper radio spectra than the classical Seyferts. Only one of the IRON objects has a spectrum flatter than $`\alpha =0.4`$, and the bulk of the objects have $`\alpha 1.11.2`$, well out on the tail of the Morganti et al. Seyfert distribution. One remarkable source, IRAS 06269$``$0543, has a spectral index of $`\alpha =2.21`$, which is steeper by far than the spectrum of any core-dominated Seyfert galaxy or radio-quiet quasar we are aware of.
Variability. It is difficult to evaluate the radio variability of the IRON sample because of resolution effects associated with the different VLA configurations used for the 20 cm observations. However, a few galaxies exhibit flux density differences that are not instrumental in nature, including IRAS 06269$``$0543 (38% variability) and IRAS 20181$``$2244 (18% variability).
## 3 Implications for the Physical Nature of NLS1s
The nuclear radio sources in the IRON galaxies tend to be compact, steep-spectrum, and, in a few cases, variable—three characteristics that are rarely found together. This unusual combination of properties can be accounted for if most of the radio flux arises from a tiny ($``$ 1 pc diameter) region near the central engine of the active nucleus. In this scenario, the variability is not intrinsic to the source, but is caused by “scintillation” as the emission passes through the interstellar medium of the Milky Way (e.g., Rickett 1990). Due to the proximity of the radio-emitting plasma to the intense optical/UV continuum source, the cooling of the electrons is dominated by inverse-Compton scattering rather than synchrotron emission, which steepens the radio spectrum. In the case of IRAS 06269$``$0543, our calculations indicate that the electron cooling time would be very short in this picture, suggesting that relativistic electrons are being continuously resupplied. This might occur if the mass accretion rate in this object is very high relative to the Eddington limit, which has been suggested to explain the steep soft X-ray spectra and rapid, large-amplitude X-ray variability observed in some NLS1s (Pounds et al. 1995; Boller et al. 1996). A thorough description of this hypothesis, which can be tested with additional radio observations, is provided in Moran et al. (2000).
|
warning/0005/quant-ph0005012.html
|
ar5iv
|
text
|
# Manipulating motional states by selective vibronic interaction in two trapped ions
\[
## Abstract
We present a selective vibronic interaction for manipulating motional states in two trapped ions, acting resonantly on a previously chosen vibronic subspace and dispersively on all others. This is done respecting technical limitations on ionic laser individual addressing. We discuss the generation of Fock states and entanglement in the ionic collective motional degrees of freedom, among other applications.
\]
Coherent manipulation of vibronic states of two or more trapped ions has become a subject of increasing interest in the last few years. A number of experimental and theoretical advances has been done recently in this field, aiming to make applications possible and to test basic features of quantum mechanics. In this context, Bell states and multiparticle entanglement of the ionic internal degrees of freedom have been studied and partially realized . Beside their fundamental importance, such states are essential ingredients for the implementation of several interesting applications, as quantum logic and quantum computing using trapped ions. One of the main problems in realizing some of the above proposals is the technical difficulty in the individual ionic laser illumination. Whereas many efforts are being done to overcome this problem , it still represents an obstacle to the existence of realistic proposals for the deterministic manipulation of the motional degrees of freedom of two or more trapped ions.
Recently, Mølmer and Sørensen have shown that if one illuminates two two-level ions with two lasers that act dispersively in the blue and red first side band transitions, it is possible to find an effective interaction, in the Lamb-Dicke limit, that causes a direct transition between the ground and the doubly excited electronic levels of the two ion system, without changing their vibrational state. This permits the deterministic generation of entangled states that are linear combinations of these two electronic states. In ref., the combination of this interaction with a carrier transition permits the generation of all electronic Bell states. This leads to more general applications, as teleportation and entanglement swapping. All these proposals are mostly concerned with the deterministic manipulation of internal states of the ions. However, it is also important to find ways for manipulating the vibrational degrees of freedom of two or more trapped ions, respecting the limitations imposed on their individual addressing.
In this article we present an effective interaction consisting of two dispersive Raman pulses simultaneously illuminating two trapped ions, which opens the possibility to manipulate coherently their vibrational motion. As the electronic Stark shifts, induced by the Raman pulses, depend on the motional state of the ions, the resulting dynamics can be described by a Jaynes-Cummings-like interaction acting distinctly on different subspaces of their vibronic Hilbert space. As we will show below, the frequencies of the two dispersive Raman pulses may be chosen in such a way that the effective interaction becomes resonant to a previously chosen vibronic subspace while remaining non-resonant to others. This enables us to excite selectively a desired subspace inside the motional Hilbert space of the ions. Vibrational state engineering of both center of mass and relative motion can then be done by means of this special property.
We consider two two-level ions of mass $`m`$, confined to move in the $`z`$ direction in a Paul trap. They are cooled down to very low temperatures and may perform small oscillations around their equilibrium positions, $`z_{10}=d/2,`$ $`z_{20}=d/2.`$ We denote by $`\widehat{Z}=(\widehat{z}_1+\widehat{z}_2)/2`$ and $`\widehat{z}=(\widehat{z}_1\widehat{z}_2)/2`$ the center of mass and relative position operators, respectively. Both ions are simultaneously illuminated by two classical homogeneous Raman effective pulses $`\stackrel{}{E}_I=\stackrel{}{E}_{0I}e^{i(\stackrel{}{q}_1\stackrel{}{r}\omega _It)}`$ and $`\stackrel{}{E}_{II}=\stackrel{}{E}_{0II}e^{i(\stackrel{}{q}_2\stackrel{}{r}\omega _{II}t)},`$ with wave vectors $`\stackrel{}{q}_1=\stackrel{}{q}_2=\stackrel{}{q},`$ parallel to the $`z`$ direction and angular frequencies $`\omega _I`$ and $`\omega _{II}`$. The Raman pulses frequencies will be chosen to be quasi-resonant with a long-living electronic transition between two ionic hyperfine levels $`|_j`$ and $`|_j`$ (j=1,2), with energy $`\mathrm{}\omega _0`$ and $`0`$, respectively. The total Hamiltonian of the system may be written, in the optical RWA approximation, as
$$\widehat{H}=\widehat{H}_0+\widehat{H}_{\mathrm{int}},$$
(1)
with
$`\widehat{H}_0`$ $`=`$ $`\mathrm{}\omega _o(\widehat{S}_{+1}\widehat{S}_1+\widehat{S}_{+2}\widehat{S}_2)+\mathrm{}\nu \widehat{a}^{}\widehat{a}+\mathrm{}\nu _r\widehat{b}^{}\widehat{b},`$ (2)
$`\widehat{H}_{\mathrm{int}}`$ $`=`$ $`\mathrm{}\mathrm{\Omega }e^{iq\widehat{Z}}(\widehat{S}_{+1}e^{iq\widehat{z}/2}+\widehat{S}_{+2}e^{iq\widehat{z}/2})\times `$ (4)
$`\left(e^{i\omega _It}+e^{i\omega _{II}t}\right)+\mathrm{H}.\mathrm{c}.`$
Here $`\widehat{S}_{+j}=|_j_j|`$ is the flip operator associated with the electronic transition $`|_j|_j`$ in the ion $`j.`$ The operators $`\widehat{a}`$ and $`\widehat{b}`$ ($`\widehat{a}^{}`$ and $`\widehat{b}^{}`$) are the annihilation (creation) operators associated with the center of mass mode of frequency $`\nu `$ and the relative vibrational mode of frequency $`\nu _r,`$ respectively. For simplicity, we have assumed that the same Rabi frequency $`\mathrm{\Omega }`$ (taken as real) is associated with both lasers.
We start by taking the frequencies $`\omega _I`$ and $`\omega _{II}`$ as:
$`\omega _I`$ $`=`$ $`\omega _0+k\nu +k_r\nu _r\delta `$ (5)
$`\omega _{II}`$ $`=`$ $`\omega _0+\delta ,`$ (6)
with $`\delta \nu ,\nu _r.`$ The frequencies are chosen so that $`\omega _I+\omega _{II}=k\nu +k_r\nu _r+2\omega _0.`$
Following the usual treatment for one single ion interacting with a laser field , we make the rotating-wave-approximation (RWA) with respect to the vibrational frequencies and select the terms that oscillate with minimum frequency. In the interaction picture we obtain, for the interaction Hamiltonian, the following expression:
$$\widehat{H}_{\mathrm{int}}^I=\mathrm{}\mathrm{\Omega }\left(\widehat{S}_+^{}\widehat{a}^k\widehat{b}^{k_r}\widehat{H}_{k,k_r}e^{i\delta t}+\widehat{S}_+^{\prime \prime }\widehat{H}_{0,0}e^{i\delta t}\right)+\mathrm{H}.\mathrm{c}.,$$
(7)
where
$`\widehat{S}_+^{}`$ $`=`$ $`\widehat{S}_{+1}e^{i\varphi _0/2}+()^{k_r}\widehat{S}_{+2}e^{i\varphi _0/2},`$ (8)
$`\widehat{S}_+^{\prime \prime }`$ $`=`$ $`\widehat{S}_{+1}e^{i\varphi _0/2}+\widehat{S}_{+2}e^{i\varphi _0/2}.`$ (9)
Here $`\varphi _0=z_0d`$ is the phase difference due to the equilibrium distance between the two ions, which, in this paper, may be taken as $`2n\pi `$ without loss of generality , and
$$\widehat{H}_{k,k_r}=(i\eta )^k(i\eta _r)^{k_r}\widehat{F}_k(\eta ^2)\widehat{G}_{k_r}(\eta _r^2),$$
(10)
with
$`\widehat{F}_k(\eta ^2)`$ $`=`$ $`{\displaystyle \underset{n}{}}e^{\eta ^2/2}{\displaystyle \frac{n!}{(n+k)!}}L_n^k(\eta ^2)|nn|,`$ (11)
$`\widehat{G}_{k_r}(\eta _r^2)`$ $`=`$ $`{\displaystyle \underset{n}{}}e^{\eta _r^2/2}{\displaystyle \frac{n_r!}{(n_r+k_r)!}}L_{n_r}^{k_r}(\eta _r^2)|n_rn_r|.`$ (12)
The parameters $`\eta =q\sqrt{\mathrm{}/4m\nu }`$ and $`\eta _r=q\sqrt{\mathrm{}/4m\nu _r}`$ are the Lamb-Dicke parameters corresponding to the CM and the relative modes, respectively. $`L_n^m`$ are associated Laguerre polynomials, whereas the states $`|n`$ and $`|n_r`$ are the eigenvectors of the number operators $`\widehat{a}^{}\widehat{a}`$ and $`\widehat{b}^{}\widehat{b}`$, respectively.
If $`\delta `$ is large enough so that $`\delta \mathrm{\Omega }`$ and for times $`t`$ such that $`\delta t1`$, it is possible to derive a two-photon effective time-independent Hamiltonian, where both CM and relative modes can be excited. It can be written as a sum of three terms:
$$\widehat{H}_{\mathrm{eff}}=\widehat{H}_1+\widehat{H}_2+\widehat{H}_3,$$
(13)
with
$`\widehat{H}_1=\mathrm{}\mathrm{\Omega }_0ϵ_{k_r}\widehat{S}_{+1}\widehat{S}_{+2}[\widehat{a}^k\widehat{b}^{k_r}\widehat{H}_{k,k_r},\widehat{H}_{0,0}]+\mathrm{H}.\mathrm{c}.`$ (14)
$`\widehat{H}_2=\mathrm{}\mathrm{\Omega }_0()^{k_r}\widehat{S}_{+1}\widehat{S}_2[\widehat{a}^k\widehat{b}^{k_r}\widehat{H}_{k,k_r},\widehat{H}_{k,k_r}\widehat{a}^k\widehat{b}^{k_r}]+\mathrm{H}.\mathrm{c}.`$ (15)
$`\widehat{H}_3={\displaystyle \frac{\mathrm{}\mathrm{\Omega }_0}{2}}{\displaystyle \underset{j=1,2}{}}\widehat{S}_{+j}\widehat{S}_j(\widehat{a}^k\widehat{b}^{k_r}\widehat{H}_{k,k_r}^2\widehat{a}^k\widehat{b}^{k_r}\widehat{H}_{0,0}^2)`$ (16)
$`{\displaystyle \underset{j=1,2}{}}\widehat{S}_j\widehat{S}_{+j}(\widehat{a}^k\widehat{b}^{k_r}\widehat{a}^k\widehat{b}^{k_r}\widehat{H}_{k,k_r}^2\widehat{H}_{0,0}^2)+\mathrm{H}.\mathrm{c}.,`$ (17)
where $`\mathrm{\Omega }_0=\mathrm{\Omega }^2/\delta .`$ The first term, $`\widehat{H}_1,`$ gives rise to an anti Jaynes-Cummings dynamics, leading to a simultaneous excitation (or de-excitation) of the electronic states of the two ions, accompanied by the creation (or annihilation) of $`k`$ vibrational quanta in the CM mode and $`k_r`$ vibrational quanta in the relative mode. The factor $`ϵ_{k_r}=(1+()^{k_r})`$ prevents excitations of odd number of quanta in the relative mode, so that the symmetry by exchange of the two ions is maintained. The second term, $`\widehat{H}_2,`$ generates a dynamics where, simultaneously, one ion undergoes a transition from the ground to the excited electronic state and the other ion makes a transition in the inverse direction. This process is not accompanied by any excitation of the vibrational modes. The third term, $`\widehat{H}_3,`$ generates motional dependent dynamical energy shifts in the electronic levels. Due to this dependence, this term turns the processes induced by $`\widehat{H}_1`$ and $`\widehat{H}_2`$ more or less resonant, depending on the particular level of excitation of the vibrational modes. Note that the sensitivity of the energy shifts to the vibrational state of the ions increases with increasing values of the Lamb–Dicke parameters. For not too small values of the these parameters, it is possible to make the interaction $`\widehat{H}_1`$ completely resonant inside a previously chosen subspace $`\{|,,N,N_r,|,,N+k,N_r+k_r\},`$ whereas remaining largely non resonant inside other subspaces. The same applies for $`\widehat{H}_2`$ in a given subspace $`\{|,,M,M_r,|,,M,M_r\}`$. This may occur, for example, when the Raman lasers frequencies, originally given by Eq. 5, are modified and correctly tuned to take into account the motional dependent energy shifts.
To describe these effects in more detail, we turn our attention to specific cases. We first consider excitations to the first blue side band of the center of mass mode $`(k=1,k_r=0).`$ If we start from the electronic ground state, only $`\widehat{H}_1`$ and $`\widehat{H}_3`$ will be effective , and we may write:
$`H_{\mathrm{eff}}=\mathrm{}\mathrm{\Omega }_0\widehat{G}_0^2\{2(i\eta )\widehat{S}_{+1}\widehat{S}_{+2}[\widehat{a}^{}\widehat{F}_1,\widehat{F}_0]+`$ (18)
$`{\displaystyle \frac{1}{2}}(\widehat{S}_{+1}\widehat{S}_1+\widehat{S}_{+2}\widehat{S}_2)(\eta ^2\widehat{a}^{}\widehat{F}_1^2\widehat{a}\widehat{F}_0^2)`$ (19)
$`{\displaystyle \frac{1}{2}}(\widehat{S}_1\widehat{S}_{+1}+\widehat{S}_2\widehat{S}_{+2})(\eta ^2\widehat{a}\widehat{a}^{}\widehat{F}_1^2\widehat{F}_0^2)+\mathrm{H}.\mathrm{c}.\}.`$ (20)
As can be easily seen from Eq.18, the energy shifts of levels $`|,,n,n_r`$ and $`|,,n+1,n_r`$ are given by:
$`\mathrm{\Delta }_{}^{n,n_r}=2\mathrm{}\mathrm{\Omega }_0g_0^2(n_r)\left[f_0^2(n)\eta ^2(n+1)f_1^2(n)\right]`$ (21)
$`\mathrm{\Delta }_{}^{n+1,n_r}=2\mathrm{}\mathrm{\Omega }_0g_0^2(n_r)\left[\eta ^2(n+1)f_1^2(n)f_0^2(n+1)\right],`$ (22)
respectively. In Eq. (21)
$`f_k(m)`$ $`=`$ $`e^{\eta ^2/2}{\displaystyle \frac{m!}{(m+k)!}}L_m^k(\eta ^2)`$ (23)
$`g_k(m)`$ $`=`$ $`e^{\eta _r^2/2}{\displaystyle \frac{m!}{(m+k)!}}L_m^k(\eta _r^2).`$ (24)
By properly adjusting the laser frequencies we may put them in resonance with the Stark shifted levels associated with a previously chosen vibronic subspace $`\{|,,N,N_r`$ $`,|,,N+1,N_r\},`$ while preventing resonant transitions in other subspaces with $`nN`$ and $`n_rN_r.`$ This can be done by modifying the laser frequencies to
$$\omega _I=\stackrel{~}{\omega }_0+\nu \delta ,\omega _{II}=\stackrel{~}{\omega }_0+\delta ,$$
(25)
where $`2\stackrel{~}{\omega }_0=2\omega _0+\mathrm{\Delta }_{}^{N+1,N_r}\mathrm{\Delta }_{}^{N,N_r}`$ is the renormalized splitting of the levels. Notice that, for very small values of the Lamb-Dicke parameters, the motional dependence of the dynamical Stark shift disappears. For this reason it is important to work beyond the Lamb-Dicke regime in order to effectively select a chosen subspace out of the whole vibronic Hilbert space. It is noteworthy to mention that for special values of the Lamb-Dicke parameter $`\eta ,`$ it may happen that energy shifts $`\mathrm{\Delta }_{}^{n+1,n_r}`$ and $`\mathrm{\Delta }_{}^{n,n_r}`$ are equal for certain values of $`n,`$ irrespectively of the state of the relative vibrational mode. For example, for $`\eta 0.51,0.42,0.24,`$ transitions inside the subspaces $`\{|,,N,N_r,|,,N+1,N_r\},`$ become resonant for $`N=1,2,8,`$ respectively. Clearly, in this case it is not necessary to correct the laser frequencies.
In order to check this model, numerical simulations were done using the time dependent Hamiltonian given in Eq. (7). Starting from the state $`|,,0,0`$ and selecting the laser frequencies as in Eq. (25), we were able to observe complete Rabi oscillations between the states $`|,,0,0`$ and $`|,,1,0,`$ in agreement with the model discussed above (see Fig. 1). In particular, for a $`\pi `$ pulse, the state $`|,,1,0`$ is generated with $`100\%`$ efficiency. For the same laser frequencies, we plot the Rabi oscillations of the population corresponding to the initial state $`|,,1,0.`$ We can see that, indeed, the Rabi oscillations for this case have a very small amplitude.
Similar results may be obtained also for transitions leading to even excitations of the relative vibrational mode only. This is done by choosing $`k=0`$ in Eq. 13 and correctly tuning the lasers to take into account the self energy terms. We checked numerically that it is possible, in this case, to drive resonantly Rabi oscillations inside selected subspaces $`\{|,,N,N_r,|,,N,N_r+k_r\}.`$
The Hamiltonian (13) could be used to generate a large set of motional states. For example, any Fock state associated with the center of mass motion and with even excitations of the relative vibrational mode could be obtained from the initial state $`|,,0,0`$ by successively applying $`\pi `$ pulses with different frequencies. However, if one is interested in generating a highly excited Fock state, this process could take an unsatisfactory long time. A more efficient, non unitary, way of producing such states, as well as engineering other vibrational states, is to start from a product of the electronic ground state and any motional state $`|\psi _{\mathrm{vib}}.`$ By selecting the laser frequencies, we excite only a chosen vibronic transition $`\{k,k_r\}`$ with a $`\pi `$ pulse. Ideally, we would end up with a superposition of the two states $`|,,N+k,N_r+k_r`$ and $`|,(|\psi _{\mathrm{vib}}|N,N_rN,N_r|\psi _{\mathrm{vib}}).`$ Measurement of the electronic levels projects out either the Fock state $`|N+k,N_r+k_r`$ or the original state with a “hole” in the $`N,N_r`$ component.
As a numerical example we have taken the initial state to be the product of the electronic ground state with a coherent state of the CM motion and the vacuum of the relative motion. We start from a coherent state with $`\overline{n}=4.0`$ and excite the center of mass transition from $`n=4`$ to $`n=5`$. A dark event in a fluorescent measurement should leave us with an state that is close to the Fock state with $`n=5.`$ In Fig. 2 we show the results for the phonon distribution. As expected, it is possible to make a “hole” in the vibrational quanta distribution ( Fig. 2a ) while creating a quasi Fock state ( Fig. 2b) using a $`\pi `$ pulse. The state with a“hole” will be associated with the ground electronic state, while the approximate Fock state of Fig. 2b will be associated with the excited electronic state. As shown in Fig. 2b, small contamination occurs around the target Fock state because transitions to levels other than $`n=5`$ are not totally suppressed. For the case studied, the probability of finding the approximated Fock state after fluorescence is about $`30\%`$. A Fock state with $`n=6,`$ for example, can now be obtained if we apply subsequently to the ions another $`\pi `$ pulse resonant to the transition $`|,,5,0`$ $``$$`|,,6,0.`$ If by measuring the electronic states we find $`|,`$ ( a priori probability of $`87\%),`$ we are left with a state very close to the Fock state $`|6,0.`$ In this case the fidelity is $`99\%`$ (See Fig. 2c). Depopulation of a region around a certain value of $`N,`$ can also be achieved by proper choice of the Lamb Dicke parameter and the detuning $`\delta `$, taking advantage of the quasi resonance character of the interaction for $`n^{}s`$ very close to $`N.`$
Another interesting application is the generation of maximally entangled states inside the subspace $`|1,0,|0,2,|1,2,|0,0`$ of the vibrational modes. Starting from the electronic and vibrational ground state, a $`\pi /2`$ pulse with a given frequency would generate the state $`\frac{1}{\sqrt{2}}\left[|,,0,0+|,,0,2\right]`$. A subsequent $`\pi `$ pulse with another frequency would then lead to the state $`\frac{1}{\sqrt{2}}|,\left[|1,0+|0,2\right]`$ which is a maximally entangled state. Similar procedures could also lead to any maximally entangled state.
Entanglement transfer may also be achieved between the internal and motional degrees of freedom. For example, assume that we start with the state $`\frac{1}{\sqrt{2}}\left[|,+|,\right]|0,0.`$ If we apply to the ions a $`\pi `$ pulse, connecting the states $`|,0,0|,,1,2`$, we obtain $`|,(|0,0+|1,2).`$ In this case the efficiency is very high since the state $`|,|0,0`$ does not couple to any state in an anti Jaynes-Cummings transition. Of course, the inverse process, where entanglement is transfered from the motional degrees of freedom to electronic one, is also possible.
In summary, we have engineered an interaction that, respecting limitations of ionic individual addressing, enhances our possibilities of manipulating and generating diverse vibronic states of two trapped ions. This interaction acts selectively in a previously chosen vibronic subspace, $`\{|,N,N_r,|,,N+k,N_r+k_r\},`$ permitting, in principle, a complete transfer of populations inside this subspace. The applications mentioned above are only a few relevant examples of what may be done by selectively addressing the initial vibronic states.
This work was partially supported by the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq), the Programa de Apoio a Núcleos de Excelência (PRONEX) and the Fundação de Amparo à Pesquisa do Estado do Rio de Janeiro (FAPERJ).
|
warning/0005/cond-mat0005138.html
|
ar5iv
|
text
|
# Electronic states in correlated three-coupled chains
Spin gap and charge gap have been studied extensively in coupled-chain systems. Theoretical study in terms of a renormalization group (RG) method shows the spin gap in two-coupled chains for small but relevant interchain hopping \[?\] where the electronic state is given by $`d_{x^2y^2}`$-like superconducting state. The spin gap has been also examined for three-coupled chains \[?, ?, ?\], whose state depends on boundary conditions in the transverse direction. In case of small interchain hopping at half-filling, there are both charge gap and spin gap for open boundary conditions (OBC) while charge gap vanishes for periodic boundary conditions (PBC). A system with infinite chains at half-filling exhibits charge gap but no spin gap \[?\].
Although these works treat the relevant interchain hopping, a theory also suggests the irrelevant one-particle interchain hopping for half-filling \[?\]. In order to understand optical experiments on quasi-one-dimensional organic conductors \[?, ?\] indicating a confinement - deconfinement transition, two-coupled chains at half-filling have been studied where a transition from deconfinement to confinement occurs for umklapp scattering being larger than a critical value \[?, ?\]. In the present paper, such a transition is examined for three-coupled chains. Although the results of two chains seem to explain the experiments, it is not yet known if such a transition can be expected for many chains.
The Hamiltonian for quarter-filled three-coupled chains with on-site repulsion ($`U`$) and interchain hopping ($`t_{}`$) is written as
$``$ $`=`$ $`{\displaystyle \underset{j,\sigma ,l}{}}\left[t+(1)^jt_\mathrm{d}\right]\left(c_{j\sigma l}^{}c_{j+1\sigma l}+\text{h.c.}\right)`$ (1)
$`t_{}{\displaystyle \underset{j,\sigma ,l}{}}\left(c_{j\sigma l}^{}c_{j\sigma l+1}+\text{h.c.}\right)+U{\displaystyle \underset{j,l}{}}n_{jl}n_{jl},`$
where $`n_{j\sigma l}=c_{j\sigma l}^{}c_{j\sigma l}`$ and $`c_{j\sigma l}`$ denotes the annihilation operator of the electron at the $`j`$-th site of the $`l`$-th chain ($`l=`$1, 2, 3) with spin $`\sigma `$($`=,`$), and $`c_{j\sigma 4}=c_{j\sigma 1}`$. In Eq. (1), $`t_\mathrm{d}`$ denotes dimerization along the chains and the case for PBC is studied.
Diagonalizing the $`t_\mathrm{d}`$-term, which leads effectively to the half-filled band \[?\], we consider an effective Hamiltonian $`^d`$ consisting of the lower band with fermion operators of $`d_{k\sigma l}`$. The terms for the interchain hopping can be diagonalized by introducing Fourier transform, $`a_{k\sigma \mu }(1/\sqrt{3})`$ $`_{l=1}^3\mathrm{exp}\left[ik_y(\mu )l\right]d_{k\sigma l}`$ with $`k_y(\mu )=(2\pi /3)\mu `$ $`(\mu =0,\pm 1)`$. The kinetic term is written as $`_K^d_{k,\sigma ,\mu }`$ $`\epsilon (k,k_y)`$ $`a_{k\sigma \mu }^{}`$ $`a_{k\sigma \mu }`$ with $`\epsilon (k,k_y)=2\sqrt{t^2\mathrm{cos}^2ka+t_\mathrm{d}^2\mathrm{sin}^2ka}2t_{}\mathrm{cos}k_y`$, which is rewritten, in terms of the linearized dispersion, as $`_K^d=_{k,p,\sigma ,\mu }`$ $`v_\mathrm{F}(pkk_{\mathrm{F}\mu })a_{kp\sigma \mu }^{}a_{kp\sigma \mu }`$ with $`p`$ being the index of the branch $`p=+`$ $`()`$ corresponding to right moving (left moving) electrons. Fermi momenta are given by $`k_{\mathrm{F0}}=k_\mathrm{F}+2t_{}/v_\mathrm{F}`$ and $`k_{\mathrm{F}\pm }=k_\mathrm{F}t_{}/v_\mathrm{F}`$ where $`v_\mathrm{F}=\sqrt{2}ta`$ $`[1(t_\mathrm{d}/t)^2]`$ $`/\sqrt{1+(t_\mathrm{d}/t)^2}`$ \[?\]. Following the conventional $`g`$-ology, coupling constants are given by $`g_1^{}=g_2^{}=g_4^{}=Ua`$, $`g_3=Ua(2t_\mathrm{d}/t)/[1+(t_\mathrm{d}/t)^2]`$ \[?\] and $`g_1^{}=g_2^{}=g_4^{}=0`$.
Applying the bosonization method, we introduce phase variables $`\theta _{\rho \mu }`$ and $`\theta _{\sigma \mu }`$ expressing fluctuations of the charge density and spin density for the $`\mu `$-band \[?\], where the conjugate phase is introduced by $`[\theta _{\nu \mu }(x),`$ $`\varphi _{\nu ^{}\mu ^{}}(x^{})]_{}`$ $`=i\pi \delta _{\nu ,\nu ^{}}`$ $`\delta _{\mu ,\mu ^{}}`$ $`\mathrm{sgn}(xx^{})`$ ($`\nu ,\nu ^{}=\rho ,\sigma `$). Thus we obtain the total Hamiltonian given by $`^d=_0+_I`$ where $`_0`$ expresses bilinear terms of density operators, and $`_I`$ denotes nonlinear terms. Here we define the new phase variables as $`Y_1(X_1+X_2+X_3)/\sqrt{3}`$, $`Y_2(X_1X_3)/\sqrt{2}`$ and $`Y_3(X_1+2X_2X_3)/\sqrt{6}`$ where $`(Y_1,Y_2,Y_3)=(\theta _\rho ,\theta _{\mathrm{C1}},\theta _{\mathrm{C2}})`$ and $`(X_1,X_2,X_3)=(\theta _{\rho +},\theta _{\rho 0},\theta _\rho )`$ and the same transformation is applied to $`\rho \sigma `$ and also to the conjugate variables $`\varphi _{\nu \mu }`$. Thus $`_0`$ is rewritten as $`_0=_\nu (v_\nu /4\pi )𝑑x[K_\nu ^1\left(\theta _\nu \right)^2+K_\nu \left(\varphi _\nu \right)^2]`$ with $`\nu =\rho ,\mathrm{C1},\mathrm{C2},\sigma ,\mathrm{S1},\mathrm{S2}`$, where $`v_{\rho (\sigma )}=v_\mathrm{F}[1+()U/\pi v_\mathrm{F}]^{1/2}`$, $`K_{\rho (\sigma )}=[1+()Ua/\pi v_\mathrm{F}]^{1/2}`$, $`v_{\mathrm{C1}}=v_{\mathrm{S1}}=v_{\mathrm{C2}}=v_{\mathrm{S2}}=v_\mathrm{F}`$ and $`K_{\mathrm{C1}}=K_{\mathrm{S1}}=K_{\mathrm{C2}}=K_{\mathrm{S2}}=1`$. The Hamiltonian, $`_I`$ is divided as $`_I=_1+_3+_2+_{}`$ where the respective term is written as ($`i=1,3,2,`$)
$`_i`$ $`=`$ $`{\displaystyle \underset{Z}{}}{\displaystyle \frac{v_\mathrm{F}}{\pi \alpha ^2}}G_{iZ}{\displaystyle 𝑑x\mathrm{cos}\mathrm{\Theta }_{iZ}h_{iZ}},`$ (2)
($`Z=AZ`$) and $`\alpha `$ is a cutoff of the order of lattice constant and $`h_{iZ}`$ denotes the product of the Majorana fermion operators introduced to retain the anticommutation relation of the field operators \[?\]. The phase variables of the backward scattering term with opposite spins, $`_1`$, are given by $`\mathrm{\Theta }_{1Z}=(2/\sqrt{3})\theta _\sigma +\stackrel{~}{\mathrm{\Theta }}_{1Z}`$ where $`\stackrel{~}{\mathrm{\Theta }}_{1A}=ϵ\sqrt{2}\theta _{\mathrm{S1}}\sqrt{2/3}\theta _{\mathrm{S2}}`$, $`\stackrel{~}{\mathrm{\Theta }}_{1B}=2\sqrt{2/3}\theta _{\mathrm{S2}}`$, $`\stackrel{~}{\mathrm{\Theta }}_{1C}=[(ϵ^{}\theta _{\mathrm{S1}}+\theta _{\mathrm{S2}}/\sqrt{3})+ϵ(\theta _{\mathrm{C1}}ϵ^{}\sqrt{3}(\theta _{\mathrm{C2}}+2\sqrt{6}t_{}x/v_\mathrm{F}))]/\sqrt{2}`$, $`\stackrel{~}{\mathrm{\Theta }}_{1D}=ϵ\sqrt{2}\theta _{\mathrm{C1}}\sqrt{2/3}\theta _{\mathrm{S2}}`$, $`\stackrel{~}{\mathrm{\Theta }}_{1E}=[(ϵ^{}\theta _{\mathrm{S1}}+\theta _{\mathrm{S2}}/\sqrt{3})+ϵ(\varphi _{\mathrm{S1}}ϵ^{}\sqrt{3}\varphi _{\mathrm{S2}})]/\sqrt{2}`$, $`\stackrel{~}{\mathrm{\Theta }}_{1F}=ϵ\sqrt{2}\varphi _{\mathrm{S1}}\sqrt{2/3}\theta _{\mathrm{S2}}`$, $`\stackrel{~}{\mathrm{\Theta }}_{1G}=(\theta _{\mathrm{S2}}+ϵ^{}\sqrt{3}\theta _{\mathrm{C1}}+ϵ3\varphi _{\mathrm{C2}}ϵϵ^{}\sqrt{3}\varphi _{\mathrm{S1}})/\sqrt{6}`$ and $`\stackrel{~}{\mathrm{\Theta }}_{1H}=[(ϵ^{\prime \prime }\theta _{\mathrm{S1}}\theta _{\mathrm{S2}}/\sqrt{3})+ϵ^{}(\theta _{\mathrm{C1}}+ϵ^{\prime \prime }\sqrt{3}(\theta _{\mathrm{C2}}+2\sqrt{6}t_{}x/v_\mathrm{F}))+ϵ(ϵ^{\prime \prime }3\varphi _{\mathrm{C1}}\sqrt{3}\varphi _{\mathrm{C2}})ϵϵ^{}(\varphi _{\mathrm{S1}}+ϵ^{\prime \prime }\sqrt{3}\varphi _{\mathrm{S2}})]/2\sqrt{2}`$. In Eq. (2), the sum is taken implicitly with respect to $`ϵ`$, $`ϵ^{}`$ and $`ϵ^{\prime \prime }`$($`=\pm `$) which lead to the distinction for $`h_{iZ}`$ but not for $`G_{iZ}`$. The umklapp scattering terms, $`_3`$, are obtained from $`_1`$ by replacing $`G_{1Z}G_{3Z}`$ and $`(\theta _\sigma ,\theta _{\mathrm{S1}},\theta _{\mathrm{S2}})(\theta _\rho ,\theta _{\mathrm{C1}},\theta _{\mathrm{C2}}+2\sqrt{6}t_{}x/v_\mathrm{F})`$. The phase variables of the forward scattering term with opposite spins, $`_2`$, are expressed as $`\mathrm{\Theta }_{2E}=[(\theta _{\mathrm{C1}}ϵ^{}\sqrt{3}(\theta _{\mathrm{C2}}+2\sqrt{6}t_{}x/v_\mathrm{F}))+ϵ(\varphi _{\mathrm{S1}}ϵ^{}\sqrt{3}\varphi _{\mathrm{S2}})]/\sqrt{2}`$, $`\mathrm{\Theta }_{2F}=\sqrt{2}\theta _{\mathrm{C1}}+ϵ\sqrt{2}\varphi _{\mathrm{S1}}`$, $`\mathrm{\Theta }_{2G}=(\theta _{\mathrm{C1}}ϵ^{}\sqrt{3}\theta _{\mathrm{S2}}ϵϵ^{}\sqrt{3}\varphi _{\mathrm{C2}}+ϵ\varphi _{\mathrm{S1}})/\sqrt{2}`$ and $`\mathrm{\Theta }_{2H}=[(\theta _{\mathrm{C1}}+ϵ^{\prime \prime }\sqrt{3}(\theta _{\mathrm{C2}}+2\sqrt{6}t_{}x/v_\mathrm{F}))ϵ^{}(ϵ^{\prime \prime }3\theta _{\mathrm{S1}}\sqrt{3}\theta _{\mathrm{S2}})ϵϵ^{}(ϵ^{\prime \prime }3\varphi _{\mathrm{C1}}\sqrt{3}\varphi _{\mathrm{C2}})+ϵ(\varphi _{\mathrm{S1}}+ϵ^{\prime \prime }\sqrt{3}\varphi _{\mathrm{S2}})]/2\sqrt{2}`$. The forward scattering term with parallel spins, $`_{}`$ is obtained from $`_2`$ by replacing $`G_{2Z}G_Z`$($`Z=EH`$) and $`(\theta _{\mathrm{S1}},\theta _{\mathrm{S2}})(\varphi _{\mathrm{S1}},\varphi _{\mathrm{S2}})`$. Coupling constants are given by $`G_{iA}=G_{iB}=G_{iC}=G_{iD}=G_{iE}=G_{iF}=G_{iG}=G_{iH}=g_i/6\pi v_\mathrm{F}`$ ($`i=1,3`$) and $`G_{iE}=G_{iF}=G_{iG}=G_{iH}=g_i/6\pi v_\mathrm{F}`$ ($`i=2,`$) where $`g_{}g_2^{}g_1^{}`$. Non-linear terms of forward scattering with the same $`p`$ branch are discarded because these effect is negligibly small. By assuming scaling invariance with respect to $`\alpha \alpha ^{}=\alpha \mathrm{e}^{dl}`$, the second order RG equation for the interchain hopping is given by
$`{\displaystyle \frac{d}{dl}}\stackrel{~}{t}_{}=\stackrel{~}{t}_{}F(\stackrel{~}{t}_{},\{G_{iZ}\})K_{\mathrm{C2}},`$ (3)
where $`\stackrel{~}{t}_{}t_{}/(v_\mathrm{F}\alpha ^1)`$, $`F(\stackrel{~}{t}_{},\{G_{iZ}\})=[G_{1C}^2+G_{2E}^2+G_E^2]J_1(6\stackrel{~}{t}_{})+[G_{1H}^2+G_{2H}^2+G_H^2]J_1(3\stackrel{~}{t}_{})+[G_{3A}^2+G_{3D}^2+G_{3F}^2]J_1(4\stackrel{~}{t}_{})/3+[G_{3C}^2+G_{3E}^2+G_{3G}^2]J_1(2\stackrel{~}{t}_{})/3+G_{3B}^2J_1(8\stackrel{~}{t}_{})/3+G_{3H}^2J_1(\stackrel{~}{t}_{})/3`$ and $`J_n`$ is $`n`$-th Bessel function. Equation (3) is solved together with RG equations for $`G_{iZ}`$ where $`l`$ is related to energy scale $`\omega `$ or temperature $`T`$ by $`l=\mathrm{ln}(W/\omega )`$ or $`\mathrm{ln}(W/T)`$ with $`W(v_\mathrm{F}\alpha ^1)`$ being of the order of band width. We take $`\alpha =2a/\pi `$ \[?\]. It is noted that the r.h.s. of eq. (3) for small $`\stackrel{~}{t}_{}`$ is reduced to $`\stackrel{~}{t}_{}[1(G_1^2+G_2^2+G_{}^2+G_3^2)/2]`$ which becomes the same as that of many chains \[?\].
In Fig. 1, the $`l`$-dependence of coupling constants is shown for $`t_\mathrm{d}/t=0.1`$ and $`t_\mathrm{d}/t=0.025`$ with $`U/t=4`$ and $`t_{}/t=0.1`$. For $`t_\mathrm{d}/t=0.025`$ (inset), the interchain hopping $`t_{}(l)`$ becomes relevant corresponding to deconfinement where $`K_\rho `$ remains finite and $`K_{\mathrm{C1}}`$, $`K_\sigma `$ and $`K_{\mathrm{S2}}`$ ($`K_{\mathrm{C2}}`$ and $`K_{\mathrm{S1}}`$) decrease to zero (become infinite). The curves are calculated, for simplicity, by setting $`J_n(l)=0`$ for $`l>l_n`$ with $`l_n`$ corresponding the first node of the Bessel function, although such a treatment gives negligible difference in the numerical results. When $`K_\nu (l)`$ decreases to zero or increases to infinity, the corresponding phase is locked leading to a formation of gap, where relevant coupling constants are $`G_{1B}(\mathrm{})`$, $`G_{1D}(\mathrm{})`$, $`G_{1F}(\mathrm{})`$, $`G_{1G}(+\mathrm{})`$, $`G_{2F}(\mathrm{})`$ and $`G_{2G}(+\mathrm{})`$. There are two kinds of gap for charge fluctuations and three kinds of gap for spin fluctuations. For $`t_\mathrm{d}/t=0.1`$(main figure), one finds confinement where $`t_{}(l)`$ decreases to zero after taking a maximum. The $`l`$-dependence of $`K_\rho (l)`$ implies charge gap in the total charge fluctuation, which is in contrast to the case of deconfinement. For $`t_\mathrm{d}/t=0.1`$, exponents $`K_\nu (l)`$ ($`\nu =`$C1, C2, $`\sigma `$, S1, S2) remain finite, i.e., $`K_\nu (l)1`$ at $`l3`$.
In Fig. 2, $`t_{}(l)/t`$ is shown with the fixed $`U/t=3,U_c/t(3.7)`$ and 4.5 where solid (dotted) curves denote $`t_{}(l)/t`$ for $`0<l<l_\mathrm{\Delta }`$ ($`l>l_\mathrm{\Delta }`$) with $`K_\rho (l_\mathrm{\Delta })K_\rho /2`$. The case for $`l>l_\mathrm{\Delta }`$ is invalid since the magnitude of the umklapp scattering increases to infinity. In the curve $`t_{}(l)`$, there is a maximum given by max$`[t_{}(l)/t]3.4`$ (0.5) at $`l4.0`$ (2.2) for $`U/t=3.0`$ (4.5) although the maximum for $`U/t=3.0`$ is located in the region of $`l>l_\mathrm{\Delta }`$. Thus the case for $`U/t=3.0`$ (4.5) corresponds to deconfinement (confinement). The boundary between confinement and deconfinement is determined by the condition that max$`[t_{}(l)/t]1`$ at $`U=U_c`$.
In Fig. 3, the $`t_{}`$-dependence of $`U_c`$ is shown for $`t_\mathrm{d}/t=0.05`$(solid curves) and 0.1(dashed curves) where confinement (deconfinement) is obtained for $`U>U_c`$ ($`U<U_c`$). For comparison, the corresponding results for two-coupled chains \[?\] are also shown. The critical values for three-coupled chains is smaller than that for two-coupled chains. Since the RG equation of two-coupled chains corresponding to Eq. (3) includes the Bessel function with only $`J_1(8\stackrel{~}{t}_{})`$, the effect of umklapp scattering for three-coupled chains is stronger than that for two-coupled chains. The effective interchain hopping $`t_{}^{\mathrm{eff}}`$ can be evaluated from $`t_{}^{\mathrm{eff}}=t\mathrm{exp}[l_{\mathrm{eff}}]`$ where $`t_{}(l_{\mathrm{eff}})/t=1`$ \[?\]. In the inset, $`t_{}^{\mathrm{eff}}/t_{}`$ is shown as a function of $`t_{}`$ on a logarithmic scale. The power-law behavior of $`t_{}^{\mathrm{eff},0}`$ $`(t_{}^{\mathrm{eff}}(g_3=0)`$) for small $`t_{}`$ is consistent with the analytical formula $`t_{}^{\mathrm{eff},0}t_{}(t_{}/W)^{\alpha _0/(1\alpha _0)}`$ with $`\alpha _0=(K_\rho +K_\rho ^1+K_\sigma +K_\sigma ^14)/4`$ \[?\]. In the presence of the dimerization, $`t_{}^{\mathrm{eff}}`$ is reduced from that of the power-law behavior and has a critical value below which the confinement occurs.
The confinement-deconfinement transition is examined in terms of the charge gap induced by the umklapp scattering. The charge gap for the single chain is obtained as the function of $`U`$ and $`t_\mathrm{d}`$ by a method $`\mathrm{\Delta }_\rho ^{1\mathrm{D}}=W\mathrm{exp}[l_\mathrm{\Delta }]`$ with $`K_\rho (l_\mathrm{\Delta })=K_\rho /2`$ for $`t_{}=0`$. Such gap has a meaning of a characteristic energy of the umklapp scattering even for the deconfined region in which the charge gap is reduced to zero due to the presence of the misfit for all the nonlinear terms of umklapp scattering. In the inset of Fig. 4, a phase diagram of confinement (I) and deconfinement (II) is shown on the plane of the bare interchain hopping $`t_{}`$ and $`\mathrm{\Delta }_\rho ^{1\mathrm{D}}`$ where the boundary for two-coupled chains \[?\] are also shown for comparison. Although the ratio of $`\mathrm{\Delta }_\rho ^{1\mathrm{D}}/t_{}`$ is nearly constant, the curve is slightly convex upward for small $`t_{}`$. In the main figure of Fig. 4, the phase diagram with the same parameter is shown on the plane of $`t_{}^{\mathrm{eff},0}/t`$ and $`\mathrm{\Delta }_\rho ^{1\mathrm{D}}/t`$. The quantity $`t_{}^{\mathrm{eff},0}`$ denotes the effective interchain hopping, which is renormalized by the intrachain interaction without umklapp scattering. The ratio of $`\mathrm{\Delta }_\rho ^{1\mathrm{D}}`$ to $`t_{}^{\mathrm{eff},0}`$ at the boundary is estimated as follows when $`0.05<t_{}/t<0.3`$. The ratio for three-coupled chains is given by $`1.0\stackrel{<}{}\mathrm{\Delta }_\rho ^{1\mathrm{D}}/t_{}^{\mathrm{eff},0}\stackrel{<}{}\mathrm{\hspace{0.17em}1.1}`$ while that for two-coupled chains is given by $`1.8\stackrel{<}{}\mathrm{\Delta }_\rho ^{1\mathrm{D}}/t_{}^{\mathrm{eff},0}\stackrel{<}{}\mathrm{\hspace{0.17em}1.9}`$. By noting that the $`t_\mathrm{d}`$-dependence of the boundary is very small, it turns out that the confinement-deconfinement transition is determined by the competition between the charge gap and the effective interchain hopping energy.
We briefly discuss the case for OBC where the RG equation for the interchain hopping takes more complicated form due to twelve coupling constants for umklapp scattering. The small difference between PBC and OBC is expected since the RG equation for the interchain hopping has the same limiting form as that of PBC for small $`\stackrel{~}{t}_{}`$. Actually, we find almost the same boundary as Fig. 3 when the solution of single chain is substituted for the RG equation of interchain hopping.
In summary, the confinement-deconfinement transition in the three-coupled Hubbard chains with dimerization has been shown for PBC when the effective interchain hopping energy becomes of the order the charge gap induced by the umklapp scattering.
The authors thank H. Yoshioka for useful discussions. This work was supported by a Grant-in-Aid for Scientific Research from the Ministry of Education, Science, Sports and Culture (Grant No.09640429), Japan.
|
warning/0005/hep-th0005206.html
|
ar5iv
|
text
|
# Stability of the GRS Model
## I Introduction
Recently, there have been lots of interest in the phenomenon of localization of gravity proposed by Randall and Sundrum (RS) (for previous relevant work see references therein). RS assumed a single positive tension 3-brane and a negative bulk cosmological constant in the five dimensional (5D) spacetime. By considering metric fluctuations from a background which is isomorphic to sections of 5D anti-de Sitter spacetime ($`AdS_5`$), they have shown that it reproduces the effect of four dimensional (4D) gravity localized on the brane without the need to compactify the extra dimension due to the “warping” in the fifth dimensional space. In more detail, the solution to linearized equations in the five dimensions results in a zero mode, which can be identified with the 4D massless graviton, and massive continuum Kaluza-Klein (KK) modes. Surprisingly, the wavefunctions of the massive continuum KK modes are suppressed at the brane for small energies, and thus ordinary gravity localized on the brane is reproduced at large distances.
Gregory, Rubakov and Sibiryakov (GRS) have recently considered a brane model which is not asymptotically $`AdS_5`$, but Minkowski flat. In the GRS model, the ordinary 4D Newtonian potential is reproduced from not the massless zero mode, but the resonance of zero mass in the continuum KK spectrum . In this sense the GRS model of “a resonance graviton” would differ from the RS model. However, there exist two potential problems with this model: One is the mismatch in polarization states and the other is the violation of weaker energy condition (WEC). It was pointed out that a massive graviton propagator with $`3`$ polarization states does not reproduce the massless graviton propagator with $`2`$ due to the missmatch of the number of polarization states . Contrary to it, Csáki, Erlich and Hollowood have argued that in the presence of localized source at $`y=0`$ the effect ($`\xi ^5`$) of the bending of the brane exactly compensates for the extra polarization in the massive graviton propagator. Thus the $`m_h^20`$ limit of the massive propagator at intermediate scales is equivalent to the massless propagator of the Einstein theory just as in the RS scenario. At ultra large scales, however, this theory includes scalar anti-gravity , which may be cured by the RG analysis . The most important fact is probably that the mechanism to cancel the extra polarization gives arise to the ghost problem . In fact, the role of the radion field with a negative kinetic term is discussed in models with metastable graviton . In order to have a well-defined theory, the ghost should disappear.
We have introduced the trace field ($`h`$) in the RS model instead of $`\xi ^5`$ in Ref. . Instead of the localized source ($`T_{\mu \nu }(x,y)=T_{\mu \nu }(x)\delta (y)`$), we introduce a matter source with uniform trace along the extra dimension ($`T_\mu ^\mu (x,y)=T_\mu ^\mu (x)`$). Fortunately, it is shown that massive graviton modes contain ghost states which can be removed by assuming a further condition on the matter source.
The WEC is a basic requirement. In the GRS model we find that this is violated. But the compatibility between the WEC and the recovery of Einstein gravity seems to be not so important. In the brane world, the first thing that we have to do is to recover the Einstein gravity. A more important thing is that the weak energy condition may be closely related to the stability of the GRS spacetime. However, the actual stability analysis of a nonlinear system of the GRS model means to assess the reliability of its linearized approximation .
In this paper we wish to study the stability of the GRS background. We require that the stability of this spacetime be given by two:
1. There are no tachyons for $`h_{\mu \nu }(x)`$.
2. $`h_{\mu \nu }(x)`$ has no ghost (i.e., no negative norm state).
We will show that although the GRS spacetime does not satisfy the WEC, this spacetime is stable. In this paper, we use the signature $`(,+,+,+,+)`$ and MTW conventions.
## II Weaker energy condition (WEC)
In this section we explicitly show that the GRS model does not satisfy the weakest form of a positive energy condition, which is the so-called null energy condition or the weaker energy condition, saying that the stress-energy tensor $`T_{MN}`$ obeys $`T_{MN}\xi ^M\xi ^N0`$ for any null vector $`\xi ^M`$. As pointed out by Witten , given such energy condition, a “holographic $`c`$-theorem” for the $`AdS_5`$ says that as one approaches to spatial infinity in the extra dimension, the bulk cosmological constant $`\mathrm{\Lambda }`$ can only become more negative. The bulk cosmological constant in the GRS model is a negative constant in the vicinity of the positive tension brane at the center, but vanishes beyond the negative tension branes.
Let us consider a five-dimensional spacetime which is described by the metric as
$$ds^2=\widehat{g}_{MN}dx^Mdx^N=e^{2A(y)}g_{\mu \nu }dx^\mu dx^\nu +dy^2,$$
(1)
where $`g_{\mu \nu }`$ is Ricci flat (i.e., $`R_{\mu \nu }(g)=0`$) and we assume that the background matter ($`T_{MN}^{(0)}`$) producing such metric is distributed only on four-dimensional domain walls (i.e., $`T_{yy}^{(0)}=0`$). Then the non-vanishing components of the Ricci tensor are
$$\widehat{R}_{\mu \nu }=\left[A^{\prime \prime }+4(A^{})^2\right]\widehat{g}_{\mu \nu },\widehat{R}_{yy}=4\left[A^{\prime \prime }+(A^{})^2\right].$$
(2)
Here $`A^{}=_yA`$. Note that the contribution to the background matter stress tensor coming from the cosmological constant does not affect the WEC because it is proportional to the metric. Hence we separate $`\mathrm{\Lambda }`$ from $`T_{MN}^{(0)}`$. Using the Einstein equation $`\widehat{G}_{MN}=\widehat{R}_{MN}\frac{1}{2}\widehat{R}\widehat{g}_{MN}=\mathrm{\Lambda }\widehat{g}_{MN}+8\pi G_5T_{MN}^{(0)}`$, we see for any null vector $`\xi ^M`$
$`T_{MN}^{(0)}\xi ^M\xi ^N`$ $`=`$ $`{\displaystyle \frac{1}{8\pi G_5}}\widehat{R}_{MN}\xi ^M\xi ^N`$ (3)
$`=`$ $`{\displaystyle \frac{1}{8\pi G_5}}\left\{\left[A^{\prime \prime }+4(A^{})^2\right]\xi _\mu \xi ^\mu 4\left[A^{\prime \prime }+(A^{})^2\right](\xi ^y)^2\right\}`$ (4)
$`=`$ $`{\displaystyle \frac{3}{8\pi G_5}}A^{\prime \prime }(\xi ^y)^2.`$ (5)
On the last line we used $`\xi _M\xi ^M=\xi _\mu \xi ^\mu +(\xi ^y)^2=0`$.
Therefore, the WEC for the background matter $`T_{MN}^{(0)}\xi ^M\xi ^N0`$ for any null vector $`\xi ^M`$ becomes equivalent to $`A^{\prime \prime }0`$. Subsequently, $`A^{}`$ must be a non-increasing monotonic function in the coordinate $`y`$. From the $`yy`$-component of the Einstein equation, we also find
$$\mathrm{\Lambda }(y)=6(A^{})^2,$$
(6)
which implies that as one goes to large $`y`$, the cosmological constant can only become more negative provided that $`A^{}`$ was a negative value at some point.
For the perfect $`AdS_5`$ spacetime which corresponds to $`\mathrm{\Lambda }=\mathrm{constant}<0`$ and $`T_{MN}^{(0)}=0`$, $`A_{AdS}(y)=y/L`$ where $`L`$ is the radius of $`AdS_5`$. Thus the perfect anti-de Sitter spacetime leads to $`A^{}=1/L`$, $`A^{\prime \prime }=0`$, and so there is no RG flow at all. For the RS spacetime, $`A_{RS}(y)=|y|/L`$. Then $`A_{RS}^{}(y)=\theta (y)/L`$, $`A_{RS}^{\prime \prime }(y)=2\delta (y)/L0`$, and so the RS model satisfies the WEC. For the GRS spacetime, however,
$$A_{GRS}(y)=\{\begin{array}{cc}|y|/L\mathrm{for}|y|y_0,& \\ y_0/L\mathrm{for}|y|y_0.& \end{array}$$
(7)
Therefore, we have
$$A_{GRS}^{}(y)=\{\begin{array}{cc}\theta (y)/L\mathrm{for}|y|y_0,& \\ 0\mathrm{for}|y|y_0,& \end{array}A_{GRS}^{\prime \prime }(y)=\frac{2}{L}\delta ^{GRS}(y),$$
(8)
where $`\delta ^{GRS}(y)\delta (y)\frac{1}{2}\delta (yy_0)\frac{1}{2}\delta (y+y_0)`$.
As is shown in Fig. 1, $`A_{GRS}^{}(y)`$ is not a monotonically decreasing function and thus the condition of $`A^{\prime \prime }0`$ does not hold. On the position of the negative branes ($`y=\pm y_0`$), one finds $`A^{\prime \prime }(y)=\delta (y\pm y_0)/L`$. Of course, this indicates the violation of $`A^{\prime \prime }(y)0`$. We wish to emphasize that $`y=\pm y_0`$ are just points at which the negative branes are introduced to obtain the Minkowski spacetime for $`|y|y_0`$. In this sense the GRS model can be considered as a regularized version of the RS model. The fact that the GRS model violates the weakest form of a positive energy condition seems to indicate that the GRS background spacetime may be unstable. However, this is not all of the story. Whether or not the GRS background spacetime is really unstable should be checked by the perturbation study, which we shall do below.
## III Linearized perturbations in GRS model
The GRS model with a positive tension domain wall at $`z=0`$ and two negative tension walls at $`z=\pm z_0`$ perpendicular to the extra fifth direction can be described by the following action:
$$I=d^4x_{\mathrm{}}^{\mathrm{}}𝑑z\left[\frac{1}{16\pi G_5}\sqrt{\widehat{g}}(\widehat{R}2\mathrm{\Lambda })\sqrt{\widehat{g}_B}\sigma (z)+_M\right].$$
(9)
Although we used the horospherical coordinates $`x^M=(x^\mu ,y)`$ in the previous section, we introduce here the conformally flat coordinates $`x^M=(x,z)`$ for our perturbative analysis. Here $`G_5`$ is the 5D Newton’s constant, $`\mathrm{\Lambda }`$ the bulk cosmological constant of five dimensinal spacetime, $`\widehat{g}_B`$ the determinant of the metric describing the brane, and the tension of the branes $`\sigma (z)=\sigma \delta ^{\mathrm{GRS}}(z)`$ with $`\sigma =3/4\pi G_5L`$. $`I_M=d^4x𝑑z_M`$ denotes the matter action, and it contributes only in the linearized level.
If we introduce a conformal factor as follows
$$ds^2=\widehat{g}_{MN}dx^Mdx^N=H^2g_{MN}dx^Mdx^N,$$
(10)
the field equation becomes
$`G_{MN}+3{\displaystyle \frac{_M_NH}{H}}3g_{MN}\left[{\displaystyle \frac{_P^PH}{H}}2{\displaystyle \frac{_PH^PH}{H^2}}\right]`$ (11)
$`=8\pi G_5\left[{\displaystyle \frac{\mathrm{\Lambda }}{8\pi G_5H^2}}g_{MN}{\displaystyle \frac{\sqrt{g_B}}{\sqrt{g}}}{\displaystyle \frac{|H|}{H^2}}\sigma (z)g_{\mu \nu }\delta _M^\mu \delta _N^\nu {\displaystyle \frac{2}{\sqrt{\widehat{g}}}}{\displaystyle \frac{\delta I_M}{\delta \widehat{g}^{MN}}}\right]`$ (12)
with the Einstein tensor $`G_{MN}`$ constructed from the metric $`g_{MN}`$. Now it is straightforward to see that, in the absence of matter source except for the domain walls themselves (i.e., $`\delta I_M/\delta \widehat{g}^{MN}=0`$), the most general solution having a 4D Poincaré symmetry is
$$ds^2=H^2(z)(\eta _{\mu \nu }dx^\mu dx^\nu +dz^2),$$
(13)
where
$$H(z)=\{\begin{array}{cc}\frac{1}{L}|z|+1,\hfill & \\ \frac{1}{L}|z_0|+1,\hfill & \end{array}\mathrm{\Lambda }(z)=\{\begin{array}{cc}\frac{6}{L^2}\mathrm{for}|z|z_0,\hfill & \\ 0\mathrm{for}|z|z_0.\hfill & \end{array}$$
(14)
Let us consider metric fluctuations around this background spacetime as follows :
$$g_{MN}=\eta _{MN}+h_{MN}.$$
(15)
Defining $`\overline{h}_{MN}=h_{MN}\frac{1}{2}\eta _{MN}h`$ where $`h=\eta ^{MN}h_{MN}`$, the linearized perturbation equation of Eq. (12) is
$`{\displaystyle \frac{1}{2}}\mathrm{}\overline{h}_{MN}+_{(M}^P\overline{h}_{N)P}{\displaystyle \frac{1}{2}}\eta _{MN}^P^Q\overline{h}_{PQ}{\displaystyle \frac{3^PH}{2H}}(_Mh_{NP}+_Nh_{MP}_Ph_{MN})`$ (18)
$`3\eta _{MN}\left[\left({\displaystyle \frac{^P^QH}{H}}+2{\displaystyle \frac{^PH^QH}{H^2}}\right)h_{PQ}{\displaystyle \frac{^QH}{H}}^P\overline{h}_{PQ}\right]3\left({\displaystyle \frac{\mathrm{}H}{H}}2{\displaystyle \frac{_PH^PH}{H^2}}\right)h_{MN}`$
$`+8\pi G_5H^2\left\{{\displaystyle \frac{\mathrm{\Lambda }(z)}{8\pi G_5}}h_{MN}+\right|H|\sigma (z)\left[{\displaystyle \frac{1}{2}}(\eta ^{\alpha \beta }h_{\alpha \beta }\eta ^{PQ}h_{PQ})\eta _{\mu \nu }\delta _M^\mu \delta _N^\nu +\delta _M^\mu \delta _N^\nu h_{\mu \nu }\right]\}`$
$`=`$ $`8\pi G_5T_{MN},`$ (19)
where the linearized matter source $`T_{MN}=\delta (2\delta I_M/\sqrt{\widehat{g}}\delta \widehat{g}^{MN})`$ is included, and $`\mathrm{}=\eta ^{MN}_M_N`$. Taking the 5D harmonic gauge condition,
$$^M\overline{h}_{MN}=0\mathrm{or}^Mh_{MN}=\frac{1}{2}_Nh,$$
(20)
the linearized equation becomes
$`\mathrm{}h_{MN}+3{\displaystyle \frac{_5H}{H}}(_Mh_{5N}+_Nh_{5M}_5h_{MN})+\left({\displaystyle \frac{8(_5H)^2}{H^2}}+{\displaystyle \frac{4_5^2H}{H}}\right)h_{55}\eta _{MN}`$ (21)
$`6{\displaystyle \frac{_5^2H}{H}}\left[{\displaystyle \frac{1}{2}}h_{55}\eta _{\mu \nu }\delta _M^\mu \delta _N^\nu h_{5\mu }(\delta _M^\mu \delta _N^5+\delta _M^5\delta _N^\mu )\right]=16\pi G_5\left(T_{MN}{\displaystyle \frac{1}{3}}\eta _{MN}T_P^P\right).`$ (22)
Notice that we do not impose the trace free condition $`h=h_M^M=0`$. In components, the above equations become
$`(\mathrm{}+3{\displaystyle \frac{_5H}{H}}_5)h_{55}+\left({\displaystyle \frac{8(_5H)^2}{H^2}}+{\displaystyle \frac{4_5^2H}{H}}\right)h_{55}={\displaystyle \frac{32\pi G_5}{3}}\left(T_{55}{\displaystyle \frac{1}{2}}T_\rho ^\rho \right),`$ (23)
$`(\mathrm{}+6{\displaystyle \frac{_5^2H}{H}})h_{5\mu }+3{\displaystyle \frac{_5H}{H}}_\mu h_{55}=16\pi G_5T_{5\mu },`$ (24)
$`(\mathrm{}3{\displaystyle \frac{_5H}{H}}_5)h_{\mu \nu }+3{\displaystyle \frac{_5H}{H}}(_\mu h_{5\nu }+_\nu h_{5\mu })+\left({\displaystyle \frac{8(_5H)^2}{H^2}}+{\displaystyle \frac{_5^2H}{H}}\right)h_{55}\eta _{\mu \nu }`$ (25)
$`=16\pi G_5\left(T_{\mu \nu }{\displaystyle \frac{1}{3}}\eta _{\mu \nu }T_P^P\right).`$ (26)
Note that $`\overline{h}=\eta ^{MN}\overline{h}_{MN}=\frac{3}{2}h`$.
For simplicity, we consider a case of $`h_{5\mu }=T_{5\mu }=0`$. Then Eq. (24) implies that $`h_{55}`$ is a function of the $`z`$-coordinate only. Thus, one can rescale the $`z`$-coordinate so that $`h_{55}=0`$. In other words, we take the Gaussian normal gauge (i.e., $`h_{5\mu }=h_{55}=0`$) without the trace free condition. Subsequently, Eq. (23) shows that it is neccessary for the matter source to satisfy the following relation
$$T_{55}=\frac{1}{2}T_\mu ^\mu .$$
(27)
We see that this is exactly the stabilization condition for the extra dimension implemented in Refs. . Note that $`T_P^P=T_\mu ^\mu +T_{55}=\frac{3}{2}T_\mu ^\mu `$. Then Eq. (26) becomes
$$(\mathrm{}3\frac{_5H}{H}_5)h_{\mu \nu }=16\pi G_5\left(T_{\mu \nu }\frac{1}{2}\eta _{\mu \nu }T\right),$$
(28)
where $`T=T_\mu ^\mu `$. Since $`h=\eta ^{\mu \nu }h_{\mu \nu }+h_{55}=h_\mu ^\mu `$, the harmonic gauge condition in Eq. (20) gives
$$^\mu (h_{\mu \nu }\frac{1}{2}\eta _{\mu \nu }h_\rho ^\rho )=0,_5h_\rho ^\rho =0.$$
(29)
Since the trace of Eq. (28) is
$$(\mathrm{}3\frac{_5H}{H}_5)h_\mu ^\mu =16\pi G_5T,$$
(30)
Eq. (28) can also be written as
$$(\mathrm{}3\frac{_5H}{H}_5)(h_{\mu \nu }\frac{1}{2}\eta _{\mu \nu }h_\rho ^\rho )=16\pi G_5T_{\mu \nu }.$$
(31)
Thus, the gauge condition in Eq. (29) strictly leads to the source conservation law
$$^\mu T_{\mu \nu }=0.$$
(32)
This is a relic of the 4D Poincaré symmetry in the linearized level. Note also that, using the second gauge condition in Eq. (29), we find from Eq. (30)
$$\mathrm{}_4h_\mu ^\mu =16\pi G_5T_\mu ^\mu \mathrm{with}\mathrm{}_4=\eta ^{\mu \nu }_\mu _\nu .$$
(33)
This means that the trace $`h`$ can propagate on the brane if one includes the matter source. Note, however this corresponds to the massless scalar propagation. Furthermore, by taking $`_5`$ on Eq. (33) and using $`_5h_\mu ^\mu =0`$, we have additional constraints for the source
$$_5T_\mu ^\mu =_5T_{55}=0.$$
(34)
Therefore, for the consistency of linearized equations, we find that $`T_\mu ^\mu `$ and $`T_{55}`$ of the matter source are to be constant in the extra dimension. Our uniform matter of $`T_\mu ^\mu =T_\mu ^\mu (x)`$ is lead to keep up the trace $`h`$ with a physical variable. In the absence of the matter source (more precisely, $`T_\mu ^\mu =0`$), the trace $`h`$ belongs to a gauge degree of freedom and thus it can be gauged away. In this case, $`h`$ is not a physical variable. Further our matter source with uniform trace cannot affect the massive KK modes because it is independent of $`z`$.
Defining $`h_{\mu \nu }(x,z)=H^{3/2}\psi (z)\widehat{h}_{\mu \nu }(x)`$, $`\psi (z)`$ satisfies the following Schrödinger-like equation
$$\left[\frac{1}{2}_5^2+\frac{15(_5H)^2}{8H^2}\frac{3_5^2H}{4H}\right]\psi (z)=\frac{1}{2}m_h^2\psi (z),$$
(35)
where the seperation constant $`m_h`$ plays as the mass of the 4D gravitational wave $`\widehat{h}_{\mu \nu }(x)`$. Then Eq. (28) becomes
$$(\mathrm{}_4m_h^2)h_{\mu \nu }=16\pi G_5(T_{\mu \nu }\frac{1}{2}\eta _{\mu \nu }T),$$
(36)
## IV No tachyonic condition
Now we are in a position to discuss “no tachyonic condition.” Defining $`2\mathrm{ln}H(z)=B(z)`$, Eq. (35), which determines the spectrum of KK excitations, can be written as
$$\frac{d^2\psi (z)}{dz^2}+[\frac{9}{16}B^{}(z)^2\frac{3}{4}B^{\prime \prime }(z)]\psi (z)=m_h^2\psi (z).$$
(37)
This can be further taken into the factorization form
$$[\frac{d}{dz}+\frac{3}{4}B^{}(z)][\frac{d}{dz}+\frac{3}{4}B^{}(z)]\psi (z)=m_h^2\psi (z)$$
(38)
which has the form of the supersymmetric quantum mechanics $`Q^{}Q\psi (z)=m_h^2\psi (z)`$, with $`Q=\frac{d}{dz}+\frac{3}{4}B^{}(z)`$.
The lowest energy state is the zero-energy state which satisfies the supersymmetric condition $`Q\widehat{\psi }_0(z)=0`$, $`\widehat{\psi }_0(z)=e^{\frac{3}{4}B(z)}=H^{3/2}(z)`$. This does not correspond to the normalizable spin-2 propagation. Hence there is no negative energy graviton modes (tachyon modes) in the GRS model. We prove that $`m_h^20`$.
## V No ghost state
Now we examine the graviton propagator on the positive brane at $`z=0`$ by considering $`h_{\mu \nu }(x,0)\widehat{h}_{\mu \nu }(x)`$ only. It requires the bilinear forms of the source with the inverse propagator to isolate the physical modes. As the present analysis is on the classical level, we express $`\widehat{h}_{\mu \nu }`$ in terms of source. Taking Fourier transformation for Eq. (36) to momentum space results in
$$\widehat{h}_{\mu \nu }(p)=\frac{16\pi G_5}{p^2+m_h^2}\left[T_{\mu \nu }(p)\frac{1}{2}\eta _{\mu \nu }T(p)\right].$$
(39)
Then the one graviton exchange amplitude for the source $`T_{\mu \nu }`$ is given by
$$A^{\mathrm{class}}=\frac{1}{4}\widehat{h}_{\mu \nu }(p)T^{\mu \nu }(p)=\frac{4\pi G_5}{p^2+m_h^2}(T^{\mu \nu }T_{\mu \nu }\frac{1}{2}T^2).$$
(40)
In order to study the massive states, it is best to use the rest frame in which
$$p_10,p_2=p_3=p_4=0.$$
(41)
Considering Eqs. (32) and (41) leads to the following source relations
$$T_{11}=T_{12}=T_{13}=T_{14}=0.$$
(42)
Thus, one obtains
$$T^{\mu \nu }T_{\mu \nu }\frac{1}{2}T^2=|T_{+2}|^2+|T_2|^2+|T_{+1}|^2+|T_1|^2+T_{44}\left[\frac{1}{2}T_{44}(T_{22}+T_{33})\right],$$
(43)
where the first two terms correspond to the exchange of graviton with helicity-2 $`T_{\pm 2}=\frac{1}{2}(T_{22}T_{33})\pm iT_{23}`$, and the third and fourth terms are the exchange of the graviphoton with helicity-1 $`T_{\pm 1}=T_{24}\pm iT_{34}`$. We note here that the last term in the above equation is not positive definite. This means that there exist ghost states (negative norm states) in general. However, if one requires
$$T_{44}=2(T_{22}+T_{33}),$$
(44)
one immediately finds that
$$T^{\mu \nu }T_{\mu \nu }\frac{1}{2}T^2=|T_{+2}|^2+|T_2|^2+|T_{+1}|^2+|T_1|^2$$
(45)
with all positive norm states and without helicity-0 states (graviscalars). In the case of $`2(T_{22}+T_{33})=aT_{44}`$ with $`a<1`$, we find no ghost states, but there exist the graviscalars which arise from the diagonal elements of $`T_{\mu \nu }`$.
In the limit of $`m_h^20`$, the graviphoton propagation can be decoupled from the brane . Hence we can neglect $`|T_{\pm 1}|^2`$-terms. Finally the amplitude takes the form
$$A_{m_h^20}^{\mathrm{class}}=\underset{m_h^20}{lim}\frac{4\pi G_5}{p_1^2+m_h^2}\left[|T_{+2}|^2+|T_2|^2\right],$$
(46)
which corresponds to the massless spin-2 amplitude.
## VI Discussion
Naively, it is conjectured that, if a certain $`AdS_5`$ spacetime does not satisfy the WEC, this may belong to an unstable manifold. However, this is not all of the story. We believe that the stability analysis of the given spacetime in curved space is based on the perturbation study around the background spacetime. This corresponds to testing the reliability of its linearized approximation. Hence we investigate the stability of the GRS spacetime along this line. Here, as in usual Minkowski background, we require two conditions for the stable background: no tachyon and no ghost states. Especially, we guarantee that the GRS background is stable against the small perturbation because it has no tachyon and no ghost states. No tachyon condition is easily checked by observing the Shrödinger equation for the massive KK modes. However, showing that there is no ghost states in the GRS model is a non-trivial task. This is because, in this model, the massless spin-2 propagation (graviton) can be only described by the massive KK modes with resonance.
The main problem is to cancel the unwanted extra polarization in the quasi-localization of 4D gravity. This is done by introducing both the trace ($`h`$) and the matter source with uniform trace ($`T_\mu ^\mu `$) at the linearized level. In the conventional RS approach, the trace ($`h`$) is just a gauge-dependent scalar and hence it can be gauged away. However, including the matter with uniform trace, this plays a role of $`\xi ^5`$ in the brane-bending model with the localized source . This is because $`h`$ ($`\xi ^5`$) satisfy the nearly same massless equations of $`\mathrm{}_4h=16\pi G_5T_\mu ^\mu `$ ($`\mathrm{}_4\xi ^5=\frac{8\pi G_5}{6}S_\mu ^\mu `$ in Ref. ). And the comparison of $`\overline{h}_{\mu \nu }=h_{\mu \nu }\frac{1}{2}\eta _{\mu \nu }h`$ with $`\overline{h}_{\mu \nu }=h_{\mu \nu }^{(m)}+2L^1\eta _{\mu \nu }\xi ^5`$ in Ref. confirms the close relationship between $`h`$ and $`\xi ^5`$. If $`T_\mu ^\mu =0`$, one finds from Eq. (40) that the massive spin-2 states have 5 polarizations with all positive norm states . Here, in the case of $`h0`$, $`T_\mu ^\mu 0`$, requiring the additional condition $`T_{44}=2(T_{22}+T_{33})`$ in Eq. (44), we find the massless spin-2 state with 2 polarizations in the limit of $`m_h^20`$. In this case the ghost states disappear.
In conclusion, it turns out that the GRS spacetime is stable against the small perturbation if $`T_\mu ^\mu 0`$, $`h0`$ with $`T_\mu ^\mu =2T_{55}=3(T_{22}+T_{33})`$. On the other hand, the RS spacetime is stable under the RS gauge ($`h=h_{5\mu }=h_{55}=0,_\mu h^{\mu \nu }=0`$) and $`T_\mu ^\mu =0`$ .
## Acknowledgments
The authors thank Hyungwon Lee for helpful discussions. This work was supported by the Brain Korea 21 Programme, Ministry of Education, Project No. D-0025.
|
warning/0005/astro-ph0005577.html
|
ar5iv
|
text
|
# BeppoSAX Observations of Bright Radio Galaxies
## 1. Introduction
The most basic classification of Active Galactic Nuclei (AGN) consists in dividing them in two classes: radio-quiet and radio-loud. This represents not only an observational distinction but a basic physical difference whose basis is still not understood. Radio morphologies (lobes, jets etc.) are obviously the product of some physical mechanisms at work in the nuclei of radio-loud objects. Rees et al. (1982) suggested, for example, that the primary engine in radio galaxies was a spinning black hole fed by thick, and hot, accretion flow. Later, Blandford (1990) and Meier (1999) have explicitly identified the black hole spin as a possible physical parameter responsible for the radio-loud and radio-quiet dichotomy. If the hole rotates faster, it is more efficient in producing the jets observed in radio-loud objects.
X-ray photons seem to be the best probe to investigate the radio loudness issue as they are produced (and reprocessed) in the inner regions of AGNs where accretion occurs. X-ray observations of radio-loud objects therefore provide unique data to test the model hypotheses. In particular, two questions can be addressed: 1) Does the jet contribute to the X-ray emission in non-blazar radio-loud AGNs? 2) Are the same accretion processes at work in radio galaxies (and in general in radio-loud AGNs) and Seyfert galaxies?
BeppoSAX is a broad band (0.1-200 keV) X-ray satellite (Scarsi 1993) which has ideal characteristics for this purpose. Here we present a BeppoSAX analysis of 6 Broad Line Radio Galaxies and compare our sources with a sample of 13 Seyfert 1 galaxies also observed by BeppoSAX (Matt these proceedings, hereafter M00). Based on the properties of optical-UV spectra, BLRGs are considered the radio-loud counterpart of Seyfert 1s (Urry and Padovani 1995). The comparison of the X-ray (nuclear) properties of these two classes of objects is then particularly appropriate to investigate the radio-loud and radio-quiet dichotomy.
## 2. Observations and Results
The radio galaxies presented here are part of a larger sample of Radio-Loud Emission-Line AGNs observed by BeppoSAX, whose analysis is still on-going. In this paper, we restrict the discussion to a subsample of 6 BLRGs, which can be directly compared with the sample of Seyfert 1s discussed by M00. The sample is shown in Table 1. All the 3C sources have been observed for about 100 ksec. PKS2152-69 and Pictor A were observed for $`17`$ and $`30`$ ksec, respectively.
### 2.1. BLRG Spectral Analysis
The spectral analysis results, obtained by simultaneously fitting the LECS (0.1-4 keV) MECS (1.5-10 keV) and PDS (15-100 keV) instruments are reported in Table 2.
A simple power law plus Galactic absorption gave good fits to the data of PKS2152-69 and Pictor A. The iron line is absent in Pictor A in agreement with the ASCA measurement (Eracleous et al. 1998). For PKS2152-69, the exposure time was too short to allow the detection of even a strong feature. For both, we could obtain only an upper limit for the iron line equivalent width. In 3C111 the continuum was also a simple power law and the iron line only marginally detected (in spite of the long exposure time). The soft photons appeared strongly absorbed. It is however possible that the cold absorber is not intrinsic to the source but associated to our Galaxy. 3C111 is in fact behind the Galactic dark cloud, Taurus B (see Reynolds et al. 1998 for more details).
More complex models were necessary to fit the other radio galaxies. Absorption in excess of the Galactic column density, iron lines, reflection (Ref.) humps, soft excesses and bendings of the high energy spectrum (Cutoff) are common spectral features, although not always simultaneously present in each source.
The soft excess was detected in two sources, namely 3C120 and 3C382. It was parameterized with a steep power law ($`\mathrm{\Gamma }^{\mathrm{𝑠𝑜𝑓𝑡}}=34`$) and contributed to the total emission in the 0.1-2 keV band more than 60$`\%`$ in both the sources. The origin of this excess is not clear. It could be related to thermal emission from a cold thin disk, to radiation coming by extended thin plasma as well as to a very soft jet emission.
All the BLRGs with detected iron line show reflection components. Although the strength of the reflection (Ref.) is not very well constrained, data seem to suggest that radio-galaxies with weak iron lines have also weak reflections (Figure 1). If confirmed, this would imply that the line is generated in the same material which reflects the X-ray primary photons.
### 2.2. Comparison between BLRGs and Seyfert 1s
We compared the spectral properties of our BLRGs with a sample of 13 Seyfert 1s observed by BeppoSAX (M00). The 2-10 keV luminosities of Seyferts (ranging from $`10^{42}`$ to $`10^{44}`$ erg sec<sup>-1</sup>) and BLRGs (see Table 1) partially overlap.
The X-ray primary power law of BLRGs is generally flat ($`<\mathrm{\Gamma }^{\mathrm{𝐵𝐿𝑅𝐺}}=1.73>`$, rms dispersion $`\sigma _{rms}^{\mathrm{𝐵𝐿𝑅𝐺}}=0.08`$). The Seyfert 1 sample is characterized by a steeper average spectral slope. However, the larger spread of the spectral indices in radio-quiet AGNs ($`<\mathrm{\Gamma }^{\mathrm{𝑆𝑒𝑦}\mathit{1}}>`$=1.87, $`\sigma _{rms}^{\mathrm{𝑆𝑒𝑦}\mathit{1}}=0.24`$), does not allow to statistically confirm any difference between Seyfert 1s and BLRGs.
Half the 3C sources in our sample shows a bending of the X-ray spectrum at high energies. When it is modeled with an exponential cutoff, the cutoff energies are similar to those observed in Seyferts.
While in radio-quiets the iron lines and the reflection components are always present, in BLRGs the reprocessed features are detected in only 3 sources. It should be noted the absence of the lines cannot be only attributed to poor statistics. In addition in radio-louds, the iron line equivalent widths are significantly smaller than in radio-quiets ($`EW^{\mathrm{𝐵𝐿𝑅𝐺}}=71`$ eV, $`\sigma _{rms}^{\mathrm{𝐵𝐿𝑅𝐺}}=45`$ eV; $`EW^{\mathit{Sey1}}=174`$ eV, $`\sigma _{rms}^{\mathit{Sey1}}=57`$ eV).
A general weakness of the reprocessed features in BLRGs is also confirmed by the XTE results presented by Sambruna and Eracleous (these proceedings). Note that Wozniak et al. (1998) have already discussed this possibility analyzing ASCA, GINGA and OSSE data.
A cold absorbing column in excess of Galactic is observed in 3 (including 3C111) out of 5 BLRGs with LECS data. No source shows absorption edges typical of warm absorber, which, on the contrary, is rather common in Seyfert 1s. It is possible that the absorbing material is different in BLRGs and Seyfert 1s, being warm in radio-quiets and cold in radio-louds (see also Sambruna Eracleous and Mushotzky 1999). This is also supported by a historical study of the column density changes in 3C390.3. The long-time (years) variability of the intrinsic N<sub>H</sub> does not appear to be correlated to the flux intensity at 1 keV (see Fig. 2 in Grandi et al. 1999). It is possible that variations in the geometry of the absorber rather than changes of its ionization state (as expected in the case of a warm absorber) are responsible for the N<sub>H</sub> long term variability.
## 3. Discussion
Two important points arise from the BeppoSAX study of BLRGs: 1) there is not a unique type of BLRG X-ray spectrum, but a variety of cases; 2) there are several differences between BLRGs and Seyfert 1 galaxies. The most impressive difference concerns the reprocessed features that are weaker in radio-louds than in radio-quiets (Fig. 1). How can these results be explained?
BLRGs are complex objects, in which at least three X-ray components, a jet, an accretion flow and a molecular torus, can contribute to the formation of the spectrum.
Jet – In some BLRGs the radio jet shows superluminal motion. It is then reasonable to suppose that Doppler-enhanced radiation also contaminates the X-ray spectra and dilutes the reprocessed spectral features when the Doppler factor of the jet ($`\delta `$=\[$`\gamma `$(1-$`\beta `$ cos)\]<sup>-1</sup>) is sufficiently large. If, in agreement with the AGN Unified Schemes, BLRGs and intrinsically powerful blazars are the same objects seen at different angles of view, also the jet output of radio galaxies should be Inverse Compton in the PDS band (Fossati et al. 1998, Ghisellini 1998). If this is the case, it is rather improbable that the observed high energy steepening in BLRGs is related to the jet. In powerful blazars, the Compton break (i.e. the Compton peak in the Spectral Energy Distribution, SED) is usually revealed at MeV-GeV energies. In order to occur at 100-200 keV as observed in BLRGs, the Doppler factor ( $`\delta `$=\[$`\gamma `$(1-$`\beta `$ cos)\]<sup>-1</sup>) should be smaller by about a factor 100 or more. This would imply a strong de-amplification of the non-thermal radiation ($`I^{obs}`$($`\nu `$)=$`\delta ^3`$I<sup>intr.</sup>($`\nu ^{})`$) and, in turn, a difficult detection of the jet.
Accretion Flow –It is not really clear how gas accretion occurs in radio-loud objects. The simplest approach is to extend the physical models developed for Seyferts to Broad Line Radio Galaxies. As in the case of radio-quiet AGNs, the accretion gas flow could be a cold geometrically thin disk with a hot corona above it (Haardt and Maraschi 1991, 1993; Petrucci et al. 2000). The role of the corona is to transform into X-ray photons, via inverse Compton scattering, the UV photons generated by the disk. Down-scattered X-rays, in turn, hit the disk and are reprocessed, generating an iron line and a reflection hump above 10 keV. A Seyfert-like accretion disk should then produce the following signatures in X-ray BLRG spectra: a soft excess (related to the thermal disk emission), a high energy thermal cutoff (related to the corona temperature), a broad red-shifted iron line (which suffers the vicinity of the strong black hole gravitational field) and a strong reflection component (produced by a disk subtending a $`2\pi `$ solid angle to the X-ray primary source, i.e. Refl$`\mathrm{\Omega }/2\pi 1`$).
Alternatively, one can assume that different physical/geometrical accretion configurations occur in AGNs which discharge large amount of energy in jets. Rees et al. (1982) speculated on the possibility that in radio-galaxies the hot accreting gas is in the shape of an ion-supported torus characterized by low radiative efficiency. The Advection Dominated Accretion Flow (ADAF) models (Narayan e al. 1998), recently proposed, follow similar lines of though. Shapiro Lightman and Eardley (SLE, 1976) found another solution for the hot flow, that resembles the ion-supported torus. However, in the SLE model the energy produced in the accreting gas by viscosity is locally radiated and not advected radially.
Then the accretion flow in BLRGs could be hot and geometrically thick in the inner region and become cold and geometrically thin only at larger radii (Chen and Halpern 1989).
Given the smaller covering factor of the cold (and reprocessing) matter to the X-ray primary source, this accretion configuration predicts less prominent soft excesses, narrower weaker iron lines and smaller reflection components than the Seyfert case. In addition, a lack of correlation between the variations of the primary X-ray source and the reprocessed features is expected; the entity of the temporal delay depending on the inner radius of the cold disk.
Molecular Torus – It is probable that a thick wall of absorbing material (perhaps in a toroidal shape) surrounds the accretion flow. The torus can produce the iron line and reflect the X-ray primary photons (Ghisellini Haardt and Matt 1994) further complicating the X-ray spectrum. In addition, if the opening angle of the torus is small, its (less thick) upper layers could be intercepted by the observer and cause the soft X-ray depletion sometimes observed in BLRGs.
Since its distance from the X-ray source is large ($`1`$ pc), iron line and reflection components should respond with a considerable delay (light-years) to the continuum variations.
It is possible and also probable that all these components are present in BLRGs. If they are mixed in different ways in different objects, the variety of X-ray spectra observed with BeppoSAX would be easily explained. As shown in Table 3 the main BeppoSAX spectral features of the BLRGS with high signal-to noise spectra (i.e. pointed for about 100 ksec) can be opportunely reproduced choosing a plausible combination of the nuclear components.
This tentative table shows that a Seyfert-like disk alone can not reproduce the observations. The weakness of the reprocessed features requires the presence either of a non-thermal radiation from a jet or of a cold disk sub-tending a small solid angle to the X-ray primary source (see figure 2).
## 4. Seyfert-like or ADAF-like Accretion?
The sources in Table 3 which seem better to fit the Seyfert-like+jet model are 3C111 and 3C120. If the model is correct, the weak (but detected) iron line in 3C120 assures that Seyfert and Blazar components have comparable intensity around $`6`$ keV, i.e. the non-thermal power law is important and can effect the continuum shape. We then tested whether the 3C120 spectrum can be fitted by a mix of beamed and un-beamed radiation on the entire BeppoSAX band (i.e. on about 3 decades in energies).
A simple power law was assumed to represent the jet (blazar) spectrum and a power law with cutoff plus iron line and reflection component were utilized to mimic the Seyfert emission. The Seyfert equivalent width was assumed EW$`{}_{}{}^{sey}=174`$ eV, the average value from the Matt sample. Different combinations of spectral slopes for the Seyfert and blazar models were tested. In all the cases, the relative normalization at 6.4 keV of the two power laws was fixed in order to reproduce the observed Fe equivalent width reported in Table 2. As shown in Figure 3 (left panel), 3C120 data can be fitted by a blazar and a Seyfert-like power law (in Figure $`\mathrm{\Gamma }^{Blazar}=2`$, $`\mathrm{\Gamma }^{Sey}=1.7`$). However the model requires very low energy cutoff ($`E_{Cut}70`$ keV) and a strong reflection (Ref$`2`$). These best fit values are rather extreme if compared to those of Seyferts. This slightly disfavors the jet model, although it cannot be completely rejected given the large uncertainties associated to the parameters.
A geometry that assumes a hot inner flow and cold disk surrounding it seems to be particularly appropriate for 3C390.3. In this source the UV bump is weak, there is not soft excess and the iron line is narrow. In this case a molecular torus (or a warped disk) also contributes to increase the strength of the iron line and the reflection component (Grandi et al. 1999).
The idea that the iron line emitting regions cannot be located near the X-ray is also supported by recent BeppoSAX observations of the very bright radio-galaxy Centaurus A (Cen A) ($`F_{\mathit{2}\mathit{10}\mathrm{𝑘𝑒𝑉}}10^{10}`$ erg cm<sup>-2</sup> sec<sup>-1</sup>). It is a FR I optically classified as Narrow Line Radio Galaxy and shows a radio-optically-X-ray jet pointed far away from us ($`i6070^0`$). It is characterized by a X-ray spectrum rather complex. However above 3 keV it is dominated by the nuclear point-like source (Turner et al. 1998).
Cen A was observed on 1997 February 20-21 (30 ksec), on 1998 January 6-7 (50 ksec) and twice for 40 ksec during the 1999 summer on July 10-11 and August 2-3. The repeated observations of this radio-galaxy, thanks to their high statistics, have allowed a detailed study of the nuclear flux variations versus the iron line flux changes. As clearly shown in Figure 3 (right panel), line and continuum do not vary together, in particular the iron line was more intense when the source was weaker (compare the 1997 and 1998 observations). Since the inclination of the jet is large in Cen A, the contribution of non-thermal radiation to the total X-ray nuclear continuum is expected to be negligible. Then Figure 3 (right panel) simply indicates that the line emitting region responds with a significant delay to the continuum variations, as it is expected if the reprocessing gas is located far from the primary X-ray source.
We conclude that, although, in principle, the two proposed accretion scenarios are both viable, the idea that a jet reduces the reprocessed features is lightly disfavored.
## 5. Conclusions
BeppoSAX analysis of 6 BLRGs has shown that a variety of X-ray spectra exits. A mix of different components, a jet, an accretion flow and a molecular torus, could explain the observations. The comparison between our 6 BLRG and a sample of 13 Seyfert 1s has pointed out important differences between radio-loud and radio-quiet AGNs. In particular, it has been shown that BLRG reprocessed features are often absent or weaker than in Seyfert 1s. This result has important implications. If the accretion flow in BLRGs and Seyferts is identical (i.e a cold thin optically thin disk with a hot above it) a strong contribution by Doppler-enhanced radiation is necessary to explain the weakness of the reprocessed features. Alternatively, if the jet emission is not important, the cold re-processing gas has to subtend a smaller solid angle to the X-ray primary source. A possibility is that the X-ray continuum is produced by a hot geometrically thick ion-supported torus that illuminates a cold thin disk at larger radii.
Choosing between the two scenarios is still premature. The new satellites (XMM, Chandra, INTEGRAL) will play an important role in solving the problem. Detailed studies of the iron line profiles in a large (and well selected) sample will allow to confirm the presence of cold matter very near to the black hole (if lines are broad and redshifted) or at large distances (if lines are narrow).
Variability studies will be also crucial. An anti-correlation is expected between the EW and the X-ray continuum flux, if the jet is dominant: the observed continuum (disk + jet) should be more variable than the Fe line (from disk). Alternatively, if no correlation is found, the Fe variations do not follow the continuum flux variations and a temporal delay effect (as observed in CenA case) is present. Finally, detailed studies of the hard X-ray continuum will be able to detect any beamed radiation. At very high energies, where the Seyfert-like power law drops, the jet, if intense, should emerge and become directly detectable (see Figure 3 left panel).
## ACKNOWLEDGEMENTS
I would like to thank all the people who, working with me on the BeppoSAX radio-galaxy project, have allowed the writing of this paper: L. Maraschi, C. M. Urry, E. Massaro, G. Matt, M. Guainazzi, F. Haardt, P. Giommi, G.G. Palumbo, G. Malaguti. An acknowledgment also go the L. Piro and the BeppoSAX CDC team for their support to this project. I also thank G. Ghisellini, L. Ferretti and C. G. Perola for the useful comments and stimulating discussions and A. Bazzano and A.J. Bird for critical reading of the manuscript. I am very grateful to M. Frutti for invaluable help in realizing Figure 2.
## REFERENCES
Blandford, R. D., 1990 in Active Galactic Nuclei, Saas-Fee Advanced Course, pag. 264
Chen, K. and Halpern, J. P., 1989, ApJ, 344, 115
Eracleous et al. 1998 ApJ, 550,557
Fossati G., Maraschi L., Celotti A., Comstri A., Ghisellini, G., 1998, MNRAS, 299, 433
Ghisellini G., Haardt F., and Matt G., 1994, MNRAS, 267, 743
Ghisellini G., Celotti A., Fossati G., Maraschi L., Comastri A., 1998, MNRAS, 301, 451
Grandi, P., et al. 1999, A$`\&`$A, 343, 40
Haardt, F., Maraschi, L., 1991, ApJ, 380, L51
Haardt F., Maraschi L., 1993, ApJ, 413, 507
Meier D. L., 1999, New Astronomy Reviews in press (astro-ph/9908283)
Narayan, R., Mahadevan, R. and Quataert, E. 1998, The Theory of Black Hole Accretion Disk, eds M.A. Abramowicz, G. Bjornsson, J.E. Pringle, pag.148
Nandra P., et al. 1997, ApJ, 477, 602
Petrucci P.O. et al. these proceedings
Reynolds C. S., 1997, MNRAS, 286, 513
Reynolds C. S., Iwasawa K., Crawford C.S., Fabian A.C., 1998, MNRAS, 299, 410
Sambruna M. R., Eracleous M., Mushotzky R. F., 1999, ApJ, 256, 60
Sambruna M. R., Eracleous M., these proceedings
Rees M.J., Begelman M.C., Blandford R.D., Phinney E.S., 1982, Nature, 295, 17
Shakura, N. I., Sunyaev, R. A. , 1973, A$`\&`$A, 24, 337
Shapiro, S. L., Lightman A. P., Eardley D. M., 1976, ApJ, 204, 187
Urry C.M. and Padovani P., 1995, PASP, 107,803
Wozniak P. R., Zdziarski A. A., Smith D., et al. 1998, MNRAS, 299, 449
|
warning/0005/hep-th0005063.html
|
ar5iv
|
text
|
# References
Diffeomorphism invariance requires that 4D quantum gravity becomes 4th order derivative theory for gravity sector. We recently showed that 4th order actions, including the Wess-Zumino (WZ) action , are uniquely determined by diffeomorphism invariance. Then, the theory also becomes renormalizable . Especially, our model satisfies the integrability condition on the WZ action discussed by Riegard, Fradkin and Tseytlin , which is generalized by the author to the form that can be applied to higher loops. A problem in 4th-order theories is that there are extra negative-metric states. Thus, the unitarity becomes obscure . In this paper we shall see that there is a posibility that diffeomorphism invariance also ensures the unitarity.
Here, we briefly explain how to realize diffeomorphism invariance. The details of the argument were discussed in our previous papers . Perturbation theory is defined by replacing the invariant measure with the measure defined on the background-metric. As a lesson from 2D quantum gravity , in order to preserve background-metric independence, or diffeomorphism invariance, we must add an action, $`S`$, which satisfies the WZ condition , as
$$Z=\frac{[d\varphi ]_{\widehat{g}}[\text{e}^hd\text{e}^h]_{\widehat{g}}[df]_{\overline{g}}}{\text{vol(diff.)}}\mathrm{exp}\left[S(\varphi ,\overline{g})I(f,g)\right],$$
(1)
where $`f`$ is a matter field and $`I`$ is an invariant action. The metric is now decomposed as $`g_{\mu \nu }=\text{e}^{2\varphi }\overline{g}_{\mu \nu }`$ and $`\overline{g}_{\mu \nu }=(\widehat{g}\text{e}^h)_{\mu \nu }`$, where $`tr(h)=0`$ . The measures of the metric fields are defined on the background-metric by the norms:
$`<d\varphi ,d\varphi >_{\widehat{g}}={\displaystyle }d^4x\sqrt{\widehat{g}}(d\varphi )^2,`$ (2)
$`<dh,dh>_{\widehat{g}}={\displaystyle }d^4x\sqrt{\widehat{g}}tr(\text{e}^hd\text{e}^h)^2.`$ (3)
The general coordinate transformation, $`\delta g_{\mu \nu }=g_{\mu \lambda }_\nu \xi ^\lambda +g_{\nu \lambda }_\mu \xi ^\lambda `$, is expressed in 4 dimensions as
$`\delta \varphi ={\displaystyle \frac{1}{4}}\widehat{}_\lambda \xi ^\lambda +\xi ^\lambda _\lambda \varphi ,`$
$`\delta \overline{g}_{\mu \nu }=\overline{g}_{\mu \lambda }\overline{}_\nu \xi ^\lambda +\overline{g}_{\nu \lambda }\overline{}_\mu \xi ^\lambda {\displaystyle \frac{1}{2}}\overline{g}_{\mu \nu }\widehat{}_\lambda \xi ^\lambda ,`$ (4)
where $`\overline{}_\lambda \xi ^\lambda =\widehat{}_\lambda \xi ^\lambda `$ is used. Under a general coordinate transformation, $`\delta I=0`$, but the WZ action is not invariant. $`\delta S`$ is proportional to the form of conformal anomaly . Diffeomorphism invariance is realized such that $`\delta S`$ cancels anomalous contributions, $`U`$, which originates from the fact that the measures defined above are no longer invariant under the transformation, as
$$\delta Z=<\delta S+U>=0.$$
(5)
More rigorously, consider the regularized 1PI effective action, $`\mathrm{\Gamma }_{\mathrm{eff}}`$, of the combined theory, $`=S+I`$, and require $`\delta \mathrm{\Gamma }_{\mathrm{eff}}=0`$, which determines $`S`$ uniquely.
In 2 dimensions we can take the conformal gauge $`h_\nu ^\mu =0`$, and hence 2D quantum gravity coupled to conformal matter can be described as a free conformal field theory . Of course, in 4 dimensions, the combined theory $`=S+I`$ can not be described as a free theory. We must take into account interactions between the conformal mode and the traceless mode in the WZ action as well as self-interactions of the traceless mode, which are ruled by the background-metric independence for the traceless mode . Thus, we must generalize the idea of 2D quantum gravity based on conformal field theories to one based on diffeomorphism invariance. The original idea on this matter is given in a study of 2D quantum dilaton gravity ,<sup>2</sup><sup>2</sup>2 There is an analogy between the dilaton field $`\phi `$ defined in and the traceless mode in our 4D model. Unfortunately, this model is unrenormalizable in the perturbation of non-minimal coupling because $`\phi `$ is a dimensionless scalar in 2 dimensions so that there are many diffeomorphism invariant counterterms like $`\phi ^n^\mu \phi _\mu \phi `$. On the other hand the dynamics of the traceless mode is ruled by the background-metric independence for the traceless mode, itself, so that the model has almost no ambiguity. and then developed to 4D quantum gravity .
The conformal-mode dynamics of the WZ action in 4 dimensions has been discussed in refs. in analogy to 2D quantum gravity. But, in their model there is no interactions for the traceless mode, because they considered the WZ action as a full effective action given after the traceless mode as well as matter fields are integrated out. From the viewpoint of full diffeomorphism invariance, it is inaccurate in 4 dimensions.
In this note we give the BRST formulation of diffeomorphism invariant quantum gravity. We first review the BRST formulation of 2D quantum gravity . We here emphasize that the nilpotence of the BRST transformation, which is equivalent to diffeomorphism invariance, is realized dynamically. We then show that similar considerations can apply to diffeomorphism invariant 4D quantum gravity. At the end we discuss the possibility of how to remove the negative-metric states in the 4D model from the viewpoint of diffeomorphism invariance. Preliminary discussions on this matter have already given in .
Our curvature conventions are $`R_{\mu \nu }=R_{\mu \lambda \nu }^\lambda `$ and $`R_{\mu \sigma \nu }^\lambda =_\sigma \mathrm{\Gamma }_{\mu \nu }^\lambda \mathrm{}`$.
2D quantum gravity
Firstly, we briefly review the BRST formulation of 2D quantum gravity. The WZ action in two dimensions, what is called the Liouville action, is given by integrating the 2D conformal anomaly as
$$S(\varphi ,\overline{g})=\frac{a}{4\pi }d^2x\sqrt{\widehat{g}}\left(\overline{g}^{\mu \nu }_\mu \varphi _\nu \varphi +\overline{R}\varphi \right).$$
(6)
In 2 dimensions we can take the gauge condition $`h_\nu ^\mu =0`$ up to the zero mode. The gauge-fixed combined action then becomes
$$=\frac{1}{4\pi }d^2x\sqrt{\widehat{g}}\left[a\left(\overline{g}^{\mu \nu }_\mu \varphi _\nu \varphi +\overline{R}\varphi \right)+_{GF+FP}+\mathrm{\Lambda }\text{e}^{\alpha \varphi }\right]+I_M(f,g),$$
(7)
where $`I_M`$ is an invariant matter action. The gauge-fixing term and the Faddeev-Popov (FP) ghost action are given by
$$_{GF+FP}=iB_{\mu \nu }(\overline{g}^{\mu \nu }\widehat{g}^{\mu \nu })+2\overline{g}^{\mu \nu }b_{\mu \lambda }\overline{}_\nu c^\lambda ,$$
(8)
where the reparametrization ghost $`c^\mu `$ is a contravariant vector. $`B_{\mu \nu }`$ and the anti-ghost $`b_{\mu \nu }`$ are covariant symmetric traceless tensors. In the following, $``$ is considered as a ”classical” action.
Consider $`N`$ massless scalars as matter fields. Then, diffeomorphism invariance requires that the coefficient, $`a`$, must be
$$a=\frac{1}{6}(25N).$$
(9)
The BRST transformation is given by
$`\delta _𝐁\overline{g}_{\mu \nu }=i(\overline{g}_{\mu \lambda }\overline{}_\nu c^\lambda +\overline{g}_{\nu \lambda }\overline{}_\mu c^\lambda \overline{g}_{\mu \nu }\widehat{}_\lambda c^\lambda ),`$
$`\delta _𝐁\varphi =ic^\lambda _\lambda \varphi +i{\displaystyle \frac{1}{2}}\widehat{}_\lambda c^\lambda ,`$
$`\delta _𝐁b_{\mu \nu }=B_{\mu \nu },\delta _𝐁B_{\mu \nu }=0,`$ (10)
$`\delta _𝐁c^\mu =ic^\lambda _\lambda c^\mu ,`$
where $`c^\lambda \overline{}_\lambda c^\mu =c^\lambda _\lambda c^\mu `$ due to the anti-commutativity of $`c^\mu `$. Note that the $`h`$-dependence appears only in $`\delta _B\overline{g}_{\mu \nu }`$. These transformations are nilpotent; $`\delta _B^2(\overline{g}_{\mu \nu },\varphi ,b_{\mu \nu },c^\mu )=0`$. Then, the gauge-fixing term and the FP ghost action can be written as $`_{GF+FP}=i\delta _𝐁\{b_{\mu \nu }(\overline{g}^{\mu \nu }\widehat{g}^{\mu \nu })\}`$ , where $`\delta _B\overline{g}^{\mu \nu }=\overline{g}^{\mu \lambda }\overline{g}^{\nu \sigma }\delta _B\overline{g}_{\lambda \sigma }`$.
The well-known form of the BRST transformation in 2D quantum gravity is given by integrating out over the $`B_{\mu \nu }`$ field. Hence, we obtain the following one:
$`\delta _𝐁\varphi =ic^\lambda _\lambda \varphi +i{\displaystyle \frac{1}{2}}\widehat{}_\lambda c^\lambda ,`$
$`\delta _𝐁b_{\mu \nu }=2i𝒯_{\mu \nu },`$ (11)
$`\delta _𝐁c^\mu =ic^\lambda _\lambda c^\mu .`$
Now, $`h_\nu ^\mu `$ becomes non-dynamical and we can set $`\overline{g}_{\mu \nu }=\widehat{g}_{\mu \nu }`$ in the expressions. To guarantee $`\delta _Bh_\nu ^\mu =0`$, $`c^\mu `$ should satisfy the conformal Killing equation, $`\widehat{}^\mu c^\nu +\widehat{}^\nu c^\mu \widehat{g}^{\mu \nu }\widehat{}_\lambda c^\lambda =0`$. $`𝒯_{\mu \nu }`$ is the modified energy-momentum tensor, which is determined by using the equation of motion for the traceless mode and tracelessness of $`b_{\mu \nu }`$, as <sup>3</sup><sup>3</sup>3 Because of the tracelessness of $`b_{\mu \nu }`$, there is an ambiguity $`\gamma `$, which is that one can add $`\gamma b_{\mu \nu }\widehat{}_\lambda c^\lambda `$ to the energy-momentum tensor for the $`bc`$-system. It is now fixed by the condition $`\delta _B_{bc}=0`$.
$`𝒯_{\mu \nu }`$ $`=`$ $`T_{\mu \nu }{\displaystyle \frac{1}{2}}\widehat{g}_{\mu \nu }T_\lambda ^\lambda `$ (12)
$`=`$ $`{\displaystyle \frac{a}{2}}\left\{_\mu \varphi _\nu \varphi {\displaystyle \frac{1}{2}}\widehat{g}_{\mu \nu }^\lambda \varphi _\lambda \varphi \left(\widehat{}_\mu \widehat{}_\nu {\displaystyle \frac{1}{2}}\widehat{g}_{\mu \nu }\widehat{\mathrm{}}\right)\varphi \right\}`$
$`+\widehat{}_{(\mu }c^\lambda b_{\nu )\lambda }+{\displaystyle \frac{1}{2}}c^\lambda \widehat{}_\lambda b_{\mu \nu }{\displaystyle \frac{1}{2}}\widehat{g}_{\mu \nu }\widehat{}^\lambda c^\sigma b_{\lambda \sigma }+T_{\mu \nu }^M.`$
Here, $`T_{\mu \nu }=\frac{2\pi }{\sqrt{\widehat{g}}}\frac{\delta }{\delta \widehat{g}^{\mu \nu }}`$ is the energy-momentum tensor of the gauge-fixed combined theory (7) with $`h_\nu ^\mu =0`$. $`T_{\mu \nu }^M`$ is the energy-momentum tensor of $`N`$ massless scalars. Since, in two dimensions, $`𝒯_\lambda ^\lambda =0`$, 2D quantum gravity can be expressed by conformal field theory.
The transformation (11) is now no longer nilpotent ”classically”. Using expression (12) and the conformal Killing equation of $`c^\mu `$, we can show that $`\delta _B^2b_{\mu \nu }`$ produces a non-vanishing quantity,
$$\delta _B𝒯_{\mu \nu }=i\frac{a}{4}\left(\widehat{}_\mu \widehat{}_\nu \frac{1}{2}\widehat{g}_{\mu \nu }\widehat{\mathrm{}}\right)\widehat{}_\lambda c^\lambda ,$$
(13)
which implies that the energy-momentum tensor, $`𝒯_{\mu \nu }`$, forms Virasoro algebra with central charges $`6a`$ classically . It reflects that the action, $``$, is not BRST-invariant: <sup>4</sup><sup>4</sup>4 Eq. (14) can be derived either by applying (10) to the action (7) or by applying (11) to the action obtained by integrating the $`B_{\mu \nu }`$ field out, where the conformal Killing equation of $`c^\mu `$ is not necessary even in the later case.
$$\delta _B=\frac{ia}{8\pi }d^2x\sqrt{\widehat{g}}\overline{R}\widehat{}_\lambda c^\lambda .$$
(14)
The BRST invariance of 2D quantum gravity is realized dynamically as follows. Quantum effects, namely, conformal anomalies give contributions to the central charge, $`N25`$, so that the total of the central charge becomes $`c_{tot}=6a+N25`$. Thus, the nilpotence of the BRST transformation is realized at the quantum level when $`a`$ is given by equation (9).
Finally, we give brief comments on physical states in 2D quantum gravity. The presence of the $`\frac{1}{2}\widehat{}_\lambda c^\lambda `$ term in the BRST transformation of $`\varphi `$ implies that the asymptotic state of the conformal mode, $`\varphi |0>`$, is not physical . On the other hand, the BRST invariant state, naively defined by $`\text{e}^{\alpha \varphi }|0>`$, where $`\alpha `$ is a real value determined by the BRST invariance, is not normalizable . Furthermore, in 2 dimensions, the normalizable Hamiltonian eigenstates are not BRST invariant because they clearly do not satisfy the Hamiltonian constraint, $`H=𝒯_0^0=0`$. Thus, there is no gravitational degrees of freedom in 2D quantum gravity.
4D quantum gravity
In 4 dimensions the WZ action becomes 4th order, parametrized by three constants $`a`$, $`b`$ and $`c`$ in the form
$`S(\varphi ,\overline{g})`$ $`=`$ $`{\displaystyle \frac{1}{(4\pi )^2}}{\displaystyle }d^4x[\sqrt{\widehat{g}}\{a\overline{F}\varphi +2b\varphi \overline{\mathrm{\Delta }}_4\varphi +b(\overline{G}{\displaystyle \frac{2}{3}}\stackrel{}{\mathrm{}}\overline{R})\varphi \}`$ (15)
$`{\displaystyle \frac{1}{36}}(2a+2b+3c)(\sqrt{g}R^2\sqrt{\widehat{g}}\overline{R}^2)],`$
where $`F`$ is the square of the Weyl tensor and $`G`$ is the Euler density defined by
$`F`$ $`=`$ $`R_{\mu \nu \lambda \sigma }R^{\mu \nu \lambda \sigma }2R_{\mu \nu }R^{\mu \nu }+{\displaystyle \frac{1}{3}}R^2,`$ (16)
$`G`$ $`=`$ $`R_{\mu \nu \lambda \sigma }R^{\mu \nu \lambda \sigma }4R_{\mu \nu }R^{\mu \nu }+R^2.`$ (17)
$`\mathrm{\Delta }_4`$ is the conformally covariant 4th order operator ,
$$\mathrm{\Delta }_4=\mathrm{}^2+2R^{\mu \nu }_\mu _\nu \frac{2}{3}R\mathrm{}+\frac{1}{3}(^\mu R)_\mu .$$
(18)
Why the number of the independent parameters is three is due to the fact that $`R^2`$ is not integrable w.r.t. conformal mode .
We consider the following invariant action including 4th order terms:
$$I(f,g)=I_4+I_{LE},$$
(19)
where
$`I_4={\displaystyle \frac{d}{(4\pi )^2}}{\displaystyle d^4x\sqrt{g}R^2},`$ (20)
$`I_{LE}={\displaystyle \frac{1}{(4\pi )^2}}{\displaystyle d^4x\sqrt{g}(m^2R+\mathrm{\Lambda })}+I_M(f,g).`$ (21)
$`I_4`$ is the 4th order action with $`d>0`$ which as well as the WZ action could be regarded as being a part of the measure. $`I_{LE}`$ is the usual 2nd order action which describes low-energy physics. $`m^2`$ is the inverse of the gravitational constant and $`\mathrm{\Lambda }`$ is the cosmological constant. $`I_M`$ is a matter action. Here, note that the presence of the $`\varphi \overline{F}`$ term in the WZ action implies that the Weyl term, $`\sqrt{g}F(=\sqrt{\widehat{g}}\overline{F})`$, can be produced by expanding around a vacuum expectation value (VEV) of $`\varphi `$.
Let us define the perturbation around VEV of $`\varphi `$ such that the Weyl term, $`\frac{1}{t^2}\overline{F}`$, is produced, where we introduce the dimensionless coupling constant, $`t`$, for the traceless mode and the $`h_\nu ^\mu `$ field is replaced with $`th_\nu ^\mu `$ in the combined action . Then, the integral region of $`\varphi `$ is effectively restricted within the region $`\frac{1}{t}<\varphi <\frac{1}{t}`$. This perturbation theory seems to be well-defined; namely, it is expected that $`t`$ is small enough in comparison with recent numerical experiments .
Let us consider the regularized 1PI effective action, $`\mathrm{\Gamma }_{\mathrm{eff}}`$. As discussed in , diffeomorphism invariance, namely, $`\delta \mathrm{\Gamma }_{\mathrm{eff}}=0`$ gives constraints on actions of gravitational fields as well as matter fields, which requires that 4D model must satisfy the following conditions:
* Matter fields must conformally couple to gravity.
* The coefficient, $`d`$, in 4th order action must be
$$d=\frac{1}{36}(2a+2b+3c).$$
(22)
The second condition means that self-interactions of the conformal mode, namely the $`R^2`$ terms, cancel out in the combined action. Then, the three coefficients $`a`$, $`b`$ and $`c`$, can be determined uniquely in the perturbation of the coupling, $`t`$, by requiring diffeomorphism invariance.<sup>5</sup><sup>5</sup>5The coefficients depend on matter contents, but the sign of $`a`$ ($`b`$) is negative-definite (positive-definite) at the zero-th order of $`t`$. $`c`$ vanishes at this order. And also the sum of them, $`d`$, becomes positive.
These conditions are more precisely represented as that, in the regularized 1PI effective action, there is no non-local correction to the WZ action like $`\varphi \stackrel{2}{\stackrel{}{\mathrm{}}}\mathrm{log}(\stackrel{}{\mathrm{}})\varphi `$ and no non-local term, $`\overline{R}\mathrm{log}(\stackrel{}{\mathrm{}})\overline{R}`$, which is associated to non-conformally invariant counterterm proportional to $`\overline{R}^2`$. This is the generalized form of the integrability condition discussed in . Here, there are two remarks. The first is that the two types of non-local corrections considered here are related to each other by the background-metric independence for the conformal mode. The second is that vanishing of the $`\overline{R}\mathrm{log}(\stackrel{}{\mathrm{}})\overline{R}`$ term does not always imply vanishing of the $`\overline{R}^2`$ divergences at higher loops.
The combined action, $`=S+I`$, including the gauge-fixing term and the FP ghost action, now becomes
$`={\displaystyle \frac{1}{(4\pi )^2}}{\displaystyle }d^4x[{\displaystyle \frac{1}{t^2}}\overline{F}+a\overline{F}\varphi +2b\varphi \overline{\mathrm{\Delta }}_4\varphi +b(\overline{G}{\displaystyle \frac{2}{3}}\stackrel{}{\mathrm{}}\overline{R})\varphi `$
$`+{\displaystyle \frac{1}{36}}(2a+2b+3c)\overline{R}^2+_{GF+FP}]+I_{LE}(f,g).`$ (23)
Here and below, we take the flat background $`\widehat{g}_{\mu \nu }=\delta _{\mu \nu }`$. The gauge-fixing term is given by .
$$_{GF}=2iB^\mu N_{\mu \nu }\chi ^\nu \zeta B^\mu N_{\mu \nu }B^\nu ,$$
(24)
where $`\chi ^\nu =^\lambda h_\lambda ^\nu `$ and $`N_{\mu \nu }`$ is a symmetric 2nd order operator. The corresponding ghost action becomes 4th order:
$$_{FP}=2i\overline{c}^\mu N_{\mu \nu }^\lambda \delta _𝐁h_\lambda ^\nu ,$$
(25)
where the BRST transformation of the traceless mode is given by
$`\delta _𝐁h_\nu ^\mu `$ $`=`$ $`i[^\mu c_\nu +_\nu c^\mu {\displaystyle \frac{1}{2}}\delta _\nu ^\mu _\lambda c^\lambda +tc^\lambda _\lambda h_\nu ^\mu `$ (26)
$`+{\displaystyle \frac{t}{2}}h_\lambda ^\mu (_\nu c^\lambda ^\lambda c_\nu )+{\displaystyle \frac{t}{2}}h_\nu ^\lambda (^\mu c_\lambda _\lambda c^\mu )+\mathrm{}].`$
This is given by replacing $`\xi ^\mu /t`$ in the transformation with the contravariant vector ghost field, $`ic^\mu `$. The kinetic term of the ghost action then becomes $`t`$-independent. The BRST transformations for other fields are given by
$`\delta _𝐁\varphi =itc^\lambda _\lambda \varphi +i{\displaystyle \frac{t}{4}}_\lambda c^\lambda ,`$
$`\delta _𝐁\overline{c}^\mu =B^\mu ,\delta _𝐁B^\mu =0,`$ (27)
$`\delta _𝐁c^\mu =itc^\lambda _\lambda c^\mu .`$
The transformations, (26) and (27), are nilpotent. Using the BRST transformation, the gauge-fixing term and the FP ghost action can be written as $`_{GF+FP}=2i\delta _𝐁\{\overline{c}^\mu N_{\mu \nu }(\chi ^\nu +\frac{i}{2}\zeta B^\nu )\}`$ .
As in 2D quantum gravity, the BRST transformation of the ”classical” action, $``$, is not BRST-invariant:
$$\delta _B=\frac{it}{4(4\pi )^2}d^4x_\lambda c^\lambda \left[a\left(\overline{F}+\frac{2}{3}\stackrel{}{\mathrm{}}\overline{R}\right)+b\overline{G}+c\stackrel{}{\mathrm{}}\overline{R}\right].$$
(28)
The BRST invariance is equivalent to diffeomorphism invariance, which is realized dynamically as mentioned before.
If the $`B^\mu `$ field is integrated out, this field is related to the energy-momentum tensor through the field equation for the traceless mode. As in 2D quantum gravity, it means that the nilpotence of the BRST transformation requires that the BRST transformation of the energy-momentum tensor vanishes at the quantum level.
Let us consider the long-distance physics of this model. The theory is asymptotically free, namely, $`a<0`$ for the coupling of the traceless mode, $`t`$. Thus, one can drop the Weyl term at the long distance. On the other hand, we leave the other three kinetic terms of gravitational fields: $`\varphi \overline{\mathrm{\Delta }}_4\varphi `$, $`\overline{R}^2`$ and $`m^2R`$. Let us compute the degrees of freedom in this case. Since $`\overline{R}^2=\chi ^\mu _\mu _\nu \chi ^\nu +o(h^3)`$, the 2nd-order operator, $`N_{\mu \nu }`$, is proportional to $`_\mu _\nu +m^2\delta _{\mu \nu }`$,<sup>6</sup><sup>6</sup>6 Although there are the off-diagonal $`\varphi h`$ terms in the $`\varphi \stackrel{}{\mathrm{}}\overline{R}`$ term and the Einstein-Hilbert term, they can be diagonalized into a 4th-order kinetic term of $`\varphi `$ and this form for $`h`$, provided that one set $`t=1`$. where $`m^2`$ comes from the Einstein-Hilbert term, which is proportional to $`m^2`$. The ghost determinant is now given by $`det^{1/2}(_\mu _\nu +m^2\delta _{\mu \nu })detM_{\mu \nu }^{GH}`$, where $`M_{\mu \nu }^{GH}`$ is the usual 2nd order ghost operator of diffeomorphism invariance given by applying the BRST transformation to $`\chi ^\mu `$ such that $`\delta _B\chi _\mu |_{h=0}=M_{\mu \lambda }^{GH}c^\lambda `$. Note that $`det(_\mu _\nu +m^2\delta _{\mu \nu })=det(\mathrm{}+m^2)|_{\mathrm{a}\mathrm{scalar}}`$, so that the number of ghost degrees of freedom is $`4+4+1=9`$. Hence, the total degrees of freedom becomes $`2\times 1+99=2`$. Thus, the negative-metric state related to the conformal mode is removed by ghosts. Really, due to the form of the BRST transformation, the conformal mode can not be a BRST invariant asymptotic state .
This result is quite suggestive, because, if there is a mechanism to remove the Weyl term, it seems that the theory becomes unitary. Recall that an exact diffeomorphism invariance implies that the integral region of $`\varphi `$ is unrestricted above and below, so that the Weyl term can be absorbed into the $`\varphi \overline{F}`$ term in the WZ action, which is just the original theory (19) with (22) adding the WZ action. Thus, we expect that diffeomorphism invariance ensures the unitarity non-perturbatively.
Acknowlegments
This work is supported in part by the Grant-in-Aid for Scientific Research from the Ministry of Education, Science and Culture of Japan.
|
warning/0005/astro-ph0005076.html
|
ar5iv
|
text
|
# ISO observations of a sample of 60 𝜇m peaker galaxies
## 1 Introduction
IRAS galaxies with spectral energy distributions peaking near 60 $`\mu `$m Vader et al. (1993) are known as Sixty Micron Peakers (SMPs or 60PKs). Vader et al. Vader et al. (1993) constructed the sample using the IRAS flux ratios $`f_{60}/f_{100}>1`$ and $`1<f_{60}/f_{25}<4`$ to select extragalactic sources with galactic latitude $`|b|>10\mathrm{°}`$ from the second version of the IRAS Point Source Catalogue. There are only 51 galaxies in the sample, constituting about 2% of the space density of 60 $`\mu `$m-selected galaxies in the range L$`_{60\mu \mathrm{m}}`$ = 10<sup>9</sup>–10<sup>12</sup>L that have been identified out to a redshift of 0.2.
All SMPs are strong emission-line galaxies Vader et al. (1993). The optical line ratio criteria of Veilleux & Osterbrock Veilleux & Osterbrock (1987) have been used to classify SMPs as H II region-like (H2) or Seyfert (Sy) galaxies. Of the 41 SMPs that have been classified, 23 are Sy2, 5 are Sy1 and 13 are H2. Approximately 60% of SMPs are therefore Seyfert galaxies. This fraction is far higher than the percentage of Seyferts found in samples of ultraluminous IRAS galaxies Heisler & Vader (1994). There are very few Sy1 galaxies in the full sample and none of the galaxies presented here is a Sy1. Only 4 of a sample of 45 SMPs have spiral optical morphologies, the majority being amorphous and peculiar Heisler & Vader (1994). The paucity of spiral galaxies is another notable feature of SMPs, since IRAS galaxies are typically spirals. The luminosity range of SMPs of about three orders of magnitude, spans the classes from dwarf galaxies to giant ellipticals.
Near-infrared (NIR) brightness profiles of H2 galaxies are well modelled by a single r<sup>1/4</sup> fit but an additional nuclear point source component is required to fit the majority of Seyfert galaxy profiles Heisler & Vader (1995). The radio continuum and H$`\alpha `$ emission is compact, indicating that the FIR radiation is also emitted from a small volume Heisler et al. (1998). The compact region has an extent of a few kpc which is comparable in size to typical narrow emission-line regions in AGN. The central region of SMPs therefore appears to be dominated by strong non-thermal or starburst emission. This emission is heavily obscured at optical wavelengths. A short-lived phase of central activity, caused by a recent interaction/merger, accounts for the morphology, the brightness near 60 $`\mu `$m and the strong non-thermal or starburst source Heisler & Vader (1995).
The $`f_{60}/f_{100}>1`$ criterion preferentially selects galaxies with warmer or more centrally located dust. The ‘cirrus’ component which is the dominant contributor to the 100 $`\mu `$m flux in spiral galaxies must therefore be weak or absent in SMPs Vader et al. (1993). Very small grains (VSGs) of carbonaceous material contribute strongly to the flux between 25 and 60 $`\mu `$m Désert et al. (1990); Laureijs (1998) and SMPs may be dominated by emission from VSGs. There is still some debate as to the actual carriers of the CH and CC aromatic bonds that produce emission features that dominate in the 3–13 $`\mu `$m spectral region of spiral galaxies Lu et al. (1999) and Seyferts Clavel et al. (1999). It is commonly assumed that the emission is produced by polycyclic aromatic hydrocarbons (PAHs) but other models exist such as the ‘Coal Model’ Papoular et al. (1989); Guillois et al. (1998). Laboratory and theoretical results based on PAH mixtures fit the observed spectra consistently Salama (1998); Boulanger et al. (1998); Moutou et al. (1998); Langhoff (1996). The features at 3.3, 6.2, 7.7, 8.6 and 11.3 $`\mu `$m are therefore referred to as ‘PAH emission’ hereafter. PAH emission is complex Peeters et al. (1999); Boulanger et al. (1998); Langhoff (1996) and varies spatially within galaxies Tielens et al. (1999); Moorwood (1999). The 2–11 $`\mu `$m spectra of SMPs are expected to be largely dominated by the ubiquitous PAH emission. Their spectra should however be modified by direct emission from non-thermal or starburst components and from hot dust (VSGs) Désert et al. (1990). The presence of high energy photons will also modify the spectrum indirectly by exciting and destroying different PAHs preferentially depending on their composition Uchida et al. (1998); Roelfsema et al. (1996).
Sect. 2 details observations of eight SMPs with the ISOPHOT photopolarimeter Lemke et al. (1996) on board the Infrared Space Observatory (ISO) Kessler et al. (1996). Results of the observations and a discussion are presented in Sect. 3. The separation of H2 and Sy galaxies is discussed in Sect. 4. Conclusions are in Sect. 5. In this paper, $`H_0`$ = 75 km s<sup>-1</sup>Mpc<sup>-1</sup> and $`q_0`$ = 0.5 are adopted.
## 2 Observations and Data Reduction
Eight of the brightest SMPs visible to ISO were selected for observation with ISOPHOT-S being roughly representative of the full SMP sample. Four of the galaxies are H2, three are Sy2, and the classification of one is uncertain. Observations were carried out between December 1997 and February 1998. Table 1 lists the R.A., Dec. and the redshift of the galaxies as well as the ISO data archive number and the date of the observation.
PHT-S consists of a dual grating spectrometer with a resolving power of $`\lambda /\mathrm{\Delta }\lambda 90`$ in two wavelength bands. Band SS covers the wavelength range 2.5–4.8 $`\mu `$m, while band SL covers the range 5.8–11.8 $`\mu `$m Laureijs et al. (1998). PHT-S measurements were made in chopped mode, using either rectangular or triangular chopping with the ISOPHOT focal-plane chopper; the mode depending on the difficulties of avoiding any nearby sources. During the chop cycle, 512 seconds were spent on-target and 512 seconds off-target – either in a single off-target pointing, or two 256 second off-target pointings. Chopper throw ranged from about 1–3′ depending on the extent of the target source and the density of surrounding field sources. The calibration of the spectrum was performed using a spectral response function derived from several calibration stars of different brightnesses observed in a mode similar to that of the observation of the target Acosta-Pulido (1999). The relative photometric uncertainty of the PHT-S spectrum is better than 20% when comparing different parts of the spectrum that are more than a few microns apart. The absolute photometric uncertainty is better than 30% for bright calibration sources Schulz (1999). All data processing was performed using the ISOPHOT Interactive Analysis (PIA V7.22) system Gabriel (1998). Data reduction consisted primarily of the removal of instrumental effects such as cosmic ray glitches. After background subtraction was performed, flux densities for the sources were determined. In order to increase the signal-to-noise ratio per channel the spectra were smoothed using:
$$as_n=(0.25\times a_{n1})+(0.5\times a_n)+(0.25\times a_{n+1})$$
(1)
where $`a_n`$ is the flux in channel n and $`as_n`$ is the smoothed flux in channel n. Using this method, the flux is effectively spread over 2 channels in a way that does not change the position of the spectral peaks and which conserves the flux. These fluxes were then corrected for redshift to obtain rest-frame spectra. The PHT-S band fluxes were derived from these spectra. An estimate of the continuum was made following the methof of Lutz et al. Lutz et al. (1998). A linear interpolation between the fluxes at 5.9 and 10.9 $`\mu `$m was used except for the sources IRAS 01475-0740, IRAS 08007-6600 and IRAS 23446+1519. For these source, a linear interpolation between the minimum flux values near 6 $`\mu `$m and 10.5 $`\mu `$m was used instead to estimate the continuum. Flux errors were determined by adding in quadrature the 1 $`\sigma `$ errors of all the bins in the feature. Upper limits were derived at the 3 $`\sigma `$ level of significance.
## 3 Results and Discussion
Spectra of eight SMP galaxies are presented in Figs. 1 and 2. The results of these observations are summarised in Tables 2 and 3. Table 2 lists ISOPHOT and IRAS fluxes, ratios of the IRAS 60 $`\mu `$m to the PHT-SL flux, the equivalent width (EW) of the 7.7 $`\mu `$m PAH feature and the PAH to continuum ratios at 7.7 $`\mu `$m (PAH L/C), as well as the optical spectral classification of the galaxies. It is clear from Table 2 that the ratio of IRAS 60 $`\mu `$m flux to the PHT-SL flux is generally higher in the H2 galaxies than it is in the Seyferts. The EW is more uncertain in galaxies with broad PAH features where the spectrum is heavily absorbed by silicate (e.g. Fig. 2d). The 7.7 $`\mu `$m PAH EWs are generally larger in the H2 galaxies. The ratio of the height of the 7.7 $`\mu `$m feature above the continuum to the continuum level at 7.7 $`\mu `$m is referred to as PAH L/C (Table 2). The continuum was obtained from a linear interpolation between points near 5.9 $`\mu `$m and 10.9 $`\mu `$m as described in the previous section. This ratio is a measure of the relative importance of PAH in the total emission from the galaxy, and has been shown to be a good discriminator between starburst and AGN in a sample of Ultraluminous Infrared Galaxies (ULIRGs) Lutz et al. (1998); Genzel et al. (1998). Applying this ratio to SMPs shows that it can discriminate between H2 and Seyfert galaxies (Table 2).
Table 3 lists the line fluxes, and an estimate of the continuum at the centre of the lines. The luminosities of the 7.7 $`\mu `$m PAH feature are also listed in Table 3. The 7.7 $`\mu `$m luminosity gives a better estimate of the amount of PAH in the galaxy than the other PAH features between 5 and 11 $`\mu `$m, which are more prone to dust-extinction Rigopoulou et al. (1999). The luminosity of this feature appears to be independent of spectral type (Table 3 and Clavel et al. 1999).
The emission feature at $``$6.7 $`\mu `$m (Table 3) exists in the spectra of two H2 and one Sy2 galaxy. It is labelled ‘PAH?’ in Figs. 1 and 2 and is identified with the weak PAH feature at 6.66 $`\mu `$m Peeters et al. (1999).
Emission features between 9 $`\mu `$m and 11 $`\mu `$m exist in the spectra of 3 H2 galaxies; IRAS 00160-0719, IRAS 02530+0211 and IRAS 23446+1519 (Figs. 1b,c and d) and in one Sy2 IRAS 01475-0740 (Fig. 2a) and may be attributable to PAH. Another possiblility however, is that these narrow peaks may be emission features from crystalline silicates (Watson et al. in preparation). Crystalline silicates have already been detected in solar system comets Hanner et al. (1994); Crovisier et al. (1997), interplanetary dust particles Bradley et al. (1998), the disks surrounding young stars Malfait et al. (1998); Waelkens et al. (1998) and in the outflows of evolved stars Waters et al. (1998). Crystalline silicate can be produced in a condensation sequence such as that around oxygen-rich AGB stars or by annealing amorphous silicate at temperatures near 1000 K Molster et al. (1999) or perhaps by a low-temperature annealing process Molster et al. (1999). For spectra with peaks near 9.1 $`\mu `$m, assuming they are produced by hypersthene at 300 K, the mass of crystalline silicate required to produce this emission is $``$1 M in IRAS 00160-0719 and $``$1.6 M in IRAS 01475-0740. Spectral observations at longer wavelengths should also reveal features due to crystalline silicate emission, further assisting identification of these minerals. Observations of these longer wavelength features are important in studies of stellar sources Waters et al. (1998).
Though the PHT-SL band extends to 11.8 $`\mu `$m, the PAH emission at $``$11.3 $`\mu `$m is difficult to detect in these spectra due to the redshift of the galaxies. It can be difficult to distinguish absorption at $``$9.7 $`\mu `$m due to silicate, from an artefact caused by broad PAH emission at $``$8 $`\mu `$m and 11.3 $`\mu `$m Roche (1989). Silicate absorption is probably present in the spectra of IRAS 04385-0828 and IRAS 03344-2103 (Figs. 2b and 2d respectively). The shape of the trough at 9.7 $`\mu `$m, the clearly rising continuum in the PHT-SS band, and the IRAS 12 $`\mu `$m fluxes all provide strong evidence that the absorption features in these two spectra are not artefacts. Noise dominates PHT-SS spectra with fluxes less than $`0.05`$ Jy. It is clear however that there is a strong signal in the long wavelength end of the PHT-SS spectra of IRAS 04385-0828 and IRAS 03344-2103.
### 3.1 H2 Galaxies
The spectra of the four H2 galaxies (Fig. 1) are all dominated by emission features associated with PAHs between 6 and 9 $`\mu `$m, though IRAS 08007-6600 (Fig. 1c) clearly has a strong MIR continuum. The PAH emission in IRAS 02530+0211 appears to be a blend of the standard 7.7 and 8.6 $`\mu `$m features in one broad peak, but absorption by silicate cannot be ruled out in this galaxy. IRAS 23446+1519 has an unusual spectrum with strong line emission near 11.0 $`\mu `$m (Fig. 1d). This line has been detected in H II regions (at 11.04 $`\mu `$m) and attributed to PAHs. Roelfsema et al. Roelfsema et al. (1996) have interpreted the existence of this feature in conjunction with an unusually strong 8.6 $`\mu `$m band (comparable to the 7.7 $`\mu `$m feature) and a 7.8 $`\mu `$m shoulder on the 7.7 $`\mu `$m feature as being indicative of the presence of non-compact PAHs. All these characteristics are present in the spectrum of IRAS 23446+1519. In older systems with PAH emission, the less stable non-compact PAHs will generally have been destroyed, but in a non-equilibrium situation, the non-compact PAHs can exist and modify the observed spectrum. Non-compact PAHs are probably present in IRAS 23446+1519. It is interesting to note that a very similar spectrum has been observed from the Wolf-Rayet (WR) galaxy NGC~1741 (McBreen et al. in preparation).
The broad 9.7 $`\mu `$m silicate absorption feature is more common in AGN than in H2 type galaxies Roche (1989). Evidence of this feature is present in the spectra of two galaxies, neither of which is H2 type. \[S IV\] emission at 10.5 $`\mu `$m was detected in three of the four H2 galaxies, but in only one Sy2 (see Table 3). Photons of a few eV are sufficient to generate standard PAH emission spectra at wavelengths shorter than 9 $`\mu `$m Uchida et al. (1998), but this is not the case with other lines emitted in the MIR. The presence of \[S IV\] emission implies a flux of hard photons in the region in which it is produced, assuming photo-ionisation. The more luminous MIR continuum in Seyferts would tend to diminish the signal-to-noise ratio of the \[S IV\] line and may explain the greater prevalence of this line in the H2 galaxies. WR stars are energetic enough to produce \[S IV\] emission and it should be noted that this line has already been observed in the WR galaxies Haro~3 Metcalfe et al. (1996), NGC~7714 O’Halloran et al. (1999) and NGC~5253 Crowther et al. (1999). Some SMPs are also classified as WR galaxies. They are NGC~5253, II Zw 40, Mrk 1210, Tol 1924-416 and possibly Tol 1345-419 Schaerer et al. (1999). A search for the optical signature of WRs in other SMPs could reveal new WR galaxies.
### 3.2 Sy2s
In general the three Sy2 galaxies (Figs. 2a, b and c) have similar 7.7 $`\mu `$m PAH luminosities (Table 3), but smaller 7.7 $`\mu `$m PAH EW and L/C values than the four H2 galaxies (Table 2). It is clear from Fig. 2 that the Sy2s have significant continuum emission thus explaining their low PAH EW and L/C values. Continuum emission from a central source above that detected in H2 galaxies is necessary to explain the NIR brightness profiles of Sy2 SMPs Heisler & Vader (1995). This implies that some of the brighter continuum observed in these Sy2 spectra could be, directly or indirectly due to the Seyfert nucleus. The spectrum of IRAS 03344-2103 (Fig. 2d) has a very large silicate absorption trough near 9.7 $`\mu `$m making it difficult to identify lines beyond $``$8 µm and making the EW of 7.7 $`\mu `$m feature difficult to determine.
There are clear discriminators between H2 and Sy2 galaxies in the mid-infrared. They are PAH L/C, EW and the IRAS 60 $`\mu `$m to the PHT-SL flux ratio (see Table 2 and Fig. 3) and are discussed in the next section.
## 4 Separation of Starbursts and AGN
In the standard model, Sy1s and Sy2s differ in their orientation to the line of sight Urry & Padovani (1995). The Seyfert nucleus is surrounded by a dusty torus that obscures the broad emission-line region (BLR) in Sy2s (which are viewed edge-on).
PAH emission is produced outside the torus and is independent of the nucleus Clavel et al. (1999). In the well-studied Sy2 Circinus galaxy, Moorwood et al. Moorwood (1999) manage explicitly to exclude the nucleus as the source of excitation for the PAH emission using ISOCAM-CVF. Clavel et al. Clavel et al. (1999) use the EW of the 7.7 $`\mu `$m feature as a discriminator between Sy1 and Sy2. Laurent et al. Laurent et al. (1999) have proposed a similar diagnostic, using the ratio of the PAH 6.2 $`\mu `$m band to the 5.1 to 6.7 $`\mu `$m continuum to distinguish AGN from starburst galaxies. These are in essence a similar measure to the PAH L/C at 7.7 $`\mu `$m since all depend on the facts that (a) PAH luminosities are independent of the active nucleus Clavel et al. (1999) and (b) the MIR continuum is stronger in Sy1 than Sy2 and brighter in AGN as a whole than in starburst galaxies. Laurent et al. Laurent et al. (1999) extend this diagnostic tool to assess the relative contributions of components from AGN, H II and photo-dissociation regions by introducing the ratio of the warm to the hot continuum. This ratio was not obtainable in the PHT-S waveband because it does not extend to 15 $`\mu `$m.
In Sy1s the BLR and the inner wall of the torus are directly visible, making the MIR continuum brighter in these galaxies. In Sy2s seen fully edge-on, our line of sight to the inner wall of the torus is blocked and the MIR continuum is suppressed Clavel et al. (1999). In the intermediate situation of Sy2 viewed at grazing incidence, one has a direct view of the inner wall of the torus but a reflected view of the BLR. This corresponds to Sy2s where the MIR continuum is strong and broad lines are observed only in polarised light Clavel et al. (1999); Heisler et al. (1996, 1997). IRAS 05189-2524 falls into this category, since it it has polarised broad-lines Young et al. (1996) and a strong MIR continuum (Fig. 2c). The 7.7 $`\mu `$m PAH EW of IRAS 05189-2524 is 0.3 $`\mu `$m falling in the range observed by Clavel et al. Clavel et al. (1999) for this sub-class of Sy2. IRAS 04385-0828 has a 7.7 $`\mu `$m EW of just 0.1 $`\mu `$m (Table 2) and is therefore another good candidate for possessing polarised broad lines. Unfortunately, though it was observed in polarised light, the signal-to-noise ratio was not sufficiently good to determine the EW of the lines in the polarised spectrum Young et al. (1996).
Dudley Dudley (1999) finds the 1–5 $`\mu `$m SED of IRAS 05189-2524 to be similar to that of the infrared-bright quasar IRAS 13349+2438 Beichman et al. (1986). He therefore proposes that hot dust heated by an AGN is responsible for the 1–5 $`\mu `$m continuum in both sources. It is clear however that two components are required to fit the 1–100 $`\mu `$m SED in IRAS 05189-2524. Dudley Dudley (1999) therefore further suggests that the 8–100 $`\mu `$m emission is better explained by emission arising from star formation than from an AGN. The observation of IRAS 05189-2524 presented here shows no evidence of a spectral break between 2.5 $`\mu `$m and 11 $`\mu `$m. The very large IRAS 12 $`\mu `$m flux for this source indicates that such a spectral break could lie near 12 $`\mu `$m (Fig. 2c). It is clear from these results that IRAS 05189-2524 is a Seyfert and that the hot inner wall of the torus is largely responsible for the 2.5 $`\mu `$m–11 $`\mu `$m continuum. Recent PHT-S observations of IRAS 13349+2438 Watson et al. (1999) while showing PAH emission, also suggest that the inner wall of the torus is the source of most of the 2.5 $`\mu `$m–11 $`\mu `$m continuum in this quasar. The similarity in the spectra of IRAS 05189-2524 and IRAS 13349+2438 may be attributed to the suggestion that they are both AGN viewed near grazing incidence to their tori Clavel et al. (1999); Wills et al. (1992).
The ratio of the PHT-SL 5.9 $`\mu `$m flux to the IRAS 60 $`\mu `$m flux is plotted against the PAH L/C values (Table 2) in Fig 3. There is an anti-correlation between the PAH L/C and the F<sub>ν</sub>(5.9 $`\mu `$m)/F$`{}_{\nu }{}^{}(60`$$`\mu `$m) ratio indicating that warmer SMPs are more AGN-dominated, and cooler SMPs are more starburst-like. The anti-correlation is significant at a probability greater than 99% (from a Pearson’s correlation statistic of -3.7). The eight SMPs divide very well at PAH L/C = 0.8 (Fig. 3). The method Lutz et al. Lutz et al. (1998) applied to a sample of ULIRGs is followed in Fig 3. The same anti-correlation was found for ULIRGs as is discovered here. Lutz et al. Lutz et al. (1998) used the criterion PAH L/C = 1 to separate AGN from starburst ULIRGs, but found this criterion tended to classify some starbursts as AGN.
IRAS 08007-6600, appears to be currently undergoing a merger and possesses two distinct nuclei Heisler & Vader (1994). It is possible that one of the nuclei may contain a hidden non-thermal source given its position on the Starburst-AGN plot (Fig. 3) with PAH L/C = 0.9, despite its optical classification as a H2 galaxy. The unclassified galaxy, IRAS 03344-2103 may also be a Seyfert since it shows evidence of silicate absorption in its spectrum (Fig. 2d) and has a PAH L/C = 0.8.
## 5 Conclusions
ISOPHOT-S spectra were obtained for a sample of eight SMPs of types H2 and Sy2. PAH emission was detected in all the spectra. The spectrum of IRAS 23446+1519 shows an unusual 8.6 $`\mu `$m feature with a height comparable to the 7.7 $`\mu `$m feature and exhibits a very bright 11.04 $`\mu `$m PAH emission line. The spectrum implies that non-compact PAHs are the source of a large proportion of the PAH emission in that galaxy. \[S IV\] emission is more prevalent in the H2 galaxies than in the Sy2s, probably because of the brighter continuum in Sy2s. Silicate absorption ($``$9.7 $`\mu `$m) was observed in IRAS 03344-2103 and in IRAS 04385-0828. The results show that the H2 galaxies have PAH L/C at 7.7 $`\mu `$m $`>0.8`$ and the Sy2 galaxies have PAH L/C $`0.8`$. The same anti-correlation exists in SMPs and ULIRGs between the PAH L/C and the F<sub>ν</sub>(5.9 $`\mu `$m)/F$`{}_{\nu }{}^{}(60`$$`\mu `$m) ratio. It is proposed that IRAS 08007-6600 may contain a hidden non-thermal source and that the previously unclassified galaxy IRAS 03344-2103 may be a Seyfert. Observations of IRAS 04385-0828 may reveal broad emission lines in polarised light.
While the galaxies presented here are diverse, observation of a larger sample of SMPs in the mid-infrared could confirm the Seyfert/H2 separation and the anti-correlation observed here. The prevalence of \[S IV\] emission in H2 SMPs implies that a search of SMP optical spectra for the WR signature may reveal new WR galaxies.
|
warning/0005/nlin0005017.html
|
ar5iv
|
text
|
# Characterization of the Emergence of Order in an Oscillated Granular Layer
## Abstract
The formation of textured patterns has been predicted to occur in two stages. The first is an early time, domain-forming stage with dynamics characterized by a disorder function $`\overline{\delta }(\beta )t^{\sigma _E}`$, with $`\sigma _E=\frac{1}{2}\beta `$; this decay is universal. Coarsening of domains occurs in the second stage, in which $`\overline{\delta }(\beta )t^{\sigma _L}`$, where $`\sigma _L`$ is a nonlinear function of $`\beta `$ whose form is system and model dependent. Our experiments on a vertically oscillated granular layer are in accord with theory, yielding $`\sigma _E0.5\beta `$, and $`\sigma _L`$ a nonlinear function of $`\beta `$.
The formation of textured spatial patterns has been studied in laboratory experiments and model systems . Several quantities have been used to describe the development of patterns represented by a scalar field $`v(𝐱)`$ with a typical wavevector $`k_0`$, including the structure factor , density of topological defects , and a recently introduced family of characterizations called the disorder function,
$$\overline{\delta }(\beta )=\frac{(2\beta )}{d^2x}\frac{d^2x|(\mathrm{}+k_0^2)v(𝐱)|^\beta }{k_0^{2\beta }|v(𝐱)|^\beta },$$
(1)
where $`|v(𝐱)|`$ denotes the mean of $`|v(𝐱)|`$, and $`\overline{\delta }(\beta )`$ ( $`0\beta <2`$) has been normalized to be scale invariant . $`\overline{\delta }(\beta )`$ describes configuration-independent aspects of textures and their formation, aspects that are independent under repetition of the experiment. It can be used to study multiple aspects of patterns just as generalized dimensions and singularity spectra can be used to describe multiple aspects of strange attractors. In experiments presented in this paper, we characterize the time evolution of patterns in a vertically oscillated granular layer using $`\overline{\delta }(\beta )`$.
Before describing our experiments, we review numerical and analytical work on the coarsening of patterns after a quench from an initial featureless (or noisy) state. Most studies have focused on solutions $`u(𝐫,t)`$ to the Swift-Hohenberg equation ,
$$\frac{u}{t}=\left(ϵ(\mathrm{}+k_0^2)^2\right)uu^3\nu (u)^2+\eta (𝐫,t),$$
(2)
where $`u(𝐫,t)`$ is a two-dimensional scalar field, $`ϵ`$ is the distance from pattern onset, $`\nu `$ is the strength of a non-variational term , and $`\eta `$ a random forcing term such that $`\eta (𝐫,t)\eta (𝐫^{},t^{})=2F\delta (𝐫𝐫^{})\delta (tt^{})`$, where $`F`$ controls the strength of the noise.
For the formation of textures in (2), the width of the structure factor $`S(t)`$ (i.e., the width of the peak in the azimuthal average of $`\stackrel{~}{u}(𝐤,t)\stackrel{~}{u}(𝐤,t))`$, has been shown to decay in two distinct stages . $`S(t)t^{\frac{1}{2}}`$ is obeyed until the peak amplitude of the field $`u(𝐱,t)`$ saturates, beyond which time the pattern coarsens and the decay becomes slower. For $`ϵ=0.25`$ and $`\nu =0`$, Cross and Meiron , Elder et al. , and Hou et al. found that in this second region $`S(t)`$ decreased as $`t^{\frac{1}{5}}`$ when $`F=0`$, and as $`t^{\frac{1}{4}}`$ when $`F0`$ . Schober et al. found that for $`F=0`$ and $`\nu =0`$, $`S(t)t^{\frac{1}{4}}`$; the discrepancy with earlier results could be due to the one-dimensionality of their model.
The characterization of pattern formation in model systems using $`\overline{\delta }(\beta )`$ has also shown the presence of two stages: the emergence of domains characterized by $`\overline{\delta }(\beta )t^{\sigma _E(\beta )}`$ with $`\sigma _E(\beta )=\frac{1}{2}\beta `$, and a slower coarsening behavior . The relaxation exponent during coarsening depends on the value of $`\nu `$ in (2), and is thus expected to be system and model dependent . We report analogous behavior in the present laboratory study of pattern formation in an oscillated granular layer and present additional differences between the two stages.
Our experiments generate patterns in a layer of 0.165 mm bronze spheres contained in a vertically oscillated circular container with diameter 14 cm . The layer is four particle diameters deep, and the cell is evacuated to 4 Pa to avoid any hydrodynamic interaction between the grains and surrounding gas. The control parameters are the frequency $`f`$ of the sinusoidal oscillations and the peak acceleration of the container, $`\mathrm{\Gamma }=(2\pi f)^2A^2/g`$, where $`A`$ is the maximum amplitude of the oscillation and $`g`$ is the gravitational acceleration. As $`f`$ and $`\mathrm{\Gamma }`$ are varied, a variety of textures including striped, square or hexagonal planforms are observed . Our analysis reported here is restricted to patterns with square planforms. In the region of the phase diagram studied, square patterns appear for increasing control parameter at $`\mathrm{\Gamma }2.75`$; the bifurcation is subcritical. The geometry of the circular container allows relaxation to an almost perfect square array through wavelength adjustment of the pattern at the container wall over a distance of less than one wavelength .
The granular surface is illuminated with a ring of LEDs surrounding the cell. The light is incident at low angles and the scattering intensity is a nonlinear function of the height of the layer; scattering from peaks (valleys) creates bright (dark) regions. The images are collected at the driving frequency and the acceleration of the container is monitored during each run. $`\mathrm{\Gamma }`$ is suddenly increased from its initial value of 2.2, where no discernible structure is observed. As the grains are not in contact with the container for part of their motion, we assume that the initiation of the quench occurs at the first layer-plate collision after the change of control parameter. The uncertainty in the time origin is the dominant source of error in our measurements.
The top row of Fig. 1 shows that local square domains emerge and coarsen to a final, almost perfect, square array. The bottom row shows this process in Fourier space. A repetition of the experiment would lead to similar, but not identical, intermediate states. Our aim is to study configuration-independent aspects of this relaxation, and to analyze their dependence on the control parameters $`f`$ and $`\mathrm{\Gamma }`$.
Patterns such as those shown in the top row of Fig. 1 can be represented by a discrete sampling of a smooth scalar field $`v(𝐱)`$. The values of the field are known on a (typically square) grid, but the analytical form of the field is unknown. The ingredients used to deduce the form of the disorder function in (1) are its invariance under arbitrary rigid motions of the texture and the nature of the local planform. Local deviations of a pattern from squares (due to curvature of the contour lines ) contribute to $`\overline{\delta }(\beta )`$ through the Laplacian, while variations of the size of squares contribute via the choice of a “global” $`k_0`$, which is obtained from the field $`v(𝐱)`$ by minimizing the value of $`\overline{\delta }(1)`$ . Unlike the information contained in the structure factor, $`|(\mathrm{}+k_0^2)v(𝐱)|`$ is a local density of irregularities in the texture, and hence distinct “moments” $`\beta `$ can be used to quantify multiple aspects of the disorder density.
The images shown in Fig. 1 have sharp changes at the edges which lead to high frequency contributions in their Fourier spectra. Their removal through simple filtering causes contamination of the pattern near the edges and leads to error in calculating $`\overline{\delta }(\beta )`$. We use a method of noise filtering that involves extending the image to a periodic one using “Distributed Approximating Functionals” (DAFs) . Fourier filtering can then be used on the extended image to eliminate high frequency noise and undesirable harmonics . The evaluation of $`\overline{\delta }(\beta )`$ requires an accurate estimation of $`\mathrm{}v(𝐱)`$, which typically amplifies any noise present in the discrete, digital experimental data. A method for this calculation has been presented in .
The behavior of $`\overline{\delta }(1)`$ for the relaxation of Fig. 1 is shown by the symbols $``$ in Fig. 2. The initial formation of local rectangular domains and the final coarsening correspond to distinct power law decays of $`\overline{\delta }(1)`$. The transition coincides with the saturation of the peak amplitude; i.e., nonlinear effects are negligible during domain formation (typically 4-5 oscillations) and become relevant during coarsening .
During the initial stage of pattern formation $`\overline{\delta }(1)t^{0.49\pm 0.02}`$. This dynamical scaling is similar to that describing the decay of the width of the structure factor and is related to the rate of domain growth in phase ordering kinetics .
Since nonlinear effects are negligible during domain formation, the evolution can be modeled by (2) with the removal of the nonlinear and stochastic terms. Numerical integration starting from states consisting of random or Gaussian noise shows that $`d^2x|u(𝐫,t)|e^{ϵt}t^{\frac{1}{4}}`$ and $`d^2x|(\mathrm{}+k_0^2)u(𝐫,t)|e^{ϵt}t^{\frac{3}{4}}`$; consequently $`\overline{\delta }(1)t^{\frac{1}{2}}`$. It has been shown analytically that the width of the structure factor for evolution of $`u(𝐫,t)`$ decays like $`S(t)t^{\frac{1}{2}}`$ until the peak amplitude saturates , providing further evidence for the interpretation that the spatiotemporal dynamics is linear during the first stage.
Furthermore, during the domain forming stage, moments of the disorder function decay as $`\overline{\delta }(\beta )t^{\sigma _E(\beta )}`$, where $`\sigma _E(\beta )\frac{1}{2}\beta `$ (see Fig. 3(a)). Analogous behavior can also be seen by numerical integration of the linear terms in (2). The linearity of $`\sigma _E(\beta )`$ and the value of the proportionality constant suggest that multiple aspects of textures, such as structure factor, curvature of contours and defect densities decay via a single mechanism during the domain forming stage.
The initiation of domain coarsening coincides with the saturation of peak heights of the granular layer (see Fig. 2); i.e., the latter stages of pattern formation correspond to nonlinear spatiotemporal dynamics of the field . The observed scaling of the disorder function is more complex. For the evolution shown in Fig. 2 at $`\mathrm{\Gamma }=2.8`$, $`\overline{\delta }(1)t^{0.18}`$ . This is close, but not identical to the rate of relaxation of the structure factor . Even though the moments $`\overline{\delta }(\beta )`$ decay (approximately) as power laws $`t^{\sigma _L(\beta )}`$, the exponent is not linearly related to $`\beta `$ as during the early phase, see Fig. 3(b). There is a tendency towards slower relaxation in Fig. 3(b) (compared to Fig. 3(a)) for larger values of $`\beta `$. Since large values of $`\beta `$ preferentially weight pattern defects, this is consistent with the slower relaxation of the density of defects than that of the structure factor .
The nonlinearity of $`\sigma _L(\beta )`$ implies that pattern relaxation occurs on more than one length scale. If the “envelope” of the texture depends on a single length scale $`L(t)`$, the field can be locally expanded as $`u(𝐱,t)=Re\left[e^{i𝐤𝐱}A(𝐗,t)\right]`$, where $`𝐗=𝐱/L(t)`$; thus
$`(\mathrm{}+k_0^2)u(𝐱,t)=Re\left[e^{i𝐤𝐱}\left(\frac{2i}{L}𝐤_𝐗A\frac{1}{L^2}\mathrm{}_XA\right)\right]`$.
Since $`L(t)k_0^1`$ during domain coarsening, the last term can be neglected, leading to
$`|(\mathrm{}+k_0^2)u(𝐱,t)|\frac{2}{L(t)}𝐤A\frac{|u|}{L(t)}`$.
Consequently $`\overline{\delta }(\beta )1/L^\beta (t)\left(\overline{\delta }(1)\right)^\beta `$. Thus the nonlinearity of $`\sigma _L(\beta )`$ implies that the relaxation during coarsening occurs on multiple time scales.
Next, we briefly consider changes in the behavior of the disorder function as the experimental system is driven further away from the onset of patterns. The decay of $`\overline{\delta }(\beta )`$ during domain formation remains unchanged, but the decay rate in the second region decreases with increasing $`\mathrm{\Gamma }`$. Similar behavior has been observed with the addition of $`\nu `$ in (2.
We have shown that the formation of texture in a vertically oscillated granular layer occurs in two distinct stages. During the initial stage the spatiotemporal dynamics is essentially linear and the disorder function obeys a universal power law $`\overline{\delta }(\beta )t^{\sigma _E(\beta )}`$, with $`\sigma _E(\beta )\frac{1}{2}\beta `$; this simple behavior is also observed in the linearized Swift-Hohenberg equation. Nonlinearity of spatiotemporal dynamics becomes relevant during domain coarsening and $`\overline{\delta }(\beta )t^{\sigma _L(\beta )}`$ where $`\sigma _L(\beta )`$ is a nonlinear function of $`\beta `$. The exponent $`\sigma _L(\beta )`$ is model and parameter dependent . Such non-universal, configuration independent characteristics of pattern formation can be used to determine the validity and limitations of model systems .
Although dynamical scaling has been reported in the formation of patterns in model systems , ours is the first reported observation of stages exhibiting trivial (i.e., a single scaling index such that $`\overline{\delta }(\beta )\left(\overline{\delta }(1)\right)^\beta `$) and nontrivial scaling during pattern formation in an experimental system. The methods introduced here are expected to have applications in studying other aspects of textures, such as the quantitative description of patterns in magnetic bubble material .
We have benefited from discussions with M. Golubitsky, J. B. Swift and P. Umbanhower. The research at the University of Texas was supported by the Engineering Research Program of the Office of Basic Energy Sciences of the U.S. Department of Energy. Additional support came from the Office of Naval Research (GHG), the Ames Laboratory of the Department of Energy (DKH), the National Science Foundation (GHG, DJK), and the R. A. Welch Foundation (DJK).
|
warning/0005/physics0005036.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
The study of transition amplitudes for molecules with initial and final states given by different electronic configurations has held great interest. In the Franck-Condon limit one models the two configurations by similar effective potentials with different geometric parameters. The transition intensities are then given by the wavefunction overlap of the two potentials (sudden approximation) \[References\].
Recently, the Franck-Condon (FC) problem of polyatomic molecules has been made tractable by the introduction of a hybrid algebraic-Schrödinger approach \[ReferencesReferences\]. In this approach a spectrum generating algebra of $`U_1(2)\times \mathrm{}\times U_k(2)`$ (where there are $`k`$ bonds) is used to obtain wavefunctions \[References\]
$$|\psi =\underset{i_1,\mathrm{},i_k}{}c_{i_1,\mathrm{},i_k}|[N_1],i_1\mathrm{}|[N_k],i_k,$$
(1.1)
by fitting spectra of a molecule in a particular configuration. In general for the two different configurations the $`i`$th bond contribution could change from the representation $`[N_i]`$ to $`[N_i^{}]`$. The Franck-Condon factors for the polyatomic case are realized in terms of the single bond factors by calculating the matrix element $`\psi ^{}|\widehat{U}|\psi `$ \[References,References\], where
$$\widehat{U}=\widehat{t}(\frac{\alpha _1}{\alpha _1^{}},\mathrm{\Delta }_1)\mathrm{}\widehat{t}(\frac{\alpha _k}{\alpha _k^{}},\mathrm{\Delta }_k)$$
(1.2)
and the operator $`\widehat{t}`$ is defined by its matrix elements as determined by Schrödinger overlaps:
$$[N^{}],n^{}|\widehat{t}(\frac{\overline{\alpha }}{\overline{\alpha }^{}},\mathrm{\Delta })|[N],n=𝑑x\psi _n^{}^N^{}(\overline{\alpha }^{};x)\psi _n^N(\overline{\alpha };x\mathrm{\Delta }).$$
(1.3)
The parameters $`\frac{\alpha }{\alpha ^{}}`$ and $`\mathrm{\Delta }`$ are typically fit from transition spectra. Due to the $`U(2)O(2)`$ chain’s correspondence with a Morse oscillator \[References\] one often uses Morse eigenfunctions with anharmonicity parameter $`\frac{1}{N+1}`$ for Schrödinger wavefunctions. It is worth noting that the heuristic approximation in \[References\] showed that this can be equivalently thought of as simple harmonic oscillator (SHO) overlaps where the scale parameters acquire a linear $`n`$ dependent correction with coefficient $`O(1/N)`$. That is, one would use the equation for the overlap of SHO wavefunctions, which has no $`N`$ dependence, and substitute a scale $`\alpha =\alpha _0(1\xi (n+\frac{1}{2}))`$. $`\alpha _0`$ was taken as the scale of the SHO best approximating the Morse and $`\xi `$ is $`O(1/N)`$. In this way one introduces geometric parameters not contained in the algebra.
Although this approach has been successful it necessitates that the initial and final configurations not be too dissimilar—in particular they must have the same normal mode expansion. This precludes, for example, a transition between a linear and bent triatomic molecule (Figure 1). In the linear configuration the center atom has two normal modes. In the bent configuration one of the normal mode solutions is spurious—corresponding to an overall rotational degree of freedom. Because of this, the treatment of vibrational modes in each configuration necessitates the freezing of a different number of degrees of freedom, whereas when studying transitions between each configuration one needs to have the same degrees active.
One may attack this problem by using more complicated algebraic models which have both geometries built in. In this way some parameters which before were artificially inserted through the hybrid method become natural—i.e. since they are built in the algebra they are determined by spectra and do not need to be fit with transition data.
However, in these more complicated algebraic models, exact Schrödinger correspondences don’t in general exist. Thus, this approach necessitates an alternate interpretation of algebraic parameters as geometric, configuration space quantities. Recently such an interpretation has been provided \[References\] by exploiting the many approximate correspondences between algebraic and Schrödinger pictures. We use these results to develop a hybrid approach for the bent to linear transitions in Figure 1. This is most easily done by using a spectrum generating algebra of $`U_1(2)\times U(3)\times U_2(2)`$ (a $`U(2)`$ for each bond and the $`U(3)`$ for the two additional degrees of freedom of the center atom). The FC transition operator would then be $`\widehat{U}=\widehat{t}_1\widehat{T}\widehat{t}_2`$, where $`\widehat{t}_1`$, $`\widehat{t}_2`$ are defined by equation 1.3. It is the purpose of this publication to motivate the definition of $`\widehat{T}`$ (eqn. 3.5).
We begin by demonstrating, via a coherent state limit analysis \[References\], that $`U(3)`$ is an appropriate algebra to describe the degrees of freedom of the center atom of the triatomic. We proceed with numerical studies to study the implications and discrepancies of this interpertation. Finally, we reconcile these discrepancies using \[References\] which naturally motivates our definition of $`\widehat{T}`$.
## 2 The Franck-Condon Problem and $`U(3)`$
### 2.1 Overview of $`U(3)`$ Statements on 2D Problem
Th algebraic approach for 2D problems was presented by \[References\]. One considers symmetric (bosonic) representations of $`u(3)`$. There are two chains of interest:
$`\begin{array}{cc}U(3)U(2)O(2)\hfill & \mathrm{I}\hfill \\ U(3)O(3)O(2)\hfill & \mathrm{II}\hfill \end{array}.`$ (2.3)
We use the same notation as \[References\] for the generators except chose a different $`O(3)`$ subgroup as explained in \[References\]. Please note that the $`O(3)`$ group is a dynamical symmetry subgroup and does not have the interpretation of a rotation in configuration space.
The general Hamiltonian of the $`U(3)`$ model is
$$H=ϵ\widehat{n}+\delta \widehat{n}(\widehat{n}+1)+\beta \widehat{l}^2A\widehat{W}^2,$$
(2.4)
where $`ϵ`$, $`\delta `$, and $`A`$ are taken as positive or $`0`$. Setting $`A=0`$ ($`ϵ=\delta =0`$) gives a Hamiltonian with dynamical symmetry I (II). The spectra for each dynamic symmetry may be determined from the well known solution to the branching problem for a symmetric representation of $`u(3)`$ labelled by $`N`$ (the eigenvalue of the $`u(3)`$ Casimir $`\widehat{n}+\widehat{n}_s`$) \[References\]. The basis corresponding to chain I is labeled by the eigenvalues of the $`u(2)`$ and $`o(2)`$ Casimirs $`n`$ and $`l`$ respectively. The basis corresponding to chain II is labeled by the $`o(3)`$ Casimir’s eigenvalues, $`\omega (\omega +1)`$, and again by $`l`$.
The spectra of each chain led the authors of \[References\] to the interpretation of each dynamical symmetry as an azimuthally symmetric potential with minimum at 0 radius (chain I) and at non-zero radius (chain II).
### 2.2 $`U(3)`$ Coherent State Limit
The interpretation of \[References\] is reaffirmed by simply studying the classical coherent state limit calculated in \[References\].
Taking the coherent state limit of the Hamiltonian 2.4, setting all momenta to zero and dropping additive constants one finds the potential in group coordinates \[References\] to be (up to a meaningless multiplicative factor):
$$\stackrel{~}{V}_{\mathrm{cl}}(r)=\eta \frac{1}{2}r^2+\frac{1}{4}r^4$$
(2.5)
where
$$\eta =\frac{ϵ+2\delta +\beta 4A(N1)}{(\delta +4A)(N1)}.$$
(2.6)
It is easy to compute the position of the potential minima
$$r_{\mathrm{min}}=\{\begin{array}{cc}0& \eta 0\\ \sqrt{\eta }& \eta <0\end{array}.$$
(2.7)
Thus we find:
$$\stackrel{~}{V}_{\mathrm{cl}}(r_{\mathrm{min}})=\{\begin{array}{cc}0& \eta 0\\ \frac{1}{4}\eta ^2& \eta <0\end{array}.$$
(2.8)
That is we have a second-order phase transition at $`\eta =0`$, or equivalently $`4A(N1)=ϵ+2\delta +\beta `$ \[References\].
The exact same analysis may be carried out in projective coordinates \[References\] revealing:
$$\stackrel{~}{V}_{\mathrm{cl}^{}}(\stackrel{~}{r}_{\mathrm{min}})=\{\begin{array}{cc}0& \eta 1\\ \frac{1}{4}\frac{(1\eta )^2}{1+\eta ^{}}& \eta <1\end{array},$$
(2.9)
where now
$$\eta =\frac{(ϵ+2\delta +\beta )}{4A(N1)},\eta ^{}=\frac{\delta }{4A},$$
(2.10)
and
$$\stackrel{~}{r}_{\mathrm{min}}^2=\{\begin{array}{cc}0& \eta 1\\ \frac{1\eta }{\eta +2\eta ^{}+1}& \eta <1\end{array}.$$
(2.11)
Thus we again find a second-order phase transition at $`4A(N1)=ϵ+2\delta +\beta `$.
In either coordinates the potential minima moves from $`r=0`$ (corresponding to a linear configuration) to $`r0`$ (corresponding to a bent configuration) at the critical point. We conclude that the algebraic model is rich enough to include both geometries depicted in Figure 1. Algebraic Hamiltonians having nearly a $`U(2)`$ dynamical symmetry correspond to the center atom in a linear triatomic, whereas those near the $`O(3)`$ limit correspond to the center atom in a bent triatomic
### 2.3 $`U(3)`$ and Schrödinger FC Connections
The FC factor for the two configurations is easily studied from a Schrödinger perspective. Assuming that, whatever the actual nature of the potentials, they may be approximated about their minima as harmonic we may numerically calculate the FC overlaps. The results of such a calculation are displayed in Figure 2. The parameters (Table 2) are chosen to be relevant to the bent to linear FC transition ($`{}_{}{}^{1}B_{2}^{}\mathrm{\Sigma }_g^+`$) of $`CS_2`$. The frequency of the linear configuration was taken from reference \[References\]. The harmonic distance scale, $`\frac{m\omega }{\mathrm{}}`$, was deduced by assuming the effective mass was that of the carbon nucleus. The radial displacement was deduced from the geometry of the bent configuration as published in reference \[References\] assuming that the heavy ($`SS`$) axis was essentially stationary. The associated distance scale of the bent configuration was assumed to be the same as the linear scale. Since $`CS_2`$ is a particularly shallow molecule we have included an additional plot (Figure 3) to demonstrate the behavior for a larger radial displacement.
If one models the potential as exactly harmonic the displaced oscillator potential has the idiosyncrasy of a ‘cusp’ at $`r=0`$. This is of little concern since the effective potential is dominated by the angular momentum barrier at this point. Approximate analytic expressions may be obtained for this limit as detailed in Appendix A.
We wish to emphasize that although these graphs serve as a valid starting point for a more complete analysis of $`CS_2`$ transition intensities their primary purpose here is heuristic. A full analysis requires a careful fitting of the bent configuration distance scale. Additionally, the bending modes considered here are known to couple strongly to symmetric stretching modes \[References\]—i.e. a full analysis would require coupling additional $`U(2)`$’s as discussed in the introduction.
One may consider more realistic bending potentials such as a Pöschl-Teller:
$$V=V_0\left[1\mathrm{cosh}^2\overline{\alpha }(rr^{})\right].$$
(2.12)
In this case the cusp at the origin still exists for the displaced oscillator (non-zero $`r^{}`$) but is tamed due to the long distance flattening of the potential. In the limit where the minima is far from the origin the cusp essentially vanishes. Figure 3 shows numerical results for a Pöschl-Teller model of a bent to linear transition. The parameters are chosen such that the the potentials are approximated to second order by exactly the harmonic plots included in the same figure. That is, the harmonic distance scale, $`\alpha ^4=2mV_0\overline{\alpha }^2/\mathrm{}^2`$ is set to the same value as the SHO FC factors. The remaining parameter, taken as $`\frac{\alpha ^2}{\overline{\alpha }^2}`$, is a unitless measure of well depth. It was chosen to be sufficiently small to emphasize differences between the FC factors of the two potentials.
The previous section’s analysis implies that the FC factors for a bent to linear configuration in the algebraic picture are given by exactly the inner product of the algebraic wavefunctions for hamiltonians near the $`O(3)`$ chain (bent configuration) and on the $`U(2)`$ chain (linear configuration). The overlaps for several such ‘bent’ hamiltonians are given in Figure 4. The algebraic ‘bent’ hamiltonian was taken to be of the form
$$H=(1\xi )\widehat{n}\frac{\xi }{(N1)}\widehat{W}^2.$$
(2.13)
The parameter $`\xi `$ was chosen to match the intensity maximum with that of several harmonic Schrödinger calculations (corresponding to unitless radial displacements of $`3`$, $`5`$, and $`7`$). The results of calculations for two significantly different irreps., $`[N]`$, are shown to emphasize that the structure is generic and not a function of any special choice of parameters.
Comparing Figures 2–4 again reaffirms the interpretation of the two $`U(3)`$ chains as bent and linear configurations of a 2D problem. Comparing the SHO and Pöschl-Teller figures one notes although not in exact agreement they are very similar given the large differences of the Schrödinger potentials. Qualitatively the $`U(3)`$ graphs also appear similar with the possible exceptions of (1) their (expected) truncation at higher $`n`$; (2) their dramatically sharper peaks than the Schrödinger FC graphs; (3) their amplitude’s diminished sensitivity to the amount of radial displacement.
### 2.4 Scale Changes
The numerical study raises two questions (1) What are the relations between the algebraic parameters (determining the hamiltonian’s proximity to either chain) and the Schrödingers (more geometric) parameters? (2) Is there an interpretation for the apparent qualitative differences (the sharper peak) between the algebraic and Schrödinger pictures?
Both these questions can be addressed by considering the results of \[References\]. The two relevant results, which we reproduce here, include the intrinsic distance scale of the harmonic approximation to a Hamiltonian near the $`O(3)`$ limit and an expression for the radial displacement for the same Hamiltonian:
$`\alpha ^22{\displaystyle \frac{\overline{m\omega }_{O(3)}}{\mathrm{}}}\left[1+{\displaystyle \frac{X}{Z}}\left({\displaystyle \frac{l^2}{N^2}}1\right){\displaystyle \frac{Y}{Z}}\left({\displaystyle \frac{l^2}{N^2}}+{\displaystyle \frac{1}{2}}\right)\right].`$ (2.14)
$$(r^{})^2\frac{N\mathrm{}}{\overline{m\omega }_{O(3)}}\left(1+\frac{X}{Z}(\frac{l^2}{(N\mathrm{})^2}1)\frac{Y}{Z}\right),$$
(2.15)
where $`X=ϵ+2\delta +\beta `$, $`Y=\delta (N1)`$, and $`Z=4A(N1)`$, and the condition that we are near the $`O(3)`$ limit implies $`\frac{X}{Z}`$ is small. The parameter $`\overline{m\omega }`$ has the interpretation of the ratio of distance to momenta scales, i.e. $`\overline{m\omega }=\overline{\alpha }(mV_0)^{\frac{1}{2}}`$ for a potential $`V=V_0f(\overline{\alpha }x)`$.
Additionally the results of \[References\] imply that Hamiltonians from either dynamical symmetry not only correspond to different geometries as implied by 2.2 but additionaly to different intrinsic scales (see Appendix C):
$`\zeta _n={\displaystyle \frac{\overline{m\omega }_{U(2)}}{\overline{m\omega }_{O(3)}}}{\displaystyle \frac{2\mathrm{log}2}{2\mathrm{log}2}}\left[1+{\displaystyle \frac{4}{\mathrm{log}2(2\mathrm{log}2)}}\left\{{\displaystyle \frac{3}{2}}{\displaystyle \frac{\mathrm{log}N}{N}}+(n\mathrm{log}\mathrm{log}2+c){\displaystyle \frac{1}{N}}\right\}\right],`$ (2.16)
$`c=\mathrm{log}\mathrm{log}2\mathrm{log}{\displaystyle \frac{2^{\frac{3}{4}}}{2\mathrm{log}2}}`$ (2.17)
This implies that the algebraic overlap of a $`U(2)`$ Hamiltonian with an $`O(3)`$ one is not simply analogous to the overlap of radially displaced oscillators, but analagous to the matrix elements of an operator which radially displaces and dilatates (much like the operator matrix elements calculated in \[References\]) changing the natural scale of the problem. The degree of dilatation depends on the proximity of the second hamiltonian to either chain. For chains near $`U(2)`$ the dilatation paramater is essentially $`1`$. As one moves nearer to the $`O(3)`$ chain the dilatation parameter increases, approaching the value given by equation 2.16. This effect must be accounted for in any algebraic or hybrid approach to the FC problem.
We saw in the introduction that in the 1D problem the scale parameters may have been thought as harmonic scale parameters with corrections linear in the quantum number $`n`$ of order $`1/N`$. The scenario is similar here—except the quantum number is now $`n`$ and there are additional corrections of the larger order $`\mathrm{log}N/N`$.
### 2.5 Schrödinger and Algebraic Parameter Relations
Equations 2.14 and 2.15 for the harmonic dilatation and radial displacement establish the needed connection between algebraic and geometrical parameters—at least in the regime where we are near the $`O(3)`$ limit. These quantities can be easily related to experimental data.
Experimentally one can find the lower energy level spacing $`\mathrm{\Delta }E_{\mathrm{exp}}`$, the reduced mass of the particle $`m_{\mathrm{exp}}`$, and from rotational spectra the distance of displacement $`r_{\mathrm{exp}}`$. In terms of these quantities one may compute the unitless distance using the harmonic oscillator dilatation
$$\alpha _{\mathrm{exp}}^2=\frac{m_{\mathrm{exp}}\mathrm{\Delta }E_{\mathrm{exp}}}{\mathrm{}^2}.$$
(2.18)
Equating $`\alpha _{\mathrm{exp}}^2r_{\mathrm{exp}}^2`$ with the previous expressions ($`\alpha ^2r_{}^{}{}_{}{}^{2}`$ given by 2.14 and 2.15) one finds
$$2N\left[1+\mathrm{corrections}\right]=\frac{m_{\mathrm{exp}}\mathrm{\Delta }E_{\mathrm{exp}}}{\mathrm{}^2}r_{\mathrm{exp}}^2.$$
(2.19)
This could be very valuable when fitting spectra. Since the $`O(3)`$ chain represents the ‘maximum’ radial displacement \[References\] this expression gives a lower bound for $`N`$. One may begin fitting data for the $`N`$ which satisfies this equation with the corrections set equal to $`0`$. If the $`O(3)`$ chain spectra with this $`N`$ doesn’t fit the experimental data then one can try higher $`N`$ and move off the $`O(3)`$ chain a corresponding amount so that the equation with the corrections is still satisfied.
## 3 2-D Franck-Condon Problem: The Prescription
### 3.1 Adding Dilatations
We are now ready to develop a prescription for calculating the FC factors for a $`U(3)`$ algebraic model. We begin by considering the FC problem of two configurations of a molecule both described by the $`U(2)`$ chain—i.e. two linear triatomics. Following the 1D procedure we propose a hybrid approach based upon calculating the matrix elements of the operator $`\widehat{T}`$ defined in terms of 2D SHO Schrödinger overlaps:
$$[N^{}],n^{},l|\widehat{T}|[N],n,l=T_{n,n^{},l}(\frac{\alpha }{\alpha ^{}})$$
(3.1)
The appropriate overlaps are calculated in Appendix D (equations D.3 and D.2). $`\widehat{T}`$ by construction does not connect subspaces of different $`ł`$.
### 3.2 Final Procedure
We have yet to add any dependance on the representation label $`N`$. One could add such dependance by hand, appealing to an analogy with the 1D case. However, when one has a bent to linear transition such an appeal is largely unnecessary due to equation 2.16.
For simplicity suppose we have a molecule with a bent configuration whose spectra is fit by a hamiltonian of the $`O(3)`$ chain in representation $`N`$ and a linear configuration whose spectra is fit by the $`U(2)`$ dynamical symmetry with $`\delta =\beta =0`$ (in this scenario the second representation label $`N^{}`$ can be arbitrary since the hamiltonian will generate identical low lying (experimentally measurable) spectra regardless of $`N^{}`$). Further, suppose the induced harmonic dilatations $`\alpha `$ and $`\alpha ^{}`$ (given by equation 2.18) can be calculated. In this instance we know that expanding the $`O(3)`$ basis in terms of the $`U(2)`$ basis is equivalent to expanding in terms of a SHO with harmonic dilatation $`\alpha _{U(2)}^2=\frac{1}{2}\zeta _n\alpha ^2`$ (where $`\zeta _n`$ is determined by 2.16). Thus the FC transitions should be the matrix elements:
$$[N^{}],n^{},l|\widehat{T}|[N],\omega ,l,$$
(3.2)
where $`\widehat{T}`$ is defined by
$$[N^{}],n^{},l|\widehat{T}|[N],n,l=T_{n,n^{},l}\left(\frac{\alpha }{\alpha ^{}}\sqrt{\frac{1}{2}\zeta _n}\right).$$
(3.3)
Expanding in large $`N`$ we see $`\sqrt{\frac{1}{2}\zeta _n}=a+bn`$ from equation 2.16 where $`b`$ is $`O(\frac{1}{N})`$ and $`a`$ has constant contributions and $`\frac{\mathrm{log}N}{N}`$ contributions. With the exception that $`a1`$ this is exactly the correction one has in the 1D case. That is, we have just let $`\alpha \alpha (a+bn)`$.
In this extreme case the FC factors can be calculated with no extra fitting parameters. However, our calculations depended on expansions in large $`N`$ \[References\] and thus we would expect them to be inaccurate for larger $`n`$. This can be compensated for phenomelogically by allowing $`b`$ to be fit to compensate for ignored terms.
Although the above situation is only for a limiting case, these results (along with insight from the 1D analysis) will imply exactly what will happen in other more realistic cases.
* Bent configuration near the $`O(3)`$ basis: One can in principle calculate the scale dependance near $`O(3)`$: $`\overline{m\omega }|_{\mathrm{Off}O(3)}\overline{m\omega }|_{O(3)}(1+\nu \frac{X}{Z}+\mu \frac{Y}{Z})`$ where $`\nu `$ and $`\mu `$ would have to be determined by repeating the calculation of \[References\] to second order. In such a calculation, it is clear that the result will go like $`\alpha \alpha (\stackrel{~}{a}+\stackrel{~}{b}n+\stackrel{~}{d}l^2)`$, where $`\stackrel{~}{a}=a(1+\gamma )`$ and $`\gamma `$ is $`O(\frac{X}{Z},\frac{Y}{Z})`$; $`\stackrel{~}{b}=b(1+\gamma ^{})`$ and $`\gamma ^{}`$ is $`O(\frac{X}{Z},\frac{Y}{Z})`$; the new parameter $`\stackrel{~}{d}`$, introducing $`l`$ dependence, is $`O(\frac{X}{ZN^2},\frac{Y}{ZN^2})`$; and we have ignored a $`n`$-$`l`$ cross term which is of a significantly smaller order. Although these coefficients could be calculated in principle, one must phenomenologically fit $`\stackrel{~}{b}`$ and $`\stackrel{~}{d}`$ in order to describe higher states anyways. Since the $`l`$ dependance is of lower order it seems likely that $`\stackrel{~}{d}`$ may not even be calculable with todays data.
* Linear configuration in $`U(2)`$ basis but $`\delta 0`$: In this instance the $`u(2)`$ requantization was approximate and inadequate for higher eigenstates. However, comparing the spectra to a Dunham expansion one sees that $`\frac{\delta }{ϵ}`$ plays the role of an anharmonicity parameter (recall equation 2.4). Thus, from the 1D case one concludes that $`\alpha ^{}\alpha ^{}(1+b^{}(n+1))`$ where $`b^{}`$ must be fit but should be of the order of $`\frac{\delta }{ϵ}`$.
* Either configuration far from both chains: Although our scheme has not allowed us to do explicit calculations in this regime, in terms of phenomologically fit parameters the result should still be clear. One should replace the induced harmonic dilatation by $`\alpha \alpha (\stackrel{~}{a}+\stackrel{~}{b}(n+1)+\stackrel{~}{d}l^2)`$. In this scenario $`\stackrel{~}{a}`$ should be between $`1`$ and $`a`$ depending on the proximity to either chain, $`\stackrel{~}{b}`$ has contributions due to both $`\delta `$ anharmonicities and the $`O(3)`$ chain, and one again expects that $`\stackrel{~}{d}`$ is of significantly smaller order. In this regime all parameters must be fit to data.
Let us recapitulate. The scheme we propose involves fitting the spectra of each configuration of the molecule in a $`U(3)`$ model to obtain wavefunctions $`|[N],\psi _E`$, $`|[N^{}],\psi _E^{}^{}`$. The FC factors are then
$$[N^{}],\psi _E^{}^{}|\widehat{T}|[N],\psi _E$$
(3.4)
where we define $`\widehat{T}`$ in terms of the $`U(2)`$ basis:
$$[N^{}],n^{},l^{}|\widehat{T}|[N],n,l=\delta _{l,l^{}}T_{n,n^{},l}\left(\frac{\alpha (\stackrel{~}{a}+\stackrel{~}{b}(n+1)+\stackrel{~}{d}l^2)}{\alpha ^{}(\stackrel{~}{a}^{}+\stackrel{~}{b}^{}(n^{}+1)+\stackrel{~}{d}^{}l^2)}\right).$$
(3.5)
In general the parameters must be fit, but whenever any of the limiting cases appear the appropriate theoretical values may be substituted. For example, if the linear configuration is highly harmonic one has $`\stackrel{~}{a}^{}=1`$ and $`\stackrel{~}{b}^{}=\stackrel{~}{d}^{}=0`$. Further, since we expect the $`l^2`$ dependance to be nearly negligible and we have a $`l=l^{}`$ selection rule, we may expand and combine $`\stackrel{~}{d}`$ and $`\stackrel{~}{d}^{}`$ into one parameter.
Note that this scheme may also be used for bent to bent transitions. This would be more useful than the 1D procedure \[References,References\] if one was interested in the $`l`$ dependance of the transitions for instance.
## 4 Conclusions
Using the result of \[References\] involving the geometry and scales of the dynamical symmetries of $`U(3)`$ we have explained the difference one sees when considering FC overlaps as described by the $`U(3)`$ algebra and FC overlaps as described by radially displaced Schrödinger oscillators. Additionally, the analysis has given us the practical result of a minimum value of $`N`$ as a function of simple experimentally measured quantities.
We observed that this scale dependency leads very naturally to a description of two dimensional FC factors. To compute the FC factors one expands the algebraic wavefunction in the $`U(2)`$ basis and formally replaces the basis element’s overlap by the 2D SHO Schrödinger overlap. Corrections for $`N`$, anharmonicities, and even $`l`$ dependance are made by letting the SHO dilatation constant have $`n`$ and $`l^2`$ dependent contributions. The resulting formulas are analogous to the 1D results obtained by expanding in anharmonicities.
## 5 Acknowledgements
This work was performed in part under the U.S. Department of Energy, Contract No. DE-FG02-91ER40608. I extend my deepest gratitude to my advisor, Prof. Franco Iachello, for introducing me to the FC problem, his suggestions and feedback on this work, and his useful comments on earlier versions of this manuscript. I am indebted to Thomas Müller and Prof. Patrick Vaccaro for useful discussions regarding the status of experimental results on bent to linear FC transitions. Finally, I would like to thank the Department of Energy’s Institute for Nuclear Theory at the University of Washington for its hospitality during the completion of this work.
Appendices
## Appendix A Approximate Analytic Expressions for 2D FC overlaps
### A.1 Linear Configuration
In the linear configuration in the harmonic limit one has, after separating out the azimuthal portion of the wavefunction, the equation:
$$\frac{1}{r}\frac{d}{dr}\left(r\frac{d}{dr}v(r)\right)\frac{l^2}{r^2}v(r)+\left(\stackrel{~}{E}\lambda ^2r^2\right)v(r)=0,$$
(A.1)
where $`\lambda `$ is related to the frequency of the oscillator by $`\lambda =\frac{m\omega }{\mathrm{}}`$ and $`\stackrel{~}{E}=\frac{2m}{\mathrm{}^2}E`$. Subject to the boundary conditions $`v(r)|_{r=\mathrm{}}=0`$, $`rv(r)|_{r=0}=0`$ the solution is well known to be \[References\]:
$$v_{n_r,l}(r)=\sqrt{\frac{2n_r!\lambda ^{|l|+1}}{(|l|+n_r)!}}r^{|l|}e^{\frac{\lambda }{2}r^2}L_{n_r}^{|l|}(\lambda r^2),$$
(A.2)
with eigenvalue
$$\stackrel{~}{E}=2\lambda (|l|+1+2n_r).$$
(A.3)
### A.2 Bent Configuration
For a system w/ equilibrium located at some $`r=r_0`$ we again assume that for the lowest states the system is well approximated by a harmonic oscillator. Our wave equation is thus:
$$\frac{1}{r}\frac{d}{dr}\left(r\frac{d}{dr}v(r)\right)\frac{l^2}{r^2}v(r)+\left(\stackrel{~}{E}^{}\lambda _{}^{}{}_{}{}^{2}(rr_0)^2\right)v(r)=0.$$
(A.4)
Notice, that this mild transformation will greatly change the form of the solutions (as opposed to the 1D case) since the laplacian is not invariant under radial displacements. Note that towards the origin the potential is ‘heightwise truncated’ (in full 2-space the potential is not differentiable at the origin). Thus, to harmonically approximate the potential we must assure that we are sufficiently far from the origin, i.e. loosely $`\lambda ^{}r_0^2`$ must be large.
To obtain approximate solutions to this bent configuration equation we make the substitution $`v(r)=\frac{u(r)}{\sqrt{r}}`$. For $`l0`$ we may expand the effective potential about its minima
$$r^{}=r_0\left(1+ϵ3ϵ^2+O(ϵ^3)\right)$$
(A.5)
where $`ϵ=\frac{1}{\lambda _{}^{}{}_{}{}^{2}}(l^2\frac{1}{4})\frac{1}{r_0^4}`$ and our large $`r_0`$ condition is refined to
$$27r_0^4>\frac{256}{\lambda _{}^{}{}_{}{}^{2}}(l^2\frac{1}{4}).$$
(A.6)
Moving the left boundary condition from $`r=0`$ to $`r=\mathrm{}`$ only introduces $`O(ϵ)`$ corrections so we may obtain a 1D SHO equation:
$$\frac{d^2}{dr^2}u(r)+\left(\mathrm{\Delta }+\overline{\lambda }^2(rr^{})^2\right)u(r)=\stackrel{~}{E}^{}u(r).$$
(A.7)
where $`\mathrm{\Delta }=r_0^2\lambda _{}^{}{}_{}{}^{2}(ϵϵ^2+O(ϵ^3))`$ and $`\overline{\lambda }^2=\lambda _{}^{}{}_{}{}^{2}(1+3ϵ12ϵ^2+O(ϵ^3))`$. Note that the $`l`$ dependence of the solution is implicit to the behavior of $`\overline{\lambda }`$. We find that the actual wavefunction for the bent configuration is:
$$v_{n_r,l}(r)\frac{\overline{\lambda }^{\frac{1}{4}}}{(\sqrt{\pi }2^{n_r}n_r!)^{\frac{1}{2}}}\frac{e^{\frac{1}{2}\overline{\lambda }r^2}}{\sqrt{r}}H_{n_r}(\sqrt{\overline{\lambda }}(rr^{})).$$
(A.8)
Although this solution is singular at the origin this is of little concern since the metric contains a factor of $`r`$ (the real objects of interest are $`\sqrt{r}v(r)`$, of which we have a good approximation).
### A.3 Franck-Condon Factors
We wish to calculate the overlap of wavefunctions from a linear and bent configuration. Given the $`SO(2)`$ symmetry of both configurations the angular part of the wavefunctions simply gives the $`\delta _{l,l^{}}`$ selection rule. Hence, the integral we wish to evaluate is
$$I_{n_r,n_r^{}}=M_{n_r^{},l}N_{n_r,l}_0^{\mathrm{}}𝑑rr^{|l|+\frac{1}{2}}e^{\frac{\lambda +\overline{\lambda }}{2}r^2}L_{n_r}^{|l|}(\lambda r^2)H_{n_r^{}}(\sqrt{\overline{\lambda }}(rr^{})),$$
(A.9)
where we introduced the short hand $`M`$ and $`N`$ for the normalizations.
We remove all dimensionful quantities by making the change of variable $`u=\sqrt{\lambda }r`$. Additionally we decide to work in the (realistic) domain where $`\lambda \lambda ^{}`$ by setting $`\frac{\lambda ^{}}{\lambda }=1+\delta `$ where $`\delta `$ is small. Using expressions from \[References\] for the expansion
$$H_s(\alpha (xx_0))=\underset{0ns}{}t_{n,s}(\alpha ,x_0)H_n(x),$$
(A.10)
assuming $`\delta `$ and $`ϵ`$ are of the same order, and ignoring quadratic terms:
$$I_{n_r,n_r^{}}\frac{M_{n_r^{},l}N_{n_r,l}}{\lambda ^{\frac{|l|}{2}+\frac{3}{4}}}\left(2\frac{(l^2\frac{1}{4})^{\frac{1}{4}}}{ϵ^{\frac{1}{4}}}\right)^{n_r^{}}_0^{\mathrm{}}𝑑uu^{|l|+\frac{1}{2}}e^{u^2}L_{n_r}^{|l|}(u^2).$$
(A.11)
Not surprisingly the dominant term is independent of $`\delta `$ (reflecting the fact that FC integrals tend to be more sensitive to changes in displacement than to dilatation). Evaluating the integral \[References\] and simplifying:
$$I_{n_r,n_r^{}}\left(\frac{2^{n_r^{}}}{2\sqrt{\pi }(|l|+n_r)!n_r^{}!n_r!}\right)^{\frac{1}{2}}\left(r_0\sqrt{\lambda ^{}}\right)^{n_r^{}}\left(\frac{|l|}{2}+\frac{1}{4}\right)_{n_r}\mathrm{\Gamma }\left(\frac{|l|}{2}+\frac{3}{4}\right).$$
(A.12)
## Appendix B Schrödinger and Algebraic Quantum Numbers
In the preceding sections for each $`l=0,\pm 1,\pm 2\mathrm{}`$ we have that $`n_r=0,1,2,\mathrm{}`$ with the energy relation $`E|l|+2n_r`$. For the $`u(2)`$ chain of the $`u(3)`$ model with hamiltonian $`H=E_0+k\widehat{n}`$ (the algebraic model most similar to the 2D SHO) we have as a spectra $`En`$. The branching rules imply that for each $`l=0,\pm 1,\pm 2\mathrm{}`$ we have $`n=|l|,|l|+2,|l|+4,\mathrm{}`$. Note that both the expressions for the spectra and the rules for the ranges of the quantum numbers agree if we make the identification:
$`n_r={\displaystyle \frac{n|l|}{2}}.`$ (B.1)
Thus the wavefunction with lowest energy within a subspace of given $`l`$ is labelled by $`n_r=0`$ in the Schrödinger picture and $`n=|l|`$ in the algebraic picture.
## Appendix C Scale Changes
Our equation 2.16 differs from the results of \[References\] by the substitution $`|l|n`$. This difference occurs because \[References\] did not compute overlaps for higher energy wavefunctions since its approximation was most valid in the low energy limit. However, their requantization technique works exactly for all levels of the $`U(2)`$ chain when $`\delta =0`$. We therefore may expect these higher energy wavefunctions to be more reliable and repeat their procedure to calculate the scale change for all levels in this larger domain. Proceeding to do so one finds that the linear $`|l|`$ dependance of equation 2.16 was an artifact of working in the ground state of each $`|l|`$ subspace and the true correspondence has linear contributions appearing in $`n`$.
## Appendix D Schrödinger FC Factors for dilatated $`r^{}=0`$ Harmonic Potentials
Using the solution to the radial part of the Schrödinger equation for the 2D SHO stated in Appendix A the FC overlaps for a dilatated SHO become:
$$2\sqrt{\frac{n_r!n_r^{}!(\lambda \lambda ^{})^{|l|+1}}{(|l|+n_r)!(|l|+n_r^{})!}}_0^{\mathrm{}}r𝑑rr^{2|l|}e^{\frac{\lambda +\lambda ^{}}{2}r^2}L_{n_r}^{|l|}(\lambda r^2)L_{n_r^{}}^{|l|}(\lambda ^{}r^2).$$
(D.1)
The integral may be evaluated in terms of a hypergeometric function \[References\] and simplified to:
$`I_{n_r,n_r^{},l}^{\mathrm{SHO}}(\lambda ,\lambda ^{})=\sqrt{{\displaystyle \frac{(\lambda \lambda ^{})^{|l|+1}n_r!n_r^{}!}{(n_r+|l|)!(n_r^{}+|l|)!}}}()^{n_r}\left({\displaystyle \frac{2}{\lambda +\lambda ^{}}}\right)^{|l|+1}`$ (D.2)
$`\times `$ $`{\displaystyle \underset{m=0}{\overset{\mathrm{min}(n_r,n_r^{})}{}}}{\displaystyle \frac{()^m(n_r+n_r^{}+|l|m)!}{m!(n_rm)!(n_r^{}m)!}}\left({\displaystyle \frac{\lambda \lambda ^{}}{\lambda +\lambda ^{}}}\right)^{n_r+n_r^{}2m},`$
where we rearranged the sum to make the formula slightly more amenable to computer implementation. With some minor algebra and judicious manipulation of Pochhammers one sees that the expression has the correct $`\lambda ^{}\lambda `$ limit of $`\delta _{n_r,n_r^{}}`$.
Given the relationships of Appendix B and that $`\lambda =\alpha ^2`$ the result needed for Section 3.1 is:
$$T_{n,n^{},l}(\frac{\alpha }{\alpha ^{}})=I_{n_r=\frac{n|l|}{2},n_r^{}=\frac{n^{}|l|}{2},l}^{\mathrm{SHO}}(\alpha ^2,\alpha _{}^{}{}_{}{}^{2})$$
(D.3)
|
warning/0005/hep-th0005288.html
|
ar5iv
|
text
|
# Untitled Document
hep-th/0005288
Note on Small Black Holes in $`AdS_p\times S^q`$
Gary T. Horowitz and Veronika E. Hubenygary@cosmic.physics.ucsb.edu, veronika@cosmic.physics.ucsb.edu
Physics Department, University of California, Santa Barbara, CA 93106, USA
Abstract
It is commonly believed that small black holes in $`AdS_5\times S^5`$ can be described by the ten dimensional Schwarzschild solution. This requires that the self-dual five-form (which is nonzero in the background) does not fall through the horizon and cause the black hole to grow. We verify that this is indeed the case: There are static solutions to the five-form field equations in a ten dimensional Schwarzschild spacetime. Similar results hold for other backgrounds $`AdS_p\times S^q`$ of interest in supergravity.
June, 2000
One of the most important consequences of the AdS/CFT correspondence \[1,,2\] is the claim that the formation and evaporation of black holes can be described by a standard unitary evolution. Since this claim is contrary to well known semiclassical arguments , it is worthwhile to carefully examine the ingredients which go into this conclusion. One such ingredient is the assumption that a small black hole in $`AdS_5\times S^5`$ will behave just like a ten dimensional Schwarzschild black hole. Intuitively, this seems reasonable since for a large AdS radius, the local description should be approximately given by the corresponding flat ten-dimensional spacetime physics; in particular, a small black hole should be approximately described by the 10-D Schwarzschild solution (sufficiently near the black hole).
However, the supergravity solution also includes a nonzero five-form. Although this acts like a cosmological constant in solutions which are products of two five dimensional spaces, in general it contains dynamical degrees of freedom. Given our experience with previous ‘no-hair’ theorems, one might worry that a small black hole will cause the five-form to fall into the horizon. Even though the local energy density in the five-form is small, if this were the case, most small black holes would grow by classically absorbing the energy density of the five-form and not quantum mechanically evaporate.
We show below that this does not occur. There exist static solutions for a self-dual five-form in the background of a ten dimensional black hole which have the correct boundary conditions at infinity to match onto the $`AdS_5\times S^5`$ solution. It is these boundary conditions which effectively stabilize the field and invalidate the ‘no-hair’ intuition. The five-form is distorted by the black hole, but does not cause it to grow. We also show that similar results hold for four-forms and seven-forms in the background of an 11-D Schwarzschild solution with the right boundary conditions to match onto $`AdS_4\times S^7`$ and $`AdS_7\times S^4`$.
We start by noting that for pure $`AdS_5\times S^5`$, the solution (in global coordinates) is given by the metric
$$ds^2=\left(\frac{\rho ^2}{R^2}+1\right)dt^2+\frac{d\rho ^2}{\frac{\rho ^2}{R^2}+1}+\rho ^2d\mathrm{\Omega }_3^2+d\chi ^2+R^2\mathrm{sin}^2\frac{\chi }{R}d\mathrm{\Omega }_4^2$$
where $`R`$ is the radius of curvature, and the five-form field strength $`\stackrel{~}{F}`$ is the sum of the volume form on $`AdS_5`$ and on $`S^5`$, normalized so that $`_{S^5}\stackrel{~}{F}=N`$. To simplify the formulas below, we will work with the rescaled five-form $`F(\pi ^3R^5/N)\stackrel{~}{F}`$, so that $`F`$ is just the sum of the volume forms<sup>1</sup> We use $`d\mathrm{\Omega }_n`$ to denote the volume $`n`$-form on unit $`S^n`$ and $`d\mathrm{\Omega }_n^2`$ to denote the metric on $`S^n`$.
$$F=\rho ^3dtd\rho d\mathrm{\Omega }_3+R^4\mathrm{sin}^4\frac{\chi }{R}d\chi d\mathrm{\Omega }_4$$
How does this solution change in the presence of a black hole? For a black hole with radius larger than $`R`$, we already know what this modification is: The metric on $`AdS_5`$ is replaced with the five-dimensional Schwarzschild-AdS solution and the metric on the $`S^5`$ is unchanged
$$ds^2=\left(\frac{\rho ^2}{R^2}+1\frac{\rho _0^2}{\rho ^2}\right)dt^2+\frac{d\rho ^2}{\frac{\rho ^2}{R^2}+1\frac{\rho _0^2}{\rho ^2}}+\rho ^2d\mathrm{\Omega }_3^2+d\chi ^2+R^2\mathrm{sin}^2\frac{\chi }{R}d\mathrm{\Omega }_4^2$$
Since this change in the metric does not effect the volume form on AdS, the five-form field strength $`F`$ remains the same. In particular, the self-duality condition is satisfied because only the combination $`dtd\rho `$ is present in this condition, so that the mass-dependence cancels out, and the “Bianchi identity” $`dF=0`$ is independent of the metric. (It is clear that $`F`$ remains smooth even at the horizon since the volume form on Schwarzschild-AdS is smooth there.)
For a small black hole, the picture becomes much less clear. The black hole is localized on the $`S^5`$ as well as in the $`AdS_5`$ , so that the metric no longer factorizes. Hence we cannot just look for a lower dimensional solution with an effective cosmological constant. Finding the appropriate exact solution to the full 10-D Einstein five-form field equations seems intractable. Since the curvature near the horizon of a small black hole should be much larger than the field strength $`F`$, to a good approximation one can ignore the backreaction and treat the five-form as a test field on a fixed background spacetime. In this approximation, the metric satisfies the vacuum equations, and the unique static, spherically symmetric black hole solution is the ten-dimensional Schwarzschild metric. However, this approximation is consistent only if there exists a static solution for a test self-dual five-form in this background, with the right boundary conditions. These boundary conditions can be understood as follows.
Very far away from the black hole, both the metric and the five-form should approach the forms given respectively by eqs. (1) and (1). Since the black hole is much smaller than $`R`$, these forms are valid even into the approximately flat region of small $`\rho `$ and $`\chi `$. We can identify this approximately flat region with the asymptotic region far from the Schwarzschild black hole. This then sets our boundary conditions “at infinity”. To be more explicit, we first write the 10-D Schwarzschild solution in convenient coordinates in which the boundary conditions are easily posed while the required symmetries are still manifest. In particular, we want to use the 10-D radial coordinate (fixed by the area of $`S^8`$), but to split $`S^8`$ into $`S^3`$ and $`S^4`$, corresponding to the rotational SO(4) symmetry of $`AdS_5`$ and the remaining (unbroken) SO(5) rotational symmetry on $`S^5`$. This is achieved by using the coordinate transformation
$$\rho =r\mathrm{sin}\theta $$
$$\chi =r\mathrm{cos}\theta $$
In these coordinates, the flat spacetime metric obtained from (1) in the limit $`\rho ,\chi R`$ takes the form
$$ds^2=dt^2+dr^2+r^2\left(d\theta ^2+\mathrm{sin}^2\theta d\mathrm{\Omega }_3^2+\mathrm{cos}^2\theta d\mathrm{\Omega }_4^2\right)$$
(The angular term in the parentheses is equivalent to $`d\mathrm{\Omega }_8^2`$.) Similarly, the five-form field strength obtained from (1) and (1) in the limit $`\rho ,\chi R`$ takes the form
$$F=r^3\mathrm{sin}^4\theta dtdrd\mathrm{\Omega }_3r^4\mathrm{sin}^3\theta \mathrm{cos}\theta dtd\theta d\mathrm{\Omega }_3$$
$$+r^4\mathrm{cos}^5\theta drd\mathrm{\Omega }_4r^5\mathrm{sin}\theta \mathrm{cos}^4\theta d\theta d\mathrm{\Omega }_4$$
One can easily recheck that $`F`$ is still closed and self-dual.
In these coordinates, the 10-D Schwarzschild metric is given by:
$$ds^2=f(r)dt^2+f^1(r)dr^2+r^2\left(d\theta ^2+\mathrm{sin}^2\theta d\mathrm{\Omega }_3^2+\mathrm{cos}^2\theta d\mathrm{\Omega }_4^2\right)$$
with $`f(r)1\left(\frac{r_+}{r}\right)^7`$. A general ansatz for the field strength with the required symmetries (namely $`F`$ being static and spherically symmetric on $`S^3`$ and $`S^4`$) can be obtained by taking each of the four terms in (1) and multiplying by arbitrary functions of $`r`$ and $`\theta `$:
$$F=g_1(r,\theta )r^3\mathrm{sin}^4\theta dtdrd\mathrm{\Omega }_3g_3(r,\theta )r^4\mathrm{sin}^3\theta \mathrm{cos}\theta dtd\theta d\mathrm{\Omega }_3$$
$$+g_2(r,\theta )r^4\mathrm{cos}^5\theta drd\mathrm{\Omega }_4g_4(r,\theta )r^5\mathrm{sin}\theta \mathrm{cos}^4\theta d\theta d\mathrm{\Omega }_4$$
Our boundary conditions require that $`g_i(r,\theta )1`$ as $`r\mathrm{}`$.<sup>2</sup> In principle, terms of the form $`\gamma _1(r,\theta )dtd\mathrm{\Omega }_4`$ and $`\gamma _2(r,\theta )drd\theta d\mathrm{\Omega }_3`$ would also be consistent with all the symmetries, and would satisfy the boundary conditions provided $`\gamma _i(r,\theta )0`$ as $`r\mathrm{}`$. However, since F is closed, $`_r\gamma _1=_\theta \gamma _1=0`$, which, along with the boundary condition $`\gamma _10`$, requires that $`\gamma _1(r,\theta )0`$. Self duality of $`F`$ then forces $`\gamma _2(r,\theta )0`$. Hence these terms will not arise, and the most general form of $`F`$ will indeed be given by (1). To determine the field strength (1) explicitly, we now impose the physical conditions that $`F`$ is closed and self-dual (with respect to the black hole metric (1)).
We first eliminate two of the four arbitrary functions $`g_i`$ appearing in (1) by imposing self-duality, $`F=F`$. The volume form associated with (1) is simply given by
$$\epsilon _{(10)}=r^8\mathrm{sin}^3\theta \mathrm{cos}^4\theta dtdrd\theta d\mathrm{\Omega }_3d\mathrm{\Omega }_4$$
Correspondingly, the dual of $`F`$ is
$$F=g_1(r,\theta )r^5\mathrm{sin}\theta \mathrm{cos}^4\theta d\theta d\mathrm{\Omega }_4+g_3(r,\theta )\frac{1}{f(r)}r^4\mathrm{cos}^5\theta drd\mathrm{\Omega }_4$$
$$g_2(r,\theta )f(r)r^4\mathrm{sin}^3\theta \mathrm{cos}\theta dtd\theta d\mathrm{\Omega }_3g_4(r,\theta )r^3\mathrm{sin}^4\theta dtdrd\mathrm{\Omega }_3$$
Self-duality then requires $`g_4=g_1`$ and $`g_3=fg_2`$, so that (1) becomes
$$F=g_1(r,\theta )\left[r^3\mathrm{sin}^4\theta dtdrd\mathrm{\Omega }_3r^5\mathrm{sin}\theta \mathrm{cos}^4\theta d\theta d\mathrm{\Omega }_4\right]$$
$$+g_2(r,\theta )\left[r^4f(r)\mathrm{sin}^3\theta \mathrm{cos}\theta dtd\theta d\mathrm{\Omega }_3+r^4\mathrm{cos}^5\theta drd\mathrm{\Omega }_4\right]$$
The condition that $`F`$ is nonsingular at the horizon requires that the arbitrary functions $`g_1(r,\theta )`$ and $`g_2(r,\theta )`$ are smooth at $`r_+`$. This can be easily seen by rewriting (1) in the ingoing Eddington coordinates, $`(v,r,\theta ,\mathrm{\Omega }_3,\mathrm{\Omega }_4)`$, which are regular at the horizon. Since $`vt+r_{}`$, where $`r_{}`$ is defined by $`dr_{}\frac{dr}{f(r)}`$, we can simply rewrite $`dt=dv\frac{dr}{f(r)}`$. The field strength is then expressed as
$$F=g_1(r,\theta )\left[r^3\mathrm{sin}^4\theta dvdrd\mathrm{\Omega }_3r^5\mathrm{sin}\theta \mathrm{cos}^4\theta d\theta d\mathrm{\Omega }_4\right]$$
$$+g_2(r,\theta )[r^4f(r)\mathrm{sin}^3\theta \mathrm{cos}\theta dvd\theta d\mathrm{\Omega }_3+r^4\mathrm{sin}^3\theta \mathrm{cos}\theta drd\theta d\mathrm{\Omega }_3$$
$$+r^4\mathrm{cos}^5\theta drd\mathrm{\Omega }_4]$$
and we see that all the terms are smooth at $`r_+`$ if $`g_i`$ are smooth at $`r_+`$.
We now require that $`F`$ is closed, $`dF=0`$. Since $`dF`$ has two nontrivial components, proportional to $`dtdrd\theta d\mathrm{\Omega }_3`$ and to $`drd\theta d\mathrm{\Omega }_4`$, we obtain two independent equations by setting each component to $`0`$:
$$r^3_\theta (g_1\mathrm{sin}^4\theta )_r(r^4fg_2)\mathrm{sin}^3\theta \mathrm{cos}\theta =0$$
$$_r(r^5g_1)\mathrm{sin}\theta \mathrm{cos}^4\theta +r^4_\theta (g_2\mathrm{cos}^5\theta )=0$$
We can simplify these partial differential equations further by separation of variables. By writing $`g_i(r,\theta )g_i(r)\stackrel{~}{g}_i(\theta )`$, the radial and angular parts decouple. By direct substitution, (1) becomes
$$\frac{\stackrel{~}{g}_1^{}(\theta )\mathrm{tan}\theta +4\stackrel{~}{g}_1(\theta )}{\stackrel{~}{g}_2(\theta )}=k=\frac{rf(r)g_2^{}(r)+4f(r)g_2(r)+rf^{}(r)g_2(r)}{g_1(r)}$$
whereas (1) yields
$$\frac{\stackrel{~}{g}_2^{}(\theta )\mathrm{cot}\theta +5\stackrel{~}{g}_2(\theta )}{\stackrel{~}{g}_1(\theta )}=l=\frac{rg_1^{}(r)+5g_1(r)}{g_2(r)}$$
where $`k`$ and $`l`$ are arbitrary separation constants. These are, however, fixed by the boundary conditions: Since $`g_i(r)\stackrel{~}{g}_i(\theta )1`$ as $`r\mathrm{}`$, each function must approach a constant, which we can require to be one, as $`r\mathrm{}`$: i.e. $`g_i(r)1`$ and $`\stackrel{~}{g}_i(\theta )1`$. The latter requirement dictates that $`\stackrel{~}{g}_i(\theta )=1`$, so that the angular part is trivial. This fixes the separation constants completely: $`k=4`$ and $`l=5`$. (We note that this is also self-consistently required by the radial parts of (1) and (1).)
Thus, we are left with the following coupled, linear, first order, ordinary differential equations for $`g_1(r)`$ and $`g_2(r)`$:
$$g_1(r)=f(r)g_2(r)+\frac{1}{4}r\frac{d}{dr}\left(f(r)g_2(r)\right)$$
$$g_2(r)=g_1(r)+\frac{1}{5}r\frac{d}{dr}g_1(r)$$
with the asymptotic boundary conditions $`g_i(r)1`$ as $`r\mathrm{}`$. Ordinarily, one would expect to be able to specify both $`g_1`$ and $`g_2`$ at $`r=r_+`$ and then integrate out to infinity. One could then hope to choose these two initial conditions to satisfy the two boundary conditions. However, (1) implies the following constraint at the horizon (using the fact that $`f(r_+)=0`$):
$$g_1(r_+)=\frac{7}{4}g_2(r_+)$$
so the solutions are determined by only one free parameter. Nevertheless, it is still possible to satisfy both boundary conditions. This is most easily seen by substituting (1) into (1) to obtain a decoupled, second order equation for $`g_1(r)`$:
$$f(r)g_1^{\prime \prime }(r)+\left(\frac{10}{r}f(r)+f^{}(r)\right)g_1^{}(r)+\left(\frac{20}{r^2}(f(r)1)+\frac{5}{r}f^{}(r)\right)g_1(r)=0$$
The asymptotic form of this equation is
$$g_1^{\prime \prime }(r)+\frac{10}{r}g_1^{}(r)=0,$$
so as $`r\mathrm{}`$, we have $`g_1(r)`$ const + $`O(1/r^9)`$. There is only a one parameter family of solutions to the second order equation (1) which are regular at the horizon since $`f(r_+)=0`$ implies
$$g_{1}^{}{}_{}{}^{}(r_+)=\frac{15}{7r_+}g_1(r_+)$$
So given $`g_1(r_+)`$, we get a unique solution of the second order equation (1). We can clearly rescale $`g_1(r_+)`$ so that $`g_11`$ at infinity. The function $`g_2`$ is then completely determined by (1), but fortunately it automatically satisfies the right boundary condition, $`g_21`$ asymptotically. This shows that a solution satisfying all boundary conditions does exist.
Fig. 1: Solutions $`g_1(r)`$ (dotted line) and $`g_2(r)`$ (dashed line), and their asymptotic value of one (solid line) for a 10-D black hole with radius $`r_+=1`$
Although we have not found the solution analytically, one can easily find it numerically.<sup>3</sup> However, we cannot integrate the solution directly from the horizon, since $`g_1^{\prime \prime }(r_+)`$ is undetermined there due to the factor of $`f(r)=0`$ at $`r=r_+`$. Instead, we must obtain new “initial conditions” near the horizon, at $`r=r_++\epsilon `$. A plot of the solution is shown in Fig. 1. We see that $`g_1`$ is enhanced and $`g_2`$ is slightly suppressed at the horizon, while both functions asymptote to the correct value, $`g_i1`$.
Even though we have found static solutions for a test five-form field strength in a 10-D Schwarzschild background (given by (1)), $`F`$ does not vanish at the horizon. So to ensure that the solution remains static when the backreaction is included, we need to check that there is no energy flux crossing the horizon. By the Raychaudhuri equation , the horizon area can remain constant only if $`R_{ab}k^ak^b=0`$, where $`k^a`$ denotes the null generators of the horizon, $`k^a=\left(\frac{}{t}\right)^a`$. Thus, a static configuration must satisfy $`T_{ab}k^ak^b=0`$ at the horizon. One can easily show that this is indeed the case for our solution:
$$T_{ab}k^ak^bk^aF_{acdem}k^bF_b^{cdem}$$
and from (1), we have (in component notation)
$$k^aF_{acdem}g_1(r)r^3\mathrm{sin}^4\theta (dr)_{[c}(d\mathrm{\Omega }_3)_{dem]}+g_2(r)r^4f(r)\mathrm{sin}^3\theta \mathrm{cos}\theta (d\theta )_{[c}(d\mathrm{\Omega }_3)_{dem]}$$
Contracting over $`c,d,e,`$ and $`m`$ yields
$$k^aF_{acdem}k^bF_b^{cdem}f(r)g_1^2(r)\mathrm{sin}^2\theta +f^2(r)g_2^2(r)\mathrm{cos}^2\theta $$
which clearly vanishes at the horizon, since $`f(r_+)=0`$ and $`g_i(r_+)`$ remain finite. Hence
$$T_{ab}k^ak^b=0$$
is indeed satisfied at the horizon.
So far, we have considered a five-form field strength in the presence of a small 10-D black hole in asymptotically $`AdS_5\times S^5`$ spacetime. We now check that the arguments of the preceeding section also apply to the other cases of interest for the AdS/CFT correspondence.
We start with the 11-D supergravity solutions $`AdS_4\times S^7`$ and $`AdS_7\times S^4`$. For conciseness, we combine these into the general case of $`AdS_p\times S^q`$, where $`(p,q)=(4,7)`$ and $`(7,4)`$. Here the logic of the argument is slightly different from the previous case, since dimensionally the field strength cannot be self-dual. Nonetheless, we shall see that the final differential equations are very similar to (1) and (1) (and are in fact identical if we set $`p=q=5`$). This will allow us to apply the same arguments as above to prove the existence of a static solution satisfying the correct boundary conditions.
As in the preceeding discussion, we start with the metric in global AdS coordinates:
$$ds^2=\left(\frac{\rho ^2}{R^2}+1\right)dt^2+\frac{d\rho ^2}{\frac{\rho ^2}{R^2}+1}+\rho ^2d\mathrm{\Omega }_{p2}^2+d\chi ^2+(\alpha R)^2\mathrm{sin}^2\frac{\chi }{\alpha R}d\mathrm{\Omega }_{q1}^2$$
where $`\alpha `$ is a numerical constant, corresponding to the ratio of the size of the sphere to the size of AdS for the given supergravity solution ($`\alpha =\frac{1}{2}`$ for $`AdS_7\times S^4`$, and $`\alpha =2`$ for $`AdS_4\times S^7`$). The flat space approximation ($`\rho ,\chi R`$) of the metric and the corresponding volume form are given by
$$ds^2=dt^2+d\rho ^2+\rho ^2d\mathrm{\Omega }_{p2}^2+d\chi ^2+\chi ^2d\mathrm{\Omega }_{q1}^2$$
$$\epsilon _{p+q}=\rho ^{p2}\chi ^{q1}dtd\rho d\mathrm{\Omega }_{p2}d\chi d\mathrm{\Omega }_{q1}$$
Hence, the $`p`$-form field strength and its $`q`$-form dual in this region are simply
$$F_{(p)}=\rho ^{p2}dtd\rho d\mathrm{\Omega }_{p2}$$
$$F_{(q)}=\chi ^{q1}d\chi d\mathrm{\Omega }_{q1}$$
Now, we use the change of coordinates (1) to write the $`(p+q)`$-dimensional Schwarzschild metric in the form
$$ds^2=f(r)dt^2+f^1(r)dr^2+r^2\left(d\theta ^2+\mathrm{sin}^2\theta d\mathrm{\Omega }_{p2}^2+\mathrm{cos}^2\theta d\mathrm{\Omega }_{q1}^2\right)$$
where $`f(r)1\left(\frac{r_+}{r}\right)^{p+q3}`$. Since the full $`(p+q)`$-dimensional volume form is independent of $`f`$, it can be obtained from (1) and (1)
$$\epsilon _{p+q}=(1)^{p1}r^{p+q2}\mathrm{sin}^{p2}\theta \mathrm{cos}^{q1}\theta dtdrd\theta d\mathrm{\Omega }_{p2}d\mathrm{\Omega }_{q1}$$
Up till now, everything was just a simple generalization of the $`AdS_5\times S^5`$ case. However, the general $`p`$-form in the presence of the localized black hole, which is consistent with all the symmetries now has only two arbitrary functions,
$$F_{(p)}=g_1(r,\theta )r^{p2}\mathrm{sin}^{p1}\theta dtdrd\mathrm{\Omega }_{p2}$$
$$g_2(r,\theta )f(r)r^{p1}\mathrm{sin}^{p2}\theta \mathrm{cos}\theta dtd\theta d\mathrm{\Omega }_{p2}$$
$`g_1(r,\theta )`$ and $`g_2(r,\theta )`$ are smooth everywhere and chosen such that they satisfy the simple flat space boundary condition<sup>4</sup> The function $`f(r)`$ was inserted into the second term for later convenience. (Note that $`f(r)1`$ as $`r\mathrm{}`$, so the asymptotic boundary conditions remain uneffected.) $`g_1(r,\theta )1`$ and $`g_2(r,\theta )1`$ as $`r\mathrm{}`$. (The other two terms which appeared in (1) for $`F_{(5)}`$ are not consistent with the dimensionality: they are $`q`$-forms rather than $`p`$-forms.) The dual $`q`$-form is then
$$F_{(q)}=g_1(r,\theta )r^q\mathrm{sin}\theta \mathrm{cos}^{q1}\theta d\theta d\mathrm{\Omega }_{q1}+g_2(r,\theta )r^{q1}\mathrm{cos}^q\theta drd\mathrm{\Omega }_{q1}$$
The differential equations for $`g_1`$ and $`g_2`$ are obtained by the condition that both the $`p`$-form field strength and its dual $`q`$-form must be closed, i.e. $`dF_{(p)}=0`$ and $`dF_{(q)}=0`$. Canceling out the angular dependence (using the same separation of variables procedure as before) yields
$$g_1(r)=f(r)g_2(r)+\frac{1}{p1}r_r\left(f(r)g_2(r)\right)$$
$$g_2(r)=g_1(r)+\frac{1}{q}r_rg_1(r)$$
Note that, as advertised, for $`p=q=5`$, these equations are identical to (1) and (1) obtained above. The second order ODE for $`g_1(r)`$ is
$$f(r)g_1^{\prime \prime }(r)+\left(\frac{p+q}{r}f(r)+f^{}(r)\right)g_1^{}(r)+\left(\frac{q(p1)}{r^2}(f(r)1)+\frac{q}{r}f^{}(r)\right)g_1(r)=0$$
The asymptotic form of this equation,
$$g_1^{\prime \prime }(r)+\frac{p+q}{r}g_1^{}(r)=0$$
implies $`g_1(r)`$ const + $`O(1/r^{p+q1})`$ as $`r\mathrm{}`$.
Solutions which are regular at the horizon are again determined by one parameter since
$$g_{1}^{}{}_{}{}^{}(r_+)=\frac{q}{r_+}\left(\frac{q2}{p+q3}\right)g_1(r_+)$$
One can choose this parameter so that $`g_11`$ asymptotically. Then $`g_2`$ is uniquely determined by (1) and automatically satisfies the right boundary condition $`g_21`$.
One can similarly show that the same conclusion will hold for another case of interest for the AdS/CFT duality, namely in IIB supergravity on $`AdS_3\times S^3\times T^4`$. In this case, the 4-torus decouples, and by a similar procedure as for $`AdS_5\times S^5`$, we arrive at equations (1) and (1), with $`p=q=3`$.
In all of the above cases one can easily verify that there is no flux of energy across the event horizon, $`T_{ab}k^ak^b=0`$. (Note that although the stress tensor now has a nonzero trace, the term proportional to the metric does not contribute when contracted with the null vectors $`k^ak^b`$.) So the solutions will remain static when backreaction is included. Thus, we have shown that in all the relevant cases, a small black hole in $`AdS_p\times S^q`$ can indeed be approximated by a $`(p+q)`$-dimensional Schwarzschild solution. The Ramond-Ramond fields will be distorted, but will remain static and not cause the black hole to grow. In retrospect, it is perhaps not surprising that static solutions do exist, since they can be viewed as higher dimensional generalizations of a black hole in a background magnetic field .
It should be noted that the validity of the Schwarzschild approximation does not imply that all small black holes will Hawking evaporate. As noted in , if one fixes the total energy, the asymptotic AdS boundary conditions ensure that certain small black holes can be in stable equilibrium with their own Hawking radiation. However, sufficiently small black holes will still evaporate, so the AdS/CFT correspondence leads one to believe that this evaporation can be described by a unitary evolution.
Acknowledgements
This work was supported in part by NSF Grant PHY95-07065.
References
relax J. Maldacena, Adv. Theor. Math. Phys. 2 (1998) 231, hep-th/9711200. relax O. Aharony, S.S. Gubser, J. Maldacena, H. Ooguri, Y. Oz, Phys. Rept. 323 (2000) 183, hep-th/9905111. relax S. Hawking, Phys. Rev. D14 (1976) 2460. relax R. Gregory and R. Laflamme, Phys. Rev. Lett. 70 (1993) 2837, hep-th/9301052; Nucl. Phys. B428 (1994) 399, hep-th/9404071. relax R. M. Wald,“General Relativity,” U. of Chicago Press (1984). relax F. J. Ernst, J. Math. Phys. 17 (1976) 54. relax G. Horowitz, Class. Quant. Grav. 17 (2000) 1107, hep-th/9910082.
|
warning/0005/gr-qc0005098.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
Dilaton fields have experienced an impressive comeback in recent years in a broad range of gravitational theories. Motivated by their appearance in string theories scalar fields play an increasingly important role in modern physics. In the context of those theories, but also as a feature of any higher-dimensional theory of gravity, the concept of compactification has become a standard method in many models that leads inevitably to the occurrence of dilatons in the reduced action. It is the aim of this paper to discuss some common properties of two-dimensional two-dilaton theories (TDT) where at least one dilaton field is produced by dimensional reduction.
Historically the dilaton was introduced for the first time by Kaluza and Klein who proposed a five-dimensional gravity theory (KKT) to unify general relativity with electrodynamics . The scalar field created by reduction to 4 dimensions inspired Fierz and Jordan to invent the first Scalar-Tensor theory (STT) in 4D where the dilaton field was interpreted as a local gravitational coupling “constant”. Already Fierz investigated the connection of this theory with usual general relativity by conformal transformations. Later work of Brans and Dicke revived the theory which in the following will be called Jordan-Brans-Dicke theory (JBD). Recently the interest in STT has increased enormously due to the observation of accelerating galaxies with high redshift, indicating a positive cosmological constant . Again the transformation of that constant into a scalar field (“quintessence”) has been proposed .
A one-dilaton theory in two dimensions emerges naturally in connection with spherically reduced general relativity (SRG). The reformulations of general one-dilaton theories in D=2 as first order theories with torsion have led to various new insights, including e.g. the discovery of a conservation law . These results have been extended to the case of SRG with a massless scalar field minimally coupled in the 4D theory, i.e. the Einstein massless Klein-Gordon model (EMKG) .
To motivate the interest in TDT we briefly summarize already existing models that belong to this category:
* The most obvious example is spherically reduced EMKG. Whenever one deals with a scalar field in ordinary general relativity and demands spherical symmetry one arrives at a TDT in 2D, where the 4D scalar field may be interpreted as one of the two dilatons.
* The polarized Gowdy model is based on the existence of two commuting space-like Killing fields in a closed Einstein universe. Then toric reduction directly leads to a TDT. This example is of particular interest because here both dilatons in the 2D theory are part of four-dimensional geometry. The polarized Gowdy model (in contrast to SRG) allows to retain one degree of freedom of the gravity waves which is transferred into one of the dilatons.
* Another example is given by KKT. Having already one dilaton in 4 dimensions one ends up again with a TDT in 2D through spherical reduction. This is not equivalent to spherically reduced EMKG since the Kaluza dilaton in the four-dimensional theory couples non-minimally to gravity.
* STT and nonlinear gravity theories are mainly rooted in KKT, thus their connection to TDT is very similar. In addition it has been shown that nonlinear gravity theories are formally equivalent to STT (cf. e.g. ).
Finally, TDT in 2D may serve as useful toy models for TDT in 4D which may be necessary to describe the appearance of various scalar fields, encountered in cosmology (Higgs, inflaton, stringy dilaton(-s), quintessence). Up to now there exists no 4D STT with a single scalar field playing the role of all of these fields and certain dimensional reductions of e.g. a 11D supergravity can yield such theories by analogy to the $`542`$ reduction of spherically reduced KKT.
Thus, a dilaton field either may be produced by dimensional reduction (like in the effective theories for the Gowdy model or KKT) or it is introduced in the action as a generic scalar field (like in STT or the EMKG).
In section 2 we present the general framework of the TDT. The notions of Einstein form and Jordan form are introduced. As examples three significant physical applications are shown to fit into this framework.
In section 3 a useful classification scheme is invented distinguishing models that are simple and/or factorizable. We further investigate how conformal transformations affect these properties.
Section 4 is devoted to (conformally) simple factorizable theories. They can be treated in a first order form where a conservation law easily can be derived. Finally we examine the scaling properties of the conserved quantity.
In the Conclusions a table summarizes the various models we consider, together with their properties. Finally possible further applications are discussed.
## 2 General framework
The ansatz for the TDT
$`S_a`$ $`=`$ $`{\displaystyle _{M_2}}d^2x\sqrt{g}[V_0(X,Y)R+V_1(X,Y)_\alpha X^\alpha X+V_2(X,Y)_\alpha Y^\alpha Y`$ (1)
$`+V_3(X,Y)_\alpha X^\alpha Y+V_4(X,Y)+V_5(X,Y)f_m(S_n,_\alpha S_n,\mathrm{})]`$
satisfies diffeomorphism invariance in 2D and the following requirements:
* Two scalar dilaton fields $`X,Y`$ should appear in the 2D action.
* The action should be linear in the scalar curvature $`R`$ since terms with higher power in $`R`$ could be accommodated by modifying the arbitrary functions in (1) just like in theories with only one dilaton field . In order to have a nontrivial dilaton geometry the factor $`V_0`$ is assumed never to be a constant.
* The dilaton fields’ first derivatives enter quadratically multiplied by an arbitrary function of the dilatons ($`V_1`$ and $`V_2`$); in general, there will be a mixing between them ($`V_30`$).
* In addition, there is an arbitrary function of the dilaton fields, $`V_4`$, henceforth called “potential”.
* Finally, there are contributions from one or more “matter fields” $`S_n`$ which couple non-minimally to the dilatons whenever $`V_5\text{const.}`$
Our paper will mainly deal with the special case $`V_5=0`$, for simplicity, although when conformal transformations are discussed we will have to reconsider the matter part since the coupling function $`V_5`$ will change in general, having important implications for geodesics and hence for the global structure of the manifold.
### 2.1 Standard forms
In the following we restrict the function $`V_0(X,Y)`$ further, in accordance with our intention to examine dimensionally reduced gravity models. The application of conformal transformations in the higher-dimensional theory reveals two standard forms, which have the advantage that all models considered in the Introduction fit into one of them. Bearing in mind that at least one of the dilatons (here called $`X`$) possesses a conformal weight $`\alpha 0`$ (being part of the higher-dimensional metric), we cannot reach a form where $`V_0=const.`$ through this kind of conformal transformation.
#### 2.1.1 Einstein form
We call the first standard form “Einstein form”(EF) because it contains as the most important representative SRG in the Einstein frame in D=4.
It is given by
$$\begin{array}{c}S_E=_{M_2}d^2x\sqrt{g}[XR+V_1^E(X,Y)_\alpha X^\alpha X+V_2^E(X,Y)_\alpha Y^\alpha Y\hfill \\ \hfill +V_3^E(X,Y)_\alpha X^\alpha Y+V_4^E(X,Y)+V_5^E(X,Y)f_m(S_n,_\alpha S_n,\mathrm{})]\end{array}$$
(2)
and contains important special cases listed in table 1 at the end of the paper. A trivial subclass are one-dilaton theories where $`V_i^E=V_i^E(X)`$ and $`V_2^E=V_3^E=V_5^E=0`$.
#### 2.1.2 The Jordan form
Motivated by spherically reduced JBD in the Jordan frame, we call the second standard form “Jordan form” (JF). It reads
$$\begin{array}{c}S_J=_{M_2}d^2x\sqrt{g}[XYR+V_1^J(X,Y)_\alpha X^\alpha X+V_2^J(X,Y)_\alpha Y^\alpha Y\hfill \\ \hfill +V_3^J(X,Y)_\alpha X^\alpha Y+V_4^J(X,Y)+V_5^J(X,Y)f_m(S_n,_\alpha S_n,\mathrm{})].\end{array}$$
(3)
Indeed its most important representatives are general spherically reduced STT. There $`V_1^J=Y/(2X)`$, $`V_2^J=wX/Y`$, $`V_3^J=2`$, $`V_4=2Y+X\widehat{V}(Y)`$ and $`V_5=X`$. In JBD $`w=const.`$ is called “Dicke parameter”. $`\widehat{V}(Y)`$ is a scalar potential (which vanishes in JBD) and $`V_5`$ is chosen such that it amounts to minimal coupling of matter fields in the Jordan frame of four-dimensional STT.
### 2.2 Applications: Specific models
In this section we consider three significant models somewhat more in detail. They are all constructed through dimensional reduction of D-dimensional gravity theories by assuming the existence of (D-2) spacelike Killing fields. In the first case we start from the spherically symmetric D-dimensional Einstein-Hilbert action with one (in D dimensions minimally coupled) massless scalar field. In the second case we apply the spherical reduction scheme to STT in D = 4 without matter where the scalar field plays the role of one of the two dilatons in two dimensions. In a last example we reduce the pure 4D polarized Gowdy model that has cylindrical symmetry and therefore one gravitational degree of freedom.
#### 2.2.1 Spherically reduced Einstein gravity with massless scalar field
The D-dimensional (D$``$4) Einstein-Hilbert action including a massless scalar field $`Y`$ reads
$$S=_{M_D}d^Dx\sqrt{g}\left[R^{(D)}\kappa _\mu Y^\mu Y\right].$$
(4)
In D=4 Einstein gravity the constant $`\kappa `$ is taken to be $`16\pi G`$ with Newton’s constant $`G`$. If the D-dimensional spacetime $`M_D`$ is spherically symmetric, its metric can be written as<sup>1</sup><sup>1</sup>1We use the metric signature $`(+,,\mathrm{})`$. The indices $`\alpha ,\beta ,\gamma `$ take the values $`0,1`$ whereas the indices $`\kappa ,\lambda `$ run from $`2`$ to $`D1`$. Indices $`\mu ,\nu `$ run from $`0`$ to $`D1`$.
$$ds^2=g_{\alpha \beta }dx^\alpha dx^\beta X^2(r,t)g_{\kappa \lambda }dx^\kappa dx^\lambda ,$$
(5)
where $`g_{\alpha \beta }`$ is the metric of a two-dimensional Lorentz manifold $`M_2`$, $`g_{\kappa \lambda }`$ the metric of a (D-2)-sphere and $`X`$ the dilaton field. The scalar curvature $`R^{(D)}`$ of $`M_D`$ can be decomposed as (cf. e.g. )
$$R^{(D)}=R\frac{(D2)(D3)}{X^2}\left[1+_\alpha X^\alpha X\right]2(D2)\frac{\mathrm{}X}{X}$$
(6)
where $`R`$ on the right side is the curvature of $`M_2`$. To integrate out the isometric angular coordinates on the unit sphere $`S^{D2}`$ we only have to substitute the scalar curvature by the above expression and the measure by
$$\sqrt{(1)^{D+1}g_{M_D}}=\sqrt{g}\sqrt{(1)^Dg_{S^{D2}}}X^{D2}.$$
(7)
For later convenience we perform a field redefinition
$$X(D2)X^{1/(D2)}.$$
(8)
Up to a constant factor, the effective 2D action thus reads
$$S=\underset{M_2}{}d^2x\sqrt{g}\left[XR+\frac{(D3)}{(D2)}\frac{(X)^2}{X}\frac{(D3)}{(D2)}X^{\frac{(D4)}{(D2)}}\kappa X(Y)^2\right]$$
(9)
and obviously is of the Einstein form (2).
#### 2.2.2 Spherically reduced Scalar-Tensor theories
The four-dimensional STT action without matter is given by ($`\varphi ^+`$)
$$S_{STT}=_{M_4}d^4x\sqrt{g}\left[\varphi Rw(\varphi )\frac{_\mu \varphi ^\mu \varphi }{\varphi }+V(\varphi )\right].$$
(10)
Here $`\varphi `$ is the (positive) scalar field (STT field) that couples non-minimally to the metric. In the original KKT this non-trivial coupling was a result from a former reduction of a five-dimensional theory. In STT one “forgets” about this fact and instead employs this coupling by hand. In JBD $`w`$ is an arbitrary constant whereas in recent quintessence theories a dependence $`w(\varphi )`$ has been proposed. A phenomenological potential $`V(\varphi )`$ can be used to describe various cosmological scenarious . Spherical reduction occurs similar to the case of Einstein gravity. Replacing the scalar curvature by (6), using the field redefinition (8) and setting D=4 we can integrate out the angular coordinates on $`S^2`$, i.e. $`\theta ,\phi `$, to obtain the 2D action
$$\begin{array}{c}S=_{M_2}d^2x\sqrt{g}[\varphi (XR+\frac{(X)^2}{2X}\frac{1}{2})+2_\alpha \varphi ^\alpha X\hfill \\ \hfill w(\varphi )X\frac{(\varphi )^2}{\varphi }+XV(\varphi )].\end{array}$$
(11)
Here we have already performed a partial integration and divided by the overall factor $`4\pi `$. It is now convenient to apply a conformal transformation
$$g_{\mu \nu }g_{\mu \nu }\varphi ^1,XX\varphi ^1,$$
(12)
together with a field-redefinition $`\varphi =A(Y)`$, $`A`$ being a solution of the ordinary differential equation
$$\frac{d\mathrm{ln}A}{dY}=\left(w(A)+\frac{3}{2}\right)^{1/2}$$
(13)
that brings the action to the Einstein form
$$S_E=_{M_2}d^2x\sqrt{g}\left[XR+\frac{(X)^2}{2X}\frac{1}{2}X(Y)^2+\frac{X}{A(Y)}V(A(Y))\right].$$
(14)
In this form the mixed term $`_\alpha \varphi ^\alpha X`$ disappears. However, in the case of interaction with matter a complicated nonminimal coupling to the STT field arises. In the matterless case the STT field is seen simply to play the role of an additional scalar field with proper (nonminimal) coupling to the geometric dilaton $`X`$ in D=2.
#### 2.2.3 Gowdy model
The four-dimensional (polarized) Gowdy metric
$`ds^2`$ $`=`$ $`e^{2a(t,\theta )}\left(dt^2d\theta ^2\right)X(t,\theta )\left(e^{Y(t,\theta )}d\sigma ^2+e^{Y(t,\theta )}d\delta ^2\right).`$ (15)
describes a 4D spacetime that has 2 commuting Killing fields spanning a flat, spacelike isometry submanifold $`T^2S^1S^1`$ (locally). Moreover it is assumed that the whole spacetime $`M_4`$ is compact. Performing the integration over the isometric coordinates $`\sigma ,\delta `$ yields an effective two-dimensional action. For this reason we have to decompose the 4D scalar curvature $`R^{(4)}`$ into terms corresponding to $`T^2`$ and $`M_2`$, which is the complementary manifold, and terms produced by the embedding. This computation is done most conveniently in the vielbein frame<sup>2</sup><sup>2</sup>2Letters from the Latin alphabet denote vielbein indices, while Greek ones are reserved for coordinate indices. $`ds^2=\eta _{ab}e^ae^b\left(e^2\right)^2\left(e^3\right)^2`$. Quantities associated to $`T^2`$ or $`M_2`$ shall be assigned a tilde. We treat $`T^2`$ as two independent one-dimensional spaces $`S_1`$. Thus the relation between the vielbeine is given by
$$e^a=\stackrel{~}{e}^a,e^2=\sqrt{X}e^{\frac{Y}{2}}\stackrel{~}{e}^2,e^3=\sqrt{X}e^{\frac{Y}{2}}\stackrel{~}{e}^3.$$
(16)
Demanding vanishing torsion and metric compatibility on $`M_4,M_2`$ and $`T^2`$ the connection 1-form on $`M_4`$ is obtained:
$`\omega _b^a=\stackrel{~}{\omega }_b^a`$ , $`\omega _{\mathrm{\hspace{0.17em}\hspace{0.17em}3}}^2=\stackrel{~}{\omega }_{\mathrm{\hspace{0.17em}\hspace{0.17em}3}}^2=0`$ (17)
$`\omega _a^2=\left(\stackrel{~}{E}_a\sqrt{X}e^{\frac{Y}{2}}\right)\stackrel{~}{e}^2`$ , $`\omega _a^3=\left(\stackrel{~}{E}_a\sqrt{X}e^{\frac{Y}{2}}\right)\stackrel{~}{e}^3.`$
This is sufficient to calculate the scalar curvature
$$R^{(4)}=R2\frac{\mathrm{}X}{X}+\frac{1}{2}\frac{_\alpha X^\alpha X}{X^2}\frac{1}{2}_\alpha Y^\alpha Y$$
(18)
where $`R`$ again is the scalar curvature of $`M_2`$. We can put this result into the 4D Einstein-Hilbert action and then integrate over the isometric coordinates while decomposing the measure as $`\sqrt{g_{M_4}}=X\sqrt{g_{M_2}}`$. The effective 2D action divided by the (finite) volume of $`T^2`$ reads
$$S=_{M_2}d^2x\sqrt{g}\left[XR+\frac{(X)^2}{2X}\frac{1}{2}X(Y)^2\right].$$
(19)
It is interesting to note that the dilaton $`X`$ acquires the scaling factor of the 4D metric while the dilaton $`Y`$ represents the gravitational degree of freedom of the Gowdy model. Clearly this action is of the Einstein form (3). It can be shown that its variation leads to the same EOM as the ones from the original 4D action when the symmetry is introduced there. This point is nontrivial as witnessed by the reduced action resulting from warped metrics in Einstein gravity .
From the Gowdy line-element alone it is not clear why to define as dilaton fields $`X`$ and $`Y`$ as in (15). In principle other gauge choices are possible. But in this representation the $`X`$-field alone carries the scale factor and hence the geometric information (“radius”) of the Gowdy spacetime, while the $`Y`$-field represents a propagating degree of freedom (“graviton”). It is only in these variables (modulo trivial field redefinitions not mixing the dilatons) that the factorizability property, as defined below, is manifest.
## 3 Classification of TDT in 2D
In this section we will introduce useful notions, with respect to which we will classify TDT. As a word of warning we would like to emphasize the following point: Our definitions are not invariant under arbitrary field redefinitions.
However, for theories with only one dilaton coming from the higher-dimensional metric field redefinitions, leading to a mixing between the geometric dilaton and the “scalar field” dilaton in the new variables, are very inconvenient for two reasons: First, in general such transformations change the geodesics of testparticles<sup>3</sup><sup>3</sup>3The Christoffel symbols in the transformed geodesic equation $`\ddot{x}^\alpha +\mathrm{\Gamma }^\alpha {}_{\beta \gamma }{}^{}(\stackrel{~}{X},\stackrel{~}{Y})\dot{x}^\beta \dot{x}^\gamma =0`$ become dependent on the new matter dilaton $`\stackrel{~}{Y}`$. and hence the geometric properties of the manifold. Second, the quantization procedure used in is spoilt by a mixing of geometric and matter variables<sup>4</sup><sup>4</sup>4In the following we will apply conformal transformations on such models. The conformal invariance (if any) then permits a “mixing” of the variables..
For the case where both dilatons stem from the higher-dimensional metric such field redefinitions correspond to the choice of a different gauge for this metric and are therefore possible. Using this feature, by choosing a specific “gauge”, we will show below that a general class of such models always fits nicely into our classification scheme.
### 3.1 Definitions
We start the classification with some new definitions which prove useful:
Definition 1: A TDT in the EF (3) is called simple iff $`V_3^E=0`$.
Simple theories are models with no dynamical mixing between the dilaton fields and can be treated like a one-dilaton theory with dilaton field $`X`$ and an additional scalar matter field $`Y`$ coupled non-minimally in general. A main reason for highlighting such models is the possibility to bring them into a first order form (see below). This means that one can easily apply the quantization scheme developed in this framework . Note that it is always possible to redefine the dilaton fields such that the diagonal term vanishes. E.g. one could use the nonlinear transformation $`X\stackrel{~}{X}^\alpha f(\stackrel{~}{\varphi }),\varphi \stackrel{~}{X}^{1\alpha }f^1(\stackrel{~}{\varphi });\alpha =(w+1)/(w+3/2)`$ to make spherically reduced JBD ((11) for $`w=const.`$) simple. However, as we have already mentioned, such a redefinition is only allowed if both dilatons stem from the higher-dimensional metric.
Definition 2: A TDT in any given form is called factorizable iff $`V_1(X,Y)=f_1(X)g(Y)`$ and $`V_0(X,Y)=f_2(X)g(Y)`$. We assume that (at least) $`X`$ is part of a higher-dimensional metric, while $`Y`$ can either be also part of this metric or a “true” scalar field.
Factorizable theories permit a simple geometrical interpretation of $`X`$ as “classical dilaton field” in the 2D model, since there is a common $`Y`$-factor $`g(Y)`$ in front of the first two terms of (1). In the EF this property translates into $`V_1^E=V_1^E(X)`$.
In the following we discuss a class of models (including all models considered in this paper) where the original D-dimensional spacetime $`M_D`$ contains one or two maximally symmetric, spacelike subspaces $`S_{(1,2)}`$(e.g. $`S^1,S^2S^3`$) such that locally $`M_DM_2S_{(1)}S_{(2)}`$.
In the first case one dilaton arises from the metric and the other from the matter Lagrangian. These models are trivially factorizable (e.g. JBD, SRG).
In the second case both dilatons stem from the metric which can then be written in the form <sup>5</sup><sup>5</sup>5As explained above this is just a specific choice of gauge for the metric of $`M_D`$.
$$ds^2=g_{\alpha \beta }dx^\alpha dx^\beta X^2[e^{2Y/d_1}d^2\mathrm{\Omega }_{(1)}+e^{2Y/d_2}d^2\mathrm{\Omega }_{(2)}]$$
(20)
where $`d^2\mathrm{\Omega }_{(1,2)}`$ are the metrics of the two subspaces with dimensions $`d_1,d_2`$ respectively, $`g_{\alpha \beta }`$ is the metric of the reduced two-dimensional spacetime and $`X(x^\alpha ),Y(x^\alpha )`$ are the dilatons. The factors in the definition of $`Y`$ are chosen such that the measure only depends on $`X`$ which carries the conformal weight:
$$\sqrt{(1)^{D+1}g_{M_D}}=\sqrt{g_{M_2}}\sqrt{(1)^{d_1+d_2}g_{(1)}g_{(2)}}X^{d_1+d_2}.$$
(21)
The reduced two-dimensional action divided by the volumes of the subspaces reads
$`S_2`$ $`=`$ $`{\displaystyle _{M_2}}\sqrt{g_{M_2}}X^{d_1+d_2}[R+(d_1+d_2)(d_1+d_21){\displaystyle \frac{(X)^2}{X^2}}`$ (22)
$`({\displaystyle \frac{1}{d_1}}+{\displaystyle \frac{1}{d_2}})(Y)^2{\displaystyle \frac{R_{(1)}}{X^2}}e^{2Y/d_1}{\displaystyle \frac{R_{(2)}}{X^2}}e^{2Y/d_2}]`$
where a $`\mathrm{}X`$ term has been partially integrated using the measure (21). $`R_{(1,2)}`$ are the (constant) scalar curvatures of the corresponding subspaces. As the kinetic term of the $`X`$-dilaton does not depend on $`Y`$ we conclude that also this class of models is factorizable, containing polarized Gowdy ($`T^2S^1S^1`$) and spherically reduced KKT ($`S^1S^2`$). From (22) one can also see that these models are simple. Thus, by field redefinitions – corresponding to the choice of adapted coordinates in the higher-dimensional metric – it is always possible to bring them into the EF and make the mixed kinetic term vanish.
From JBD we know that it is conformally equivalent to Einstein gravity modulo the aforementioned problem of coupling to matter and a potential change of geodesic behavior. We will call such theories conformally related:
Definition 3: Two theories are called conformally related, iff there exists a conformal transformation (in the higher-dimensional theory) between them.
It is an interesting task to investigate whether a TDT given in the JF is conformally related to a simple model, since such models are particularly easy to treat and interpret. However, not all models allow a simplification through conformal transformations.
Definition 4: If a non-simple model is conformally related to a simple model we call it conformally simple.
By explicit calculation we will show below that all factorizable, non-simple models where the functions $`V_0,V_3`$ are monomials are conformally simple, provided that only one of the dilatons carries a conformal weight.
We would like to emphasize that conformally related theories represent dynamically inequivalent models in general. A simple example is the CGHS-model which can either be introduced by complete spherical reduction from an infinite-dimensional Einstein-Hilbert action (cf. eq. (9) for $`D\mathrm{}`$) or by the requirement of scale-invariance in the 2D action:
$`V_0^{CGHS}`$ $`=`$ $`X,`$ $`V_3^{CGHS}`$ $`=`$ $`0,`$ $`V_1^{CGHS}`$ $`=`$ $`1/X,`$ $`V_4^{CGHS}`$ $`=`$ $`X,`$ $`V_2^{CGHS}`$ $`=`$ $`0,`$ $`V_5^{CGHS}`$ $`=`$ $`0.`$ (23)
This action is invariant under a constant rescaling $`X\lambda X`$. Through an intrinsically 2D conformal transformation with a conformal factor $`\mathrm{\Omega }=X^{1/2}`$ one can get rid of the $`V_1`$-term and the transformed theory describes flat spacetime. Thus the conformally related global structures are profoundly different: The Black Hole singularity of CGHS has disappeared. Also for any other theory important properties of the spacetime such as the 2D curvature and geodesic (in)completeness can be changed by a conformal transformation .
Despite of this, conformal transformations are frequently used in the literature on quantization of 2D dilaton gravity (cf. e.g. ) or JBD (cf. e.g. ) although by now even some of the proponents of this method have (re)discovered this subtlety . The issue of (in)equivalence of conformal frames has a long history of confusion, as pointed out in (see also references therein and references 28,29 of for positive and negative examples).
It is necessary to bear in mind that most 2D models are dimensionally reduced theories which follow from a physically motivated higher-dimensional model. Another alternative is that they are merely toy models. In both cases a conformal transformation changing the global structure leads to a different theory. In the first case no longer a 2D equivalent of the original theory (the “correct” conformal frame is known) is described. In the second case, one could have started from the transformed toy model instead of introducing an “auxiliary” toy model (one could have introduced the “correct” conformal frame from the very beginning). Of course, from a technical point of view, conformal transformations are very useful and indeed will be employed amply below, if they only represent an intermediate step (especially in the context of a classical theory: For the quantum case the frame where the quantization is performed must be the “correct” one under all circumstances).
In the following we consider conformal transformations
$$g_{\alpha \beta }g_{\alpha \beta }\mathrm{\Omega }^2,\sqrt{g}\sqrt{g}\mathrm{\Omega }^2,XX\mathrm{\Omega }^\alpha ,YY,$$
(24)
assuming that $`X`$ has a conformal weight of $`\alpha ^{}`$ and $`Y`$ has conformal weight zero. This applies to models in the JF where only $`X`$ stems from dimensional reduction. To include a larger class of models we generalize the JF by allowing $`V_0`$ to be an arbitrary function of $`Y`$:
$$V_0(X,Y)=Xv_0(Y).$$
(25)
Since we want to answer the question whether a model is conformally related to a simple one we have to impose the condition $`V_3^E=0`$ after the (higher-dimensional) conformal transformation. The choice of the conformal factor
$$\mathrm{\Omega }=(v_0(Y))^{1/\alpha }$$
(26)
is necessary to bring the action in the EF. The first term in (1) as a consequence of the identity, valid under the conformal transformation (24),
$$RR\mathrm{\Omega }^2+2\frac{_\gamma ^\gamma \mathrm{\Omega }}{\mathrm{\Omega }}2\frac{_\gamma \mathrm{\Omega }^\gamma \mathrm{\Omega }}{\mathrm{\Omega }^2}$$
(27)
produces the first term in the EF (2), plus further “kinetic” terms. Note that $`\mathrm{\Omega }`$ must be $`C^1`$, manifestly positive and invertible with respect to $`Y`$. Its inverse will be denoted by
$$Y=f(\mathrm{\Omega }).$$
(28)
### 3.2 Conformally simple TDT
In the following steps we will restrict ourselves to a smaller subset of TDT having the advantage of simplifying calculations drastically while still being general enough to include all “physical” models considered so far (and more). Using (27) and dropping a boundary term the action (1) after the conformal transformation (24) with conformal factor (26) becomes
$$\begin{array}{c}S_c=_{M_2}d^2x\sqrt{g}\hfill \\ \hfill [XR2_\gamma \mathrm{\Omega }^\gamma X+V_1\mathrm{\Omega }^{2\alpha }(\alpha ^2X^2\frac{(\mathrm{\Omega })^2}{\mathrm{\Omega }^2}2\alpha X\frac{_\gamma X^\gamma \mathrm{\Omega }}{\mathrm{\Omega }}+(X)^2)\\ \hfill +V_2f_{}^{}{}_{}{}^{2}(\mathrm{\Omega })^2\alpha V_3Xf^{}\mathrm{\Omega }^{\alpha 1}(\mathrm{\Omega })^2+V_3\mathrm{\Omega }^\alpha f^{}_\gamma X^\gamma \mathrm{\Omega }+V_4\mathrm{\Omega }^2].\end{array}$$
(29)
Note that in $`V_i(X,Y)`$ one has to replace $`XX\mathrm{\Omega }^\alpha `$ and $`Yf(\mathrm{\Omega })`$ as defined by (28).
Conformal simplicity requires the vanishing of the mixed term $`_\gamma Y^\gamma X`$. This yields a first order differential equation for the function $`f`$,
$$f^{}(\mathrm{\Omega })=\frac{2}{V_3}\left(\alpha V_1(X\mathrm{\Omega }^\alpha ,\mathrm{\Omega })X\mathrm{\Omega }^{\alpha 1}+v_0(\mathrm{\Omega })\mathrm{\Omega }^1\right),$$
(30)
which already restricts the functions $`V_1,V_3`$ severely because the l.h.s. of (30) is $`X`$-independent by construction.
The convenient ansatz, to be used as of now,
$$V_1=X^1v_1(\mathrm{\Omega }),V_3=v_3(\mathrm{\Omega })$$
(31)
is sufficient to satisfy (30) although not necessary. Next we impose factorizability on the original model which together with (31) implies
$$V_1(X,\mathrm{\Omega })=v_1X^1\mathrm{\Omega }^\alpha ,v_1.$$
(32)
Assuming monomiality for $`V_3`$ by
$$v_3(\mathrm{\Omega })=v_3\mathrm{\Omega }^\beta ,\beta ,v_3^{}$$
(33)
the differential equation (30) establishes a four-parameter solution<sup>6</sup><sup>6</sup>6The solution for the conformal factor of course only depends on two real parameters, $`c`$ and $`\gamma `$, but the original action and the transformed one contain all four parameters $`\alpha `$, $`\beta `$, $`v_1`$ and $`v_3`$.
$$Y=f(\mathrm{\Omega })=2\mathrm{\Omega }^{\alpha \beta }\frac{1+\alpha v_1}{v_3(\alpha \beta )}=:c\mathrm{\Omega }^\gamma ,\alpha \beta .$$
(34)
This equation implies monomiality of $`v_0(Y)`$, too:
$$v_0(Y)=c^\delta Y^\delta ,\delta :=\frac{\alpha }{\alpha \beta }.$$
(35)
Thus, with the assumptions made above $`v_0(Y)`$ is completely determined.
The resulting action may be written as
$`S_{cs}`$ $`=`$ $`{\displaystyle _{M_2}}d^2x\sqrt{g}[XR+v_1{\displaystyle \frac{(X)^2}{X}}+V_4({\displaystyle \frac{X}{v_0(Y)}},Y)v_0(Y)^{2/\alpha }`$ (36)
$`+(Y)^2[V_2({\displaystyle \frac{X}{v_0(Y)}},Y)\alpha \delta ^2{\displaystyle \frac{X}{Y^2}}(2+v_1\alpha )]],`$
which is conformally related to the original TDT action
$`S_{TDT}`$ $`=`$ $`{\displaystyle _{M_2}}d^2x\sqrt{g}[Xv_0(Y)R+v_0(Y)v_1{\displaystyle \frac{(X)^2}{X}}+V_4(X,Y)`$ (37)
$`+V_2(X,Y)(Y)^2+v_3{\displaystyle \frac{cv_0(Y)}{Y}}_\gamma X^\gamma Y].`$
We recall the meaning of the four parameters: $`\alpha `$ is the conformal weight of the (geometric) dilaton $`X`$, $`\delta `$ (or $`\beta `$ or $`\gamma `$) defines the power of the monomial $`v_0(Y)`$, $`v_1`$ and $`v_3`$ are constants entering the corresponding functions ($`c`$ is defined in (34) and depends on all these constants).
Thus we have shown that a TDT satisfying (24)-(35) is conformally simple, provided that $`Y`$ is positive everywhere. We emphasize that factorizability is conserved under this conformal transformation, as can be seen from (36), and hence is an independent property indeed.
The most important examples are STT, which are known to be conformally simple :
$$\alpha =2,\delta =1(\beta =0),v_1=1/2,v_3=2.$$
(38)
The relation between the conformal factor and the STT field is $`f(\mathrm{\Omega })=Y=\mathrm{\Omega }^2`$, a well-known result. Note that the whole class of STT is given by a single point in the four-dimensional parameter space of possible conformally simple actions with conformal factor given by (34). Thus, despite of the various restrictions which led to (34), the set of conformally simple theories described by the action (36), resp. (37), is very large.
It is straightforward to construct toy models which possess all possible combinations of factorizability and (conformal) simplicity.
## 4 First Order Formalism
Dilaton models that are (conformally) simple as well as factorizable have the important property that they may be written in an equivalent first order form . This is manifest in the EF. In this case one of the dilatons (namely the scalar field $`Y`$) is disentangled from the gravitational sector, in the sense that no mixed kinetic term appears. The geometric part of the Lagrangian (including the dilaton $`X`$) can be brought to first order in its derivatives by introducing Cartan variables and auxiliary fields $`X^a`$. The zweibein basis is expressed in light-cone coordinates<sup>7</sup><sup>7</sup>7We choose a representation $`(0,1)(,+)`$. Light-cone indices are adorned with bars. $`e^{}=\frac{1}{\sqrt{2}}\left(e^0e^1\right)`$. The invariant volume element in this frame is given by $`d^2x\sqrt{g}=d^2x(e)=e^{}e^+`$. The Levi-Civitá symbol $`\epsilon ^{\overline{a}\overline{b}}`$ is defined by<sup>8</sup><sup>8</sup>8The $`\epsilon `$-symbol in ordinary coordinates is defined by $`\epsilon ^{01}=1`$. $`\epsilon ^+=1`$. The connection 1-form $`\omega _{\overline{b}}^{\overline{a}}`$ which is proportional to $`\epsilon _{\overline{b}}^{\overline{a}}`$ becomes $`\omega _{\overline{b}}^{\overline{a}}=\omega \text{diag}(1,1)`$. Thus the 2D scalar curvature can be written as $`d^2x\sqrt{g}R=2d\omega `$. According to the second reference of we add the terms $`X_{\overline{a}}T^{\overline{a}}=X^{}(d+\omega )e^++X^+(d\omega )e^{}`$ to the action where $`T^{\overline{a}}`$ is the torsion associated with the connection $`\omega `$. The EF action (2) divided by $`(2)`$ becomes equivalent to
$$\begin{array}{c}S_{FO}=\underset{M_2}{}[X^{}(d+\omega )e^++X^+(d\omega )e^{}+Xd\omega +e^{}e^+V_1^E\left(X\right)X^{}X^+\hfill \\ \hfill \frac{1}{2}V_2^E(X,Y)dYdY\frac{1}{2}e^{}e^+(V_4^E(X,Y)+V_5^E(X,Y)f_m)],\end{array}$$
(39)
where we have included also the matter term. The fields $`X^{}`$ and $`X`$ are determined from the EOM produced by the variation of the Cartan variables. The whole set of EOM derived from (39) is equivalent to the one obtained from the original action . Actually $`X^\pm `$ and $`\omega `$ may be eliminated by algebraic EOM from (39).
For a theory in the EF the corresponding first order formulation has many advantages, especially at the quantum level, where e.g. the geometric degrees of freedom of SRG can be integrated exactly . Here we will only use one important result, namely the existence of a conservation law that can be derived in a particularly simple way in this context . Taking appropriate linear combinations of the EOM derived from (39) with an integrating factor $`I(X)=e^{^XV_1^E\left(X^{}\right)𝑑X^{}}`$ we obtain a relation of the type
$$I(X)_\alpha \left(X^{}X^+\right)I(X)_\alpha X\left(V_1^E\left(X\right)X^{}X^+\frac{V_4^E(X,Y)}{2}\right)+W_\alpha =0.$$
(40)
Splitting the potential $`V_4^E`$ into two terms $`V_4^E=V_4^{\left(g\right)}\left(X\right)+V_4^{\left(Y\right)}(X,Y)`$ we obtain the conservation law
$$_\alpha 𝒞=_\alpha 𝒞^{\left(g\right)}+W_\alpha =0$$
(41)
where $`𝒞^{\left(g\right)}=X^{}X^+I(X)+\frac{1}{2}^XV_4^{\left(g\right)}\left(X^{}\right)I(X^{})𝑑X^{}`$. From (41) the 1-form $`W_\alpha =W_\alpha ^{(Y)}+W_\alpha ^{(m)}`$ is trivially exact. Its separation into matter and $`Y`$-terms depends on the coupling function $`V_5^E`$ that can have an arbitrary $`Y`$-dependence. The components $`W_\alpha ^{\left(Y\right)}`$ are given by
$$\begin{array}{c}W_\alpha ^{\left(Y\right)}=I(X)[\frac{_\alpha X}{2}V_4^{\left(Y\right)}(X,Y)+\hfill \\ \hfill +\frac{V_2^E(X,Y)}{\left(e\right)^2}\{Y^{}Y^+\left(_\alpha X\right)\left(e\right)(Y^{}X^++Y^+X^{})_\alpha Y\}]\end{array}$$
(42)
where $`Y^{}=\epsilon ^{\alpha \beta }e_\beta ^{}\left(_\alpha Y\right)`$. The analogous expression for the matter part becomes
$$W_\alpha ^{\left(m\right)}=I(X)\frac{V_5^E(X,Y)}{2}\left[\left(_\alpha X\right)f_m\left(e\right)\epsilon _{\alpha \beta }\left(X^{}\frac{f_m}{e_\beta ^{}}+X^+\frac{f_m}{e_\beta ^+}\right)\right].$$
(43)
It has been shown that this conservation law is connected to the energy conservation of the model considered. More precisely, the geometric part $`𝒞^{(g)}`$ is proportional to a mass-aspect function $`m_{eff}(r,t)`$ which is the sum of the ADM mass and the energy fluxes given by the matter- and $`Y`$-contributions. Since we have not specified as yet the functions $`V_1^E,V_2^E,V_4^E,V_5^E`$ we have generalized that conservation law from EMKG to all factorizable (conformally) simple theories.
It is interesting to check its behavior under a (constant) Weyl-rescaling $`g_{\alpha \beta }\lambda g_{\alpha \beta }`$ in the higher-dimensional theory (taking into account the conformal weight of the geometric dilaton $`X`$). For general spherically reduced models we obtain a scaling weight
$$\text{sw}(𝒞)|_{SR}=D3,$$
(44)
where D is the higher dimension (e.g. D=4 for ordinary SRG), provided that the matter part and $`Y`$-part couple linearly to the dilaton $`X`$. In all other cases (the coupling to matter or $`Y`$ is different or the potentials differ from SRG) the conserved quantity does not have a well defined conformal weight with respect to a global conformal transformation in the higher-dimensional theory (e.g. the CGHS model (23) contains minimally coupled scalar fields instead of linearly coupled ones).
## 5 Conclusions
In this paper we have investigated TDT produced by dimensional reduction. First we introduced two standard forms, namely the Einstein form (2) and the Jordan form (3), covering all models considered in this paper. The useful properties factorizability, simplicity and conformal simplicity have been defined. Since there seems to be still confusion in the literature (for a selected list of such papers cf. e.g. the review article ) we have emphasized the physical inequivalence of conformally related theories.
We could show that all investigated models can be derived by dimensional reduction of a spacetime with one or two maximally symmetric subspaces<sup>9</sup><sup>9</sup>9In some cases like JBD one “forgets” about the origin of one of the dilatons. This manifests itself in setting its conformal weight equal to zero.. From this we conclude that:
* All these models are factorizable.
* Models with one geometric dilaton are at least conformally simple.
* Models with two geometric dilatons are simple (in adapted coordinates).
The second result was obtained by explicitly mapping a non-simple TDT in the JF (with $`V_0,V_3`$ monomials) onto a simple TDT in the EF through a conformal transformation. Thereby we could also show that factorizability is conserved under such a conformal transformation. It is not yet clear if theories with more complicated Killing-orbits are still factorizable. This could be the subject of further work.
Our subsequent investigations were restricted to the subclass of (conformally) simple factorizable theories. In the EF these models allow a first order formulation, and by straightforward application of previous work an absolute conservation law (41) could be established.
By investigating the scaling weight of the conserved quantity under global conformal transformations in the higher-dimensional theory for SRG the intuitively expected result (44) was obtained. Moreover it was clarified that a necessary condition for a definite scaling weight of the conserved quantity was linear coupling of the matter fields and the second dilaton $`Y`$ to the (geometric) dilaton $`X`$, i.e. to the dilaton with a non-vanishing scaling weight.
Apart from the obvious applications (namely a 2D description of various higher-dimensional models considered in this paper) TDT serve as toy models for D=4 theories with two dilaton fields and as a basis for generalizations to models with more than two dilatons. Compactification of e.g. D=11 supergravity can yield two or more dilaton fields, and up to now no satisfactory cosmological theory with a single scalar field (which serves e.g. as inflaton and quintessence) is known. Here one may hope that – as in the case of the nonvanishing cosmological constant (or quintessence?) – further input may be provided by the enormously increasing amount of astrophysical data to be expected for the near future. If the need for more dilaton fields should arise we believe that similar structures in the classification of such models will appear.
Although the physically relevant TDT are related to dimensional reduction it could be of interest to focus on an intrinsically 2D treatment of TDT. Such an investigation would e.g. involve a conformal transformation where no conformal weight is attributed to both dilatons and suggest the introduction of a “true” Jordan frame, i.e. a conformal frame where no dilaton at all is coupled to the scalar curvature.
At the quantum level the next step should be a Hamiltonian analysis and BRST quantization. Similarities to the analysis of non-minimally coupled scalars interacting with a one-dilaton theory which is based upon the simpler results obtained for the minimally coupled case may well occur. In fact, for simple factorizable theories the constraint algebra is already known and differs only slightly from the simpler algebra obtained in . Non-simple, but conformally simple factorizable models fit into this theoretical frame only through a conformal transformation. Thus it will be an interesting task to investigate the action of a conformal transformation on the constraint algebra. This would provide a basis of (path integral) quantization of all conformally simple factorizable TDT.
## Acknowledgements
The authors have benefitted from discussions with D. Schwarz. One of the authors (D.G.) would like to thank P. Hübner for drawing his attention to Gowdy models. We are also grateful to the referee for his detailed criticism which led to an improved manuscript.
This research has been supported by Project P14650-TPH of the Austrian Science Foundation (Österreichischer Fonds zur Förderung der wissenschaftlichen Forschung).
|
warning/0005/math0005079.html
|
ar5iv
|
text
|
# Classification of Equivariant Real Vector Bundles over a Circle
## 1. Introduction
In \[CKMS99\] we classified equivariant *complex* vector bundles over a circle, and in this paper we classify equivariant *real* ones. The argument developed in this paper is similar to that in \[CKMS99\] but is rather more complicated. The complexity arises from two aspects: one is topology and the other is representation theory. For instance, any (nonequivariant) complex line bundle over a circle is trivial while there are two non-isomorphic real line bundles, that is, the Hopf line bundle and the trivial one. This is an evidence of the topological complexity in the real case. As for representation theory, it is recognized in general that real representation theory is more complicated than complex representation theory.
Let us introduce some notation to state our results. Let $`G`$ be a compact Lie group and let $`\rho :GO(2)`$ be an orthogonal representation. The unit circle of the corresponding $`G`$-module is denoted by $`S(\rho )`$. It is well known that any circle with $`G`$-action is equivalent to $`S(\rho )`$ for some $`\rho `$. We set $`H=\rho ^1(1)`$, so that $`H`$ acts trivially on $`S(\rho )`$ and the *fiber $`H`$-module* of a real $`G`$-vector bundle over $`S(\rho )`$ is determined uniquely up to isomorphism.
Let $`\mathrm{Irr}(H)`$ be the set of characters of irreducible real $`H`$-modules. It has a $`G`$-action defined as follows: For $`\chi \mathrm{Irr}(H)`$ and $`gG`$, $`{}_{}{}^{g}\chi \mathrm{Irr}(H)`$ is defined by $`{}_{}{}^{g}\chi (h)=\chi (g^1hg)`$ for $`hH`$. Since a character is a class function, the isotropy subgroup $`G_\chi `$ of $`G`$ at $`\chi \mathrm{Irr}(H)`$ contains $`H`$. We choose and fix a representative from each $`G`$-orbit in $`\mathrm{Irr}(H)`$ and denote the set of those representatives by $`\mathrm{Irr}(H)/G`$. Denote by $`\mathrm{Vect}_G(S(\rho ))`$ the set of isomorphism classes of real $`G`$-vector bundles over $`S(\rho )`$ and by $`\mathrm{Vect}_{G_\chi }(S(\rho ),\chi )`$ the subset of $`\mathrm{Vect}_{G_\chi }(S(\rho ))`$ with a multiple of $`\chi `$ as the character of fiber $`H`$-modules. They are semi-groups under Whitney sum. The decomposition of a $`G`$-vector bundle into the $`\chi `$-isotypical components induces an isomorphism
$$\mathrm{Vect}_G(S(\rho ))\underset{\chi \mathrm{Irr}(H)/G}{}\mathrm{Vect}_{G_\chi }(S(\rho ),\chi ),$$
see \[CKMS99, Section 2\]. This reduces the study of $`\mathrm{Vect}_G(S(\rho ))`$ to that of $`\mathrm{Vect}_{G_\chi }(S(\rho ),\chi )`$, and since $`\chi `$ is $`G_\chi `$-invariant and $`G_\chi `$ is again a compact Lie group, we are led to study $`\mathrm{Vect}_G(S(\rho ),\chi )`$ where $`\chi `$ is $`G`$-invariant, namely $`{}_{}{}^{g}\chi =\chi `$ for all $`gG`$.
Let $`E\mathrm{Vect}_G(S(\rho ),\chi )`$. The fiber $`H`$-module of $`E`$ has a multiple of $`\chi `$ as the character by definition. In fact, the fiber $`E_z`$ of $`E`$ over a point $`zS(\rho )`$ is a real module of the isotropy subgroup $`G_z`$ at $`z`$, and unless $`\rho (G)SO(2)`$, $`G_z`$ properly contains $`H`$ for some $`z`$. It turns out that these fiber $`G_z`$-modules almost distinguish elements in $`\mathrm{Vect}_G(S(\rho ),\chi )`$. To be more specific, we shall introduce some more notation. Unless $`\rho (G)SO(2)`$, $`\rho (G)`$ is $`O(2)`$ or a dihedral group $`D_n`$ of order $`2n`$ for some positive integer $`n`$. We identify $`S(\rho )`$ with the unit circle of the complex line $``$, so that the dihedral group $`D_n`$ is generated by the rotation through an angle $`2\pi /n`$ and the reflection about the $`x`$-axis. Then the isotropy subgroups $`G_z`$ at $`z=1`$ (and $`z=e^{\pi i/n}`$ when $`\rho (G)=D_n`$) contain $`H`$ as an index two subgroup unless $`\rho (G)SO(2)`$. For a group $`K`$ containing $`H`$ we denote by $`\mathrm{Rep}(K,\chi )`$ the set of isomorphism classes of real $`K`$-modules whose characters restricted to $`H`$ are multiples of $`\chi `$. The set $`\mathrm{Rep}(K,\chi )`$ is a semi-group under direct sum. Restriction of elements in $`\mathrm{Vect}_G(S(\rho ),\chi )`$ to fibers at $`1`$ (and $`\mu =e^{\pi i/n}`$ when $`\rho (G)=D_n`$) yields a semi-group homomorphism
$$\mathrm{\Gamma }:\mathrm{Vect}_G(S(\rho ),\chi )\{\begin{array}{cc}\mathrm{Rep}(H,\chi ),\hfill & \text{if }\rho (G)SO(2),\hfill \\ \mathrm{Rep}(G_1,\chi ),\hfill & \text{if }\rho (G)=O(2),\hfill \\ \mathrm{Rep}(G_1,G_\mu ,\chi ),\hfill & \text{if }\rho (G)=D_n,\hfill \end{array}$$
where $`\mathrm{Rep}(G_1,G_\mu ,\chi )`$ denotes the subsemi-group of $`\mathrm{Rep}(G_1,\chi )\times \mathrm{Rep}(G_\mu ,\chi )`$ consisting of pairs of the same dimension. The map $`\mathrm{\Gamma }`$ can also be defined in the complex case and is proved to be an isomorphism in \[CKMS99, Proposition 6.2\]. However, $`\mathrm{\Gamma }`$ is not always an isomorphism in the real case, for instance the two non-isomorphic line bundles over a circle (when $`G`$ is trivial) mentioned before have obviously the same $`\mathrm{\Gamma }`$ image. Nevertheless it turns out that $`\mathrm{\Gamma }`$ is an isomorphism in most cases. Remember that $`\chi `$ is called of *real*, *complex*, or *quaternion* type if the $`H`$-endomorphism algebra of the irreducible real $`H`$-module with $`\chi `$ as the character is isomorphic to $``$, $``$, or $``$ respectively, and that $`\chi `$ is called $`K`$-*extendible* if it extends to a character of a group $`K`$ containing $`H`$. There are two cases in which the classification works somewhat exceptionally.
*Case A:* $`\rho (G)SO(2)`$ and $`\chi `$ is of real type.
*Case B:* $`\rho (G)=D_n`$, $`\chi `$ is of real type and neither $`G_1`$\- nor $`G_\mu `$-extendible.
###### Theorem 1.1.
Except for Cases A and B, the semi-group homomorphism $`\mathrm{\Gamma }`$ is an isomorphism. In Case A or B, $`\mathrm{\Gamma }`$ is a two to one map; more precisely, there is a free involution on $`\mathrm{Vect}_G(S(\rho ),\chi )`$ given by tensoring with a nontrivial $`G`$-line bundle (with trivial fiber $`H`$-module) and $`\mathrm{\Gamma }`$ induces an isomorphism on the orbit space.
Theorem 1.1 reduces the study of $`\mathrm{Vect}_G(S(\rho ),\chi )`$ to that of real representation theory, especially to the study of $`\mathrm{Rep}(K,\chi )`$ where $`K`$ is a compact Lie group containing $`H`$ as an index two subgroup and $`\chi `$ is $`K`$-invariant. The complexity of real representation theory emerges here. Namely, the number of $`K`$-extensions of $`\chi `$ can be zero, one, or two, while it is always two in the complex case. Combining this observation with Theorem 1.1, one sees that the semi-group structures on $`\mathrm{Vect}_G(S(\rho ),\chi )`$ are of five types depending on $`\rho (G)`$ and $`\chi `$ (see Theorem 5.1) while they are of three types in the complex case (see \[CKMS99, Theorem B\]).
The paper is organized as follows. In Section 2 we shall determine the semi-group structure of the target space of the semi-group homomorphism $`\mathrm{\Gamma }`$. For this we need to study some representation theory, especially on extensions of representations. The semi-group structure of the target space of $`\mathrm{\Gamma }`$ is given in Lemma 2.2 and Lemma 2.3.
In section 3 we prove Theorem 1.1 except for Cases A and B. Indeed, Proposition 3.2 shows that $`\mathrm{\Gamma }`$ is always surjective, and Proposition 3.3 shows that $`\mathrm{\Gamma }`$ is injective except for Cases A and B.
The proof for Cases A and B are treated in Section 4. Case A can be easily proved by reducing the case to nonequivariant case. For Case B we consider possible complex $`G`$ vector bundle structures on an element of $`\mathrm{Vect}_G(S(\rho ),\chi )`$. Then we use the classification results of complex $`G`$ vector bundles over a circle in \[CKMS99\], and some counting argument to finish the proof for Case B.
The semi-group structure of $`\mathrm{Vect}_G(S(\rho ),\chi )`$ is given in Theorem 5.1 of Section 5 by figuring out the generators with their relations. Theorem 6.2 in Section 6 shows which of the generators of $`\mathrm{Vect}_G(\mathrm{\S }(\rho ),\chi )`$ are trivial.
## 2. Ingredients from representation theory
We shall determine the semi-group structure of the target space of $`\mathrm{\Gamma }`$. For that, the following lemma from representation theory plays a key role. Throughout this section, $`H`$ is an index two normal subgroup of a group $`K`$ and $`U`$ is a real irreducible $`H`$-module with $`K`$-invariant character. By the type of $`U`$ we mean the type of the character of $`U`$. As is well known, any $`K`$-extension of $`U`$ appears in $`\mathrm{ind}_H^KU`$ as a direct summand at least once, which follows from the Frobenius reciprocity, and $`\mathrm{res}_H\mathrm{ind}_H^KU2U`$ because $`H`$ is of index two and the character of $`U`$ is $`K`$-invariant.
###### Lemma 2.1.
(1) Suppose $`U`$ is $`K`$-extendible.
1. If $`U`$ is of real type, there are two mutually non-isomorphic $`K`$-extensions of real type.
2. If $`U`$ is of complex type, there are either two mutually non-isomorphic $`K`$-extensions of complex type or unique $`K`$-extension of real type, not both.
3. If $`U`$ is of quaternionic type, there are either two mutually non-isomorphic $`K`$-extensions of quaternionic type or unique $`K`$-extension of complex type, not both.
In any case, if $`U`$ has two $`K`$-extensions, then one is isomorphic to the tensor product of the other with the nontrivial real $`K`$-module of dimension one with trivial $`H`$-action.
(2) Suppose $`U`$ is not $`K`$-extendible. Then $`U`$ is not of quaternionic type, and $`\mathrm{ind}_H^KU`$ is irreducible of complex or quaternionic type according as $`U`$ is of real or complex type, respectively. Moreover, any $`K`$-extension of $`2U`$ is isomorphic to $`\mathrm{ind}_H^KU`$.
###### Proof.
(1) Let $`W`$ be a $`K`$-extension of $`U`$. Since $`U`$ is irreducible, so is $`W`$. Set $`d(W)=dim_{}\mathrm{Hom}_K(W,W)`$ and $`d(U)=dim_{}\mathrm{Hom}_H(U,U)`$. Note that $`d(U)=1,2`$, or $`4`$ according as $`U`$ is of real, complex, or quaternionic type, respectively. Since $`\mathrm{res}_HWU`$, it follows from the Frobenius reciprocity that
(\*)
$$m(W)d(W)=dim_{}\mathrm{Hom}_K(W,\mathrm{ind}_H^KU)=dim_{}\mathrm{Hom}_H(\mathrm{res}_HW,U)=d(U),$$
where $`m(W)`$ denotes the multiplicity of $`W`$ in $`\mathrm{ind}_H^KU`$. On the other hand, since
$$dim_{}\mathrm{ind}_H^KU=2dim_{}U=2dim_{}W,$$
the multiplicity $`m(W)`$ must be either $`1`$ or $`2`$.
If $`m(W)=1`$ (i.e., $`d(W)=d(U)`$), then $`\mathrm{ind}_H^KUWW^{}`$ for some $`K`$-module $`W^{}`$ which is not isomorphic to $`W`$. Since $`\mathrm{res}_H\mathrm{ind}_H^KU2U`$, $`\mathrm{res}_HWU`$ is isomorphic to $`\mathrm{res}_HW^{}`$, so that $`W^{}`$ is also a $`G`$-extension of $`U`$. Since $`W^{}`$ appears in $`\mathrm{ind}_H^KU`$ once, $`m(W^{})=1`$. Therefore the identity (\*) applied to $`W^{}`$ implies $`d(W^{})=d(U)`$. This together with the equality $`d(W)=d(U)`$ implies that the two mutually non-isomorphic $`K`$-extensions $`W`$ and $`W^{}`$ of $`U`$ are of the same type as $`U`$.
If $`m(W)=2`$, then $`\mathrm{ind}_H^KU2W`$. Since any $`K`$-extension of $`U`$ is contained in $`\mathrm{ind}_H^KU`$ as a direct summand, $`W`$ is the unique $`K`$-extension of $`U`$. The type of $`W`$ can be read from the equality $`d(W)=d(U)/2`$. This equality in particular implies that $`d(U)`$ must be even, in other words, $`U`$ is not of real type when $`m(W)=2`$.
These prove the statement (1) except the last statement. The last statement can be seen as follows. If $`U`$ has two $`K`$-extensions $`W`$ and $`W^{}`$, then $`\mathrm{Hom}_H(W,W^{})`$ is a nontrivial real $`K`$-module of dimension one with trivial $`H`$-action, and $`W\mathrm{Hom}_H(W,W^{})`$ is isomorphic to $`W^{}`$, proving the last statement.
(2) Suppose $`U`$ has no $`K`$-extension. It is known that $`U`$ is $`K`$-extendible if $`U`$ is of quaternionic type, see the remark following Proposition 4.5 in \[CKS99\]. If $`\mathrm{ind}_H^KU`$ is reducible, then each direct summand of $`\mathrm{ind}_H^KU`$ is a $`K`$-extension of $`U`$ which contradicts the assumption that $`U`$ has no $`K`$-extension. Therefore, $`\mathrm{ind}_H^KU`$ is irreducible. Noting that $`\mathrm{res}_H\mathrm{ind}_H^KU2U`$, it follows from the Frobenius reciprocity that
$$d(\mathrm{ind}_H^KU)=dim_{}\mathrm{Hom}_K(\mathrm{ind}_H^KU,\mathrm{ind}_H^KU)=dim_{}\mathrm{Hom}_H(2U,U)=2d(U),$$
which implies the statement on the type of $`\mathrm{ind}_H^KU`$. The last statement in (2) follows again from the Frobenius reciprocity. ∎
The following lemma follows from Lemma 2.1.
###### Lemma 2.2.
Let $`\chi `$ be the $`K`$-invariant character of $`U`$, and let $`e`$ be the number of $`K`$-extensions of $`U`$. Then $`e=0`$, $`1`$, or $`2`$, and
1. if $`e=0`$, then $`\mathrm{Rep}(K,\chi )`$ is generated by $`\mathrm{ind}_H^KU`$,
2. if $`e=1`$, then $`\mathrm{Rep}(K,\chi )`$ is generated by a $`K`$-module of the same dimension as $`U`$,
3. if $`e=2`$, then $`\mathrm{Rep}(K,\chi )`$ is generated by two $`K`$-modules of the same dimension as $`U`$ such that one is isomorphic to the tensor product of the other with the nontrivial real $`K`$-module of dimension one with trivial $`H`$-action. ∎
Let $`K_1`$ and $`K_2`$ be two groups containing $`H`$ as an index two subgroup, and let $`\chi `$ be a real irreducible character of $`H`$ which is $`K_1`$\- and $`K_2`$-invariant. We next consider the subset $`\mathrm{Rep}(K_1,K_2,\chi )`$ of $`\mathrm{Rep}(K_1,\chi )\times \mathrm{Rep}(K_2,\chi )`$ consisting of pairs of the same dimension. Denote by $`e_1`$ and $`e_2`$ the number of $`K_1`$\- and $`K_2`$-extensions, respectively.
###### Lemma 2.3.
The semi-group $`\mathrm{Rep}(K_1,K_2,\chi )`$ is generated by
$$\{\begin{array}{cc}\text{one element }R_\chi ,\hfill & \text{if }(e_1,e_2)=(0,0),(1,0),(0,1)\text{, or }(1,1),\hfill \\ \text{two elements }R_\chi ^\pm ,\hfill & \text{if }(e_1,e_2)=(2,1)\text{ or }(1,2),\hfill \\ \text{three elements }\stackrel{~}{R}_\chi ^0,\stackrel{~}{R}_\chi ^\pm ,\hfill & \text{if }(e_1,e_2)=(2,0)\text{ or }(0,2),\hfill \\ \text{four elements }R_\chi ^{\pm \pm },\hfill & \text{if }(e_1,e_2)=(2,2),\hfill \end{array}$$
with relations $`2\stackrel{~}{R}_\chi ^0=\stackrel{~}{R}_\chi ^++\stackrel{~}{R}_\chi ^{}`$ and $`R_\chi ^{++}+R_\chi ^{}=R_\chi ^++R_\chi ^+`$.
###### Proof.
For $`i=1`$ and $`2`$, denote by $`\stackrel{~}{R}_i`$, $`R_i`$, and $`R_i^\pm `$ the set of generators of $`\mathrm{Rep}(K_i,\chi )`$ according to $`e_i=0`$, $`1`$, and $`2`$, respectively. Note that the dimension of $`\stackrel{~}{R}_i`$ is twice that of $`R_i`$ and $`R_i^\pm `$. Then it is easy to find the generators and relations of $`\mathrm{Rep}(K_1,K_2,\chi )`$ according to $`(e_1,e_2)`$ as in Table 1. ∎
###### Remark.
For $`i=1`$ and $`2`$, denote by $`_i^+`$ and $`_i^{}`$, respectively, the trivial and the nontrivial real $`K_i`$-module of dimension one with trivial $`H`$-action. Then the set of pairs $`(_1^\pm ,_2^\pm )`$ forms a group isomorphic to $`/2\times /2`$ under tensor product on each factor, and it acts by the same operation on the generators in Lemma 2.3. The action is transitive except for the third case where $`\mathrm{Rep}(K_1,K_2,\chi )`$ is generated by three elements. In that case, $`\stackrel{~}{R}_\chi ^\pm `$ constitute one orbit and $`\stackrel{~}{R}_\chi ^0`$ is fixed by the action of the pairs $`(_1^\pm ,_2^\pm )`$.
We recall some facts on the extension of an $`H`$-module, which will be used in Section 6.
###### Lemma 2.4.
Let $`H`$ be a normal subgroup of $`G`$ and let $`U`$ be a real irreducible $`H`$-module with $`G`$-invariant character.
1. Suppose $`G/H`$ is finite cyclic of odd order. Then $`U`$ has a $`G`$-extension, and the $`G`$-extension is unique if $`U`$ is of real type.
2. Suppose $`G/H`$ is a dihedral group of order $`2n`$ for odd $`n`$. Then
1. $`U`$ has a $`G`$-extension if and only if it has a $`K`$-extension for some subgroup $`K`$ of $`G`$ which contain $`H`$ as an index two subgroup,
2. $`2U`$ always has a $`G`$-extension.
###### Proof.
See \[CKS99, Proposition 4.5 and 4.6\] for the former statement in (1) and (2-a).
To see the uniqueness in (1), we note that if $`U_1`$ and $`U_2`$ are $`G`$-extensions of $`U`$, then $`\mathrm{Hom}_H(U_1,U_2)`$. Since $`H`$ acts trivially on $`\mathrm{Hom}_H(U_1,U_2)`$ and $`G/H`$ is of odd order, $`\mathrm{Hom}_H(U_1,U_2)`$ must be a trivial $`G`$-module. Therefore $`\mathrm{Hom}_G(U_1,U_2)`$ is also isomorphic to $``$, which means that $`U_1`$ and $`U_2`$ are isomorphic as $`G`$-modules. This proves the uniqueness of the $`G`$-extension of $`U`$.
It remains to prove (2-b). Let $`P`$ be a normal subgroup of $`G`$ which contains $`H`$ and $`P/H`$ is a normal cyclic subgroup of $`G/H`$ of order $`n`$. By (1) above, $`U`$ has a $`P`$-extension, say $`W`$. Then $`\mathrm{ind}_P^GW`$ is a $`G`$-extension of $`2U`$. ∎
## 3. Fiber modules
In this section we prove Theorem 1.1 except for Cases A and B. The following two propositions can be proved by the same argument as in the complex case, see Theorem A and its subsequent remark in \[CKMS99\].
###### Proposition 3.1.
A real $`H`$-module is the fiber $`H`$-module of a real $`G`$-vector bundle over $`S(\rho )`$ if and only if its character is $`G`$-invariant and $`G_1`$-extendible (and $`G_\mu `$-extendible when $`\rho (G)=D_n`$). ∎
###### Proposition 3.2.
Given $`G_z`$-extensions $`V_z`$ of a real $`H`$-module with $`G`$-invariant character for $`z=1`$ (and $`\mu `$ when $`\rho (G)=D_n`$), there exists a real $`G`$-vector bundle $`E`$ over $`S(\rho )`$ such that the fiber $`E_z`$ of $`E`$ over $`z`$ is isomorphic to $`V_z`$ as $`G_z`$-modules. ∎
Proposition 3.1 gives a characterization of the fiber $`H`$-module of a real $`G`$-vector bundle over $`S(\rho )`$, and Proposition 3.2 shows that the semi-group homomorphism $`\mathrm{\Gamma }`$ in the introduction is surjective. On the other hand, the following proposition shows that $`\mathrm{\Gamma }`$ is injective except for Cases A and B, which proves Theorem 1.1 except for Cases A and B.
###### Proposition 3.3.
Let $`\chi `$ be a real irreducible character of $`H`$ which is $`G`$-invariant. Except for Cases A and B, two real $`G`$-vector bundles $`E`$ and $`E^{}`$ in $`\mathrm{Vect}_G(S(\rho ),\chi )`$ are isomorphic if and only if the fiber $`G_z`$-modules $`E_z`$ and $`E_z^{}`$ at $`zS(\rho )`$ are isomorphic for $`z=1`$ (and for $`z=\mu `$ when $`\rho (G)=D_n`$).
###### Proof.
The proof of \[CKMS99, Theorem 6.1\] holds in the real category with slight modification. For reader’s convenience we shall give the argument when $`\rho (G)`$ is finite. The case when $`\rho (G)`$ is infinite is easy since the action of $`G`$ on $`S(\rho )`$ is transitive, see \[CKMS99, Proposition 2.3\] for details.
We first note that if there exists an equivariant isomorphism $`\mathrm{\Psi }:EE^{}`$, then it must satisfy the equivariance condition $`\mathrm{\Psi }_{\rho (g)z}=g\mathrm{\Psi }_zg^1`$ for any $`gG`$ where $`\mathrm{\Psi }_z=\mathrm{\Psi }|_{E_z}`$.
Suppose $`\rho (G)SO(2)`$. Then $`G_1=H`$, $`\rho (G)`$ is finite cyclic, say, of order $`n`$, and since Case A is excluded, $`\chi `$ is not of real type. Choose an element $`aG`$ such that $`\rho (a)`$ is the rotation through the angle $`2\pi /n`$. By the assumption we have an $`H`$-linear isomorphism $`\mathrm{\Psi }_1:E_1E_1^{}`$. Set $`\mathrm{\Psi }_{\rho (a)1}=a\mathrm{\Psi }_1a^1`$, which is also an $`H`$-linear isomorphism. Then we connect $`\mathrm{\Psi }_1`$ and $`\mathrm{\Psi }_{\rho (a)1}`$ continuously in the set of $`H`$-linear isomorphisms of the fiber $`H`$-module along the arc of $`S(\rho )`$ joining $`1`$ and $`\rho (a)1=e^{2\pi i/n}`$. This is possible because the set of $`H`$-linear isomorphisms of the fiber $`H`$-module is homeomorphic to a general linear group over $``$ or $``$ depending on the type of $`\chi `$ (remember that $`\chi `$ is not of real type), and it is arcwise connected. Thus we have a bundle isomorphism between $`E`$ and $`E^{}`$ restricted to the arc of $`S(\rho )`$ joining $`1`$ and $`\rho (a)1`$. We extend this isomorphism to an entire isomorphism over $`S(\rho )`$ using the equivariance condition $`\mathrm{\Psi }_{\rho (a)z}=a\mathrm{\Psi }_za^1`$.
When $`\rho (G)=D_n`$, we choose a $`G_1`$-linear isomorphism $`\mathrm{\Psi }_1`$ and a $`G_\mu `$-linear isomorphism $`\mathrm{\Psi }_\mu `$. Similarly to the above, we connect $`\mathrm{\Psi }_1`$ and $`\mathrm{\Psi }_\mu `$ as $`H`$-linear isomorphisms along the arc of $`S(\rho )`$ joining $`1`$ and $`\mu =e^{\pi i/n}`$, and then extend it to an isomorphism over $`S(\rho )`$ using the equivariance condition. But it is not always possible to connect $`\mathrm{\Psi }_1`$ and $`\mathrm{\Psi }_\mu `$ when $`\chi `$ is of real type because the set of $`H`$-linear isomorphisms of the fiber $`H`$-module, which is homeomorphic to $`\mathrm{GL}(m,)`$, is not arcwise connected. In this case, however, we have another assumption that $`\chi `$ is $`G_z`$-extendible for $`z=1`$ or $`\mu `$ since Case B is excluded. By Lemma 2.1 (1-a) $`\chi `$ has two $`G_z`$-extensions, say $`\stackrel{~}{\chi }_1`$ and $`\stackrel{~}{\chi }_2`$. Thus the character of $`E_zE_z^{}`$ as a $`G_z`$-module is of the form $`m_1\stackrel{~}{\chi }_1+m_2\stackrel{~}{\chi }_2`$ for some nonnegative integers $`m_1`$ and $`m_2`$ with $`m=m_1+m_2`$, so that the set of $`G_z`$-linear isomorphisms between $`E_z`$ and $`E_z^{}`$ is homeomorphic to $`\mathrm{GL}(m_1,)\times \mathrm{GL}(m_2,)`$. Since the inclusion map from $`\mathrm{GL}(m_1,)\times \mathrm{GL}(m_2,)`$ to $`\mathrm{GL}(m,)`$ induces a surjection on the $`\pi _0`$ level, it is possible to choose a $`G_z`$-linear isomorphism $`\mathrm{\Psi }_z`$ so that $`\mathrm{\Psi }_1`$ and $`\mathrm{\Psi }_\mu `$ can be connected in the set of $`H`$-linear isomorphisms of the fiber $`H`$-module. ∎
## 4. Topological complexity: Cases A and B
Propositions 3.2 and 3.3 show that the map $`\mathrm{\Gamma }`$ is an isomorphism except for Cases A and B. In this section we investigate the structure on $`\mathrm{Vect}_G(S(\rho ),\chi )`$ for Cases A and B, and complete the proof of Theorem 1.1.
### Case A
The case where $`\rho (G)SO(2)`$ and $`\chi `$ is of real type. In this case one can reduce the study of $`\mathrm{Vect}_G(S(\rho ),\chi )`$ to the nonequivariant case.
###### Lemma 4.1.
In Case A, the semi-group $`\mathrm{Vect}_G(S(\rho ),\chi )`$ is generated by two elements $`N_\chi ^\pm `$ with relation $`2N_\chi ^+=2N_\chi ^{}`$. Moreover, $`N_\chi ^\pm `$ have $`\chi `$ as the character of the fiber $`H`$-modules, and they are related in such a way that $`N_\chi ^{}`$ is obtained from $`N_\chi ^+`$ by tensoring with a nontrivial real $`G`$-line bundle over $`S(\rho )`$ with trivial fiber $`H`$-module.
###### Proof.
There is an element $`L`$ in $`\mathrm{Vect}_G(S(\rho ),\chi )`$ with $`\chi `$ as the character of the fiber $`H`$-module by Proposition 3.1, and we have the semi-group isomorphisms
$$\mathrm{Vect}_G(S(\rho ),\chi )\mathrm{Vect}_{G/H}(S(\rho ))\mathrm{Vect}(S^1),$$
where the former isomorphism is given by sending $`E`$ to $`\mathrm{Hom}_H(L,E)`$ and the latter is given by taking orbit spaces by the $`G/H`$-action. In fact, the map sending $`F\mathrm{Vect}_{G/H}(S(\rho ))`$ to $`LF`$ is the inverse of the former isomorphism, where $`F`$ is viewed as a $`G`$-vector bundle through the quotient map from $`G`$ to $`G/H`$ (see Lemma 2.2 in \[CKMS99\] for details), and the latter is an isomorphism because the action of $`G/H`$ on $`S(\rho )`$ is free. As is well known, $`\mathrm{Vect}(S^1)`$ is generated by the trivial line bundle $`ϵ`$ and the Hopf line bundle $`\eta `$ with relation $`2ϵ=2\eta `$. Therefore, if we denote by $`N_\chi ^\pm `$ the two generators of $`\mathrm{Vect}_G(S(\rho ),\chi )`$ corresponding to $`ϵ`$ and $`\eta `$ in $`\mathrm{Vect}(S^1)`$ through the above isomorphism, then the lemma follows except the last statement. To see the last statement, we note that $`\mathrm{Hom}_H(N_\chi ^+,N_\chi ^{})`$ is a nontrivial real $`G`$-line bundle over $`S(\rho )`$ with trivial fiber $`H`$-module and that
$$N_\chi ^+\mathrm{Hom}_H(N_\chi ^+,N_\chi ^{})N_\chi ^{},$$
proving the last statement. ∎
### Case B
The case where $`\rho (G)=D_n`$, $`\chi `$ is of real type, and neither $`G_1`$\- nor $`G_\mu `$-extendible. In this case we investigate complex structures on the bundles in $`\mathrm{Vect}_G(S(\rho ),\chi )`$.
Let $`𝔽=`$ or $``$, and set
$$𝒥(𝔽^k)\{J\mathrm{GL}(k,𝔽)J^2=I\},$$
which is the set of complex structures on $`𝔽^k`$. Needless to say, $`𝒥(^k)`$ is empty unless $`k`$ is even. Viewing $``$ as $`^2`$ in a natural way induces an injective homomorphism from $`\mathrm{GL}(k,)`$ to $`\mathrm{GL}(2k,)`$, so that it induces an injection from $`𝒥(^k)`$ to $`𝒥(^{2k})`$ and we view $`𝒥(^k)`$ as a subset of $`𝒥(^{2k})`$ through this map.
###### Lemma 4.2.
1. $`𝒥(^k)`$ has $`k+1`$ connected components.
2. $`𝒥(^{2k})`$ has two connected components.
3. If $`k`$ is odd, then each connected component of $`𝒥(^{2k})`$ contains $`(k+1)/2`$ connected components of $`𝒥(^k)`$, while if $`k`$ is even, then one connected component of $`𝒥(^{2k})`$ contains $`k/2`$ and the other contains $`k/2+1`$ connected components of $`𝒥(^k)`$.
###### Proof.
(1) We note that $`\mathrm{GL}(k,)`$ acts on $`𝒥(^k)`$ by conjugation. Since the minimal polynomial of any element in $`𝒥(^k)`$ has distinct root it is diagonalizable. So two elements in $`𝒥(^k)`$ are in the same orbit if and only if they have the same eigenvalues which are $`\pm i`$ because $`J^2=I`$. This implies that $`𝒥(^k)`$ has exactly $`k+1`$ connected components because there are $`k+1`$ possibilities of the $`k`$ eigenvalues.
(2) $`\mathrm{GL}(2k,)`$ acts transitively on $`𝒥(^{2k})`$, and the isotropy subgroup at an element of $`𝒥(^{2k})`$ is isomorphic to $`\mathrm{GL}(k,)`$; so $`𝒥(^{2k})`$ is homeomorphic to a homogeneous space $`\mathrm{GL}(2k,)/\mathrm{GL}(k,)`$ which has two connected components (see \[MS98, Proposition 2.48\] for more details).
(3) As observed in (1) above, $`k+1`$ elements
$$\mathrm{diag}(i,i,\mathrm{},i),\mathrm{diag}(i,i,\mathrm{},i),\mathrm{},\mathrm{diag}(i,i,\mathrm{},i)$$
respectively lie in the $`k+1`$ different connected components of $`𝒥(^k)`$. Through the inclusion map from $`𝒥(^k)`$ to $`𝒥(^{2k})`$, they respectively are mapped to
$$\mathrm{diag}(J_0,J_0,\mathrm{},J_0),\mathrm{diag}(J_0,J_0,\mathrm{},J_0),\mathrm{},\mathrm{diag}(J_0,J_0,\mathrm{},J_0)$$
where $`J_0`$ is the $`2\times 2`$ matrix $`\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right)`$. Since $`J_0`$ and $`J_0`$ are conjugate by $`\left(\begin{array}{cc}0& 1\\ 1& 0\end{array}\right)`$ whose determinant is negative, the $`k+1`$ elements above in $`𝒥(^k)`$ are in a same connected component of $`𝒥(^{2k})`$ if and only if the number of $`J_0`$’s as entries are congruent modulo $`2`$. This implies (3). ∎
For a real $`G`$-module $`V`$, we denote the set of $`G`$-invariant complex structures on $`V`$ by
$$𝒥(V)^G\{J\mathrm{GL}(V)^GJ^2=I\},$$
where $`\mathrm{GL}(V)^G`$ denotes the $`G`$-linear automorphisms of $`V`$. A pair $`(V,J)`$ is a complex $`G`$-module whose realification is $`V`$. We note that $`\mathrm{GL}(V)^G`$ acts on $`𝒥(V)^G`$ by conjugation and that two complex $`G`$-modules $`(V,J)`$ and $`(V,J^{})`$ are isomorphic if and only if $`J`$ and $`J^{}`$ are in the same orbit of the $`\mathrm{GL}(V)^G`$ action.
We consider the following setting for later use.
###### Lemma 4.3.
Let $`K`$ be a group and let $`H`$ be a normal subgroup of $`K`$. Suppose
1. $`W`$ is an irreducible real $`K`$-module of complex type,
2. $`U`$ is an irreducible real $`H`$-module of real type,
3. $`\mathrm{res}_HW2U`$.
Then, for any positive integer $`k`$, $`𝒥(kW)^K`$ can naturally be viewed as a subspace of $`𝒥(2kU)^H`$, and we have
1. $`𝒥(kW)^K`$ has $`k+1`$ connected components,
2. $`𝒥(2kU)^H`$ has two connected components,
3. if $`k`$ is odd, then each connected component of $`𝒥(2kU)^H`$ contains $`(k+1)/2`$ connected components of $`𝒥(kW)^K`$, while if $`k`$ is even, then one connected component of $`𝒥(2kU)^H`$ contains $`k/2`$ and the other contains $`k/2+1`$ connected components of $`𝒥(kW)^K`$.
###### Proof.
It follows from the assumptions (a) and (b) that $`\mathrm{GL}(kW)^K\mathrm{GL}(k,)`$ and $`\mathrm{GL}(2kU)^H\mathrm{GL}(2k,)`$. Therefore the lemma follows from Lemma 4.2. ∎
We return to the original setting of Case B. Denote by $`U`$ a real irreducible $`H`$-module with $`\chi `$ as its character. Since $`\chi `$ is neither $`G_1`$\- nor $`G_\mu `$-extendible and of real type, $`\mathrm{ind}_H^{G_z}U`$ is the unique $`G_z`$-extension of $`2U`$ of complex type for $`z=1`$ and $`\mu `$ by Lemma 2.1 (2). Therefore we are in a setting to which Lemma 4.3 can be applied. Moreover this shows that an element $`E`$ in $`\mathrm{Vect}_G(S(\rho ),\chi )`$ must have the fibers at $`z=1`$ and $`\mu `$ isomorphic to $`k(\mathrm{ind}_H^{G_z}U)`$ for some integer $`k`$. In particular, its fiber $`H`$-module is $`2kU`$.
A $`G`$-invariant complex structure on $`E`$ is a $`G`$-vector bundle automorphism $`J`$ of $`E`$ such that $`J^2=I`$. A pair $`(E,J)`$ is a complex $`G`$-vector bundle whose realification is $`E`$. We say that two $`G`$-invariant complex structures $`J`$ and $`J^{}`$ on $`E`$ are equivalent if $`(E,J)`$ and $`(E,J^{})`$ are isomorphic as complex $`G`$-vector bundles, in other words, if $`J`$ and $`J^{}`$ are conjugate by a real $`G`$-vector bundle automorphism of $`E`$.
###### Lemma 4.4.
The number of inequivalent $`G`$-invariant complex structures on $`E`$ is $`(k+1)^2/2`$ if $`k`$ is odd, and $`k(k/2+1)`$ or $`k(k/2+1)+1`$ if $`k`$ is even.
###### Proof.
Let $`𝒥(E_z)`$ be the set of (not necessarily invariant) complex structures on the fiber $`E_z`$. The collection $`𝒥(E)`$ of $`𝒥(E_z)`$ over $`zS(\rho )`$ forms a $`G`$-fiber bundle over $`S(\rho )`$, the $`G`$-action on $`𝒥(E)`$ being induced from that on $`E`$. Then a $`G`$-invariant complex structure on $`E`$ can be viewed as a continuous $`G`$-equivariant cross section of the $`G`$-fiber bundle. The image of the cross section lies in $`𝒥(E)^H`$ because $`H`$ acts trivially on $`S(\rho )`$.
In order to construct a continuous $`G`$-equivariant cross section of $`𝒥(E)S(\rho )`$, we choose a pair of points from $`𝒥(E_1)^{G_1}`$ and $`𝒥(E_\mu )^{G_\mu }`$ (i.e., one point from each), which can be connected by a continuous cross section of $`𝒥(E)^H`$ restricted to the arc $`R`$ in $`S(\rho )`$ joining $`1`$ and $`\mu =e^{\pi i/n}`$. Not all pairs of those points are connected by such a cross section as observed later. But, once we find such a cross section, we can extend it to an entire $`G`$-equivariant cross section using the equivariance as is done in the proof of Proposition 3.3. On the other hand, we know in \[CKMS99, Theorem 6.1\] that isomorphism classes of complex $`G`$-vector bundles over $`S(\rho )`$ are distinguished by the complex fiber $`G_1`$\- and $`G_\mu `$-modules. Therefore, the number $`CS(E)^G`$ of inequivalent $`G`$-invariant complex structures on $`E`$ is equal to the number of pairs of connected components in $`𝒥(E_1)^{G_1}`$ and $`𝒥(E_\mu )^{G_\mu }`$ which are connected through $`𝒥(E)^H|_R`$.
Suppose $`k`$ is even. Denote by $`C_z^1`$ and $`C_z^2`$, for $`z=1`$ and $`\mu `$, the connected components of $`𝒥(E_z)^H|_R`$ containing $`k/2`$ and $`k/2+1`$ connected components of $`𝒥(E_z)^{G_z}`$, respectively. If $`C_1^1`$ and $`C_\mu ^1`$ are connected through $`𝒥(E)^H|_R`$, then so are $`C_1^2`$ and $`C_\mu ^2`$. Counting the number of choices of pairs of connected components in $`𝒥(E_1)^{G_1}`$ and $`𝒥(E_\mu )^{G_\mu }`$ which are connected through $`𝒥(E)^H|_R`$, one has
$$CS(E)^G=\left(\frac{k}{2}\right)^2+\left(\frac{k}{2}+1\right)^2=k(k/2+1)+1.$$
On the other hand, if $`C_1^1`$ and $`C_\mu ^2`$ are connected through $`𝒥(E)^H|_R`$, then so are $`C_1^2`$ and $`C_\mu ^1`$ and one has
$$CS(E)^G=\frac{k}{2}\left(\frac{k}{2}+1\right)+\frac{k}{2}\left(\frac{k}{2}+1\right)=k(k/2+1).$$
For $`k`$ odd, a similar argument proves that
$$CS(E)^G=\left(\frac{k+1}{2}\right)^2+\left(\frac{k+1}{2}\right)^2=(k+1)^2/2.$$
###### Lemma 4.5.
In Case B, the semi-group $`\mathrm{Vect}_G(S(\rho ),\chi )`$ is generated by two elements $`M_\chi ^\pm `$ with relation $`2M_\chi ^+=2M_\chi ^{}`$. Moreover, $`M_\chi ^\pm `$ have $`2\chi `$ as the character of the fiber $`H`$-modules, and they are related in such a way that $`M_\chi ^{}`$ is obtained from $`M_\chi ^+`$ by tensoring with a nontrivial $`G`$-line bundle over $`S(\rho )`$ with trivial fiber $`H`$-module.
###### Proof.
Since $`\chi `$ is of real type, the character of $`U`$ is also $`\chi `$; so we may view $`\chi `$ as a complex irreducible character of $`H`$. We have proved in \[CKMS99, Theorem B\] that the semi-group $`\mathrm{Vect}_G^{}(S(\rho ),\chi )`$ of isomorphism classes of complex $`G`$-vector bundles over $`S(\rho )`$ with multiples of $`\chi `$ as the character of fiber $`H`$-modules is generated by four elements $`L_\chi ^{\pm \pm }`$ with relation $`L_\chi ^{++}+L_\chi ^{}=L_\chi ^++L_\chi ^+`$, where $`L_\chi ^{\pm \pm }`$ are complex $`G`$-vector bundles over $`S(\rho )`$ with $`U`$ as the fiber $`H`$-module such that the fiber $`G_1`$-modules (resp. $`G_\mu `$-modules) of $`L_\chi ^{st}`$ and $`L_\chi ^{s^{}t^{}}`$, where $`s`$, $`s^{}`$, $`t`$, and $`t^{}`$ denote $`+`$ or $``$, agree if and only if $`s=s^{}`$ (resp. $`t=t^{}`$). In fact, the two non-isomorphic fiber $`G_1`$-modules (resp. $`G_\mu `$-modules) of $`L_\chi ^{\pm \pm }`$ are complex conjugate to each other, so the complex conjugate (or dual) bundles of $`L_\chi ^{++}`$ and $`L_\chi ^+`$ are respectively $`L_\chi ^{}`$ and $`L_\chi ^+`$.
Let $`\mathrm{\Phi }:\mathrm{Vect}_G^{}(S(\rho ),\chi )\mathrm{Vect}_G(S(\rho ),\chi )`$ be the realification map. It is surjective by Lemma 4.4. Since any complex $`G`$-vector bundle is isomorphic to its complex conjugate bundle as real $`G`$-vector bundles, $`\mathrm{\Phi }(L_\chi ^{++})=\mathrm{\Phi }(L_\chi ^{})`$ and $`\mathrm{\Phi }(L_\chi ^+)=\mathrm{\Phi }(L_\chi ^+)`$. Therefore the relation $`L_\chi ^{++}+L_\chi ^{}=L_\chi ^++L_\chi ^+`$ on $`\mathrm{Vect}_G^{}(S(\rho ),\chi )`$ reduces to $`2\mathrm{\Phi }(L_\chi ^{++})=2\mathrm{\Phi }(L_\chi ^+)`$ on $`\mathrm{Vect}_G(S(\rho ),\chi )`$. It follows that for each fixed fiber dimension there are at most two elements in $`\mathrm{Vect}_G(S(\rho ),\chi )`$. We claim that there is no other relation. It suffices to show that there are exactly two elements in $`\mathrm{Vect}_G(S(\rho ),\chi )`$ for a fixed fiber dimension. If there is only one element for a fixed fiber dimension, say $`2kdimU`$, then the unique bundle must have $`(k+1)^2`$ inequivalent $`G`$-invariant complex structures because the number of elements in $`\mathrm{Vect}_G^{}(S(\rho ),\chi )`$ of (real) fiber dimension $`2kdimU`$ is exactly $`(k+1)^2`$ \[CKMS99, Corollary 5.2\]. This contradicts Lemma 4.4.
It remains to show that the two generators $`\mathrm{\Phi }(L_\chi ^{++})`$ and $`\mathrm{\Phi }(L_\chi ^+)`$ are related by tensoring with a nontrivial $`G`$-line bundle with trivial fiber $`H`$-module. The fiber $`G_1`$-modules of $`L_\chi ^+`$ and $`L_\chi ^{++}`$ at $`1`$ are isomorphic but the fiber $`G_\mu `$-modules of them at $`\mu `$ are not, more precisely, they are related through the tensor product with the nontrivial real $`1`$-dimensional $`G_\mu `$-module defined by $`G_\mu G_\mu /H\{\pm 1\}`$, see Lemma 2.1 (1). Therefore $`\mathrm{\Phi }(L_\chi ^+)`$ is obtained from $`\mathrm{\Phi }(L_\chi ^{++})`$ by tensoring with a real $`G`$-line bundle with trivial fiber $`H`$-module, whose fiber at $`1`$ is the trivial $`G_1`$-module and the fiber at $`\mu `$ is the nontrivial $`G_\mu `$-module. The existence of such line bundle is guaranteed by Proposition 3.2. ∎
###### Proof of Theorem 1.1.
The map $`\mathrm{\Gamma }`$ is surjective by Proposition 3.2 and injective except for Cases A and B by Proposition 3.3. In both Cases A and B the target of $`\mathrm{\Gamma }`$ is a semi-group generated by one element by Lemmas 2.2 and 2.3 while the domain of $`\mathrm{\Gamma }`$ is generated by two elements with the relation as in Lemmas 4.1 and 4.5. This implies that $`\mathrm{\Gamma }`$ is two to one.
Finally, we note that tensoring elements in $`\mathrm{Vect}_G(S(\rho ),\chi )`$ with a nontrivial $`G`$-line bundle with trivial $`H`$-module does not change the fiber $`G_1`$-modules (resp. fiber $`G_1`$\- and $`G_\mu `$-modules) in Case A (resp. Case B). This implies the last statement in the theorem. ∎
## 5. The semi-group structure on $`\mathrm{Vect}_G(S(\rho ),\chi )`$
In this section we determine the semi-group structure on $`\mathrm{Vect}_G(S(\rho ),\chi )`$. Let $`e_1`$ and $`e_\mu `$ denote the numbers of $`G_1`$\- and $`G_\mu `$-extensions of $`\chi `$, respectively. When $`\rho (G)`$ agrees with $`O(2)`$ or is contained in $`SO(2)`$, we define $`e_\mu `$ to be $`1`$ for convenience. In both real and complex category, the semi-group structure on the target of $`\mathrm{\Gamma }`$ is determined by the numbers $`e_1`$ and $`e_\mu `$. The numbers $`e_1`$ and $`e_\mu `$ depend only on the types of $`\rho (G)`$ in the complex category, but this is not true in the real category. This is another complexity in our study arising from real representation theory.
The possible values of $`e_1`$ and $`e_\mu `$ according to $`\rho (G)`$ and the type of $`\chi `$ are given by Table 2 and 3.
We state here the semi-group structure on $`\mathrm{Vect}_G(S(\rho ),\chi )`$ according to the values of $`e_1`$ and $`e_\mu `$.
###### Theorem 5.1.
Except for Cases A and B, the semi-group $`\mathrm{Vect}_G(S(\rho ),\chi )`$ is generated by
1. one element $`L_\chi `$, if $`(e_1,e_\mu )=(0,0),(1,0),(0,1)`$ or $`(1,1)`$,
2. two elements $`L_\chi ^\pm `$, if $`(e_1,e_\mu )=(2,1)`$ or $`(1,2)`$,
3. three elements $`\stackrel{~}{L}_\chi ^0,\stackrel{~}{L}_\chi ^\pm `$ with relation $`2\stackrel{~}{L}_\chi ^0=\stackrel{~}{L}_\chi ^++\stackrel{~}{L}_\chi ^{}`$, if $`(e_1,e_\mu )=(2,0)`$ or $`(0,2)`$,
4. four elements $`L_\chi ^{\pm \pm }`$ with relation $`L_\chi ^{++}+L_\chi ^{}=L_\chi ^++L_\chi ^+`$, if $`(e_1,e_\mu )=(2,2)`$.
In Case A or B, it is generated by
1. two elements $`\stackrel{~}{\stackrel{~}{L}}_\chi ^\pm `$ with relation $`2\stackrel{~}{\stackrel{~}{L}}{}_{\chi }{}^{+}=2\stackrel{~}{\stackrel{~}{L}}_\chi ^{}`$.
Moreover, except for $`\stackrel{~}{L}_\chi ^0`$ in the case (3), all generators are related through tensor product with real $`G`$-line bundles over $`S(\rho )`$ with trivial fiber $`H`$-module.
###### Proof.
The statements (1)–(4) follow from Lemmas 2.2, 2.3 and Theorem 1.1, and the statement (5) follows from Lemmas 4.1 and 4.5.
We now prove the last statement in the theorem. After setting $`K_1=G_1`$ and $`K_2=G_\mu `$, it is obvious that the inverse images of the pairs $`(_1^\pm ,_2^\pm )`$ in the remark after Lemma 2.3 by the semi-group homomorphism $`\mathrm{\Gamma }`$ in Theorem 1.1 are real $`G`$-line bundles with trivial fiber $`H`$-module. Moreover, $`\mathrm{\Gamma }`$ preserves the two tensor product operations, one on $`\mathrm{Vect}_G(S(\rho ),\chi )`$ with real $`G`$-line bundles and the other on $`\mathrm{Rep}(G_1,G_\mu ,\chi )`$ by the pairs $`(_1^\pm ,_2^\pm )`$. Therefore, Proposition 3.3 implies that, except for Cases A and B, the generators of $`\mathrm{Vect}_G(S(\rho ),\chi )`$ are related through tensor product with real $`G`$-line bundles with trivial fiber $`H`$-module. The same argument also holds for $`\mathrm{Rep}(G_1,\chi )`$ by Lemma 2.2. For Cases A and B, the statement follows from the last statement of Lemmas 4.1 and 4.5. ∎
###### Corollary 5.2.
Let $`N`$ be the number of isomorphism classes of real $`G`$-vector bundles over $`S(\rho )`$ with $`m\chi `$ as the character of the fiber $`H`$-modules. In case that $`m`$ is odd, $`N`$ is zero if $`e_1=0`$ or $`e_\mu =0`$. Otherwise, except for Cases A and B,
$$N=\{\begin{array}{cc}1,\hfill & \text{if }(e_1,e_\mu )=(0,0),(1,0),(0,1)\text{, or }(1,1),\hfill \\ m+1,\hfill & \text{if }(e_1,e_\mu )=(2,0),(0,2),(2,1)\text{, or }(1,2),\hfill \\ (m+1)^2,\hfill & \text{if }(e_1,e_\mu )=(2,2).\hfill \end{array}$$
In Case A or B, the number $`N`$ is exactly two.
###### Proof.
The proof is elementary and left to the reader. ∎
## 6. Triviality of real $`G`$-vector bundles over a circle
In this section we investigate triviality of the generators in Theorem 5.1 when $`\rho (G)`$ is finite. Triviality of a $`G`$-vector bundle is closely related to the existence of a $`G`$-extension of the fiber $`H`$-module in the following sense: For a given $`H`$-module $`V`$, there exists at least one trivial $`G`$-vector bundle with $`V`$ as its fiber $`H`$-module if $`V`$ extends to a $`G`$-module. In the following we denote by $`Z_n`$ the finite cyclic subgroup of $`SO(2)`$ generated by the rotation through an angle $`2\pi /n`$. Then $`\rho (G)=Z_n`$ for some $`n`$ if $`\rho (G)SO(2)`$. Denote by $`1`$ the trivial real $`H`$-module of dimension one, in other words, $`H`$ acts trivially on $`1`$. In the notation of Lemma 4.1 and Theorem 5.1, real $`G`$-line bundles over $`S(\rho )`$ with trivial fiber $`H`$-module are denoted by $`N_1^\pm `$ and $`L_1^{\pm \pm }`$ according as $`\rho (G)=Z_n`$ and $`D_n`$, respectively.
###### Lemma 6.1.
(1) Suppose $`\rho (G)=Z_n`$. If $`n`$ is even, then $`N_1^\pm `$ are both trivial. If $`n`$ is odd, then one of them, say $`N_1^+`$, is trivial and $`N_1^{}`$ is nontrivial.
(2) Suppose $`\rho (G)=D_n`$. If $`n`$ is even, then $`L_1^{\pm \pm }`$ are all trivial. If $`n`$ is odd, then two of them, say $`L_1^{++}`$ and $`L_1^{}`$, are trivial and the other two are nontrivial.
###### Proof.
(1) Since $`G/H`$ acts freely on $`S(\rho )`$, every real $`G`$-line bundle over $`S(\rho )`$ with trivial fiber $`H`$-module is the pull-back of a real line bundle over $`S^1`$ by the quotient map $`\pi :S(\rho )S(\rho )/GS^1`$. Suppose $`n`$ is even. Then $`\pi ^{}:H^1(S(\rho )/G,/2)H^1(S(\rho ),/2)`$ is trivial, so pullback line bundles by $`\pi `$ have trivial first Whitney classes, which means that the underlying line bundles over $`S(\rho )`$ are trivial. According to \[KM94, Proposition 1.1\], an equivariant line bundle is trivial if and only if its underlying bundle is trivial. Thus, $`N_1^\pm `$ are both trivial when $`n`$ is even.
If $`n`$ is odd, then $`\pi ^{}`$ above is an isomorphism. Therefore, exactly one of $`N_1^\pm `$ has trivial first Whitney class. This together with the result in \[KM94\] mentioned above shows that exactly one of $`N_1^\pm `$ is trivial equivariantly.
(2) Set $`P=\rho ^1(Z_n)`$. Since $`P/H`$ acts freely on $`S(\rho )`$, $`L_1^{\pm \pm }`$ are pullback of real $`G/P`$-line bundles over $`S(\rho )/P`$ by the quotient map $`\pi :S(\rho )S(\rho )/P`$. Here $`G/P`$ is of order two and acts on the circle $`S(\rho )/P`$ as reflection, so we may think of $`G/P`$ as $`D_1`$. According to Theorem 5.1 (or Corollary 5.2) there are four real $`D_1`$-line bundles over $`S(\rho )/P`$. Since the map $`\mathrm{\Gamma }`$ is an isomorphism in this case, they are distinguished by their fiber $`D_1`$-modules over the points $`\pm 1S(\rho )/P`$. More precisely, there are two possibilities for the fiber $`D_1`$-modules at $`1`$ and $`1`$ respectively since there are two real one-dimensional $`D_1`$-modules (the trivial one and the nontrivial one), and hence altogether there are four real $`D_1`$-line bundles over $`S(\rho )/P`$. Moreover, $`D_1`$-line bundles are trivial if and only if the fiber $`D_1`$-modules at $`\pm 1`$ are isomorphic (see also \[Kim94\]).
If $`n`$ is even, then all pullback line bundles by $`\pi `$ are trivial as discussed in (1); so $`L_1^{\pm \pm }`$ are all trivial. If $`n`$ is odd, then the pullback by $`\pi `$ preserves the triviality of line bundles because $`\pi ^{}:H^1(S(\rho )/P;/2)H^1(S(\rho );/2)`$ is an isomorphism. Since there are exactly two trivial $`D_1`$-line bundles over $`S(\rho )/P`$, two of $`L_1^{\pm \pm }`$ are trivial and the other two are nontrivial. ∎
###### Remark.
Suppose $`\rho (G)=D_n`$. For $`z=1`$ and $`\mu `$, denote by $`_z^+`$ and $`_z^{}`$, respectively, the trivial and the nontrivial real $`G_z`$-module of dimension one with trivial $`H`$-action, see also the remark after Lemma 2.3. Then we may assume without loss of generality that the images of $`L_1^{st}`$ by $`\mathrm{\Gamma }`$ in Theorem 1.1 are $`(_1^s,_\mu ^t)`$, where $`s`$ and $`t`$ denote a sign $`+`$ or $``$.
###### Theorem 6.2.
Let $`\rho (G)=Z_n`$ or $`D_n`$, and let $`\chi `$ be a real irreducible character of $`H`$ which is $`G`$-invariant. If $`n`$ is even, then the generators in Theorem 5.1 except for $`\stackrel{~}{L}_\chi ^0`$ are all trivial or all nontrivial in each case. If $`n`$ is odd, then
1. $`L_\chi `$ is trivial,
2. $`L_\chi ^\pm `$ are both trivial,
3. $`\stackrel{~}{L}_\chi ^0`$ and $`\stackrel{~}{L}_\chi ^\pm `$ are all trivial,
4. two of $`L_\chi ^{\pm \pm }`$ are trivial and the other two are nontrivial,
5. one of $`\stackrel{~}{\stackrel{~}{L}}_\chi ^\pm `$ is trivial and the other is nontrivial.
###### Proof.
Recall from the last statement in Theorem 5.1 that all generators are related through tensor product with the real $`G`$-line bundles $`N_1^\pm `$ and $`L_1^{\pm \pm }`$ according as $`\rho (G)=Z_n`$ and $`D_n`$, respectively. These line bundles are all trivial if $`n`$ is even by Lemma 6.1. So the existence of one trivial generator implies triviality of the other generators, and this finishes the proof in case that $`n`$ is even.
In the following we assume that $`n`$ is odd. Denote by $`U`$ a real irreducible $`H`$-module with $`\chi `$ as its character. Recall that the fiber $`H`$-module of a generator is $`U`$ if both $`e_1`$ and $`e_\mu `$ are nonzero, and $`2U`$ otherwise. In case $`\rho (G)=D_n`$, we choose elements $`a`$ and $`b`$ in $`G`$ such that $`\rho (a)`$ is the rotation through the angle $`2\pi /n`$ and $`\rho (b)`$ is the reflection about the $`x`$-axis. Then $`G_1`$ (resp. $`G_\mu `$) is generated by $`H`$ and $`b`$ (resp. $`ab`$).
(1) It suffices to show that the fiber $`H`$-module of a generator extends to a $`G`$-module. In case that $`e_1=e_\mu =1`$, the fiber $`H`$-module of $`L_\chi `$ is $`U`$ and it is $`G`$-extendible by Lemma 2.4. The other case is that either $`e_1=0`$ or $`e_\mu =0`$ and in this case the fiber $`H`$-module of $`L_\chi `$ is $`2U`$ which is $`G`$-extendible by Lemma 2.4 (2-b).
(2) In this case $`\rho (G)=D_n`$ by Table 2 and the fiber $`H`$-modules of generators are $`U`$ which is $`G`$-extendible by Lemma 2.4 (2). So there is at least one trivial generator, say $`L_\chi ^+`$. Since $`(e_1,e_\mu )=(2,1)`$ or $`(1,2)`$, the tensor product of $`L_\chi ^+`$ with $`L_1^{}`$ has different fiber $`G_z`$-module from that of $`L_\chi ^+`$ at the point $`z`$ such that $`e_z=2`$. So we get the other generator $`L_\chi ^{}L_\chi ^+L_1^{}`$. Since $`L_1^{}`$ is trivial by Lemma 6.1, so is $`L_\chi ^{}`$.
(3) In this case $`\rho (G)=D_n`$ by Table 2 and the fiber $`H`$-modules of generators are $`2U`$ because either $`e_1`$ or $`e_\mu `$ is zero. Set $`P=\rho ^1(Z_n)`$. Then $`U`$ has a $`P`$-extension, say $`V`$, by Lemma 2.4 (1). Note that the fiber modules of $`\stackrel{~}{L}_\chi ^0`$ at $`1`$ and $`\mu `$ are isomorphic to $`\mathrm{ind}_H^{G_1}U`$ and $`\mathrm{ind}_H^{G_\mu }U`$, respectively. Thus $`\stackrel{~}{L}_\chi ^0`$ is isomorphic to the product bundle $`S(\rho )\times \mathrm{ind}_P^GV`$ by Proposition 3.3.
We next consider triviality of the generators $`\stackrel{~}{L}_\chi ^\pm `$. It suffices to show that at least one generator, say $`\stackrel{~}{L}_\chi ^+`$, is trivial. Then so is the other generator $`\stackrel{~}{L}_\chi ^{}\stackrel{~}{L}_\chi ^+L_1^{}`$. We assume that $`(e_1,e_\mu )=(2,0)`$. The other case $`(e_1,e_\mu )=(0,2)`$ can be proved similarly.
###### Claim.
$`\chi `$ is of real type.
###### Proof of Claim.
Since $`\chi `$ is not of quaternionic type by Table 3, it suffices to prove that $`\chi `$ is not of complex type. Suppose that $`\chi `$ is of complex type. Then there is a complex $`H`$-module $`V`$ such that $`UV\overline{V}`$ and $`V\overline{V}`$ as complex $`H`$-modules. We note that the realifications of $`V`$ and $`\overline{V}`$ are $`U`$, and since $`\chi `$ is $`G`$-invariant, $`{}_{}{}^{g}\chi _{V}^{}=\chi _V`$ or $`\chi _{\overline{V}}`$ for $`gG`$ where $`\chi _V`$ and $`\chi _{\overline{V}}`$ denote the characters of $`V`$ and $`\overline{V}`$ respectively.
Since $`e_1=2`$ and $`e_\mu =0`$ by assumption, $`U`$ has two $`G_1`$-extensions of complex type by Lemma 2.1 but no $`G_\mu `$-extension. It follows that $`V`$ is $`G_1`$-extendible but not $`G_\mu `$-extendible, so $`\chi _V`$ is $`G_1`$-invariant but not $`G_\mu `$-invariant. Namely, $`{}_{}{}^{b}\chi _{V}^{}=\chi _V`$ and $`{}_{}{}^{ab}\chi _{V}^{}=\chi _{\overline{V}}`$, so that $`{}_{}{}^{a}\chi _{V}^{}=\chi _{\overline{V}}`$. Therefore $`{}_{}{}^{a^n}\chi _{V}^{}=\chi _{\overline{V}}`$ because $`n`$ is odd. On the other hand, since $`a^n`$ is an element of $`H`$, $`{}_{}{}^{a^n}\chi _{V}^{}=\chi _V`$. Therefore $`\chi _V=\chi _{\overline{V}}`$, but this contradicts that $`V\overline{V}`$. Thus $`\chi `$ must be of real type. ∎
Since $`\chi `$ is of real type by the claim above, $`U`$ is irreducible and its character is $`G`$-invariant. It follows that there is a trivial complex $`G`$-vector bundle $`F`$ over $`S(\rho )`$ with $`U`$ as the fiber $`H`$-module, see \[CKMS99, Theorem C (3)\]. Since $`e_1=2`$, there are two $`G_1`$-extensions of $`U`$, say $`\stackrel{~}{U}_1`$ and $`\stackrel{~}{U}_2`$. Their complexifications $`\stackrel{~}{U}_1`$ and $`\stackrel{~}{U}_2`$ are non-isomorphic because $`\stackrel{~}{U}_1\stackrel{~}{U}_2`$. Moreover these modules are both $`G_1`$-extensions of $`U`$. Thus the fiber $`G_1`$-module, say $`F_1`$, of $`F`$ at $`1`$ must be either $`\stackrel{~}{U}_1`$ or $`\stackrel{~}{U}_2`$. It follows that the realification of $`F_1`$ is either $`2\stackrel{~}{U}_1`$ or $`2\stackrel{~}{U}_2`$. Therefore the realification of $`F`$, which is trivial, is isomorphic to one of $`\stackrel{~}{L}_\chi ^\pm `$.
(4) In this case $`\rho (G)=D_n`$ by Table 2 and the fiber $`H`$-modules of the generators are $`U`$. By a similar argument to the case (2) there are two trivial generators, $`L_\chi ^{++}`$ and $`L_\chi ^{}L_\chi ^{++}L_1^{}`$. It suffices to show that the other two generators $`L_\chi ^+L_\chi ^{++}L_1^+`$ and $`L_\chi ^+L_\chi ^{++}L_1^+`$ are nontrivial. Consider the following isomorphisms
(\**)
$$\mathrm{Hom}_H(L_\chi ^{++},L_\chi ^+)\mathrm{Hom}_H(L_\chi ^{++},L_\chi ^{++}L_1^+)\mathrm{Hom}_H(L_\chi ^{++},L_\chi ^{++})L_1^+.$$
Note that $`\mathrm{Hom}_H(L_\chi ^{++},L_\chi ^{++})`$ is isomorphic to the product bundle $`S(\rho )\times ^k`$, where $`k=1,2`$, or $`4`$ according to the type of $`\chi `$. It follows that $`\mathrm{Hom}_H(L_\chi ^{++},L_\chi ^+)kL_1^+`$.
###### Claim.
Both $`kL_1^+`$ and $`kL_1^+`$ are nontrivial for all $`k>0`$.
###### Proof of Claim.
Note that the fiber $`G_1`$-module of $`L_1^+`$ at $`1S(\rho )`$ is the trivial $`G_1`$-module $`_1^+`$ while the fiber $`G_\mu `$-module at $`\mu `$ is the nontrivial $`G_\mu `$-module $`_\mu ^{}`$ by the remark after Lemma 6.1. Then $`b`$ (resp. $`ab`$) acts on $`_1^+`$ (resp. $`_\mu ^{}`$) as multiplication by $`1`$ (resp. $`1`$). Recall that $`H`$ acts on both $`_1^+`$ and $`_\mu ^{}`$ trivially, i.e., as multiplication by $`1`$.
Assume that $`kL_1^+`$ is trivial. Then there exists a $`G`$-module $`W`$ such that $`\mathrm{res}_{G_1}Wk_1^+`$ and $`\mathrm{res}_{G_\mu }Wk_\mu ^{}`$. Thus $`b`$ and $`ab`$ act on $`W`$ as multiplication by $`1`$ and $`1`$, respectively. Hence $`a`$ acts on $`W`$ as multiplication by $`1`$, and since $`n`$ is odd, $`a^n`$ also acts on $`W`$ as multiplication by $`1`$. But this contradicts that $`a^nH`$ acts trivially on $`W`$. In the same way we can prove that $`kL_1^+`$ is also nontrivial. ∎
Since $`L_\chi ^{++}`$ is trivial, $`L_\chi ^+`$ must be nontrivial by the equation (\**) and the claim above. Replacing $`L_\chi ^+`$ by $`L_\chi ^+`$ we can similarly prove that $`L_\chi ^+`$ is nontrivial.
(5) *Case A:* In this case $`\stackrel{~}{\stackrel{~}{L}}_\chi ^\pm `$ are $`N_\chi ^\pm `$ in Lemma 4.1. Since $`n`$ is odd, $`U`$ has a $`G`$-extension by Lemma 2.4 (1). So we may assume that one generator, say $`N_\chi ^+`$, is trivial. Then the following isomorphisms
$$\mathrm{Hom}_H(N_\chi ^+,N_\chi ^{})\mathrm{Hom}_H(N_\chi ^+,N_\chi ^+N_1^{})\mathrm{Hom}_H(N_\chi ^+,N_\chi ^+)N_1^{}N_1^{}$$
imply that $`N_\chi ^{}`$ is nontrivial since $`N_1^{}`$ is nontrivial by Lemma 6.1 (1).
*Case B:* In this case $`\stackrel{~}{\stackrel{~}{L}}_\chi ^\pm `$ are $`M_\chi ^\pm `$ in Lemma 4.5. Remember that $`M_\chi ^+=\mathrm{\Phi }(L_\chi ^{++})=\mathrm{\Phi }(L_\chi ^{})`$ and $`M_\chi ^{}=\mathrm{\Phi }(L_\chi ^+)=\mathrm{\Phi }(L_\chi ^+)`$, see the proof of Lemma 4.5. Since $`L_\chi ^{++}\mathrm{Vect}_G^{}(S(\rho ),\chi )`$ is trivial by Theorem C in \[CKMS99\], $`M_\chi ^+`$ is also trivial. In the following we shall prove that $`M_\chi ^{}`$ is nontrivial.
Assume that $`M_\chi ^{}`$ is trivial, i.e., it is isomorphic to the product bundle $`S(\rho )\times W`$ for some $`G`$-extension $`W`$ of the fiber $`H`$-module $`2U`$.
###### Claim.
$`W`$ is of real type.
###### Proof of Claim.
If $`W`$ is not of real type, then we may view $`M_\chi ^{}`$ as the realification of a complex product bundle $`S(\rho )\times W`$, but this contradicts that $`M_\chi ^{}`$ is the realification of the nontrivial bundles $`L_\chi ^+`$ and $`L_\chi ^+`$ in $`\mathrm{Vect}_G^{}(S(\rho ),\chi )`$. ∎
Denote by $`\chi _W`$ the character of $`W`$. Every fiber $`G_z`$-module of $`M_\chi ^{}`$, which is $`\mathrm{res}_{G_z}W`$, is isomorphic to $`\mathrm{ind}_H^{G_z}U`$ and it is irreducible of complex type by Lemma 2.1 (2). It is well known in representation theory that the character of $`\mathrm{res}_{G_z}W\mathrm{ind}_H^{G_z}U`$ is zero on $`G_zH`$. Thus $`\chi _W`$ is always zero on $`_{zS(\rho )}G_zH=GP`$, where $`P=\rho ^1(Z_n)`$. It follows that we have
$$1=_G\chi _W(g)^2𝑑g=\frac{1}{2}_P\chi _W(p)^2𝑑p+\frac{1}{2}_{GP}\chi _W(p)^2𝑑p=\frac{1}{2}_P\chi _W(p)^2𝑑p,$$
so $`{\displaystyle _P}\chi _W(p)^2𝑑p=2`$. This implies that $`\mathrm{res}_PW`$ is either irreducible of complex type or reducible with different direct summands of real type. In the sequel we show that neither case occurs.
It is easy to see that the latter case does not occur because if it does, then each summand of $`\mathrm{res}_PW`$ is a $`P`$-extension of $`U`$ which contradicts the uniqueness of the $`P`$-extension of $`U`$ by Lemma 2.4 (1).
Now, suppose $`\mathrm{res}_PW`$ is irreducible and of complex type. We claim that the set $`𝒥(W)^G`$ of $`G`$-invariant complex structures on $`W`$ is not empty. Then it contradicts that $`W`$ is of real type. Since $`\mathrm{res}_{G_1}W`$ is irreducible and of complex type, there exists a $`G_1`$-invariant complex structure on $`W`$, i.e., $`(𝒥(W)^H)^{G_1/H}=𝒥(W)^{G_1}\mathrm{}`$. This means that each connected component of $`𝒥(W)^H𝒥(^2)`$ is invariant under the action of $`G_1`$ because $`𝒥(W)^H𝒥(^2)`$ has two connected components. On the other hand, since the order $`n`$ of $`P/H`$ is odd and the number of connected components of $`𝒥(W)^H`$ is two, each connected component is also invariant under the $`P/H`$-action. Therefore, it is invariant under the $`G/H`$-action because $`P`$ and $`G_1`$ generate $`G`$. Now we note that each connected component of $`𝒥(W)^H𝒥(^2)`$ is homeomorphic to $`^2`$ (see \[MS98, Exercise 2.57\]) and that any smooth action of a finite group on $`^2`$ is linear, so the $`G/H`$-action on $`𝒥(W)^H`$ has a fixed point, i.e., $`𝒥(W)^G\mathrm{}`$. ∎
|
warning/0005/cond-mat0005226.html
|
ar5iv
|
text
|
# Caught Active Walkers
## 1 INTRODUCTION
Studies of the interaction in many-particle systems has revealed a wide range of fascinating phenomena. Pattern formation and self-organization has been found in chemistry, physics and biology . The same ideas and principles can even be applied in sociology and economics . The analytical approach usually models a system by differential equations. Thus fluids, reaction-diffusion systems and many other systems have been studied. Since several years, the enhancing capacity of computers allows to pursue also the discrete, numerical approach. Attention has also been paid to the simulation of traffic and pedestrian motion .
Microsimulations of traffic, pedestrian and also granular flow compute the motion of each distinct object . For example, the motion of a granular bead is determined by its interaction with the environment, i.e. walls, other beads, electrical fields and possibly further more. Striking correspondence between simulation and experiment can be found .
## 2 Active Walker Model
The Active Walker Model is based on the social force concept . According to this idea, environmental influences causing behaviorial changes of an individual are modelled by a social force. Thus the change in velocity $`d\stackrel{}{v}/dt`$ of a walker is determined by the social force $`\stackrel{}{F}_s`$. $`\stackrel{}{F}_s`$ depends not only on the environment but also on the preferences and aims of the individual. $`\stackrel{}{F}_s`$ is often calculated from a social or environmental potential $`V_{\mathrm{env}}`$. The walker is called “active” because he is able to change the environment locally. For example, walkers may increase the walking comfort by their footprints and thus establish a beaten path .
## 3 The Discrete Model
Let the world be a two-dimensional lattice and let time pass in discrete steps. At each time $`t`$, a walker has a certain position $`\stackrel{}{x}(t)`$. Furthermore, we assume that this walker is on his way to a distinct destination $`\stackrel{}{d}`$ on our lattice. If reaching his destination were his one and only will, he would take the direct path along $`\stackrel{}{d}\stackrel{}{x}`$. But even in a two-dimensional lattice world life may become more complicated. Let each lattice site $`\stackrel{}{x}=(x,y)`$ has a set
$$𝒩(\stackrel{}{x})=\left\{\begin{array}{c}\stackrel{}{n}_1=(x1,y),\hfill \\ \stackrel{}{n}_2=(x+1,y),\hfill \\ \stackrel{}{n}_3=(x,y1),\hfill \\ \stackrel{}{n}_4=(x,y+1)\hfill \end{array}\right\}$$
(1)
of four neighbouring sites. If $`\stackrel{}{x}(t)`$ is the position of the walker at time $`t`$, $`p(\stackrel{}{n})`$ is the probability that $`\stackrel{}{n}𝒩(\stackrel{}{x}(t))`$ will become the next position $`\stackrel{}{x}(t+1)`$ of the walker. The neighbouring site with the highest value of $`p`$ becomes the position of the walker at time $`t+1`$. $`p(\stackrel{}{n})`$ is the mathematical expression of the walker’s aims. If the walker just wants to reach his destination without taking the environment into account, the following expression for $`p(\stackrel{}{n})`$ would be reasonable:
$$p(\stackrel{}{n})=|\stackrel{}{x}(t)\stackrel{}{d}||\stackrel{}{n}\stackrel{}{d}|.$$
(2)
In general, the influence of the environment can be modelled by a environmental potential $`V_{\mathrm{env}}(\stackrel{}{x},t)`$. Thus, the probability $`p(\stackrel{}{n})`$ becomes
$$p(\stackrel{}{n})=w_{\mathrm{dist}}(|\stackrel{}{x}(t)\stackrel{}{d}||\stackrel{}{n}\stackrel{}{d}|)+w_{\mathrm{pot}}V_{\mathrm{env}}(\stackrel{}{n},t),$$
(3)
i.e. we calculate the weighted sum of the aim to reach the destination and to maximize the environmental potential. The condition $`w_{\mathrm{dist}}+w_{\mathrm{pot}}=1`$ has to be fulfilled.
The environmental potential $`V_{\mathrm{env}}(\stackrel{}{x},t)`$ determines how comfortable the walker feels at site $`\stackrel{}{x}`$ at time $`t`$. It thus depends heavily on the individual conditions and aims of the walker. For example, if a walker runs down a street in order to reach a bus station at the end of this street and he is also interested in books, he will possibly try to pass book stores to catch a glimpse. So the environmental potential $`V_{\mathrm{env}}(\stackrel{}{x},t)`$ near book stores would be higher than near clothes stores (of course!).
In our model, $`V_{\mathrm{env}}(\stackrel{}{x},t)`$ reflects the walking comfort $`C(\stackrel{}{x},t)`$ at site $`\stackrel{}{x}`$ at time $`t`$. If a walker passes a site $`\stackrel{}{x}`$, the walking comfort $`C(\stackrel{}{x},t)`$ will increase (because hindering vegetation is being damaged). Several walkers running the same way may cause a beaten path on the long term. On the other hand, $`C(\stackrel{}{x},t)`$ decreases by time as beaten paths vanish if they are not used. Thus, the following differential equation holds for $`C(\stackrel{}{x},t)`$:
$$\frac{\mathrm{d}C(\stackrel{}{x},t)}{\mathrm{d}t}=1/\tau [C_{min}C(\stackrel{}{x},t)]+I(\stackrel{}{x},t)[C_{\mathrm{max}}C(\stackrel{}{x},t)].$$
(4)
$`C_{min}`$ is the minimum walking comfort and $`C_{\mathrm{max}}`$ is the maximum one. $`I(\stackrel{}{x},t)`$ describes the intensity site $`\stackrel{}{x}`$ is frequented by walkers at time $`t`$. $`1/\tau `$ quantifies how fast a beaten path weathers.
The environmental potential $`V_{\mathrm{env}}(\stackrel{}{x},t)`$ is mainly but not exclusively determined by the walking comfort $`C(\stackrel{}{x},t)`$ at site $`\stackrel{}{x}`$. If we used only $`C(\stackrel{}{x},t)`$, the walker would not be able to recognize a comfortable site at some distance and move to it. Clearly this is not realistic. So we calculate $`V_{\mathrm{env}}(\stackrel{}{x},t)`$ as the distance-weighted sum of the walking comforts of all lattice sites:
$$V_{\mathrm{env}}(\stackrel{}{x},t)]=_{\stackrel{}{y}\stackrel{}{x}}e^{|\stackrel{}{y}\stackrel{}{x}|/s(\stackrel{}{x},t)}C(\stackrel{}{y},t),$$
(5)
where $`s(\stackrel{}{x},t)`$ indicates how far one can see at site $`\stackrel{}{x}`$ at time $`t`$.
## 4 First Results
Fig. 1 illustrates how our model works. We use a $`(40,40)`$–lattice and let a walker start at $`(1|36)`$ with destination $`(40|6)`$. All lattice sites has the same initial walking comfort $`C_0`$ apart from the sites on the beaten path in the middle between $`(1,21)`$ and $`(40|21)`$ shown as a solid line in Fig. 1. If we use $`w_{\mathrm{pot}}=0.005`$ in our discrete active walker model, the walker’s path is hardly influenced by the environment. The walker looks for the fastest way to his destination. As Fig. 1 reveals, the “fastest” way on our two-dimensional lattice is not the same as in a continuous two-dimensional world. Initially, the walker moves horizontally. Afterwards, he follows a diagonal path to his destination. This strange behaviour results from the combination of restricted movements (horizontal and vertical steps only) and the usage of the Euclidian distance in equ. 2. Nevertheless, this effect is not important for the results presented in this paper. A close look on Fig. 1 shows that the walker follows the beaten path just for 3 steps.
If $`w_{\mathrm{pot}}`$ is increased to $`0.05`$, the walkers quits its “direct” way to the destination in order to follow the beaten path. One can also see the effect of the distance-weighted sum in equ. 5 for $`V_{\mathrm{env}}`$: The walker is actually attracted by the beaten path. Finally, for $`w_{\mathrm{pot}}=0.85`$ the influence of the beaten path increases again.
We define the mean difference $`\overline{\mathrm{\Delta }}`$ between the actual path of a walker and the “direct” path as
$$\overline{\mathrm{\Delta }}=\frac{1}{n}\underset{\mathrm{i}=1}{\overset{n}{}}|\stackrel{}{x}_\mathrm{i}\stackrel{}{x}_{\mathrm{i},\mathrm{dir}}|,$$
(6)
where $`\stackrel{}{x}_\mathrm{i},\mathrm{i}=1,\mathrm{},\mathrm{n}`$ are the $`\mathrm{n}`$ actual positions of the walker on its way and $`\stackrel{}{x}_{\mathrm{i},\mathrm{dir}},\mathrm{i}=1,\mathrm{},\mathrm{n}`$ are the positions on the direct path to the walker’s destination. $`\stackrel{}{x}_{\mathrm{i},\mathrm{dir}}`$ depends on $`\stackrel{}{x}_\mathrm{i}`$ by equal x values: $`x_{\mathrm{i},\mathrm{dir}}=x_\mathrm{i}`$. $`\overline{\mathrm{\Delta }}`$ describes how strong the walker swerved from the direct way to its destination. Fig. 2 shows the reasonable result that $`\overline{\mathrm{\Delta }}`$ increases if the influence of the environment becomes higher.
Let us do now a more interesting simulation. Initially, all sites have the same walking comfort. We use a $`(50,50)`$–lattice. A walker starts from a random position on the left edge to a random position on the right edge. He will definitely use the shortest way to his destination because there are no beaten paths so far. After the first walker has reached his destination, we continue with a second walker somewhere on the left border of the lattice and again a random destination on the right. This walker might be affected by the beaten path the first one has left on the lattice (see fig. 3). My initial interest was to study the lattice and the walkers after say 1000 iterations:
1. Are there persistent beaten paths? How long do they remain?
2. What quantity may describe the change of the beaten path pattern from one iteration to the next?
3. What is the influence of the parameters $`s`$, $`\tau `$ and $`w_{\mathrm{pot}}`$?
## 5 Caught Walkers
At some point in my studies, I wondered why a walker seems to never reach its destination. I had a closer look and found out that he was bouncing from one site to a contiguous one and vice versa all the time. First, I thought of an error in the implementation of the model. But I could not find any. So I checked the algorithm by calculating $`p(\stackrel{}{n})`$ by hand - and astonishingly, my program was correct.The explanation is simple. In direction of his destination the sites offer much less walking comfort than the site he has been before. Although this artefact obviously diminishes the applicability of the model, it is worth a close look.
It was obvious that the number of walkers having reached their destinations successfully before a walker is being caught differed from one simulation run to the next. Thus I was interested in the probability distribution $`p(n)`$ of the number of walkers $`n`$ reaching their destination. Further on I investigated how this probability distribution $`p(n)`$ depended on the parameters of the model, i.e. $`w_{\mathrm{pot}}`$, $`\tau `$ or even the size of the lattice. The distributions $`p(n)`$ shown below are normalized:
$$_0^{n_{\mathrm{max}}}p(n)\mathrm{dn}\underset{n=0}{\overset{n_{\mathrm{max}}}{}}p(n)\mathrm{\Delta }n=1.$$
(7)
$`\mathrm{\Delta }n=1`$ is the bin width of the discrete probability distribution and $`n_{\mathrm{max}}`$ is the maximum number of walkers.
Fig. 4 elucidates the dependency of $`p(n)`$ on $`w_{\mathrm{pot}}`$. Generally speaking, if the influence of the environment and thus $`w_{\mathrm{pot}}`$ is increased, the walkers are being caught faster. $`p(n)`$ shows a distinct maximum. The plots for $`w_{\mathrm{pot}}=0.055`$ and $`0.5`$ show a peak at $`n=220`$. Because the simulation was stopped after 220 walkers had reached their destinations, the probability $`p(n220)`$ is aggregated in $`p(220)`$. Only for small values of $`w_{\mathrm{pot}}`$ $`p(n220)`$ is neglectable.
The half-logarithmic plot in fig. 4a shows that $`p(n)`$ decreases exponentially for increasing $`n`$.
Where are the “prisons” of the walkers located on the lattice? In order to answer this question, I determined the normalized probability distribution $`p(\stackrel{}{x})`$ for the locations $`\stackrel{}{x}`$ of the prisons. $`p(\stackrel{}{x}=(x|y))`$ fulfills the equation
$$_x_yp(x,y)\mathrm{dxdy}\underset{x=0}{\overset{x_{\mathrm{max}}}{}}\underset{y=0}{\overset{y_{\mathrm{max}}}{}}p(x,y)\mathrm{\Delta }x\mathrm{\Delta }y=1,$$
(8)
where $`x_{\mathrm{max}}=y_{\mathrm{max}}=40`$ is the size of the lattice and $`\mathrm{\Delta }x=\mathrm{\Delta }y=1`$ the size of the bins.
Fig. 5 shows that $`p(\stackrel{}{x})`$ is nearly symmetrical to $`y=20`$. Please note that for $`x<12`$ and $`x>32`$ no prisons are found at all. The highest values for $`p(\stackrel{}{x})`$ are found for $`x=32`$. Astonishingly, walkers are not being caught next to the right border of the lattice.
The probability distributions $`p(n)`$ for different $`\tau `$ and different lattice sizes are shown in fig. 6 in half-logarithmic plots. They are very similar to the curves discussed above and also show exponential decay for increasing $`n`$.
Now we have a closer look on the maxima of the plots in fig. 4a. The maxima are found at different positions $`n_{\mathrm{max}}`$ and have different heights $`p_{\mathrm{max}}`$. Fig. 7 shows $`n_{\mathrm{max}}`$ and $`p_{\mathrm{max}}`$ for different values of $`w_{\mathrm{pot}}`$. Obviously, $`n_{\mathrm{max}}`$ decreases for higher $`w_{\mathrm{pot}}`$ whereas the maxima becomes higher. Especially $`p_{\mathrm{max}}`$ shows nearly a linear dependency on $`w_{\mathrm{pot}}`$.
## 6 Outlook
The described “catchment” phenomenon is an artefact of the introduced discrete active walker model. Thus the model has to be improved in order to allow real-life simulations. On the other hand, it would be interesting to compare the described statistical characteristics of the catchment phenomenon with those of real-world catchment effects for example in particle physics.
———————————————————————-
|
warning/0005/cond-mat0005036.html
|
ar5iv
|
text
|
# New Class of Magnetoresistance Oscillations: Interaction of a Two-Dimensional Electron Gas with Leaky Interface Phonons
thanks: Present address: Stanford Picosecond Free Electron Laser Center, Stanford University, Stanford, CA 94305
## Abstract
We report on a new class of magnetoresistance oscillations observed in a high-mobility two-dimensional electron gas (2DEG) in GaAs-Al<sub>x</sub>Ga<sub>1-x</sub>As heterostructures. Appearing in a weak magnetic field ($`B<`$ 0.3 T) and only in a narrow temperature range (2 K $`<T<`$ 9 K), these oscillations are periodic in $`1/B`$ with a frequency proportional to the electron Fermi wave vector, $`k_F`$. We interpret the effect as a magnetophonon resonance of the 2DEG with leaky interface-acoustic phonon modes carrying a wave vector $`q=2k_F`$. Calculations show a few branches of such modes on the GaAs-Al<sub>x</sub>Ga<sub>1-x</sub>As interface, and their velocities are in quantitative agreement with the data.
There are several classes of transverse magnetoresistance (MR) oscillations known to exist in a two-dimensional homogeneous electron gas (2DEG). The most common of these are the Shubnikov-de Haas oscillations (SdH), which arise from a magnetic field $`B`$-induced modulation of the density of states at the Fermi level $`E_F`$. They become more pronounced with decreasing temperature $`T`$. The magnetophonon resonance (MPR) mpr ; tsui is a source of another class of oscillations resulting from the absorption of bulk longitudinal optical phonons. These resonances appear under the condition $`\omega _{LO}=l\omega _c`$, where $`\omega _{LO}`$ and $`\omega _c=eB/mc`$ are the optical phonon and cyclotron frequencies respectively, $`l`$ is an integer, and $`m`$ is the effective mass of the carriers. These oscillations are only seen at relatively high $`T100180`$K tsui . Both SdH and MPR are periodic in $`1/B`$, but the SdH frequency (reciprocal period) scales with electron density as $`n_e`$, whereas MPR is $`n_e`$-independent.
In this Letter, we report on a new class of MR oscillations zudov observed in a high-mobility 2DEG in GaAs-Al<sub>x</sub>Ga<sub>1-x</sub>As heterostructures. Unrelated to either of the above origins, these novel oscillations are still periodic in $`1/B`$, but they appear only in a narrow temperature range (2 K $`<T<`$ 9 K), and their frequency scales with $`\sqrt{n_e}`$. We interpret the data in terms of a magnetophonon resonance mediated by thermally excited leaky interface-acoustic phonon (LIP) modes. In principle, the surface modes might provide a good explanation as well, but in our case 2DEG is located so far from the surface ($`0.5`$ $`\mu `$m) that no such interaction is possible.
The leaky interface modes have been studied for a few decades in connection with the Earth’s crust maradudin . The term “leaky” shows that the waves propagate at a small angle with the interface, so that the energy radiates away from the boundary. For some specific parameters these waves may not be leaky stoneley , but for the interface under study all of them are leaky. Despite the fact that LIP is commonly present in layered material systems zinin , it has so far not been considered on the GaAs-Al<sub>x</sub>Ga<sub>1-x</sub>As interface. Due to radiation of energy, the frequency and velocity of leaky waves are complex: $`u=\omega /q=u_Riu_I`$ with $`u_Iu_R`$.
The novel oscillations can be explained by a simple momentum selection rule which is derived later in the paper. It states that at high Landau levels (LLs) the electrons interact predominantly with the interface phonons carrying a wave vector $`q=2k_F`$, where $`k_F`$ is the Fermi wave vector of the 2DEG at zero $`B`$ field. The condition for resonant absorption or emission of an interface phonon is then given by
$$2k_Fu_R=l\omega _c,l=1,2,3,\mathrm{}.$$
(1)
We claim that Eq. (1) determines the values of $`B`$ for the maxima in these new MR oscillations. It shows that the oscillations are periodic in $`1/B`$ with a frequency $`f=2k_Fumc/e`$. Evidently, the bulk phonons can not account for the resonance, since their frequency depends on $`q_z`$, while the selection rule includes lateral momentum only.
Our primary samples are lithographically defined Hall bars cleaved from modulation-doped GaAs-Al<sub>0.3</sub>Ga<sub>0.7</sub>As heterostructures of high-mobility $`\mu 3\times 10^6`$ cm<sup>2</sup>/Vs. The wafers are grown by molecular-beam epitaxy on the (001) GaAs substrate. At low $`T`$, the density of the 2DEG, $`n_e`$ (in units of $`10^{11}`$ cm<sup>-2</sup> throughout the text), can be tuned by a combination of illumination from light-emitting diode and the NiCr front gate potential. The experiments were performed in a variable-temperature <sup>4</sup>He cryostat equipped with a superconducting magnet, employing a standard low-frequency lock-in technique for resistance measurement.
In Fig. 1 we show the normalized low-field magnetoresistivity $`\rho _{xx}(B)/\rho _{xx}(0)`$ measured at $`T=4`$ K, for the electron density $`n_e=`$ 2.05, 2.27, and 2.55, respectively. In addition to the damped SdH commonly seen in a 2DEG at this $`T`$, the traces reveal new oscillations that appear only at $`B<0.3`$ T. The amplitude of the oscillations is about 2-3 % in these traces.
Three aspects of the observation should be highlighted. First, the oscillations are roughly periodic in inverse magnetic field, $`1/B`$. The arrows next to the traces indicate the $`\rho _{xx}(B_l)`$ maxima (indexed as $`l`$ = 1, 2, 3, 4) in this oscillatory structure. In the inset we plot the order of the oscillations, $`l`$ (and $`d^2\rho _{xx}/dB^2`$), vs. $`1/B`$ for $`n_e=`$ 2.55 and observe a linear dependence. Such periodic oscillations have been seen for all $`n_e`$ (from $`1.5`$ to 3) studied. Second, with increasing $`n_e`$ the features shift orderly towards higher $`B`$. Finally, the oscillatory structure is accompanied by a negative MR background, apparently in the same $`B`$ range where the oscillations take place.
We have now measured over a dozen specimen (from five wafers), of both the Hall bar (width from 10 $`\mu `$m to 500 $`\mu `$m) and the square (5 mm x 5 mm) geometries, and consistently observed similar oscillatory structures. On the other hand, the significance of the ubiquitous MR background remains unclear. Either negative or positive MR has been observed, and its strength (and even the sign) is largely specimen and cooling-cycle dependent. In the following we shall focus on the analysis of the oscillatory structure, in particular, its dependence on $`n_e`$ and $`T`$.
To further quantify our results, we have performed fast Fourier transform (FFT) on the resistance data. As an example, Fig. 2 shows the FFT power spectra obtained from the three traces in Fig. 1 fft . Surprisingly, such analysis has uncovered two frequencies, marked by $`A`$ and $`B`$. Peak $`A`$ corresponds to the main period, conforming to the simple fit in Fig. 1. Peak $`B`$ is somewhat weaker, and occurs at $`f_B1.5f_A`$. The shift of the doublet with increasing $`n_e`$ is marked by three arrows for the main peak. The FFT data have revealed a striking linear relation between the frequencies of oscillations and the electron Fermi wave vector. We plot (see the inset) $`f^2`$ of the FFT peaks against the electron density, $`n_e`$, which has been varied from 1.47 to 2.95 in the same specimen. Since $`k_F=\sqrt{2\pi n_e}`$, the observed linearity indicates that $`fk_F`$.
Such a linear dependence distinguishes the new oscillations from SdH, as $`f_{SdH}k_F^2`$, and is exactly what one expects from the phonon resonance scenario proposed here. As such, the oscillatory structure must be viewed as resulting from the resonance of the 2DEG with two branches of the interface modes. Using Eq. 1 and a single known material parameter, the GaAs band electron mass $`m0.068m_e`$, we fit the data (solid line in the inset) and deduce a velocity for the slow (fast) mode $`u_A`$ 2.9 km/s ($`u_B4.4`$ km/s). Within the experimental error of 10% the data from several specimen collapse on the same lines, indicating that the new oscillations are generic in high-mobility 2DEG in GaAs-Al<sub>x</sub>Ga<sub>1-x</sub>As heterostructures.
The $`T`$-dependence of the oscillations is consistent with a thermally excited phonon-scattering model. Fig. 3 shows the $`\rho _{xx}(B)`$ at selected temperatures (1.9 K $`<T<`$ 9.1 K ) where the evolution of the oscillations is clearly seen. Notice first (see inset) that $`\rho _{xx}(0)`$ grows linearly with $`T`$, indicating that acoustic-phonon scattering dominates the electron mobility in this temperature range mendez ; stormer . Considering the interface phonon modes of interest here, we use the value of the slow mode $`u_A=2.9`$ km/s to estimate a characteristic temperature, $`T_c`$, from $`k_BT_c=\mathrm{}u_A(2k_F)`$. The value of $`T_c`$ 5 K can qualitatively account for the temperature dependence of the main features of the oscillations. While the SdH gradually diminishes as $`T`$ increases, the oscillations are best developed at $`T37`$ K and are strongly damped at both higher and lower $`T`$. At $`TT_c`$ the number of interface phonons carrying $`q=2k_F`$ becomes small and therefore the amplitudes diminish. At high $`T`$ the smearing of the LLs prevails and the oscillations disappear as well.
We now turn to the details of the theoretical explanation of the novel oscillations. We have performed calculations efros of LIP modes for the GaAs-Al<sub>0.3</sub>Ga<sub>0.7</sub>As interface on the basal (001) plane. In the anisotropic case the speed of LIPs depends on angle between $`q`$ and the direction. Using the elastic moduli of the bulk lattices bulk we found a series of modes with weak anisotropy and a small imaginary part of the velocity ($`u_I/u_R<0.03`$). We have studied the modes within the interval of velocities 2.4$``$6.0 km/s. Two close groups of modes have been found, one within the interval of 3$``$3.5 km/s and the other within 4.2$``$4.5 km/s. These modes may be responsible for the two periods of oscillations which have been observed. The frequencies of the other modes found are too high to be detected in our experiment. Note that different modes may interact with electrons with different strengths.
To calculate the transverse conductivity due to the scattering of the 2DEG by the LIPs, we employ a 2D analog of the formula, first derived by Titeica titeica :
$`\sigma _{xx}={\displaystyle \frac{4\pi e^2}{Am^2k_BT\omega _c^2}}{\displaystyle \underset{n,n^{}}{}}{\displaystyle \underset{k_y,k_y^{}}{}}{\displaystyle \underset{q_x,q_y}{}}|I_{nn^{}}(q\lambda )|^2q_y^2|C(q)|^2`$
$`\times N_lf_n(1f_n^{})\delta _{k_yk_y^{}+q_y}\delta (\omega _c(n^{}n)qu).`$ (2)
Here $`A`$ is the area, $`N_l=\left(\mathrm{exp}(\mathrm{}\omega /k_BT)1\right)^1`$, $`f_n=\left(\mathrm{exp}\left((E_n\mu )/k_BT\right)+1\right)^1`$, $`\lambda =\sqrt{\mathrm{}c/eB}`$ is the magnetic length, and $`|C(q)|^2v(q)/A`$ is the square modulus of the 2DEG-LIP interaction, which has a power law dependence on $`q`$. This formula can be interpreted in the following way. A 2D electron in a magnetic field has a wave function which is a product of a plane wave in the $`y`$ direction and an oscillatory wave function, centered at the position $`x_0=c\mathrm{}k_y/eB`$: $`\mathrm{\Psi }=\mathrm{exp}(ik_yy)\varphi _n(xx_0)`$, where $`n`$ is the LL index. In the absence of scattering the electric current may flow only in the $`y`$-direction, providing the Hall effect. A transverse conductivity appears because an electron transfers wave vector $`q_y=k_y^{}k_y`$ to a scatterer. This is equivalent to a jump in the $`x`$-direction at a distance $`\mathrm{\Delta }x_0=c\mathrm{}q_y/eB`$. In Eq. (New Class of Magnetoresistance Oscillations: Interaction of a Two-Dimensional Electron Gas with Leaky Interface Phonons) this physics is applied to electron-interface phonon scattering. The mechanism of the 2DEG-LIP interaction, which may be either deformation potential or piezoelectric interaction, is not particularly important for our purpose. The overall scattering is of the same order as the bulk phonon scattering, since the energy densities of both excitations are of the same order in the vicinity of the interface.
If the interface phonon has no attenuation, the square of matrix element $`I_{nn^{}}`$ is given by sdh
$`|I_{n,n+l}(b)|^2`$ $`=`$ $`\left|{\displaystyle _{\mathrm{}}^+\mathrm{}}e^{iq_xx}\varphi _n(xx_0)\varphi _{n+l}(xx_0^{})𝑑x\right|^2`$ (3)
$`=`$ $`{\displaystyle \frac{n!}{(n+l)!}}\left({\displaystyle \frac{b^2}{2}}\right)^le^{\frac{b^2}{2}}\left[L_n^l\left({\displaystyle \frac{b^2}{2}}\right)\right]^2,`$
where $`b=q\lambda `$ and $`L_n^l(x)`$ is the generalized Laguerre polynomial. Substituting summation over wave vectors by integration in Eq. (New Class of Magnetoresistance Oscillations: Interaction of a Two-Dimensional Electron Gas with Leaky Interface Phonons) one obtains
$`\sigma _{xx}={\displaystyle \frac{e^2}{2\pi \mathrm{}}}{\displaystyle \frac{1}{mk_BT\omega _c}}{\displaystyle \underset{n,l}{}}N_lf_n(1f_{n+l})`$
$`\times {\displaystyle _0^{\mathrm{}}}dqq^3v(q)|I_{nn+l}(q\lambda )|^2\delta (\omega _clqu).`$ (4)
Taking into account the imaginary part of the LIP frequency, $`\omega =q(u_Riu_I)`$, we can substitute for the $`\delta `$-function in Eq. (New Class of Magnetoresistance Oscillations: Interaction of a Two-Dimensional Electron Gas with Leaky Interface Phonons) a Gaussian distribution with appropriate dispersion $`\sigma =qu_I`$. Since the dispersion is small we can set $`q=\omega _cl/u_R`$ everywhere except for the strongly oscillating function $`|I_{nl}(q\lambda )|^2`$. Then after averaging we obtain for the transverse conductivity
$$\sigma _{xx}=\frac{e^2/2\pi \mathrm{}}{mu\omega _ck_BT}\underset{l,n}{}v\left(\frac{\omega _cl}{u}\right)F_{nl}N_lf_n(1f_{n+l}),$$
(5)
where the function $`F_{nl}`$ can be expressed as a series of Hermite polynomials of imaginary argument:
$`F_{nl}`$ $`=`$ $`{\displaystyle \frac{(\omega _cl/u)^3}{\sqrt{1+\alpha ^1}}}\mathrm{exp}\left({\displaystyle \frac{\alpha }{1+\alpha }}{\displaystyle \frac{\mathrm{}\omega _cl^2}{2mu^2}}\right)`$
$`\times {\displaystyle \underset{k,j}{\overset{n}{}}}{\displaystyle \frac{n!(1)^l}{(n+l)!k!j!}}\left(\begin{array}{c}n+l\\ nk\end{array}\right)\left(\begin{array}{c}n+l\\ nj\end{array}\right)`$
$`\times {\displaystyle \frac{H_{2(k+j+l)}\left(il\sqrt{\frac{\mathrm{}\omega _c}{2mu^2}}\frac{\alpha }{\sqrt{1+\alpha }}\right)}{[2\sqrt{1+\alpha }]^{2(k+l+j)}}},`$
with $`\alpha =(u/\sigma \lambda )^2`$. Hereafter, we assume that $`u`$ is the real part of the LIP velocity. In Fig. 4 we plot $`F_{nl}`$ for $`n=17`$ and $`l=1`$ as a function of $`B`$ for LIP with $`\sigma =\omega _cu_I/u_R`$ (solid line) and in the limit $`\sigma =0`$ (dashed line).
As we can see, once attenuation is introduced, only one strong peak remains that corresponds to Eq. (1) at $`l=1`$. This means that only phonons with wave vector $`q=2\sqrt{2n}/\lambda `$ effectively interact with electrons under the condition $`n1`$. In fact, due to the Fermi distribution in Eq. (5) only the values of $`nE_F/\mathrm{}\omega _c`$ are important. Then, indeed, $`2\sqrt{2n}/\lambda =2\sqrt{2\pi n_e}=2k_F`$, and we arrive at Eq. (1) for $`l=1`$. The same conclusion holds for any $`ln`$.
This is an important result of our work. It can be interpreted from the following semi-classical consideration (see the inset in Fig. 4). Let us consider $`nn^{}1`$. Since the square of the matrix element in Eq. (3) depends on $`q`$ only, we can put $`q_x=0`$. Then the integrand in (3) is an overlap of two oscillatory wave functions shifted with respect to each other. In the vicinity of the turning point the wave function always has a maximum since the momentum is small and the particle spends most of its time there. There are three possibilities. Cases 1 and 3 in the inset show situations when turning points are apart from each other, and case 2 occurs when the turning points coincide in space. Obviously, in case 2 the overlap integral has a maximum. This occurs when $`m\omega _c^2(\mathrm{\Delta }x_0)^2/8=n\mathrm{}\omega _c`$, which is equivalent to the above condition $`q\lambda 2\sqrt{2n}`$. Note that the other maxima in Fig. 4 can be smeared very easily because their widths are proportional to $`n^{1/2}`$, while the first maximum near the turning point can be approximated by an Airy function and its width is independent of $`n`$. Thus, the maxima in $`F_{nl}(B)`$ for different $`l`$ give rise to oscillations in $`\rho _{xx}(B)`$.
As a whole, the results provide good agreement with our experimental data. In particular, the slow mode $`u_A`$ can be identified with the lower bunch of modes calculated here. Within the experimental uncertainty we are unable to find any anisotropy for the velocity, therefore, the data must be viewed as an average over all directions. Likewise, the velocities of fast modes coincide with $`u_B=`$ 4.4 km/s, but this should be taken with caution. Since our experiments have so far been centered on a temperature range around 5 K, a positive identification of the fast mode awaits for a more detailed $`T`$-dependence study at higher temperatures.
In conclusion, we have discovered a new class of magneto-oscillations in a high-mobility 2DEG and interpreted it as a magnetophonon resonance with leaky interface-acoustic phonons. Owning to their 2D characteristics, the leaky interface modes play a unique role in the scattering of 2D electrons in GaAs-Al<sub>x</sub>Ga<sub>1-x</sub>As heterostructures and quantum wells. This role has never been studied before.
The experimental work (M.A.Z. and R.R.D.) is supported by NSF grant DMR-9705521. R.R.D. also acknowledges an Alfred P. Sloan Research Fellowship and thanks M. E. Raikh for helpful conversations. The theoretical work (I.V.P. and A.L.E.) is supported by a seed grant of the University of Utah. A.L.E. is grateful to R. L. Willett for insightful discussions. The work at Sandia is supported by the US DOE under contract DE-AC04-94AL85000.
|
warning/0005/physics0005019.html
|
ar5iv
|
text
|
# Clase 1
## Clase 1
### Generalidades
Los campos eléctrico ($`\stackrel{}{E}`$) y de inducción magnética ($`\stackrel{}{B}`$) se introdujeron originalmente a través de la fuerza ejercida por cargas ($`q^{}`$) o corrientes ($`I`$) sobre una carga de prueba ($`q`$):
$`\stackrel{}{F}_q=q\stackrel{}{E}`$ $``$ $`\stackrel{}{E}={\displaystyle \frac{1}{q}}\stackrel{}{F}_q,`$
$`d\stackrel{}{F}_I=Id\stackrel{}{l}\times \stackrel{}{B}`$ $``$ $`B={\displaystyle \frac{|d\stackrel{}{F}_I|}{Idl\mathrm{sin}\alpha }}.`$
De acuerdo con esto, $`\stackrel{}{E}`$ se interpreta como fuerza por unidad de carga y $`\stackrel{}{B}`$ como fuerza por unidad de corriente. Sin embargo ambos campos tienen significado propio, independiente del tipo de fuente que los genera.
Ahora bien, $`\stackrel{}{E}`$ y $`\stackrel{}{B}`$ no son los únicos campos importantes en la electrodinámica. En la mayoría de las sustancias:
$`\stackrel{}{D}=ϵ_0\stackrel{}{E}+\stackrel{}{P},`$
$`\stackrel{}{H}={\displaystyle \frac{1}{\mu _0}}\stackrel{}{B}\stackrel{}{M}.`$
$`\stackrel{}{D}`$ se conoce como desplazamiento eléctrico, $`\stackrel{}{H}`$ campo magnético, $`\stackrel{}{P}`$ y $`\stackrel{}{M}`$ son, respectivamente, las polarizaciones eléctrica y magnética (i. e., representan el promedio macroscópico de dipolos eléctricos$`/`$magnéticos en el material en presencia de campos); $`ϵ_0=8.85\times 10^{12}C/Nm^2`$, $`\mu _0=4\pi \times 10^7H/m`$ y $`ϵ_0\mu _0=c^2`$. La conexión entre los vectores $`(\stackrel{}{P},\stackrel{}{E})`$ y $`(\stackrel{}{M},\stackrel{}{H})`$ está determinada por las propiedades de cada sustancia. Para medios anisotrópicos la aproximación lineal en los campos es:
$`P_i=ϵ_0\alpha _{ik}E_k,`$
$`M_i=\kappa _{ik}H_k,`$
con $`i,k=1,2,3`$; $`\alpha `$ es el tensor de polarizabilidad y $`\kappa `$ el tensor de magnetización. Entonces
$`D_i=ϵ_{ik}E_k,`$
$`B_i=\mu _{ik}H_k,`$
donde $`ϵ_{ik}ϵ_0(\delta _{ik}+\alpha _{ik})`$, $`\mu _{ik}\mu _0(\delta _{ik}+\kappa _{ik})`$. Para medios isotrópicos:
$`\stackrel{}{P}=ϵ_0\alpha \stackrel{}{E}`$
$`\stackrel{}{M}=\kappa \stackrel{}{H},ϵϵ_0(1+\alpha )`$
$`\stackrel{}{D}=ϵ\stackrel{}{E},\mu \mu _0(1+\kappa )`$
$`\stackrel{}{B}=\mu \stackrel{}{H}.`$
Una vez definidos estos vectores, podemos presentar las ecuaciones de Maxwell (1873), que son en electrodinámica lo que las leyes de Newton en mecánica. Las ecuaciones de Maxwell en forma diferencial son
$`\times \stackrel{}{H}=\stackrel{}{j}+{\displaystyle \frac{\stackrel{}{D}}{t}},`$ (1)
$`\stackrel{}{B}=0,`$ (2)
$`\times \stackrel{}{E}={\displaystyle \frac{\stackrel{}{B}}{t}},`$ (3)
$`\stackrel{}{D}=\rho `$ (4)
($`\rho `$ es la densidad de carga y $`\stackrel{}{j}`$ la densidad de corriente). La forma integral de estas ecuaciones es
$`{\displaystyle _C}\stackrel{}{H}𝑑\stackrel{}{l}={\displaystyle _S}\left(\stackrel{}{j}+{\displaystyle \frac{\stackrel{}{D}}{t}}\right)\widehat{n}𝑑A,`$
$`{\displaystyle _S}\stackrel{}{B}\widehat{n}𝑑A=0`$
$`{\displaystyle _C}\stackrel{}{E}𝑑\stackrel{}{l}={\displaystyle _S}{\displaystyle \frac{\stackrel{}{B}}{t}}\widehat{n}𝑑A,`$
$`{\displaystyle _S}\stackrel{}{D}\widehat{n}𝑑A={\displaystyle _V}\rho 𝑑V.`$
De estas últimas se obtienen las condiciones de frontera entre dos medios:
$`(\stackrel{}{D}_2\stackrel{}{D}_1)\widehat{n}_{1,2}=\sigma ,`$ (5)
$`\widehat{n}_{1,2}\times (\stackrel{}{E}_1\stackrel{}{E}_2)=0,`$ (6)
$`(\stackrel{}{B}_2\stackrel{}{B}_1)\widehat{n}_{1,2}=0,`$ (7)
$`\widehat{n}_{1,2}\times (\stackrel{}{H}_2\stackrel{}{H}_1)=\stackrel{}{i},`$ (8)
donde
$$|\stackrel{}{i}|=\left|\frac{dI}{dS}\right|.$$
### Electrostática
Estudiamos un problema de electrostática si se satisfacen las condiciones
* No hay dependencia temporal en los campos.
* No existen cargas en movimiento.
Con esto, las ecuaciones de Maxwell (1 \- 4) se reducen a
$`\times \stackrel{}{E}=0,`$ (9)
$`\stackrel{}{D}=\rho .`$ (10)
En vista de (9), del cálculo vectorial sabemos que
$$\stackrel{}{E}=\mathrm{\Phi }.$$
De esta forma se introduce el potencial electrostático ($`\mathrm{\Phi }`$). Considerando medios isotrópicos (i. e., $`\stackrel{}{D}=ϵ\stackrel{}{E}`$) la ecuación (10) se reduce a
$$^2\mathrm{\Phi }=\frac{\rho }{ϵ}$$
que se conoce como ecuación de Poisson (en ausencia de cargas se obtiene la ecuación de Laplace).
Por otra parte, las condiciones de frontera (Red triangular \- 8) se reducen a
$`\mathrm{\Phi }_1`$ $`=`$ $`\mathrm{\Phi }_2,`$
$`ϵ_1(\mathrm{\Phi }\widehat{n})_1`$ $``$ $`ϵ_2(\mathrm{\Phi }\widehat{n})_2=\sigma .`$
En el caso de un conductor, dado que en su interior el campo eléctrico es nulo, se tiene
$$\mathrm{\Phi }_{\text{conductor}}=\text{const},$$
y así, la densidad superficial de carga en el mismo es
$$\sigma =ϵ(\mathrm{\Phi }\widehat{n})_{\text{afuera}}.$$
### Magnetostática
Las condiciones para hablar de magnetostática son:
* No hay dependencia temporal en los campos.
* Hasta hoy no se han detectado los monopolos magnéticos.
Bajo estas consideraciones las ecuaciones de Maxwell (1 \- 4) se simplifican a
$`\times \stackrel{}{H}=\stackrel{}{j},`$ (11)
$`\stackrel{}{B}=0,`$ (12)
y las condiciones de frontera (Red triangular \- 8)
$`(\stackrel{}{D}_1\stackrel{}{D}_2)\widehat{n}_{1,2}=0,`$
$`\widehat{n}_{1,2}\times (\stackrel{}{H}_2\stackrel{}{H}_1)=\stackrel{}{i}.`$
Al igual que en el caso electrostático, a partir del cálculo vectorial y (12) se introduce el potencial vectorial magnético como
$$\stackrel{}{B}=\times \stackrel{}{A}$$
el cual, para materiales homogéneos e isotrópicos ($`\stackrel{}{B}=\mu \stackrel{}{H}`$), se obtiene de (11) como
$$^2\stackrel{}{A}=\mu \stackrel{}{j}$$
(13)
(NOTA: cabe aclarar que, dada su definición, da lo mismo tomar $`\stackrel{}{A}`$ que $`\stackrel{}{A}+\phi `$; por ello se elige el potencial vectorial tal que $`(\stackrel{}{A})=0`$, obteniendo así (13) a partir de (11)). Si se conoce $`\stackrel{}{j}`$, la solución a (13) es
$$\stackrel{}{A}(\stackrel{}{r})=\frac{\mu }{4\pi }_V\frac{\stackrel{}{j}(\stackrel{}{r}^{})}{|\stackrel{}{r}\stackrel{}{r}^{}|}𝑑V^{}$$
y para $`rr_{\text{sistema}}`$
$$\stackrel{}{A}(\stackrel{}{r})=\frac{\mu }{4\pi }\frac{\stackrel{}{m}\times \stackrel{}{r}}{r^3}$$
donde $`\stackrel{}{m}`$ es el momento magnético del sistema, dado como
$$\stackrel{}{m}=\frac{1}{2}_V\stackrel{}{r}\times \stackrel{}{j}(r)𝑑V.$$
Por último, la energía de un campo magnético estático es
$`W_{\text{mag}}`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \stackrel{}{B}\stackrel{}{H}𝑑V}`$
$`=`$ $`{\displaystyle \frac{\mu }{8\pi }}{\displaystyle \frac{\stackrel{}{j}(\stackrel{}{r})\stackrel{}{j}(\stackrel{}{r}^{})}{|\stackrel{}{r}\stackrel{}{r}^{}|}𝑑V𝑑V^{}}.`$
Para un sistema de conductores
$$W_{\text{mag}}=\frac{1}{2}\underset{i,k}{}L_{ik}I_iI_k,$$
donde se define el coeficiente de inducción magnética entre las corrientes $`\stackrel{}{j}_i`$ y $`\stackrel{}{j}_k`$ como
$$L_{ik}=\frac{\mu }{4\pi I_iI_k}\frac{\stackrel{}{j}_k(\stackrel{}{r}_k)\stackrel{}{j}_i(\stackrel{}{r}_i)}{|\stackrel{}{r}_k\stackrel{}{r}_i|}𝑑V_k𝑑V_i.$$
## Clase 2
En el caso en el cual los campos varían lentamente en el tiempo, o sea son funciones $`f(at)`$ donde se satisfacen las condiciones
$$a1\omega \frac{\sigma _c}{ϵ}l\lambda $$
con $`\sigma _c`$ conductividad, $`\omega ,\lambda `$ carácterísticas de las oscilaciones electromagnéticas, $`l`$ dimensiones lineales del sistema.
Las ecuaciones de Maxwell toman la forma
$$\times 𝐇=𝐉\times 𝐄=\frac{𝐁}{t}$$
$$𝐁=0𝐃=\rho .$$
Obsérvese que se ha despreciado el término $`\frac{𝐃}{t}`$.
En el caso de campos variables arbitrarios para situaciones en las cuales no existen corrientes ni cargas presentes las ecuaciones de Maxwell toman la siguiente forma
$$\times 𝐇=\frac{𝐃}{t}\times 𝐄=\frac{𝐁}{t}$$
$$𝐁=0𝐃=0.$$
Se tienen soluciones tipo ondas planas
$$𝐄=𝐄_oe^{i(𝐤𝐫\omega t)}$$
$$𝐇=𝐇_oe^{i(𝐤𝐫\omega t)}.$$
Las notaciones usadas son las usuales, $`\omega `$ es la frecuencia, $`𝐤=\frac{\omega }{c}`$ es el vector de onda, la dirección del cual, en medios isotrópicos, coincide con la dirección de la energía. El vector que justamente nos da el flujo de energía es el llamado vector de Poynting (S), definido por
$$𝐒=𝐄\times 𝐇.$$
(1)
Para campos variables la conexión entre los campos y los potenciales es de la forma
$$𝐄=\varphi \frac{𝐀}{t}$$
$$𝐁=\times 𝐀.$$
(2)
En general, los potenciales no son observables directamente (sino por sus efectos, E,B). Entre ellos existe una condición muy importante (de consistencia de la teoría electromagnética) que se llama condición de “gauge” que puede ser diferente en función de la condición considerada.
Una de las condiciones de “gauge” más frecuentes es la de Lorentz
$$𝐀+ϵ\mu \frac{\varphi }{t}=0.$$
(3)
Esta condición “gauge” es muy usada porque permite una simple generalización de las ecuaciones laplacianas del caso estático
$$\mathrm{}\varphi =\frac{\rho }{ϵ}$$
$$\mathrm{}𝐀=\mu 𝐉$$
(4)
estas ecuaciones son llamadas D’Alembertianas y $`\mathrm{}^2\frac{^2}{t^2}`$. Los potenciales que son solución de estas ecuaciones son llamados potenciales retardados (y no precisamente porque sean muy tontos) los cuales tienen la forma
$$\varphi (𝐫,t)=\frac{1}{4\pi ϵ}\frac{\rho (𝐫^{},t\frac{𝐫𝐫^{}}{v})}{𝐫𝐫^{}}𝑑V^{}$$
(5)
$$𝐀(𝐫,t)=\frac{\mu }{4\pi }\frac{𝐉(𝐫^{},t\frac{𝐫𝐫^{}}{v})}{𝐫𝐫^{}}𝑑V^{}.$$
(6)
A grandes distancias del sistema de cargas ($`r>>\lambda `$) y en el vacío, B,E y A se pueden escribir como sigue
$$𝐁=\frac{1}{c}\dot{𝐀}\times 𝐧$$
(7)
$$𝐄=c𝐁\times 𝐧=(\dot{𝐀}\times 𝐧)\times 𝐧$$
(8)
$$𝐀=\frac{\mu }{4\pi r}𝐉(𝐫^{},t\frac{𝐫𝐫^{}}{v})𝑑V^{}$$
(9)
donde n$`=\frac{𝐫}{r}`$ es el versor en la dirección de la radiación. Si además $`\lambda >>l`$, con $`l`$ la dimensión del sistema radiante, se puede usar la llamada aproximación multipolar, es decir, la radiación se puede representar como una sumatoria de los campos emitidos por los dipolos, cuadrupolos, etc., que forman el sistema. Para el caso dipolar se tiene
$$𝐁=\mu _0\frac{\ddot{𝐩}\times 𝐧}{4\pi cr}$$
(10)
$$𝐄=\frac{\mu _0}{4\pi r}(\ddot{𝐩}\times 𝐧\times 𝐧)$$
(11)
donde p es el momento dipolar del sistema. La intensidad de la radiación de un dipolo es
$$I=\frac{\ddot{𝐩}^2}{6\pi ϵ_0c^3}.$$
(12)
### Magnetohidrodinámica
La magnetohidrodinámica estudia el comportamiento de los líquidos o los gases conductores (plasmas) en campos electromagnéticos. Se usan los conceptos hidrodinámicos: densidad, velocidad, presión, viscosidad. Las ecuaciones básicas son:
$$\frac{\rho _m}{t}+(\rho _m𝐯)=0$$
$$\rho _m\frac{𝐯}{t}+\rho _m(𝐯)𝐯=P+𝐣\times 𝐁+\eta ^2𝐯+\rho _m𝐠$$
$$\times 𝐄=\frac{𝐁}{t}\times 𝐇=𝐣𝐣=\sigma _e(𝐄+𝐯\times 𝐁)$$
mas la ecuación de estado del fluido.
### Relatividad Especial
La teoría de la relatividad especial surgió en la electrodinámica y se basa en dos postulados fundamentales
* La velocidad de la luz en el vacío $`c=2.99793\times 10^8m/s`$ es una constante en todos los sistemas de referencia inerciales.
* Las leyes de la Física tienen la misma forma en todos los sistemas inerciales (covariancia de las leyes naturales).
Las transformaciones de Lorentz en una dimensión se escriben así
$$x_1^{}=\frac{x_1+i\beta x_4}{\sqrt{1\beta ^2}}$$
$$x_2^{}=x_2x_3^{}=x_3$$
$$x_4^{}=\frac{x_4i\beta x_1}{\sqrt{1\beta ^2}}$$
donde $`x_4=ict`$, $`x_4^{}=ict^{}`$, $`\beta =\frac{v}{c}`$. Las velocidades $`u^{}`$ de un cuerpo en K’ con respecto a las velocidades $`u`$ del mismo cuerpo en K están relacionadas mediante las siguientes expresiones
$$u_x^{}=\frac{u_xv}{1\frac{vu_x}{c^2}},u_y=\frac{u_y\sqrt{1\beta ^2}}{1\frac{vu_x}{c^2}},u_z^{}=\frac{u_z\sqrt{1\beta ^2}}{1\frac{vu_x}{c^2}}.$$
La segunda ley para partículas relativistas se escribe
$$𝐅=\frac{d𝐩}{dt}=\frac{d}{dt}\left(\frac{m𝐯}{\sqrt{1\beta ^2}}\right).$$
Cantidades del tipo $`(𝐩,i/cE)`$ son llamadas cuadrivectores, el anterior se llama cuadrivector de impulso-energía. Otros ejemplos de cuadrivectores son $`(𝐤,i/c\omega )`$; $`(𝐣,ic\rho )`$; $`(𝐀,i/c\phi )`$. Existen también objetos llamados cuadritensores por extensión de lo anterior, algunos ejemplos de ellos son
$$F_{\alpha \beta }=\left(\begin{array}{cccc}\hfill 0& cB_z& cB_y\hfill & iE_x\hfill \\ \hfill cB_z& 0& cB_x\hfill & iE_y\hfill \\ \hfill cB_y& cB_x& 0\hfill & iE_z\hfill \\ \hfill iE_x& iE_y& iE_z\hfill & 0\hfill \end{array}\right)$$
$$T_{\alpha \beta }=\mathrm{\Sigma }_0(F_{\alpha \mu }F_{\beta \mu }\frac{1}{4}\delta _{\alpha \beta }F_{\mu \eta }F^{\mu \eta }).$$
## Clase 3
### Fuerza de Lorentz como fuerza lagrangiana
Las ecuaciones del movimiento de Euler-Lagrange son
$$Q_k=\frac{L}{q_k}+\frac{d}{dt}\left(\frac{L}{\dot{q_k}}\right)$$
(1)
donde las $`Q_k`$ son las fuerzas externas o fuerzas generalizadas y $`L=TU`$ .
Por otra parte, las ecuaciones de Maxwell en unidades de Gauss son
$`\left(M1\right)\times \stackrel{}{E}+{\displaystyle \frac{1}{c}}{\displaystyle \frac{\stackrel{}{B}}{t}}`$ $`=`$ $`0\left(M3\right)\stackrel{}{D}=4\pi \rho `$
$`\left(M2\right)\times \stackrel{}{H}{\displaystyle \frac{1}{c}}{\displaystyle \frac{\stackrel{}{D}}{t}}`$ $`=`$ $`{\displaystyle \frac{4\pi }{c}}\stackrel{}{j}\left(M4\right)\stackrel{}{B}=0.`$
Ahora con $`\stackrel{}{F}=q\stackrel{}{E}=q\phi `$ sólo en electrostatica en general la fuerza es la ley de Lorentz o sea
$$\stackrel{}{F}_L=q\left(\stackrel{}{E}+\frac{1}{c}\stackrel{}{v}\times \stackrel{}{B}\right).$$
(2)
Ahora de la $`\left(M4\right)`$ encontramos que $`\stackrel{}{B}=\times \stackrel{}{A}`$ y sustituyendo en $`\left(M1\right)`$ encontramos
$$\times \stackrel{}{E}+\frac{1}{c}\frac{}{t}\left(\times \stackrel{}{A}\right)=0$$
(3)
por tanto
$$\times \left(\stackrel{}{E}+\frac{1}{c}\frac{}{t}\stackrel{}{A}\right)=0$$
(4)
de aqui que podemos definir una función escalar tal que
$$\mathrm{\Phi }=\stackrel{}{E}+\frac{1}{c}\frac{}{t}\stackrel{}{A}$$
(5)
entonces
$$\stackrel{}{F}_L=q\left(\mathrm{\Phi }\frac{1}{c}\frac{}{t}\stackrel{}{A}+\frac{1}{c}\stackrel{}{v}\times \left(\times \stackrel{}{A}\right)\right)$$
(6)
donde el doble producto vectorial lo podemos expresar de la siguiente forma
$$\stackrel{}{v}\times \left(\times \stackrel{}{A}\right)=\left(\stackrel{}{v}\stackrel{}{A}\right)\frac{d\stackrel{}{A}}{dt}+\frac{\stackrel{}{A}}{t}$$
(7)
por tanto
$`\stackrel{}{F}_L`$ $`=`$ $`q\left(\mathrm{\Phi }{\displaystyle \frac{1}{c}}{\displaystyle \frac{}{t}}\stackrel{}{A}+{\displaystyle \frac{1}{c}}\left(\left(\stackrel{}{v}\stackrel{}{A}\right){\displaystyle \frac{d\stackrel{}{A}}{dt}}+{\displaystyle \frac{\stackrel{}{A}}{t}}\right)\right)`$
$`=`$ $`q\left(\left[\mathrm{\Phi }{\displaystyle \frac{1}{c}}\stackrel{}{v}\stackrel{}{A}\right]{\displaystyle \frac{1}{c}}{\displaystyle \frac{d}{dt}}\left[_\stackrel{}{v}\left(\stackrel{}{v}\stackrel{}{A}\right)\right]\right)`$
lo que hace que $`\stackrel{}{F}_L`$ se pueda escribir como fuerza lagrangiana
$$\stackrel{}{F}_L=U+\frac{d}{dt}\frac{U}{\stackrel{}{v}}$$
(8)
con $`U=q\mathrm{\Phi }\frac{q}{c}\stackrel{}{v}\stackrel{}{A}`$ .
### Electrodinámica no líneal
Para la electrodinamica no líneal la constante dielectrica se expresa como
$$\epsilon _v=\frac{\epsilon _o}{\left(1+\frac{1}{b^2}\left(c^2B^2E^2\right)^{\frac{1}{2}}\right)}$$
(9)
y la permeabilidad se escribe como
$$\mu _v=\mu _o\left(1+\frac{1}{b^2}\left(c^2B^2E^2\right)^{\frac{1}{2}}\right)$$
(10)
donde $`b`$ en ambos casos es un parametro que fija una intensidad máxima de los campos.
Al menos para campos que varían lentamente, en función de los tensores de permeabilidades eléctricas y magnética del vacío tenemos
$$D_i=\underset{k}{}\epsilon _{ik}E_k\mathrm{y}B_i=\underset{k}{}\mu _{ik}H_k$$
(11)
donde
$`\epsilon _{ik}`$ $`=`$ $`\epsilon _o\left[\delta _{ik}+{\displaystyle \frac{e^4\mathrm{}}{45\pi m^4c^7}}2\left(E^2c^2B^2\right)\delta _{ik}+7c^2B_iB_k\right]+\mathrm{}.`$ (12)
$`\mu _{ik}`$ $`=`$ $`\mu _o\left[\delta _{ik}+{\displaystyle \frac{e^4\mathrm{}}{45\pi m^4c^7}}2\left(B^2{\displaystyle \frac{E^2}{c^2}}\right)\delta _{ik}+7E_iE_k/c^2\right]+\mathrm{}..`$ (13)
para el límite clásico hacemos $`\mathrm{}0`$ y estos efectos no lineales desaparecen al comparar con la expresión clásica en (9) y (10) encontramos
$$b_q=\frac{\sqrt{45\pi }}{2}\sqrt{\frac{e^2}{4\pi \epsilon _o\mathrm{}c}}\frac{e}{4\pi \epsilon _or_o^2}0.51\frac{e}{4\pi \epsilon _or_o^2}=0.51\frac{e_G}{r_o^2}$$
(14)
por tanto
$$r_o=\frac{e_G^2}{mc^2}2.8\times 10^{15}\mathrm{metros}$$
(15)
este es el radio clásico del electrón.
Ahora si tenemos varias cargas
$$\stackrel{}{E}\stackrel{}{n}da=\frac{1}{\epsilon _o}\underset{i}{}q_i$$
(16)
y si tenemos distribuciones de carga
$$_S\stackrel{}{E}\stackrel{}{n}da=\frac{1}{\epsilon _o}_V\rho \left(\stackrel{}{x}\right)dV$$
(17)
donde $`V`$ es el vólumen enserrado por la superficie, ahora el teorema de la divergencia nos dice que
$$_S\stackrel{}{v}\stackrel{}{n}da=_V\stackrel{}{v}dV$$
(18)
entonces aplicando este teorema en la ley de Gauss encontramos
$$\stackrel{}{E}=\frac{\rho }{\epsilon _o}$$
(19)
y esta es la forma diferencial de la ley de Gauss.
## Clase 5
### Energía potencial electrostática y densidad de energía; capacitancia
Imaginemos el caso en que una carga $`q_i`$ es traída desde al infinito hasta el punto $`\stackrel{}{x}_i`$, localizado en una región del espacio donde se conoce el potencial electrostático $`\mathrm{\Phi }(\stackrel{}{x})`$. El trabajo realizado sobre esta carga es
$$W_i=q_i\mathrm{\Phi }(\stackrel{}{x}_i).$$
Ahora bien, si este potencial es provocado por la presencia de otras $`n1`$ cargas, se tiene
$$\mathrm{\Phi }(\stackrel{}{x}_i)=\frac{1}{4\pi ϵ_0}\underset{j=1}{\overset{n1}{}}\frac{q_j}{|\stackrel{}{x}_i\stackrel{}{x}_j|}$$
y por tanto
$$W_i=\frac{q_i}{4\pi ϵ_0}\underset{j=1}{\overset{n1}{}}\frac{q_j}{|\stackrel{}{x}_i\stackrel{}{x}_j|}.$$
(1)
Por un proceso mental similar, se puede ver que el trabajo total necesario para obtener el arreglo de $`n`$ cargas, trayendo cada una desde infinito a una región del espacio originalmente vacía, es
$$W_{\text{total}}=\frac{1}{8\pi ϵ_0}\underset{i}{}\underset{j}{}\frac{q_iq_j}{|\stackrel{}{x}_i\stackrel{}{x}_j|}$$
(2)
donde $`i,j`$ toman todos los valores entre $`1`$ y $`n`$, excepto $`i=j`$ (autoenergías).
En el caso de una distribución continua de cargas es claro que
$$W_{\text{total}}=\frac{1}{8\pi ϵ_0}\frac{\rho (\stackrel{}{x})\rho (\stackrel{}{x}^{})}{|\stackrel{}{x}\stackrel{}{x}^{}|}d^3xd^3x^{},$$
(3)
expresión que puede reescribirse de varias formas:
* En términos del potencial
$$W_{\text{total}}=\frac{1}{2}\rho (\stackrel{}{x})\mathrm{\Phi }(\stackrel{}{x})d^3x.$$
(4)
* Utilizando la ecuación de Poisson:
$$W_{\text{total}}=\frac{ϵ_0}{2}\mathrm{\Phi }^2\mathrm{\Phi }d^3x.$$
(5)
Integrando por partes la última expresión se obtiene
$`W_{\text{total}}`$ $`=`$ $`{\displaystyle \frac{ϵ_0}{2}}{\displaystyle |\mathrm{\Phi }|^2d^3x}`$ (6)
$`=`$ $`{\displaystyle \frac{ϵ_0}{2}}{\displaystyle |\stackrel{}{E}|^2d^3x}.`$
Por la forma de la última integral, se define la densidad volumétrica de energía como
$$w=\frac{ϵ_0}{2}|\stackrel{}{E}|^2.$$
Notemos que esta densidad de energía es no negativa, y por tanto el trabajo total tampoco será negativo. Sin embargo, de (1) se ve que el trabajo para hacer un arreglo con dos cargas de signo contrario es negativo; esta contradicción surge porque en las expresiones (3 \- 5) se incluyen las autoenergías en el trabajo total, mientras que en el caso discreto (2) se las excluye.
Por último, como siempre, se puede calcular la fuerza a partir de los cambios que sufre la energía ante desplazamientos virtuales pequeños.
Consideremos un sistema de $`n`$ conductores, el $`i`$ésimo de ellos con carga $`Q_i`$ y potencial $`V_i`$. Dada la relación lineal que existe entre el potencial y la carga, podemos escribir
$$V_i=\underset{j=1}{\overset{n}{}}p_{ij}Q_j,$$
donde $`p_{ij}`$ depende sólo del arreglo geométrico de los conductores. Invirtiendo las ecuaciones anteriores se obtiene
$$Q_j=\underset{i=1}{\overset{n}{}}C_{ji}V_i.$$
Los coeficientes $`C_{ii}`$ son las capacitancias, y $`C_{ij}`$ ($`ij`$) los coeficientes de inducción.
De esta forma
$`W_{\text{total}}`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \underset{i=1}{\overset{n}{}}}Q_iV_i`$
$`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \underset{i,j=1}{\overset{n}{}}}C_{ij}V_iV_j.`$
### Aproximación variacional a la solución de las ecuaciones de Poisson y Laplace
El uso de métodos variacionales es muy popular en Física. La electrodinámica no es la excepción. En efecto, la idea de considerar funcionales cuyos extremales satisfagan ecuaciones de movimiento tipo Poisson o Laplace es muy sugestiva (sobre todo por la elegancia del método variacional).
Consideremos la funcional
$$I[\psi ]=\frac{1}{2}_V\psi \psi d^3x_Vg\psi d^3x,$$
(7)
sujeta a la condición tipo Dirichlet $`\delta \psi (S)=0`$ ($`S`$ es la superficie cerrada que contiene a $`V`$). Es fácil ver que $`\delta I=I[\psi +\delta \psi ]I[\psi ]=0`$ conduce a la ecuación de movimiento
$$^2\psi =g.$$
Se ve que este problema no es otro que resolver la ecuación de Poisson con condiciones de frontera tipo Dirichlet.
Similarmente, para condiciones de frontera tipo Neumann, se plantea el funcional
$$I[\psi ]=\frac{1}{2}_V\psi \psi d^3x_Vg\psi d^3x_Sf\psi d^3x,$$
(8)
con
$$\left(\frac{\psi }{n}\right)_S=f(S).$$
Es fácil probar que $`\delta I[\psi ]=0`$ conduce a las ecuaciones
$`^2\psi =g,`$
$`\left({\displaystyle \frac{\psi }{n}}\right)_S=f(S).`$
Resulta lógico preguntar si este método variacional de obtener la ecuación de Poisson sirve para algo, o es sólo un juego matemático. Para contestar, notemos que una vez conocida la forma de los funcionales (7, 8) aún es necesario encontrar $`\psi `$ (o sea resolver la ecuación de Poisson); por tanto el problema es el mismo. Sin embargo, se pueden proponer soluciones $`\psi =A\mathrm{\Psi }(\stackrel{}{x},\alpha ,\beta ,\mathrm{})`$ que satisfagan las condiciones de frontera dadas, para después variacionalmente encontrar las constantes indeterminadas (notar que con esta elección $`I=I[A,\alpha ,\beta ,\mathrm{}]`$). En este sentido el método variacional sirve para encontrar soluciones aproximadas.
## Clase 6
## Método de las imágenes
Este método se refiere a problemas de cargas puntuales en la presencia de superficies a potencial cero o constante. Las condiciones de frontera se simulan con cargas puntuales de valores y posiciones bien determinadas conocidas como “cargas imágenes”.
### Carga puntual con esfera a $`\varphi =0`$
El potencial asociado a la carga real y la carga imagen es
$$\varphi (𝐱)=\frac{1}{4\pi ϵ_0}[\frac{q}{𝐱𝐲}+\frac{q^{}}{𝐱𝐲^{}}].$$
La condición de frontera es que el potencial se anule en $`𝐱=a`$. Introducimos dos vectores unitarios n,n’, uno en la dirección de x y el otro en la dirección de y, de manera que el potencial se puede expresar
$$\varphi (𝐱)=\frac{1}{4\pi ϵ_0}[\frac{q}{x𝐧y𝐧^{}}+\frac{q^{}}{x𝐧y𝐧^{}}].$$
Factorizando $`x`$ del primer término, $`y^{}`$ del segundo y valuando en $`x=a`$
$$\varphi (x=a)=\frac{1}{4\pi ϵ_0}[\frac{q}{a𝐧\frac{y}{a}𝐧^{}}+\frac{q^{}}{y^{}𝐧^{}\frac{a}{y^{}}𝐧}].$$
Se observa que para que el potencial se anule en la frontera de la esfera se debe satisfacer
$$\frac{q}{a}=\frac{q^{}}{y^{}},\frac{y}{a}=\frac{a}{y^{}}$$
resolviendo estas ecuaciones se encuentra
$$q^{}=\frac{y^{}}{a}q=\frac{a}{y}q,y^{}=\frac{a^2}{y}$$
$`q^{}`$ es la carga total de inducción sobre la superficie de la esfera, podemos observar además lo siguiente
$$yaq^{}q$$
$$y\mathrm{}q^{}0$$
La densidad superficial de carga está dada por
$$\sigma =ϵ_0\frac{\varphi }{x}_{x=a}=\frac{q}{4\pi a^2}\frac{a}{y}\frac{1\frac{a^2}{y^2}}{(1+\frac{a^2}{y^2}2\frac{a}{y}\mathrm{cos}\gamma )^{3/2}}.$$
Es posible calcular también la fuerza de atracción hacia la esfera, la magnitud de la cual está dada por
$$𝐅=\frac{1}{4\pi ϵ_0}\frac{q^2}{a^2}\frac{a^3}{y^3}\left(1\frac{a^2}{y^2}\right)^2.$$
### Carga q en presencia de una esfera conductora cargada a Q, aislada
El potencial para esta configuración se puede expresar así
$$\varphi (𝐱)=\frac{1}{4\pi ϵ_0}\left[\frac{q}{𝐱𝐲}\frac{aq}{y𝐱\frac{a^2}{y^2}𝐲}+\frac{Q+\frac{a}{y}q}{𝐱}\right].$$
La fuerza de atracción en este caso es
$$𝐅(𝐲)=\frac{1}{4\pi ϵ_0}\frac{q𝐲}{y^3}\left[Q\frac{qa^3(2y^2a^2)}{y(y^2a^2)^2}\right].$$
### Carga q cerca de una esfera conductora a potencial constante
Para la situación presente el potencial adopta la forma
$$\varphi (𝐱)=\frac{1}{4\pi ϵ_0}\left[\frac{q}{𝐱𝐲}\frac{aq}{y𝐱\frac{a^2}{y^2}𝐲}\right]+\frac{Va}{𝐱}.$$
La fuerza de atracción está dada por
$$𝐅(𝐲)=\frac{q𝐲}{y^3}\left[Va\frac{1}{4\pi ϵ_0}\frac{qay^3}{(y^2a^2)^2}\right].$$
### Esfera conductora en un campo eléctrico uniforme
Un campo eléctrico uniforme es producido por ejemplo por dos cargas puntuales $`\pm `$Q localizadas en $`z=\pm R`$ para $`R\mathrm{}`$. Si ahora una esfera conductora es colocada en el origen, el potencial será el debido a las cargas $`\pm `$Q en $``$R y sus imágenes $`\frac{Qa}{R}`$ en $`z=\frac{a^2}{R}`$
$$\varphi =\frac{1}{4\pi ϵ_0}\left[\frac{Q}{(r^2+R^2+2rR\mathrm{cos}\theta )^{1/2}}\frac{Q}{(r^2+R^22rR\mathrm{cos}\theta )^{1/2}}\right]+$$
$$\frac{1}{4\pi ϵ_0}\left[\frac{aQ}{R\left(r^2+\frac{a^4}{R^2}+\frac{2a^2r}{R}\mathrm{cos}\theta \right)^{1/2}}+\frac{aQ}{R\left(r^2+\frac{a^4}{R^2}\frac{2a^2r}{R}\mathrm{cos}\theta \right)^{1/2}}\right].$$
Como $`R>>r`$ podemos desarrollar los denominadores
$$\varphi =\frac{1}{4\pi ϵ_0}\left[\frac{2Q}{R^2}r\mathrm{cos}\theta +\frac{2Q}{R^2}\frac{a^3}{r^2}\mathrm{cos}\theta \right]+\mathrm{}$$
Para $`R\mathrm{}`$, $`\frac{2Q}{4\pi ϵ_0R^2}`$ es el campo aplicado de manera que el potencial en este límite toma la forma
$$\varphi _R\mathrm{}=E_0\left(r\frac{a^3}{r^2}\right)\mathrm{cos}\theta =E_0z+\frac{a^3}{r^3}E_0z,$$
donde el último término es el del ”dipolo imagen”. La densidad superficial de carga está dada por
$$\sigma =ϵ_0\frac{\varphi }{r}_{r=a}=3ϵ_0E_0\mathrm{cos}\theta ,$$
la cual se anula al integrarla sobre la superficie
$$\sigma 𝑑a=0.$$
### Función de Green para la esfera conductora
Para problemas de Dirichlet con conductores $`G(𝐱,𝐱^{})/4\pi ϵ_0`$ puede ser interpretada como el potencial debido a la distribución superficial de carga inducida sobre la superficie por la presencia de una carga puntual (fuente) en el punto x’. Por definición la función de Green $`G(𝐱,𝐱^{})`$ satisface la ecuación
$$^2G(𝐱,𝐱^{})=4\pi \delta (𝐱𝐱^{}).$$
Para el caso de la esfera la función de Green está dada por
$$G_{esf}(𝐱,𝐱^{})=\frac{1}{𝐱𝐱^{}}\frac{a}{x^{}𝐱\frac{a^2}{x^2}𝐱^{}}.$$
En coordenadas esféricas lo anterior es
$$G_{esf}(𝐱,𝐱^{})=\frac{1}{\left(x^2+x^22xx^{}\mathrm{cos}\gamma \right)^{1/2}}\frac{1}{\left(\frac{x^2x^2}{a^2}+a^22xx^{}\mathrm{cos}\gamma \right)^{1/2}},$$
$$\frac{G}{n^{}}_{x^{}=a}=\frac{x^2a^2}{a\left(x^2+a^22ax\mathrm{cos}\gamma \right)^{3/2}}\sigma .$$
Recordando la solución de la ecuación de Poisson con condiciones de Dirichlet para el potencial
$$\varphi (𝐱)=\frac{1}{4\pi ϵ_0}_V\rho (𝐱^{})G_D(𝐱,𝐱^{})d^3x\frac{1}{4\pi }_S\varphi (𝐱^{})\frac{G_D}{n^{}}𝑑a^{}$$
usando esto, podemos escribir la solución general para el potencial de la esfera conductora para la cual conocemos el potencial en la frontera
$$\varphi _{esf}(𝐱)=\frac{1}{4\pi }\varphi (a,\theta ^{},\phi ^{})\frac{a(x^2a^2)}{\left(x^2+a^22ax\mathrm{cos}\gamma \right)^{3/2}}𝑑\mathrm{\Omega }^{},$$
donde $`\mathrm{cos}\gamma =\mathrm{cos}\theta \mathrm{cos}\theta ^{}+\mathrm{sin}\theta \mathrm{sin}\theta ^{}\mathrm{cos}(\phi \phi ^{})`$. Para el interior de la esfera $`x^2a^2a^2x^2`$, y en el caso en el que se tienen distribuciones volumétricas de carga se tiene que tomar en cuenta la contribución de la integral de volumen.
## Clase 9
### Análisis de elemento finito para resolver la ecuación de Poisson
A continuación presentamos una breve introducción al análisis de elemento finito para resolver la ecuación de Poisson. Por simplicidad en la presentación sólo consideramos problemas bidimensionales.
Primeramente esbozamos el método de Galerkin para replantear la ecua-ción de Poisson, y dividir la región de estudio en una red cuyo número de celdas es finito. Por último presentamos dos tipos particulares de redes: cuadriculada regular y triangular.
### El método de Galerkin
Sea una región bidimensional $`R`$ limitada por una curva cerrada $`C`$; consideremos en $`R`$ la ecuación de Poisson
$$^2\psi =g$$
(1)
con condiciones de frontera tipo Dirichlet; multiplicamos (1) por una función de prueba $`\varphi (x,y)`$ que sea continua a trozos en $`R`$ y tal que $`\varphi (C)=0`$; después integramos sobre $`R`$, obteniendo
$$_R[\varphi ^2\psi +g\varphi ]𝑑x𝑑y=0.$$
A continución, utilizando la primera identidad de Green (bidimensional), la integral anterior se reescribe como
$$_R[\varphi \psi g\varphi ]𝑑x𝑑y=0.$$
(2)
El siguiente paso es dividir la región $`R`$ por medio de una red con $`N`$ celdas, y definir un conjunto de funciones $`\{\varphi _i(x,y),i=1,2,\mathrm{},N\}`$ tal que cada una de ellas es no nula sólo en una celda particular de la red. A continuación se expresa $`\psi `$ como
$$\psi (x,y)\underset{i=1}{\overset{N}{}}\mathrm{\Psi }_i\varphi _i(x,y);$$
sustituyendo lo anterior en (2) y escogiendo $`\varphi =\varphi _j`$ se obtiene
$$\underset{i=1}{\overset{N}{}}\mathrm{\Psi }_i\varphi _i(x,y)\varphi _j(x,y)=g_0_R\varphi _i(x,y)𝑑x𝑑y,$$
donde se ha supuesto que las celdas son suficientemente pequeñas como para que $`g(x,y)g_0`$ dentro de ellas (el valor de $`g_0`$ varía de celda a celda). Con esto, (2) se reduce a la ecuación matricial
$$𝐊\mathrm{\Psi }=G$$
(3)
aquí $`𝐊`$ es una matriz $`N\times N`$ con elementos
$$k_{ij}_R\varphi _i_jdxdy$$
$`\mathrm{\Psi }`$ es la matriz columna formada con los coeficientes $`\mathrm{\Psi }_i`$; $`G`$ es una matriz columna con elementos
$$G_ig_i_R\varphi _i(x,y)𝑑x𝑑y.$$
El poder del método de Galerkin radica en que, por la forma como se escojen las $`\varphi _i`$, la matriz $`𝐊`$ es dispersa, i.e., sólo pocos de sus elementos son diferentes de cero, y por ello es relativamente fácil conocer $`\mathrm{\Psi }`$ a partir de (3), lo cual nos da la solución a la ecuación tipo Poisson (1).
### Casos particulares
### Red cuadriculada regular
Se escoge una red de cuadros, cada uno de lado $`h`$; sean $`(x_i,y_j)`$ las coordenadas de los vértices. Se toman las funciones $`\varphi _{ij}(x,y)`$ tales que $`\varphi _{ij}0`$ sólo en una vecindad de área $`h^2`$ alrededor de $`(x_i,y_j)`$, y las $`\varphi _{ij}`$’s son linealmente independientes entre sí. Con esto, de acuerdo al método de Galerkin
$$\psi \underset{k,l=1}{\overset{(N_0)}{}}\mathrm{\Psi }_{kl}\varphi _{kl}(x,y)$$
donde se supone que el total de celdas es $`N_0`$; los coeficientes $`\mathrm{\Psi }_{kl}`$ se obtienen a partir de (3) con
$`𝐊=\left({\displaystyle _R}\varphi _{ij}\varphi _{kl}dxdy\right),`$
$`(G)=\left(g(x_i,y_j){\displaystyle _R}\varphi _{ij}𝑑x𝑑y\right)`$
$`(\mathrm{\Psi })=(\mathrm{\Psi }_i).`$
La inconveniencia del uso de redes como ésta es que se presentan casos donde el potencial varía de formas diferentes en diferentes regiones, y por ello sería más conveniente utilizar celdas irregulares. A continuación se presenta una de ellas.
### Red triangular
Las redes triangulares son las más utilizadas en el análisis de elemento finito, por las razones expuestas al final de la sección anterior. Para este tipo de redes se asume que el elemento triangular ($`e`$) es lo suficientemente pequeño como para que $`\psi `$ cambie poco en su interior y de hecho pueda ser aproximado de forma lineal en cada dirección:
$$\psi (x,y)\psi _e(x,y)=A+Bx+Cy.$$
Sean $`(x_i,y_i)`$ ($`i=1,2,3`$) las coordenadas de cada vértice del triángulo. Entonces las constantes $`(A,B,C)`$ quedan determinadas por los valores de $`\psi `$ en cada uno de ellos.
Con el fin de sistematizar el procedimiento, es conveniente definir las funciones de forma $`N_j(x,y)`$ (una por cada vértice), tales que $`N_j^{(e)}(x_j,y_j)=1`$, $`N_j^{(ee_j)}(x,y)=0`$ y $`N_j^{(e)}(x,y)=0`$ si $`xx_j`$, $`yy_j`$. Por la linealidad de $`\psi `$ dentro de $`e`$, tomamos $`N_j^{(e)}(x,y)=a_j+b_jx+c_jy`$. De aquí, para $`j=1`$
$`a_1+b_1x_1+c_1y_1=1`$
$`a_1+b_1x_2+c_1y_2=0`$ (4)
$`a_1+b_1x_3+c_1y_3=0`$
de donde
$`a_1={\displaystyle \frac{1}{2S_e}}(x_2y_3x_3y_2)`$
$`b_1={\displaystyle \frac{1}{2S_e}}(y_2y_3)`$
$`c_1={\displaystyle \frac{1}{2S_e}}(x_2x_3)`$
donde $`S_e`$ es el área del triángulo $`e`$.
Ahora, siguiendo el método de Galerkin, tomamos $`\varphi _i=N_i^{(e)}`$. De esta forma, expresamos $`\psi `$ como
$$\psi (x,y)\underset{f,j}{}\mathrm{\Psi }_j^{(f)}N_j^{(f)}(x,y),$$
(5)
donde la suma se realiza para todos los triángulos ($`f`$) y todos los vértices de cada triángulo ($`j`$); $`\mathrm{\Psi }_j^{(f)}`$ es el valor de $`\psi `$ en el vértice $`j`$ del triángulo $`f`$. Estos coeficientes se encuentran, para cada triángulo, a partir de una ecuación similar a (3):
$$\underset{j=1}{\overset{3}{}}k_{ij}^{(e)}\mathrm{\Psi }_j^{(e)}=\frac{1}{3}S_eg_e$$
con $`k_{ij}^{(e)}S_e(b_ib_j+c_ic_j)`$ (coeficentes de acoplamiento); $`g_eg(\overline{x}_e,\overline{y}_e)`$, y $`(\overline{x}_e,\overline{y}_e)`$ son las coordenadas del centro de gravedad del triángulo. A continuación sólo falta incluir todos los triángulos de la red. Para ello, considerando que los vértices interiores a $`C`$ son $`N`$, y el total de vértices (interiores a $`C`$ y sobre ella) es $`N_0`$, los índices corren de $`1`$ a $`N`$ para los vértices internos, y de $`N+1`$ a $`N_0`$ para los que están sobre la frontera. Con esto, se obtiene la ecuación equivalente a (3) para toda la red es con
$`𝐊=(k_{ij}),k_{ii}={\displaystyle \underset{T}{}}k_{ii}^{(e)},k_{ij}={\displaystyle \underset{E}{}}k_{ij},ij,`$
$`G_i={\displaystyle \frac{1}{3}}{\displaystyle \underset{T}{}}S_eg_e{\displaystyle \underset{j=N+1}{\overset{N_0}{}}}k_{ij}^{(e)}\mathrm{\Psi }_j^{(e)};`$
$`T`$ indica que la suma es sobre los triángulos con vértice común $`i`$; $`E`$ que la suma es sobre triángulos con lados entre los vértices $`i`$, $`j`$.
Como ya se dijo, $`𝐊`$ es una matriz dispersa, y por tanto la solución a (1) se puede encontrar como
$$\psi (x,y)\underset{f,j}{}\mathrm{\Psi }_j^{(f)}N_j^{(f)}(x,y).$$
|
warning/0005/cond-mat0005094.html
|
ar5iv
|
text
|
# Anomalous crossover between thermal and shot noise in macroscopic diffusive conductors
## Abstract
We predict the existence of an anomalous crossover between thermal and shot noise in macroscopic diffusive conductors. We first show that, besides thermal noise, these systems may also exhibit shot noise due to fluctuations of the total number of carriers in the system. Then we show that at increasing currents the crossover between the two noise behaviors is anomalous, in the sense that the low frequency current spectral density displays a region with a superlinear dependence on the current up to a cubic law. The anomaly is due to the non-trivial coupling in the presence of the long range Coulomb interaction among the three time scales relevant to the phenomenon, namely, diffusion, transit and dielectric relaxation time.
Shot noise and thermal noise are the two prototypes of noise present in nature. Thermal noise is displayed by a conductor at or near equilibrium, and is associated with its conductance through Nyquist theorem $`S_I^{ther}(0)=4k_BTG`$, where $`S_I^{ther}(0)`$ is the low frequency current spectral density, $`k_B`$ the Boltzman constant, $`T`$ the temperature and $`G`$ the conductance. Shot noise is due to the discreteness of the carriers charge, and displays a low frequency spectral density of current fluctuations in the form $`S_I^{shot}(0)=\gamma 2q\overline{I}`$, where $`\overline{I}`$ is the average dc current, $`q`$ the carrier charge and $`\gamma `$ the so called Fano Factor. Being an excess noise, it can only be observed under non-equilibrium conditions and provides information not available from linear response coefficients such as conductance. Following Landauer’s ideas, these two types of noise are special forms of a more general noise formula representing different manifestations of the same underlying microscopic mechanisms. As a result, for systems displaying shot noise one should expect a continuous and smooth transition between the equilibrium thermal noise and the non-equilibrium shot noise. Two examples of such transitions are provided by the expressions
$$S_I(0)=2q\overline{I}\mathrm{coth}\left(\frac{qV}{2k_BT}\right)\text{,}$$
(1)
and
$$S_I(0)=4k_BTG\left[(1\gamma )+\gamma \frac{qV}{2k_BT}\mathrm{coth}\left(\frac{qV}{2k_BT}\right)\right]\text{,}$$
(2)
which represent standard transitions for a classical and a quantum system, respectively. In previous equations $`V`$ is the applied voltage. In both cases one obtains $`S_I^{ther}(0)`$ at or near equilibrium, when $`\left|qV/k_BT\right|1`$, and $`S_I^{shot}(0)`$ far from equilibrium, when $`\left|qV/k_BT\right|1`$.
A variety of classical and quantum physical systems exhibit the above $`coth`$-like cross-over. Among them we remind $`pn`$ junctions, Schottky barrier diodes, tunnel diodes, mesoscopic diffusive conductors with coherent, and semiclassical transport , etc. We remark that an essential feature of the above formulae is to predict a monotonic increase of the spectral density with current which never exceeds a linear dependence. Finally, we note that it is common believe that macroscopic conductors do not display shot noise.
The aim of this article is to prove that macroscopic conductors can display shot noise and that the transition between thermal and shot noise shows a remarkable deviation from the standard $`coth`$-like behavior. In particular, the region of cross-over evidences a current spectral density which increases more than linearly with current, up to a cubic dependence.
The system under consideration is a macroscopic homogeneous diffusive conductor of length $`L`$, (henceforth shortly referred to as macroscopic diffusive conductor). The conductor is considered to be macroscopic in the sense that the sample length $`L`$ satisfies $`Ll_{in},l_e`$, where $`l_{in}`$ and $`l_e`$ are the inelastic and elastic mean free paths, respectively. Moreover, homogeneous conditions implies that the stationary electric field and charge density profiles are homogeneous. Although at first sight it seems surprising that macroscopic diffusive conductors are able to display shot noise, see for instance Ref. , it is easy to convince oneself that this is indeed the case. The key argument is provided by the fact that the diffusion of carriers through the sample, a part from velocity fluctuations, also induce fluctuations of the total number of particles inside the sample. These number fluctuations are related to the fact that the time a carrier spends to cross the sample depends on the particular succession of scattering events, thus giving rise to fluctuations in the instantaneous value of the total number of particles inside the sample. As a consequence, besides the usual thermal noise associated with velocity fluctuations, we will have an excess noise associated with number fluctuations. Note, that existing arguments against the presence of shot noise in macroscopic conductors are always based on the assumption that number fluctuations are negligible, what is not always true in macroscopic diffusive conductors, as will be shown below.
That number fluctuations can give rise to shot noise can be seen as follows. The excess noise associated with number fluctuations can be characterized as
$$S_I^{ex}(0)=\left(\frac{\overline{I}}{\overline{N}}\right)^2S_N(0)\text{,}$$
(3)
where $`S_N(0)=2_{\mathrm{}}^+\mathrm{}𝑑t\overline{\delta N(0)\delta N(t)}`$ is the low frequency spectral density of number fluctuations and $`\overline{N}`$ the average number of carriers inside the system. Furthermore, within an exponential model for the decay of number fluctuations one assumes $`S_N(0)=\overline{\delta N^2}\tau _N`$ where $`\overline{\delta N^2}`$ is the variance and $`\tau _N`$ the relaxation time for such a fluctuations. If the relaxation of number fluctuations takes place on a time scale of the order of the transit time $`\tau _T`$, then one has $`\tau _N\tau _T`$. By using in Eq.(3) that for a diffusive conductor $`\tau _T=L/v=q\overline{N}/\overline{I}`$, where $`v`$ is the drift velocity, and where we have used that $`\overline{I}=qA\overline{n}v`$, with $`A`$ being the cross sectional area and $`\overline{n}`$ the average carrier density, we obtain $`S_I^{ex}(0)q\left(\overline{\delta N^2}/\overline{N}\right)\overline{I}`$, which is shot noise like.
Therefore, macroscopic diffusive conductors offer a new and simple example in which to investigate in detail the transition between thermal and shot noise. To this purpose, we need an explicit expression for the current spectral density valid, in particular, in the transition region between thermal and shot noise. This explicit expression can be obtained by solving the appropriate equations for the fluctuations. For simplicity the sample is assumed to have a transversal size sufficiently thick to allow a one dimensional electrostatic treatment in the $`x`$ direction and to neglect the effects of boundaries in the $`y`$ and $`z`$ directions. Furthermore, since we are interested in the low frequency noise properties (beyond $`1/f`$ noise), we will neglect the displacement current. Accordingly the standard drift-diffusion Langevin equation for a macroscopic diffusive conductor reads
$$\frac{I(t)}{A}=qn\mu E+qD\frac{dn}{dx}+\frac{\delta I_x(t)}{A}\text{,}$$
(4)
which after linearization around the stationary homogeneous state gives
$`{\displaystyle \frac{\delta I(t)}{A}}`$ $`=`$ $`q\mu \overline{E}\eta \delta n_x(t)+q\overline{n}\mu \delta E_x(t)+`$ (6)
$`qD{\displaystyle \frac{d\delta n_x(t)}{dx}}+{\displaystyle \frac{\delta I_x(t)}{A}}\text{.}`$
Here, $`\delta E_x(t)`$ and $`\delta n_x(t)`$ refer to the fluctuations of electric field and number density at point $`x`$, respectively, while $`\delta I(t)`$ refers to the fluctuations of the total current. Moreover, $`\mu `$ is the mobility, $`\overline{E}`$ the average electric field, $`D`$ the diffusion coefficient and the bar denotes a time average. We assume that $`\mu `$ and $`D`$ may depend on $`\overline{n}`$, in order to include in the model also degenerate conductors. The numerical factor $`\eta =\left(1+\frac{\mu _N^{}}{\mu /\overline{n}}\right)`$ , with $`\mu _N^{}=\frac{\mu }{\overline{n}}`$, accounts for the possible dependence of the mobility on the number density and $`\delta I_x(t)`$ is a Langevin noise source, which accounts for the fluctuations of current due to the diffusion of carriers inside the sample. It has zero mean and correlation function,
$$\delta I_x(t)\delta I_x^{^{}}(t^{})=\frac{1}{2}K\delta (xx^{})\delta (tt^{})\text{,}$$
(7)
where $`K=4qAk_BT\mu \overline{n}`$ is the strength of the fluctuations. Equation (6) must be supplemented with the Poisson equation
$$\frac{d\delta E_x(t)}{dx}=\frac{q}{ϵ}\delta n_x(t)\text{,}$$
(8)
where $`ϵ`$ is the electric permittivity. Generally, Eqs.(6) and (8) are combined into a single equation for the electric field fluctuation of the form
$$\left(\frac{d^2}{dx^2}+\frac{1}{L_E}\frac{d}{dx}\frac{1}{L_D^2}\right)\delta E_x(t)=\frac{(\delta I_x(t)\delta I(t))}{ϵAD}\text{,}$$
(9)
where $`L_E=D/\eta \mu \overline{E}`$ and $`L_D=\left(Dϵ/\mu q\overline{n}\right)^{1/2}`$.Here, $`L_E/L`$ characterizes the ratio between a characteristic carrier energy and the energy supplied by the applied voltage, and $`L_D`$ is the Debye screening length. The ratio $`L/L_D`$ constitutes a relevant indicator of the effects of the long range Coulomb interaction on the current fluctuations, since for $`L/L_D1`$, one can neglect the term proportional to $`\delta E_x(t)`$ in Eq.(6), and the equation for the current fluctuations becomes uncoupled from the Poisson equation. Moreover, since contact effects are negligible we will use as boundary conditions $`\delta n_0=\delta n_L=0`$, which gives,
$$\frac{d\delta E_x(t)}{dx}|_0=\text{ }\frac{d\delta E_x(t)}{dx}|_L=0\text{.}$$
(10)
Equation (9), together with Eqs. (7) and Eq.(10), constitute a complete set of equations to analyze the noise properties of macroscopic diffusive conductors. In the present form, they can be used to describe both degenerate as well as non-degenerate conductors. The fact that the same underlying scattering mechanisms are responsible for the noise properties of the system is reflected by the presence of a unique Langevin source in the model. Being Eq.(9) a second order differential equation with constant coefficients, its solution can be obtained in a closed analytical form. Hence, from the expression of $`\delta E_x(t)`$ one can compute the voltage fluctuation under fixed current conditions $`\delta _IV(t)=_0^L𝑑x\delta E_x(t)`$ (where one uses $`\delta I(t)=0`$), from where the current spectral density can be obtained as $`S_I(0)=G^22_{\mathrm{}}^+\mathrm{}𝑑t\overline{\delta V_I(0)\delta _IV(t)}`$, with $`G=qA\mu \overline{n}/L`$. After simple but cumbersome algebra, the final result can be written in the form
$$S_I(0)=S_I^{ther}(0)+S_I^{ex}(0)\text{,}$$
(11)
where
$$S_I^{ther}(0)=\frac{K}{L}=4k_BTG\text{,}$$
(12)
and where
$`S_I^{ex}(0)`$ $`=`$ $`K{\displaystyle \frac{(\lambda _2^2\lambda _1^2)}{2L^2\lambda _1^2\lambda _2^2}}{\displaystyle \frac{\left(e^{\lambda _1L}1\right)\left(e^{\lambda _2L}1\right)}{\left(e^{\lambda _2L}e^{\lambda _1L}\right)^2}}\times `$ (15)
$`[\lambda _2(e^{\lambda _2L}+1)(e^{\lambda _1L}1)`$
$`\lambda _1(e^{\lambda _1L}+1)(e^{\lambda _2L}1)]\text{.}`$
Here, $`\lambda _1`$and $`\lambda _2`$ are the two eigenvalues of Eq.(9) and are given by
$$\lambda _{1,2}=\frac{1}{2L_E}\left(1\pm \sqrt{1+4\frac{L_E^2}{L_D^2}}\right)\text{.}$$
(16)
Equations (11)-(15)constitute the general expression for the low frequency current spectral density of a macroscopic diffusive conductor, and represent the main result of the present paper. In Eq.(11) we distinguish two different contributions. The first one, $`S_I^{ther}(0)`$, corresponds to thermal noise. The second one, $`S_I^{ex}(0)`$, constitutes an excess noise and it is directly related to carrier number fluctuations. This can be proved directly by computing $`S_N(0)`$ from the solution of Eq.(9) by considering that the number fluctuations are given through $`\delta N(t)=A_0^L𝑑x\delta n_x=A\frac{ϵ}{q}(\delta E_0(t)\delta E_L(t))`$. One then obtains the identity
$$S_I^{ex}(0)=\left(\frac{\overline{I}}{\overline{N}}\right)^2\eta ^2S_N(0)\text{.}$$
(17)
Equation (17) is of the form of Eq.(3) except for the presence of $`\eta `$ which accounts for the possible dependence of the mobility on carrier density. From Eqs.(15) and (17), it can be shown that when $`L_D^2/L_ELL_D`$ or $`L_DLL_E`$ one has
$$S_I^{ex}(0)=2\gamma q\overline{I}\text{,}$$
(18)
where $`\gamma =\eta k_BT\frac{\mathrm{ln}\overline{N}}{E_F}`$. This result proves the possibility for macroscopic diffusive conductors to display shot noise. By defining a characteristic time associated to number fluctuations through $`\tau _N=S_N(0)/\overline{\delta N^2}^{eq}`$, with $`\overline{\delta N^2}^{eq}=\overline{N}k_BT\frac{\mathrm{ln}\overline{N}}{E_F}`$ being the variance of number fluctuation at equilibrium, Eq.(18) corresponds to a situation in which $`\tau _N(2/\eta )\tau _T`$, thus confirming that when number fluctuations relax on the time scale given by the transit time they give rise to shot noise.
Now we are in a position to investigate the properties of the transition between thermal and shot noise. In Fig.1 we display the current spectral density for an ohmic conductor obtained from Eqs.(11)-(15), as a function of current for different sample lengths. The current is normalized to $`I_R=GV_R`$ where $`V_R=\frac{\overline{n}}{\eta q}\frac{E_F}{\overline{n}}`$. In the present units the curves corresponding to $`L/L_D<1`$ are indistinguishable from the curve corresponding to $`L/L_D=1`$. In the figure we can easily identify the thermal and shot noise regimes as the constant and proportional to current behaviors, respectively. Also depicted for comparison is the current spectral density of Eq.(1) that represents the standard transition between thermal and shot noise for a classical system (empty squares). Remarkably, while the transition between thermal and shot noise follows the standard form for $`L<L_D`$, in the opposite case $`L>L_D`$ it is anomalous. The anomaly is characterized by a spectral density which at most increases with the third power of the current tending asymptotically to
$$\frac{S_I(0)}{S_I^{ther}(0)}=\left[1+\frac{1}{2}\left(\frac{L_D}{L}\right)^4\left(\frac{\overline{I}}{I_R}\right)^3\right]\text{,}$$
(19)
which holds for $`0I\left(L/L_D\right)^2I_R`$ as can be seen in Fig.1, where the filled circles represent Eq.(19). Since this anomalous crossover is absent for $`L<L_D`$ , i.e. when the long range Coulomb interaction does not affect the current fluctuations, we conclude that this interaction plays a central role in this unexpected behavior.
To better understand the role of the long range Coulomb interaction in the origin of this anomaly, we will analyze how the three characteristic times in the system combine to yield $`\tau _N`$. For the present case the following characteristic times can be identified: the diffusion time $`\tau _D=L^2/D`$, the dielectric relaxation time, $`\tau _d=ϵ/q\overline{n}\mu `$ and the, already defined, transit time $`\tau _T`$.
In Fig. 2 we plot $`\tau _N`$ as obtained from our theory as a function of current for different sample lengths. Here, we clearly identify two different behaviors for $`\tau _N`$ depending on whether $`L/L_D1`$ or $`L/L_D1`$. For $`L/L_D1`$ we observe a smooth transition between the equilibrium value $`\tau _N1/3\tau _D`$ and the far from equilibrium value $`\tau _N(2/\eta )\tau _T`$. This result shows that when the long range Coulomb interaction is not effective, only $`\tau _D`$ and $`\tau _T`$ are relevant. As a consequence, near equilibrium we have $`\tau _D\tau _T`$ and number fluctuations are governed by diffusion, while far from equilibrium we have $`\tau _D\tau _T`$ and they are governed by the transit time, thus giving rise to shot noise. On the other hand, when $`L/L_D1`$ the transition between the equilibrium value $`\tau _N4(\tau _d/\tau _D)^{1/2}\tau _d`$ and the far from equilibrium value $`\tau _N(2/\eta )\tau _T`$ is mediated by a region in which $`\tau _N2\eta \tau _d^2/\tau _T`$. The far from equilibrium behavior, being dominated by the transit time gives rise to shot noise, while in the intermediate region $`\tau _N`$ is proportional to the current thus giving rise to the cubic dependence of the current spectral density. Notice that the transition between the intermediate and the shot noise region takes place when $`\tau _N\tau _d\tau _T`$. From these results we conclude that the origin of the anomalous transition between thermal and shot noise can be found in the non-trivial coupling between the different characteristic times in the presence of long range Coulomb interaction.
From the previous analysis we argue that there are two possible ways of providing an experimental test of our theory. The first way is an indirect test to be performed at or neat equilibrium. It consists in proving the non-trivial coupling of the characteristic times in the presence of the long range Coulomb interaction. In this case, when $`L/L_D1`$, one should obtain a characteristic time for number fluctuations in agreement with the relationship $`\tau _N4\left(\tau _d/\tau _D\right)^{1/2}\tau _d=4(ϵ/q\mu \overline{n})^{3/2}D^{1/2}/L`$. The second way is a direct test, which consists in observing the current dependence of the current spectral density. According to Eq.(19) one should observe the anomalous transition for $`LL_D`$ when $`\overline{I}(L/L_D)^{4/3}I_R`$, or analogously for $`\overline{V}/L=\overline{E}(L/L_D)^{4/3}V_R/L`$. To this end, non-degenerate semiconductor systems offer the best possibilities. For a non-degenerate semiconductor, with typical parameters $`\overline{n}10^{14}cm^3`$, $`T300K`$, $`ϵ10ϵ_0`$, one has $`L_D0.4\mu m`$ and $`V_R=k_BT/q=0.0259V`$. Therefore for $`L=50L_D=20\mu m`$ one enters the anomalous regime for $`\overline{E}2kV/cm`$. This value of the electric field is experimentally accessible. In addition, when $`\overline{I}(L/L_D)^2I_R`$, that is for $`\overline{V}/L=\overline{E}\frac{q}{ϵ}\overline{n}L`$, one should enter the regime of shot noise. For the parameters chosen above we obtain the condition $`\overline{E}35kV/cm`$ which is still experimentally accessible.
In summary, we have proven that a macroscopic diffusive conductor can display shot noise, and that the transition between thermal and shot-noise is anomalous when the length of the sample is much longer than the Debye screening length. The anomaly of the transition consists in a nonlinear dependence of the low frequency spectral density of current fluctuations upon the current, which can lead up to a cubic behavior. The origin of this unexpected behavior is related to the non-trivial coupling among diffusion, dielectric relaxation and drift in the presence of the long range Coulomb interaction.
Partial support from the Spanish SEUID and from the EC Improving Human Research Potential program through contract HPMF-CT-1999-00140, are gratefully acknowledged.
|
warning/0005/math0005201.html
|
ar5iv
|
text
|
# Untitled Document
Gerbes of chiral differential operators. III
Vassily Gorbounov, Fyodor Malikov, Vadim Schechtman
Introduction
This note is a sequel to \[GII\]. Its aim is to ”switch on an exterior vector bundle” into the framework of op. cit.
Let $`X`$ be a smooth scheme over a fixed ground ring $`k`$ containing $`1/2`$ and $`E`$ be a vector bundle (i.e. a locally free $`𝒪_X`$-module) of finite rank over $`X`$. Consider the exterior algebra $`\mathrm{\Lambda }E=_{i=0}^{rk(E)}\mathrm{\Lambda }^iE`$ (over $`𝒪_X`$); this is a sheaf of commutative superalgebras over $`X`$, where by definition $`𝒪_X`$ is purely even and the parity of a component $`\mathrm{\Lambda }^iE`$ is equal to the parity of $`i`$.
In this note we study the chiral counterparts of the sheaf $`𝒟_{\mathrm{\Lambda }E}`$ of superalgebras of differential operators acting on $`\mathrm{\Lambda }E`$. Similarly to op. cit., these chiral sheaves of differential operators on $`\mathrm{\Lambda }E`$ exist locally and are by no means unique; the corresponding categories form a champ en groupoids $`𝔇_{\mathrm{\Lambda }E}`$ over $`X`$, called the gerbe of chiral differential operators on $`\mathrm{\Lambda }E`$.
Our first main result (see Theorem 5.9) says that the characteristic class $`c(𝔇_{\mathrm{\Lambda }E})`$ lies in the second hypercohomology group $`H^2(X;\mathrm{\Omega }_X^{[2,3})`$ (i.e. in the same group where $`c(𝔇_X)`$ lies) and is equal to
$$c(𝔇_{\mathrm{\Lambda }E})=c(\mathrm{\Theta }_{X/k})c(E)=c(\mathrm{\Omega }_{X/k}^1)c(E)$$
$`(0.1)`$
where $`c(E)`$ is the ”Atiyah-Chern-Simons” class defined in \[GII\], 7.6. Here $`\mathrm{\Theta }_{X/k}`$ is the tangent bundle. Recall that $`\mathrm{\Omega }_X^{[2,3}`$ denotes the length $`1`$ complex of sheaves $`\mathrm{\Omega }_{X/k}^2\mathrm{\Omega }_{X/k}^{3,closed}`$, with $`\mathrm{\Omega }_{X/k}^2`$ living in degree $`0`$.
As usually, we obtain in fact a stronger statement, namely the equality (0.1) ”on the level of cocycles”. As a corollary of this, we conclude that for $`E=\mathrm{\Theta }_{X/k}`$ or $`E=\mathrm{\Omega }_{X/k}^1`$ our gerbes admit a canonical global section. In other words, there exist canonically defined the sheaves of chiral do $`𝒟_{\mathrm{\Lambda }\mathrm{\Theta }_X}^{ch}`$ and $`𝒟_{\mathrm{\Omega }_X^{}}^{ch}`$.
Section 6 is devoted to the study of the last sheaf, which is nothing but (the underlying sheaf of) chiral de Rham complex from \[MSV\]. We obtain the transformation laws of $`4`$ local generators of $`N=2`$ supersymmetry $`Q,J,G`$ and $`L`$, see Theorem 6.25. In particular, the component $`Q_0`$ of the field $`Q(z)`$ is a globally defined square zero derivation of $`𝒟_{\mathrm{\Omega }_X^{}}^{ch}`$, which is the chiral de Rham differential from op. cit.
This completes an alternative construction of the chiral de Rham complex sketched in Section 6 of \[MSV\]. Its difference from the original construction is that it does not use Wick theorem and the arguments of ”formal geometry”.
In the last section we show that as a simple consequence of the Poincaré-Birkhoff-Witt theorem for $`𝒟_{\mathrm{\Omega }_X^{}}^{ch}`$ and the Lefschetz fixed point theorem one gets a ”moonshine style” formula, cf. Theorem 7.9.
The work was done while V.S. visited IHES. He is grateful to the Institute for the excellent working atmosphere.
§1. Preliminaries
1.1. We keep the assumptions of \[GII\]. We assume that the ground ring $`k`$ contains $`1/2`$. For a $`k`$-supermodule $`M`$, we denote by $`M^{ev}`$ (resp. $`M^{odd}`$) the submodule of even (resp. odd) elements, so that $`M=M^{ev}M^{odd}`$. For a homogeneous element $`aM`$, we denote by $`p(a)/2`$ its parity. When we speak about graded $`k`$-supermodules $`M=_{iI}M_i`$ we imply that the $`I`$-grading is compatible with the parity, i.e. $`M^x=_iM_i^x`$ where $`M_i^x=M^xM_i,x=ev`$ or $`odd`$.
Let $`A`$ be a commutative $`k`$-superalgebra. A Lie superalgebroid over $`A`$ is a Lie superalgebra over $`k`$ equipped with a structure of an $`A`$-module, such that the identities \[GII\] (0.2.1) and (0.2.2) hold true.
1.2. A $`_0`$-graded vertex superalgebra (over $`k`$) is a $`_0`$-graded $`k`$-supermodule $`V=_{i0}V_i`$ equipped with a distinguished even vector $`\text{1}V_0`$ (vacuum vector) and a family of bilinear operations
$${}_{(n)}{}^{}:V\times VV,n,$$
such that
$$p(a_{(n)}b)=p(a)+p(b);V_{i(n)}V_jV_{i+jn1}$$
$`(\mathrm{1.2.1})`$
The following properties must hold:
$$\text{1}_{(n)}a=\delta _{n,1}a;a_{(n)}\text{1}=0\text{for }n0,a_{(1)}\text{1}=a$$
$`(\mathrm{1.2.2})`$
and
$$\underset{j=0}{\overset{\mathrm{}}{}}\left(\genfrac{}{}{0pt}{}{m}{j}\right)(a_{(n+j)}b)_{(m+lj)}c=$$
$$=\underset{j=0}{\overset{\mathrm{}}{}}(1)^j\left(\genfrac{}{}{0pt}{}{n}{j}\right)\left\{a_{(m+nj)}b_{(l+j)}c(1)^{n+p(a)p(b)}b_{(n+lj)}a_{(m+j)}c\right\}$$
$`(\mathrm{1.2.3})`$
for all $`m,n,l`$. A particular case of (1.2.3) corresponding to $`m=0`$:
$$(a_{(n)}b)_{(l)}c=\underset{j=0}{\overset{\mathrm{}}{}}(1)^j\left(\genfrac{}{}{0pt}{}{n}{j}\right)\left\{a_{(nj)}b_{(l+j)}c(1)^{n+p(a)p(b)}b_{(n+lj)}a_{(j)}c\right\}$$
$`(\mathrm{1.2.4})`$
Setting $`n=l=1`$ we get
$$(a_{(1)}b)_{(1)}c=\underset{j=0}{\overset{\mathrm{}}{}}\left\{a_{(1j)}b_{(1+j)}c+(1)^{p(a)p(b)}b_{(2j)}a_{(j)}c\right\}$$
$`(\mathrm{1.2.5})`$
In the sequel we shall work only with $`_0`$-graded vertex superalgebras, and call them simply vertex superalgebras. This $`_0`$-grading will be called the grading by conformal weight.
1.3. Let $`V`$ be a vertex superalgebra. The even operators $`^{(j)}:VV`$ of degree $`j`$ ($`j_0`$) are defined in the same manner as in \[GII\], (0.5.5), and they satisfy \[GII\], (0.5.7), (0.5.8), (0.5.10) and (0.5.11). The ”supercommutativity” formula reads as
$$a_{(n)}b=(1)^{n+p(a)p(b)+1}\underset{j0}{}(1)^j^{(j)}(b_{(n+j)}a)$$
$`(\mathrm{1.3.1})`$
and we have the usual OPE formula
$$[a_{(m)},b_{(n)}]=\underset{j0}{}\left(\genfrac{}{}{0pt}{}{m}{j}\right)(a_{(j)}b)_{(m+nj)}$$
$`(\mathrm{1.3.2})`$
where in the left hand side stands the supercommutator
$$[a_{(m)},b_{(n)}]:=a_{(m)}b_{(n)}(1)^{p(a)p(b)}b_{(n)}a_{(m)}$$
$`(\mathrm{1.3.3})`$
§2. Vertex Superalgebroids
2.1. An extended Lie superalgebroid (over $`A`$) is a quintuple $`𝒯=(A,T,\mathrm{\Omega },,,)`$ where $`T`$ is a Lie superalgebroid over $`A`$, $`\mathrm{\Omega }`$ is an $`A`$-module equipped with a structure of a module over the Lie superalgebra $`T`$, $`:A\mathrm{\Omega }`$ an even $`A`$-derivation and a morphism of $`T`$-modules, $`,:T\times \mathrm{\Omega }A`$ is an even $`A`$-bilinear pairing.
The following identities must be true $`(aA,\tau ,\nu T,\omega \mathrm{\Omega })`$:
$$\tau ,a=\tau (a)$$
$`(\mathrm{2.1.1})`$
$$\tau (a\omega )=\tau (a)\omega +(1)^{p(\tau )p(a)}a\tau (\omega )$$
$`(\mathrm{2.1.2})`$
$$(a\tau )(\omega )=a\tau (\omega )+\tau ,\omega a$$
$`(\mathrm{2.1.3})`$
$$\tau (\nu ,\omega )=[\tau ,\nu ],\omega +(1)^{p(\tau )p(\nu )}\nu ,\tau (\omega )$$
$`(\mathrm{2.1.4})`$
For example to a Lie superalgebroid $`T`$ one associates canonically an extended superalgebroid with $`\mathrm{\Omega }=Hom_A(T,A)`$, as in \[GII\], 1.2.
2.2. De Rham - Chevalley complex. Let $`𝒯=(A,T,\mathrm{\Omega },\mathrm{})`$ be an extended Lie $`A`$-superalgebroid. Let us define $`A`$-modules $`\mathrm{\Omega }^i=\mathrm{\Omega }^i(𝒯),i_0`$, as follows. Set $`\mathrm{\Omega }^0=A,\mathrm{\Omega }^1=\mathrm{\Omega }`$. For $`i2`$, $`\mathrm{\Omega }^i`$ is the submodule of the module of $`A`$-polylinear homomorphisms $`h`$ from $`T^{i1}`$ to $`\mathrm{\Omega }`$ such that the function $`\tau _1,h(\tau _2,\mathrm{},\tau _i)`$ is skew symmetric (in the graded sense) with respect to all permutations of $`(\tau _1,\mathrm{},\tau _i)`$.
For example, if $`𝒯`$ is associated with a Lie superalgebroid $`T`$ as above then $`\mathrm{\Omega }^i=Hom_A(\mathrm{\Lambda }_A^iT,A)`$.
Let us define the even maps $`d_{DR}=d_{DR}^i:\mathrm{\Omega }^i\mathrm{\Omega }^{i+1}`$ as follows. For $`i=0`$ we set $`d_{DR}a=a`$. For $`i1`$ we set
$$d_{DR}h(\tau _1,\mathrm{},\tau _i)=d_{Lie}h(\tau _1,\mathrm{},\tau _i)(1)^{p(h)p(\tau _1)}\tau _1,h(\tau _2,\mathrm{},\tau _i)$$
$`(\mathrm{2.2.1})`$
where
$$d_{Lie}h(\tau _1,\mathrm{},\tau _i)=\underset{j=1}{\overset{i}{}}(1)^{j+1+p(\tau _j)(p(\tau _1)+\mathrm{}+p(\tau _{j1}))}\tau _j(h(\tau _1,\mathrm{},\widehat{\tau }_j,\mathrm{},\tau _i))+$$
$$+\underset{1j<li}{}(1)^{j+l+p(\tau _j)(p(\tau _1)+\mathrm{}+p(\tau _{j1}))+p(\tau _l)(p(\tau _1)+\mathrm{}+\widehat{p(\tau _j)}+\mathrm{}+p(\tau _{l1}))}\times $$
$$\times h([\tau _j,\tau _l],\tau _1,\mathrm{},\widehat{\tau }_j,\mathrm{},\widehat{\tau }_l,\mathrm{},\tau _i)$$
$`(\mathrm{2.2.2})`$
For example,
$$d_{DR}\omega (\tau )=(1)^{p(\omega )p(\tau )}\left\{\tau (\omega )\tau ,\omega \right\},$$
$`(\mathrm{2.2.3})`$
for $`\omega \mathrm{\Omega }^1=\mathrm{\Omega }`$; and
$$d_{DR}h(\tau _1,\tau _2)=h([\tau _1,\tau _2])+(1)^{p(h)p(\tau _1)}\tau _1(h(\tau _2))$$
$$(1)^{p(\tau _2)(p(h)+p(\tau _1)}\tau _2(h(\tau _1))(1)^{p(h)p(\tau _1)}\tau _1,h(\tau _2),$$
$`(\mathrm{2.2.4})`$
for $`h\mathrm{\Omega }^2`$.
Let us introduce the action of the Lie algebra $`T`$ on the modules $`\mathrm{\Omega }^i`$ by
$$\tau (h)(\tau _1,\mathrm{},\tau _{i1})=\tau (h(\tau _1,\mathrm{},\tau _{i1}))$$
$$\underset{j=1}{\overset{i1}{}}(1)^{p(\tau )(p(\tau _1)+\mathrm{}+p(\tau _{j1}))}h(\tau _1,\mathrm{},[\tau ,\tau _j],\mathrm{},\tau _i)$$
$`(\mathrm{2.2.5})`$
Let us define the convolution operators $`\tau ,:\mathrm{\Omega }^i\mathrm{\Omega }^{i1}`$ by
$$\tau ,h(\tau _1,\mathrm{},\tau _{i2}=(1)^{p(\tau )p(h)}h(\tau ,\tau _1,\mathrm{},\tau _{i2})$$
$`(\mathrm{2.2.6})`$
The maps $`\{d_{DR}^i\}`$ may be characterized as a unique collection of maps such that $`d_{DR}^0=`$ and the Cartan formula
$$\tau (h)=\tau ,d_{DR}h+d_{DR}\tau ,h$$
$`(\mathrm{2.2.7})`$
holds true.
The maps $`d_{DR}`$ commute with the action of $`T`$. One checks that $`d_{DR}^2=0`$, so we get a complex $`(\mathrm{\Omega }^{}(𝒯),d_{DR})`$ called the de Rham-Chevalley complex of $`𝒯`$.
2.3. A vertex superalgebroid is a septuple $`𝒜=(A,T,\mathrm{\Omega },,\gamma ,,,c)`$ where $`A`$ is a supercommutative $`k`$-algebra, $`T`$ is a Lie superalgebroid over $`A`$, $`\mathrm{\Omega }`$ is an $`A`$-module equipped with an action of the Lie superalgebra $`T`$, $`:A\mathrm{\Omega }`$ is an even derivation commuting with the $`T`$-action,
$$,:(T\mathrm{\Omega })\times (T\mathrm{\Omega })A$$
is a supersymmetric even $`k`$-bilinear pairing equal to zero on $`\mathrm{\Omega }\times \mathrm{\Omega }`$ and such that $`𝒯_𝒜=(A,T,\mathrm{\Omega },,,|_{T\times \mathrm{\Omega }})`$ is an extended Lie superalgebroid over $`A`$; $`c:T\times T\mathrm{\Omega }`$ is a skew supersymmetric even $`k`$-bilinear pairing and $`\gamma :A\times T\mathrm{\Omega }`$ is an even $`k`$-bilinear map.
The following axioms must hold $`(a,bA;\tau ,\tau _iT)`$:
$$\gamma (a,b\tau )=\gamma (ab,\tau )a\gamma (b,\tau )$$
$$(1)^{p(\tau )(p(a)+p(b))}\tau (a)b(1)^{p(a)p(b)+p(\tau )p(a)+p(\tau )p(b)}\tau (b)a$$
$`(A1)`$
$$a\tau _1,\tau _2=a\tau _1,\tau _2+\gamma (a,\tau _1),\tau _2(1)^{p(a)(p(\tau _1)+p(\tau _2))}\tau _1\tau _2(a)$$
$`(A2)`$
$$c(a\tau _1,\tau _2)=ac(\tau _1,\tau _2)+\gamma (a,[\tau _1,\tau _2])$$
$$(1)^{p(\tau _2)(p(\tau _1)+p(a))}\gamma (\tau _2(a),\tau _1)+(1)^{p(\tau _2)(p(\tau _1)+p(a))}\tau _2(\gamma (a,\tau _1))$$
$$(1)^{p(a)(p(\tau _1)+p(\tau _2))}\frac{1}{2}\tau _1,\tau _2a+(1)^{p(a)(p(\tau _1)+p(\tau _2))}\frac{1}{2}\tau _1\tau _2(a)$$
$$(1)^{p(\tau _2)(p(a)+p(\tau _1))}\frac{1}{2}\tau _2,\gamma (a,\tau _1)$$
$`(A3)`$
$$[\tau _1,\tau _2],\tau _3+(1)^{p(\tau _1)p(\tau _2)}\tau _2,[\tau _1,\tau _3]=\tau _1(\tau _2,\tau _3)$$
$$(1)^{p(\tau _1)p(\tau _2)}\frac{1}{2}\tau _2(\tau _1,\tau _3)(1)^{p(\tau _3)(p(\tau _1)+p(\tau _2))}\frac{1}{2}\tau _3(\tau _1,\tau _2)+$$
$$+(1)^{p(\tau _1)p(\tau _2)}\tau _2,c(\tau _1,\tau _3)+(1)^{p(\tau _3)(p(\tau _1)+p(\tau _2))}\tau _3,c(\tau _1,\tau _2)$$
$`(A4)`$
$$d_{Lie}c(\tau _1,\tau _2,\tau _3)=\frac{1}{2}\{[\tau _1,\tau _2],\tau _3+(1)^{p(\tau _2)p(\tau _3)}[\tau _1,\tau _3],\tau _2$$
$$(1)^{p(\tau _1)(p(\tau _2)+p(\tau _3))}[\tau _2,\tau _3],\tau _1\tau _1(\tau _2,\tau _3)+(1)^{p(\tau _1)p(\tau _2)}\tau _2(\tau _1,\tau _3)$$
$$(1)^{p(\tau _3)(p(\tau _1)+p(\tau _2))}2\tau _3,c(\tau _1,\tau _2)\}$$
$`(A5)`$
where $`d_{Lie}`$ is defined by (2.2.2).
2.4. All the constructions of \[GII\] generalize to the $`/(2)`$-graded case in an obvious manner.
§3. Some formulas
3.1. Let $`A`$ be a smooth $`k`$-algebra of relative dimension $`n`$, such that the $`A`$-module $`T=Der_k(A)`$ is free and admits a base $`\{\overline{\tau }_i\}`$ consisting of commuting vector fields. Let $`E`$ be a free $`A`$-module of rank $`m`$, with a base $`\{\varphi _\alpha \}`$. We shall call the set $`𝔤=\{\overline{\tau }_i;\varphi _\alpha \}AE`$ a frame of $`(A,E)`$.
Consider a commutative $`A`$-superalgebra $`\mathrm{\Lambda }E=_{i=0}^m\mathrm{\Lambda }_A^i(E)`$ where the parity of $`\mathrm{\Lambda }_A^i(E)`$ is equal to the parity of $`i`$. Each frame $`𝔤`$ as above gives rise to a $`\mathrm{\Lambda }E`$-base $`\{\tau _i;\psi _\alpha \}`$ of the Lie superalgebroid $`T_{\mathrm{\Lambda }E}=Der_k(\mathrm{\Lambda }E)`$, defined as follows. We extend the fields $`\overline{\tau }_i`$ to derivations $`\tau _i`$ of the whole superalgebra $`\mathrm{\Lambda }E`$ by the rule
$$\tau _i(a)=\overline{\tau }_i(a);\tau _i(a_\alpha \varphi _\alpha )=\overline{\tau }_i(a_\alpha )\varphi _\alpha $$
$`(\mathrm{3.1.1})`$
(Note that this extension depends on a choice of a base $`\{\varphi _\alpha \}`$ of the module $`E`$.) The fields $`\{\tau _i\}`$ form a $`\mathrm{\Lambda }E`$-base of the even part $`T_{\mathrm{\Lambda }E}^{ev}`$.
We define the odd vector fields $`\psi _\alpha T_{\mathrm{\Lambda }E}^{odd}`$ by
$$\psi _\alpha (a_\nu \varphi _\nu )=a_\alpha ;\psi _\alpha (a)=0$$
$`(\mathrm{3.1.2})`$
These fields form a $`\mathrm{\Lambda }E`$-base of $`T_{\mathrm{\Lambda }E}^{odd}`$.
Let $`\{\omega _i;\rho _\alpha \}`$ be the dual base of the module of $`1`$-superforms $`\mathrm{\Omega }_{\mathrm{\Lambda }E}=Hom_{\mathrm{\Lambda }E}^{ev}(T_{\mathrm{\Lambda }E},\mathrm{\Lambda }E)`$, defined by
$$\tau _i,\omega _j=\delta _{ij};\psi _\alpha ,\rho _\beta =\delta _{\alpha \beta };\tau _i,\rho _\alpha =\psi _\alpha ,\omega _i=0$$
$`(\mathrm{3.1.3})`$
3.2. Let us describe the effect of a change of frame. Let $`𝔤^{}=\{\overline{\tau }_i^{};\varphi _\alpha ^{}\}`$ be another frame, with $`\overline{\tau }_i^{}=g^{ij}\overline{\tau }_j;\varphi _\alpha ^{}=A^{\alpha \beta }\varphi _\beta `$, $`g=(g^{ij})GL_n(A),A=(A^{\alpha \beta })GL_m(A)`$.
The corresponding new bases $`\tau _i^{},`$ etc. look as follows.
$$\tau _i^{}=g^{ip}\tau _p+g^{i\alpha \gamma }\varphi _\gamma \psi _\alpha $$
$`(\mathrm{3.2.1})`$
where
$$g^{i\alpha \gamma }=g^{iq}\tau _q(A^{1\alpha \mu })A^{\mu \gamma }$$
$`(\mathrm{3.2.2})`$
Next,
$$\psi _\alpha ^{}=A^{1\mu \alpha }\psi _\mu $$
$`(\mathrm{3.2.3})`$
$$\omega _i^{}=g^{1pi}\omega _p$$
$`(\mathrm{3.2.4})`$
$$\rho _\alpha ^{}=\tau _i(A^{\alpha \gamma })\varphi _\gamma \omega _i+A^{\alpha \mu }\rho _\mu $$
$`(\mathrm{3.2.5})`$
Formulas for the inverse transformation:
$$\tau _q=g^{1qi}\tau _i^{}+\tau _q(A^{\alpha \gamma })\varphi _\gamma \psi _\alpha ^{}$$
$`(\mathrm{3.2.6})`$
$$\psi _\beta =A^{\alpha \beta }\psi _\alpha ^{}$$
$`(\mathrm{3.2.7})`$
$$\omega _j=g^{pj}\omega _p^{}$$
$`(\mathrm{3.2.8})`$
$$\rho _\beta =A^{1\beta \alpha }\rho _\alpha ^{}+g^{p\beta \gamma }\varphi _\gamma \omega _p^{}$$
$`(\mathrm{3.2.9})`$
These formulas show that $`T`$ is canonically an $`A`$-module quotient of $`T_{\mathrm{\Lambda }E}`$ and $`\mathrm{\Omega }=Hom_A(T,A)`$ is canonically an $`A`$-submodule of $`\mathrm{\Omega }_{\mathrm{\Lambda }E}`$. In fact the whole de Rham complex $`\mathrm{\Omega }_A^{}`$ is canonically the subcomplex of $`\mathrm{\Omega }_{\mathrm{\Lambda }E}^{}`$.
3.3. Recall that
$$g^{ip}\tau _p(g^{jq})=g^{jp}\tau _p(g^{iq})$$
$`(\mathrm{3.3.1})`$
$$g^{ip}\tau _q\tau _p(g^{jq})=g^{jq}\tau _p\tau _q(g^{ip})$$
$`(\mathrm{3.3.2})`$
and
$$\tau _p(g^{1qr})=\tau _q(g^{1pr})$$
$`(\mathrm{3.3.3})`$
see \[GII\], 5.4.
It is easy to see that
$$tr\left\{\tau _p(A)\tau _q(A^1)\right\}=tr\left\{\tau _q(A)\tau _p(A^1)\right\}$$
$`(\mathrm{3.3.4})`$
Using (3.3.1) and (3.3.4) one sees easily that
$$g^{ip}\tau _p(g^{j\nu \nu })=g^{jq}\tau _q(g^{i\nu \nu })$$
$`(\mathrm{3.3.5})`$
3.4. Let $`𝒜=𝒜_{\mathrm{\Lambda }E;𝔤}`$ be the vertex superalgebroid corresponding to the frame $`𝔤`$.
We have the following identities in $`𝒜`$:
$`\gamma `$-formulas
$$\gamma (a,b\tau _i)=\tau _i(a)b\tau _i(b)a$$
$`(\mathrm{3.4.1})`$
$$\gamma (a,b\varphi _\nu \psi _\mu )=\delta _{\nu \mu }ba$$
$`(\mathrm{3.4.2})`$
$$\gamma (a\varphi _\beta ,b\psi _\mu )=\delta _{\beta \mu }ab$$
$`(\mathrm{3.4.3})`$
$`,`$-formulas
$$a\tau _i,b\tau _j=b\tau _i\tau _j(a)a\tau _j\tau _i(b)\tau _i(b)\tau _j(a)$$
$`(\mathrm{3.4.4})`$
$$a\varphi _\alpha \psi _\beta ,b\tau _i=\delta _{\alpha ,\beta }b\tau _i(a)$$
$`(\mathrm{3.4.5})`$
$$a\varphi _\alpha \psi _\beta ,\varphi _\alpha ^{}\psi _\beta ^{}=ab\delta _{\beta \alpha ^{}}\delta _{\beta ^{}\alpha }$$
$`(\mathrm{3.4.6})`$
$`c`$-formulas
$$c(a\tau _i,b\tau _j)=\frac{1}{2}\{\tau _i(b)\tau _j(a)\tau _j(a)\tau _i(b)\}+\frac{1}{2}\{b\tau _i\tau _j(a)a\tau _j\tau _i(b)\}$$
$`(\mathrm{3.4.7})`$
$$c(a\varphi _\alpha \psi _\mu ,b\varphi _\beta \psi _\nu )=\frac{\delta _{\mu \beta }\delta _{\nu \alpha }}{2}\left\{abba\right\}$$
$`(\mathrm{3.4.8})`$
$$c(a\varphi _\alpha \psi _\mu ,b\tau _i)=\frac{\delta _{\mu \alpha }}{2}\left\{b\tau _i(a)\right\}$$
$`(\mathrm{3.4.9})`$
$$c(a\tau _i,b\psi _\alpha )=c(a\varphi _\alpha \psi _\mu ,b\psi _\nu )=0$$
$`(\mathrm{3.4.10})`$
3.5. Let $`𝔤^{}`$ be another frame as in 3.2. We have
$$\gamma (a,\tau _p^{})=\gamma (a,g^{pq}\tau _q+g^{p\mu \nu }\varphi _\nu \psi _\mu )=$$
$$=\tau _q(a)g^{pq}\tau _q(g^{pq})a+g^{p\mu \mu }a$$
$`(\mathrm{3.5.1})`$
$$\gamma (a\varphi _\mu ^{},\psi _\alpha ^{})=\gamma (aA^{\mu \beta }\varphi _\beta ,A^{1\nu \alpha }\psi _\nu )=aA^{\mu \beta }A^{1\beta \alpha }$$
$`(\mathrm{3.5.2})`$
Next,
$$\tau _i^{},\tau _j^{}=g^{ip}\tau _p+g^{i\mu \alpha }\varphi _\alpha \psi _\mu ,g^{jq}\tau _q+g^{j\nu \beta }\varphi _\beta \psi _\nu =$$
$$=2g^{ip}\tau _q\tau _p(g^{jq})\tau _p(g^{jq})\tau _q(g^{ip})+$$
$$+2g^{ip}\tau _p(g^{j\nu \nu })+g^{ip}g^{jq}\tau _p(A^{1\mu \beta })A^{\beta \gamma }\tau _q(A^{1\gamma \sigma })A^{\sigma \mu }$$
$`(\mathrm{3.5.3})`$
and
$$\tau _i^{},\psi _\alpha ^{}=\psi _\alpha ^{},\psi _\beta ^{}=0$$
$`(\mathrm{3.5.4})`$
Finally,
$$c(\tau _i^{},\tau _j^{})=\frac{1}{2}\left\{\tau _p(g^{jq})\tau _q(g^{ip})\tau _q(g^{ip})\tau _p(g^{jq})\right\}+\frac{1}{2}\left\{g^{i\mu \nu }g^{j\nu \mu }g^{j\nu \mu }g^{i\mu \nu }\right\}$$
$`(\mathrm{3.5.5})`$
and
$$c(\tau _i^{},\psi _\alpha ^{})=c(\psi _\alpha ^{},\psi _\beta ^{})=0$$
$`(\mathrm{3.5.6})`$
§4. Chern-Simons term
This Section is parallel to \[GII\], Section 5.
4.1. We keep the setup of the previous section. Let $`𝒜=𝒜_{\mathrm{\Lambda }E;𝔤},𝒜^{}=𝒜_{\mathrm{\Lambda }E;𝔤^{}}`$ (resp., $`=_{\mathrm{\Lambda }E;𝔤},^{}=_{\mathrm{\Lambda }E;𝔤^{}}`$) be the vertex superalgebroids (resp. prealgebroids) corresponding to our frames.
As in \[GII\], 5.5 we have a canonical isomorphism
$$g=g_{𝔤,𝔤^{}}=(Id_{\mathrm{\Lambda }E},Id_{T_{\mathrm{\Lambda }E}},Id_{\mathrm{\Omega }_{\mathrm{\Lambda }E}},h):^{}\stackrel{}{}$$
$`(\mathrm{4.1.1})`$
where
$$h=h_{𝔤,𝔤^{}}:T_{\mathrm{\Lambda }E}\mathrm{\Omega }_{\mathrm{\Lambda }E}$$
$`(\mathrm{4.1.2})`$
is defined by the condition
$$x^{},h(y^{})=\frac{1}{2}x^{},y^{},x^{},y^{}\{\tau _i^{}\}\{\psi _\alpha ^{}\}$$
$`(\mathrm{4.1.3})`$
Using (3.5.3) and (3.5.4) we find the following explicit formulas for $`h`$:
$$h(\tau _i^{})=h^{ij}\omega _j;h(\psi _\alpha ^{})=0$$
$`(\mathrm{4.1.4})`$
where $`h^{ij}=h_\mathrm{\Omega }^{ij}h_E^{ij}`$,
$$h_\mathrm{\Omega }^{ij}=\tau _p\tau _j(g^{ip})+\frac{1}{2}\tau _q(g^{ip})\tau _p(g^{rq})g^{1jr}$$
$`(\mathrm{4.1.5})`$
cf. \[GII\], (5.7.2), and
$$h_E^{ij}=\tau _j(g^{i\nu \nu })+\frac{1}{2}g^{iq}\tau _j(A^{1\mu \beta })A^{\beta \gamma }\tau _q(A^{1\gamma \nu })A^{\nu \mu }$$
$`(\mathrm{4.1.6})`$
The meaning of the notation $`h_\mathrm{\Omega }`$ will become clear below, see §6.
4.2. We have
$$𝒜=g_{}𝒜^{}\stackrel{}{+}𝔟$$
$`(\mathrm{4.2.1})`$
where the closed $`3`$-form $`𝔟\mathrm{\Omega }_{\mathrm{\Lambda }E}^{3,cl}`$ is defined by
$$𝔟(x^{},y^{})=c(x^{},y^{})x^{}(h(y^{}))+(1)^{p(x^{})p(y^{})}y^{}(h(x^{})),$$
$`(\mathrm{4.2.2})`$
$`x^{},y^{}\{\omega _i^{}\}\{\psi _\alpha ^{}\}`$, cf. \[GII\], (5.7.3).
It is easy to see that $`\psi _\alpha ^{}(h(\tau _i^{}))=0`$; on the other hand we know already that $`h(\psi _\alpha ^{})=0`$ and $`c(\psi _\alpha ^{},y^{})=0`$. It follows that $`𝔟\mathrm{\Omega }^{3,cl}\mathrm{\Omega }_{\mathrm{\Lambda }E}^{3,cl}`$.
Next, we have
$$\tau _i^{}(h(\tau _j^{}))=(g^{ip}\tau _p+g^{i\alpha \gamma }\varphi _\gamma \psi _\alpha )(h^{jq}\omega _q)=$$
(note that the second summand is zero)
$$=(g^{ip}\tau _p)(h_\mathrm{\Omega }^{jq}\omega _q)+(g^{ip}\tau _p)(h_E^{jq}\omega _q)=g^{ip}\tau _p(h_\mathrm{\Omega }^{jq})\omega _q+h_\mathrm{\Omega }^{jp}g^{ip}$$
$$g^{ip}\tau _p(h_E^{jq})\omega _qh_E^{jp}g^{ip}$$
$`(\mathrm{4.2.3})`$
It follows that $`𝔟=𝔟_\mathrm{\Omega }𝔟_E`$ where $`𝔟_\mathrm{\Omega },𝔟_E\mathrm{\Omega }^3`$ are given by
$$𝔟_\mathrm{\Omega }(\tau _i^{},\tau _j^{})=\frac{1}{2}\left\{\tau _p(g^{jq})\tau _q(g^{ip})\tau _q(g^{ip})\tau _p(g^{jq})\right\}$$
$$g^{ip}\tau _p(h_\mathrm{\Omega }^{jq})\omega _qh_\mathrm{\Omega }^{jp}g^{ip}+g^{jp}\tau _p(h_\mathrm{\Omega }^{iq})\omega _q+h_\mathrm{\Omega }^{ip}g^{jp}$$
$`(\mathrm{4.2.4})`$
and
$$𝔟_E(\tau _i^{},\tau _j^{})=\frac{1}{2}\left\{g^{i\mu \nu }g^{j\nu \mu }g^{j\nu \mu }g^{i\mu \nu }\right\}$$
$$g^{ip}\tau _p(h_E^{jq})\omega _q+g^{jp}\tau _p(h_E^{iq})\omega _qh_E^{jp}g^{ip}+h_E^{ip}g^{jp}$$
$`(\mathrm{4.2.5})`$
The form $`𝔟_\mathrm{\Omega }`$ has already been computed in \[GII\], Magic Lemma 5.6 and Theorem 6.4 (b), and is equal to
$$𝔟_\mathrm{\Omega }(\tau _i^{},\tau _j^{})=\frac{1}{2}tr\left\{g^1\tau _i^{}(g)g^1\tau _j^{}(g)g^1\tau _r^{}(g)g^1\tau _j^{}(g)g^1\tau _i^{}(g)g^1\tau _r^{}(g)\right\}\omega _r^{}$$
$`(\mathrm{4.2.6})`$
cf. loc.cit. (5.5.3) and (6.4.2). Note that $`𝔟_\mathrm{\Omega }`$ is closed, hence $`𝔟_E`$ is closed.
4.3. Magic Lemma. We have
$$𝔟_E(\tau _i^{},\tau _j^{})=\frac{1}{2}tr\left\{A^1\tau _i^{}(A)A^1\tau _j^{}(A)A^1\tau _r^{}(A)A^1\tau _j^{}(A)A^1\tau _i^{}(A)A^1\tau _r^{}(A)\right\}\omega _r^{}$$
$`(\mathrm{4.3.1})`$
Proof. Let us denote the six terms in (4.2.5) by $`A,A^{},B,B^{},C`$ and $`C^{}`$. We have
$$A=\frac{1}{2}g^{iq}\tau _q(A^{1\mu \alpha })A^{\alpha \nu }\tau _r\left\{g^{jp}\tau _p(A^{1\nu \beta })A^{\beta \mu }\right\}\omega _r=$$
$$=\frac{1}{2}g^{iq}\tau _q(A^{1\mu \alpha })A^{\alpha \nu }\{\tau _r(g^{jp})\tau _p(A^{1\nu \beta })A^{\beta \mu }\omega _r+g^{jp}\tau _r\tau _p(A^{1\nu \beta })A^{\beta \mu }\omega _r+$$
$$+g^{jp}\tau _p(A^{1\nu \beta })\tau _r(A^{\beta \mu })\omega _r\}$$
Next,
$$B=\frac{1}{2}g^{ip}\tau _p\left\{g^{jq}\tau _r(A^{1\mu \beta })A^{\beta \gamma }\tau _q(A^{1\gamma \nu })A^{\nu \mu }\right\}\omega _rg^{ip}\tau _p\tau _r(g^{j\nu \nu })\omega _r=$$
$$=\frac{1}{2}g^{ip}\tau _p(g^{jq})\tau _r(A^{1\mu \beta })A^{\beta \gamma }\tau _q(A^{1\gamma \nu })A^{\nu \mu }\omega _r\frac{1}{2}g^{ip}g^{jq}\tau _p\tau _r(A^{1\mu \beta })A^{\beta \gamma }\tau _q(A^{1\gamma \nu })A^{\nu \mu }\omega _r$$
$$\frac{1}{2}g^{ip}g^{jq}\tau _r(A^{1\mu \beta })\tau _p(A^{\beta \gamma })\tau _q(A^{1\gamma \nu })A^{\nu \mu }\omega _r\frac{1}{2}g^{ip}g^{jq}\tau _r(A^{1\mu \beta })A^{\beta \gamma }\tau _p\tau _q(A^{1\gamma \nu })A^{\nu \mu }\omega _r$$
$$\frac{1}{2}g^{ip}g^{jq}\tau _r(A^{1\mu \beta })A^{\beta \gamma }\tau _q(A^{1\gamma \nu })\tau _p(A^{\nu \mu })\omega _rg^{ip}\tau _p\tau _r(g^{j\nu \nu })\omega _r$$
Finally,
$$C=\frac{1}{2}g^{jq}\tau _p(A^{1\mu \beta })A^{\beta \gamma }\tau _q(A^{1\gamma \nu })A^{\nu \mu }\tau _r(g^{ip})\omega _r\tau _r(g^{ip})\tau _p(g^{j\nu \nu })\omega _r$$
We see that $`A1=C1^{},A2=B^{}2,B1=B^{}1`$ (by (3.4.6)), $`B2=A^{}2,B4=B^{}4`$ and $`C1=A^{}1`$. Next,
$$B6+C2=\tau _r\left\{g^{ip}\tau _p(g^{j\nu \nu })\right\},$$
so $`B6+C2=B^{}6C^{}2`$ by (3.3.5).
So we are left with three terms: $`A3,B3`$ and $`B5`$ and their primed partners. It is easy to see that
$$A3=B3=B5=\frac{1}{2}tr\left\{A^1\tau _i^{}(A)A^1\tau _j^{}(A)A^1\tau _r^{}(A)\right\}\omega _r^{},$$
which implies the Lemma. $``$
§5. Atiyah term
This section is parallel to \[GII\], Section 6.
5.1. We keep the setup of the previous section. Let us denote by $`𝒯_{\mathrm{\Lambda }E}=(\mathrm{\Lambda }E,T_{\mathrm{\Lambda }E},\mathrm{\Omega }_{\mathrm{\Lambda }E},)`$ the extended vertex superalgebroid Lie corresponding to our data.
Let $`𝔤^{\prime \prime }=\{\overline{\tau }_i^{\prime \prime },\varphi _\alpha ^{\prime \prime }\}`$ be a third frame of $`(A,E)`$, with $`\overline{\tau }_i^{\prime \prime }=g^{ij}\overline{\tau }_j^{},\varphi _\alpha ^{\prime \prime }=A^{\alpha \beta }\varphi _\beta ^{}`$. We have the corresponding new bases of $`T_{\mathrm{\Lambda }E}`$ and $`\mathrm{\Omega }_{\mathrm{\Lambda }E}`$ given by
$$\tau _i^{\prime \prime }=g^{ip}\tau _p^{}+g^{i\alpha \gamma }\varphi _\gamma ^{}\psi _\alpha ^{}$$
$`(\mathrm{5.1.1})`$
where
$$g^{i\alpha \gamma }=g^{iq}\tau _q^{}(A^{1\alpha \mu })A^{\mu \gamma }$$
$`(\mathrm{5.1.2})`$
$$\psi _\alpha ^{\prime \prime }=A^{1\mu \alpha }\psi _\mu ^{}$$
$`(\mathrm{5.1.3})`$
$$\omega _i^{\prime \prime }=g^{1pi}\omega _p^{}$$
$`(\mathrm{5.1.4})`$
$$\rho _\alpha ^{\prime \prime }=\tau _i^{}(A^{\alpha \gamma })\varphi _\gamma ^{}\omega _i^{}+A^{\alpha \mu }\rho _\mu ^{}$$
$`(\mathrm{5.1.5})`$
5.2. Let $`𝒜^{\prime \prime }=𝒜_{\mathrm{\Lambda }E,𝔤^{\prime \prime }}`$ (resp. $`^{\prime \prime }=_{\mathrm{\Lambda }E;𝔤^{\prime \prime }}`$) be the vertex superalgebroid (resp. prealgebroid) corresponding to the third frame. We have canonical isomorphisms
$$^{\prime \prime }\stackrel{g_{𝔤^{},𝔤^{\prime \prime }}}{\stackrel{}{}}^{}\stackrel{g_{𝔤,𝔤^{}}}{\stackrel{}{}}$$
as well as the morphism $`g_{𝔤,𝔤^{\prime \prime }}:^{\prime \prime }\stackrel{}{}`$ over $`Id_{𝒯_{\mathrm{\Lambda }E}}`$, given by functions $`h_{𝔤,𝔤^{}},h_{𝔤^{},𝔤^{\prime \prime }},h_{𝔤,𝔤^{\prime \prime }}`$, and we are aiming to compute the discrepancy
$$𝔞=𝔞_{𝔤,𝔤^{},𝔤^{\prime \prime }}:=h_{𝔤,𝔤^{}}+h_{𝔤^{},𝔤^{\prime \prime }}h_{𝔤,𝔤^{\prime \prime }}\mathrm{\Omega }_{\mathrm{\Lambda }E}^2$$
$`(\mathrm{5.2.1})`$
Note that
$$\gamma (a,b\psi _\mu )=\psi _\mu (a)b\psi _\mu (b)a=0$$
$`(\mathrm{5.2.2})`$
(in $`𝒜`$), hence
$$h_{𝔤,𝔤^{}}(a\psi _\mu ^{})=ah_{𝔤,𝔤^{}}(\psi _\mu ^{})\gamma (a,\psi _\mu ^{})=0$$
$`(\mathrm{5.2.3})`$
therefore
$$h_{𝔤,𝔤^{}}(\psi _\alpha ^{\prime \prime })=h_{𝔤,𝔤^{}}(A^{1\mu \alpha }\psi _\mu ^{})=0$$
$`(\mathrm{5.2.4})`$
It follows that
$$𝔞(\psi _\alpha ^{\prime \prime })=0$$
$`(\mathrm{5.2.5})`$
5.3. Let us denote for brevity $`h:=h_{𝔤,𝔤^{}},h^{}=h_{𝔤^{},𝔤^{\prime \prime }},h^{\prime \prime }=h_{𝔤,𝔤^{\prime \prime }}`$.
We have
$$h(a\tau _p^{})=ah(\tau _p^{})\gamma (a,\tau _p^{})=$$
$$=ah^{pr}\omega _r+\tau _q(a)g^{pq}+\tau _q(g^{pq})ag^{p\mu \mu }a$$
$`(\mathrm{5.3.1})`$
and
$$h(a\varphi _\gamma ^{}\psi _\alpha ^{})=aA^{\gamma \beta }A^{1\beta \alpha }$$
$`(\mathrm{5.3.2})`$
Thus we get
$$h(\tau _i^{\prime \prime })=h(g^{ip}\tau _p^{}+g^{i\alpha \gamma }\varphi _\gamma ^{}\psi _\alpha ^{})=$$
$$=g^{ip}h_\mathrm{\Omega }^{pr}\omega _rg^{ip}h_E^{pr}\omega _r+\tau _q(g^{ip})g^{pq}+\tau _q(g^{pq})g^{ip}$$
$$g^{p\mu \mu }g^{ip}g^{i\alpha \gamma }A^{\gamma \beta }A^{1\beta \alpha }$$
$`(\mathrm{5.3.3})`$
It follows that
$$𝔞=𝔞_\mathrm{\Omega }𝔞_E$$
$`(\mathrm{5.3.4})`$
where
$$𝔞_\mathrm{\Omega }(\tau _i^{\prime \prime })=g^{ip}h_\mathrm{\Omega }^{pr}\omega _r+\tau _q(g^{ip})g^{pq}+\tau _q(g^{pq})g^{ip}+h_\mathrm{\Omega }^{ip}\omega _p^{}h_\mathrm{\Omega }^{\prime \prime ir}\omega _r$$
$`(\mathrm{5.3.5})`$
and
$$𝔞_E(\tau _i^{\prime \prime })=g^{ip}h_E^{pr}\omega _r+g^{p\mu \mu }g^{ip}+g^{i\alpha \gamma }A^{\gamma \beta }A^{1\beta \alpha }+h_E^{ip}\omega _p^{}h_E^{\prime \prime ir}\omega _r$$
$`(\mathrm{5.3.6})`$
The form $`𝔞_\mathrm{\Omega }`$ has already been computed in \[GII\], Section 6. Namely, by Theorem 6.4 (a) from op. cit.,
$$𝔞_\mathrm{\Omega }(\tau _i^{\prime \prime })=\frac{1}{2}tr\{g^1\tau _i^{\prime \prime }(g^{})\tau _s^{\prime \prime }(g)g^1g^1\tau _s^{\prime \prime }(g^{})\tau _i^{\prime \prime }(g)g^1\}\omega _s^{\prime \prime }$$
$`(\mathrm{5.3.7})`$
5.4. Let us compute $`𝔞_E(\tau _i^{\prime \prime })`$. Let us denote the five terms in the rhs of (5.3.6) by $`𝔄,𝔅,,𝔇`$ and $`𝔈`$. Thus,
$$𝔄=g^{ip}h_E^{pr}\omega _r=\frac{1}{2}g^{ip}g^{pq}\tau _r(A^{1\mu \beta })A^{\beta \gamma }\tau _q(A^{1\gamma \nu })A^{\nu \mu }\omega _r+g^{ip}\tau _r(g^{p\nu \nu })\omega _r$$
$`(\mathrm{5.4.1})`$
$$𝔅=g^{p\mu \mu }g^{ip}=g^{pq}\tau _q(A^{1\mu \alpha })A^{\alpha \mu }\tau _r(g^{ip})\omega _r$$
$`(\mathrm{5.4.2})`$
$$=g^{i\alpha \gamma }A^{\gamma \beta }A^{1\beta \alpha }=g^{ip}\tau _p^{}(A^{1\alpha \mu })A^{\mu \gamma }A^{\gamma \beta }\tau _r(A^{1\beta \alpha })\omega _r$$
$`(\mathrm{5.4.3})`$
$$𝔇=h_E^{ij}\omega _j^{}=\frac{1}{2}g^{is}\tau _j^{}(A^{1\mu \beta })A^{\beta \gamma }\tau _s^{}(A^{1\gamma \nu })A^{\nu \mu }g^{1rj}\omega _r+\tau _j^{}(g^{i\nu \nu })g^{1rj}\omega _r=$$
$$=\frac{1}{2}g^{is}\tau _r(A^{1\mu \beta })A^{\beta \gamma }g^{sq}\tau _q(A^{1\gamma \nu })A^{\nu \mu }\omega _r+\tau _r\left\{g^{ip}g^{pq}\tau _q(A^{1\nu \mu })A^{\mu \nu }\right\}\omega _r$$
$`(\mathrm{5.4.4})`$
and
$$𝔈=h^{\prime \prime ir}\omega _r=\frac{1}{2}(g^{}g)^{iq}\tau _r\left((A^{}A)^{1\mu \beta }\right)(A^{}A)^{\beta \gamma }\tau _q\left((A^{}A)^{1\gamma \nu }\right)(A^{}A)^{\nu \mu }\omega _r\tau _r(g^{\prime \prime i\nu \nu })=$$
$$=\frac{1}{2}(g^{}g)^{iq}\{\tau _r(A^{1\mu \sigma })A^{1\sigma \beta }+A^{1\mu \sigma }\tau _r(A^{1\sigma \beta })\}A^{\beta \rho }A^{\rho \gamma }\times $$
$$\times \{\tau _q(A^{1\gamma \delta })A^{1\delta \nu }+A^{1\gamma \delta }\tau _q(A^{1\delta \nu })\}A^{\nu ϵ}A^{ϵ\mu }\omega _r\tau _r(g^{\prime \prime i\nu \nu })=$$
$$=\frac{1}{2}g^{ip}g^{pq}\{\tau _r(A^{1\mu \sigma })A^{\sigma \gamma }\tau _q(A^{1\gamma \delta })A^{\delta \mu }\omega _r+\tau _r(A^{1\mu \sigma })\tau _q(A^{1\sigma \nu })A^{\nu ϵ}A^{ϵ\mu }\omega _r+$$
$$+\tau _r(A^{1\sigma \beta })A^{\beta \rho }A^{\rho \gamma }\tau _q(A^{1\gamma \sigma })\omega _r+\tau _r(A^{1\sigma \beta })A^{\beta \rho }\tau _q(A^{1\rho \nu })A^{\nu \sigma }\omega _r\}$$
$$\tau _r\left\{(g^{}g)^{iq}\tau _q\left((A^{}A)^{1\nu \mu }\right)(A^{}A)^{\mu \nu }\right\}\omega _r$$
$`(\mathrm{5.4.5})`$
We see that $`𝔄1=𝔈1,=2𝔈2,𝔇1=𝔈4`$. It is easy to see that $`𝔄2+𝔇2+𝔈5=𝔅`$.
Finally,
$$+𝔈2=\frac{1}{2}tr\left\{A^1\tau _i^{\prime \prime }(A^{})\tau _r^{\prime \prime }(A)A^1\right\}\omega _r^{\prime \prime }$$
$`(\mathrm{5.4.6})`$
and
$$𝔈3=\frac{1}{2}tr\left\{A^1\tau _r^{\prime \prime }(A^{})\tau _i^{\prime \prime }(A)A^1\right\}\omega _r^{\prime \prime }$$
$`(\mathrm{5.4.7})`$
So, we have proven
5.5. Lemma. The form $`𝔞_E`$ is given by
$$𝔞_E(\tau _i^{\prime \prime })=\frac{1}{2}tr\left\{A^1\tau _i^{\prime \prime }(A^{})\tau _r^{\prime \prime }(A)A^1A^1\tau _r^{\prime \prime }(A^{})\tau _i^{\prime \prime }(A)A^1\right\}\omega _r^{\prime \prime }$$
$`(\mathrm{5.5.1})`$
Combining 4.3 and 5.5 we get
5.6. Theorem. (a) The cocycle $`𝔞_{𝔤,𝔤^{},𝔤^{\prime \prime }}`$ is given by
$$𝔞_{𝔤,𝔤^{},𝔤^{\prime \prime }}(\tau _i^{\prime \prime })=\frac{1}{2}tr\{g^1\tau _i^{\prime \prime }(g^{})\tau _r^{\prime \prime }(g)g^1g^1\tau _r^{\prime \prime }(g^{})\tau _i^{\prime \prime }(g)g^1\}\omega _r^{\prime \prime }$$
$$\frac{1}{2}tr\left\{A^1\tau _i^{\prime \prime }(A^{})\tau _r^{\prime \prime }(A)A^1A^1\tau _r^{\prime \prime }(A^{})\tau _i^{\prime \prime }(A)A^1\right\}\omega _r^{\prime \prime }$$
$`(\mathrm{5.6.1})`$
(b) The $`3`$-form $`𝔟_{𝔤,𝔤^{}}`$ is given by
$$𝔟_{𝔤,𝔤^{}}(\tau _i^{},\tau _j^{})=\frac{1}{2}tr\left\{g^1\tau _i^{}(g)g^1\tau _j^{}(g)g^1\tau _r^{}(g)g^1\tau _j^{}(g)g^1\tau _i^{}(g)g^1\tau _r^{}(g)\right\}\omega _r^{}+$$
$$+\frac{1}{2}tr\left\{A^1\tau _i^{}(A)A^1\tau _j^{}(A)A^1\tau _r^{}(A)A^1\tau _j^{}(A)A^1\tau _i^{}(A)A^1\tau _r^{}(A)\right\}\omega _r^{}$$
$`(\mathrm{5.6.2})`$
5.7. Lemma. Let $`E^{}=Hom_A(E,A)`$ be the dual module. We have
$$(𝔞_E^{},𝔟_E^{})=(𝔞_E,𝔟_E)$$
$`(\mathrm{5.7.1})`$
Indeed, this follows from the easy identities
$$tr\left\{A^t\tau _i((A^t)^1)A^t\tau _j((A^t)^1)A^t\tau _r((A^t)^1)\right\}=tr\left\{A^1\tau _r(A)A^1\tau _j(A)A^1\tau _i(A)\right\}$$
$`(\mathrm{5.7.2})`$
and
$$tr\{A^t\tau _i((A^t)^1\tau _j((B^t)^1)B^t\}=tr\left\{A^1\tau _i(A)\tau _j(B)B^1\right\}$$
$`(\mathrm{5.7.3})`$
5.8. Let us pass to the global situation. Let $`X`$ be a smooth variety over $`k`$ and $`E`$ be a vector bundle over $`X`$. As in \[GII\], we define the gerbe $`𝔇_{\mathrm{\Lambda }E}`$ of chiral differential operators on $`\mathrm{\Lambda }E`$ over $`X`$.
Its characteristic class $`c(𝔇_{\mathrm{\Lambda }E})`$ will belong to the second hypercohomology $`H^2(X;\mathrm{\Omega }_{\mathrm{\Lambda }E}^{[2,3})`$ (in obvious notations). Recall that we have a canonical imbedding of de Rham complexes
$$\mathrm{\Omega }_X^{}\mathrm{\Omega }_{\mathrm{\Lambda }E}^{}$$
$`(\mathrm{5.8.1})`$
In \[GII\], 7.6 we have defined the ”Atiyah-Chern-Simons” characteristic class $`c(E)H^2(X;\mathrm{\Omega }_X^{[2,3})`$; let us denote by $`c(E)_{\mathrm{\Lambda }E}`$ its image in $`H^2(X;\mathrm{\Omega }_{\mathrm{\Lambda }E}^{[2,3})`$.
The theorem below is an immediate consequence Theorem 5.6 and Lemma 5.7.
5.9. Theorem. The class $`c(𝔇_{\mathrm{\Lambda }E})`$ is equal to
$$c(𝔇_{\mathrm{\Lambda }E})=c(\mathrm{\Theta }_X)c(E)=c(\mathrm{\Omega }_X^1)c(E)$$
$`(\mathrm{5.9.1})`$
where $`\mathrm{\Theta }_X`$ is the tangent bundle.
§6. Chiral de Rham complex
6.1. Let us return to the local situation 3.1, 4.1. Let $`E`$ be equal to the module of vector fields $`T`$. Given a base $`\{\overline{\tau }_i\}`$ consisting of commuting vector fields, we get a frame $`𝔤=\{\overline{\tau }_i;\varphi _i:=\overline{\tau }_i\}`$ of $`(A,E)`$. Let us call such frames natural.
Let $`𝔤,𝔤^{}`$ be two natural frames, with transition matrices as in 3.2. By definition, $`(A^{rs})=(g^{rs})`$. Therefore the coefficients $`g^{i\alpha \gamma }`$ (3.2.2) are given by
$$g^{i\alpha \gamma }=g^{iq}\tau _q(g^{1\alpha \mu })g^{\mu \gamma }=g^{iq}\tau _\alpha (g^{1q\mu })g^{\mu \gamma }=$$
$$=\tau _\alpha (g^{iq})g^{1q\mu }g^{\mu \gamma }=\tau _\alpha (g^{i\gamma })$$
$`(\mathrm{6.1.1})`$
where we have used (3.3.3). Consequently the function $`h_E`$ (4.1.6) is given by
$$h_E^{ij}=\tau _j\tau _\nu (g^{i\nu })+\frac{1}{2}g^{iq}\tau _j(g^{1\mu \beta })g^{\beta \gamma }\tau _q(g^{1\gamma \nu })g^{\nu \mu }$$
$`(\mathrm{6.1.2})`$
The second summand is equal to
$$\frac{1}{2}g^{iq}\tau _j(g^{1\mu \beta })\tau _q(g^{\beta \gamma })g^{1\gamma \nu }g^{\nu \mu }=\frac{1}{2}g^{iq}\tau _j(g^{1\mu \beta })\tau _q(g^{\beta \mu })=$$
$$=\frac{1}{2}g^{iq}\tau _\mu (g^{1j\beta })\tau _q(g^{\beta \mu })=\frac{1}{2}\tau _\mu (g^{1j\beta })g^{\beta q}\tau _q(g^{i\mu })=\frac{1}{2}g^{1j\beta }\tau _\mu (g^{\beta q})\tau _q(g^{i\mu })$$
$`(\mathrm{6.1.3})`$
We see that the first summand in (6.1.2) is equal to minus the first summand of (4.1.5), and second summand of (6.1.2) is equal to the second summand of (4.1.5). Thus
$$h^{ij}=2\tau _p\tau _j(g^{ip})$$
$`(\mathrm{6.1.4})`$
If $`𝔤,𝔤^{},𝔤^{\prime \prime }`$ are natural frames of $`(A,T)`$ then Theorem 5.6 says that $`𝔞_{𝔤,𝔤^{},𝔤^{\prime \prime }}=𝔟_{𝔤,𝔤^{}}=0`$. This means that
the chiral superalgebroid $`𝒜_{\mathrm{\Lambda }T;𝔤}`$ does not depend, up to a canonical isomorphism, on the choice of the base $`\{\overline{\tau }_i\}`$. In other words, we have a canonically defined chiral algebroid $`𝒜_{\mathrm{\Lambda }T}`$.
Passing to chiral envelopes, we get a canonically defined chiral (vertex) superalgebra $`D_{\mathrm{\Lambda }T}^{ch}`$ of chiral differential operators on $`T`$.
It follows that
for each smooth variety $`X`$ we have a canonically defined sheaf of chiral superalgebras $`𝒟_{\mathrm{\Lambda }\mathrm{\Theta }_X}^{ch}`$. These Zariski sheaves form in fact a sheaf in the étale topology.
The gluing functions for this sheaf are given explicitly by (6.1.4).
6.2. Lemma. In the situation 4.1, consider the function $`h_E`$
$$h_E^{ij}=\tau _j(g^{i\nu \nu })+\frac{1}{2}g^{iq}\tau _j(A^{1\mu \beta })A^{\beta \gamma }\tau _q(A^{1\gamma \nu })A^{\nu \mu }$$
$`(\mathrm{6.2.1})`$
The function $`h_E^{}`$ associated with the dual module $`E^{}`$ is given by
$$h_E^{}^{ij}=\tau _j(g^{i\nu \nu })+\frac{1}{2}g^{iq}\tau _j(A^{1\mu \beta })A^{\beta \gamma }\tau _q(A^{1\gamma \nu })A^{\nu \mu }$$
$`(\mathrm{6.2.2})`$
This follows from the identities
$$tr\left\{\tau _i(A^t)A^1\right\}=tr\left\{\tau _i(A^1)A\right\}$$
$`(\mathrm{6.2.3})`$
and
$$tr\left\{\tau _i(A^t)A^1\tau _j(A^t)A^1\right\}=tr\left\{\tau _i(A^1)A\tau _j(A^1)A\right\}$$
$`(\mathrm{6.2.4})`$
(cf. (5.7.2), (5.7.3)).
6.3. Let $`E`$ be the module of $`1`$-forms $`\mathrm{\Omega }=\mathrm{\Omega }_{A/k}^1`$; its exterior algebra is the de Rham algebra of differential forms $`\mathrm{\Omega }^{}=\mathrm{\Omega }_{A/k}^{}`$. Frames of the form $`𝔤=\{\overline{\tau }_i,\varphi _i:=\omega _i\}`$ will be called natural.
If $`𝔤,𝔤^{}`$ are natural frames then formulas (6.1.2) and (6.1.3), together with the previous lemma, show that $`h_E^{ij}=h_\mathrm{\Omega }^{ij}`$ where $`h_\mathrm{\Omega }`$ is given by (4.1.5). (Of course one easily checks this directly.) This explains the notation for $`h_\mathrm{\Omega }^{ij}`$).
In other words, we arrive at an interesting conclusion.
6.4. Theorem. The matrices $`h=(h^{ij})`$ defined in 4.1 are equal to $`0`$ if $`E=\mathrm{\Omega }`$ and frames $`𝔤,𝔤^{}`$ are natural.
6.4.1. Warning. The functions $`h_{𝔤,𝔤^{}}`$ are nonzero since they are not linear.
6.5. On the other hand, Theorem 5.6 together Lemma 5.7 say that $`𝔞_E`$ and $`𝔟_E`$ are $`0`$ for $`E=\mathrm{\Omega }`$ (for natural frames $`𝔤,𝔤^{},𝔤^{\prime \prime }`$ of $`(A,\mathrm{\Omega })`$). This gives us a canonically defined chiral superalgebroid $`𝒜_\mathrm{\Omega }^{}`$. Its vertex envelope will be denoted $`D_\mathrm{\Omega }^{}^{ch}`$ and called the chiral algebra of differential operators on $`\mathrm{\Omega }^{}`$.
This implies
6.6. Theorem. For each smooth variety $`X`$ the construction 6.2 - 6.5 gives a canonically defined sheaf of chiral superalgebras $`𝒟_{\mathrm{\Omega }_X^{}}^{ch}`$. These Zariski sheaves form a sheaf in the étale topology.
6.7. The de Rham differential may be considered as an odd first order differential operator acting on $`\mathrm{\Omega }_X^{}`$ (and commuting with itself).
In coordinates, if $`𝔤=\{\overline{\tau }_i;\varphi _i:=\omega _i\}`$ is a natural frame of $`(A,\mathrm{\Omega })`$, it is given by
$$Q_𝔤^{cl}=\varphi _i\tau _i$$
$`(\mathrm{6.7.1})`$
($`cl`$ is for ”classical”). Let us check the independence of (6.7.1) on the choice of a frame. Let $`𝔤^{}=\{\overline{\tau }_i^{};\omega _i^{}\}`$ be another natural frame, $`\overline{\tau }_i^{}=g^{ij}\overline{\tau }_i;\omega _i^{}=g^{1pi}\omega _p`$ (cf. (3.2.4)).
Using (3.2.1) we have
$$Q_𝔤^{}^{cl}=\varphi _i^{}\tau _i^{}=g^{1pi}\varphi _p\{g^{iq}\tau _q+g^{irs}\varphi _s\psi _r\}$$
$`(\mathrm{6.7.2})`$
where
$$g^{ipq}=g^{ir}\tau _r(g^{sp})g^{1qs}=g^{sr}\tau _r(g^{ip})g^{1qs}=\tau _q(g^{ip})$$
$`(\mathrm{6.7.3})`$
(we have used (3.3.1)). So,
$$Q_𝔤^{}^{cl}=\varphi _p\tau _p+g^{1pi}g^{irs}\varphi _s\psi _r=Q_𝔤^{cl}+g^{1pi}\tau _s(g^{ir})\varphi _p\varphi _s\psi _r$$
$`(\mathrm{6.7.4})`$
Note that
$$g^{1pi}\tau _s(g^{ir})=\tau _s(g^{1pi})g^{ir}$$
which is symmetric under the permutation of $`p`$ with $`s`$, due to (3.3.3); therefore the second summand in (6.7.4) is zero, i.e. $`Q_𝔤^{cl}=Qcl_𝔤^{}`$.
Thus, $`Q^{cl}`$ is a correctly defined odd element of $`D_\mathrm{\Omega }^{}`$. It is obvious from (6.7.1) that $`[Q^{cl},Q^{cl}]=0`$.
6.8. Let us investigate the chiral counterpart of $`Q^{cl}`$. Let us define an odd element $`Q_𝔤`$ (of conformal weight $`1`$) of the vertex superalgebra $`D_{\mathrm{\Omega }^{};𝔤}^{ch}:=U𝒜_{\mathrm{\Omega }^{};𝔤}`$ by
$$Q_𝔤=\varphi _{i(1)}\tau _i$$
$`(\mathrm{6.8.1})`$
Let $`𝔤^{}`$ be another natural frame as in 6.7. Due to Theorem 6.4, the element $`Q_𝔤^{}`$ goes under the canonical isomorphism $`D_{\mathrm{\Omega }^{};𝔤^{}}^{ch}=D_{\mathrm{\Omega }^{};𝔤}^{ch}`$ to
$$Q_𝔤^{}=\varphi _{i(1)}^{}\tau _i^{}=\varphi _i^{}\tau _i^{}\gamma (\varphi _i^{},\tau _i^{})$$
$`(\mathrm{6.8.2})`$
cf. \[GII\], (3.3.1).
6.9. Lemma. We have (in $`D_{\mathrm{\Omega }^{};𝔤}`$)
$$\gamma (\varphi _i^{},\tau _i^{})=\left\{tr\left(\tau _r(g)g^1\right)\varphi _r\right\}$$
$`(\mathrm{6.9.1})`$
6.10. Before the proof, let us write down useful formulas
$$\gamma (a\varphi _r,b\tau _i)=\tau _i(a)\varphi _rb\tau _i(b)(a\varphi _r)$$
$`(\mathrm{6.10.1})`$
and
$$\gamma (a\varphi _r,b\varphi _s\psi _p)=\delta _{rp}a(b\varphi _s)+\delta _{sp}b(a\varphi _r)$$
$`(\mathrm{6.10.2})`$
6.11. Proof of 6.9. We have
$$\gamma (\varphi _i^{},\tau _i^{})=\gamma (g^{1qi}\varphi _q,g^{ip}\tau _p+g^{isr}\varphi _r\psi _s)$$
where
$$\gamma (g^{1qi}\varphi _q,g^{ip}\tau _p)=\tau _p(g^{1qi})\varphi _qg^{ip}\tau _p(g^{ip})(g^{1qi}\varphi _q)$$
$`(\mathrm{6.11.1})`$
and
$$\gamma (g^{1qi}\varphi _q,g^{isr}\varphi _r\psi _s)=g^{1qi}(g^{iqr}\varphi _r)+g^{irr}(g^{1qi})\varphi _q$$
$`(\mathrm{6.11.2})`$
Since $`g^{irr}=\tau _r(g^{ir})`$, the second summands in (6.11.1) and (6.11.2) cancel out. On the other hand, the first term in (6.11.1) is equal to
$$\tau _q(g^{1ri})\tau _s(g^{iq})\omega _s\varphi _r=\tau _r(g^{1si})\tau _s(g^{iq})\omega _s\varphi _r=$$
$$=g^{1qa}\tau _r(g^{ab})g^{1bi}\tau _s(g^{iq})\omega _s\varphi _r=\tau _r(g^{ab})\tau _s(g^{1ba})\omega _s\varphi _r=\tau _r(g^{ab})(g^{1ba})\varphi _r$$
Therefore
$$\gamma (\varphi _i^{},\tau _i^{})=\tau _r(g^{ab})(g^{1ba})\varphi _rg^{1ba}\left\{\tau _r(g^{ab})\varphi _r\right\}=\left\{\tau _r(g^{ab})g^{1ba}\varphi _r\right\},$$
QED.
6.12. From (6.8.1) we have $`Q_𝔤=\varphi _i\tau _i`$, and from 6.7 $`\varphi _i^{}\tau _i^{}=\varphi _i\tau _i`$. Therefore, (6.8.2) and Lemma 6.9 imply
6.13. Theorem. We have
$$Q_𝔤^{}=Q_𝔤+\left\{tr\left(\tau _r(g)g^1\right)\varphi _r\right\}$$
$`(\mathrm{6.13.1})`$
6.14. Consider the field $`Q_𝔤(z)`$ acting on the vertex algebra $`D_\mathrm{\Omega }^{}^{ch}`$. Due to (6.13.1), its zeroth component $`Q_{𝔤0}`$ does not depend on the choice of the frame $`𝔤`$. Therefore we get a canonical operator $`Q_0`$ acting on $`D_\mathrm{\Omega }^{}^{ch}`$.
Since it is a zeroth component of a field, it is a derivation of the vertex algebra, and it is obvious from the local definition (6.8.1) that $`[Q_0,Q_0]=0`$.
Consequently, for each smooth variety $`X`$ we get a canonical odd derivation $`Q_{0X}`$ of the sheaf $`𝒟_{\mathrm{\Omega }_X^{}}^{ch}`$, such that $`[Q_{0X},Q_{0X}]=0`$. The pair $`(𝒟_{\mathrm{\Omega }_X^{}}^{ch},Q_{0X})`$ is the chiral de Rham complex from \[MSV\].
Our Theorem 6.13 is a version of op. cit., (4.1c).
6.15. In the situation 6.4, consider an even element $`J_𝔤`$ of conformal weight $`1`$ of the algebra $`D_{\mathrm{\Omega }^{};𝔤}^{ch}`$, given by
$$J_𝔤=\varphi _{i(1)}\psi _i=\varphi _i\psi _i$$
$`(\mathrm{6.15.1})`$
After a change of frame as in loc. cit., we get an element
$$J_𝔤^{}=\varphi _{i(1)}^{}\psi _i^{}=\varphi _i^{}\psi _i^{}\gamma (\varphi _i^{},\psi _i^{})$$
$`(\mathrm{6.15.2})`$
where we have again used Theorem 6.4. We have
$$\varphi _i^{}\psi _i^{}=g^{1pi}\varphi _pg^{iq}\psi _q=\varphi _p\psi _p=J_𝔤$$
(see (3.2.3)). On the other hand, by (3.4.3)
$$\gamma (\varphi _i^{},\psi _i^{})=\gamma (g^{1pi}\varphi _p,g^{iq}\psi _q)=\delta _{pq}g^{1pi}g^{iq}=g^{1pi}g^{ip}=tr(g^1g)$$
Thus
$$J_𝔤^{}=J_𝔤tr(g^1g)$$
$`(\mathrm{6.15.3})`$
6.16. Consider an odd element $`G_𝔤`$ of conformal weight $`2`$ given by
$$G_𝔤=\psi _{i(1)}\omega _i$$
$`(\mathrm{6.16.1})`$
In the frame $`𝔤^{}`$,
$$G_𝔤^{}=(g^{iq}\psi _q)_{(1)}(g^{1si}\omega _s)$$
Note that $`\psi _{q(j)}a=0`$ for $`j0`$ (everything happens in $`D`$), hence it follows from the commutativity formula (1.3.1) that
$$a\psi _q=a_{(1)}\psi _q=\psi _{q(1)}a$$
$`(\mathrm{6.16.2})`$
Therefore by ”associativity” (1.2.5)
$$(a\psi _q)_{(1)}(b\omega _s)=(\psi _{q(1)}a)_{(1)}(b\omega _s)=\psi _{q(1)}a_{(1)}b\omega _s=\psi _{q(1)}(ab\omega _s)$$
$`(\mathrm{6.16.3})`$
Therefore
$$G_𝔤^{}=\psi _{q(1)}g^{iq}g^{1si}\omega _s=\psi _{s(1)}\omega _s=G_𝔤$$
$`(\mathrm{6.16.6})`$
6.17. Let us investigate the Virasoro element. Define an even element $`L_𝔤`$ of conformal weight $`2`$ by
$$L_𝔤=L_{(b)𝔤}+L_{(f)𝔤}$$
$`(\mathrm{6.17.1})`$
where
$$L_{(b)𝔤}=\omega _{i(1)}\tau _i$$
$`(\mathrm{6.17.2})`$
($`(b)`$ is for ”bosonic”) and
$$L_{(f)𝔤}=\rho _{i(1)}\psi _i$$
$`(\mathrm{6.17.3})`$
($`(f)`$ is for ”fermionic”), cf. \[MSV\], (2.3a).
6.18. We have
$$(a\omega _s)_{(1)}(b\tau _p)=\omega _{s(1)}\left\{ab\tau _p+\tau _p(a)b+\tau _p(b)a\right\}\omega _{s(1)}b\tau _p(a)$$
$`(\mathrm{6.18.1})`$
Indeed, it follows from ”associativity” (1.2.5) that
$$(a\omega _s)_{(1)}(b\tau _p)=(\omega _{s(1)}a)_{(1)}(b\tau _p)=\omega _{s(1)}a_{(1)}(b\tau _p)+\omega _{s(2)}a_{(0)}(b\tau _p)+$$
$$+a_{(2)}\omega _{s(0)}(b\tau _p)$$
Next,
$$a_{(1)}(b\tau _p)=ab\tau _p\gamma (a,b\tau _p)=ab\tau _p+\tau _p(a)b+\tau _p(b)a$$
(see \[GII\] (3.3.1));
$$a_{(0)}(b\tau _p)=b\tau _{p(0)}a=b\tau _p(a)$$
(see \[GII\] (3.3.2)). Finally
$$\omega _{s(0)}(b\tau _p)=(b\tau _p)_{(0)}\omega _s+b\tau _p,\omega _s$$
where
$$(b\tau _p)_{(0)}\omega _s=(b\tau _p)(\omega _s)=\tau _p,\omega _sb=\delta _{ps}b$$
by (1.1.3), since
$$\tau _p(\omega _s)=0,$$
$`(\mathrm{6.18.2})`$
and $`b\tau _p,\omega _s=b\delta _{ps}`$. This implies
$$\omega _{s(0)}(b\tau _p)=0$$
$`(\mathrm{6.18.3})`$
Formula (6.18.1) follows.
6.19. We have
$$(a\omega _s)_{(1)}(b\varphi _\alpha \psi _\beta )=\omega _{s(1)}\left\{ab\varphi _\alpha \psi _\beta \delta _{\alpha \beta }ba\right\}$$
$`(\mathrm{6.19.1})`$
Indeed, by commutativity and ”associativity” (1.2.5)
$$(a\omega _s)_{(1)}(b\varphi _\alpha \psi _\beta )=(\omega _{s(1)}a)_{(1)}(b\varphi _\alpha \psi _\beta )=\omega _{s(1)}a_{(1)}(b\varphi _\alpha \psi _\beta )$$
On the other hand
$$a_{(1)}(b\varphi _\alpha \psi _\beta )=ab\varphi _\alpha \psi _\beta \gamma (a,b\varphi _\alpha \psi _\beta )=ab\varphi _\alpha \psi _\beta \delta _{\alpha \beta }ba,$$
see (3.4.2). This implies (6.19.1).
6.20. We have
$$L_{(b)𝔤^{}}=L_{(b)𝔤}+\omega _{s(1)}\left\{\tau _p(g^{1si})g^{ip}+g^{1si}\tau _\beta (g^{i\alpha })\varphi _\beta \psi _\alpha \right\}$$
$$\omega _{s(2)}g^{ip}\tau _p(g^{1si})$$
$`(\mathrm{6.20.1})`$
Indeed, due to Theorem 6.4
$$L_{(b)𝔤^{}}=(g^{1si}\omega _s)_{(1)}\left\{g^{ip}\tau _p+g^{i\alpha \beta }\varphi _\beta \psi _\alpha \right\}$$
According to (6.18.1)
$$(g^{1si}\omega _s)_{(1)}(g^{ip}\tau _p)=\omega _{s(1)}\{g^{1si}g^{ip}\tau _p+\tau _p(g^{1si})g^{ip}+\tau _p(g^{ip})g^{1si}\}$$
$$\omega _{s(2)}g^{ip}\tau _p(g^{1si})$$
By (6.18.2)
$$(g^{1si}\omega _s)_{(1)}(g^{i\alpha \beta }\varphi _\beta \psi _\alpha )=(g^{1si}\omega _s)_{(1)}(\tau _\beta (g^{i\alpha })\varphi _\beta \psi _\alpha )=$$
$$=\omega _{s(1)}\left\{g^{1si}\tau _\beta (g^{i\alpha })\varphi _\beta \psi _\alpha \delta _{\alpha \beta }\tau _\beta (g^{i\alpha })g^{1si}\right\}$$
The third term in the first expression cancels out the second term in the second one, and we get (6.20.1).
6.21. We have
$$(a\rho _\mu )_{(1)}(b\psi _\nu )=\rho _{\mu (1)}ab\psi _\nu $$
$`(\mathrm{6.21.1})`$
Indeed,
$$(a\rho _\mu )_{(1)}(b\psi _\nu )=(\rho _{\mu (1)}a)_{(1)}(b\psi _\nu )=\rho _{\mu (1)}ab\psi _\nu +a_{(2)}\rho _{\mu (0)}b\psi _\nu $$
and by commutativity
$$\rho _{\mu (0)}b\psi _\nu =(b\psi _\nu )(\rho _\mu )\left\{(b\psi _\nu )_{(1)}\rho _\mu \right\}=\delta _{\nu \mu }b\delta _{\nu \mu }b=0,$$
cf. (6.18.3). This implies (6.21.1).
6.22. We have
$$(a\varphi _\gamma \omega _i)_{(1)}(b\psi _\nu )=\omega _{i(1)}\left\{ab\varphi _\gamma \psi _\nu \delta _{\gamma \nu }ab\right\}+\delta _{\nu \gamma }\omega _{i(2)}ab$$
$`(\mathrm{6.22.1})`$
Indeed,
$$(a\varphi _\gamma \omega _i)_{(1)}(b\psi _\nu )=(\omega _{i(1)}a\varphi _\gamma )_{(1)}(b\psi _\nu )=\omega _{i(1)}(a\varphi _\gamma )_{(1)}b\psi _\nu +\omega _{i(2)}(a\varphi _\gamma )_{(0)}b\psi _\nu $$
where
$$(a\varphi _\gamma )_{(1)}b\psi _\nu =a\varphi _\gamma b\psi _\nu \gamma (a\varphi _\gamma ,b\psi _\nu )=ab\varphi _\gamma \psi _\nu \delta _{\gamma \nu }ab,$$
cf. (3.4.3), and
$$(a\varphi _\gamma )_{(0)}b\psi _\nu =(b\psi _\nu )(a\varphi _\gamma )=ab\delta _{\nu \gamma }$$
6.23. We have
$$L_{(f)𝔤^{}}=L_{(f)𝔤}+\omega _{i(1)}\left\{\tau _i(g^{1\gamma \alpha })g^{\alpha \nu }\varphi _\gamma \psi _\nu \tau _i(g^{1\gamma \alpha })g^{\alpha \gamma }\right\}+\omega _{i(2)}\tau _i(g^{1\gamma \alpha })g^{\alpha \gamma }$$
$`(\mathrm{6.23.1})`$
Indeed,
$$L_{(f)𝔤^{}}=\rho _{\alpha (1)}^{}\psi _\alpha ^{}=\left\{g^{1\mu \alpha }\rho _\mu +\tau _i(g^{1\gamma \alpha })\varphi _\gamma \omega _i\right\}_{(1)}(g^{\alpha \nu }\psi _\nu )$$
By (6.21.1)
$$(g^{1\mu \alpha }\rho _\mu )_{(1)}(g^{\alpha \nu }\psi _\nu )=\rho _{\mu (1)}g^{1\mu \alpha }g^{\alpha \nu }\psi _\nu =L_{(f)𝔤}$$
and by (6.22.1)
$$(\tau _i(g^{1\gamma \alpha })\varphi _\gamma \omega _i)_{(1)}(g^{\alpha \nu }\psi _\nu )=\omega _{i(1)}\left\{\tau _i(g^{1\gamma \alpha })g^{\alpha \nu }\varphi _\gamma \psi _\nu \tau _i(g^{1\gamma \alpha })g^{\alpha \gamma }\right\}+$$
$$+\omega _{i(2)}\tau _i(g^{1\gamma \alpha })g^{\alpha \gamma }$$
6.24. Comparing (6.20.1) and (6.23.1) we see easily that
$$L_𝔤^{}=L_{(b)𝔤^{}}+L_{(f)𝔤^{}}=L_{(b)𝔤}+L_{(f)𝔤}=L_𝔤$$
$`(\mathrm{6.24.1})`$
Let us collect our computations of transformation rules.
6.25. Theorem. Let $`𝔤,𝔤^{}`$ be two natural frames of $`(A,\mathrm{\Omega })`$. Consider $`4`$ elements of the vertex superalgebra $`D_{\mathrm{\Omega }^{};𝔤}^{ch}`$ given by
$$Q_𝔤=\varphi _{i(1)}\tau _i$$
$`(\mathrm{6.25.1})`$
(an odd element of conformal weight $`1`$)
$$J_𝔤=\varphi _{i(1)}\psi _i$$
$`(\mathrm{6.25.2})`$
(an even element of conformal weight $`1`$)
$$G_𝔤=\psi _{i(1)}\omega _i$$
$`(\mathrm{6.25.3})`$
(an odd element of conformal weight $`2`$) and
$$L_𝔤=\omega _{i(1)}\tau _i+\rho _{i(1)}\psi _i$$
$`(\mathrm{6.25.4})`$
(an even element of conformal weight $`2`$).
After the canonical identification $`D_{\mathrm{\Omega }^{};𝔤^{}}^{ch}=D_{\mathrm{\Omega }^{};𝔤}^{ch}`$ these elements are transformed as follows
$$Q_𝔤^{}=Q_𝔤+\left\{tr\left(g^1\tau _r(g)\right)\varphi _r\right\}$$
$`(\mathrm{6.25.5})`$
$$J_𝔤^{}=J_𝔤tr(g^1g)$$
$`(\mathrm{6.25.6})`$
$$G_𝔤^{}=G_𝔤$$
$`(\mathrm{6.25.7})`$
and
$$L_𝔤^{}=L_𝔤$$
$`(\mathrm{6.25.8})`$
This is a version of \[MSV\], Theorem 4.2.
§7. Poincaré-Birkhoff-Witt
7.1 Let $`X`$ be a smooth variety and $`𝒟_{\mathrm{\Omega }_X^{}}^{ch}`$ be the sheaf discussed in the previous section, cf. Theorem 6.6. It is a sheaf of $`_0`$-graded vertex algebras, so
$$𝒟_{\mathrm{\Omega }_X^{}}^{ch}=_{n_0}𝒟_{\mathrm{\Omega }_X^{};n}^{ch}$$
$`(\mathrm{7.1.1})`$
where $`𝒟_{\mathrm{\Omega }_X^{};n}^{ch}`$ denotes the component of conformal weight $`n`$.
According to Theorem 6.25 (see (6.25.8)) we have a canonical global section $`L`$ of $`𝒟_{\mathrm{\Omega }_X^{};2}^{ch}`$. Let $`L(z)=L_nz^{n2}`$ be the corresponding field.
7.1.1. Claim. A local section $`\alpha 𝒟_{\mathrm{\Omega }_X^{}}^{ch}`$ belongs to $`𝒟_{\mathrm{\Omega }_X^{};n}^{ch}`$ if and only if $`L_0(\alpha )=n\alpha `$.
In a uniform notation $`L(z)=L_{(n)}z^{n1}`$ we have $`L_0=L_{(1)}`$. We shall check 7.1.1 simultaneously with
7.1.2. Claim. The operator $`L_1=L_{(0)}`$ coincides with the canonical derivation $``$ of the vertex algebra $`𝒟_{\mathrm{\Omega }_X^{}}^{ch}`$.
Both statements are local, so we may assume we are in the local situation 6.17. Note that the operator $`L_{(0)}`$ is a derivation with respect to the operation <sub>(-1)</sub>:
$$L_{(0)}(x_{(1)}y)=(L_{(0)}x)_{(1)}y+x_{(1)}L_{(0)}y$$
$`(\mathrm{7.1.2})`$
Therefore it suffices to check 7.1.2 on the generators $`a,\tau _i,\omega _i,\psi _i,\rho _i`$ of the vertex algebroid $`𝒜_{\mathrm{\Omega }^{};𝔤}`$, which is done by a simple explicit computation.
It follows from the associativity formula (1.2.4) that
$$L_{(1)}y_{(1)}z=(L_{(1)}y)_{(1)}z+y_{(1)}L_{(1)}z+L_{(0)}y_{(0)}zy_{(0)}L_{(0)}z=$$
$$=(L_{(1)}y)_{(1)}z+y_{(1)}L_{(1)}z$$
$`(\mathrm{7.1.3})`$
since
$$L_{(0)}y_{(0)}zy_{(0)}L_{(0)}z=(y_{(0)}z)y_{(0)}z=0$$
In other words, $`L_{(1)}`$ is a derivation of <sub>(-1)</sub>. Therefore it suffices to check 7.1.1 on the generators of $`𝒜_{\mathrm{\Omega }^{};𝔤}`$ as above, which is straightforward.
7.2. In the local situation 6.1, consider the local algebra $`D_{\mathrm{\Omega }^{};𝔤}^{ch}=U𝒜_{\mathrm{\Omega }^{};𝔤}`$. Let us introduce a $``$-grading
$$D_{\mathrm{\Omega }^{};𝔤}^{ch}=_{pZ}D_{\mathrm{\Omega }^{};𝔤}^{ch;p}$$
$`(\mathrm{7.2.1})`$
to be called fermionic charge. For an element $`xD_{\mathrm{\Omega }^{};𝔤}^{ch}`$ let us denote its fermionic charge (to be defined) by $`F(x)`$. It is defined uniquely by the following conditions:
(a) $`F(a)=F(\tau _i)=F(\omega _i)=0;F(\varphi _i)=F(\rho _i)=F(\psi _i)=1`$;
(b) $`F(x_{(1)}y)=F(x)+F(y)`$.
Due to the transformation formulas (3.2.1) - (3.2.5) this grading is obviously preserved under a change of frames. Therefore in the situation of 7.1 the sheaf $`𝒟_{\mathrm{\Omega }_X^{}}^{ch}`$ gets a canonical $``$-grading
$$𝒟_{\mathrm{\Omega }_X^{}}^{ch}=_p𝒟_{\mathrm{\Omega }_X^{}}^{ch;p}$$
$`(\mathrm{7.2.2})`$
Note that parity is equal to fermionic charge modulo $`2`$.
Here is another way to define the grading (7.2.2). First notice a simple
7.2.1. Lemma. Let $`𝒜=(A,T,\mathrm{\Omega },,\mathrm{})`$ be a vertex (super)algebroid. For every invertible element $`aA`$ the operator $`(a^1a)_{(0)}`$ acting on $`U𝒜`$ is trivial.
Proof. Obviously this operator is trivial on $`A=U𝒜_0`$. Let $`xU𝒜_1`$ and $`\tau T`$ be its image under the canonical projection $`U𝒜_1T`$. We have
$$(a^1a)_{(0)}x=x_{(0)}a^1a+(x_{(1)}a^1a)=\tau (a^1a)+\tau ,a^1a=$$
$$=a^2\tau (a)aa^1\tau (a)+(a^1\tau (a))=0,$$
so $`(a^1a)_{(0)}`$ is trivial on $`U𝒜_1`$. Therefore it is trivial on the whole algebra $`U𝒜`$ since $`(\mathrm{?})_{(0)}`$ is a derivation of the operation <sub>(-1)</sub>. $``$
Applying this lemma to $`a=det(g)`$ in the formula (6.25.6) we see that the component $`J_{0;𝔤}`$ of the field $`J_𝔤(z)=J_{n;𝔤}z^{n1}`$ is preserved under the change of frames. Consequently it gives rise to a well defined endomorphism $`J_0`$ of the sheaf $`𝒟_{\mathrm{\Omega }_X^{}}^{ch}`$.
7.2.2. Claim. A local section $`\alpha 𝒟_{\mathrm{\Omega }_X^{}}^{ch}`$ belongs to $`𝒟_{\mathrm{\Omega }_X^{}}^{ch;p}`$ if and only if $`J_0(\alpha )=p\alpha `$.
Indeed, the function $`F(\alpha )`$ defined by $`J_0(\alpha )=F(\alpha )\alpha `$ obviously satisfies the condtions (a) and (b) above.
7.3. The two gradings (7.1.1) and (7.2.2) are compatible: if we denote
$$𝒟_{\mathrm{\Omega }_X^{};n}^{ch;p}:=𝒟_{\mathrm{\Omega }_X^{};n}^{ch}𝒟_{\mathrm{\Omega }_X^{}}^{ch;p}$$
$`(\mathrm{7.3.1})`$
then
$$𝒟_{\mathrm{\Omega }_X^{}}^{ch}=_{n_0;p}𝒟_{\mathrm{\Omega }_X^{};n}^{ch;p}$$
$`(\mathrm{7.3.2})`$
For a fixed $`n`$, only a finite number of sheaves $`𝒟_{\mathrm{\Omega }_X^{};n}^{ch;p}`$ are nonzero.
If the ground ring $`k`$ is a field of characteristic $`0`$ then the sheaves $`𝒟_{\mathrm{\Omega }_X^{};n}^{ch;p}`$ and $`𝒟_{\mathrm{\Omega }_X^{};n}^{ch;np}`$ are in a certain sense dual to each other, see \[MS\].
7.4. Starting from this point we assume that $`k`$. According to a (superversion of) the PBW theorem, \[GII\], Theorem 9.18, the sheaf $`𝒟_{\mathrm{\Omega }_X^{}}^{ch}`$ admits a canonical filtration such that the associated graded sheaf is canonically isomorphic to
$$gr(𝒟_{\mathrm{\Omega }_X^{}}^{ch})=Sym_{\mathrm{\Omega }_X^{}}\left\{\left(_{n1}\mathrm{\Theta }_{\mathrm{\Omega }_X^{}(n)}\right)\left(_{n1}\mathrm{\Omega }_{\mathrm{\Omega }_X^{}(n)}^1\right)\right\}$$
$`(\mathrm{7.4.1})`$
Here $`\mathrm{\Theta }_{\mathrm{\Omega }_X^{}}`$ (resp. $`\mathrm{\Omega }_{\mathrm{\Omega }_X^{}}^1`$) denotes the tangent (resp. the cotangent) bundle of the supervariety $`(X,\mathrm{\Omega }_X^{})`$, and $`(\mathrm{?})_{(n)}`$ means that this bundle is put into the conformal weight $`n`$.
The endomorphisms $`L_0`$ and $`J_0`$ respect the canonical filtration; hence we get a canonical finite filtration $`F_{}`$ on each homogeneous component $`𝒟_{\mathrm{\Omega }_X^{};n}^{ch;p}`$. The graded quotients $`F_i𝒟_{\mathrm{\Omega }_X^{};n}^{ch;p}/F_{i+1}𝒟_{\mathrm{\Omega }_X^{};n}^{ch;p}`$ are locally free $`𝒪_X`$-modules of finite rank (we shall see this in the course of computations below). This allows us to introduce the elements of the Grothendieck group $`K(X)`$ of vector bundles
$$[𝒟_{\mathrm{\Omega }_X^{};n}^{ch;p}]:=\underset{i}{}[F_i𝒟_{\mathrm{\Omega }_X^{};n}^{ch;p}/F_{i+1}𝒟_{\mathrm{\Omega }_X^{};n}^{ch;p}]K(X)$$
$`(\mathrm{7.4.2})`$
Here $`[E]`$ in the right hand side denotes the class of a vector bundle $`E`$ in $`K(X)`$. Consider the generating function
$$cl(𝒟_{\mathrm{\Omega }_X^{}}^{ch})(y,q):=\underset{p,n}{}[𝒟_{\mathrm{\Omega }_X^{};n}^{ch;p}]y^pq^nK(X)[y,y^1][[q]]$$
$`(\mathrm{7.4.3})`$
7.5. For a vector bundle $`E`$ over $`X`$ and an indeterminate $`x`$ we introduce the notations
$$[S_xE]=\underset{i=0}{\overset{\mathrm{}}{}}[Sym_{𝒪_X}^iE]K(X)[[x]]$$
$`(\mathrm{7.5.1})`$
and
$$[\mathrm{\Lambda }_xE]=\underset{i=0}{\overset{\mathrm{}}{}}[\mathrm{\Lambda }_{𝒪_X}^iE]K(X)[x]$$
$`(\mathrm{7.5.2})`$
The following fact was noticed in \[BL\] (cf. also \[W\]).
7.6. Theorem (L. Borisov - A. Libgober) We have
$$cl(𝒟_{\mathrm{\Omega }_X^{}}^{ch})(y,q)=[\mathrm{\Lambda }_y\mathrm{\Omega }_X^1]\underset{n=1}{\overset{\mathrm{}}{}}\left\{[S_{q^n}\mathrm{\Theta }_X][S_{q^n}\mathrm{\Omega }_X^1][\mathrm{\Lambda }_{y^1q^n}\mathrm{\Theta }_X][\mathrm{\Lambda }_{yq^n}\mathrm{\Omega }_X^1]\right\}$$
$`(\mathrm{7.6.1})`$
7.7. Proof. Let us understand the bundles $`\mathrm{\Theta }_{\mathrm{\Omega }_X^{}}`$ and $`\mathrm{\Omega }_{\mathrm{\Omega }_X^{}}^1`$ a little bit more attentively.
Let us consider the local situation 3.1, with $`E=\mathrm{\Omega }`$, so that $`\mathrm{\Lambda }E=\mathrm{\Omega }^{}`$. All our frames $`𝔤`$ will be natural. Let $`T_\psi T_\mathrm{\Omega }^{}`$ be the $`A`$-submodule with the base $`\{\psi _i\}`$. The coordinate change formula (3.2.3) shows that it is a well defined $`A`$-submodule of $`T`$ independent on the choice of a frame, canonically isomorphic to $`T`$.
We set
$$T_{\psi \mathrm{\Omega }^{}}=\mathrm{\Omega }^{}_AT_\psi T_\mathrm{\Omega }^{}$$
We denote by $`T_{\tau \mathrm{\Omega }^{}}`$ the quotient $`\mathrm{\Omega }^{}`$-module $`T_\mathrm{\Omega }^{}/T_{\psi \mathrm{\Omega }^{}}`$. Let $`T_\tau T_{\tau \mathrm{\Omega }^{}}`$ be the $`A`$-submodule generated by all $`\tau _i`$. The formula (3.2.1) shows that $`T_\tau `$ is a well defined $`A`$-module canonically isomorphic to $`T`$, and we have
$$T_{\tau \mathrm{\Omega }^{}}=\mathrm{\Omega }^{}_AT_\tau $$
Returning to our variety $`X`$, we see that we get two vector bundles $`\mathrm{\Theta }_\psi `$ and $`\mathrm{\Theta }_\tau `$ both isomorphic to $`\mathrm{\Theta }_X`$ and a canonical short exact sequence
$$0\mathrm{\Theta }_{\psi \mathrm{\Omega }_X^{}}\mathrm{\Theta }_{\mathrm{\Omega }_X^{}}\mathrm{\Theta }_{\tau \mathrm{\Omega }_X^{}}0$$
$`(\mathrm{7.7.1})`$
with
$$\mathrm{\Theta }_{\psi \mathrm{\Omega }_X^{}}=\mathrm{\Omega }_X^{}_{𝒪_X}\mathrm{\Theta }_\psi ;\mathrm{\Theta }_{\tau \mathrm{\Omega }_X^{}}=\mathrm{\Omega }_X^{}_{𝒪_X}\mathrm{\Theta }_\tau $$
$`(\mathrm{7.7.2})`$
Note that $`\mathrm{\Theta }_\psi `$ has fermionic charge $`1`$ and $`\mathrm{\Theta }_\tau `$ has fermionic charge $`0`$.
Dually, we have two vector bundles $`\mathrm{\Omega }_\rho ^1`$ and $`\mathrm{\Omega }_\omega ^1`$ both isomorphic to $`\mathrm{\Omega }_X^1`$ and a canonical short exact sequence
$$0\mathrm{\Omega }_{\omega \mathrm{\Omega }_X^{}}^1\mathrm{\Omega }_{\mathrm{\Omega }_X^{}}^1\mathrm{\Omega }_{\rho \mathrm{\Omega }_X^{}}^10$$
$`(\mathrm{7.7.3})`$
with
$$\mathrm{\Omega }_{\rho \mathrm{\Omega }_X^{}}^1=\mathrm{\Omega }_X^{}_{𝒪_X}\mathrm{\Omega }_\rho ^1;\mathrm{\Omega }_{\omega \mathrm{\Omega }_X^{}}^1=\mathrm{\Omega }_X^{}_{𝒪_X}\mathrm{\Omega }_\omega ^1$$
$`(\mathrm{7.7.4})`$
The bundles $`\mathrm{\Omega }_\rho ^1`$, $`\mathrm{\Omega }_\omega ^1`$ have fermionic charges $`1`$, $`0`$ respectively.
Note that if $`E`$ is a vector bundle then
$$Sym_{\mathrm{\Omega }_X^{}}(\mathrm{\Omega }_X^{}_{𝒪_X}E)=\mathrm{\Omega }_X^{}_{𝒪_X}Sym_{𝒪_X}E$$
$`(\mathrm{7.7.5})`$
Returning to PBW formula (7.4.1) we see that these remarks imply (7.6.1). $``$
7.8. Starting from this point let us assume that $`k=`$. Consider the formal power series
$$\theta (y,q)=i^1(y^{1/2}y^{1/2})q^{1/8}\underset{n=1}{\overset{\mathrm{}}{}}\left\{(1q^n)(1yq^n)(1y^1q^n)\right\}$$
$`(\mathrm{7.8.1})`$
It is nothing but the theta function $`\theta _1(h,z)`$ as defined in \[HC\], II, 2, §10, formula (3), p. 204, with $`q=h^2`$ and $`y=z^2`$.
If $`fGL(V)`$ is an automorphism of a $`d`$-dimendional vector space $`V`$ with eigenvalues $`\lambda _1,\mathrm{},\lambda _d`$, we shall denote by $`\theta _f(y,q)`$ the power series
$$\theta _f(y,q)=\frac{_{i=1}^d\theta (\lambda _iy,q)}{_{i=1}^d\theta (\lambda _i,q)}$$
$`(\mathrm{7.8.2})`$
Let $`X`$ be a proper smooth $`d`$-dimensional algebraic variety; let $`g:XX`$ be a simple automorphism, which means by definition that the graph $`\mathrm{\Gamma }_gX\times X`$ is transversal to the diagonal. This implies that the set $`X^g`$ if its fixed points is finite.
For each $`xX^g`$ denote by $`g_x`$ the induced endomorphism of the cotangent space $`\mathrm{\Omega }_{X;x}^1`$. All eigenvalues of $`g_x`$ are distinct from $`1`$.
7.9. Theorem. Consider the power series
$$T_{X;g}(y,q):=y^{d/2}\underset{a,b,n}{}(1)^{a+b}Tr(g;H^a(X;𝒟_{\mathrm{\Omega }_X^{};n}^{ch;b}))y^bq^n$$
$`(\mathrm{7.9.1})`$
We have
$$T_{X;g}(y,q)=\underset{xX^g}{}\theta _{g_x}(y,q)$$
$`(\mathrm{7.9.2})`$
7.10. Proof. Recall that according to the Atiyah-Bott holomorphic Lefschetz fixed point formula, if $`E`$ is a $`g`$-equivariant vector bundle over $`X`$ then
$$\underset{i}{}(1)^iTr(g;H^i(X;E))=\underset{xX^g}{}\frac{Tr(g;E_x)}{det(1g_x)}$$
$`(\mathrm{7.10.1})`$
see \[AB\], Theorem 4.12. Note that if $`(V,f)`$ are as in 7.8 then
$$Tr(g;Sym_x(V))=\underset{i=1}{\overset{d}{}}(1\lambda _ix)^1=Tr(g;\mathrm{\Lambda }_x(V))^1$$
$`(\mathrm{7.10.2})`$
The proof of 7.6 shows that each sheaf $`𝒟_{\mathrm{\Omega }_X^{};n}^{ch;p}`$ carries a canonical filtration whose quotients are vector bundles and the associated graded sheaf is given by the formula (7.6.1). Therefore we may apply the Lefschetz formula (7.10.1).
Note that since in the expression (7.9.1) the fermionic charge $`a+b`$ is taken into account, we should apply the Lefschetz formula to the element $`cl(𝒟_{\mathrm{\Omega }_X^{}}^{ch})(y,q)`$. Due to (7.10.2) each fixed point $`x`$ gives a contribution
$$y^{d/2}\underset{i=1}{\overset{d}{}}\left\{\frac{1\lambda _iy}{1\lambda _i}\underset{n=1}{\overset{\mathrm{}}{}}\frac{(1\lambda _iyq^n)(1\lambda _i^1y^1q^n)}{(1\lambda _iq^n)(1\lambda _i^1q^n)}\right\}=\theta _{g_x}(y,q)$$
where $`\lambda _i`$ are the eigenvalues of $`g_x`$. This implies the theorem. $``$.
The reader may wish to compare (7.9.2) with the explicit formulas for the trace of certain automorphisms of the Frenkel-Lepowsky-Meurman Monster vertex algebra, cf. \[FLM\].
References
\[AB\] M.F. Atiyah, R. Bott, A Lefschetz fixed point formula for elliptic complexes: II. Applications, Ann. Math., 88, No. 3 (1968), 451-491.
\[BL\] L. Borisov, A. Libgober, Elliptic genera and applications to mirror symmetry, math.AG/9904126.
\[FLM\] I. Frenkel, J. Lepowsky, A. Meurman, Vertex operator algebras and the Monster, Pure and Applied Mathematics, 134, Academic Press, Boston, 1988.
\[GI\] V. Gorbounov, F. Malikov, V. Schechtman, Gerbes of chiral differential operators, math.AG/9906117; Math. Research Letters, 7, 1-12 (2000).
\[GII\] V. Gorbounov, F. Malikov, V. Schechtman, Gerbes of chiral differential operators. II, math.AG/0003170.
\[HC\] A. Hurwitz, R. Courant, Funktionentheorie, Vierte Auflage, Springer-Verlag, Berlin-Göttingen-Heidelberg-New York, 1964.
\[K\] V. Kac, Vertex algebras for beginners, Second Edition, University Lecture Series, 10, American Mathematical Society, Providence, Rhode Island, 1998.
\[MSV\] F. Malikov, V. Schechtman, A. Vaintrob, Chiral de Rham complex, Comm. Math. Phys., 204 (1999), 439-473.
\[MS\] F. Malikov, V. Schechtman, Chiral Poincaré duality, Math. Research Letters, 6 (1999), 533-546.
\[W\] E. Witten, Elliptic genera and quantum field theory, Comm. Math. Phys. 109 (1987), 525-536.
V.G.: Department of Mathematics, University of Kentucky, Lexington, KY 40506, USA; vgorbms.uky.edu
F.M.: Department of Mathematics, University of Southern California, Los Angeles, CA 90089, USA; fmalikovmathj.usc.edu
V.S.: IHES, 35 Route de Chartres, 91440 Bures-sur-Yvette, France; vadikihes.fr
|
warning/0005/hep-ph0005021.html
|
ar5iv
|
text
|
# UT-STPD-3/00 FTUAM 00-09 Cold Dark Matter and 𝑏→𝑠𝛾 in the Hořava-Witten Theory
## Abstract
The minimal supersymmetric standard model with complete, partial or no Yukawa unification and radiative electroweak breaking with boundary conditions from the Hořava-Witten theory is considered. The parameters are restricted by constraining the lightest sparticle relic abundance by cold dark matter considerations and requiring the $`b`$-quark mass after supersymmetric corrections and the branching ratio of $`bs\gamma `$ to be compatible with data. Complete Yukawa unification can be excluded. Also, $`tb`$ Yukawa unification is strongly disfavored since it requires almost degenerate lightest and next-to-lightest sparticle masses. However, the $`b\tau `$ or no Yukawa unification cases avoid this degeneracy. The latter with $`\mu <0`$ is the most natural case. The lightest sparticle mass, in this case, can be as low as about $`77\mathrm{GeV}`$.
Recently, it has been realized that the five existing perturbative string theories (type I open strings, type IIA and IIB closed strings, and the $`E_8\times E_8^{}`$ and $`SO(32)`$ closed heterotic strings) and the 11-dimensional supergravity correspond to different vacua of a unique underlying theory, called M-theory. Hořava and Witten have shown that the strong coupling limit of the $`E_8\times E_8^{}`$ heterotic string theory is equivalent to the low energy limit of M-theory compactified on $`S^1/Z_2`$ which is a line segment of length $`\rho `$. As $`\rho 0`$, the weakly coupled heterotic string is recovered. The observable $`E_8`$ gauge fields reside in one (10-dimensional) end of this segment, while the hidden sector $`E_8^{}`$ gauge fields reside in its other end. Gravitational fields propagate in the 11-dimensional bulk.
The main success of the Hořava-Witten theory is that it solves, in an elegant way, the gauge coupling unification problem, i.e., the discrepancy between the supersymmetric (SUSY) grand unified theory (GUT) scale $`M_X2\times 10^{16}\mathrm{GeV}`$ (consistent with the data on the low energy gauge coupling constants) and the string unification scale $`M_{str}5\times 10^{17}\mathrm{GeV}`$ calculated in the weakly coupled string theory. Before M-theory, there were several proposals (such as large threshold corrections, intermediate scales, and extra particles) for explaining this discrepancy but none was totally satisfactory. In the strongly coupled heterotic string theory, the extra Kaluza-Klein states do not affect the running of the gauge coupling constants, which live on the boundary of the 11-dimensional spacetime. On the contrary, they accelerate the running of the gravitational coupling constant and, thus, reduce $`M_{str}`$ to $`M_X`$. Moreover, SUSY breaking in M-theory naturally leads to gaugino masses of the order of the gravitino mass in contrast to the weakly coupled heterotic string case where the gaugino masses were tiny.
Similarly to the weakly coupled heterotic string, the compactification of the Hořava-Witten theory can lead to the spontaneous breaking of $`E_8`$ to phenomenologically more interesting groups. The simplest breaking of $`E_8`$ to $`E_6`$ is achieved by the so-called standard embedding (SE), where the holonomy group of the spin connection of a Calabi-Yau three-fold is identified with a $`SU(3)`$ subgroup of $`E_8`$. Further breaking of $`E_6`$ to semi-simple groups such as the trinification group $`SU(3)_c\times SU(3)_L\times SU(3)_R`$ and the flipped $`SU(6)\times U(1)`$ group can be performed via Wilson loops. The trinification group contains $`SU(2)_R`$. Assuming then that the Higgs doublets and the third family right-handed quarks form $`SU(2)_R`$ doublets, one obtains the ‘asymptotic’ Yukawa coupling relation $`h_t=h_b`$ and, hence, large $`\mathrm{tan}\beta m_t/m_b`$. The flipped $`SU(6)`$, for certain embeddings of the minimal supersymmetric standard model (MSSM) fields, contains $`SU(4)_c`$. Requiring that the third family lepton doublet belongs to $`SU(6)`$ 15-plets and the right-handed $`b`$-quark as well as the Higgs doublet coupling to the down-type quarks belong to $`SU(6)`$ $`\overline{6}`$-plets, one gets ‘asymptotic’ $`b\tau `$ Yukawa unification ($`h_b=h_\tau `$).
In the strongly coupled case, the SE is not special . Non-standard embeddings (NSE) may lead to simple gauge groups such as $`SU(5)`$ or $`SO(10)`$ which could yield $`b\tau `$ or complete ($`h_t=h_b=h_\tau `$) Yukawa unification. However, in general, we do not obtain Higgs superfields in the adjoint representation. Further gauge symmetry breaking then requires Wilson loops and, thus, (partial) Yukawa unification is lost. This may be avoided by employing special constructions with higher Kac-Moody level . Complete Yukawa unification can be obtained in the Pati-Salam gauge group $`SU(4)_c\times SU(2)_L\times SU(2)_R`$ which may arise in NSE. This group contains both $`SU(4)_c`$ and $`SU(2)_R`$ and does not require Wilson loops for its breaking. Furthermore, in string theories where the couplings have a common origin, partial or complete Yukawa unification can be realized even without a unified gauge group . Thus all four possibilities with complete, partial ($`tb`$ or $`b\tau `$) or no Yukawa unification are in principle allowed.
The soft SUSY breaking in the SE and NSE cases has been studied in Ref.. One obtains universal boundary conditions, i.e., a common scalar mass $`m_0`$, a common gaugino mass $`M_{1/2}`$ and a common trilinear coupling $`A_0`$ given by (with zero vacuum energy density and no CP violating phases)
$$m_0^2=m_{3/2}^2\frac{3m_{3/2}^2}{(3+ϵ)^2}\left(ϵ(6+ϵ)\mathrm{sin}^2\theta +(3+2ϵ)\mathrm{cos}^2\theta 2\sqrt{3}ϵ\mathrm{cos}\theta \mathrm{sin}\theta \right),$$
(1)
$$M_{1/2}=\frac{\sqrt{3}m_{3/2}}{1+ϵ}(\mathrm{sin}\theta +\frac{ϵ}{\sqrt{3}}\mathrm{cos}\theta ),$$
(2)
$$A_0=\frac{\sqrt{3}m_{3/2}}{3+ϵ}\left((32ϵ)\mathrm{sin}\theta +\sqrt{3}ϵ\mathrm{cos}\theta \right),$$
(3)
where $`m_{3/2}`$ is the gravitino mass, $`\theta `$ ($`0<\theta <\pi /2`$) is the goldstino angle, and the parameter $`ϵ`$ lies between 0 ($`1`$) and 1 in the SE (NSE) case . The range of $`ϵ`$ is the only difference between the two embeddings at the level of soft SUSY breaking.
In this paper, we will study the MSSM which results from the Hořava-Witten theory. We will assume radiative electroweak symmetry breaking with the universal boundary conditions in Eqs.(1)-(3) and examine all cases with complete, partial ($`tb`$ or $`b\tau `$) or no Yukawa unification. Our main aim is to restrict the parameter space by simultaneously imposing a number of phenomenological and cosmological constraints. In particular, the $`b`$-quark mass after including SUSY corrections and the branching ratio of $`bs\gamma `$ should be compatible with data. Also, the lightest supersymmetric particle (LSP) is required to provide the cold dark matter (CDM) in the universe. Its relic abundance must then be consistent with either of the two available cosmological models with zero/nonzero cosmological constant, which provide the best fits to all the data (see Refs.).
The GUT scale $`M_X`$ and gauge coupling constant are determined by using the 2-loop SUSY renormalization group equations (RGEs) for the gauge and Yukawa coupling constants between $`M_X`$ and a common SUSY threshold $`M_S\sqrt{m_{\stackrel{~}{t}_1}m_{\stackrel{~}{t}_2}}`$ ($`\stackrel{~}{t}_{1,2}`$ are the stop quark mass eigenstates), which minimizes the radiative corrections to $`\mu `$ and $`m_A`$ (see e.g., Ref.). Between $`M_S`$ and $`m_Z`$, we take the standard model (SM) 1-loop RGEs. The $`t`$-quark and $`\tau `$-lepton masses are fixed to their central experimental values $`m_t(m_t)=166\mathrm{GeV}`$ and $`m_\tau (m_\tau )=1.78\mathrm{GeV}`$. The asymptotic values of $`h_t`$, $`h_\tau `$ are then determined for each $`\mathrm{tan}\beta `$ at $`M_S`$ and $`h_b`$ is derived from $`tb`$ or $`b\tau `$ Yukawa unification. The resulting $`m_b(m_Z)`$ is compared to its experimental value $`m_b(m_Z)2.67\pm 0.98\mathrm{GeV}`$ (with a $`95\%`$ confidence margin) after 1-loop SUSY corrections. For complete Yukawa unification, $`\mathrm{tan}\beta `$ at $`M_S`$ is fixed. For no Yukawa unification, $`h_b`$ is adjusted so that the corrected $`m_b(m_Z)=2.67\mathrm{GeV}`$. $`M_S`$ is specified consistently with the SUSY spectrum.
We next integrate the 1-loop RGEs for the soft SUSY breaking terms assuming universal boundary conditions given by Eqs.(1)-(3). At $`M_S`$, we impose the minimization conditions to the tree-level renormalization group improved potential and calculate the Higgsino mass $`\mu `$ (up to its sign). The sparticle spectrum is evaluated at $`M_S`$. The LSP, which is the lightest neutralino ($`\stackrel{~}{\chi }`$), turns out to be bino-like with purity $`>98\%`$ for almost all values of the parameters. The next-to-lightest supersymmetric particle (NLSP) is the lightest stau ($`\stackrel{~}{\tau }_2`$). Since we consider large $`\mathrm{tan}\beta `$’s too, we are obliged to include the third generation sfermion mixing. The mixing of the lighter generation sfermions, however, remains negligible due to the small masses of the corresponding fermions. Furthermore, we take into account the 2-loop radiative corrections to the CP-even neutral Higgs boson masses $`m_h`$, $`m_H`$, which turn out to be sizeable for the lightest boson $`h`$.
Our calculation depends on the following free parameters: $`\mathrm{sign}\mu `$, $`\mathrm{tan}\beta `$, $`m_{3/2}`$, $`ϵ`$, $`\theta `$. The relation found in Ref. between the CP-odd Higgs boson mass $`m_A`$ and the asymptotic scalar and gaugino masses, takes, in our case, the form
$$m_A^2c_{3/2}m_{3/2}^2+c_sm_{3/2}^2\mathrm{sin}^2\theta +c_{2s}m_{3/2}^2\mathrm{sin}2\theta m_Z^2,$$
(4)
where the coefficients $`c_{3/2}0.1`$, $`c_s,c_{2s}1`$ depend on $`\mathrm{tan}\beta `$, $`ϵ`$, and $`M_S`$. We verified that this relation holds with an accuracy better than $`0.02\%`$. We use it to express $`m_{3/2}`$ in terms of $`m_A`$ for fixed $`\mathrm{sign}\mu `$, $`\mathrm{tan}\beta `$, $`ϵ`$ and $`\theta `$ ($`M_S`$ is determined self-consistently from the SUSY spectrum). The free parameter $`m_{3/2}`$ can, thus, be replaced by $`m_A`$.
In practice, the number of free parameters can be reduced by one. To see this, we fix $`\mathrm{sign}\mu `$, $`\mathrm{tan}\beta `$ and $`m_A`$ and observe that, along the lines in the $`ϵ\theta `$ plane where $`m_0`$ and $`M_{1/2}`$ remain constant, $`A_0`$ varies only by a few per cent. Consequently, the whole sparticle spectrum (except the gravitino mass) remains essentially unchanged along these lines which we call equispectral lines. Thus, for all practical purposes, $`ϵ`$ and $`\theta `$ can be replaced by a single parameter which we choose to be the relative mass splitting between the LSP and the NLSP $`\mathrm{\Delta }_{NLSP}=(m_{\stackrel{~}{\tau }_2}m_{\stackrel{~}{\chi }})/m_{\stackrel{~}{\chi }}`$. Our final free parameters then are $`\mathrm{sign}\mu `$, $`\mathrm{tan}\beta `$, $`m_A`$, $`\mathrm{\Delta }_{NLSP}`$. Note that, for fixed $`ϵ`$, $`\mathrm{\Delta }_{NLSP}`$ increases as $`\theta `$ decreases. Also, for fixed $`\theta >\pi /6(<\pi /6)`$, $`\mathrm{\Delta }_{NLSP}`$ decreases (increases) as $`ϵ`$ increases. Finally, we find that $`\mathrm{\Delta }_{NLSP}`$ is maximized, generally, at $`\theta =\pi /9`$ and $`ϵ1`$. Our calculation is performed at an appropriate value of $`ϵ`$ in each case so that all relevant $`\mathrm{\Delta }_{NLSP}`$’s can be obtained.
An important constraint results from the inclusive branching ratio of $`bs\gamma `$ , which is calculated here by using the formalism of Ref.. The dominant contributions, besides the SM one, come from the charged Higgs bosons ($`H^\pm `$) and the charginos. The former interferes constructively with the SM contribution, while the latter interferes constructively (destructively) with the other two contributions when $`\mu >0`$ ($`\mu <0`$). The SM contribution, which is factorized out in the formalism of Ref., includes the next-to-leading order (NLO) QCD and the leading order (LO) QED corrections. The NLO QCD corrections to the charged Higgs boson contribution are taken from the first paper in Ref.. The SUSY contribution is evaluated by including only the LO QCQ corrections using the formulae in Ref.. NLO QCD corrections to the SUSY contribution have also been discussed in Ref., but only under certain very restrictive conditions which never hold in our case since the chargino and lightest stop quark masses are comparable to the masses of the other squarks and the gluinos. We, thus, do not include these corrections in our calculation.
The branching ratio $`\mathrm{BR}(bs\gamma )`$ is first evaluated with central values of the input parameters and the renormalization and matching scales. We find that, for each $`\mathrm{sign}\mu `$, $`\mathrm{tan}\beta `$ and $`\mathrm{\Delta }_{NLSP}`$, there exists a value of $`m_A`$ above which the $`\mathrm{BR}(bs\gamma )`$ enters and remains in the experimentally allowed region : $`2\times 10^4\stackrel{_<}{_{}}\mathrm{BR}(bs\gamma )\stackrel{_<}{_{}}4.5\times 10^4`$. This lower bound on $`m_A`$ corresponds to the upper (lower) bound on the branching ratio for $`\mu >0`$ ($`\mu <0`$) and, for most of the parameter space, is its absolute minimum. For relatively small $`\mathrm{tan}\beta `$’s, however, the absolute minimum of $`m_A`$ comes from the experimental bound $`m_h\stackrel{_>}{_{}}113.4\mathrm{GeV}`$. We take $`\mathrm{tan}\beta 2.3`$ since otherwise $`m_h`$ is too small.
The lower bound on $`m_A`$ can be considerably reduced if the theoretical uncertainties entering into the calculation of $`\mathrm{BR}(bs\gamma )`$ are taken into account. These uncertainties originating from the experimental errors in the input parameters and the ambiguities in the renormalization and matching scales are known to be quite significant. The SM and charged Higgs contributions generate an uncertainty of about $`\pm 10\%`$ (see first paper in Ref.). The uncertainty from the SUSY contribution cannot be reliably calculated at the moment since the NLO QCD corrections to this contribution are not known in our case. Fortunately, the SUSY contribution is pretty small in all cases which are crucial for our qualitative conclusions. Be that as it may, we take the uncertainty from this contribution, evaluated at the LO in QCD, to be about $`\pm 30\%`$.
For large or intermediate $`\mathrm{tan}\beta `$’s, a severe restriction arises from the sizable SUSY corrections to the $`b`$-quark mass. The dominant contributions are from the sbottom-gluino and stop-chargino loops and are calculated by using the simplified formulae of Ref.. We find here that the size of these corrections practically depends only on $`\mathrm{tan}\beta `$ (compare with Refs.). Also, their sign is opposite to the one of $`\mu `$ in contrast to the chargino contribution to the $`\mathrm{BR}(bs\gamma )`$ which, as mentioned, has the sign of $`\mu `$.
An additional restriction comes from the LSP cosmic relic abundance. We calculate this abundance by closely following the formalism of Ref. where $`\stackrel{~}{\chi }\stackrel{~}{\tau }_2`$ coannihilations have been consistently included for all values of $`\mathrm{tan}\beta `$. However, coannihilations of these sparticles with the lighter generation right-handed sleptons $`\stackrel{~}{e}_R`$, $`\stackrel{~}{e}_R^{}`$, $`\stackrel{~}{\mu }_R`$, $`\stackrel{~}{\mu }_R^{}`$ (considered degenerate), which were ignored in Ref., are now important and must be included since our calculation here extends to small ($`15`$) $`\mathrm{tan}\beta `$’s too . The effective cross section entering into the Boltzmann equation then becomes
$`\sigma _{eff}`$ $`=`$ $`\sigma _{\stackrel{~}{\chi }\stackrel{~}{\chi }}r_{\stackrel{~}{\chi }}r_{\stackrel{~}{\chi }}+4\sigma _{\stackrel{~}{\chi }\stackrel{~}{\tau }_2}r_{\stackrel{~}{\chi }}r_{\stackrel{~}{\tau }_2}+2(\sigma _{\stackrel{~}{\tau }_2\stackrel{~}{\tau }_2}+\sigma _{\stackrel{~}{\tau }_2\stackrel{~}{\tau }_2^{}})r_{\stackrel{~}{\tau }_2}r_{\stackrel{~}{\tau }_2}+8(\sigma _{\stackrel{~}{\tau }_2\stackrel{~}{e}_R}+\sigma _{\stackrel{~}{\tau }_2\stackrel{~}{e}_R^{}})r_{\stackrel{~}{\tau }_2}r_{\stackrel{~}{e}_R}`$ (5)
$`+`$ $`8\sigma _{\stackrel{~}{\chi }\stackrel{~}{e}_R}r_{\stackrel{~}{\chi }}r_{\stackrel{~}{e}_R}+4(\sigma _{\stackrel{~}{e}_R\stackrel{~}{e}_R}+\sigma _{\stackrel{~}{e}_R\stackrel{~}{e}_R^{}})r_{\stackrel{~}{e}_R}r_{\stackrel{~}{e}_R}+4(\sigma _{\stackrel{~}{e}_R\stackrel{~}{\mu }_R}+\sigma _{\stackrel{~}{e}_R\stackrel{~}{\mu }_R^{}})r_{\stackrel{~}{e}_R}r_{\stackrel{~}{e}_R}.`$ (6)
Here $`\sigma _{ij}`$ ($`i,j=\stackrel{~}{\chi }`$, $`\stackrel{~}{\tau }_2`$, $`\stackrel{~}{\tau }_2^{}`$, $`\stackrel{~}{e}_R`$, $`\stackrel{~}{e}_R^{}`$, $`\stackrel{~}{\mu }_R`$, $`\stackrel{~}{\mu }_R^{}`$) is the total cross section for particle $`i`$ to annihilate with particle $`j`$ averaged over initial spin and particle-antiparticle states and the $`r_i`$’s can be found from Ref.. The Feynman graphs for $`\sigma _{\stackrel{~}{\chi }\stackrel{~}{\chi }}`$, $`\sigma _{\stackrel{~}{\chi }\stackrel{~}{\tau }_2}`$, $`\sigma _{\stackrel{~}{\tau }_2\stackrel{~}{\tau }_2}`$, and $`\sigma _{\stackrel{~}{\tau }_2\stackrel{~}{\tau }_2^{}}`$ are listed in Table I of Ref.. From these diagrams, we can also obtain the ones for $`\sigma _{\stackrel{~}{\chi }\stackrel{~}{e}_R}`$, $`\sigma _{\stackrel{~}{e}_R\stackrel{~}{e}_R}`$, $`\sigma _{\stackrel{~}{e}_R\stackrel{~}{e}_R^{}}`$ by replacing $`\stackrel{~}{\tau }_2`$ by $`\stackrel{~}{e}_R`$ and $`\tau `$ by $`e`$ and ignoring diagrams with $`\stackrel{~}{\tau }_1`$ exchange. The processes $`\stackrel{~}{\tau }_2\stackrel{~}{e}_R\tau e`$, $`\stackrel{~}{\tau }_2\stackrel{~}{e}_R^{}\tau \overline{e}`$, $`\stackrel{~}{e}_R\stackrel{~}{\mu }_Re\mu `$ and $`\stackrel{~}{e}_R\stackrel{~}{\mu }_R^{}e\overline{\mu }`$ are realized via a t-channel $`\stackrel{~}{\chi }`$ exchange. The calculation of the $`a_{ij}`$’s and $`b_{ij}`$’s given in Ref. is readily extended to include these extra processes too.
The main contribution to the LSP (almost pure bino) annihilation cross section generally arises from stau exchange in the t- and u-channel leading to $`\tau \overline{\tau }`$ in the final state. We do not include s-channel exchange diagrams. So our results are not valid for values of $`m_{\stackrel{~}{\chi }}`$ very close to the poles at $`m_Z/2`$, $`m_h/2`$, $`m_H/2`$ or $`m_A/2`$ where the annihilation cross section is enhanced and the relic density drops considerably. The expressions for $`a_{\stackrel{~}{\chi }\stackrel{~}{\chi }}`$ and $`b_{\stackrel{~}{\chi }\stackrel{~}{\chi }}`$ can be found in Ref. (with the final state lepton masses neglected).
The most important contribution to coannihilation arises from the $`a_{ij}`$’s. (The contribution of the $`b_{ij}`$’s ($`ij\stackrel{~}{\chi }\stackrel{~}{\chi }`$), although included in the calculation, is in general negligible.) The contributions of the various coannihilation processes to the $`a_{ij}`$’s and $`b_{ij}`$’s ($`ij\stackrel{~}{\chi }\stackrel{~}{\chi }`$) are calculated using techniques and approximations similar to the ones in Ref.. In particular, the contributions to the $`a_{ij}`$’s from the processes with
* $`\stackrel{~}{\chi }\stackrel{~}{\tau }_2`$, $`\stackrel{~}{\tau }_2\stackrel{~}{\tau }_2`$, $`\stackrel{~}{\tau }_2\stackrel{~}{\tau }_2^{}`$ in the initial state are listed in Table II of Ref..
* $`\stackrel{~}{\chi }\stackrel{~}{e}_R`$, $`\stackrel{~}{e}_R\stackrel{~}{e}_R^{}`$ in the initial state can be obtained from the formulae in Tables II and IV of Ref. by the replacement $`\stackrel{~}{\tau }_2\stackrel{~}{e}_R`$ and putting $`\theta =0`$, $`m_\tau =0`$.
* $`\stackrel{~}{\tau }_2\stackrel{~}{e}_R`$, $`\stackrel{~}{\tau }_2\stackrel{~}{e}_R^{}`$, $`\stackrel{~}{e}_R\stackrel{~}{e}_R`$, $`\stackrel{~}{e}_R\stackrel{~}{\mu }_R`$, $`\stackrel{~}{e}_R\stackrel{~}{\mu }_R^{}`$ in the initial state are listed in the following Table:
TABLE. Contributions to the Coefficients $`a_{ij}`$
| Process | Contribution to the Coefficient $`a_{ij}`$ |
| --- | --- |
| $`\stackrel{~}{\tau }_2\stackrel{~}{e}_R\tau e`$ | $`e^4Y_R^4\mathrm{cos}^2\theta m_{\stackrel{~}{\chi }}^2(m_{\stackrel{~}{e}_R}+m_{\stackrel{~}{\tau }_2})^2/`$ |
| | $`8\pi c_W^4m_{\stackrel{~}{e}_R}m_{\stackrel{~}{\tau }_2}(m_{\stackrel{~}{\chi }}^2+m_{\stackrel{~}{e}_R}m_{\stackrel{~}{\tau }_2})`$ |
| $`\stackrel{~}{\tau }_2\stackrel{~}{e}_R^{}\tau \overline{e}`$ | $`e^4Y_L^2Y_R^2\mathrm{sin}^2\theta m_{\stackrel{~}{\chi }}^2(m_{\stackrel{~}{e}_R}+m_{\stackrel{~}{\tau }_2})^2/`$ |
| | $`8\pi c_W^4m_{\stackrel{~}{e}_R}m_{\stackrel{~}{\tau }_2}(m_{\stackrel{~}{\chi }}^2+m_{\stackrel{~}{e}_R}m_{\stackrel{~}{\tau }_2})`$ |
| $`\stackrel{~}{e}_R\stackrel{~}{e}_Ree`$ | $`e^4Y_R^4m_{\stackrel{~}{\chi }}^2/\pi c_W^4\mathrm{\Sigma }_e^2`$ |
| $`\stackrel{~}{e}_R\stackrel{~}{\mu }_Re\mu `$ | $`e^4Y_R^4m_{\stackrel{~}{\chi }}^2/2\pi c_W^4\mathrm{\Sigma }_e^2`$ |
| $`\stackrel{~}{e}_R\stackrel{~}{\mu }_R^{}e\overline{\mu }`$ | $`e^4Y_R^4m_{\stackrel{~}{e}_R}^2/12\pi c_W^4\mathrm{\Sigma }_e^2`$ |
where $`\theta `$ is the stau mixing angle (not to be confused with the goldstino angle), $`c_W=\mathrm{cos}\theta _W`$, $`Y_{L(R)}=1/2(1)`$ is the hypercharge of the left(right)-handed leptons and $`\mathrm{\Sigma }_e=m_{\stackrel{~}{\chi }}^2+m_{\stackrel{~}{e}_R}^2`$ with $`m_{\stackrel{~}{e}_R}`$ being the common mass of $`\stackrel{~}{e}_R`$, $`\stackrel{~}{\mu }_R`$.
The LSP relic abundance $`\mathrm{\Omega }_{LSP}h^2`$, which remains practically constant on the equispectral lines, can now be evaluated for any $`\mathrm{sign}\mu `$, $`\mathrm{tan}\beta `$, $`m_A`$ and $`\mathrm{\Delta }_{NLSP}`$. We find that, away from the poles, $`\mathrm{\Omega }_{LSP}h^2`$ increases with $`m_A`$ (or $`m_{\stackrel{~}{\chi }}`$). Also, for fixed $`m_{\stackrel{~}{\chi }}`$, it increases with $`\mathrm{\Delta }_{NLSP}`$, since coannihilation becomes less efficient. The mixed or the pure cold (in the presence of a nonzero cosmological constant) dark matter scenarios for large scale structure formation require $`0.09\stackrel{_<}{_{}}\mathrm{\Omega }_{LSP}h^2\stackrel{_<}{_{}}0.22`$ , which restricts $`\mathrm{\Delta }_{NLSP}`$ .
We will first examine the case with no Yukawa unification. As already mentioned, the asymptotic value of $`h_b`$ is specified, in this case, by requiring that $`m_b(m_Z)`$, after SUSY corrections, coincides with its central experimental value. For $`\mu >0`$, $`m_A`$ (and, thus, $`m_{\stackrel{~}{\chi }}`$) is forced to be quite large in order to have the $`\mathrm{BR}(bs\gamma )`$ reduced below its upper experimental limit. Thus, the LSP and NLSP masses are required to be relatively close to each other so that coannihilation is more efficient and the bounds on $`\mathrm{\Omega }_{LSP}h^2`$ can be satisfied. For $`\mu <0`$, smaller $`m_A`$’s are needed for enhancing the $`bs\gamma `$ branching ratio so as to overtake its lower bound. Thus, in some regions of the parameter space, one can get cosmologically acceptable LSP relic densities even without invoking coannihilation. For $`\mu >0`$, $`\mathrm{tan}\beta 38`$ or $`\mu <0`$, the Higgs sector turns out to be heavier than the LSP and NLSP ($`m_A450\{400\}\mathrm{GeV}`$ for $`\mu >0`$, $`\mathrm{tan}\beta 38`$ and $`m_A340\{310\}\mathrm{GeV}`$ for $`\mu <0`$) implying that processes with $`\tau H`$, $`\tau A`$, $`hH`$, $`HH`$, $`H^+H^{}`$, $`AA`$ in the final state are, generally, kinematically blocked. Here and below, the limiting values obtained by including the theoretical uncertainty in $`\mathrm{BR}(bs\gamma )`$ are indicated in curly brackets.
We start by constructing the regions in the $`m_{\stackrel{~}{\chi }}\mathrm{\Delta }_{NLSP}`$ plane allowed by the CDM and $`bs\gamma `$ considerations for each $`\mathrm{sign}\mu `$ and $`\mathrm{tan}\beta `$. A typical example of such a region is shown in Fig.1 and corresponds to $`\mu >0`$ and $`\mathrm{tan}\beta 10`$. Here, we fixed $`ϵ=0.65`$ and regulated $`\mathrm{\Delta }_{NLSP}`$ via $`\theta `$. The lower bound on $`m_{\stackrel{~}{\chi }}`$ (almost vertical line) comes from the upper bound ($`4.5\times 10^4`$) on $`\mathrm{BR}(bs\gamma )`$. The lower (upper) curved boundary of the allowed region corresponds to $`\mathrm{\Omega }_{LSP}h^20.09(0.22)`$ and the horizontal boundary to $`\mathrm{\Delta }_{NLSP}=0`$. The maximal $`m_{\stackrel{~}{\chi }}`$ ($`\mathrm{\Delta }_{NLSP}`$) is obtained at the lower right (upper left) corner of this region. The value of $`m_{\stackrel{~}{\chi }}`$ can vary between about $`169\{123\}`$ and $`575\mathrm{GeV}`$. So, the LSP is relatively heavy and the maximal allowed $`\mathrm{\Delta }_{NLSP}`$ is small ($`0.096\{0.19\}`$). Coannihilation is important in the whole allowed region. On the contrary, for $`\mu <0`$ and $`\mathrm{tan}\beta 10`$, we find lighter LSPs. Specifically, $`m_{\stackrel{~}{\chi }}`$ varies between about $`85\{79\}`$ and $`572\mathrm{GeV}`$. So, the maximal allowed $`\mathrm{\Delta }_{NLSP}`$ is much larger ($`0.6\{0.71\}`$) now, and there is a region ($`85\{79\}\mathrm{GeV}m_{\stackrel{~}{\chi }}120\mathrm{GeV}`$) where coannihilation is negligible. The lower bound on $`m_{\stackrel{~}{\chi }}`$, for $`\mu <0`$, corresponds to the lower bound on $`\mathrm{BR}(bs\gamma )`$ or $`m_h`$.
For $`\mu <0`$, there exist $`\mathrm{tan}\beta `$’s where the maximal $`\mathrm{\Delta }_{NLSP}`$ is not obtained at the minimal $`m_{\stackrel{~}{\chi }}`$. This is illustrated in Fig.2 depicting the allowed region in the $`m_{\stackrel{~}{\chi }}\mathrm{\Delta }_{NLSP}`$ plane for $`\mu <0`$ and $`\mathrm{tan}\beta 35.3`$. Here, we fixed $`ϵ=0.99`$. The LSP mass can vary between about $`203`$ and $`614\mathrm{GeV}`$ with the lower bound corresponding to the lower bound on $`\mathrm{BR}(bs\gamma )`$. For the minimal $`m_{\stackrel{~}{\chi }}`$, the maximal $`\mathrm{\Delta }_{NLSP}`$ ($`0.045`$) does not correspond to $`\mathrm{\Omega }_{LSP}h^20.22`$. It is, rather, the absolute maximum of $`\mathrm{\Delta }_{NLSP}`$ for the given values of $`\mathrm{sign}\mu `$, $`\mathrm{tan}\beta `$ and $`m_A`$ which is obtained at $`\theta =\pi /9`$ as indicated earlier and corresponds to $`\mathrm{\Omega }_{LSP}h^20.114`$. Increasing $`m_{\stackrel{~}{\chi }}`$, this absolute maximum of $`\mathrm{\Delta }_{NLSP}`$ increases (along the inclined part of the left boundary) and $`\mathrm{\Omega }_{LSP}h^2`$ becomes $`0.22`$ at $`\mathrm{\Delta }_{NLSP}0.064`$, which is the overall maximal allowed $`\mathrm{\Delta }_{NLSP}`$ in this case. Including the theoretical uncertainty in $`\mathrm{BR}(bs\gamma )`$, we see that the vertical part of the boundary disappears and the minimal value of $`m_{\stackrel{~}{\chi }}`$ is reduced to about $`198\mathrm{GeV}`$ corresponding to $`\mathrm{\Delta }_{NLSP}0.038`$.
The maximal allowed $`\mathrm{\Delta }_{NLSP}`$’s can be found for all possible $`\mathrm{tan}\beta `$’s and any sign of $`\mu `$ by repeating the above analysis. The results are displayed in Fig.3, which shows the allowed regions in the $`\mathrm{tan}\beta \mathrm{\Delta }_{NLSP}`$ plane for $`\mu >0`$ (between the solid and dashed lines) and $`\mu <0`$ (between the solid and dot-dashed lines). Here, the bold (faint) lines are obtained by ignoring (including) the theoretical errors in $`\mathrm{BR}(bs\gamma )`$, and $`ϵ`$ is chosen for each $`\mathrm{sign}\mu `$ and $`\mathrm{tan}\beta `$ so that it lies in the domain of all relevant equispectral lines. We found that $`\mathrm{\Delta }_{NLSP}=0`$ can be achieved at the maximal LSP mass ($`600700\mathrm{GeV}`$) corresponding to each $`\mathrm{tan}\beta `$ between 2.3 and $`43.9\{44.3\}`$. So, the minimal allowed $`\mathrm{\Delta }_{NLSP}`$ is always zero. Regarding the maximal allowed $`\mathrm{\Delta }_{NLSP}`$’s, we can distinguish the cases:
* For $`\mu >0`$ ($`<0`$) and $`6.5\{8.6\}(9.2)\mathrm{tan}\beta 43.9\{44.3\}(34.5)`$, the maximal $`\mathrm{\Delta }_{NLSP}`$ corresponds to the lower bound on $`m_{\stackrel{~}{\chi }}`$ found from the experimental limits on $`\mathrm{BR}(bs\gamma )`$. The allowed regions are of the type in Fig.1 and the upper curves in Fig.3 are obtained from the upper left corners of these regions as we vary $`\mathrm{tan}\beta `$. For $`\mu >0`$, the lower curved boundary of the allowed regions disappears at high enough $`\mathrm{tan}\beta `$’s and, eventually, at $`\mathrm{tan}\beta 43.9\{44.3\}`$, the allowed region shrinks to a point with $`m_{\stackrel{~}{\chi }}730\{740\}\mathrm{GeV}`$ and $`\mathrm{\Delta }_{NLSP}0`$.
* For $`\mu >0`$ ($`<0`$) and $`2.3\mathrm{tan}\beta 6.5\{8.6\}(9.2\{9.8\})`$, the lower bound on $`m_{\stackrel{~}{\chi }}`$ is found from the experimental limit on $`m_h`$. This mass comes out too small for small $`m_A`$’s. So, bigger $`m_A`$’s (and, thus, $`m_{\stackrel{~}{\chi }}`$’s) are required to raise $`m_h`$ above $`113.4\mathrm{GeV}`$. The allowed regions are again typically as in Fig.1 (with or without the curved lower boundary) and the maximal $`\mathrm{\Delta }_{NLSP}`$ rapidly decreases with $`\mathrm{tan}\beta `$.
* For $`\mu <0`$ and $`\mathrm{tan}\beta `$ between about $`34.5\{9.8\}`$ and 41, the maximal $`\mathrm{\Delta }_{NLSP}`$ does not correspond to the minimal $`m_{\stackrel{~}{\chi }}`$ from the lower limit on $`\mathrm{BR}(bs\gamma )`$ or $`m_h`$. The obtained allowed regions are of the type in Fig.2 (with or without the vertical part of the boundary). As $`\mathrm{tan}\beta `$ increases above $`34.5\{9.8\}`$, the inclined part of their left boundary moves to the right and the vertical part eventually disappears. At even higher $`\mathrm{tan}\beta `$’s, the curved lower boundary also disappears and, finally, the region shrinks to a point at $`\mathrm{tan}\beta 41`$ with $`\mathrm{\Delta }_{NLSP}0`$ and $`m_{\stackrel{~}{\chi }}640\mathrm{GeV}`$. For low $`\mathrm{tan}\beta `$’s, the bino purity of the LSP decreases from above $`98\%`$ to $`95\%`$ and our calculation, which assumes a bino-like LSP, becomes less accurate.
In conclusion, in the case of no Yukawa unification and for $`\mu >0`$ ($`<0`$), the maximal $`\mathrm{\Delta }_{NLSP}0.16\{0.25\}(0.68\{0.73\})`$ is achieved at $`\mathrm{tan}\beta 6.5\{8.6\}(9.2\{9.8\})`$. Also, $`138\{114\}(84\{77\})\mathrm{GeV}m_{\stackrel{~}{\chi }}730\{740\}(640)\mathrm{GeV}`$. The minimal $`m_{\stackrel{~}{\chi }}`$ corresponds to the maximal $`\mathrm{\Delta }_{NLSP}`$ except $`m_{\stackrel{~}{\chi }}77\mathrm{GeV}`$ which is obtained at $`\mathrm{tan}\beta 20.4`$.
We now turn to the case of $`b\tau `$ Yukawa unification. To keep $`\stackrel{~}{\tau }_2`$ heavier than $`\stackrel{~}{\chi }`$, we must take $`\mathrm{tan}\beta 45`$. For $`\mu <0`$, the values of $`m_b(m_Z)`$, obtained from this unification assumption, turn out to be larger than the experimental upper limit after including the SUSY corrections. This forces us to take $`\mu >0`$. In Fig.4, we plot the tree-level (dotted line) and the corrected (solid line) $`m_b(m_Z)`$ versus $`\mathrm{tan}\beta `$ for $`\mathrm{\Delta }_{NLSP}0`$ and the minimal value of $`m_A`$ which corresponds to $`\mathrm{BR}(bs\gamma )4.5\times 10^4`$ for $`6.5\mathrm{tan}\beta 45`$ or $`m_h113.4\mathrm{GeV}`$ for $`2.3\mathrm{tan}\beta 6.5`$. This choice is not crucial, because $`m_b(m_Z)`$, for fixed $`\mathrm{tan}\beta `$, turns out to be almost independent from $`m_A`$ and $`\mathrm{\Delta }_{NLSP}`$. The corrected $`m_b(m_Z)`$ increase as $`\mathrm{tan}\beta `$ decreases and reaches a maximum of about $`3.65\mathrm{GeV}`$ at $`\mathrm{tan}\beta 4.7`$. The SUSY corrections decrease with $`\mathrm{tan}\beta `$. We find that, in the entire range $`2.3\mathrm{tan}\beta 45`$, the corrected $`m_b(m_Z)`$ is within the experimental limits.
Due to the relatively heavy LSP obtained with $`\mu >0`$, coannihilation is generally important for reducing $`\mathrm{\Omega }_{LSP}h^2`$ to an acceptable level. For $`38\mathrm{tan}\beta 45`$, the maximal allowed $`m_{LSP}`$ is raised to $`790\mathrm{GeV}`$ due to the fact that the processes with $`\tau H`$, $`\tau A`$ in the final state are kinematically allowed. Thus, coannihilation is strengthened and larger $`m_{LSP}`$’s are allowed. On the contrary, for $`2.3\mathrm{tan}\beta 34`$, these processes are blocked and the upper bound on $`m_{LSP}`$ decreases to $`580\mathrm{GeV}`$. $`\mathrm{\Delta }_{NLSP}`$ ranges between 0 and $`0.16\{0.25\}`$ with its maximum achieved at $`\mathrm{tan}\beta 6.5\{8.6\}`$ corresponding to the lowest possible $`m_{LSP}141\{115\}\mathrm{GeV}`$. Finally, in the range $`34\mathrm{tan}\beta 38`$, $`m_{LSP}`$ can get close to $`m_A/2`$, $`m_H/2`$ for certain $`m_A`$’s and $`\mathrm{\Omega }_{LSP}h^2`$ can be considerably reduced. Thus, the maximal $`\mathrm{\Delta }_{NLSP}`$ and $`m_{LSP}`$ can be very large in isolated regions of the parameter space. This also applies in the no Yukawa unification case with $`\mu >0`$.
In the case of $`tb`$ Yukawa unification the corrected $`m_b(m_Z)`$, for $`\mu <0`$, again turns out to be larger than the experimental upper limit, so we must still choose $`\mu >0`$. We find that, for $`34.3\mathrm{tan}\beta `$, the corrected $`m_b(m_Z)`$ is compatible with the experimental limits after including its theoretical uncertainties ($`6\%`$). This provides the lower bound on $`\mathrm{tan}\beta `$ if the theoretical uncertainties in $`\mathrm{BR}(bs\gamma )`$ are included. Without these uncertainties, however, the lower bound on $`\mathrm{tan}\beta `$ is 43.7 below which the allowed region in the $`m_{LSP}\mathrm{\Delta }_{NLSP}`$ plane disappears. To keep $`\stackrel{~}{\tau }_2`$ heavier than $`\stackrel{~}{\chi }`$, we must take $`\mathrm{tan}\beta 48.5`$. So there is an allowed range $`43.7\{33.5\}\mathrm{tan}\beta 48.5`$ in which the minimal $`m_{LSP}`$ is about $`730\{507\}\mathrm{GeV}`$ with the maximal $`\mathrm{\Delta }_{NLSP}`$ being $`0\{0.01\}`$.
For complete Yukawa unification, the lightest stau turns out to be lighter than the neutralino (by at least $`11\%`$). So, this case is excluded.
Theoretical errors from the implementation of the radiative electroweak breaking, the renormalization group analysis and the radiative corrections to (s)particle masses, and inclusion of experimental margins of various quantities can only further widen the allowed parameter ranges which we obtained. They will also produce a larger uncertainty in Eq.(4). However, all these ambiguities are not expected to change our qualitative conclusions, especially the exclusion of complete Yukawa unification.
Neutralinos could be detected via their elastic scattering with nuclei. For an almost pure bino, however, the cross section is expected to lie well below the reported sensitivity \[$`(110)\times 10^6\mathrm{Pb}`$\] of current experiments (DAMA). The reason is that the channels with Higgs and $`Z`$ boson (squark) exchange are suppressed (by the squark mass).
In summary, we studied the MSSM with radiative electroweak breaking and boundary conditions from the Hořava-Witten theory. We assumed complete, partial or no Yukawa unification. The parameters were restricted by assuming that the CDM consists of the LSP and requiring $`m_b`$, after SUSY corrections, and $`\mathrm{BR}(bs\gamma `$) to be compatible with data. We found that complete Yukawa unification is excluded. Also, $`tb`$ Yukawa unification is strongly disfavored since it requires the LSP and NLSP masses to be almost degenerate. This can be avoided with $`b\tau `$ or no Yukawa unification which, for $`\mu <0`$, is the most natural case and allows the LSP mass to be as low as $`77\mathrm{GeV}`$.
We thank M. Gómez and C. Muñoz for discussions. S. K. is supported by the Spanish Ministerio de Educacion y Cultura and C. P. by the Greek State Scholarship Institution (I. K. Y.). This work was supported by the EU under TMR contract No. ERBFMRX–CT96–0090 and the Greek Government research grant PENED/95 K.A.1795.
|
warning/0005/astro-ph0005295.html
|
ar5iv
|
text
|
# Constraints on diffuse neutrino background from primordial black holes
## I Introduction
Some recent inflation models (e.g., the hybrid inflationary scenario ) predict the ”blue” power-spectrum of primordial density fluctuations. In turn, as is well known, the significant abundance of primordial black holes (PBHs) is possible just in the case when the density fluctuations have an $`n>1`$ spectrum ($`n`$ is the spectral index of the initial density fluctuations, $`n>1`$ spectrum is, by definition, the ”blue perturbation spectrum”).
Particle emission from PBHs due to the evaporation process predicted by Hawking may lead to observable effects. Up to now, PBHs have not been detected, so the observations have set limits on the initial PBH abundance or on characteristics of a spectrum of the primordial density fluctuations. In particular, PBH evaporations contribute to the extragalactic neutrino background. The constraints on an intensity of this background (and, correspondingly, on an PBH abundance) can be obtained from the existing experiments with atmospheric and solar neutrinos. The obtaining of such constraints is a main task of the present paper.
The spectrum and the intensity of the evaporated neutrinos depend heavily on the PBH’s mass. Therefore, the great attention should be paid to the calculation of the initial mass spectrum of PBHs. We use in this paper the following assumptions leading to a prediction of the PBH’s mass spectrum.
1. The formation of PBHs begins only after an inflation phase when the universe returns to the ordinary radiation-dominated era. The reheating process is such that an equation of state of the universe changes almost instantaneously into the radiation type (e.g., due to the parametric resonance ) after the inflation.
2. It is assumed, in accordance with analytic calculations that a critical size of the density contrast needed for the PBH formation, $`\delta _c`$, is about $`1/3`$. Further, it is assumed that all PBHs have mass roughly equal to the horizon mass at a moment of the formation, independently of the perturbation size.
3. Summation over all epochs of the PBH formation can be done using the Press-Schechter formalism . This formalism is widely used in the standard hierarchial model of the structure formation for calculations of the mass distribution functions (see, e.g., ).
It was shown recently that near the threshold of a black hole formation the gravitational collapse behaves as a critical phenomenon . In this case the initial mass function will be quite different from that follows from the standard calculations of refs. .The first calculations of the PBH initial mass function (e.g., in ref.) have been done under the assumption that all PBHs form at the same horizon mass (by other words, that all PBHs form at the smallest horizon scale immediatly after reheating). The initial PBH mass spectrum for the case of the critical collapse, based on the Press-Schechter formalism, was obtained in the recent work .
The calculations in the present paper are based on the standard picture of the gravitational collapse leading to a PBH formation. The case of the critical collapse will be considered in a separate work.
The plan of the paper is as follows.
In Sec.II we give, for completeness, the brief derivation of a general formula for the initial PBH mass spectrum. The final expression is presented in a form, which is valid for an arbitrary relation between three physical values: the initial PBH mass, the fluctuation mass (i.e., the mass of the perturbed region) at a moment of the collapse, and the density contrast in the perturbed region (also at a moment of the collapse). This expression contains the corresponding results obtained in refs. as particular cases. As in refs. , the derivation is based on the linear perturbation theory and on the assumption that a power spectrum of the primordial fluctuations can be described by a power law.
In Sec.III we derive the approximate formula for a calculation of the extragalactic neutrino background from PBH evaporations. We do not use cosmological models of the inflation and of the spectrum of primordial fluctuations, so there are two free parameters: the reheating temperature and the spectral index. At the end of the section, some examples of instantaneous neutrino spectra from the evaporation of an individual black hole are presented.
In Sec.IV we determine the explicit dependence of an background intensity on the spectral index (by normalization of standard deviation of the density contrast at horizon crossing on COBE data). After this, numerical calculations of the neutrino background become possible. For background neutrino energies $`10100MeV`$ we studied relative contributions to the background intensity of different cosmological redshifts (and it is shown that at high reheating temperatures the characteristic values of the redshifts are very large). Further, several results of numerical calculations of the neutrino background spectra are presented. Possibilities of constraining the spectral index using experiments with neutrinos of natural origin are discussed.
In Sec.V the spectral index constraints, followed from our calculations and from available data of the neutrino experiments, are given. It is proved that, almost in all intervals of the reheating temperatures considered in this work, effects of an background neutrino absorption in the space are small and can be neglected.
## II The initial mass spectrum of PBHs
As it pointed out in the Introduction, we use the Press-Schechter formalism which allows to carry out the summation over all epochs of PBHs formation. According to this formalism, the mass spectrum of density fluctuations ( i.e., the number density of regions with mass between $`M`$ and $`M+dM`$ ) is calculated by the formula
$$n(M,\delta _c)dM=\frac{\rho _i}{M}\left|\frac{\beta }{M}(M,\delta _c)\right|dM.$$
(1)
Here, $`\beta (M,\delta _c)`$ is the fraction of regions having sizes larger than $`R`$ and density contrast larger than $`\delta _c`$ ,
$$\beta (M,\delta _c)=2_{\delta _c}^{\mathrm{}}P(M,\delta )𝑑\delta ,$$
(2)
$$P(M,\delta )=\frac{1}{\sqrt{2\pi }\sigma _R(M)}\mathrm{exp}\left(\frac{\delta ^2}{2\sigma _R^2(M)}\right).$$
(3)
Here, $`\delta `$ is the initial density contrast, $`\sigma _R`$ is the standard deviation of the density contrast of the regions with size $`R`$ and mass $`M`$.
It is convenient to introduce the double differential distribution $`n(M,\delta )`$, the integral over which gives the total number of fluctuated regions,
$$n=n(M,\delta )dMd\delta ,n(M,\delta _c)=n(M,\delta )d\delta .$$
(4)
Using Eqs.(1-3), for $`n(M,\delta )`$ one has the expression
$$n(M,\delta )=\sqrt{\frac{2}{\pi }}\frac{\rho _i}{M}\frac{1}{\sigma _R^2(M)}\frac{\sigma _R}{M}\left|\left(\frac{\delta ^2}{\sigma _R^2(M)}1\right)\right|\mathrm{exp}\left(\frac{\delta ^2}{2\sigma _R^2(M)}\right).$$
(5)
To obtain the mass spectrum of PBHs one must introduce the $`M_{BH}`$ variable. Besides, we will use the variable $`\delta ^{}`$, connected with $`\delta `$ by the relation
$$\delta ^{}=\delta \left(\frac{M}{M_i}\right)^{2/3}.$$
(6)
Here, $`M_i`$ is the horizon mass at the moment of a beginning of the growth of density fluctuations. This new variable is the density contrast at the moment of the collapse. The new distribution function is
$$n_{BH}(M_{BH},\delta ^{})=n(M_{BH},\delta )\frac{d\delta }{d\delta ^{}}\frac{dM}{dM_{BH}}.$$
(7)
Further, we assume that there is some functional connection between $`M_{BH}`$, $`M`$ and $`\delta ^{}`$:
$$M_{BH}=f(M,\delta ^{}).$$
(8)
In this case one can rewrite Eq.(7) in the form:
$$n_{BH}(M_{BH},\delta ^{})=n(M,\delta ^{}\left(\frac{M}{M_i}\right)^{2/3})\left(\frac{M}{M_i}\right)^{2/3}\frac{1}{df(M,\delta ^{})/dM}$$
(9)
and the PBH mass spectrum is given by the integral
$$n_{BH}(M_{BH})=n_{BH}(M_{BH},\delta ^{})𝑑\delta ^{}.$$
(10)
Now, to connect the mass spectrum with the spectral index one can use the relations
$$\sigma _R=\sigma _H(M)\left(\frac{M}{M_i}\right)^{2/3};\sigma _H(M)M^{\frac{1n}{6}}$$
(11)
(see Sec.IV for details). Using Eqs.(11) in the expression (5) for $`n(M,\delta )`$ one obtains:
$$n(M,\delta )=\frac{n+3}{6}\sqrt{\frac{2}{\pi }}\frac{\rho _i}{M_i^2}\frac{1}{\sigma _H}\left(\frac{M}{M_i}\right)^{4/3}\left|\frac{\delta ^2}{\sigma _R^2}1\right|e^{\delta ^2/2\sigma _R^2}.$$
(12)
Substituting this expression in the r.h.s. of the Eq.(9) and using Eq(10) one has:
$$n_{BH}(M_{BH})=\frac{n+3}{6}\sqrt{\frac{2}{\pi }}\rho _i\frac{1}{M^2\sigma _H}\left|\frac{\delta _{}^{}{}_{}{}^{2}}{\sigma _H^2}1\right|e^{\frac{\delta _{}^{}{}_{}{}^{2}}{2\sigma _H^2}}\frac{1}{df(M,\delta ^{})/dM}𝑑\delta ^{}.$$
(13)
Values of $`M`$ in r.h.s of Eq.(13) are expressed through $`M_{BH}`$, $`\delta ^{}`$ by the relation (8).
Formula (13) is the final expression for the PBH mass spectrum, the main result of this Section. It is valid for any relation between PBH mass $`M_{BH}`$, mass of the original overdense region $`M`$ and the density contrast $`\delta ^{}`$.
In the particular case of a near critical collapse
$$f(M,\delta ^{})=M_i^{1/3}M^{2/3}k(\delta ^{}\delta _c)^{\gamma _k}\xi M_i^{1/3}M^{2/3}$$
(14)
and
$$M=M_{BH}^{3/2}M_i^{1/2}\xi ^{3/2};df/dM=\frac{2}{3}M_i^{1/2}M_{BH}^{1/2}\xi ^{3/2}.$$
(15)
Eq.(14) can be rewritten in the form, derived in the ref. , using the connection between $`M`$ and the horizon mass $`M_h`$ (which is equal to a fluctuation mass at the moment when the fluctuation crosses horizon):
$$M_h=M_i^{1/3}M^{2/3}.$$
(16)
The resulting formula for the spectrum,
$$n(M_{BH})=\frac{n+3}{4}\sqrt{\frac{2}{\pi }}\sqrt{M_i}M_{BH}^{5/2}_{\delta _c}^1\frac{1}{\sigma _H}\left|\frac{\delta ^{}^2}{\sigma _H^2}1\right|e^{\frac{\delta ^2}{2\sigma _H^2}}\xi ^{3/2}𝑑\delta ^{},$$
(17)
coincides, as can be easily proved, with the expression derived in ref. .
The mass spectrum formula for the Carr-Hawking collapse can be obtained analogously, with the relation $`f(M,\delta ^{})=\gamma ^{1/2}M_i^{1/3}M^{2/3}`$, or from Eq.(17) using the substitutions
$$\gamma _k0,k\gamma ^{1/2},\delta _c\gamma $$
(18)
and the approximate relation
$$_\gamma ^1𝑑\delta ^{}\left(\frac{\delta ^{}^2}{\sigma _H^2}1\right)e^{\frac{\delta ^{}^2}{2\sigma _H^2}}\gamma e^{\frac{\gamma ^2}{2\sigma _H^2}}.$$
(19)
In this case one obtains the expression obtained in ref. :
$$n_{BH}(M_{BH})=\frac{n+3}{4}\sqrt{\frac{2}{\pi }}\gamma ^{7/4}\rho _iM_i^{1/2}M_{BH}^{5/2}\sigma _H^1\mathrm{exp}\left(\frac{\gamma ^2}{2\sigma _H^2}\right).$$
(20)
One can see from Eqs.(17) and (20) that the PBH mass spectrum following from the Press-Shechter formalism has quasi power form in both considered cases $`\left(M_{BH}^{5/2}\right)`$. In Carr-Hawking case $`M_{BH}^{min}M_h`$ , in contrast with this in the critical collapse case it is possible that $`M_{BH}M_h`$ . However , the low mass part of the PBH spectrum is suppressed by the factor $`\xi ^{3/2}`$ in the integral in Eq. (17).
## III Neutrino diffuse background from PBHs
The starting formula for a calculation of the cosmological background from PBH evaporations is
$$S(E)=n_{com}\frac{1}{4\pi a_0^2\rho _{}^2}f\left(E(1+z)\right)𝑑V_{com}.$$
(21)
Here , $`n_{com}`$ is the comoving number density of the sources (in our case the source is an evaporating PBH of the definite mass $`m`$ ), $`a_0`$ is the scale factor at present time, $`t=t_0`$, $`f(E)`$ is a differential energy spectrum of the source radiation, $`V_{com}`$ is a comoving volume of the space filled by sources, therefore
$$dV_{com}=a_0^3\frac{\rho ^2d\rho }{\sqrt{1k\rho ^2}}d\mathrm{\Omega }.$$
(22)
Here, $`k`$ is the curvature coefficient, and $`\rho `$ is the radial comoving coordinate. Using the change of the variable,
$$\frac{d\rho }{\sqrt{1k\rho ^2}}=\frac{dt}{a},$$
(23)
the comoving number density can be expressed via the initial density $`n_i`$,
$$n_{com}=n_{phys}(t_0)=n_i\left(\frac{a_i}{a}\right)^3\left(\frac{a}{a_0}\right)^3=n_i\left(\frac{a_i}{a_0}\right)^3.$$
(24)
Substituting Eqs. (22) - (24) in Eq. (21) one obtains
$$S(E)=n_i𝑑t\frac{a_0}{a}\left(\frac{a_i}{a_0}\right)^3f\left(E(1+z)\right).$$
(25)
In our concrete case the source of the radiation is a Hawking evaporation:
$$n_if\left(E(1+z)\right)=𝑑mn_{BH}(m,t)f_H(E(1+z),m).$$
(26)
Here , $`n_{BH}(m,t)`$ is the PBH mass spectrum at any moment of time, $`f_H(E,m)`$ is the Hawking function ,
$$f_H(E,m)=\frac{1}{2\pi \mathrm{}}\frac{\mathrm{\Gamma }_s(E,m)}{exp\left(\frac{8\pi GEm}{\mathrm{}c^3}\right)(1)^{2s}}.$$
(27)
Here, $`\mathrm{\Gamma }_s(E,m)`$ is the coefficient of the absorption by a black hole of a mass $`m`$, for an particle having spin s and energy $`E`$.
Initial spectrum of PBHs is given by Eq. (20). The minimum value of PBH mass is $`\gamma _{}^{1/2}M_i^{}`$, so we must add to the initial spectrum expression the step factor $`\mathrm{\Theta }(m_{BH}^{}\gamma _{}^{1/2}M_i^{})`$ . The connection of the initial mass value $`M_{BH}`$ and the value at any moment $`t`$ is determined by the solution of the equation :
$$\frac{dm}{dt}=\frac{\alpha (m)}{m^2}.$$
(28)
The function $`\alpha (m)`$ accounts for the degrees of freedom of evaporated particles and determines the lifetime of a black hole. In the approximation $`\alpha =const`$ the solution of Eq.(28) is:
$$M_{BH}\left(3\alpha t+m^3\right)^{1/3}.$$
(29)
This decrease of PBH mass leads to the corresponding evolution of a form of the PBH mass spectrum. At any moment one has
$$n_{BH}(m,t)dm=\frac{m^2}{\left(3\alpha t+m^3\right)^{2/3}}n_{BH}\left((3\alpha t+m^3)^{1/3}\right)\times \mathrm{\Theta }\left[m\left((\gamma ^{1/2}M_i)^33\alpha t\right)^{1/3}\right]dm.$$
(30)
Substituting Eqs. (27), (30) in the integral in Eq.(26) , we obtain the final expression for the spectrum of the background radiation:
$`S(E)={\displaystyle \frac{c}{4\pi }}{\displaystyle }dt{\displaystyle \frac{a_0}{a}}\left({\displaystyle \frac{a_i}{a_0}}\right)^3{\displaystyle }dm{\displaystyle \frac{m^2}{(3\alpha t+m^3)^{1/3}}}n_{BH}\left[(3\alpha t+m^3)^{1/3}\right]\times `$ (31)
(32)
$`\mathrm{\Theta }[(m((\gamma ^{1/2}M_i)^33\alpha t)^{1/3}]f_H(E(1+z),m).`$ (33)
One should note that the corresponding expressions for the spectrum in refs. contain the factor $`\left(\frac{a_i}{a}\right)^3`$ instead of the correct factor $`\left(\frac{a_i}{a_0}\right)^3`$. It leads to a strong overestimation of large $`z`$ contributions in $`S(E)`$ (see below, Fig.3).
It is convenient to use in Eq. (31) the variable $`z`$ instead of $`t`$. In our case ($`\mathrm{\Omega }_\mathrm{\Lambda }=\mathrm{\Omega }_K=0`$) we have
$`{\displaystyle \frac{dt}{dz}}={\displaystyle \frac{1}{H_0(1+z)}}\left(\mathrm{\Omega }_{m_0}(1+z)^3+\mathrm{\Omega }_{r_0}(1+z)^4\right)^{1/2},`$ (34)
(35)
$`\mathrm{\Omega }_{r_0}=(2.2510^4h^2)^1,\mathrm{\Omega }_{m_0}=1\mathrm{\Omega }_{r_0}.`$ (36)
The factor $`(\frac{a_i}{a_0})^3`$ can be expressed through the value of $`t_{eq}`$:
$$\left(\frac{a_i}{a_0}\right)^3(1+z_{eq})^3\left(\frac{t_i}{t_{eq}}\right)^{3/2}H_0^{3/2}\left(2.2510^4h^2\right)^{3/4}t_i^{3/2}.$$
(37)
Integrating over PBH’s mass in Eq. (31) one obtains finally, after the change of the variable $`t`$ on $`z`$, the integral over $`z`$:
$$S(E)=d\mathrm{log}_{10}(z+1)F(E,z).$$
(38)
In analogous calculations of the photon diffuse background integral over $`z`$ in the expression for $`S(E)`$ is cut off at $`z=z_0700`$ because for larger $`z`$ the photon optical depth will be larger than unity . In contrast with this , interactions of neutrinos with the matter can be neglected up to very high values of $`z`$. Therefore the neutrino diffuse background from PBH evaporations is much more abundant. The neutrino absorption effects are estimated below, in Sec.V.
The evaporation process of a black hole with not too small initial mass is almost an explosion. So, for a calculation of spectra of evaporated particles with acceptable accuracy it is enough to know the value of $`\alpha `$ for an initial value of the PBH mass only. Taking this into account and having in mind the steepness of the PBH mass spectrum, we use the approximation
$$\alpha (m)=\alpha (M_{BH}^{min})=\alpha (\gamma ^{1/2}M_i),$$
(39)
and just this value of $`\alpha `$ is meant in the expressions (29)-(31).
The very detailed calculation of the function $`\alpha (m)`$ was carried out in the works . Here we use the simplified approach in which $`\alpha (m)`$ is represented by the dependence
$$\alpha =\alpha _0+\underset{i}{}a_i10^{25}\mathrm{\Theta }(b_ilog_{10}(m)).$$
(40)
Here, $`\mathrm{\Theta }(x)`$ is the Heaviside step function. Coefficient $`\alpha _0`$ gives the summary contribution of $`e^+`$, $`e^{}`$, $`\nu `$ and $`\gamma `$ and is equal to $`8.4210^{25}g^3sec^1`$ . All other coefficients are collected in the Table I.
The coefficients $`b_i`$ determine the value of the PBH mass beginning from which particles of $`i`$-type can be evaporated. This value, $`M_{BH}^b`$, is obtained from the relations
$$\frac{10^{13}}{M_{BH}^b(g)}T_{BH}^b(GeV)\frac{m_i}{3}.$$
(41)
So,
$$b_i=log_{10}M_{BH}^b.$$
(42)
The resulting function $`\alpha (m)`$ is shown on Fig.1 For comparison, the corresponding dependence from the work (drawn using the Eq.7 of ) is also shown.
The formula (31), as it stands, takes into account the contribution to the neutrino background solely from a direct process of the neutrino evaporation. If we suppose that particles evaporated by a black hole propagate freely, the calculation of other contributions to the neutrino background can be performed using our knowledge of particle physics .
On Fig.2 the typical result of our calculation of instantaneous neutrino spectra from evaporating black hole is shown. The spectrum of the straightforward (direct) $`(\nu _e+\stackrel{~}{\nu }_e)`$ \- emission is given by the Hawking function $`f_H(E,m)`$, Eq.(27). The $`(\nu _e+\stackrel{~}{\nu }_e)`$ spectrum arising from decays of $`(\mu ^++\mu ^{})`$ evaporated directly is calculated by the formula
$$f^{(\mu )}(E,m)=f_H(E_\mu ,m)\frac{dn_\nu (E_\mu ,E)}{dE}𝑑E_\mu ,$$
(43)
where $`dn_\nu ^\mu /dE`$ is the neutrino spectrum in a $`\mu `$ \- decay. To evaluate the electron neutrino spectrum resulted from fragmentations of evaporated quarks, we used the simplest chain:
$$(u,d)\text{quarks}\pi \mu \nu _e,$$
(44)
and the formula
$$f^{(q)}(E,m)=f_H(E_q,m)\frac{dn_\pi ^q(\xi )}{d\xi }\frac{d\xi }{dE_q}\frac{dn_\mu ^\pi (E_\mu ,E_\pi )}{dE_\mu }\frac{dn_\nu ^\mu (E_\mu ,E)}{dE}𝑑E_q𝑑E_\pi 𝑑E_\mu .$$
(45)
Here, $`dn_\pi ^q/d\xi `$ is the $`q\pi `$ fragmentation function, for which the simple parametrization was taken:
$$\frac{dn_\pi ^q}{d\xi }=\frac{15}{16}(\xi 1)^2\xi ^{3/2},\xi =\frac{E_\pi }{E_q},$$
(46)
and $`dn_\mu ^\pi /dE_\mu `$ is the $`\mu `$ \- spectrum in a decay $`\pi \mu +\nu _\mu `$.
Analogous formulas are used for calculations of the neutrino spectrum from other channels of the neutrino production, for instance from the decays of evaporated $`W`$-bosons ($`We+\nu _e,`$ $`W\mu \nu _e`$).
The relative contribution of different channels to the total $`\nu _e`$ spectrum depends on the black hole temperature. One can see from Fig.2, that at high temperature decays of massive particles evaporated by the black hole become very important at high energy tail of the spectrum (if the corresponding branching ratios are not too small).
The total instantaneous neutrino spectrum from a black hole evaporation is given by the sum
$$f(E,m)=f_H(E,m)+f^{(\mu )}(E,m)+f^{(q)}(E,m)+\mathrm{},$$
(47)
and the total background neutrino spectrum is given by the same Eq.(31), except the change $`f_H(E,m)f(E,m)`$.
## IV Constraints on the spectral index
The spectral index of initial density fluctuations is defined by the relations
$`\sigma _r^2={\displaystyle \frac{1}{V_W^2}}{\displaystyle \frac{d^3k}{(2\pi )^3}\left|\delta _k\right|^2W_k^2(r)},`$ (48)
(49)
$`\left|\delta _k\right|^2=Ak^n.`$ (50)
Here, $`\delta _k`$ and $`W_k`$ are Fourier transforms of the density field $`\delta (\stackrel{}{x})`$ and the window function of comoving size $`r`$, respectively, $`V_r`$ is the effective volume filtered by $`W_r`$.
One obtains from Eq.(48):
$$\sigma _R^2(t)A\frac{k_{fl}^3}{2\pi ^2}k_{fl}^n,$$
(51)
where $`k_{fl}`$ is the comoving wave number, characterizing the perturbed region,
$$k_{fl}=\frac{a(t)}{R},$$
(52)
and $`R`$ is the physical size of this region at arbitrary moment $`t`$. Using the connection of $`k_{fl}`$ with the fluctuation mass,
$$M(t)=\frac{4}{3}\pi \left(\frac{a(t)}{k_{fl}}\right)^3\rho (t),$$
(53)
and introducing the horizon mass $`M_h`$ (which is equal to the fluctuation mass at the moment when the fluctuated region crosses horizon) we can rewrite Eq.(48) in the form:
$$\sigma _R(t)=\left(\frac{M_{hor}(t)}{M(t)}\right)^{2/3}\sigma _H(M_h).$$
(54)
Here, $`M_{hor}`$ is the horizon mass at $`t`$ , $`\sigma _H(M_h)`$ is the standard deviation at horizon crossing.
At the initial moment of time one has :
$$M_{hor}(t_i)=M_i,M(t_i)=M.$$
(55)
The form of the $`\sigma _H(M_h)`$ function depends on the spectral index:
$$\sigma _H\{\begin{array}{ccc}M_h^{\frac{1n}{4}}\hfill & ,& \text{radiation dominance}\hfill \\ M_h^{\frac{1n}{6}}\hfill & ,& \text{matter dominance}.\hfill \end{array}$$
(56)
Eqs.(11) in Sec.II are obtained from Eqs. (54)-(56).
From COBE data we know the normalization of $`\sigma _H(M_h)`$ at present horizon size :
$$\sigma _H(M_{h,0})=9.510^5;M_{h,0}=10^{56}g.$$
(57)
Now one can easily show that $`\sigma _H(M_h)`$ for radiation dominance case is connected with $`\sigma _H(M_{h,0})`$ by the following approximate relation:
$$\sigma _H(M_h)=\sigma _H(M_{h,0})\left(\frac{M_{h,0}}{M_{h,eq}}\right)^{\frac{n1}{6}}\left(\frac{M_h}{M_{h,eq}}\right)^{\frac{1n}{4}}.$$
(58)
Our calculation of neutrino spectra from evaporating PBHs contains two parameters: a spectral index $`n`$ and a time of an end of the inflation $`t_i`$ (which, by assumption, is a time when density fluctuations develop). We assume that at $`t_i`$ the universe has as a result of the reheating the equilibrium temperature $`T_{RH}`$. The connection of $`T_{RH}`$ and $`t_i`$ is given by the standard model (in formulas of this section we use the convention $`\mathrm{}=c=1`$):
$$t_i=0.301g_{}^{1/2}\frac{M_{pl}}{T_{RH}^2}\frac{0.24}{T_{RH}^2(MeV)}s$$
(59)
($`g_{}100`$ is the number of the degrees of freedom in the early universe).
On Fig.3 the integrand of the background spectrum integral (Eq.(38)) is shown as a function of redshift $`z`$ for several values of the parameter $`T_{RH}`$. Each curve on Fig.3 has a strong cut off on some redshift value. This feature is connected with the existence in our model the minimum value of PBH mass,
$$M_{BH}^{min}=\gamma ^{1/2}M_i.$$
(60)
The PBH mass spectrum is steeply falling function of the mass, so the masses near minimum give a largest contribution to the neutrino background. The moment of their evaporation is, approximately,
$$t_{ev}\frac{\left(M_{PBH}^{min}\right)^3}{3\alpha },$$
(61)
and the corresponding redshift is determined by the relation
$$z_{ev}+1(z_{eq}+1)\left(\frac{t_{eq}}{t_{ev}}\right)^{1/2}.$$
(62)
Larger masses evaporate at larger times and smaller redshifts. If, for instance, $`T_{RH}=10^{10}\text{GeV}`$, one has $`t_i=0.2410^{26}\text{s}`$ ; $`M_i=7,510^{10}\text{g}`$ , $`z_{ev}10^7`$. So, at $`T_{RH}=10^{10}GeV`$ the redshifts of order of $`10^7`$ give a largest contribution to the neutrino background, and this is clearly seen at Fig3.
The cut off value $`z_{ev}`$ strongly depends on $`T_{RH}`$:
$$z_{ev}\frac{1}{\sqrt{t_{ev}}}\left(M_{BH}^{min}\right)^{3/2}t_i^{3/2}T_{RH}^3.$$
(63)
On Figs.4 spectra of electron neutrino background are shown separately for several channels of the neutrino production. One can see from these figures that the contribution to the summary background from the direct $`\nu _e`$ emission is dominant (at least, for $`E10\text{ MeV}`$) and the relative importance of different channels changes with an increase of $`T_{RH}`$. It is seen also that the contribution of the quark fragmentation channel is very small at $`E10\text{ MeV}`$ $`(1\%)`$, so the evident underestimation of this channel in our calculation, connected with the neglect of the contribution of heavy quark fragmentations, has no particular importance.
Some typical results of electron neutrino background calculations are shown on Figs.5,6 The main features of such spectra have been revealed in previous works : $`E^3`$ dependence above $`100\text{ MeV}`$, the flattening out of the spectra at lower energies and the shift of the turnover energy with a change of the $`T_{RH}`$ value . According to Figs.5,6 the turnover energy is about $`1\text{ MeV}`$ at $`T_{RH}=10^9\text{GeV}`$.
For an obtaining of the constraints on the spectral index we use three types of neutrino experiments.
1. Radiochemical experiments for the detection of solar neutrinos. There are data from the famous Davis experiment and the $`GaGe`$ experiment . The cross section for the neutrino absorption via a bound-bound transition was calculated using the approximate formula
$$\sigma =\{\begin{array}{ccc}\frac{G_F^2}{\pi }\left(1^2+\left(\frac{g_a}{g_v}\right)^2\sigma ^2\right)p_eE_e\hfill & ,& E<100MeV\hfill \\ \text{const.}\hfill & ,& E>100MeV.\hfill \end{array}$$
(64)
In the case of the Davis experiment ($`ClAr`$ reaction) we take into account the super-allowed transition only, for which
$$1^2=3,\sigma ^2=0.2,E_{thr}5MeV.$$
(65)
In the $`GaGe`$ case the main contribution gives the ground state - ground state transition ($`E_{thr}=0.242MeV`$). The corresponding cross section is, for $`E<100MeV`$,
$$\sigma \left(\nu +GaGe+e\right)0.64610^{44}p_eE_e.$$
(66)
The number of target atoms $`N^T`$ is about $`2.210^{30}`$ for the $`ClAr`$ experiment and $`10^{29}`$ for the $`GaGe`$ experiments . The average statistics is $`1.5day^1`$ ($`ClAr`$) and $`1day^1`$ ($`GaGe`$). The constraint is calculated using the relation
$$4\pi N^T10^5S(E)\sigma (E)𝑑E<1.$$
(67)
2. The experiment on a search of an antineutrino flux from the Sun . In some theoretical schemes (e.g., in the model of a spin - flavor precession in a magnetic field) the Sun can emit rather large flux of antineutrinos. LSD experiment sets the upper limit on this flux, $`\mathrm{\Phi }_{\stackrel{~}{\nu }}/\mathrm{\Phi }_\nu 1.7\%`$. In this experiment the neutrino detection is carried out using the reaction
$$\stackrel{~}{\nu }_e+pn+e^+.$$
(68)
The number of target protons is $`8.610^{28}`$ per 1 ton of the scintillation detector, and the obtained upper limit is $`0.28`$ antineutrino events per year per ton . The cross section of the reaction (68) is well known (see, e.g.,). It grows with the neutrino energy up to $`E2GeV`$, and is a constant ($`0.510^{38}cm^2`$) at larger energies.The product of this cross section and a PBH antineutrino background spectrum has the maximum at $`E100MeV`$. The constraint is determined from the condition that the calculated effect in the LSD detector is smaller than the upper limit obtained in ref.
3. The Kamiokande experiment on a detection of atmospheric electron neutrinos . In this experiment the electrons arising in the reaction
$$\nu _e^{atm}+np+e^{}$$
(69)
in the large water Cherenkov detector were detected and, moreover, their energy spectrum was measured. This spectrum has a maximum at the energy about $`300MeV`$. The spectrum of the atmospheric electron neutrinos is calculated(see, e.g.,) with a very large accuracy (assuming an absence of the neutrino oscillations) and the experimentally measured electron spectrum coincides, more or less, with the theoretical prediction. The observed electron excess at $`E100MeV`$ (which is a possible consequence of the oscillations) is not too large. We use the following condition for an obtaining the our constraint : the absolute differential intensity of the PBH neutrino background at the neutrino energy $`E0.3GeV`$ cannot exceed the theoretical differential intensity of the atmospheric electron neutrinos at the same energy (otherwise the total electron energy spectrum is strongly different from the observed one).
## V Results and discussions
Fig.7 shows our results for the spectral index constraints. It is seen that the best constraints are obtained using the Kamiokande atmospheric neutrino data and the LSD upper limit on an antineutrino flux from the Sun.
The behavior of the constraint curve $`n(T_{RH})`$ is sharply different from that was obtained in ref. (authors of ref. used the initial PBH mass function following from the near critical collapse scenario with a domination of the earliest epoch of PBHs formation and diffuse extragalactic photon data).
The slightly non-monotonic behavior of the curves Fig.7 near $`T_{RH}310^8\text{GeV}`$ is connected with the sharp increase of the function $`\alpha (m)`$ at $`m10^{14}g`$ due to a switching on the quark evaporations (see Fig.1). Indeed, according to the relations Eq.(63), the minimum value of PBH mass in the initial mass spectrum is inversely proportional to $`T_{RH}^2`$. More exactly, the relation between $`M_{BH}^{min}`$ and $`T_{RH}`$ is
$$M_{BH}^{min}(g)=\frac{7.210^{30}}{T_{RH}^2(GeV)}.$$
(70)
Thus, when the parameter $`T_{RH}`$ passes over the value $`310^8GeV`$, $`M_{BH}^{min}`$ becomes smaller than $`10^{14}g`$, and the value of $`\alpha `$ used in Eqs.(30-31) increases on $`20`$ units.
One can compare our spectral index constraints with the corresponding results of ref. , where the same initial mass spectrum of PBHs had been used. Some difference in resulting constraints ($`\mathrm{\Delta }n0.02`$) may be connected simply with the fact that authors of ref. use slightly different formula for $`\sigma _H(M_h)`$, namely,
$$\sigma _H(M_h)=\sigma _H(M_{h,0})\left(\frac{M_h}{M_{h,0}}\right)^{\frac{1n}{4}}.$$
(71)
In other respects the constraints are quite similar although in ref. they were based on diffuse extragalactic photon background data.
Constraints on the spectral index, obtained using the neutrino background calculations, are sensitive to a possible existence of a QCD photosphere around black holes only in one case: if the density of the outward-propagating plasma is so high that it is opaque for neutrinos. It was shown above (see fig4) that the fragmentations of evaporated quarks are inessential and can be neglected when calculating the electron neutrino background (in spite of the fact that neutrinos from these fragmentations are very abundant (fig.2)). The reason is simple: neutrino energy is strongly redshifted during cosmological expansion. It is clear that the fragmentations of photosphere’s quarks are also inessential especially as energies of these quarks are, on average, lower than energies of quarks evaporated directly. Thus, only the neutrino opacity of the photosphere can modify our results. According to a general phylosophy of a photosphere’s model an opaque ”neutrino photosphere” eventually forms , but it may happen too late, when the black hole mass is already too small to give noticeable contribution to the integral in Eq.(31).
A precise calculation of the PBH neutrino background must include also a taking into account of the neutrino absorption during a travelling in the space. In the case of the photon background the most important absorption process is a pair production on neutral matter due to which the photons from PBHs evaporated earlier than $`z700`$ are absent today. In our case, the analog of an optical depth of the universe for the neutrino emitted at a redshift z and having today an energy $`E`$ is given by the integral
$$\tau (z,E)=c\underset{0}{\overset{z}{}}\sigma \left(E(1+z^{})\right)n(z^{})\frac{dt}{dz^{}}𝑑z^{}.$$
(72)
Here, $`\sigma (E)`$ is the neutrino interaction cross section, $`n(z)`$ is a number density of the target particles.
Two processes are potentially ”dangerous”: neutrino-nucleon inelastic scattering growing linearly with an energy, and annihilations with neutrinos of the relic background:
$`\nu _e+Ne^{}+anything,`$ (73)
(74)
$`\nu _e+\stackrel{~}{\nu }_e(relic){\displaystyle \underset{i}{}}(f_i+\stackrel{~}{f}_i).`$ (75)
Here, $`f_i`$ are charged fermions (leptons and quarks). As is known, the relic neutrino background exists, with a Planck distribution, from the epoch of neutrino decoupling ($`z10^{10}`$).
For estimations of the neutrino absorption it is enough to use some characteristic value of a neutrino energy. For the neutrino processes in the detectors, discussed in the previous Section, such value is about $`100MeV`$. In the case of the $`\nu N`$-scattering one has, approximately,
$$\sigma _{\nu N}(E)10^{38}\left(\frac{E}{\text{GeV}}\right)cm^2,$$
(76)
$$\frac{n_N(z)}{(1+z)^3}=n_{0N}(10^610^7)cm^3.$$
(77)
Using these values and the formula (30) for $`dt/dz`$, one can easily estimate the integral (72) for $`\tau (z,E)`$. One obtains the important result:
$$\tau _{\nu N}(z10^7,E=100MeV)1.$$
(78)
In the case of the absorption through annihilations of evaporated neutrinos with relic neutrinos the integral (72) reduces to
$$\tau _{\nu \stackrel{~}{\nu }}(z,E)0.310^{34}\left(\frac{E}{\text{TeV}}\right)\frac{1}{H_0}\underset{0}{\overset{z}{}}(1+z^{})^5\frac{dt}{dz^{}}𝑑z^{}.$$
(79)
This gives
$$\tau _{\nu \stackrel{~}{\nu }}(z210^6,E=100\text{MeV})1.$$
(80)
From here one can conclude that the neutrino absorption becomes important (for the problems considered in this paper) at $`z210^6`$. It was shown in Sec.IV that at $`T_{RH}=10^{10}GeV`$ the redshifts $`10^7`$ give a maximum contribution to the neutrino background (see Fig.3). Besides, it was shown that, in general, a redshift value corresponding to such a maximum is proportional to $`T_{RH}^3`$ (Eq.(63)). So, the value $`z=210^6`$ corresponds to $`T_{RH}610^9GeV`$, and, therefore, for $`T_{RH}>610^9GeV`$ the neutrino absorption effects are important.
One should note, at the end, that usually the calculations of these constraints are accompanied by the calculation of the bounds based on requirement that the energy density in PBHs does not overclose the universe at any epoch ($`\mathrm{\Omega }_{BH}<1`$). For a setting of such bounds one must consider the cosmological evolution of the system PBHs + radiation. We intend to carry out these calculations in a separate paper.
###### Acknowledgements.
We wish to thank G. V. Domogatsky for valuable discussions and comments. We are grateful also to H.I.Kim for informing us about his work on the same problem and for useful remarks.
|
warning/0005/nucl-th0005022.html
|
ar5iv
|
text
|
# Effects of the Neutron Spin-Orbit Density on Nuclear Charge Density in Relativistic Models
## Abstract
The neutron spin-orbit density contributes to the nuclear charge density as a relativistic effect. The contribution is enhanced by the effective mass stemming from the Lorentz-scalar potential in relativistic models. This enhancement explains well the difference between the cross sections of elastic electron scattering off <sup>40</sup>Ca and <sup>48</sup>Ca which was not reproduced in non-relativistic models. The spin-orbit density will be examined in more detail in electron scattering off unstable nuclei which would be available in the future.
At present there are two kinds of phenomenological models which explain well nuclear structure and reactions. The one is a conventional non-relativistic model which assumes phenomenological interactions between nucleons, like Skyrme forces. The other is a relativistic model which takes into account explicitly meson-exchanges between nucleons, as the $`\sigma \omega `$ model. The relationship between these models, however, is not clear. In the relativistic model the effective mass coming from the Lorentz scalar potential, which is absent in non-relativistic models, plays a crucial role in understanding nuclear properties. For example, the effective mass enhances the spin-orbit force which is responsible for the spin-orbit splitting of the single-particle levels and polarization phenomena of hadron-nucleus scattering. The nuclear saturation density and binding energy are also dominated by the effective mass. In non-relativistic models those quantities are explained by taking into account other kinds of many-body correlations. It is one of the important questions in nuclear physics which model is realistic. The best way to answer this question is to find fundamental physical quantities which can be explained in the one model, but not in the other model. So far, however, such quantities are not found, as far as the authors know.
The purpose of this paper is to show that the neutron spin-orbit density is very sensitive to the effective mass of the relativistic model. The effective mass yields an additional spin-orbit density as a relativistic correction, which contributes to the nuclear charge density, and the effects are seizable in the form factors of elastic electron scattering. This correction explains well the existing data which were not reproduced in previous non-relativistic models. We will also show that the effective mass effects will be able to be explored in more detail in unstable nuclei.
The neutron spin-orbit charge density is due to the Pauli current. It was first discussed by Bertozzi et al. in a non-relativistic framework. The nuclear wave function was obtained in a two-component model and the relativistic correction was derived by expanding the free nucleon current in terms of $`1/M`$, $`M`$ being the mass of the free nucleon. They calculated the cross sections for elastic electron scattering off <sup>40</sup>Ca and <sup>48</sup>Ca, and found that the relativistic correction was not negligible, but not enough to explain the difference between the two cross sections. Later Miller analyzed the same data using a relativistic model, but again could not reproduce the experimental data. The Pauli current made rather worse the agreement between his results and experiment.
Our model discussed below is in principle the same as Miller’s one, but we will calculate the spin-orbit density, using a different relativistic model developed later by Horowitz and Serot. Moreover we will make clear the relationship between the relativistic model and non-relativistic models. The effective mass effects, which are peculiar to the relativistic model, on the spin-orbit current will be clarified.
We calculate the cross section for elastic electron scattering using phase shift analyses. For this purpose we have to obtain the nuclear charge density, which is given by
$$\rho _c(r)=\frac{d^3q}{(2\pi )^3}\mathrm{exp}(i𝐪𝐫)0|\widehat{\rho }(𝐪)|0,$$
(1)
where $`𝐪`$ denotes the momentum transfer from the electron to the nucleus. In the relativistic theory, the ground-state expectation value of the time-component of the nuclear current is given by
$`0|\widehat{\rho }(𝐪)|0`$ $`=`$ $`0|{\displaystyle \underset{k}{}}\mathrm{exp}(i𝐪𝐫_k)`$ (3)
$`\times \left(F_{1k}(𝐪^2)+{\displaystyle \frac{\mu _k}{2M}}F_{2k}(𝐪^2)𝐪𝜸_k\right)|0,`$
where $`F_{1k}(𝐪^2)`$ and $`F_{2k}(𝐪^2)`$ stand for the Dirac and Pauli form factors of the nucleon, respectively, and $`\mu _k`$ the anomalous magnetic moment. The above equation is rewritten by using the Sachs form factor, $`G_E(𝐪^2)`$, as
$`0|\widehat{\rho }(𝐪)|0`$ $`=`$ $`{\displaystyle d^3x\mathrm{exp}(i𝐪𝐱)}`$ (5)
$`\times {\displaystyle \underset{\tau }{}}(G_{E\tau }(𝐪^2)\rho _\tau (x)+F_{2\tau }(𝐪^2)W_\tau (x))`$
$`=`$ $`{\displaystyle d^3xd^3y\mathrm{exp}(i𝐪(𝐱+𝐲))}`$ (7)
$`\times {\displaystyle \underset{\tau }{}}(G_{E\tau }(y)\rho _\tau (x)+F_{2\tau }(y)W_\tau (x)),`$
where the sum of $`\tau `$ is performed with respect to the proton and the neutron, $`\tau =p,n`$. The functions, $`G_{E\tau }(y)`$ and $`F_{2\tau }(y)`$, are obtained by the inverse Fourier transformation of the Sachs and Pauli form factors, respectively. The nucleon density, $`\rho _\tau (x)`$, and the spin-orbit density, $`W_\tau (x)`$, are given by
$`\rho _\tau (r)`$ $`=`$ $`0|{\displaystyle \underset{k}{}}\delta (𝐫𝐫_k)|0,`$ (8)
$`W_\tau (r)`$ $`=`$ $`{\displaystyle \frac{\mu _\tau }{2M}}({\displaystyle \frac{1}{2M}}\mathbf{}^2\rho _\tau (r)`$ (10)
$`+i\mathbf{}0|{\displaystyle \underset{k}{}}\delta (𝐫𝐫_k)𝜸_k|0),`$
where the sum over $`k`$ is performed up to $`Z`$ for $`\tau =p`$ and up to $`N`$ for $`\tau =n`$. By inserting Eq. (7) into Eq. (1), the nuclear charge density is given by
$$\rho _c(r)=\underset{\tau }{}\left(\rho _{c\tau }(r)+W_{c\tau }(r)\right),$$
(11)
where the nucleon charge density, $`\rho _{c\tau }(r)`$, and the spin-orbit charge density, $`W_{c\tau }(r)`$, are written as
$`\rho _{c\tau }(r)`$ $`=`$ $`{\displaystyle \frac{1}{r}}{\displaystyle _0^{\mathrm{}}}𝑑xx\rho _\tau (x)`$ (13)
$`\times \left(g_\tau (|rx|)g_\tau (r+x)\right),`$
$`W_{c\tau }(r)`$ $`=`$ $`{\displaystyle \frac{1}{r}}{\displaystyle _0^{\mathrm{}}}𝑑xxW_\tau (x)`$ (15)
$`\times \left(f_{2\tau }(|rx|)f_{2\tau }(r+x)\right)`$
with
$`g_\tau (x)`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle _{\mathrm{}}^{\mathrm{}}}𝑑qe^{iqx}G_{E\tau }(𝐪^2),`$
$`f_{2\tau }(x)`$ $`=`$ $`{\displaystyle \frac{1}{2\pi }}{\displaystyle _{\mathrm{}}^{\mathrm{}}}𝑑qe^{iqx}F_{2\tau }(𝐪^2).`$
In order to calculate the nucleon and the spin-orbit density, we take the Horowitz-Serot model, where the ground state wave function is described with the mean field approximation. The single-particle wave function is written as
$`\psi _{\alpha m}=\left(\begin{array}{c}i{\displaystyle \frac{G_\alpha (r)}{r}}𝒴_{\mathrm{}jm}\\ \\ {\displaystyle \frac{F_\alpha (r)}{r}}{\displaystyle \frac{𝝈𝐫}{r}}𝒴_{\mathrm{}jm}\end{array}\right),`$
The large and small components in the present model satisfy the Dirac equation:
$`{\displaystyle \frac{dG_\alpha }{dr}}`$ $`=`$ $`{\displaystyle \frac{\kappa _\alpha }{r}}G_\alpha +\left(\epsilon _\alpha U_\tau (r)+M^{}(r)\right)F_\alpha ,`$ (16)
$`{\displaystyle \frac{dF_\alpha }{dr}}`$ $`=`$ $`{\displaystyle \frac{\kappa _\alpha }{r}}F_\alpha \left(\epsilon _\alpha U_\tau (r)M^{}(r)\right)G_\alpha ,`$ (17)
where $`\kappa _\alpha =(1)^{j\mathrm{}+1/2}(j+1/2)`$ denotes the eigenvalue of $`(1+𝝈\mathbf{})`$, and $`M^{}(r)`$ the nucleon effective mass given by
$`M^{}(r)=MU_s(r).`$
The Lorentz scalar potential, $`U_s(r)`$, comes from the $`\sigma `$-meson exchanges between nucleons, while the Lorentz vector potential, $`U_\tau (r)`$, is due to the $`\omega `$\- and $`\rho `$-mesons and photons in the present model. Then the nucleon density in Eq. (8) is given by
$`\rho _\tau (r)={\displaystyle \underset{\alpha }{}}{\displaystyle \frac{2j_\alpha +1}{4\pi r^2}}\left(G_\alpha ^2+F_\alpha ^2\right).`$ (18)
On the other hand, the spin-orbit density, $`W_\tau (r)`$, in Eq. (10) is described as
$`W_\tau (r)`$ $`=`$ $`{\displaystyle \frac{\mu _\tau }{M}}{\displaystyle \underset{\alpha }{}}{\displaystyle \frac{2j_\alpha +1}{4\pi r^2}}{\displaystyle \frac{d}{dr}}({\displaystyle \frac{MM^{}(r)}{M}}G_\alpha F_\alpha `$ (20)
$`+{\displaystyle \frac{\kappa _\alpha +1}{2Mr}}G_\alpha ^2{\displaystyle \frac{\kappa _\alpha 1}{2Mr}}F_\alpha ^2).`$
The relationship between the relativistic model and non-relativistic models is very clear. In non-relativistic models, usually the neutron charge and spin-orbit charge densities are neglected in estimating the electron scattering cross section. Bertozzi et al. took into account the neutron charge density and a part of the spin-orbit charge density in the non-relativistic framework. Their spin-orbit density is obtained from Eq. (20) by setting $`M^{}(r)=M`$ and neglecting $`F_\alpha ^2`$-term,
$`W_\tau (r)`$ $``$ $`{\displaystyle \frac{\mu _\tau }{2M^2r^2}}{\displaystyle \frac{d}{dr}}r{\displaystyle \underset{\alpha }{}}{\displaystyle \frac{2j_\alpha +1}{4\pi r^2}}\left(\kappa _\alpha +1\right)G_\alpha ^2`$ (21)
$``$ $`{\displaystyle \frac{1}{r^2}}{\displaystyle \frac{d}{dr}}r0|{\displaystyle \frac{\mu _\tau }{2M^2}}{\displaystyle \underset{k}{}}\delta (𝐫𝐫_k)𝝈_k\mathbf{}_k|0.`$ (22)
We will show that the spin-orbit density due to the effective mass in Eq. (20) is very important in the relativistic model for reproducing the experimental data.
The nucleon form factors used in the present calculations are the dipole type according to Ref. with more recent experimental data for the neutron. The Sachs form factor for the proton is given by
$`G_{Ep}={\displaystyle \frac{1}{\left(1+r_p^2𝐪^2/12\right)^2}},r_p=r^2^{1/2}=0.81\text{fm},`$
while the one for the neutron by
$`G_{En}`$ $`=`$ $`{\displaystyle \frac{1}{\left(1+r_+^2𝐪^2/12\right)^2}}{\displaystyle \frac{1}{\left(1+r_{}^2𝐪^2/12\right)^2}},`$
$`r_\pm ^2`$ $`=`$ $`(0.9)^20.06\text{fm}^2.`$
The Pauli form factors for the proton and the neutron are
$`F_{2p}={\displaystyle \frac{G_{Ep}}{1+𝐪^2/4M^2}},F_{2n}={\displaystyle \frac{G_{Ep}G_{En}/\mu _n}{1+𝐪^2/4M^2}}.`$
The values of the anomalous magnetic moment are given by $`\mu _p=1.793`$ and $`\mu _n=\mathrm{\hspace{0.17em}1.913}`$.
Now, we calculate the differential cross sections $`\sigma (\theta )`$ for elastic electron scattering off <sup>40</sup>Ca and <sup>48</sup>Ca and compare their difference $`D(\theta )`$ given by
$$D(\theta )=\frac{\sigma _{40}(\theta )\sigma _{48}(\theta )}{\sigma _{40}(\theta )+\sigma _{48}(\theta )},$$
(23)
with experiment. First we show in Fig. 1 results of the non-relativistic models, since Bertozzi et al. did not compare explicitly their results including both the nucleon charge density and the spin-orbit charge density with experiment. In Fig. 1, the thin solid curve shows the result taking into account the only proton charge density, which is calculated with the Skyrme force I in the Hartree-Fock approximation. This is almost the same as the result of Bertozzi et al.. When we include also the neutron charge density and the proton and neutron spin-orbit charge density in Eq. (22), we obtain the solid curve in Fig. 1. It is seen that the the discrepancy between the theory and the experiment is a little reduced, in particular, at the electron scattering angle around $`\theta =60^{}`$ to 90. The improvement is mainly due to the spin-orbit charge density from the neutrons in the $`1f_{7/2}`$ shell, but is not enough to explain the experimental data.
Fig. 2 shows the results of the Horowitz-Serot model. The thin solid curve is obtained by taking into account the only proton charge density, while the solid curve by the full density, Eq. (11). In the relativistic model, the proton charge density itself improves results of the non-relativistic model, and the experimental data are almost reproduced by taking into account the neutron spin-orbit charge density enhanced by the effective mass. In this figure, we show, as a reference, by the dotted curve the results of the PWBA calculations using the full density and the effective momentum transfer to simulate DWBA. We note that both in non-relativistic and relativistic models, the center of mass correction to the cross section are negligible in Ca isotopes.
As seen in Eq. (20), the effects of the spin-orbit charge density appears, when the sub-shell is occupied by the neutrons; In closed shell nuclei, the effects disappear. Moreover, if protons also occupy the subshell as in <sup>208</sup>Pb, the proton and neutron spin-orbit densities almost cancel each other as in non-relativistic model, since the anomalous magnetic moment of the proton has the opposite sign to that of the neutron. Another interesting result of the spin-orbit density is found in neutron rich nuclei. In Fig. 3 shown the results with respect to <sup>40</sup>Ca and <sup>52</sup>Ca in the same designation as in Fig. 2. We see that effects from the spin-orbit charge density of the $`1f_{7/2}`$ neutrons are almost cancelled by those from the $`2p_{3/2}`$ neutrons. The similar results are obtained in Zr isotopes. The effect of the neutron spin-orbit charge density is enhanced in the cross section of <sup>90</sup>Zr, compared with the one of <sup>80</sup>Zr, but disappears in <sup>96</sup>Zr. It is interesting to observe experimentally these predictions of the relativistic model in electron scattering off unstable nuclei which is planned in RIKEN.
In conclusion, the effective mass due to the Lorentz scalar potential is a necessary ingredient of the relativistic model. It enhances the neutron spin-orbit charge density in a peculiar way to this model. The enhanced density explains well the difference between the cross sections of elastic electron scattering off <sup>40</sup>Ca and <sup>48</sup>Ca, which was not reproduced in previous non-relativistic models. Electron scattering off unstable nuclei is desirable in order to explore in more detail the effective mass effects. More detailed discussions on the results of non-relativistic and relativistic models will be published elsewhere.
###### Acknowledgements.
The authors would like to thank Dr. T. Suda for useful discussions.
|
warning/0005/astro-ph0005047.html
|
ar5iv
|
text
|
# A Candidate Substellar Companion to HR 7329
## 1 Introduction
The discovery of substellar objects in stellar systems is a key goal in contemporary astronomy, and an essential element in furthering our knowledge of the mass function of binary star and planetary system formation. The substellar mass range from 10 to 80 Jupiter masses (0.01 $``$ 0.08 M) is crucial to our understanding of the bridge between the lowest mass stars and the giant planets. To this end, the Near Infrared Camera and Multi-Object Spectrometer (NICMOS) Instrument Definition Team (IDT) has conducted an infrared coronagraphic survey of young, main-sequence stars to search for substellar companions. Substellar objects cool with age because they do not sustain hydrogen fusion and are more difficult to detect with time as they become fainter (e.g. Burrows et al. 1997). Using independently determined ages and distances for the target stars, the masses of newly-detected secondaries can be ascertained from infrared fluxes and theoretical evolutionary tracks on the H$``$R diagram. Follow-up spectroscopy further constrains the effective temperature and probability of companionship. Here we present a spectrum obtained with the Space Telescope Imaging Spectrograph (STIS) of a substellar companion candidate, HR 7329B, from our NICMOS imaging survey. Previously, this survey revealed TWA5B, a $``$ 20 Jupiter mass brown dwarf companion to TWA 5A (Lowrance et al. 1999).
## 2 NICMOS
### 2.1 Observations
HR 7329 (HD 181296; A0V; d$``$48pc; V=5.05; RA=19h 22m 51.2s, DEC=$``$54 25 26<sup>′′</sup> (J2000.0)) was observed with NICMOS on 1998 June 29, 16:15 $``$ 17:17 UT. We obtained multiple-exposure images with the star behind the coronagraph (radius = 0.3$`\mathrm{}`$) on Camera 2 (pixel scale = $``$0.076 arcsec pixel<sup>-1</sup>) and a wide-band F160W filter (central wavelength: 1.5940 $`\mu `$m, $`\mathrm{\Delta }\lambda `$ = 0.4030 $`\mu `$m), which corresponds closely to the central wavelength of a Johnson H-band photometric filter. Five standard NICMOS STEP16 MultiAccum (non-destructive read) integrations (MacKenty et al. 1997) totaling 719.6s were executed at each of two orientations differing by 29.9 degrees. While the stellar point-spread-function (PSF), the instrumental scattering function, and detector artifacts rotate with the aperture, any real features in the unocculted area of the detector will be unaffected by a change in the camera orientation. Subtraction of these two images has been shown to significantly reduce residual PSF background light (Schneider et al. 1998). The NICMOS coronagraphic images were reduced and processed utilizing calibration darks and flat-fields created by the NICMOS IDT from on-orbit observations following the method described in Lowrance et al. (1999).
### 2.2 Results
Subtraction and analysis of the NICMOS coronagraphic images reveal a stellar-like object (HR 7329B) at a separation of 4.17$`\mathrm{}`$ $`\pm `$ 0.05, and a position angle of 166.8 $`\pm `$ 0.2 from HR 7329 (HR 7329A) (Figure 1). This secondary is point-like with a Full Width Half Maximum (FWHM) of 0.$`\mathrm{}`$15 (the diffraction limit is 0.$`\mathrm{}`$14) with the first Airy ring apparent in Figure 1. Since the target star is occulted in the NICMOS coronagraphic images, its position is ascertained from the target acquisition image and located behind the coronagraph by a known telescope offset.
The secondary fell near the edge of the field of view in the second orientation, so the magnitude of HR 7329B was measured using a 12 pixel radius circular aperture centered on the companion in the subtracted image from the first orientation. A correction factor of 9.66$`\%`$, determined from coronagraphic photometric aperture corrections developed by the NICMOS IDT, was applied to compensate for the flux which fell out of this aperture. The \[F160W\] magnitude of HR 7329B is then 11.90 $`\pm `$ 0.06 mag using a conversion factor for the F160W filter of 2.19 $`\times 10^6`$ Jy ADU<sup>-1</sup>s, and 1083 Jy corresponding to an H magnitude of zero in the Vega system (Rieke 1999) where the majority of the uncertainty is dominated in NICMOS’s calibration in relation to standard stars. The F160W filter is $`30\%`$ wider than ground-based Johnson H-band filters which necessitates a careful conversion from F160W to H band for cool temperature objects. For six M dwarfs between spectral types M6 and M9 with measured F160W and ground-based H-band magnitudes, we find a mean difference of 0.03 $`\pm `$ 0.02 mag. For HR 7329B, a M7.5, we thus expect the H-\[F160W\] color to be about 0.03 mag. Making this color correction, we estimate an H magnitude of 11.93 $`\pm `$ 0.06 mag.
The \[F187N\] magnitude of HR 7329A was determined from aperture photometry of the two calibrated target acquisition images (at each of the two spacecraft orientations) processed as described in Lowrance et al. (1999). Within the uncertainties, the two measurements agreed and were averaged to yield \[F187N\] $`=`$ 5.0 $`\pm `$ 0.1 mag.
## 3 STIS
### 3.1 Observations
HR 7329 was acquired into the STIS 52$`\mathrm{}\times `$0.$`\mathrm{}`$2 slit on 20 May 1999 and then offset by $`0.95\mathrm{}`$ in right ascension and $`4.06\mathrm{}`$ in declination (based on the NICMOS astrometric results) to place the secondary into the slit. To keep the primary as far out of the slit as possible, we employed a slit position angle of 252.06 degrees so that the line joining the primary and secondary was approximately perpendicular to the slit, thereby minimizing contamination from scattered primary light. Spectral imaging sequences were completed in one orbit with the G750M grating in three tilt settings with central wavelengths of 8311, 8825 and 9336Å (resolution $``$ 0.55Å) for total integration times of 340, 172, and 150 seconds, respectively. At each tilt setting we executed a two position dither of 0.$`\mathrm{}`$35 along the slit to allow replacement of bad or hot pixels, and the exposures were split for cosmic ray removal. Thus, we obtained four spectra at each tilt setting. After each set of four spectral images, we obtained flat fields required to calibrate the known effects of fringing which appear longward of $``$ 7500Å . Due to a failure of HST to acquire one of the two guide stars, there was a small differential pointing error of about 0.04″, or 1 pixel. This caused the secondary to be marginally de-centered and as a result a small percentage of the target flux fell out of the slit.
### 3.2 Results
The STIS spectra were calibrated, averaged, binned to a resolution of $``$ 6Å and normalized to the flux (in ergs/s/cm<sup>2</sup>/Å) at 8500Å. We compared the final, total spectrum to those of standard low-temperature dwarf and giant star spectra with a resolution = 18 Å, a factor of three lower than our STIS spectrum (Kirkpatrick et al. 1991; Kirkpatrick et al. 1997) (see Figure 2). The HR 7329B spectrum contains an absorption line near 8200Å which we attribute to the Na I doublet, which does not appear in late-type giant stars, but is nicely fit in the dwarfs. Also, as seen in Figure 2, the slope of the spectrum from 8600 to 8850Å is small, as in the dwarf spectra, whereas it rises sharply for giant stars. The NaI line is fit very well by the M8 V standard, but the TiO absorption near 8860Å is best fit by the M7 V spectrum. We therefore assign HR 7329B a spectral type of M7.5 V with an uncertainty of 0.5 spectral type.
The diffraction spikes from the primary star also fall in the slit above and below the secondary, and we used the relative positions of the three resulting spectra to determine the primary-secondary separation. The result of 4.13 $`\mathrm{}`$ $`\pm `$ 0.05 agrees, to within the uncertainties, with the NICMOS measured separation reported in Section 2.2.
## 4 Discussion
### 4.1 Likelihood of Companionship
From its H-magnitude and M7.5V spectral type, HR 7329B can be either a background object, a foreground main-sequence M star, or a companion to HR 7329A. A main-sequence M7.5V star has M<sub>H</sub> $`=`$ 10.3 (Kirkpatrick & McCarthy 1994), so HR 7329B is too bright to be a background main-sequence star. If it were on the main-sequence, its photometric distance would be 19 parsecs. Henry (1991), in a volume limited infrared survey, finds six objects with M<sub>H</sub> $`>`$ 9.5 within five pc from the sun. If we assume a spherical distribution of low mass stars in the solar neighborhood, we can extrapolate the results within 5 pc to expect 1000 such objects out at 25 pc, so the a priori probability of finding one in projection within a 4$`\mathrm{}`$ radius circle is $`10^7`$.
Proper motion measurements of the companion and primary in the time between the NICMOS and STIS observations could be used to further constrain the probability of companionship. Unfortunately, the positional errors are too large. However, we can further constrain the probability that the object is not a foreground M dwarf. Searching the Tycho cataloque, we find that the mean proper motion of 1000 stars between 16 and 25 parsecs is 0.373$`\mathrm{}\pm `$ 0.277. Therefore, if we assume a gaussian distribution of proper motions about this mean, almost 80% of foreground stars have moved more than the half-width of the STIS slit (0.1$`\mathrm{}`$)(taking into account angles along the slit), and would not be visible in the second epoch.
Given these arguments, it is unlikely ($`10^8`$) HR 7329B is a foreground object and for the remainder of the paper, we assume it is physically associated with HR 7329A.
### 4.2 Age of System
It is difficult to determine an age for A-type stars, but HR 7329 appears to be young ($`<`$ 40 Myr) based on its rotation, and more importantly, location on an H$``$R diagram. For massive stars, rotational velocities decline with age; HR 7329 has an especially large v$`sini`$ ($`=`$ 330 km/s) (Abt & Morrel 1995) which is considerably above the majority of A-type stars ($``$ 100 km/s). Figure 3 reproduces the H$``$R diagram from Jura et al (1998) for A stars from the Yale Bright Star Catalog and overplots nearby, young clusters. There seem to be common areas of similar age stars; the 50-90 Myr IC2391 and Alpha Per clusters lie below the older (600 Myr) Hyades and Preasepe. There is a large scatter in the Pleaides (70-125 Myr), which could be due to a range of distances and ages as well as unresolved binaries. HR 7329 lies on a line located below the Alpha Per and IC 2391 cluster which intersects $`\beta `$ Pic, HR 4796 and HD 141569. The latter stars have recently been assigned ages from their late-type companions of 20, 8, and 4 Myr respectively (Barrado Y Navascues et al. 1999; Stauffer, Hartmann, & Barrado Y Navascues 1995; and Weinberger et al. 2000). This suggests that HR 7329 is between 10 and 30 Myrs old. Finally, it has recently been suggested that HR 7329 is found within a young co-moving cluster much like the TW Hydrae Association with an age of $``$40 Myr (Zuckerman & Webb 2000; Webb et al. 2000).
### 4.3 Effective Temperature and Bolometric Luminosity
An effective temperature of HR 7329B is required to position it on an H$``$R diagram, but the temperature scale for late, young M-dwarfs is uncertain (Allard et al. 1997). Luhman & Rieke (1998) extrapolate from Leggett et al.’s (1996) model fits to derive 2670 K and 2505 K for M7 V and M8 V respectively, which agrees with the newer models used by Leggett, Allard, & Hauschildt (1998) with an uncertainty of about 100 K. With this uncertainty for late M dwarf stars and the added uncertainty due to the spectral type, we plot the derived temperatures for each spectral class (Figure 4) and their associated uncertainty which overlaps and gives a possible range from 2405 K to 2770 K.
The parallactic distance measured to HR 7329A by the Hipparcos mission is 47.67 $`\pm `$ 1.6 pc. With a derived H magnitude of 11.93 for HR7329B, and a distance modulus of 3.39, we calculate an absolute H magnitude of 8.54 mag. There exists a number of bolometric corrections in the literature for M7 V and M8 V stars (Tinney et al. 1993; Kirkpatrick et al. 1993; Bessel, Castelli, & Plez 1998) based on I and K magnitudes. However, none give the BC in the H band. We have used the BC at the other bandpasses and the colors of late-type stars as a function of spectral type from Kirkpatrick & McCarthy (1994) to find a relationship between BC(H) and spectral type. For M7 and M8 we find a range of BC<sub>H</sub> from 2.54 to 2.78. Using a solar M<sub>bol</sub> of 4.75, we derive a luminosity for HR 7329B of 0.0026 $`\pm `$ 0.0003 L$``$, with an uncertainty that includes the 0.5 spectral type range, bolometric correction, and distance errors.
### 4.4 Derived Mass
We place HR 7329B on pre-main sequence evolutionary tracks (Baraffe et al. 1998) to infer a mass (Figure 4). Assuming only companionship, and therefore distance, indicates a mass of less than 50 M<sub>Jup</sub> (less than 35M<sub>Jup</sub> is not covered in Baraffe’s models) and an age of less than 30 Myr. This supports the young age attributed to HR 7329A from its position on the H$``$R diagram, other youth indicators and possible membership in a young moving group. Evolutionary tracks from different authors do differ somewhat due to the different model atmospheres used. The tracks of DAntona & Mazitelli (1997) indicate a mass range of 40 M<sub>Jup</sub> or less for this luminosity and temperature. Burrows et al.’s (1997) models predict a 40 M<sub>Jup</sub> brown dwarf will have an effective temperature of 2800K and a luminosity of 0.0023 L$``$ at an age of 22 Myr.
## 5 Limits on Disk Detection
To look for possible reflected light from a circumstellar disk around the primary, we subtracted an observed coronagraphic PSF from each roll of HR 7329. The NICMOS PSF is time variable, exhibiting small-amplitude structural changes over multi-orbit timescales (Kulkarni et al. 1999). To find the best matched coronagraphic PSF to HR 7329, we tested each observation of the 40 other stars in our NICMOS program to see which gave the lowest noise subtraction as measured in the diffraction spikes and in annuli from 0.3$``$4$`\mathrm{}`$. There was no evidence of excess scattered light from a disk in any of the subtractions, but the first visit of the star HD 17925 observed on 26 Sep 1998 gave the lowest subtraction residuals. This K1 V star is 2.3 times brighter than HR 7329 at F160W. A plot of the azimuthally averaged residual surface brightness after subtraction is shown in the top panel of Figure 5. The error bars represent the standard deviation of all of the pixels (not including pixels obscured by diffraction spikes) averaged at each radius. The residuals are everywhere consistent with zero, i.e. no disk detection. The lower panel shows these uncertainties, multiplied by three and converted to F160W magnitudes as a measure of the disk flux which could have been detected at each radius.
HR 7329 appears in the IRAS Point Source catalog as having excess thermal infrared emission, indicating orbiting dust (Mannings & Barlow 1998). After color correcting the catalog fluxes for spectral index and subtracting the stellar photospheric contribution, the flux densities are F<sub>12μm</sub>=0.25$`\pm `$0.09 Jy, F<sub>25μm</sub>=0.36$`\pm `$0.05 Jy, F<sub>60μm</sub>=0.52$`\pm `$0.05 Jy, and an upper limit of 1 Jy at 100$`\mu `$m. These give a total dust optical depth $`\tau `$=$`L_{IR}/L_{star}3.5\times 10^4`$, which is an order of magnitude smaller than other similar stars at comparable distances such as HR 4796 and HD 141569 around which NICMOS imaged disks (Schneider et al. 1999, Weinberger et al. 1999).
## 6 Discussion
We present high signal-to-noise near-infrared photometry and optical spectroscopy of a probable companion (HR 7329B) at a projected distance of 200 AU from HR 7329 (A). We suggest the mass of B is less than 40 M<sub>Jup</sub> . The derived age of less than 30 Myr for this companion supports the very young age of the primary A0V star indicated by its placement on the H$``$R diagram of nearby A-type stars. We do not detect any 1.6$`\mathrm{\mu m}`$ scattered light from the far-infrared emitting dust seen by IRAS around HR 7329A.
The HR 7329 system stands out from other binaries in that it has a very high mass ratio, q$``$0.01. Zuckerman and Becklin (1992) found that around $``$ 200 white dwarf stars whose progenitors are F and A main sequence stars, the percentage of systems with low-mass M star companions (M$``$0.1M, q$``$0.06) was 5 to 10%, and the number of detected brown dwarfs was one (GD 165B), or $`<`$1%. The small percentage of white dwarfs with detectable brown dwarf companions is probably the result of the decline in brown dwarf luminosity with age. The discovery of the brown dwarf, HR 7329B, among a small sample of young A and F stars ($``$10) observed by NICMOS suggests that the number of companion brown dwarfs and low mass stars may not be too different. In the field (Reid 1999), and the Pleiades cluster (Zapatero Osorio et al 1997), the relative number of low mass stars and brown dwarfs per log mass interval is also about equal, suggesting a flat inital-mass-function (IMF) for single stars. Clearly, greater statistics are needed before firm conclusions can be reached about the IMF of secondaries.
We would like to thank M. Jura, C. Chen and J. Patience for their invaluable help and assistance. We thank the anonymous referee for comments which help to clarify the presentation. This work is supported in part by NASA grants NAG 5-4688 to UCLA and NAG 5-3042 to the University of Arizona NICMOS Instrument Definition Team. This paper is based on observations obtained under grant GO-8176.01-97A with the NASA/ESA Hubble Space Telescope at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc. under NASA contract NAS 5-26555.
|
warning/0005/physics0005014.html
|
ar5iv
|
text
|
# Multi-frequency evaporative cooling to BEC in a high magnetic field.
## Abstract
We demonstrate a way to circumvent the interruption of evaporative cooling observed at high bias field for <sup>87</sup>Rb atoms trapped in the $`(F=2,m=+2)`$ ground state. Our scheme uses a 3-frequencies-RF-knife achieved by mixing two RF frequencies. This compensates part of the non linearity of the Zeeman effect, allowing us to achieve BEC where standard 1-frequency-RF-knife evaporation method did not work. We are able to get efficient evaporative cooling, provided that the residual detuning between the transition and the RF frequencies is smaller than the power broadening of the RF transitions at the end of the evaporation ramp.
Forced evaporative cooling of atoms in a magnetic trap is at the moment the only known way to achieve Bose-Einstein condensation . Particles with energy significantly larger than the average thermal energy are removed from the trap and the remaining ones thermalize to a lower temperature by elastic collisions. For that, a radiofrequency (RF) magnetic field is used to induce a multi-photon transition from a trapping state to a non-trapping state via all intermediate Zeeman sublevels. Atoms moving in the trap with sufficient energy can reach the resonance point (RF knife) and exit the trap. If the RF-frequency is decreased slowly enough, and no other process is hampering the forced-evaporation, the increase of the phase space density obtained by this method eventually leads to Bose-Einstein condensation.
In a previous publication , we reported that RF forced evaporative cooling of <sup>87</sup>Rb atoms in the $`(F=2,m=+2)`$ ground state in a magnetic trap with a high bias field is hindered and eventually interrupted. Our interpretation of this phenomenon is based on the non-linear terms of the Zeeman effect that lift the degeneracy of transition frequencies between adjacent Zeeman sublevels. This interpretation is supported by numerical calculations . Interrupted evaporative cooling in a large magnetic field is a serious problem in several situations, interesting for practical reasons `-` like the use of permanent magnets or of an iron core electromagnet as the one described in . High magnetic field evaporation is also important in connection with Feshbach resonances . In this paper, we demonstrate that it is possible to achieve efficient evaporative cooling in a high magnetic field, by use of a multi-frequency RF knife allowing a multi-photon transition to take place across non equidistant levels. We show that, for our range of magnetic fields, it is possible to use a simple experimental scheme where the three required frequencies are obtained by RF frequency mixing yielding a carrier and two sidebands.
We focus in this paper on <sup>87</sup>Rb in the $`F=2`$ manifold of the electronic ground state. Atoms are initially trapped in the $`m=+2`$ state. Our high bias field magnetic trap follows the Ioffe-Pritchard scheme. To the second order in position (see eq. 1 in ), the magnetic field modulus $`B`$ has a 3D quadratic dependence allowing trapping, plus a bias field $`B_0`$ between 50 and 200 Gauss. This is much larger than in most other experiments where $`B_0`$ can be independently adjusted, and is set typically at 1 Gauss . In a large magnetic field, the non linear terms are not negligible in the Zeeman shifts given by the Breit-Rabi formula
$$E_m(B)=mg_\mathrm{I}\mu _\mathrm{n}B+\frac{\mathrm{}\omega _{\mathrm{HF}}}{2}\left(\sqrt{1+m\xi +\xi ^2}1\right)$$
(1)
with
$$\xi =\frac{(g_\mathrm{S}\mu _\mathrm{B}+g_\mathrm{I}\mu _\mathrm{n})B}{\mathrm{}\omega _{\mathrm{HF}}}.$$
Here $`g_\mathrm{S}2.002`$ and $`g_\mathrm{I}1`$ are respectively the Landé factor for the electron and the nucleus, $`\mu _B`$ and $`\mu _n`$ are the Bohr magneton and the nucleus magneton, and $`\omega _{\mathrm{HF}}`$ ($`2\pi \times 6834.7`$ MHz) is the hyperfine splitting.
Compared to the low magnetic field case , the evaporation process changes drastically. At a given magnetic field, the spacings between adjacent sublevels ($`|\mathrm{\Delta }m|=1`$) are not equal and the direct multi-photon transition from trapping to non-trapping states becomes negligible. Evaporation of hot atoms can only happen via a sequence of one-photon transitions of limited efficiency (see fig. 8 in ) separated in space. This results in long lasting atoms in the $`m=+1`$ and $`m=0`$ states responsible for hindered evaporative cooling. Moreover, transitions to non-trapping states are suppressed at the end of the evaporation ramp, leading to an interruption of cooling before BEC is reached.
To overcome these limitations, 3 distinct RF fields can be used to induce a direct three photon transition from the $`m=+2`$ trapping state to the $`m=1`$ non trapping state. At a magnetic field $`B`$, the three RF frequencies must match the transition frequencies defined by:
$`\omega _0\delta \omega _0^{}`$ $`=`$ $`(E_2E_1)/\mathrm{}`$ (2)
$`\omega _0`$ $`=`$ $`(E_1E_0)/\mathrm{}`$ (3)
$`\omega _0+\delta \omega _0`$ $`=`$ $`(E_0E_1)/\mathrm{}`$ (4)
with $`E_m`$ taken from eq.(1).
Fig. 1 represents all possible transitions induced by these three RF frequencies in the magnetic trap. At position $`K`$, each RF field is resonant with a given transition : the smallest RF frequency with the $`(m=+2)(m=+1)`$ transition, the intermediate frequency with the $`(m=+1)(m=0)`$ transition, and the largest frequency with the $`(m=0)(m=1)`$ transition; this is where the 3-photon transition occurs. Because of the ordering of the three RF frequencies, the points where one-photon transitions can be induced from $`m=+2`$ to $`m=+1`$ by the two larger frequencies are located beyond $`K`$ (the multi-photon knife). Consequently, during the evaporation, hot atoms will first encounter the three photon knife and be expelled from the trap, provided that the RF power is large enough to enable efficient multi-photon adiabatic passage to the non-trapping state $`m=1`$.
The discussion above shows that in principle, the multi-frequency evaporation requires a synchronized non trivial sweep of three different frequencies in the 100 MHz range, with an accuracy of a few kHz (see below). We have rather implemented a simplified scheme where the three frequencies are obtained by mixing a carrier at frequency $`\omega _{\mathrm{RF}}`$ with a smaller frequency $`\delta \omega _{\mathrm{RF}}`$. We then obtain three equally spaced radiofrequency fields : $`\omega _{\mathrm{RF}}\delta \omega _{\mathrm{RF}}`$, $`\omega _{\mathrm{RF}}`$, $`\omega _{\mathrm{RF}}+\delta \omega _{\mathrm{RF}}`$, of approximately the same power (as checked with a spectrum analyzer). Since in general $`\delta \omega _0`$ and $`\delta \omega _0^{}`$ are slightly different, the RF frequencies will not exactly match the transition frequencies of eq.(3). Nonetheless, they compensate the second order (quadratic) term of the Zeeman shift, and should work under certain condition discussed hereafter.
At the position where the three-photon transition is resonant, the carrier frequency $`\omega _{\mathrm{RF}}`$ will verify
$$3\omega _0+\delta \omega _0\delta \omega _0^{}=3\omega _{\mathrm{RF}}$$
(5)
but there will be a residual detuning for each one photon step of the multi-photon transition. For example, the optimum $`\delta \omega _{\mathrm{RF}}`$ that maximizes the multi-photon transition probability will be
$$\delta \omega _{\mathrm{RF}}=\frac{\delta \omega _0+\delta \omega _0^{}}{2}$$
(6)
and the residual detunings for each intermediate steps of the three photons transition are both equal to
$$\mathrm{\Delta }=\frac{\delta \omega _0\delta \omega _0^{}}{6}.$$
(7)
If the Rabi frequency $`\mathrm{\Omega }_{\mathrm{RF}}`$ associated with each one photon transition is significantly larger than the residual detuning $`\mathrm{\Delta }`$, the multi-photon transition is quasi resonant in the intermediate levels, leading to an effective Rabi frequency $`\mathrm{\Omega }_{\mathrm{eff}}\mathrm{\Omega }_{\mathrm{RF}}`$. If on the other hand $`\mathrm{\Omega }_{\mathrm{RF}}`$ is smaller than $`\mathrm{\Delta }`$, the effective Rabi frequency is
$$\mathrm{\Omega }_{\mathrm{eff}}\frac{\mathrm{\Omega }_{\mathrm{RF}}^3}{\mathrm{\Delta }^2}$$
(8)
and the multi-photon transition is inefficient for evaporation ; we are then in the scheme of hindered and interrupted evaporation. We therefore expect that our scheme will be efficient for small enough magnetic field when the residual detuning $`\mathrm{\Delta }`$ is smaller than the one-photon Rabi frequency $`\mathrm{\Omega }_{\mathrm{RF}}`$.
Table I gives the values of the Zeeman shifts and the difference $`\delta \omega _0\delta \omega _0^{}`$ for various magnetic fields. For the RF power used in this scheme, the one photon Rabi frequency $`\mathrm{\Omega }_{\mathrm{RF}}`$ is of the order of 10 kHz, and the discussion above shows that our simplified 3-knives evaporation scheme should work for magnetic fields significantly less than a hundred Gauss. This is what we observe: it is impossible to achieve BEC in bias fields of 207 Gauss and 110 Gauss, but BEC is obtained in a trap with a bias field of 56 Gauss, by using an appropriate sideband splitting $`\delta \omega _{\mathrm{RF}}`$ kept constant while ramping down the carrier frequency $`\omega _{\mathrm{RF}}`$.
Figure 2 shows the effect of the sideband splitting $`\delta \omega _{\mathrm{RF}}`$ at a bias field value of 56 Gauss. We have plotted the number of condensed atoms as a function of $`\delta \omega _{\mathrm{RF}}`$, all other parameters being kept unchanged. This is a good indication of the efficiency of the evaporation. The curve shows a maximum at $`\delta \omega _{\mathrm{RF}}=2\pi \times 0.45`$ MHz. This value verifies equation (6) for a magnetic field of 56.6 Gauss. This magnetic field corresponds to the position of the RF knife at the end of the ramp. We conclude that frequency matching is mostly important in the last part of the radiofrequency ramp. The width of the curve is about 10 kHz (HWHM) which corresponds to power broadening .
Table II report experimental data, showing quantitatively the efficiency of our simplified 3-knives scheme, without which BEC could not be obtained at 56 Gauss. It is interesting to note that even when the magnetic field is too large to allow our simplified 3-knives scheme to reach BEC, it is nevertheless more efficient than a simple 1-frequency knife, since it allows us to reach a significantly lower temperature. It is also remarkable that an efficient evaporation was obtained at a bias field of 56 Gauss, since the beginning of the evaporation takes place in a larger magnetic field (of the order of 200 Gauss) where the condition (6) does not hold, and the detuning of the intermediate one photon transitions is much larger than the Rabi frequency $`\mathrm{\Omega }_{\mathrm{RF}}`$. Although it has not been much noticed, a similar situation is encountered in most BEC experiments (using 1-frequency knife evaporation) : the non linear Zeeman effect at the beginning of the evaporation is often much larger than the Rabi frequency, and the evaporation hampering described in is certainly happening then. The success of these experiments as well as of our 3-frequencies scheme shows that whether the evaporation is hindered or not only matters at the end of the evaporation ramp. To understand qualitatively this observation, we can note that the heating induced by the atoms populating the intermediate levels should not vary drastically with the temperature of the cooled cloud. At the beginning of the evaporation, i.e. “high” temperatures, the relative heating stays negligible . Close to the end, i.e. “low” temperature, when heating should give rise to hampered evaporative cooling, evaporation is fully efficient and the intermediate levels are completely depleted. This could explain the success of BEC experiments. To verify these assumptions, more theoretical work, for instance in the spirit of , is needed.
In conclusion, we have demonstrated a scheme to circumvent the hindrance and interruption of evaporative cooling in the presence of non linear Zeeman effect. We implement a 3-frequency evaporative knife by a modulation of the RF field, yielding two sidebands. This scheme allows us to obtain BEC of <sup>87</sup>Rb atoms in the $`(F=2,m=+2)`$ ground state in a bias field of 56 Gauss, where the standard 1-frequency RF evaporation scheme fails. Our observations also support the physical ideas presented in our previous work to explain the hindrance and interruption of evaporative cooling in a high magnetic field, as well as the qualitative discussions of this paper.
The success of this simplified scheme and the complementary observations reported in this paper, indicate that a more sophisticated multi-frequency evaporation scheme should work at larger bias field, provided that the resonance in the intermediate steps of the multi-photon transition is achieved within the Rabi frequency of the one photon transitions, at the end of the evaporative ramp.
###### Acknowledgements.
The authors thank S. Rangwala for helpfull discussions and M. Lécrivain for the elaboration of the iron-core electromagnet. This work is supported by CNRS, MENRT, Région Ile de France, DGA and the European Community. SM acknowledges support from Ministère des Affaires Étrangères. YLC acknowledges support from DGA.
|
warning/0005/cond-mat0005451.html
|
ar5iv
|
text
|
# Landau Level Mixing and Skyrmion Stability in Quantum Hall Ferromagnets
## I Introduction
It has been recognized that exotic magnetic excitations known as skyrmions may exist in a two-dimensional electron gas in a strong homogeneous magnetic field (quantum Hall system) near spin polarized groundstates. These are excitations of a two-dimensional spontaneous ferromagnet, the physics of which is relevant to this system (despite the presence of a strong magnetic field), because of the small Landé $`g`$-factor in GaAs systems (where most experiments take place), which makes the Zeeman coupling very small compared to other energy scales (Coulomb interaction, cyclotron energy) in the problem. Skyrmions are spin configurations with a non-trivial winding number (Pontryagin index). They were first discussed in the context of four-dimensional field theories , and were later recognized as states occurring in the non-linear sigma model description of two-dimensional ferromagnets . For filling factors $`\nu N/N_\varphi `$ close to one ($`N`$ is the number of electrons and $`N_\varphi `$ the number of magnetic flux quanta penetrating the system), these turn out to be the lowest energy quasiparticles under typical experimental circumstances. Skyrmions can thus be introduced into the groundstate by adding or removing charge from the system. .
Experimentally, the case for the existence of skyrmions in a system close to $`\nu =1`$ is quite strong. NMR experiments show a degrading of the spin polarization with deviation of filling factor from one that is in remarkably good agreement with Hartree-Fock theory . The quasiparticle spin measured in transport experiments are also reasonably well accounted for by Hartree-Fock calculations . Electromagnetic absorption experiments further support that doping away from $`\nu =1`$ injects skyrmions into the system.
In weaker magnetic fields, near filling factor $`\nu =3`$, early experiments suggested that spin-polarized quasiparticles are lower in energy than skyrmions, so the effects seen near $`\nu =1`$ would not be present at higher filling. This is consistent with calculations of skyrmion energies near $`\nu =3`$ that include finite thickness corrections , which indicate that skyrmions will be present only at much smaller Zeeman couplings than realized in typical experiments. The size of the skyrmion may be quantified by a number $`K`$, the difference in the spin component $`S_z`$ between the skyrmion and the spin-polarized quasiparticle. Because of the necessarily small Zeeman coupling, stable skyrmions close to $`\nu =3`$ have large values of $`K`$. They also become unstable with respect to spin-polarized quasiparticles at a finite value of $`K`$. (For $`\nu =1`$, $`K0`$ as the Zeeman coupling reaches the maximum value for which the system supports skyrmions; i.e., the skyrmion state smoothly goes into the spin-polarized quasiparticle state.) For a two-dimensional electron gas (2DEG) with width of about 2$`\mathrm{}`$, where $`\mathrm{}=\sqrt{\mathrm{}c/eB}`$ is the magnetic length, the minimum $`K`$ expected is approximately 4.
Recently however, NMR experiments have uncovered evidence that some anomalous degrading of spin polarization $`does`$ occur as one dopes away from $`\nu =3`$ at relatively high Zeeman couplings. These experiments further indicate that the number of overturned spins per quasiparticle is quite small, $`K1`$. The simple models usually considered are inconsistent with this, and one is naturally led to inquire as to what other ingredients might change the critical Zeeman coupling and smallest $`K`$ observable near $`\nu =3`$. Two possible answers are Landau level mixing and screening by filled Landau levels. It should be noted that these are not distinct effects: screening by filled Landau levels occurs because they may admix high (unoccupied) Landau levels to smooth fluctuations due to external potentials and/or inhomogeneous electron densities in partially filled levels. Conversely, the states which may be used for Landau level mixing in a partially filled level are limited to those that are not occupied by electrons in other levels, due to Pauli exclusion. Thus, a correct treatment of either screening or Landau level mixing near $`\nu =3`$ must include both these effects.
In this work, we present a method by which these may be incorporated into the Hartree-Fock description of skyrmion states.
Our principal conclusions may be summarized as follows: (i) For skyrmions near $`\nu =1`$, Landau level mixing tends to lower the quasiparticle energy, although not enough to quantitatively explain the activation energies seen in experiment . Introduction of a finite width of one magnetic length lowers the energy of the skyrmion by approximately 40%, and inclusion of Landau level mixing lowers the energy by approximately another 20% for $`\mathrm{}\omega _c/(e^2/\kappa \mathrm{})0.6`$, where $`\omega _c=eB/m^{}c`$ is the cyclotron frequency of the electrons, and $`\kappa `$ is the dielectric constant of the host crystal. The resulting quasiparticle gap is approximately a factor of 2 larger than what is found in experiment . This result agrees qualitatively with that of another study of Landau level mixing effects on $`\nu =1`$ skyrmions . (ii) For $`\nu =1`$, Landau level mixing tends to suppress quasihole-like skyrmions (i.e., lowering the maximum Zeeman coupling for which they are stable), while enhancing the stability of quasielectron (anti-)skyrmions. (iii) For $`\nu =3`$ and higher, we find that a sufficiently realistic model of the effective electron-electron interaction, as modified by the finite thickness of the electronic wavefunctions, is necessary to obtain reliable results. Use of a simple potential due to Zhang and Das Sarma grossly overestimates the stability of $`\nu =3`$ skyrmions; more realistic potentials allow skyrmions only for very small Zeeman couplings. (iv) Screening and Landau level mixing for $`\nu =3`$ and $`\nu =5`$ tend to lower the energy of spin-polarized quasiparticles more than that of skyrmions, making the latter even less stable. (v) The results of Ref. song cannot be understood solely on the basis of Hartree-Fock states for skyrmions.
The remainder of this article is organized as follows. In Section II below, we discuss the method used to allow screening and Landau level mixing to be included in our calculations. Section III gives details of our results, and we conclude with a summary of our findings in Section IV.
## II Hartree-Fock Method with Landau Level Mixing
Most previous Hartree-Fock studies of skyrmions have relied on Landau level representation of the single particle states . We choose instead to construct the wavefunctions in real space. This enables us to include in the model Landau level mixing occurring in weak magnetic fields, without having to expand over the large number of Landau levels necessary in the former approach. We thus trade the calculational convenience of working with the functions given in closed analytic form (Landau levels) for a closer description of the single-particle states by representing them on a real-space grid.
In this calculation we aim to model all the participating particles, including the ones in the filled levels. Our Hartree-Fock wavefunction is a Slater determinant composed of single-particle states which have $`L_z\pm S_z`$ as a good quantum number but whose radial form is to be determined self-consistently:
$`|\mathrm{\Psi }_{skyrmion}={\displaystyle \underset{i,m}{}}\gamma _{im}^{}|0`$ (1)
$`\stackrel{}{r},\sigma |\gamma _{im}^{}|0=\left[\begin{array}{c}f_{im}(r)\\ g_{im}(r)e^{\pm i\theta }\end{array}\right]e^{im\theta }.`$ (4)
Here $`r`$ and $`\theta `$ are polar coordinates, $`m`$ is the angular momentum quantum number, and $`i`$ labels different states of the same $`m`$. The sign $`\pm `$ corresponds to two families of solutions, $`+`$ for antiskyrmion (or quasielectron spin structured solution) and $``$ for skyrmion (quasihole).
In very strong magnetic fields, the functions $`f(r)`$ and $`g(r)`$ take the form expected for Landau level states. When the strength of the magnetic field is lowered to bring the ratio of cyclotron and Coulomb energy scales close to 1 , the form of the radial part relaxes toward some modified form, as dictated by the interactions in the system.
Using the trial form of the wavefunction (1), the many-body Schrödinger equation with the Hamiltonian
$`H={\displaystyle \frac{1}{2m}}{\displaystyle d^2r\underset{\sigma }{}\mathrm{\Psi }_\sigma ^{}(\stackrel{}{r})|\frac{\mathrm{}}{i}\stackrel{}{}\frac{e}{c}\stackrel{}{A}|^2\mathrm{\Psi }_\sigma (\stackrel{}{r})}`$ (5)
$`+`$ $`{\displaystyle \frac{1}{2}}g\mu B{\displaystyle d^2r[\mathrm{\Psi }_{}^{}(\stackrel{}{r})\mathrm{\Psi }_{}(\stackrel{}{r})\mathrm{\Psi }_{}^{}(\stackrel{}{r})\mathrm{\Psi }_{}(\stackrel{}{r})]}`$ (6)
$`+`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \underset{\sigma \sigma ^{}}{}}{\displaystyle d^2rd^2r^{}\mathrm{\Psi }_\sigma ^{}(\stackrel{}{r})\mathrm{\Psi }_\sigma ^{}(\stackrel{}{r}^{})v(\stackrel{}{r}\stackrel{}{r}^{})\mathrm{\Psi }_\sigma ^{}^{}(\stackrel{}{r}^{})\mathrm{\Psi }_\sigma (\stackrel{}{r})}`$ (7)
(where $`\sigma `$ denotes spin and $`v(\stackrel{}{r}\stackrel{}{r}^{})`$ the Coulomb interaction), upon variation with respect to the functions $`f`$ and $`g`$, gives a system of mean-field single-particle equations:
$`D_m(r)f_{im}(r){\displaystyle \frac{1}{2}}g\mu _BBf_{im}(r)`$ (8)
$`+`$ $`{\displaystyle _0^{\mathrm{}}}r^{}𝑑r^{}V^H(r,r^{})[\rho (r^{})\rho _0]f_{im}(r)`$ (9)
$``$ $`{\displaystyle _0^{\mathrm{}}}r^{}𝑑r^{}{\displaystyle \underset{m^{}}{}}V_{mm^{}}^{ex}(r,r^{})\rho _m^{}^{}(r^{},r)f_{im}(r^{})`$ (10)
$``$ $`{\displaystyle _0^{\mathrm{}}}r^{}𝑑r^{}{\displaystyle \underset{m^{}}{}}V_{mm^{}}^{ex}(r,r^{})\rho _m^{}^{}(r^{},r)g_{im\pm 1}(r^{})`$ (11)
$`=`$ $`ϵ_if_{im}(r)`$ (12)
together with the analogous equation for the function $`g(r)`$. Here is a dictionary of the notation accompanying Eq. 8: the operator $`D_m`$ is given by
$`D_m`$ $`=`$ $`{\displaystyle \frac{\mathrm{}^2}{2m^{}}}[{\displaystyle \frac{1}{r}}{\displaystyle \frac{d}{dr}}r{\displaystyle \frac{d}{dr}}{\displaystyle \frac{m^2}{r^2}}]`$ (13)
$``$ $`m{\displaystyle \frac{\mathrm{}\omega _c}{2}}+{\displaystyle \frac{(m^{})^2}{4}}{\displaystyle \frac{\omega _c^2r^2}{2m^{}}}`$ (14)
with $`\omega _c=\frac{eB}{m^{}c}`$, $`B`$ the magnitude of the external magnetic field, and $`m^{}`$ the effective mass of the electron. $`g`$ is the Landé g-factor, and $`\mu _B`$ is the Bohr magneton. $`\rho `$’s denote generalized densities:
$`\rho _m^{}^{}(r^{},r)`$ $`=`$ $`f_m^{}^{}(r^{})f_m^{}(r)`$ (15)
$`\rho _m^{}^{}(r^{},r)`$ $`=`$ $`g_{m^{}\pm 1}^{}(r^{})g_{m^{}\pm 1}(r)`$ (16)
$`\rho _m^{}^{}(r^{},r)`$ $`=`$ $`g_{m^{}\pm 1}^{}(r^{})f_m^{}(r)`$ (17)
$`\rho (r)`$ $`=`$ $`{\displaystyle \underset{m^{}}{}}[\rho _m^{}^{}(r,r)+\rho _m^{}^{}(r,r)],`$ (18)
and $`\rho _0`$ is the uniform background density. $`V^H`$ and $`V^{ex}`$ are the following integrals of the Coulomb potential over the azimuthal variable:
$`V^H(r,r^{})`$ $`=`$ $`{\displaystyle _0^{2\pi }}𝑑\theta {\displaystyle _0^{2\pi }}𝑑\theta ^{}v(\stackrel{}{r}\stackrel{}{r^{}})`$ (19)
$`V_{mm^{}}^{ex}(r,r^{})`$ $`=`$ $`{\displaystyle _0^{2\pi }}𝑑\theta {\displaystyle _0^{2\pi }}𝑑\theta ^{}e^{i(mm^{})(\theta \theta ^{})}v(\stackrel{}{r}\stackrel{}{r^{}}),`$ (20)
and, finally, $`ϵ_i`$ stands for the single-particle Hartree-Fock energy.
The finite width of the sample is modeled using the form of the in-plane potential due to Cooper : we replace the Coulomb interaction $`v(\stackrel{}{r}\stackrel{}{r^{}})`$ in the Eq. (19) by
$`v_C(r)={\displaystyle _{\mathrm{}}^{\mathrm{}}𝑑z𝑑z^{}\frac{e^{(z^2+z^2)/2w^2}}{2\pi w^2}\frac{1}{\sqrt{r^2+(zz^{})^2}}}.`$ (21)
The symbol $`w`$ denotes the width of system in the direction perpendicular do the plane of the system, and $`z`$, $`z^{}`$ are coordinates in that direction.
To handle the boundaries of the system, we assume the electron states with angular momentum $`m>m_{max}`$ have the ferromagnetic groundstate form (i.e. Landau levels with well defined spin). The states with $`mm_{max}`$ are explicitly included in the calculation. For $`K`$ not too large we find it is sufficient to allow variations of the states with $`m`$ of up to 30 for $`\nu =3`$ and up to 50 for $`\nu =5`$. In practice, including boundary electrons from the states with $`m`$ between 31 and 100 ($`\nu =3`$) and between 51 and 120 ($`\nu =5`$) describes the effect of the system edge with precision matching the rest of the calculation.
Understanding Eq. (8) as a system of coupled eigenproblems, we look for the self-consistent single particle solutions. (Discretization will turn each eigenequation into a matrix diagonalization problem which can be handled using standard methods.) The results we thus obtain will be largely presented as comparisons between energies of the spin-polarized quasiparticle and energies of the corresponding skyrmion.
To assess the energy of the skyrmion in the region of parameters where it is not stable, we add to the Hamiltonian a term of the form $`\lambda (\widehat{S}_zS_0)^2`$, $`\widehat{S}_z`$ being the spin operator, and $`\lambda `$ a tunable parameter. This term favors a state with total spin $`S_0`$, but is insensitive to the detailed form of the wavefunctions. This allows the variational scheme to pick out the lowest energy Slater determinant of the form given in Eq.(1) within the space of states with the same fixed value of $`K`$.
## III Results
Based on the calculation described in the previous section, we present some of the results the method allows us to obtain; we focus mainly on the singly charged excitations in the first three Landau levels. Consideration of higher Landau levels is also possible, but computation of the potential lookup tables becomes prohibitively expensive, and, as the results so far indicate, leads to no new insight. In the following we shall take the unit of energy to be $`e^2/\kappa \mathrm{}`$, and the unitless Zeeman splitting to be $`\stackrel{~}{g}=\frac{g\mu _BB}{e^2/\kappa \mathrm{}}`$, where $`e`$ is the electron charge, $`\kappa `$ is the dielectric constant of the host material, $`\mathrm{}`$ is the magnetic length in the field $`B`$ and $`\mu _B`$ is the Bohr magneton.
### A Skyrmion vs. polarized quasiparticle
In Fig. 1 we show the energy difference between the spin-polarized quasiparticle and the skyrmion of size $`K`$, $`V_KV_{K=0}`$. This quantity is a pure interaction energy (i.e. Zeeman energy is not included), and represents the energy gained or lost in deforming a spin-polarized quasiparticle into a skyrmion when Zeeman coupling is absent. Of particular importance is the slope (negative slopes indicate that skyrmion is stable for some value of $`g`$) and the curvature (concave curves will support small-sized skyrmions). For concave curves the largest Zeeman splitting that supports skyrmions is the negative of the initial slope of the curve .
For large cyclotron energies our results are essentially identical to those obtained using the single Landau level method . Note the quasielectron and quasihole excitations are precisely degenerate in this case, due to particle-hole symmetry. For smaller values of $`\mathrm{}\omega _c`$, the two curves split; surprisingly, the quasihole skyrmion is suppressed by Landau level mixing, whereas the quasielectron skyrmion is enhanced. (The former result is in agreement with Ref. melik ).
The energy gaps (Fig. 2) which result from creation of skyrmion-antiskyrmion pairs when Landau level mixing and finite thickness corrections are included are considerably smaller than what is found for two-dimensional layers and no mixing . However, the resulting energies are still almost a factor of two larger than what is found in experiment . The discrepancy is likely to be due to disorder.
Figs. (3) and (4) present analogous results for $`\nu =3`$ and $`\nu =5`$. Note the considerably smaller energy scales in these figures, indicating that skyrmions can only be stable (if ever) for small values of $`\stackrel{~}{g}`$ . It is apparent that the introduction of Landau level mixing and screening destabilizes the skyrmion. Evidently, spin-polarized particles are better able to take advantage of the admixture of higher Landau levels than skyrmions.
### B Effect of Finite Well Width
Quasiparticle energies depend on the well width, as illustrated in Figs. 5 and 6. As expected , we find that for narrower wells the difference in energy is less favorable for the skyrmion (Fig. 6). Note that the width used in Fig. 5 is close to an experimentally reported value .
It is worth remarking at this juncture that a reasonably realistic model of the electron-electron interaction with finite sample thickness corrections is needed to obtain qualitatively correct results. Fig. 7 shows that the use of a simpler model potential (Zhang and Das Sarma )
$`v_{ZdS}(\stackrel{}{r}\stackrel{}{r}^{})={\displaystyle \frac{1}{\sqrt{|\stackrel{}{r}\stackrel{}{r}^{}|^2+w^2}}}`$ (22)
which is commonly used in studying quantum Hall systems (see for example Refs. melik and shankar ) gives substantially different results then those presented above (Fig. 3).
The principal difference between $`v_C(r)`$ and $`v_{ZdS}(r)`$ is the behavior at small r; the former diverges logarithmically, whereas the latter is regular. Other divergent potentials give results consistent with Figs. 3 and 4; it is likely that the oversimplified behavior at short distances is responsible for the poor performance of $`v_{ZdS}(r)`$ in this problem.
### C Phase Diagram
Based on results in Figs. 1, 3, and 4 we can construct the phase diagrams of skyrmion stability for the filling factors of $`\nu =1,3,`$ and $`5`$. Large $`\omega _c`$ and small $`g_c`$ is the region favoring the spin structured excitations. We see that the region “shrinks” as one moves to the higher filling factors. Also, according to this calculation the breaking of symmetry between the quasihole and quasielectron excitations upon lowering the cyclotron energy is quite spectacular in the lowest Landau level, whereas it plays no significant role in the higher ones.
In the paper by Song *et al*. the reported excitation at the parameter values of $`w=1.73\mathrm{},\omega _c=0.56`$, and $`\stackrel{~}{g}=1.72\times 10^2`$ falls well outside the boundary expected from the Hartree-Fock calculation.
### D Effect of Impurities
It is tempting to speculate that inclusion of impurity effects can stabilize the skyrmions at values of $`g`$ bigger than allowed in a pure sample. To test this idea we can include a simple model of an impurity in our calculation: a point charge (impurity) at a distance $`z_0`$ above the central plane of the system. It is replaced by an effective non-uniform charge density in the plane producing the same potential,
$`{\displaystyle \frac{1}{|\stackrel{}{r}(\stackrel{}{r}_0+\stackrel{}{z}_0)|}}={\displaystyle d^2r^{}\rho _{eff}(\stackrel{}{r}^{})\frac{1}{|\stackrel{}{r}\stackrel{}{r}^{}|}}.`$ (23)
The effective density can be found to be
$`\rho _{eff}(\stackrel{}{r})={\displaystyle \frac{1}{2\pi }}{\displaystyle \frac{z_0}{(z_0^2+R^2)^{3/2}}}`$ (24)
for $`z_0>0`$. Results with and without such an impurity are illustrated in Fig. 9. As may be seen, the impurity favors spin-polarized quasiparticle over skyrmion. A similar result is expected for a short-range (e.g. delta-function) impurity potential. Apparently the simplest models of disorder are not likely to explain the results of Ref. song . It is probable that more complicated impurities (e.g. multiply charged or magnetic ones) could stabilize the small-spin skyrmions at $`\nu =3`$. However, in the absence of data indicating such types of disorder in real samples, an investigation of this phenomenon is left for future work.
## IV Conclusion
In this paper we have presented a real-space method for computing Hartree-Fock states and energies of two-dimensional systems in magnetic fields, appropriate for systems with circular symmetry in which Landau level mixing may be important. The method was applied to compute the effects of Landau level mixing and screening on skyrmion states. It was found that in most cases these tend to destabilize skyrmions, with a notable exception occurring for the case of the quasielectron (antiskyrmion) around $`\nu =1`$. The calculations indicate that Hartree-Fock states cannot account for the results of Ref. song (reporting skyrmions at $`\nu =3`$). This is in agreement with earlier studies where Landau level mixing and screening were not included.
## ACKNOWLEDGMENTS
This work was supported by NSF Grant Nos.DMR98-70681 and PHY94-07194, the Research Corporation, and the Center for Computational Sciences of the University of Kentucky.
|
warning/0005/hep-ph0005138.html
|
ar5iv
|
text
|
# MODELLING BOSE-EINSTEIN EFFECT FROM ASYMMETRIC SOURCES IN MONTE CARLO GENERATORSaafootnote aPresented by K. Fiałkowski at the XXXVth Rencontres de MORIOND, ”QCD and High Energy Hadronic Interactions”, to be published in the Proceedings.
## 1 Introductory remarks and data
Recently we observe a renewal of interest in analysing the space-time structure of sources in multiparticle production by means of Bose-Einstein (BE) interference. Such analysis followed the example of astrophysical investigations of Hanbury-Brown and Twiss $`^\mathrm{?}`$. By improving the standard approach $`^\mathrm{?}`$ it became possible to model this effect in Monte Carlo generators: as the ”afterburner” for which the original MC provides a source $`^{\mathrm{?},\mathrm{?}}`$, by shifting the momenta $`^\mathrm{?}`$ or by adding weights to generated events $`^{\mathrm{?},\mathrm{?}}`$.
The analysis of BE effect in 3 dimensions is supposed to reflect the spatial source asymmetry. Such analysis was done for the LEP data at the $`Z^0`$ peak $`^\mathrm{?}`$ which have very high statistics and good accuracy. In the following we concentrate our attention on the L3 data $`^\mathrm{?}`$, as the DELPHI data $`^\mathrm{?}`$ are parametrized with only two radii, and the OPAL data $`^\mathrm{?}`$ use the like/unlike ratio which requires a cut off of the resonance affected regions even in double ratios.
As in the L3 paper $`^\mathrm{?}`$ we use for each pair of identical pions three components of the invariant $`Q^2=(p_1p_2)^2`$: $`Q_L^2,Q_{out}^2,Q_{side}^2`$ defined in the LCMS, where the sum of three - vector momenta is perpendicular to the thrust axis. Similarly we define a ”double ratio” using a reference sample from mixed events:
$$R_2(p_1,p_2)=\frac{\rho _2}{\rho _2^{mix}}/\frac{\rho _2^{MC}}{\rho _2^{mix,MC}}.$$
This ”double ratio” is parametrized by
$$R_2(Q_L,Q_{out},Q_{side})=\gamma [1+\delta Q_L+ϵQ_{out}+\zeta Q_{side}]$$
$$[1+\lambda exp(R_L^2Q_L^2R_{out}^2Q_{out}^2R_{side}^2Q_{side}^22\rho _{L,out}R_LR_{out}Q_LQ_{out})]$$
The first bracket reflects possible traces of long-distance correlations; the last term in the second bracket seems to be negligible when fitting data and will be omitted in the following.
By fitting the parameters $`R_L`$ and $`R_{side}`$ we get some information on the geometric radii in the longitudinal and transverse directions (respective to the thrust axis). $`R_{out}`$ reflects both the spatial extension and time duration of the emission process.
In the L3 data the fit region in all three variables extends to 1.04 GeV and is divided into 13 bins, which gives 2197 points fitted with 8 parameters. The fit parameters $`\delta ,ϵ`$ and $`\zeta `$ are rather small; this means the observed BE enhancement is rather well approximated with a Gaussian.
The fitted values of radii (in fm) are as follows:
$$R_L=0.74\pm 0.02_{0.03}^{+0.04},R_{out}=0.53\pm 0.02_{0.06}^{+0.05},R_{side}=0.59\pm 0.01_{0.13}^{+0.03}$$
We see clear evidence for source elongation: $`R_{side}/R_L`$ is smaller than one by more than four standard deviations.
## 2 Asymmetric effects from symmetric models
The geometric interpretation of data requires a comparison with the results from the standard MC procedures modelling the BE effect. In the L3 paper such an analysis is given for the standard LUBOEI procedure built into the JETSET Monte Carlo generator.This procedure modifies the final state by a shift of momenta for each pair of identical pions. The shift is calculated to enhance low values of $`Q^2`$ and to reproduce the experimental ratio in this variable. The superposition of the procedure for all the pairs and subsequent rescaling (restoring the energy conservation) makes the connection between the parameters of the shift and the resulting ratio rather indirect.
Using the JETSET parameters adjusted to all the L3 data and the LUBOEI parameters fitted to describe the BE ratio in $`Q^2`$ the authors of the L3 paper calculated the same quantities as measured in the experiment. The projections of $`R_2`$ look qualitatively very similar to the experimental ones. However, the fit to the 3-dimensional distribution gives results different from data. The ratio $`R_{side}/R_L`$ is not smaller but greater than one; the fitted values (in fm) are:
$$R_L=0.71\pm 0.01,R_{out}=0.58\pm 0.01,R_{side}=0.75\pm 0.01.$$
We confirmed these numbers in our calculations. We found also that the results are sensitive to the JETSET parameters. Using the default values instead of the L3 values we obtained a significantly smaller value of $`R_{out}`$ (below 0.5) and significantly smaller $`\lambda `$. Other values are less affected and $`R_{side}/R_l`$ still bigger than 1.
We have checked also how the results depend on the source radius $`R`$ and incoherence parameter $`\lambda _{in}`$ assumed in the LUBOEI input function $`R_{BE}(Q)=1+\lambda _{in}exp(R^2Q^2)`$. In all cases we get $`R_{side}>R_L>Rout`$, although the input function was obviously symmetric. The values of $`R_{side}`$ and $`R_L`$ are proportional to $`R`$, whereas $`R_{out}`$ changes much less; the dependence on $`\lambda _{in}`$ is very weak. The output value of $`\lambda `$ decreases quite strongly with increasing $`R`$ and increases with $`\lambda _{in}`$. No choice of input parameters gives the values of $`R_i`$ compatible with data. This is shown in Fig. 1.
Another interesting observation is that to fit the L3 data one needs $`\lambda =1.5`$, which is beyond the physically acceptable value of 1. This supports our doubts about the usefulness of the LUBOEI procedure in understanding the experimental results (although certainly it is the most practical description of data).
Therefore we have also compared the data with the results from another procedure modelling the BE effect - the weight method $`^\mathrm{?}`$. In this method each event gets a weight calculated by summing the products of two particle weight factors, which are just 1 for equal momenta and vanish for large separation in momentum space. A reasonable description of the effect in $`Q^2`$ is obtained with a simple gaussian form of the weight factor
$$w_2(p,q)=exp[(pq)^2R_0^2/2],$$
(1)
or, even simpler, $`\theta `$ \- function form with $`w_2=1`$ for some range of $`(pq)^2<1/R_0^2`$ and $`w_2=0`$ outside.
In this method we may repeat the same calculation as done for the LUBOEI procedure. The resulting double ratios are not that smooth and monotically decreasing as in the data or from the LUBOEI procedure (which is the usual drawback of the weight methods). However, the major features are surprisingly similar: with weight factors depending only on $`Q^2`$ we get different values of fitted $`R_i`$ parameters. Moreover, the hierarchy of parameters is the same: $`R_{side}>R_L`$. This suggests that the assymetry is generated by the jet-like structure of final states and not by any specific features of the procedure modelling the BE effect. In Fig. 1b we show the values of the fit parameters as functions of $`R_0`$ for a Gaussian as well as the $`\theta `$-like weight factors. Again, no choice of input parameters allows to describe the data.
These results suggest also that one should be careful with the geometric interpretation of the data. If one gets asymmetric distributions from the generator without assuming explicitly space asymmetry of the source, it is not clear how the assumed asymmetry will be reflected in the results.
We tried to get some information on this problem within the asymmetric weight method, i.e. introducing weight factors which depend in a different way on $`Q_L`$, $`Q_{side}`$ and $`Q_{out}`$. This work is progress. However, it seems rather difficult to reproduce the data even with two more parameters.
## 3 Conclusions and outlook
In this note we investigated the asymmetry of the BE effect in two procedures imitating this effect in the Monte Carlo generators and compared them to the data at $`Z^0`$ peak. Both procedures give surprisingly similar results and disagree with data. The work on the possible introduction of asymmetry in the weight method is in progress.
## Acknowledgments
This work was partially supported by the KBN grants No 2 P03B 086 14, 2 P03B 010 15 and 2 P03B 019 17.
## References
|
warning/0005/math0005253.html
|
ar5iv
|
text
|
# Un th or me de Cartier-Milnor-Moore-Quillen pour les big bres dendriformes et les alg bres braces
## Introduction
On tudie ici un analogue du couple form par une alg bre de Lie et son alg bre associative enveloppante. Le r le de la notion d’alg bre associative est jou par celle d’alg bre dendriforme, introduite par Loday comme structure duale au sens des op rades de celle de dig bre. Le r le de la notion d’alg bre de Lie est tenu par les alg bres braces, apparues d’abord dans les articles de Kadeishvili et Getzler sur l’homologie des alg bres $`A_{\mathrm{}}`$, et plus r cemment dans les travaux sur le complexe de Hochschild et la conjecture de Deligne, notamment ceux de Kontsevich et Soibelman et de Gerstenhaber et Voronov . La notion d’alg bre pr -Lie issue des travaux de Gerstenhaber et Vinberg joue aussi un r le dans la suite.
On r interpr te en termes d’op rades des r sultats de Mar a Ronco qui montrent qu’une alg bre dendriforme est en particulier une alg bre brace , en introduisant l’op rade des alg bres braces et un morphisme d’op rades dans celle des alg bres dendriformes.
On introduit une notion de big bre dendriforme, voisine de la notion d’alg bre de Hopf dendriforme introduite ind pendamment par Ronco dans , mais qui en diff re par la pr sence d’une unit . La pr sence de cette unit offre des avantages techniques, dont celui de simplifier certaines d monstrations.
On d finit ensuite l’alg bre dendriforme enveloppante d’une alg bre brace $`B`$, not e $`\stackrel{~}{𝒰}(B)`$. Par analogie avec le fait que l’alg bre associative enveloppante d’une alg bre de Lie est une big bre, on montre que l’alg bre dendriforme enveloppante d’une alg bre brace est une big bre dendriforme.
Si $`W`$ est une big bre dendriforme, il est d montr dans que l’ensemble $`\mathrm{Prim}(W)`$ des l ments primitifs pour la structure de cog bre sous-jacente est une sous-alg bre brace de $`W`$.
On montre que, si $`B`$ est une alg bre brace, alors $`B=\mathrm{Prim}\stackrel{~}{𝒰}(B)`$ et que, r ciproquement, si $`D`$ est une big bre dendriforme connexe (comme cog bre), alors $`D\stackrel{~}{𝒰}(\mathrm{Prim}(D))`$. On obtient ainsi une quivalence de cat gories qui forme un analogue dendriforme/brace du th or me de Cartier-Milnor-Moore-Quillen.
Le plan de l’article est le suivant :
La premi re partie contient des rappels sur les diff rents types d’alg bres utilis s et sur certaines relations entre eux.
Dans la seconde partie, on introduit l’op rade $`\mathrm{Brace}`$ qui d crit les alg bres braces en termes d’arbres enracin s plans, on en donne une pr sentation par g n rateurs et relations et on d finit un morphisme d’op rades de $`\mathrm{Brace}`$ dans $`\mathrm{Dend}`$. On retrouve ainsi le fait que toute alg bre dendriforme est une alg bre brace, tabli par . On d montre enfin deux propositions techniques qui seront utilis es dans la suite de l’article.
La troisi me partie est consacr e l’introduction des notions d’alg bre dendriforme unitaire et de big bre dendriforme. On rappelle que, d’apr s Ronco, les l ments primitifs d’une big bre dendriforme forment une sous-alg bre brace, puis on nonce deux lemmes sur les big bres dendriformes, qui joueront un r le important dans la suite. On montre ensuite que l’alg bre dendriforme libre est une big bre dendriforme, pour le coproduit introduit par Loday et Ronco dans .
La quatri me partie contient des nonc s g n raux sur les suites exactes courtes d’op rades, les id aux d’op rades et les adjoints des foncteurs d’oubli entre cat gories d’alg bres induits par les morphismes d’op rades, pour lesquels on n’a pas trouv de r f rences dans la litt rature.
Dans la cinqui me partie, on d finit l’alg bre dendriforme enveloppante d’une alg bre brace $`L`$, et l’on montre que c’est une big bre dendriforme, au sens de la seconde partie. On applique cette construction aux cas des alg bres braces libres et des alg bres braces de produits nuls.
Enfin, dans la sixi me partie, on montre que le foncteur alg bre dendriforme enveloppante $`\stackrel{~}{𝒰}`$ tablit une quivalence de cat gories entre la cat gorie des alg bres braces et celle des big bres dendriformes connexes, dont le quasi-inverse (inverse au sens des quivalences de cat gories) est le foncteur $`\mathrm{Prim}`$ des l ments primitifs.
## 1 Rappels sur les alg bres dendriformes et pr -Lie
On rappelle ici, pour la commodit du lecteur et pour les notations, la d finition des diff rents types d’alg bres utilis s. Les notions d’alg bre dendriforme et d’alg bre de Leibniz duale ont t introduites et d velopp es par Loday, voir . Les alg bres pr -Lie ont t introduites ind pendamment par Gerstenhaber et Vinberg . On renvoie aussi pour des r sultats sur l’op rade correspondante.
Une alg bre dendriforme est un espace vectoriel $`W`$ muni de deux produits not s $``$ et $``$ de $`WW`$ dans $`W`$ tels que
$`(xy)z`$ $`=x(yz)+x(yz),`$ (1)
$`(xy)z`$ $`=x(yz),`$ (2)
$`x(yz)`$ $`=(xy)z+(xy)z.`$ (3)
On note $`D(V)`$ l’alg bre dendriforme libre sur un espace vectoriel $`V`$, voir \[14, 2.4\] pour sa d finition en termes de l’op rade $`\mathrm{Dend}`$. Si $`(W,,)`$ est une alg bre dendriforme, on note $``$ le produit d fini par $`xy=xy+xy`$. C’est un produit associatif. On note $`W_{\mathrm{As}}`$ l’alg bre associative $`(W,)`$. Cette construction d finit un morphisme d’op rades $`\mathrm{As}\mathrm{Dend}`$.
Une alg bre pr -Lie est un espace vectoriel $`W`$ muni d’une op ration $``$ de $`WW`$ dans $`W`$ telle que
$$(xy)zx(yz)=(xz)yx(zy).$$
(4)
On note $`PL(V)`$ l’alg bre pr -Lie libre sur un espace vectoriel $`V`$, voir pour sa d finition en termes de l’op rade $`\mathrm{PreLie}`$. Si $`(W,)`$ est une alg bre pr -Lie, on note $`[,]`$ le produit d fini par $`[x,y]=xyyx`$. C’est un crochet de Lie. On note $`W_{\mathrm{Lie}}`$ l’alg bre de Lie $`(W,[,])`$. Cette construction d finit un morphisme d’op rades $`\mathrm{Lie}\mathrm{PreLie}`$.
Une alg bre de Leibniz duale ou $`\mathrm{Zin}`$-alg bre est un espace vectoriel $`W`$ muni d’un produit not $``$ de $`WW`$ dans $`W`$ tel que
$$(xy)z=x(yz)+x(zy).$$
(5)
On note $`Z(V)`$ la $`\mathrm{Zin}`$-alg bre libre sur un espace vectoriel $`V`$, voir \[14, 7.1\] pour sa d finition en termes de l’op rade $`\mathrm{Zin}`$. Si $`(W,)`$ est une $`\mathrm{Zin}`$-alg bre, on note $``$ le produit d fini par $`xy=xy+yx`$. C’est un produit associatif et commutatif. On note $`W_{\mathrm{Com}}`$ l’alg bre associative commutative $`(W,)`$. Cette construction d finit un morphisme d’op rades $`\mathrm{Com}\mathrm{Zin}`$.
Si $`(W,,)`$ est une alg bre dendriforme, on note $``$ le produit d fini par $`xy=xyyx`$. C’est un produit pr -Lie. On note $`W_{\mathrm{PreLie}}`$ l’alg bre pr -Lie $`(W,)`$. Cette construction d finit un morphisme d’op rades $`\mathrm{PreLie}\mathrm{Dend}`$.
En posant $`yx=xy=xy`$, on peut consid rer une $`\mathrm{Zin}`$-alg bre $`W`$ comme une alg bre dendriforme, qui est sym trique au sens o $`xy=yx`$, pour tous $`x,y`$ dans $`W`$. Ceci d finit un morphisme d’op rades $`\mathrm{Dend}\mathrm{Zin}`$.
R ciproquement, on observe que toute alg bre dendriforme sym trique au sens ci-dessus provient d’une $`\mathrm{Zin}`$-alg bre par cette construction. Il y a donc quivalence entre la notion d’alg bre dendriforme sym trique et celle de $`\mathrm{Zin}`$-alg bre. La condition de sym trie entra ne la commutativit de l’alg bre associative $`W_{\mathrm{As}}`$.
On rappelle les morphismes d’op rades bien connus $`\mathrm{Lie}\mathrm{As}\mathrm{Com}`$. Les foncteurs correspondants entre cat gories d’alg bres consistent respectivement associer une alg bre associative l’alg bre de Lie dont le crochet est le commutateur du produit associatif et consid rer une alg bre commutative comme une alg bre associative.
En r sum , on a le diagramme suivant de morphismes d’op rades :
$$\begin{array}{ccccc}\mathrm{Zin}& & \mathrm{Dend}& & \mathrm{PreLie}\\ & & & & & & \\ \mathrm{Com}& & \mathrm{As}& & \mathrm{Lie}.\end{array}$$
(6)
On v rifie ais ment qu’il est commutatif.
Si $`𝒫=(𝒫(k))_{k1}`$ est une op rade telle que $`𝒫(1)`$ soit engendr par l’identit , alors il existe un morphisme naturel de $`𝒫`$ dans l’op rade triviale (r duite l’identit ). On note $`𝒫_+=(𝒫(k))_{k2}`$ le noyau de ce morphisme d’augmentation.
Les deux lignes du diagramme (6) sont exactes au sens suivant : on a $`\mathrm{Com}\mathrm{As}/<\mathrm{Lie}_+>`$ et $`\mathrm{Zin}\mathrm{Dend}/<\mathrm{PreLie}_+>`$, o $`<>`$ d signe l’id al engendr . Ceci signifie que les alg bres commutatives sont exactement les alg bres associatives dont le crochet de Lie sous-jacent est nul. La proposition similaire pour les $`\mathrm{Zin}`$-alg bres est l’identification d crite plus haut avec les alg bres dendriformes sym triques.
Si $`𝒫`$ est une op rade, on notera $`𝒫\mathrm{Alg}`$ la cat gorie des $`𝒫`$-alg bres.
## 2 L’op rade des alg bres braces
On d finit une op rade $`\mathrm{APE}`$ sur les arbres enracin s plans, qui a t implicitement introduite par Kontsevich et Soibelman . On en donne ensuite une pr sentation par g n rateurs et relations qui permet de l’identifier avec l’op rade $`\mathrm{Brace}`$ des alg bres braces utilis es par Gerstenhaber et Voronov . Cette description en termes d’arbres pr sente l’avantage de rendre visuelle la combinatoire des produits braces. Les produits braces ont t g n ralis s par Akman en des produits multi-braces, dont l’ ventuelle relation avec ce qui suit reste clarifier. On montre que le morphisme $`\mathrm{PreLie}\mathrm{Dend}`$ se factorise via l’op rade $`\mathrm{Brace}`$, le morphisme $`\mathrm{Brace}\mathrm{Dend}`$ tant une interpr tation en termes d’op rades des r sultats de . La section se termine par des propositions sur les id aux de $`\mathrm{Dend}`$ engendr s par les images de $`\mathrm{Brace}_+`$ et $`\mathrm{PreLie}_+`$.
### 2.1 D finition
Soit $`I`$ un ensemble fini non vide. On appelle arbre enracin plan sur $`I`$ la donn e d’un graphe connexe sans boucles sur l’ensemble de sommets $`I`$, muni d’un sommet distingu appel la racine et d’un plongement dans le plan. On dessine les arbres avec la racine en bas et on consid re les ar tes comme orient es vers la racine.
Supposons l’arbre $`T`$ dessin dans le demi-disque sup rieur ouvert
$$𝒟_+=\{(x,y)^2y>0\text{ et }x^2+y^2<1\},$$
sauf la racine plac e au point $`x=y=0`$. On appelle angle de $`T`$ une paire $`(s,\alpha )`$ o $`s`$ est un sommet de $`T`$ et $`\alpha `$ une composante connexe de $`B_ϵ(s)(𝒟_+T)`$ o $`B_ϵ(s)`$ est un petit disque de centre $`s`$. On note $`\mathrm{Angles}(T)`$ l’ensemble des angles de $`T`$. On munit naturellement $`\mathrm{Angles}(T)`$ d’un ordre total de gauche droite, de la fa on suivante. En consid rant un angle comme une direction issue d’un sommet, on peut tracer un chemin de chaque angle vers un point du cercle unit . On peut aussi supposer que ces chemins ne se coupent pas. On obtient alors un point du demi-cercle associ chaque angle. L’ordre de ces points de gauche droite ne d pend pas des chemins choisis pourvu qu’ils ne se coupent pas. Ceci d finit l’ordre total voulu, voir la figure 1.
Cette notion d’angle est due Kontsevich , \[11, p. 29-30\], de m me que la composition de l’op rade qui suit.
On note $`𝒜𝒫(I)`$ l’ensemble des arbres enracin s plans sur $`I`$ et $`\mathrm{APE}(I)`$ le $``$-module libre sur $`𝒜𝒫(I)`$. Soit $`T𝒜𝒫(I)`$. Si $`j`$ est un sommet de $`T`$, on note $`\mathrm{Entr}(T,j)`$ l’ensemble totalement ordonn de gauche droite des ar tes entrantes en $`j`$. On peut alors consid rer l’ensemble des fonctions croissantes au sens large de $`\mathrm{Entr}(T,j)`$ dans $`\mathrm{Angles}(S)`$.
On d finit une op rade $`\mathrm{APE}`$ sur les modules $`\mathrm{APE}(I)`$. La composition d’un arbre $`S𝒜𝒫(I)`$ au sommet $`j`$ d’un arbre $`T𝒜𝒫(J)`$ est d finie par
$$T_jS=\underset{f:\mathrm{Entr}(T,j)\mathrm{Angles}(S)}{}T_j^fS,$$
(7)
o $`f`$ est une application croissante au sens large et $`T_j^fS`$ est l’arbre obtenu par substitution de l’arbre $`S`$ au sommet $`j`$ de $`T`$, les ar tes de $`T`$ entrantes dans $`j`$ tant greff es sur $`S`$ selon l’application $`f`$. On donne un exemple dans la figure 2, avec $`11`$ angles pour $`S`$ et $`2`$ ar tes entrantes dans le sommet $`1`$ de $`T`$, ce qui donne $`66`$ termes dans la composition.
###### Proposition 1
La composition ci-dessus d finit une op rade $`\mathrm{APE}`$.
Preuve : On v rifie ais ment les axiomes d’unit , d’associativit et d’ quivariance. $`\mathrm{}`$
### 2.2 Pr sentation
On montre maintenant que l’op rade $`\mathrm{APE}`$ admet une pr sentation par g n rateurs et relations qui permet de l’identifier l’op rade des alg bres braces.
On note $`\mathrm{Corl}_n(1;2,\mathrm{},n+1)`$ ou simplement $`\mathrm{Corl}_n`$ la corolle $`n`$ feuilles index e par $`\{1,\mathrm{},n+1\}`$, la racine portant l’indice $`1`$, les feuilles les indices $`2,\mathrm{},n+1`$ dans l’ordre de gauche droite. On va prendre des l ments $`C_n`$ correspondant ces corolles comme g n rateurs. Les relations entre ces corolles sont celles entre les op rations $`x_1\{x_2,\mathrm{},x_{n+1}\}`$ des alg bres braces, voir \[7, Eq. (6)\] par exemple. Ces relations expriment une composition de corolles la racine comme une somme de compositions multiples aux feuilles.
###### Proposition 2
L’op rade $`\mathrm{APE}`$ est isomorphe au quotient de l’op rade libre sur la famille de g n rateurs $`C_n`$ en $`n+1`$ variables pour $`n1`$ (on note $`\{x_1x_2,\mathrm{},x_{n+1}\}`$ les op rations correspondantes) par les relations qui s’expriment en termes d’op rations de la fa on suivante :
$$\begin{array}{c}\{\{zx_1,\mathrm{},x_n\}y_1,\mathrm{},y_m\}=\hfill \\ \hfill \{zY_0,\{x_1Y_1\},Y_2,\{x_2Y_3\},Y_4\mathrm{},Y_{2n2},\{x_nY_{2n1}\},Y_{2n}\},\end{array}$$
(8)
o la somme porte sur les partitions de l’ensemble ordonn $`\{y_1,\mathrm{},y_m\}`$ en intervalles successifs ventuellement vides $`Y_0Y_1\mathrm{}Y_{2n}`$.
Preuve : On note $`\mathrm{Brace}`$ l’op rade quotient consid r e. Elle d crit exactement les alg bres braces au sens de . On d finit un morphisme d’op rades $`\kappa `$ de l’op rade libre sur les $`C_n`$ dans $`\mathrm{APE}`$ en posant $`\kappa (C_n)=\mathrm{Corl}_n`$. Comme les corolles v rifient les relations (8), le morphisme $`\kappa `$ passe au quotient et d finit un morphisme d’op rades $`\beta :\mathrm{Brace}\mathrm{APE}`$.
La surjectivit de $`\kappa `$ est claire, car tout arbre s’obtient par composition de corolles aux feuilles. Le morphisme $`\beta `$ est donc surjectif.
D’autre part, on d duit des relations (8) que l’on peut d finir un morphisme surjectif de $``$-modules $`\gamma `$ de $`\mathrm{APE}(n)`$ sur $`\mathrm{Brace}(n)`$. La compos e $`\beta \gamma `$ est surjective, donc un isomorphisme de $``$-modules libres. Par cons quent $`\gamma `$ est injective, donc bijective et $`\beta `$ est un isomorphisme d’op rades. $`\mathrm{}`$
Cette proposition permet d’identifier les op rades $`\mathrm{APE}`$ et $`\mathrm{Brace}`$. On appelle alg bres braces les $`\mathrm{Brace}`$-alg bres. On note $`B(V)`$ l’alg bre brace libre sur un espace vectoriel $`V`$.
### 2.3 Un morphisme de $`\mathrm{Brace}`$ dans $`\mathrm{Dend}`$
On d finit dans cette section un morphisme d’op rades de $`\mathrm{Brace}`$ dans $`\mathrm{Dend}`$.
On d finit par r currence des l ments $`_n`$ et $`^n`$ de $`\mathrm{Dend}`$ :
$$_1(x_1)=x_1,_n(x_1,\mathrm{},x_n)=x_1_{n1}(x_2,\mathrm{},x_n),$$
(9)
et
$$^1(x_1)=x_1,^{n+1}(x_1,\mathrm{},x_{n+1})=^n(x_1,\mathrm{},x_n)x_{n+1}.$$
(10)
Par la suite, on notera simplement $``$ et $``$ en omettant le nombre d’arguments lorsque cela ne pr te pas confusion.
On a alors la proposition suivante :
###### Proposition 3
Il existe un unique morphisme d’op rades $`\psi `$ de $`\mathrm{Brace}`$ dans $`\mathrm{Dend}`$ tel que pour tout $`n`$, $`\psi (\mathrm{Corl}_n)\mathrm{Dend}(n+1)`$ soit l’op ration qui $`x_1,\mathrm{},x_{n+1}`$ associe
$$\underset{i=1}{\overset{n+1}{}}(1)^i(x_2,\mathrm{},x_i)x_1(x_{i+1},\mathrm{},x_{n+1}),$$
(11)
expression sans ambigu t par la formule $`(\text{2})`$, et o on convient que les termes pour $`i=1`$ et $`i=n+1`$ sont respectivement
$$x_1(x_2,\mathrm{},x_{n+1})\text{ et }(x_2,\mathrm{},x_{n+1})x_1.$$
(12)
Preuve : L’unicit est claire puisque que les op rations $`\mathrm{Corl}_n`$ engendrent $`\mathrm{Brace}`$.
On peut prouver la compatibilit aux relations (8) en adaptant la preuve du th or me 3.4 de . On peut aussi en donner une preuve l g rement diff rente, qu’on ne d taillera pas ici, bas e sur la description de l’op rade $`\mathrm{Dend}`$ en termes d’arbres binaires plans . $`\mathrm{}`$
### 2.4 Un morphisme de $`\mathrm{PreLie}`$ dans $`\mathrm{Brace}`$
On d finit dans cette section un morphisme de $`\mathrm{PreLie}`$ dans $`\mathrm{Brace}`$. On renvoie le lecteur pour la description de l’op rade $`\mathrm{PreLie}`$ en termes d’arbres non-plans enracin s et la d finition pr cise de cette notion.
###### Proposition 4
Il existe un unique morphisme d’op rades $`\varphi `$ de $`\mathrm{PreLie}`$ dans $`\mathrm{Brace}`$ tel que
$$\varphi (x_1x_2)=\{x_1x_2\}.$$
(13)
Ce morphisme est donn par la sym trisation des arbres : si $`T`$ est un arbre enracin non-plan index par un ensemble $`I`$, $`\varphi (T)`$ est la somme des arbres enracin s plans isomorphes $`T`$ comme arbres enracin s non-plans.
Preuve : L’unicit r sulte du fait que $`\mathrm{PreLie}`$ est engendr e par l’op ration $`x_1x_2`$. L’existence est un calcul imm diat avec la relation qui d finit $`\mathrm{PreLie}`$. Pour la seconde assertion, il suffit alors de v rifier que l’application de sym trisation d finit bien un morphisme d’op rades, ce qui est clair par la forme des compositions en termes d’arbres dans les op rades $`\mathrm{PreLie}`$ et $`\mathrm{Brace}`$, voir section 2.1 et . $`\mathrm{}`$
Il est clair que le morphisme compos $`\psi \varphi `$ co ncide avec le morphisme de $`\mathrm{PreLie}\mathrm{Dend}`$ d fini dans la premi re section.
### 2.5 Id al engendr par l’image de $`\mathrm{Brace}_+`$ dans $`\mathrm{Dend}`$
La proposition suivante permet d’identifier l’op rade quotient de $`\mathrm{Dend}`$ par l’id al engendr par $`\psi (\mathrm{Brace}_+)`$.
###### Proposition 5
Les images de $`\mathrm{PreLie}_+`$ et de $`\mathrm{Brace}_+`$ dans $`\mathrm{Dend}`$ engendrent le m me id al. L’op rade quotient de $`\mathrm{Dend}`$ par cet id al est l’op rade $`\mathrm{Zin}`$.
Preuve : On a d j vu dans la section 1 que $`\mathrm{Zin}`$ est le quotient de $`\mathrm{Dend}`$ par l’id al engendr par $`\mathrm{PreLie}_+`$. Il reste donc montrer la premi re assertion.
Comme $`\varphi `$ fait de $`\mathrm{PreLie}`$ une sous-op rade de $`\mathrm{Brace}`$, on a clairement une inclusion $`<\psi \varphi (\mathrm{PreLie}_+)><\psi (\mathrm{Brace}_+)>`$. Pour montrer la r ciproque, il suffit de montrer que $`\psi (\mathrm{Corl}_n)`$ appartient $`<\psi \varphi (\mathrm{PreLie})>`$, pour tout $`n1`$.
On rappelle que $`\psi (\mathrm{Corl}_n)`$ est, par d finition, l’op ration qui $`x_1,\mathrm{},x_{n+1}`$ associe
$$\begin{array}{c}x_1(x_2,\mathrm{},x_{n+1})x_1+(1)^{n+1}(x_2,\mathrm{},x_{n+1})\hfill \\ \hfill +\underset{i=2}{\overset{n}{}}(1)^i(x_2,\mathrm{},x_i)x_1(x_{i+1},\mathrm{},x_{n+1}).\end{array}$$
(14)
Pour chaque terme de la somme de $`i=2`$ $`i=n`$, on utilise la relation
$$\begin{array}{c}(x_2,\mathrm{},x_i)x_1(x_{i+1},\mathrm{},x_{n+1})\hfill \\ \hfill ((x_2,\mathrm{},x_i)(x_{i+1},\mathrm{},x_{n+1}))x_1,\end{array}$$
modulo $`<\psi \varphi (\mathrm{PreLie})>`$. On r crit aussi le terme correspondant $`i=1`$ :
$$x_1(x_2,\mathrm{},x_{n+1})(x_2,\mathrm{},x_{n+1})x_1,$$
modulo $`<\psi \varphi (\mathrm{PreLie})>`$. On obtient donc au total
$$\begin{array}{c}(x_2,\mathrm{},x_{n+1})x_1+(1)^{n+1}(x_2,\mathrm{},x_{n+1})x_1\hfill \\ \hfill +\underset{i=2}{\overset{n}{}}(1)^i((x_2,\mathrm{},x_i)(x_{i+1},\mathrm{},x_{n+1}))x_1,\end{array}$$
c’est- -dire une expression de la forme $`Zx_1`$, o
$$\begin{array}{c}Z=(x_2,\mathrm{},x_{n+1})+(1)^{n+1}(x_2,\mathrm{},x_{n+1})\hfill \\ \hfill +\underset{i=2}{\overset{n}{}}(1)^i((x_2,\mathrm{},x_i)(x_{i+1},\mathrm{},x_{n+1})).\end{array}$$
Le lemme 2.6 de dit exactement que $`Z`$ est nul dans $`\mathrm{Dend}`$, ce qui termine la d monstration. $`\mathrm{}`$
### 2.6 Id al gauche engendr par l’image de $`\mathrm{Brace}`$
On montre dans cette section que l’id al gauche (au sens d fini ci-dessous) engendr par l’image de $`\mathrm{Brace}_+`$ dans $`\mathrm{Dend}`$ est un id al.
On rappelle qu’un $`𝕊`$-module $`E`$ est une collection de $`𝔖_n`$-modules $`E(n)`$ pour $`n1`$. Si $`e_m𝒫(m)`$, $`f_n𝒫(n)`$ et $`i\{1,\mathrm{},m\}`$, on note $`e_m_if_n`$ la composition de $`f_n`$ la place $`i`$ de $`e_m`$. On renvoie le lecteur peu familier avec les op rades la r f rence standard .
###### D finition 1
Un id al gauche dans une op rade $`𝒫`$ est un sous-$`𝕊`$-module $`𝒥`$ tel que pour tous $`f_n𝒥(n)`$, $`e_m𝒫(m)`$ et $`i\{1,\mathrm{},m\}`$, on ait $`e_m_if_n𝒥`$.
On peut rapprocher cette notion de celle d’id al gauche dans un anneau associatif. En particulier, un id al est aussi un id al gauche.
Voici un crit re simple pour montrer qu’un id al gauche est un id al. La d monstration est une cons quence facile des axiomes d’op rades et est laiss e au lecteur.
###### Lemme 1
Soient $`𝒫`$ une op rade et $`𝒥`$ un id al gauche de $`𝒫`$. Soient $`E`$ un syst me de g n rateurs de $`𝒫`$ comme op rade et $`F`$ un syst me de g n rateurs de $`𝒥`$ comme id al gauche. Alors $`𝒥`$ co ncide avec l’id al engendr par $`F`$ si et seulement si, pour tout $`e_n`$ dans $`E(n)`$, tout $`f_m`$ dans $`F(m)`$ et tout $`i\{1,\mathrm{},m\}`$, on a $`f_m_ie_n𝒥`$.
Soit $`𝒥`$ l’id al gauche engendr par l’image de $`\mathrm{Brace}_+`$ dans $`\mathrm{Dend}`$. On aura besoin de deux lemmes pr liminaires sur le $`𝕊`$-module quotient de $`\mathrm{Dend}`$ par $`𝒥`$.
###### Lemme 2
Pour tout entier non nul $`n`$, on a dans $`\mathrm{Dend}(n)`$ modulo $`𝒥`$,
$$(x_n,\mathrm{},x_1)(x_1,\mathrm{},x_n).$$
(15)
Preuve : Par r currence sur $`n`$. La formule est vraie pour $`n=1`$. Soit maintenant $`n2`$ fix et supposons la formule v rifi e pour tous les entiers strictement inf rieurs $`n`$. On a $`(x_n,\mathrm{},x_1)x_n(x_{n1},\mathrm{},x_1)`$, donc par hypoth se de r currence, $`(x_n,\mathrm{},x_1)x_n(x_1,\mathrm{},x_{n1})`$. Ceci se r crit modulo $`𝒥`$, en ajoutant $`\psi (\mathrm{Corl}_{n1}(x_n;x_1,\mathrm{},x_{n1}))`$,
$$\underset{i=1}{\overset{n1}{}}(1)^i(x_1,\mathrm{},x_i)x_n(x_{i+1},\mathrm{},x_{n1}).$$
(16)
On va utiliser le sous-lemme suivant dans une seconde r currence d croissante sur $`k`$.
###### Sous-lemme 1
Pour $`k2`$, modulo $`𝒥`$, on a
$$\begin{array}{c}\underset{i=1}{\overset{k}{}}(1)^i(x_1,\mathrm{},x_i)z(x_{i+1},\mathrm{},x_k)\hfill \\ \hfill \underset{i=1}{\overset{k1}{}}(1)^i(y_1,\mathrm{},y_i)z(y_{i+1},\mathrm{},y_{k1}).\end{array}$$
(17)
o $`y_1=x_1x_2`$ et $`y_i=x_{i+1}`$ pour $`2ik1`$.
Preuve du sous-lemme : La diff rence entre les deux membres est
$$x_1(\psi (\mathrm{Corl}_{k1}(z;x_2,\mathrm{},x_k))),$$
qui appartient bien $`𝒥`$. C.Q.F.D.
Par application r p t e du sous-lemme la formule (16), on la r crit $`(x_1,\mathrm{},x_n)`$, ce qui termine la d monstration du lemme. $`\mathrm{}`$
On note $`\mathrm{Pli}(p,q)`$ l’ensemble des permutations de $`\{1,\mathrm{},p+q\}`$ telles que $`\sigma (p)<\mathrm{}<\sigma (1)`$ et $`\sigma (p+1)<\mathrm{}<\sigma (p+q)`$. Ces permutations sont des battages o on a renvers l’ordre d’une des deux parties avant de battre.
Le lemme suivant permet de r crire toute op ration de $`\mathrm{Dend}`$ modulo $`𝒥`$ comme une somme sur certains ensembles de battages.
###### Lemme 3
Pour tous entiers strictement positifs $`p,q`$, on a modulo $`𝒥`$,
$$(x_p,\mathrm{},x_1)z(x_{p+q},\mathrm{},x_{p+1})\underset{\sigma \mathrm{Pli}(p,q)}{}(x_{\sigma (1)},\mathrm{},x_{\sigma (p+q)},z).$$
Preuve : Par r currence sur le couple $`(p,q)`$. La formule est vraie pour $`(p,q)=(1,1)`$. Par le lemme 2, on a, modulo $`𝒥`$,
$$z(x_{p+q},\mathrm{},x_{p+1})(x_{p+1},\mathrm{},x_{p+q})z(x_{p+q},\mathrm{},x_{p+1})z.$$
On en d duit que
$$\begin{array}{c}(x_p,\mathrm{},x_1)z(x_{p+q},\mathrm{},x_{p+1})\hfill \\ \hfill ((x_p,\mathrm{},x_2)x_1(x_{p+q},\mathrm{},x_{p+1}))z\\ \hfill +((x_p,\mathrm{},x_1)x_{p+q}(x_{p+q1},\mathrm{},x_{p+1}))z.\end{array}$$
Ceci permet de conclure sans difficult . $`\mathrm{}`$
On va maintenant d montrer la proposition suivante :
###### Proposition 6
L’id al gauche $`𝒥`$ engendr par l’image de $`\mathrm{Brace}_+`$ dans $`\mathrm{Dend}`$ est un id al.
Preuve : D’apr s le lemme 1, et en tenant compte de l’automorphisme de $`\mathrm{Dend}`$ donn par la sym trie des arbres plans, il suffit de v rifier que $`\psi (\mathrm{Corl}_n)_j`$ est dans $`𝒥`$ pour tout $`j`$. On distingue deux cas, selon que la composition est dans la racine ou dans une feuille de la corolle, c’est- -dire $`j=1`$ ou $`j>1`$.
On traite d’abord le cas de la composition la racine, c’est- -dire
$$\psi (\mathrm{Corl}_n(z;x_1,\mathrm{},x_n))_z(wt),$$
(18)
soit encore
$$\underset{i=0}{\overset{n}{}}(1)^i(x_1,\mathrm{},x_i)((wt)(x_{i+1},\mathrm{},x_n)),$$
(19)
ce qui est gal modulo $`𝒥`$, par application du lemme 2,
$$\underset{i=0}{\overset{n}{}}(1)^i(x_1,\mathrm{},x_i)(tw(x_{i+1},\mathrm{},x_n)).$$
(20)
On utilise deux fois le lemme 3 pour r crire ceci comme une somme sur $`i`$ et sur un ensemble de battages. Chaque battage s’obtient alors pour deux indices $`i`$ successifs, donc contribue par z ro la somme, qui est donc nulle.
On traite de fa on similaire le cas de la composition dans une feuille, en se ramenant une somme de battage l’aide des lemmes 2 et 3. $`\mathrm{}`$
## 3 Big bres dendriformes
On se place dans cette section et dans toute la suite sur un corps $`𝕂`$. On d finit des notions d’alg bre dendriforme unitaire et de big bre dendriforme. Apr s avoir rappel que, d’apr s , les l ments primitifs d’une big bre dendriforme $`D`$ forment une sous-alg bre brace de $`D_{\mathrm{Brace}}`$, on nonce quelques lemmes sur les big bres dendriformes qui seront utiles dans la suite. On donne une premi re famille d’exemples de big bres dendriformes : les alg bres dendriformes libres.
### 3.1 D finitions
On introduit d’abord une notion d’alg bre dendriforme unitaire, voir \[14, p. 32, (5.5.4)\]. Je remercie Mar a Ronco pour m’avoir signal une erreur importante dans une version pr c dente de cette d finition.
###### D finition 2
On appelle alg bre dendriforme unitaire un espace vectoriel $`W`$ muni d’une d composition $`W=𝕂1V`$ et de deux applications $`,:VW+WVV`$ telles que $`(V,,)`$ soit une alg bre dendriforme et que, pour tout $`xV`$, on ait
$$\begin{array}{c}\hfill 1x=x1=0,\\ \hfill x1=1x=x.\end{array}$$
(21)
Si $`D`$ est une alg bre dendriforme, alors $`\stackrel{~}{D}:=𝕂1D`$ est naturellement munie d’une structure d’alg bre dendriforme unitaire. On associe une alg bre associative unitaire augment e une alg bre dendriforme unitaire en posant $`xy=xy+xy`$ et $`11=1`$.
A l’aide de la remarque qui identifie les $`\mathrm{Zin}`$-alg bres aux alg bres dendriformes sym triques, on d finit de mani re analogue la notion de $`\mathrm{Zin}`$-alg bre unitaire.
On appelle alg bre dendriforme unitaire filtr e (resp. gradu e) une alg bre dendriforme unitaire $`W`$ munie d’une filtration (resp. d’une graduation) $`(W_n)_{n0}`$ telle que $`1W_0`$ et si $`xW_p`$ et $`yW_q`$, alors $`xyW_{p+q}`$ et $`xyW_{p+q}`$. L’espace gradu associ une alg bre dendriforme unitaire filtr e est naturellement une alg bre dendriforme unitaire gradu e.
Si $`W`$ est une alg bre dendriforme unitaire, on note $`W^+`$ le second facteur de la d composition $`W=𝕂1W^+`$. Un morphisme d’alg bres dendriformes unitaires $`f:D_1D_2`$ est une application qui respecte les d compositions et telle que $`f(1)=1`$ et $`f:D_1^+D_2^+`$ soit un morphisme dendriforme. Ceci d finit la cat gorie des alg bres dendriformes unitaires, not e $`\mathrm{Dend}\mathrm{Alg}u`$. Le lemme suivant est vident.
###### Lemme 4
Le foncteur $`D\stackrel{~}{D}`$ de $`\mathrm{Dend}\mathrm{Alg}`$ dans $`\mathrm{Dend}\mathrm{Alg}u`$ est une quivalence de cat gories, de quasi-inverse $`WW^+`$.
On introduit maintenant une notion de big bre dendriforme. Cette d fi-nition est voisine de celle donn e par Ronco dans sous le nom d’alg bre de Hopf dendriforme. La diff rence r side dans la pr sence d’une unit , qui simplifie quelque peu la forme des axiomes et de certaines d mons-trations, au prix d’un certain abus de notation.
###### D finition 3
On appelle big bre dendriforme une alg bre dendriforme unitaire $`W=𝕂1V`$ munie d’un coproduit coassociatif $`\mathrm{\Delta }:WWW`$ v rifiant les conditions suivantes :
* $`\mathrm{\Delta }(1)=11`$,
* la projection sur $`𝕂1`$ parall lement $`V`$ est une co nit .
Ces conditions entra nent, pour $`vV`$, $`\mathrm{\Delta }(v)1v+v1+VV`$.
* Le coproduit est compatible aux produits dendriformes, au sens suivant : pour tous $`x,yV`$, avec la notation de Sweedler,
$$\mathrm{\Delta }(xy)=(x_{(1)}y_{(1)})(x_{(2)}y_{(2)})(xy)(11)+(xy)1,$$
(22)
et
$$\mathrm{\Delta }(xy)=(x_{(1)}y_{(1)})(x_{(2)}y_{(2)})(xy)(11)+(xy)1,$$
(23)
o la soustraction des termes en $`11`$ et $`11`$ est une notation commode pour signifier que dans la sommation de Sweedler, on doit omettre l’unique terme correspondant $`x_{(2)}=y_{(2)}=1`$.
On utilisera par la suite sans davantage de commentaires cette convention sur les termes non-existants en $`11`$ et $`11`$.
Si $`(W,,,\mathrm{\Delta })`$ est une big bre dendriforme, alors $`W`$ est en particulier une big bre pour le produit associatif $``$ et le coproduit $`\mathrm{\Delta }`$.
On appelle big bre dendriforme filtr e (resp. gradu e) une big bre dendriforme munie d’une filtration (resp. d’une graduation) qui en fait la fois une alg bre dendriforme unitaire filtr e (resp. gradu e) et une cog bre filtr e (resp. gradu e). L’espace gradu associ une big bre dendriforme filtr e est naturellement une big bre dendriforme gradu e.
### 3.2 l ments primitifs d’une big bre dendriforme
Si $`C`$ est une cog bre, on note $`\mathrm{Prim}(C)`$ le sous-espace des l ments primitifs de $`C`$. On rappelle le fait remarquable, tabli par Ronco , que les l ments primitifs d’une big bre dendriforme $`D`$ forment une sous-alg bre brace de $`D_{\mathrm{Brace}}^+`$.
On aura besoin du lemme suivant dans les sections 5 et 6.
###### Lemme 5
Soient $`(W,,,\mathrm{\Delta })`$ une big bre dendriforme et $`x_1,\mathrm{},x_n`$ des l ments primitifs de $`W`$. Alors
$$\mathrm{\Delta }((x_1,\mathrm{},x_n))=\underset{i=0}{\overset{n}{}}(x_1,\mathrm{},x_i)(x_{i+1},\mathrm{},x_n),$$
(24)
o on convient que $`()=1`$.
Preuve : Par r currence, l’aide des formules (22) et (23), voir \[20, 2.7\], pour plus de d tails. $`\mathrm{}`$
###### Proposition 7
Si $`W=𝕂1V`$ est une big bre dendriforme, l’ensemble $`\mathrm{Prim}(W)`$ des l ments primitifs de $`W`$ est une sous-alg bre brace de $`V_{\mathrm{Brace}}`$.
Preuve : Il r sulte aussit t de la d finition que $`\mathrm{Prim}W`$ est contenu dans $`V`$. On renvoie \[20, Prop 2.8\] pour la d monstration du fait que $`\mathrm{Prim}W`$ est une sous-alg bre brace. $`\mathrm{}`$
### 3.3 Lemmes
Les lemmes suivants, qui sont des cons quences directes des axiomes de big bre dendriforme, jouent un r le crucial pour la suite.
###### Lemme 6
Si $`W=𝕂1V`$ est une big bre dendriforme et $`IV`$ un coid al de $`W`$, alors l’id al dendriforme de $`V`$ engendr par $`I`$ est un coid al de $`W`$.
Preuve : Soit $`I`$ un coid al de $`W`$ contenu dans $`V`$. Il r sulte des formules (22) et (23) que $`𝔰(I)=I+VI+IV+VI+IV`$ est aussi un coid al de $`W`$ contenu dans $`V`$. On d finit par r currence
$$𝔰^n(I):=𝔰(𝔰^{n1}(I)).$$
Les coid aux $`𝔰^n(I)`$ sont inclus dans $`V`$ et forment une suite croissante. Soit alors $`𝔖(I)`$ l’union des $`𝔰^n(I)`$ pour $`n1`$. C’est encore un coid al de $`W`$ et il est clair que c’est l’id al dendriforme de $`V`$ engendr par $`I`$. Ceci termine la d monstration. $`\mathrm{}`$
###### Lemme 7
Soient $`D_1=𝕂1D_1^+,D_2=𝕂1D_2^+`$ deux big bres dendriformes et $`f:D_1D_2`$ un morphisme d’alg bres dendriformes unitaires. Si $`D_1^+`$ est engendr e par $`V`$ comme alg bre dendriforme et si $`\mathrm{\Delta }f=(ff)\mathrm{\Delta }`$ sur $`V`$, alors $`f`$ est un morphisme de big bres dendriformes.
Preuve : On a clairement $`\mathrm{\Delta }(f(1))=(ff)\mathrm{\Delta }(1)`$. Consid rons le sous-espace
$$K:=\{xD_1^+\mathrm{\Delta }(f(x))=(ff)\mathrm{\Delta }(x)\}.$$
Par hypoth se, $`K`$ contient $`V`$. On montre sans difficult l’aide des formules (22) et (23) que $`K`$ est une sous-alg bre dendriforme de $`D_1^+`$. Comme $`D_1^+`$ est engendr e par $`V`$, on a $`K=D_1^+`$, ce qui termine la d monstration. $`\mathrm{}`$
### 3.4 Alg bres dendriformes libres
On montre que les alg bres dendriformes libres sont des big bres dendriformes. Ce r sultat a galement t obtenu par Mar a Ronco . On en pr sente ici une d monstration l g rement diff rente qui utilise la pr sence d’une unit . On consid re l’alg bre dendriforme unitaire $`\stackrel{~}{D}(V):=𝕂1D(V)`$ o $`D(V)`$ est l’alg bre dendriforme libre sur $`V`$.
Le coproduit est celui introduit par Loday et Ronco dans qui, stricto sensu, ne traite que le cas de l’alg bre libre sur un g n rateur. L’extension des r sultats de au cas g n ral est toutefois imm diate, voir pour plus de d tails. On obtient ainsi la description suivante du coproduit $`\mathrm{\Delta }`$.
Pour $`x,y\stackrel{~}{D}(V)`$ et $`vV`$, on pose d’abord
$$x\stackrel{v}{}y=xvy.$$
Alors, le coproduit $`\mathrm{\Delta }`$ v rifie, pour $`x,y\stackrel{~}{D}(V)`$ et $`vV`$,
$$\mathrm{\Delta }(x\stackrel{v}{}y)=(x\stackrel{v}{}y)1+(x_{(1)}y_{(1)})(x_{(2)}\stackrel{v}{}y_{(2)}),$$
(25)
et il est enti rement d termin par ces relations et la condition $`\mathrm{\Delta }(1)=11`$. On remarque que le second terme du membre de droite contient $`1(x\stackrel{v}{}y)`$.
###### Proposition 8
L’alg bre dendriforme unitaire $`\stackrel{~}{D}(V)`$, munie du coproduit $`\mathrm{\Delta }`$, est une big bre dendriforme.
Preuve : On d duit de l’identit $`x(yz)=(xy)z`$ dans $`D(V)`$ la formule suivante, pour $`x,w,zD(V)`$ et $`vV`$ :
$$x(w\stackrel{v}{}z)=(xw)\stackrel{v}{}z.$$
(26)
Elle est encore valable pour $`x,w,z\stackrel{~}{D}(V)`$.
Soient maintenant $`x,yD(V)`$. Comme on suppose que $`yD(V)`$, on peut crire de mani re unique $`y=w\stackrel{v}{}z`$ avec $`w,z\stackrel{~}{D}(V)`$ et $`vV`$. On a alors
$$\begin{array}{c}\mathrm{\Delta }(xy)=\mathrm{\Delta }(x(w\stackrel{v}{}z))=\mathrm{\Delta }((xw)\stackrel{v}{}z)\hfill \\ \hfill =((xw)\stackrel{v}{}z)1+((xw)_{(1)}z_{(1)})((xw)_{(2)}\stackrel{v}{}z_{(2)}).\end{array}$$
Comme $`\mathrm{\Delta }`$ est un morphisme d’alg bres associatives, ceci est gal
$$((xw)\stackrel{v}{}z)1+(x_{(1)}w_{(1)}z_{(1)})((x_{(2)}w_{(2)})\stackrel{v}{}z_{(2)}).$$
(27)
D’autre part, en utilisant la convention des termes fant mes en $`11`$ (voir Def. 3), on a
$$\begin{array}{c}(x_{(1)}y_{(1)})(x_{(2)}y_{(2)})(xy)(11)+(xy)1=\hfill \\ \hfill (x_{(1)}(w\stackrel{v}{}z)_{(1)})(x_{(2)}(w\stackrel{v}{}z)_{(2)})(x(w\stackrel{v}{}z))(11)+(x(w\stackrel{v}{}z))1\\ \hfill =(x(w\stackrel{v}{}z))(11)+(x_{(1)}w_{(1)}z_{(1)})(x_{(2)}(w_{(2)}\stackrel{v}{}z_{(2)}))\\ \hfill (x(w\stackrel{v}{}z))(11)+(x(w\stackrel{v}{}z))1\\ \hfill =((xw)\stackrel{v}{}z)1+(x_{(1)}w_{(1)}z_{(1)})((x_{(2)}w_{(2)})\stackrel{v}{}z_{(2)}).\end{array}$$
(28)
Ceci termine la preuve de la formule (22). La formule (23) s’en d duit par l’automorphisme de sym trie gauche-droite des arbres plans. $`\mathrm{}`$
## 4 Foncteurs adjoints et suites exactes courtes
### 4.1 Foncteurs adjoints
La d finition de l’alg bre dendriforme enveloppante passe d’abord par celle de l’adjoint gauche du foncteur d’oubli $`()_{\mathrm{Brace}}`$ qui associe une alg bre dendriforme l’alg bre brace sous-jacente. La proposition suivante a pour vocation essentielle de rappeler la construction g n rale de ce type de foncteur adjoint, qui est sans difficult et sans doute bien connue.
###### Proposition 9
Soit $`\gamma :𝒫𝒬`$ un morphisme d’op rades. Le foncteur d’oubli $`\mathrm{\Gamma }:𝒬\mathrm{Alg}𝒫\mathrm{Alg}`$ admet un adjoint gauche, not $`U_\mathrm{\Gamma }`$.
Preuve : Soit $`P`$ une $`𝒫`$-alg bre, on d finit $`U_\mathrm{\Gamma }(P)`$ comme le quotient de la $`𝒬`$-alg bre libre sur $`P`$, not e $`F_𝒬P`$, par le $`𝒬`$-id al engendr par les relations
$$\gamma (e_n)(p_1,\mathrm{},p_n)e_n(p_1,\mathrm{},p_n)$$
(29)
pour tous $`n1`$, $`e_n𝒫(n)`$ et $`p_1,\mathrm{},p_n`$ dans $`P`$. On observe qu’il revient au m me de quotienter par ces relations pour des $`e_n`$ d crivant seulement un ensemble $`E`$ de g n rateurs de $`𝒫`$.
Par construction, l’application naturelle de $`P`$ dans $`F_𝒬P`$ fournit un $`𝒫`$-morphisme $`\tau _P`$ de $`P`$ dans $`U_\mathrm{\Gamma }(P)`$. Il faut v rifier qu’il a bien la propri t universelle voulue.
Soient $`D`$ une $`𝒬`$-alg bre et $`\varphi :P\mathrm{\Gamma }(D)`$ un $`𝒫`$-morphisme. Alors $`\varphi `$ se prolonge en un $`𝒬`$-morphisme $`\mathrm{\Phi }:F_𝒬PD`$. Comme $`\varphi `$ est un un $`𝒫`$-morphisme, $`\mathrm{\Phi }`$ se factorise en un $`𝒬`$-morphisme $`\overline{\mathrm{\Phi }}:U_\mathrm{\Gamma }(P)D`$. On a clairement $`\overline{\mathrm{\Phi }}\tau _P=\varphi `$. Un $`𝒬`$-morphisme qui v rifie cette relation est unique, car $`U_\mathrm{\Gamma }(P)`$ est engendr e comme $`𝒬`$-alg bre par $`\tau _P(P)`$.
On a donc une bijection, donn e par la composition avec $`\tau _P`$,
$$\mathrm{Hom}_{𝒫\mathrm{Alg}}(P,\mathrm{\Gamma }(D))\mathrm{Hom}_{𝒬\mathrm{Alg}}(U_\mathrm{\Gamma }(P),D).$$
On v rifie sans peine que l’on obtient ainsi une quivalence de bifoncteurs. Ceci d montre la proposition. $`\mathrm{}`$
La cas particulier qui nous int resse ici est le suivant :
###### Corollaire 1
Le foncteur $`()_{\mathrm{Brace}}`$ de $`\mathrm{Dend}\mathrm{Alg}`$ dans $`\mathrm{Brace}\mathrm{Alg}`$ admet un adjoint gauche, not $`𝒰`$.
Explicitement, si $`B`$ est une alg bre brace, $`𝒰(B)`$ est le quotient de l’alg bre dendriforme libre $`D(B)`$ par l’id al dendriforme engendr par les l ments
$$\psi (\mathrm{Corl}_n)(x_1,x_2,\mathrm{},x_{n+1})\mathrm{Corl}_n(x_1,x_2,\mathrm{},x_{n+1}),$$
(30)
pour $`n2`$ et $`x_1,\mathrm{},x_{n+1}B`$.
Le cas des alg bres braces libres $`B(V)`$ est particuli rement simple. On le d veloppe ici pour un usage ult rieur.
###### Lemme 8
Les alg bres dendriformes $`𝒰(B(V))`$ et $`D(V)`$ sont isomorphes.
Preuve : Le foncteur d’oubli $`\mathrm{Dend}\mathrm{Alg}\mathrm{Vect}_𝕂`$ est la compos e des foncteurs d’oubli $`\mathrm{Dend}\mathrm{Alg}\mathrm{Brace}\mathrm{Alg}`$ et $`\mathrm{Brace}\mathrm{Alg}\mathrm{Vect}_𝕂`$. Comme l’adjoint d’un foncteur compos est la compos e des foncteurs adjoints, et que pour toute op rade $`𝒫`$, le foncteur $`𝒫`$-alg bre libre est adjoint au foncteur d’oubli de $`𝒫\mathrm{Alg}\mathrm{Vect}_𝕂`$, on obtient le r sultat. $`\mathrm{}`$
### 4.2 Suites exactes courtes d’op rades
Soit $`P`$ une $`𝒫`$-alg bre. Alors, l’ensemble $`𝒥`$ des op rations de $`𝒫`$ qui sont nulles dans $`P`$, c’est- -dire le noyau du morphisme $`𝒫\mathrm{End}(P)`$, forme un id al (bilat re) au sens des op rades. On cherche ici sous quelles conditions la nullit de certaines op rations de $`𝒫`$ sur $`P`$ peut se d duire de leur nullit sur un sous-espace $`V`$ de $`P`$ engendrant $`P`$ comme $`𝒫`$-alg bre.
On omettra ici les d monstrations des deux prochains lemmes, qui sont des exercices de manipulation des axiomes d’op rades, sans surprises. Le lemme suivant justifie l’introduction de la notion d’id al gauche.
###### Lemme 9
Soit $`P`$ une $`𝒫`$-alg bre et $`V`$ un sous-espace de $`P`$. L’ensemble des op rations de $`𝒫`$ nulles sur $`V`$ forme un id al gauche de $`𝒫`$.
###### Lemme 10
Soit $`P`$ une $`𝒫`$-alg bre, $`V`$ un sous-espace de $`P`$ qui engendre $`P`$ comme $`𝒫`$-alg bre et $`𝒥`$ un id al de $`𝒫`$. Si les op rations de $`𝒥`$ sont nulles sur $`V`$, elles sont nulles sur $`P`$.
Soient $`𝒫`$ une op rade augment e, $`𝒫_+`$ son id al d’augmentation et $`\gamma :𝒫𝒬`$ un morphisme d’op rades. On note comme ci-dessus $`U_\mathrm{\Gamma }`$ l’adjoint du foncteur $`\mathrm{\Gamma }`$ de $`𝒬\mathrm{Alg}`$ dans $`𝒫\mathrm{Alg}`$ induit par $`\gamma `$. On a alors la
###### Proposition 10
On suppose que $`𝒥`$, l’id al gauche de $`𝒬`$ engendr par $`\gamma (𝒫_+)`$ est un id al. On note $`𝒞`$ l’op rade quotient de $`𝒬`$ par $`𝒥`$. Si $`P`$ est une $`𝒫`$-alg bre annul e par $`𝒫_+`$, alors $`U_\mathrm{\Gamma }(P)`$ est une $`𝒞`$-alg bre, isomorphe la $`𝒞`$-alg bre libre sur $`P`$.
Preuve : Soit $`\tau `$ l’inclusion de $`P`$ dans $`U_\mathrm{\Gamma }(P)`$. Il r sulte de l’hypoth se et de la d finition de $`U_\mathrm{\Gamma }(P)`$ que les op rations $`\gamma (𝒫_+)`$ sont nulles sur $`\tau (P)`$. Donc, d’apr s le lemme 9, les op rations de $`𝒥`$ sont nulles sur $`\tau (P)`$. Comme $`U_\mathrm{\Gamma }(P)`$ est engendr e comme $`𝒬`$-alg bre par $`\tau (P)`$ et comme, par hypoth se, $`𝒥`$ est un id al, il r sulte du lemme 10 que les op rations de $`𝒥`$ sont nulles sur $`U_\mathrm{\Gamma }(P)`$. Par cons quent, $`U_\mathrm{\Gamma }(P)`$ est une $`𝒞`$-alg bre.
Il reste montrer que la $`𝒞`$-alg bre $`U_\mathrm{\Gamma }(P)`$, poss de la propri t universelle de la $`𝒞`$-alg bre libre sur $`P`$. Soit $`Z`$ une $`𝒞`$-alg bre et $`\varphi :PZ`$ un morphisme d’espaces vectoriels. Comme $`Z`$ est aussi une $`𝒬`$-alg bre, il existe un unique $`𝒬`$-morphisme $`\stackrel{~}{\varphi }:F_𝒬PZ`$ prolongeant $`\varphi `$. Comme $`Z`$ est une $`𝒞`$-alg bre, ce morphisme passe au quotient pour donner un $`𝒬`$-morphisme $`\overline{\varphi }:U_\mathrm{\Gamma }(P)Z`$ tel que $`\overline{\varphi }\tau =\varphi `$. Comme $`U_\mathrm{\Gamma }(P)`$ et $`Z`$ sont des $`𝒞`$-alg bres, $`\overline{\varphi }`$ est un $`𝒞`$-morphisme. Enfin, un tel morphisme est unique, car $`P`$ engendre $`U_\mathrm{\Gamma }(P)`$ comme $`𝒬`$-alg bre, donc aussi comme $`𝒞`$-alg bre. $`\mathrm{}`$
## 5 Alg bre dendriforme enveloppante
### 5.1 D finition et coproduit
On d finit l’alg bre dendriforme enveloppante d’une alg bre brace $`B`$, not e $`\stackrel{~}{𝒰}(B)`$, comme l’alg bre dendriforme unitaire associ e $`𝒰(B)`$.
###### Proposition 11
$`\stackrel{~}{𝒰}(B)`$ est une big bre dendriforme.
Preuve : Comme $`\mathrm{Prim}(\stackrel{~}{D}(B))`$ est une sous $`\mathrm{Brace}`$-alg bre de $`D(B)`$, les l ments apparaissant dans (30) appartiennent $`\mathrm{Prim}(\stackrel{~}{D}(B))`$, et donc forment un coid al de $`\stackrel{~}{D}(B)`$. Par le lemme 6, l’id al dendriforme de $`D(B)`$ engendr par ces l ments est encore un coid al de $`\stackrel{~}{D}(B)`$, donc la structure de big bre dendriforme de $`\stackrel{~}{D}(B)`$ passe au quotient. $`\mathrm{}`$
On note $`B(V)`$ l’alg bre brace libre sur un espace vectoriel $`V`$.
Par le lemme 8, l’alg bre dendriforme enveloppante $`\stackrel{~}{𝒰}(B(V))`$ est isomorphe $`\stackrel{~}{D}(V)`$. On retrouve ainsi la big bre dendriforme des arbres binaires plans, dans la terminologie de , voir le paragraphe 3.4.
### 5.2 Cas des alg bres braces de produits nuls
On appelle alg bre brace triviale une alg bre brace dont tous les produits $`\{\mathrm{?}\mathrm{?}\}_n`$ sont nuls.
On donne une description en termes d’objets classiques de l’alg bre dendriforme enveloppante d’une alg bre brace triviale.
On note $`T^c(V)`$ la cog bre tensorielle sur $`V`$, voir \[19, 1.4,1.5\] pour la d finition. On rappelle que $`T^c(V)`$ poss de une structure de big bre, o le coproduit est donn par la d concatenation et le produit commutatif par les battages, et telle que $`\mathrm{Prim}(T^c(V))=V`$, cf \[19, loc. cit.\]. On rappelle deux propri t s classiques : $`T^c(V)`$ est la cog bre coassociative connexe colibre sur $`V`$ et $`T^c(V)`$ est duale au sens gradu de l’alg bre tensorielle $`T(V)`$.
On note $`\stackrel{~}{Z}(V)`$ la $`\mathrm{Zin}`$-alg bre unitaire $`𝕂1Z(V)`$. D’apr s \[14, 7.1\], $`\stackrel{~}{Z}(V)_{\mathrm{Com}}`$ s’identifie comme alg bre commutative $`T^c(V)`$. Par cons quent, $`T^c(V)`$ est naturellement une alg bre dendriforme sym trique et $`\stackrel{~}{Z}(V)_{\mathrm{Com}}`$ une big bre.
Soit maintenant $`V`$ une alg bre brace triviale. On a alors la description suivante de $`\stackrel{~}{𝒰}(V)`$ :
###### Proposition 12
Soit $`V`$ une alg bre brace triviale.
* Il existe un unique isomorphisme d’alg bres dendriformes unitaires $`\theta :\stackrel{~}{𝒰}(V)\stackrel{~}{Z}(V)`$ qui est l’identit sur $`V`$.
* De plus, $`\theta `$ est un isomorphisme de cog bres.
Preuve : Par la proposition 6, on peut appliquer la proposition 10 au morphisme $`\psi :\mathrm{Brace}\mathrm{Dend}`$. Par la proposition 5, l’op rade quotient est $`\mathrm{Zin}`$. Par la proposition 10, il existe donc un unique isomorphisme dendriforme de $`𝒰(V)`$ dans $`Z(V)`$ qui est l’identit sur $`V`$. Par le lemme 4, ce morphisme se rel ve de mani re unique en un isomorphisme d’alg bres dendriformes unitaires de $`\stackrel{~}{𝒰}(V)`$ dans $`\stackrel{~}{Z}(V)`$. On obtient la premi re assertion.
Il reste v rifier que le coproduit sur $`\stackrel{~}{𝒰}(V)`$ d fini par la proposition 11 co ncide avec le coproduit de d concat nation sur $`\stackrel{~}{Z}(V)=T^c(V)`$. Explicitons pour cela l’unique $`\mathrm{Zin}`$-morphisme de $`Z(V)`$ dans $`𝒰(V)`$ qui fixe $`V`$. On montre sans difficult que l’image du tenseur $`x_1x_2\mathrm{}x_n`$ est la classe de l’ l ment $`(x_1,\mathrm{},x_n)`$.
Il r sulte alors du lemme 5 que le coproduit sur $`\stackrel{~}{𝒰}(V)`$ correspond par l’isomorphisme la d concatenation dans $`\stackrel{~}{Z}(V)`$. $`\mathrm{}`$
On en d duit imm diatement le corollaire suivant, qui a aussi t not par Ronco .
###### Corollaire 2
Il existe sur $`T^c(V)`$ une structure de big bre dendriforme dont la structure de big bre sous-jacente est la structure usuelle de la big bre des battages.
## 6 Une quivalence de cat gories
On montre dans cette section que le foncteur $`\stackrel{~}{𝒰}`$ est une quivalence de la cat gorie des alg bres braces dans celle des big bres dendriformes connexes, de quasi-inverse le foncteur $`\mathrm{Prim}`$. Ceci forme un quivalent dendriforme/brace du th or me de Cartier-Milnor-Moore-Quillen sur les big bres cocommutatives connexes et les alg bres de Lie .
On dit qu’une big bre dendriforme sur $`𝕂`$ est connexe si la cog bre sous-jacente est connexe au sens de Quillen , c’est- -dire si le coradical est $`𝕂`$ et si la filtration par le coradical est exhaustive.
Soit $`D`$ une big bre dendriforme et notons $`B`$ le sous-espace des l ments primitifs de $`D`$. D’apr s la proposition 7, $`B`$ est une sous-alg bre brace de $`D_{\mathrm{Brace}}^+`$. Par cons quent, d’apr s la propri t d’adjonction du foncteur $`𝒰`$, on obtient un morphisme d’alg bres dendriformes de $`𝒰(B)`$ dans $`D^+`$, qu’on rel ve en un morphisme d’alg bres dendriformes unitaires $`\theta `$ de $`\stackrel{~}{𝒰}(B)`$ dans $`D`$. On a la proposition suivante.
###### Proposition 13
$`\theta `$ est un morphisme de cog bres.
Preuve : La restriction de $`\theta `$ $`B`$ v rifie $`\mathrm{\Delta }\theta =1\theta +\theta 1=(\theta \theta )\mathrm{\Delta }`$. Comme $`B`$ engendre $`𝒰(B)`$ comme alg bre dendriforme, le lemme 7 permet de conclure. $`\mathrm{}`$
On rappelle le lemme classique suivant.
###### Lemme 11
Soit $`f:CC^{}`$ un morphisme de cog bres. Si $`C`$ est connexe et si $`f`$ restreint aux primitifs de $`C`$ est injectif, alors $`f`$ est injectif.
Preuve : Voir \[18, Appendice B, Prop. 3.2\] ou \[1, Th. 2.4.11\]. $`\mathrm{}`$
La proposition suivante joue un r le crucial dans la suite.
###### Proposition 14
Soit $`B`$ une alg bre brace, alors $`\stackrel{~}{𝒰}(B)`$ est isomorphe comme cog bre $`T^c(B)`$. Par cons quent, $`\mathrm{Prim}\stackrel{~}{𝒰}(B)=B`$.
Preuve : On d finit une application lin aire $`f`$ de $`T^c(B)`$ dans $`\stackrel{~}{𝒰}(B)`$ par $`f(x_1\mathrm{}x_n)=(x_1,\mathrm{},x_n)`$. Par le lemme 5, $`f`$ est un morphisme de cog bres. Comme la cog bre $`T^c(B)`$ est connexe et $`f`$ injectif sur $`B`$, le lemme 11 entra ne que $`f`$ est injectif.
D’autre part, la big bre dendriforme $`\stackrel{~}{𝒰}(B)`$ h rite comme quotient de $`\stackrel{~}{D}(B)`$ d’une structure de big bre dendriforme filtr e. En notant $`𝒥`$ le noyau du morphisme $`\stackrel{~}{D}(B)\stackrel{~}{𝒰}(B)`$, c’est- -dire l’id al dendriforme de $`D(B)`$ engendr par les $`\psi (\mathrm{Corl}_n)\mathrm{Corl}_n`$, on voit aussit t que l’id al $`gr𝒥`$ de $`D(B)=grD(B)`$ contient les l ments $`\psi (\mathrm{Corl}_n)`$. Par cons quent, $`gr\stackrel{~}{𝒰}(B)`$ est un quotient de $`\stackrel{~}{𝒰}(B_0)`$, o $`B_0`$ d signe $`B`$ munie de la structure brace nulle. Comme $`\stackrel{~}{𝒰}(B_0)T^c(B)`$ comme big bre dendriforme, il en r sulte que $`gr\stackrel{~}{𝒰}(B)`$ est engendr comme espace vectoriel par les l ments $`(x_1,\mathrm{},x_n)`$. Par cons quent, $`\stackrel{~}{𝒰}(B)`$ est aussi engendr comme espace vectoriel par ces l ments, donc $`f`$ est surjective. $`\mathrm{}`$
On peut maintenant d montrer la proposition suivante. La preuve s’inspire de celle donn e par Quillen dans \[18, Appendice B, Th. 4.5\].
###### Proposition 15
Soit $`D`$ une big bre dendriforme suppos e connexe comme cog bre et $`B:=\mathrm{Prim}(D)`$ l’alg bre brace des primitifs de $`D`$. Alors on a un isomorphisme naturel de big bres dendriformes $`\stackrel{~}{𝒰}(B)D`$.
Preuve : Il s’agit de montrer que le morphisme de big bres dendriformes naturel $`\theta `$ de $`\stackrel{~}{𝒰}(B)`$ dans $`D`$ est bijectif.
Comme $`\stackrel{~}{𝒰}(B)T^c(B)`$ comme cog bre, $`\stackrel{~}{𝒰}(B)`$ est connexe. Comme $`\theta `$ est injectif sur $`B`$, $`\theta `$ est injectif par le lemme 11. Par cons quent, en choisissant arbitrairement un suppl mentaire de $`\stackrel{~}{𝒰}(B)`$ dans $`D`$, on voit qu’il existe une application lin aire $`\pi :DB`$ telle que $`\pi \theta `$ soit la projection canonique $`p:\stackrel{~}{𝒰}(B)B`$. Mais, comme $`\stackrel{~}{𝒰}(B)T^c(B)`$ comme cog bre, elle a la propri t universelle de la cog bre connexe colibre, c’est- -dire il existe un unique morphisme de cog bres $`g:D\stackrel{~}{𝒰}(B)`$ tel que $`pg=\pi `$. Comme $`D`$ est connexe par hypoth se et que $`g`$ restreint aux primitifs $`B`$ de $`D`$ est injectif, $`g`$ est injectif par le lemme 11.
D’autre part, $`g\varphi `$ est un endomorphisme de la cog bre $`\stackrel{~}{𝒰}(B)`$ tel que $`g\varphi p=p`$, donc par la propri t universelle, $`g\varphi =\mathrm{Id}`$, donc $`g`$ est aussi surjectif. On conclut que $`g`$ est bijectif d’inverse $`\varphi `$, donc $`\varphi `$ est un isomorphisme de big bres dendriformes, ce qui termine la d monstration. $`\mathrm{}`$
En combinant les propositions 14 et 15, on obtient le th or me principal.
###### Th or me 1
Le foncteur $`B\stackrel{~}{𝒰}(B)`$ est une quivalence entre la cat gorie des alg bres braces et celle des big bres dendriformes connexes, le foncteur quasi-inverse tant $`D\mathrm{Prim}(D)`$.
Remerciements. Je voudrais remercier ici Jim Stasheff pour son inter t et ses remarques, et mon directeur de th se Patrick Polo pour son soutien constant et ses suggestions sur la derni re partie.
Fr d ric CHAPOTON
quipe Analyse alg brique, case 82
Institut de Math matiques, Universit Pierre et Marie Curie
4, Place Jussieu, 75252 Paris Cedex 05, FRANCE
Mel. : chapoton$`\mathrm{@}`$math.jussieu.fr
|
warning/0005/cond-mat0005295.html
|
ar5iv
|
text
|
# Hall coefficient and ARPES in systems with strong pair fluctuations
## I INTRODUCTION
The anomalous temperature and doping behavior of transport coefficients such as the dc resistivity $`\rho `$, the Hall coefficient $`R_H`$, the magnetoresistance $`\mathrm{\Delta }\rho `$ and the Hall angle $`cot\theta _H=\rho /R_H`$ in the normal state of high-$`T_c`$ cuprate superconductors (HTS) are amongst the clearest indications for non-Fermi liquid behavior in these materials. Quite generally $`R_H`$ shows a rapid decrease with increasing temperature and in certain overdoped samples can change from positive to small negative values. This, together with the Hall angle showing an unexpected quadratic temperature variation in the underdoped regime, has led to speculations about different transport mechanisms in presence and in absence of a magnetic field, based on the phenomenological introduction of two different scattering rates - perpendicular and parallel to the Fermi surface. Such a description may hold in non-Fermi liquids, provided they retain sufficient Fermi liquid properties so that the concept of scattering rates and a description in terms of a Boltzmann equation is still meaningful. When this is not guaranteed, the interpretation of the anomalous transport properties has to be sought in the single-particle spectral features themselves, strongly affected by many-body effects. The Mott-Hubbard scenario of a strongly correlated system as well as the $`tJ`$ spin correlation picture, which are believed to capture certain features of the underdoped HTS, have been investigated in this connection and show qualitative similarities with the experimental results.
We shall in this Letter address the question of the anomalous magneto-transport processes in terms of a scenario where the Fermi liquid properties get destroyed in the normal state due to dynamical electron pair formation showing up in the appearence of a pseudogap. The decrease of $`R_H`$ with increasing temperature is then expected to be related to thermal excitations across such a pseudogap feature in the single-particle spectra. The present study aims to examine the underdoped and the overdoped systems on equal footing.
It is presently a matter of debate whether the onset of the pseudogap at $`T^{}`$ is a precursor of the superconducting state or not. If not, pair fluctuations are a necessary but not a sufficient condition for the system to become superconducting upon lowering the temperature. The present experimental situation seems to suggest that the opening of the pseudogap is not triggered by superconducting precursor effects. If it were, then superconducting correlations should show up in the entire temperature regime of the pseudogap phase, giving rise to an enhancement of the dc conductivity, the diamagnetism etc. into a Meissner state. Such features are indeed observed, but only in a very reduced fraction of the pseudogap temperature regime relatively close to $`T_c`$, where such pair fluctuations exist on a time scale of the order of $`T_c`$. Recent experiments on the frequency dependence of the Meissner screening seem also to confirm that.
Presently there are two phenomenological models, the negative-U Hubbard model and the Boson-Fermion model (BFM), which are studied in connection with those experimental findings. In the negative-U Hubbard model the HTS physics is treated as a cross-over phenomenon between a BCS state in the overdoped regime and something close to a Bose-Einstein condensation (BEC) of preformed pairs in the underdoped regime. Going from one extreme doping situation to the other one is monitored by a variation of the attractive interaction U between the electrons. In the BFM the HTS physics is described by a mechanism by which pairing in an electronic subsystem is induced by resonant scattering of the electrons in and out of localized tightly bound pair electron (bosonic) states. Doping is monitored by changing the position of the bosonic energy level which alters the relative occupation between the electrons and the bosons. The bosonic states in the BFM have their counterpart in the negative-U Hubbard system in form of two-particle resonant states above the chemical potential.
The BFM was originally introduced and studied in some detail in the past as regards the single-particle spectral properties of the electrons. The qualitative differences which exist between the underdoped and the overdoped regime warrant to examine whether and how these differences can possibly be held responsible for the unusual temperature and doping dependence of certain magneto-transport properties seen in the HTS. Such a study must be done in a way which relates the above mentioned transport coefficients to the single-particle properties in a rigorous fashion. In the limit of infinite dimensions this can be achieved within the dynamical mean-field theory approach which we shall adopt here.
The hamiltonian for the BFM is given by
$`H`$ $`=`$ $`\epsilon _0{\displaystyle \underset{i,\sigma }{}}c_{i\sigma }^{}c_{i\sigma }t{\displaystyle \underset{ij,\sigma }{}}c_{i\sigma }^{}c_{j\sigma }`$ (2)
$`+E_0{\displaystyle \underset{i}{}}b_i^{}b_i+g{\displaystyle \underset{i}{}}[b_i^{}c_ic_i+c_i^{}c_i^{}b_i].`$
$`c_{i\sigma }^{()}`$ denote annihilation (creation) operators for electrons with spin $`\sigma `$ at some effective sites $`i`$ (involving molecular units rather than individual atoms) and $`b_i^{()}`$ denote hard-core bosonic operators describing tightly bound electron pairs. $`t`$, $`D`$, $`\mathrm{\Delta }_B`$ and $`g`$ denote respectively the bare hopping integral for the electrons, the bare electronic half bandwidth, the boson energy level and the boson-fermion pair-exchange coupling constant. Furthermore we put $`\epsilon _0=D\mu `$ and $`E_0=\mathrm{\Delta }_B2\mu `$ and assume the chemical potential $`\mu `$ to be common to fermions and bosons (up to a factor 2 for the bosons) in order to guarantee charge conservation. These parameters are fixed in such a way that at temperatures large compared to the interaction $`g`$ the number of fermions $`n_F=_{i,\sigma }c_{i\sigma }^{}c_{i\sigma }`$ lies in the interval $`[1,0.75]`$ which covers the typical experimental doping regime. We furthermore choose $`n=n_F+2n_B`$ ($`n_B=\frac{1}{N}_ib_i^{}b_i`$ denoting the number of bosonic electron pairs) to lie in the interval $`[1,2]`$ in order to account for the appearence of a pseudogap phase. For the present analysis we choose $`n=1.5`$ as a representative value, having verified that in the regime $`n=[1.2,1.8]`$ the various physical quantities which we shall discuss here exhibit qualitatively the same behavior. Finally, in order to have a $`T^{}`$ of the order of a few hundred degrees K we choose $`g=0.2`$ (all energies are in units of the bare electronic bandwidth $`2D`$). From our previous studies we know that the anomalous electronic properties in the normal state are largely determined by those on an atomic or molecular level. The dynamical mean-field approach permits to properly deal with the local electronic structure which is renormalized by the itinerancy of the electrons via a Weiss mean-field.
## II ON THE ORIGIN OF THE PSEUDOGAP
In order to highlight the physics leading up to the pseudogap phenomenon, let us first consider the above Hamiltonian in the atomic limit i.e., $`t_{ij}=0`$. In that case we derive the following set of eigenstates $`|n`$ and eigenvalues $`E_n`$:
$`|1`$ $`=`$ $`|0,E_1=\mathrm{\hspace{0.33em}0}`$ (3)
$`|2`$ $`=`$ $`|,E_2=\epsilon _0`$ (4)
$`|3`$ $`=`$ $`|,E_3=\epsilon _0`$ (5)
$`|4`$ $`=`$ $`u|v|,E_4=\epsilon _0+E_0/2\gamma `$ (6)
$`|5`$ $`=`$ $`v|+u|,E_5=\epsilon _0+E_0/2+\gamma `$ (7)
$`|6`$ $`=`$ $`|,E_6=\epsilon _0+E_0`$ (8)
$`|7`$ $`=`$ $`|,E_7=\epsilon _0+E_0`$ (9)
$`|8`$ $`=`$ $`|,E_8=\mathrm{\hspace{0.33em}2}\epsilon _0+E_0`$ (10)
The presence on a given site of an electron with spin up and down, respectively, is denoted by corresponding arrows and the presence of a Boson is denoted by a dot. Moreover we have
$`u^2`$ $`=`$ $`{\displaystyle \frac{1}{2}}\left[\mathrm{\hspace{0.33em}1}{\displaystyle \frac{\epsilon _0E_0/2}{\gamma }}\right],v^2={\displaystyle \frac{1}{2}}\left[\mathrm{\hspace{0.33em}1}+{\displaystyle \frac{\epsilon _0E_0/2}{\gamma }}\right]`$ (11)
$`\gamma `$ $`=`$ $`\left[(\epsilon _0E_0/2)^2+g^2\right]^{1/2},uv={\displaystyle \frac{g}{2\gamma }}`$ (12)
The corresponding atomic Green’s function takes the form
$`G_0(i\omega _n)={\displaystyle _0^\beta }𝑑\tau e^{i\omega _n\tau }<T[c_{}(\tau )c_{}^{}]>=`$ (13)
$`{\displaystyle \frac{Z_f}{i\omega _n\epsilon _0}}+\left(1Z_f\right)\left[{\displaystyle \frac{u^2}{i\omega _nE_4+\epsilon _0}}+{\displaystyle \frac{v^2}{i\omega _nE_5+\epsilon _0}}\right]`$ (14)
where the spectral weight of the non-bonding single-particle excitations is given by
$$Z_f=\frac{1}{Z_0}\left(1+e^{\beta \epsilon _0}+e^{\beta (\epsilon _0+E_0)}+e^{\beta (2\epsilon _0+E_0)}\right)$$
(15)
and $`Z_0=_nexp(E_n/k_BT)`$ denotes the partition function. The spectral weights for the bonding and antibonding two-particle states are given by $`(1Z_f)u^2`$ and $`(1Z_f)v^2`$, respectively. We stress that the form of $`G_0(i\omega _n)`$ is formally equivalent to that of a BCS Green’s function, where $`g^2`$ plays the role of the usual gap function, which, of course, in this lowest order approximation does not depend on temperature. $`Z_f`$ in general decreases with decreasing temperature, which is the crucial effect that controls the pseudogap physics in this model. It leads, already on a purely local electronic consideration, to a situation where the spectral weight of the single-particle states diminishes with decreasing temperature, while at the same time the spectral weight of the two-particle bonding and antibonding states increases. These features are reminiscent of the situation of a BCS superconductor below $`T_c`$ where the role of the two-particle states is played by the Cooper pairs.
## III DYNAMICAL MEAN FIELD THEORY APPROACH TO THE BFM
In developing a theoretical description which accounts for the itinerancy of the electrons, care must be taken in fully taking into account the local electronic structure which contains the essential ingredients of the pseudogap phase. This can be done by using as a starting basis the atomic limit of the model and then introducing the effects of the electron itinerancy within the Dynamical Mean Field Theory (DMFT) . In this approach a many-body system is seen as a purely local system coupled to a ”medium”, representing a Weiss field to be determined self-consistently. The problem is cast into the form of a single-impurity Anderson problem described by an effective Hamiltonian
$`H={\displaystyle \underset{\sigma }{}}\epsilon _0c_\sigma ^{}c_\sigma +E_0b^{}b+g[c_{}^{}c_{}^{}b+b^{}c_{}c_{}]`$ (16)
$`+{\displaystyle \underset{k,\sigma }{}}w_kd_{k\sigma }^{}d_{k\sigma }+{\displaystyle \underset{k,\sigma }{}}v_k[d_{k\sigma }^{}c_\sigma +c_\sigma ^{}d_{k\sigma }]`$ (17)
where $`c_\sigma ^{()}`$ and $`b^{()}`$ denote the original fermion and boson operators, respectively, at a single impurity site and $`d_{k\sigma }^{()}`$ denote the auxiliary Fermionic operators associated with the Weiss field. The evaluation of their energy spectrum $`w_k`$ and their coupling $`v_k`$ to the impurity electrons has to be performed in a self-consistent way.
It can be shown that the impurity Green’s function can be cast into the form
$$G_{imp}(\omega _n)=\frac{1}{i\omega _n\epsilon _0\mathrm{\Sigma }_W(\omega _n)\mathrm{\Sigma }_{int}(\omega _n)}$$
(18)
which makes explicit two momentum-independent contributions to the self-energy: one contribution being due to the boson-fermion exchange coupling and denoted by $`\mathrm{\Sigma }_{int}(\omega )`$, and one contribution being due to the hybridization of the impurity center with the medium and denoted by $`\mathrm{\Sigma }_W(\omega )`$. This latter quantity, depending on the parameters $`v_k`$ and $`w_k`$ and generally referred to as the Weiss self-energy, is determined self-consistently by equating the impurity Green’s function (18) to the integral over the $`𝐤`$ space of the lattice Green’s function
$`G_{lat}(\omega _n,\epsilon _𝐤)={\displaystyle \frac{1}{i\omega _n+\mu \epsilon _𝐤\mathrm{\Sigma }_{int}(\omega _n)}}`$ (19)
where $`\epsilon _𝐤`$ denotes the bare electron dispersion. Upon replacement of the integration over $`𝐤`$ by an integration over energy, the integrated lattice Green’s function takes the form
$$G_{lat}(\omega _n)=𝑑\epsilon \frac{\rho (\epsilon )}{i\omega _n+\mu \epsilon \mathrm{\Sigma }_{int}(\omega _n).}$$
(20)
Assuming, as is usually done, a Bethe lattice in infinite dimensions, the density of states (DOS) for the bare electrons appearing in eq.(20) is the semi-circular DOS $`\rho (\epsilon )=(1/2\pi t^2)\sqrt{\epsilon (4t\epsilon )}`$ of width $`2D=4t`$. The self-consistency condition $`G_{imp}(\omega _n)=G_{lat}(\omega _n)`$ implies
$$\mathrm{\Sigma }_W(\omega _n)=t^2G_{imp}(\omega _n)$$
(21)
or alternatively
$$\mathrm{\Delta }(\omega )=t^2A_F(\omega )$$
(22)
where $`A_F(\omega )=2\mathrm{Im}G_{imp}(\omega _n=\omega +i\delta )`$ is the fermionic DOS and
$$\mathrm{\Delta }(\omega )=\mathrm{\hspace{0.33em}2}\pi \underset{k}{}v_k^2\delta (\omega w_k)$$
(23)
is the spectral function associated with the self-energy of the auxiliary fermions.
The above impurity problem is solved here within the so-called Non Crossing Approximation (NCA) , along the lines presented and discussed in Ref.. Within the DMFT framework, NCA has recently been applied to several other models, such as a multiband Hubbard model for perovskites , the Anderson lattice model with correlated conduction electrons and the Kondo lattice model with correlated conduction electrons .
One serious drawback of the DMFT+NCA procedure is that it does not recover the non-interacting limit, corresponding in the BFM to the case $`g=0`$. The point is that this approach leads to some kinematic interactions which result in a self-energy $`\mathrm{\Sigma }_{int}(\omega )`$ which does not vanish as the coupling constant $`g`$ tends to zero. In this section we highlight this shortcoming for the fully symmetric case, realized when the boson site energy and the chemical potential are both pinned in the middle of the fermionic band, giving $`n=2`$ for all temperatures. In the top figure of Fig.1 we plot $`\mathrm{Im}\mathrm{\Sigma }_{int}(\omega )`$ at $`T=0.04`$ for different coupling constants $`g`$, as obtained within the conventional NCA approach. We notice a strong dependence on $`g`$ for the frequency regime around the chemical potential, characterized by a rapid decrease as $`g`$ tends to zero, as it should be. On the contrary, for larger frequencies $`\mathrm{Im}\mathrm{\Sigma }_{int}(\omega )`$ is sizeable and shows little dependence on the value of $`g`$. This fact indicates the existence of kinematical interactions which, contrary to what one would expect, are effective even for $`g=0`$ and thus should be subtracted out in the calculation procedure. In order to do that, we first consider the case without interaction ($`g=0`$) for a given temperature and calculate self-consistently the interaction self-energy $`\mathrm{\Sigma }_{int}^{g=0}(\omega )`$ within the standard DMFT+NCA approach. As just pointed out, this leads to an unphysical expression which over a large range of energies does not vanish, as it ought to in the absence of interaction. We then repeat the whole calculation for finite boson-fermion coupling at the same temperature, obtaining a given expression $`\mathrm{\Sigma }_{int}(\omega )`$ for the interaction self-energy. Finally, we replace in the integrated lattice Green’s function (20) the self-energy $`\mathrm{\Sigma }_{int}(\omega )`$ calculated at finite $`g`$ by $`\mathrm{\Sigma }_{int}(\omega )\mathrm{\Sigma }_{int}^{g=0}(\omega )`$, thus subtracting out the unphysical part of the interaction self-energy coming from the non-interacting case. In this way eq.(20) is rewritten in the form
$$G_{lat}^{}(\omega )=𝑑\epsilon \frac{\rho (\epsilon )}{\omega +\mu \epsilon \mathrm{\Sigma }_{int}(\omega )+\mathrm{\Sigma }_{int}^{g=0}(\omega )},$$
(24)
thus leading to a redefined integrated lattice Green’s function which in the non-interacting case correctly gives back the fermionic DOS $`\rho (\epsilon )`$ for free electrons.
In the bottom panel of Fig.1 we illustrate the behavior of $`\mathrm{Im}\mathrm{\Sigma }_{int}(\omega )`$ as evaluated within this redefined NCA approach. From a comparison with the curves reported in the upper panel, we notice that as far as the low frequency behavior around the chemical potential is concerned, the two methods give results which are in better and better agreement as increasing values of $`g`$ are considered. Outside this frequency regime the self-energy is in general small when evaluated within the redefined NCA, going to zero for $`g=0`$, as it actually ought to be the case.
This difference in the self-energy also affects the behavior of the density of states. In Fig.2 we compare the fermionic DOS evaluated within the conventional NCA, given by $`A_F(\omega )=2\mathrm{Im}G_{lat}(\omega )`$, with that evaluated within the redefined NCA approach, given by $`A_F^{}(\omega )=2\mathrm{Im}G_{lat}^{}(\omega )`$. We again notice that for small values of $`g`$ the discrepancy between the two approaches is noticeable, with unphysical tails in the DOS obtained from the standard NCA which should not be present. As higher values of the coupling $`g`$ are considered, the agreement between the two NCA approaches is increasingly improved, in particular in the frequency regime close to the chemical potential.
The situation away from the fully symmetric case is numerically more involved. We have nonetheless verified that the results obtained in a number of specific cases for $`n2`$ are in agreement with our findings for the fully symmetric case. Since here we are interested in analyzing cases in which the hybridization constant $`g`$ is of the order of $`0.2`$, we conclude that we can safely work with the conventional NCA approach, which in this regime is expected to give correct results. This will be done for a variety of cases in the next sections. In particular, our aim is to evaluate the single-particle fermionic Green’s function from which measureable quantities can be derived, such as the angle-resolved direct and inverse photoemission spectra (ARPES and ARIPES, respectively), and certain transport coefficients for which vertex corrections, in the limit of infinite dimensions, can be neglected.
## IV THE ARPES AND INVERSE ARPES SPECTRUM
We examine first of all the doping and temperature dependence of the spectral function of the lattice Green’s function for fermions $`A_F(\epsilon _𝐤,\omega )=2\mathrm{Im}G_{lat}(\epsilon _𝐤,\omega )`$. Within the BFM scenario the doping process is primarily controlled by a variation of the position of the bosonic level. For a given $`n`$ this implies a relative change in concentration between the bosons and fermions which leads to a reduction of $`n_F`$ with decreasing $`\mathrm{\Delta }_B`$. We should stress that doping in HTS is an extremely complicated process which involves charge transfer from the dielectric layers into the conduction plane. Monitoring the doping process by changing the position of the Bosonic energy level represents only a crude mechanism for it (though the most significant one for this model). Other features such as anisotropy effects and electronic correlations of course play a role, but are neglected here.
The temperature and doping dependence of the single-particle electron spectral function $`A_F(\epsilon _k,\omega )`$ is studied for $`\epsilon _k=\epsilon _{k_F}`$, where $`\epsilon _{k_F}`$ is determined from the condition that the distribution function $`n_F(\epsilon _k)=𝑑\omega A_F(\epsilon _k,\omega )`$ is independent on the temperature. The behavior of $`A_F(\epsilon _{k_F},\omega )`$ is easily interpreted in terms of the level spectrum of the BFM in the atomic limit. As demonstrated in Section 3, this spectrum consists of bonding and antibonding two-particle states having energies $`E_4`$ and $`E_5`$, plus a non-bonding single-particle state having energy $`\epsilon _0`$. The dynamical mean field, mimicking the electron itinerancy, broadens these levels into a continuous spectrum with peaks approximately centered around $`\epsilon _{}=E_4\epsilon _0`$, $`\epsilon _+=E_5\epsilon _0`$ and $`\epsilon _0`$ (see Fig.3). We notice that there is a qualitative difference between the underdoped ($`\mathrm{\Delta }_B=1.2`$) and the overdoped ($`\mathrm{\Delta }_B=1`$) situation (here $`\mathrm{\Delta }_B/2`$ gives the position of the energy level of the two-particle bosonic state with respect to the bare electronic band, as measured from its bottom). In the case of an underdoped system we can clearly distinguish spectral features (Fig.3a) with a predominant peak corresponding to the non-bonding state which, as the temperature decreases, has its intensity transferred to the lower-lying two-particle bonding state and thereby leads to the opening of a pseudogap around the chemical potential. This is quite different from what happens in the overdoped situation (Fig.3b) where the spectral weight is roughly equally distributed between the bonding and the antibonding two-particle states and no clear feature for the opening of a pseudogap is visible. This effect is particularly evident when plotting (Fig.4) the spectral function at low temperature, together with the corresponding intensity for Angle-Resolved Photoemission Spectroscopy (ARPES) given by $`A_F(\epsilon _{k_F},\omega )f(\omega )`$ ($`f(\omega )`$ denoting the Fermi function) for different doping levels (different values of $`\mathrm{\Delta }_B`$). The results in Fig.4b are in good agreement with the experimental ARPES indication of a pseudogap, well opened in the underdoped regime and practically not visible in the overdoped one. In Fig.4c we plot the intensity for Angle-Resolved Inverse Photoemission Spectroscopy (ARIPES) given by $`A_F(\epsilon _{k_F},\omega )(1f(\omega ))`$ which clearly indicates the contribution from the antibonding two-particle states above the chemical potential. With the rapid improvement of photemission spectroscopy experiments such a feature might be resolved in some near future.
Let us now turn to the discussion of the ordinary dc $`(\sigma )`$ and Hall $`(\sigma _H)`$ conductivity and show to what extent they are influenced by the anomalous single-particle spectral properties discussed above. In the limit of infinite dimensionality these quantities reduce to
$`\sigma ={\displaystyle \frac{\pi e^2}{2d}}{\displaystyle 𝑑\epsilon \rho (\epsilon )𝑑\omega \left(\frac{f(\omega )}{\omega }\right)A_F(\epsilon ,\omega )^2}`$ (25)
$`\sigma _H={\displaystyle \frac{\pi ^2e^3H}{3d^2}}{\displaystyle 𝑑\epsilon \epsilon \rho (\epsilon )𝑑\omega \left(\frac{f(\omega )}{\omega }\right)A_F(\epsilon ,\omega )^3}.`$ (26)
From the above expressions we evaluate, with the help of the single-particle spectral functions $`A_F(\epsilon ,\omega )`$ determined before, the temperature and doping dependence of the Hall coefficient $`R_H=\sigma _H/\sigma ^2H`$. In Fig.5 we plot $`R_H`$ as a function of temperature for different doping levels. As we decrease $`\mathrm{\Delta }_B`$ for a fixed temperature, $`R_H`$ decreases, showing a sign change in the high-temperature regime below $`T0.17`$. In the underdoped regime ($`\mathrm{\Delta }_B1.2`$) $`R_H`$ increases rapidly as the temperature decreases, while a similar temperature variation in the overdoped limit ($`\mathrm{\Delta }_B1`$) changes $`R_H`$ only moderately. With decreased doping (increasing $`\mathrm{\Delta }_B`$) the temperature at which the sign change occurs moves upwards. The sign change of $`R_H`$ suggests a change-over from negative charge carriers at high temperatures to positive charge carriers at low temperature. While at high temperatures $`R_H`$ is roughly determined by the concentration of electrons $`n_F`$, at low temperatures on the contrary it is roughly controlled by $`1n_F`$ and hence scales with doping.
Although experiments seem to suggest that for underdoped systems $`R_H`$ is always positive, our results indicate that as doping is reduced the sign change occurs for higher and higher temperatures, possibly not reachable in real experiments. At first sight it is tempting to interprete the results of Fig.5 as a change-over from a large Fermi surface at high temperatures to small Fermi surface pockets at low temperatures. However, our study of the single-particle spectral function indicates that the Fermi surface is always large, in agreement with experimental findings.
We finally address ourselves to the temperature dependence of the Hall angle $`cot\theta _H`$, which in the underdoped regime shows a $`T^2`$ behavior for a large variety of HTS, sometimes extending up to 400 K. In Fig.6 we plot our calculated Hall angle as a function of temperature for three different doping levels. We notice that deep inside the underdoped regime we do indeed find a $`T^2`$ behavior while with increased doping, going from $`\mathrm{\Delta }_B=1.26`$ to $`\mathrm{\Delta }_B=1.18`$, the range of this $`T^2`$ behavior gets more and more restricted to low temperatures, in qualitative agreement with the experiments.
## V CONCLUSIONS
The non-Fermi liquid properties of the cuprate HTS seem now to be established. Under these circumstances the evaluation of transport coefficients, such as the ones discussed here, can no longer be based on the standard Boltzmann-type relaxation time approach. The transport coefficients are no longer determined exclusively by scattering processes in the immediate vicinity of the Fermi surface, as is the case for standard Fermi liquids. On the contrary, scattering processes cover a wide regime in frequency and wave vector, in accordance with the experimental evidence for very broadened single-particle spectral functions. In this paper we have examined the question to what extent the anomalous properties of the Hall coefficient $`R_H`$ can be linked to the anomalous single-particle properties and in particular to the pseudogap features of them. Our main result is a strongly temperature dependent $`R_H`$ which, for low temperatures, is positive (hole-like charge carriers) and scales with the number of holes as measured from the half-filled band case. As the temperature is increased, $`R_H`$ rapidly drops and saturates at very small negative values for temperatures above that where the pseudogap opens. Correspondingly, Fermi liquid properties begin to be recovered. Since we do not take into account electronic correlations, this result is what one should expect. Taking account of correlations should alter this result as far as the doping dependence of $`R_H`$ is concerned and should make $`R_H`$ scale with the number of holes (measured from the half-filled band case) even for high temperatures. This, however, requires a treatment of correlations and pair fluctuations on the same footing, which is beyond the present study. Finally, a $`T^2`$ behavior of the Hall angle is obtained, but exclusively for very low doped systems.
## VI ACKNOWLEDGEMENTS
We would like to thank T. Domanski and K. Matho for stimulating discussions.
|
warning/0005/physics0005030.html
|
ar5iv
|
text
|
# The modifications of customary filtrational equation
## Abstract
The usable limits of the customary and relaxational filtrational theories are considered. The questions of applicable the locality and local thermodynamical equilibrium principles to depict the nonstationary flows are discussed. The experimental procedures are proposed to determine the filtrational flows relaxation times.
PACS number: 47.10.+g
The theoretical and experimental investigations of the filtrational processes in porous media makes for a long time and the complexity of such systems does not enables to descript their evolution in a simple manner. The nonlinear effects are an essetial in some situations. The locality and local thermodynamical equilibrium principles applicability remains to be investigated also. Below we will be take into consideration the linear theories only.
The Darcy equation
$$\stackrel{}{W}=\frac{k}{\mu }P$$
(1)
was obtain from experiments under stationary filtration conditions. To descript the nonstationary processes usually used the continuously and the state equations in form
$$m(P)=m_0+\beta _m(PP_0)$$
(2)
$$\rho (P)=\rho _0(1+\beta _f(PP_0))$$
(3)
$$\frac{(m\rho )}{t}+div(\rho \stackrel{}{W})=0$$
(4)
Now can produced the customary filtrational equation now as
$$\frac{P}{t}\mathrm{\Delta }P=0$$
(5)
where æ - piezoconductivity, k - permeability, $`\mu `$ \- viscosity, P - pressure, $`\stackrel{}{W}`$ \- filtration velosity , $`\rho `$ \- fluid density, $`m`$ \- porosity, $`\beta _m`$ and $`\beta _f`$ \- compressibility of porous matrix and fluid respectively.
The fundamental solution of (5) for onedimesional system is
$$P(x,t)=\frac{\mathrm{\Theta }(t)}{\sqrt{4\pi t}}exp(\frac{x^2}{4t})$$
(6)
where $`\mathrm{\Theta }(t)`$ is the Heaviside function.
We can see from (6) that the customary filtrational equation leads to infinity phase and group velosities paradox like equations for classical heat conductivity and diffusion.
It should be mentioned that the questions of locality and local thermodynamical equilibrium principles applicability for the systems under investigation are dicussed seldom. In this aspect let us assume the solution of equation (5) for the case of plane parallel onedimensional filtration with the constant pressure difference by the frontiers ($`P_f`$)
$$P(x,t)=P_f(1x/L\underset{n=0}{\overset{\mathrm{}}{}}(2/(\pi n))Sin(\pi nx/L)exp(\pi ^2n^2t/L^2))$$
(7)
The multiexponential dependences pressure from time make it possible to introduce the characteristic time of the transition to the stationary state as $`\tau ^{}=0.1L^2/`$, where L- the distance between frontiers. Now we can estimate this times. Let L=100 meters, $`=1m^2/sec`$, then $`\tau ^{}10^3sec`$. If L=1 m, $`=1m^2/sec`$, $`\tau ^{}0.1`$, when $`L<10^2`$, $`=1m^2/sec`$, $`\tau ^{}<10^5sec`$, and in the last case we have a situation when the velosity of stationary state establising becomes more than the sound velosity in this media. It is a strange conclution. Where is the time and space usable limits of the filtrational theories?
One of the effective attempt to resolve this situation is the relaxational theory . This theory takes into account that the local equilibriun is established in time with the according the next relaxational equation ($`\tau _w`$ \- time of relaxation)
$$\stackrel{}{W}+\tau _w\frac{\stackrel{}{W}}{t}=\frac{k}{\mu }P$$
(8)
Actually this is the local nonequilibrium procedure. In according (8) we come to the hiperbolic equation
$$\frac{P}{t}+\tau _w\frac{^2}{t^2}P\mathrm{\Delta }P=0$$
(9)
with the finite phase and group velocities $`(V_{ph}=\sqrt{(}/\tau ))`$.
In some cases author written relaxation equation in the double relaxational form
$$\stackrel{}{W}+\tau _w\frac{\stackrel{}{W}}{t}=\frac{k}{\mu }(P+\tau _P\frac{P}{t})$$
(10)
and in that event we returns to parabolic form the filtrational equation with the infinite group and phase velosities
$$\frac{P}{t}+\tau _w\frac{^2}{t^2}P\mathrm{\Delta }(P+\tau _p\frac{P}{t})=0$$
(11)
$$V_{ph}=Re(\sqrt{\omega }\sqrt{\frac{1+i\omega \tau _p}{i+\omega \tau _w}})$$
(12)
To test the validity the relaxational filtration theory we may to carry out the experiments with so calles filtrational waves, when the harmonic oscillations of pressure is created in porous media. In case $`\omega \tau <<1`$ the relatation theory leads to the declination $`\omega \tau /2`$ in phase velosities relative to customary equations (5). But in high frequences the relaxation theory have to tends to Biot theory for waves in saturated porous media. So, it is nesessary to explore the investigations in this area. For instance we can investigate this process by means of molecular dynamics simulation and produce the filtrational law averiging the Navier and Stokes equation .
1. Sobolev S. L. Phys.Rev. E, 55 4, 1997.
2. Molokovich Yu.M. Izvestia vuzov. Mathematics, 1977, 8, pp.49-55.
3. Biot M.A. J.Acoust.Soc.Amer., vol.88, 1956, pp.168-186.
4. Ovchinnikov M.N. (in press).
5. Sanchez-Palencia E. Int.Jour.Eng.Sci., 1974, 12, pp.331-351.
|
warning/0005/hep-th0005148.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
In recent years there has been much interest in the study of D-branes on group manifolds (see for instance ). String theory on group manifolds is governed by a WZW model, which has two distinct descriptions: the current algebra and the sigma model realization. Since the WZW model is a typical example of an exact string background, whose CFT is known explicitely, one approach to find possible D-brane configurations is to impose gluing conditions on the chiral currents $`J^a(z)`$ and $`\overline{J}^a(\overline{z})`$ in terms of which the CFT is defined. Actually the boundary state approach has been applied widely to find D-brane configurations on group manifolds . In it was found that the D-branes in WZW models associated with the gluing condition $`J^a=\overline{J}^a`$ along the boundary are configurations of quantized conjugacy classes. As the gluing conditions in the boundary state approach are defined on the chiral currents rather than on the spacetime fields there is a lack of an obvious geometric interpretation of WZW boundary states and in particular of the corresponding D-brane configurations. Since the WZW model provides also an example of a string background with a sigma model description, which allows a complementary study of the D-brane configurations, it is interesting to compare the D-brane configurations obtained from the sigma model realization with those from the boundary state approach (CFT) in order to see how they match with each other.
The other motivation for this work was to see the quantization of the worldvolume $`U(1)`$ flux on the spherical D2-brane. In it was suggested that the $`U(1)`$ worldvolume flux $`F`$ rather than that of $`[(2\pi \alpha ^{})^1B+F]`$ should be quantized. In , the quantization problem was mainly discussed from the Born-Infeld theory, so it is quite interesting to see whether we can study it from the worldsheet perspective. Since the $`U(1)`$ gauge field appears in the action of the sigma model, we wonder what the D-brane configurations constructed from the sigma model approach has to say about this problem.
Motivated by the above, in this paper we study WZW D-branes on the group manifold of $`SU(2)`$ from the sigma model point of view and compare the results with the boundary state approach. Our strategy is that we turn the boundary conditions of the spacetime fields into gluing condition of the chiral current at the boundary and try to adjust the $`U(1)`$ gauge field to make the gluing matrices field independent in order to check chiral Kac-Moody symmetry. For the spherical D2-branes we find that in order to keep the infinite-dimensional symmetry of the current algebra, the $`U(1)`$ worldvolume gauge field strength has to take the form $`F=\frac{\kappa }{2\pi }\psi _0ϵ_2`$, where $`\kappa `$ is the integer level of the associated current algebra and $`\psi _0`$ describes the radius of the spherical D2-branes. Imposing the quantization condition on the flux $`F`$ that follows from a consistent definition of the sigma model action we will recover the quantization of the brane positions $`\psi _0^{(n)}=n\pi /\kappa `$ . For other D-branes we find it impossible to adjust the $`U(1)`$ gauge field to make the gluing matrices $`R_b^a`$ field independent (the gluing matrix is defined by the gluing condition $`J^a(z)+R_b^a\overline{J}^a(\overline{z})=0`$ at the boundary). The dependence of the gluing matrices $`R_b^a`$ on the spacetime fields certainly breaks the chiral Kac-Moody symmetry, but we find that conformal invariance is maintained for these D-branes, at least in our classical approximation. The straightforward extension our result to twisted conjugacy classes is briefly discussed in the last section.
## 2 Parametrization of the sigma model action and chiral currents
We start with the $`SU(2)`$ WZW action with gauge bundles $`A`$ defined on the brane subspaces of the group manifold :
$$S=\frac{\kappa }{8\pi }_\mathrm{\Sigma }tr(g^1_+gg^1_{}g)+\frac{1}{2\pi \alpha ^{}}_\mathrm{\Sigma }g^{}B+_\mathrm{\Sigma }g^{}A,$$
(1)
where we use the string normalization of target space fields. For world sheets with boundary the WZ part of the action is defined in terms of the field strengths $`H=dB`$ and $`F=dA`$ by employing a closed auxiliary surface $`\mathrm{\Sigma }^{}=M`$ that is the union of $`\mathrm{\Sigma }`$ with $`n`$ disks $`D^{(i)}`$ if the boundary $`\mathrm{\Sigma }`$ has $`n`$ connected components, and by extending $`g`$ to the 3-dimensional manifold $`M`$:
$$S_{WZ}=\frac{\kappa }{12\pi }_Mtr(g^1dg)^3\underset{i=1}{\overset{n}{}}\frac{1}{2\pi \alpha ^{}}_{D^{(i)}}g^{}(B+2\pi \alpha ^{}F).$$
(2)
The $`U(1)`$ field strength $`F=dA`$ is defined on the D-brane submanifold (i.e. at the allowed positions of the boundaries of the embedded world sheet; in case of several branes we have to introduce an independent gauge connection $`A`$ for every brane and the respective field strength has to be used for each disk). Moreover, we need to choose a gauge where $`B`$ is not singular on the brane.
Independence of all quantum amplitudes of the various choices involved in this definition implies that the level $`\kappa `$ and all integrals $`_{S^2}F/2\pi `$ of the respective field strengths $`F`$ over any 2-spheres $`S^2`$ embedded in the branes have to be integers : If we close some component of the boundary of $`\mathrm{\Sigma }`$ with two different disks $`D`$ and $`D^{}`$, then the difference in the action is $`\frac{1}{2\pi \alpha ^{}}((_M^{}_M)H(_D^{}_D)(B+2\pi \alpha ^{}F))`$. Since $`H=dB`$ globally on the respective brane, $`B`$ drops out and we are left with an integral of $`F`$ over the 2-sphere $`DD^{}`$.
We can also think about the action (1) in the following way: The gauge transformation $`BB+d\mathrm{\Lambda }`$ leads to surface terms that can only be compensated if we introduce a gauge field $`A`$ that transforms as $`AA2\pi \alpha ^{}\mathrm{\Lambda }`$ at the boundary. The gauge symmetry $`AA+d\lambda `$ just corresponds to the trivial part of the reducible $`\mathrm{\Lambda }`$-transformation that leaves $`B`$ invariant. Hence only the field strength $`H`$ is physical outside the branes, whereas at the allowed positions of the boundary of the world sheet $`F+B/2\pi \alpha ^{}`$ also becomes observable.
To proceed we choose the parametrization
$$g=\left(\begin{array}{cc}\mathrm{cos}\psi i\mathrm{sin}\psi \mathrm{sin}\theta \mathrm{sin}\varphi & \mathrm{sin}\psi \mathrm{sin}\theta \mathrm{cos}\varphi i\mathrm{sin}\psi \mathrm{cos}\theta \\ \mathrm{sin}\psi \mathrm{sin}\theta \mathrm{cos}\varphi i\mathrm{sin}\psi \mathrm{cos}\theta & \mathrm{cos}\psi +i\mathrm{sin}\psi \mathrm{sin}\theta \mathrm{sin}\varphi \end{array}\right)$$
(3)
of the group manifold with
$$0\psi \pi ,0\theta \pi ,0\varphi 2\pi .$$
(4)
In these coordinates the metric and the NS three-form field are given by
$`ds^2=\kappa \alpha ^{}[d\psi ^2+\mathrm{sin}^2\psi (d\theta ^2+\mathrm{sin}^2\theta d\varphi ^2)]`$ (5)
$`H={\displaystyle \frac{1}{6}}\kappa \alpha ^{}\text{tr}(g^1dg)^3=2\kappa \alpha ^{}\mathrm{sin}^2\psi \mathrm{sin}\theta d\psi d\theta d\varphi `$ (6)
and the action turns into
$`S`$ $`=`$ $`{\displaystyle \frac{1}{2\pi \alpha ^{}}}{\displaystyle }d\tau d\sigma \{{\displaystyle \frac{1}{2}}\eta ^{\alpha \beta }\kappa \alpha ^{}(_\alpha \psi _\beta \psi +\mathrm{sin}^2\psi _\alpha \theta _\beta \theta +\mathrm{sin}^2\psi \mathrm{sin}^2\theta _\alpha \varphi _\beta \varphi )`$ (7)
$`+B_{\theta \varphi }(_\tau \theta _\sigma \varphi _\sigma \theta _\tau \varphi )+B_{\psi \theta }(_\tau \psi _\sigma \theta _\sigma \psi _\tau \theta )`$
$`+B_{\psi \varphi }(_\tau \psi _\sigma \varphi _\sigma \psi _\tau \varphi )\}+{\displaystyle }_\mathrm{\Sigma }g^{}A`$
where $`\eta ^{\alpha \beta }=\text{diag}(1,1)`$.
The WZW model has conserved chiral currents (with $`_\pm =_\tau \pm _\sigma `$),
$$J=_+gg^1,\overline{J}=g^1_{}g$$
(8)
Inserting the parametrization (3) into (8) we have
$$J^a=\overline{e}_\mu ^a_+X^\mu ,\overline{J}^a=e_\mu ^a_{}X^\mu $$
(9)
$$\overline{e}_\mu ^a=\left(\begin{array}{ccccc}\mathrm{cos}\theta & & \mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta & & \mathrm{sin}^2\psi \mathrm{sin}^2\theta \\ & & & & \\ \mathrm{sin}\theta \mathrm{cos}\varphi & & \mathrm{sin}\psi \mathrm{cos}\psi \mathrm{cos}\theta \mathrm{cos}\varphi & & \mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta \mathrm{sin}\varphi \\ & & \mathrm{sin}^2\psi \mathrm{sin}\varphi & & \mathrm{sin}^2\psi \mathrm{sin}\theta \mathrm{cos}\theta \mathrm{cos}\varphi \\ & & & & \\ \mathrm{sin}\theta \mathrm{sin}\varphi & & \mathrm{sin}\psi \mathrm{cos}\psi \mathrm{cos}\theta \mathrm{sin}\varphi & & \mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta \mathrm{cos}\varphi \\ & & +\mathrm{sin}^2\psi \mathrm{cos}\varphi & & \mathrm{sin}^2\psi \mathrm{sin}\theta \mathrm{cos}\theta \mathrm{sin}\varphi \end{array}\right)$$
(10)
$$e_\mu ^a=\left(\begin{array}{ccccc}\mathrm{cos}\theta & & \mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta & & \mathrm{sin}^2\psi \mathrm{sin}^2\theta \\ & & & & \\ \mathrm{sin}\theta \mathrm{cos}\varphi & & \mathrm{sin}\psi \mathrm{cos}\psi \mathrm{cos}\theta \mathrm{cos}\varphi & & \mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta \mathrm{sin}\varphi \\ & & \mathrm{sin}^2\psi \mathrm{sin}\varphi & & \mathrm{sin}^2\psi \mathrm{sin}\theta \mathrm{cos}\theta \mathrm{cos}\varphi \\ & & & & \\ \mathrm{sin}\theta \mathrm{sin}\varphi & & \mathrm{sin}\psi \mathrm{cos}\psi \mathrm{cos}\theta \mathrm{sin}\varphi & & \mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta \mathrm{cos}\varphi \\ & & +\mathrm{sin}^2\psi \mathrm{cos}\varphi & & \mathrm{sin}^2\psi \mathrm{sin}\theta \mathrm{cos}\theta \mathrm{sin}\varphi \end{array}\right)$$
(11)
where $`X^1=\psi ,X^2=\theta ,X^3=\varphi `$, and $`J^a,\overline{J}^a`$ are defined by
$$J^a=\frac{i}{2}\text{tr}(\sigma ^a_+gg^1),\overline{J}^a=\frac{i}{2}\text{tr}(\sigma ^ag^1_{}g)$$
(12)
and $`\sigma ^a`$ are the Pauli matrices. The vielbein matrices $`e`$ and $`\overline{e}`$ satisfy $`e^Te=\overline{e}^T\overline{e}=G`$, where $`G`$ is the metric
$$G_{\mu \nu }=\text{diag}(1,\mathrm{sin}^2\psi ,\mathrm{sin}^2\psi \mathrm{sin}^2\theta ).$$
(13)
## 3 D-brane configurations constructed in the sigma model approach
When we vary the action (7) we get the equations of motion. In addition we work out the boundary conditions from which we can construct possible D-brane configurations. We first consider the solution
$$(B+2\pi \alpha ^{}F)_{\theta \varphi }=\kappa \alpha ^{}(\psi \frac{\mathrm{sin}2\psi }{2}+f)\mathrm{sin}\theta $$
(14)
to $`H=dB+2\pi \alpha ^{}F`$ with constant $`f`$ and with all other components vanishing, which is suggested by the symmetry of our choice of coordinates.<sup>1</sup><sup>1</sup>1 Two other simple choices are $`(B+2\pi \alpha ^{}F)_{\psi \theta }=2\kappa \alpha ^{}(\varphi \mathrm{sin}^2\psi \mathrm{sin}\theta +f^{})`$ and $`(B+2\pi \alpha ^{}F)_{\psi \varphi }=2\kappa \alpha ^{}(\mathrm{sin}^2\psi \mathrm{cos}\theta +f^{\prime \prime })`$ with all other components vanishing in both cases. For $`f0`$ they are, however, singular at $`\theta =0`$ mod $`\pi `$ for all $`\psi `$ and therefore not useful for $`D2`$ branes. The 2-form $`\mathrm{sin}\theta d\theta d\varphi `$ is singular at $`\psi =0`$ and at $`\psi =\pi `$, which suggests to associate the term proportional to $`f`$ with the gauge field strength $`F`$. The remaining $`B`$ field is then regular everywhere except at the point $`\psi =\pi `$.
At the present stage of our discussion $`f`$ is an undetermined parameter. We can formaly extend the domain of $`F`$ and effectively obtain a family of choices for the $`B`$ field. Obviously $`ff\pi `$ then shifts the $`B`$ field into another gauge with a singularity in a single point, but this time at $`\psi =0`$, the unit element of the group (at the same time the flux $`F/2\pi `$ through a sphere at fixed $`\psi =\psi _0`$ is shifted by the integer $`\kappa `$). It turns out that $`f`$ can not be fixed by the leading order condition of conformal invariance at the boundary given in
$$_\mu [\sqrt{G}G^{\mu \nu }G^{\rho \sigma }(B+2\pi \alpha ^{}F)_{\nu \rho }]=0$$
(15)
where the metric is given by (13).
With the choice (14) we can read off the boundary condition by varying the action (7) and we find
$`(\delta \psi _\sigma \psi )|_\mathrm{\Sigma }`$ $`=`$ $`0`$
$`\delta \theta \left(\mathrm{sin}^2\psi _\sigma \theta (\psi {\displaystyle \frac{\mathrm{sin}2\psi }{2}}+f)\mathrm{sin}\theta _\tau \varphi \right)|_\mathrm{\Sigma }`$ $`=`$ $`0`$
$`\delta \varphi \left(\mathrm{sin}^2\psi \mathrm{sin}^2\theta _\sigma \varphi +(\psi {\displaystyle \frac{\mathrm{sin}2\psi }{2}}+f)\mathrm{sin}\theta _\tau \theta \right)|_\mathrm{\Sigma }`$ $`=`$ $`0`$ (16)
By exploiting (3) we can look for D-brane configurations of various dimensions by considering the following simple boundary conditions.
D0-brane:
$$\psi |_\mathrm{\Sigma }=\psi _0,\theta |_\mathrm{\Sigma }=\theta _0,\varphi |_\mathrm{\Sigma }=\varphi _0$$
(17)
D1-branes:
$`\psi |_\mathrm{\Sigma }=\psi _0,\theta |_\mathrm{\Sigma }=\theta _0,_\sigma \varphi |_\mathrm{\Sigma }=0`$ (18)
$`\psi |_\mathrm{\Sigma }=\psi _0,_\sigma \theta |_\mathrm{\Sigma }=0,\varphi |_\mathrm{\Sigma }=\varphi _0`$ (19)
$`_\sigma \psi |_\mathrm{\Sigma }=0,\theta |_\mathrm{\Sigma }=\theta _0,\varphi |_\mathrm{\Sigma }=\varphi _0`$ (20)
Spherical D2-brane:<sup>2</sup><sup>2</sup>2When $`\psi _0=0`$ and $`\pi `$, the spherical D2-branes reduce to D0-branes, and the D0-branes described by (17) can be derived from the D0-branes of $`\psi _0=0`$ and $`\pi `$ by an inner automorphism.
$`\psi |_\mathrm{\Sigma }`$ $`=`$ $`\psi _0`$
$`\left(\mathrm{sin}^2\psi _\sigma \theta (\psi {\displaystyle \frac{\mathrm{sin}2\psi }{2}}+f)\mathrm{sin}\theta _\tau \varphi \right)|_\mathrm{\Sigma }`$ $`=`$ $`0`$
$`\left(\mathrm{sin}^2\psi \mathrm{sin}^2\theta _\sigma \varphi +(\psi {\displaystyle \frac{\mathrm{sin}2\psi }{2}}+f)\mathrm{sin}\theta _\tau \theta \right)|_\mathrm{\Sigma }`$ $`=`$ $`0`$ (21)
where $`\psi _0,\theta _0,\varphi _0`$ are arbitrary constants. The last two D1-brane candidates still have to be closed by the antipodal halfs of the respective circles but we have to drop them from our considerations anyway because our ansatz for $`B+F`$ is singular at their location (this is not a big loss, however, because these configurations can be obtaind from global rotations of the group manifold, as we will discuss below). Replacing the Dirichlet boundary condition in (3) by $`_\sigma \psi |_\mathrm{\Sigma }=0`$ we formally get a D3-brane, but this is inconsistent because we cannot have a B-field without singularity on the group manifold.
## 4 Comparison of the D-brane configurations between two approaches and quantized $`U(1)`$ worldvolume flux on $`S^2`$
Now we compare the D-brane configurations derived from the above sigma model with those from the boundary state approach. To do so, we construct the gluing condition $`J^a(z)+R_b^a\overline{J}^a(\overline{z})|_\mathrm{\Sigma }=0`$ from the boundary condition of the spacetime fields $`\psi ,\theta ,\varphi `$ for various D-brane configurations. We try to adjust the undetermined parameter $`f`$ to see whether we can get spacetime field independent gluing matrices $`R_b^a`$ in order to check the infinite-dimensional symmetry of the current algebra.
For the following comparison, we need the explicit expressions for $`J^a`$ and $`\overline{J}^a`$. Using (9)-(11) we rewrite them as
$`J^1`$ $`=`$ $`\mathrm{cos}\theta _\tau \psi +\mathrm{cos}\theta _\sigma \psi \mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta _\sigma \theta +\mathrm{sin}^2\psi \mathrm{sin}^2\theta _\tau \varphi `$
$`+\mathrm{sin}\theta (\mathrm{sin}^2\psi \mathrm{sin}\theta _\sigma \varphi \mathrm{sin}\psi \mathrm{cos}\psi _\tau \theta )`$
$`J^2`$ $`=`$ $`\mathrm{sin}\theta \mathrm{cos}\varphi _\tau \psi +\mathrm{sin}\theta \mathrm{cos}\varphi _\sigma \psi \mathrm{sin}^2\psi \mathrm{sin}\varphi _\tau \theta +\mathrm{sin}\psi \mathrm{cos}\psi \mathrm{cos}\theta \mathrm{cos}\varphi _\sigma \theta `$
$`\mathrm{sin}^2\psi \mathrm{sin}\theta \mathrm{cos}\theta \mathrm{cos}\varphi _\tau \varphi \mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta \mathrm{sin}\varphi _\sigma \varphi `$
$`\mathrm{sin}\varphi (\mathrm{sin}^2\psi _\sigma \theta +\mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta _\tau \varphi )`$
$`\mathrm{cos}\theta \mathrm{cos}\varphi (\mathrm{sin}^2\psi \mathrm{sin}\theta _\sigma \varphi \mathrm{sin}\psi \mathrm{cos}\psi _\tau \theta )`$
$`J^3`$ $`=`$ $`\mathrm{sin}\theta \mathrm{sin}\varphi _\tau \psi +\mathrm{sin}\theta \mathrm{sin}\varphi _\sigma \psi +\mathrm{sin}^2\psi \mathrm{cos}\varphi _\tau \theta +\mathrm{sin}\psi \mathrm{cos}\psi \mathrm{cos}\theta \mathrm{sin}\varphi _\sigma \theta `$ (22)
$`\mathrm{sin}^2\psi \mathrm{sin}\theta \mathrm{cos}\theta \mathrm{sin}\varphi _\tau \varphi +\mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta \mathrm{cos}\varphi _\sigma \varphi `$
$`+\mathrm{cos}\varphi (\mathrm{sin}^2\psi _\sigma \theta +\mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta _\tau \varphi )`$
$`\mathrm{cos}\theta \mathrm{sin}\varphi (\mathrm{sin}^2\psi \mathrm{sin}\theta _\sigma \varphi \mathrm{sin}\psi \mathrm{cos}\psi _\tau \theta )`$
$`\overline{J}^1`$ $`=`$ $`\mathrm{cos}\theta _\tau \psi +\mathrm{cos}\theta _\sigma \psi \mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta _\sigma \theta +\mathrm{sin}^2\psi \mathrm{sin}^2\theta _\tau \varphi `$
$`\mathrm{sin}\theta (\mathrm{sin}^2\psi \mathrm{sin}\theta _\sigma \varphi \mathrm{sin}\psi \mathrm{cos}\psi _\tau \theta )`$
$`\overline{J}^2`$ $`=`$ $`\mathrm{sin}\theta \mathrm{cos}\varphi _\tau \psi +\mathrm{sin}\theta \mathrm{cos}\varphi _\sigma \psi \mathrm{sin}^2\psi \mathrm{sin}\varphi _\tau \theta +\mathrm{sin}\psi \mathrm{cos}\psi \mathrm{cos}\theta \mathrm{cos}\varphi _\sigma \theta `$
$`\mathrm{sin}^2\psi \mathrm{sin}\theta \mathrm{cos}\theta \mathrm{cos}\varphi _\tau \varphi \mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta \mathrm{sin}\varphi _\sigma \varphi `$
$`+\mathrm{sin}\varphi (\mathrm{sin}^2\psi _\sigma \theta +\mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta _\tau \varphi )`$
$`+\mathrm{cos}\theta \mathrm{cos}\varphi (\mathrm{sin}^2\psi \mathrm{sin}\theta _\sigma \varphi \mathrm{sin}\psi \mathrm{cos}\psi _\tau \theta )`$
$`\overline{J}^3`$ $`=`$ $`\mathrm{sin}\theta \mathrm{sin}\varphi _\tau \psi +\mathrm{sin}\theta \mathrm{sin}\varphi _\sigma \psi +\mathrm{sin}^2\psi \mathrm{cos}\varphi _\tau \theta +\mathrm{sin}\psi \mathrm{cos}\psi \mathrm{cos}\theta \mathrm{sin}\varphi _\sigma \theta `$ (23)
$`\mathrm{sin}^2\psi \mathrm{sin}\theta \mathrm{cos}\theta \mathrm{sin}\varphi _\tau \varphi +\mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta \mathrm{cos}\varphi _\sigma \varphi `$
$`\mathrm{cos}\varphi (\mathrm{sin}^2\psi _\sigma \theta +\mathrm{sin}\psi \mathrm{cos}\psi \mathrm{sin}\theta _\tau \varphi )`$
$`+\mathrm{cos}\theta \mathrm{sin}\varphi (\mathrm{sin}^2\psi \mathrm{sin}\theta _\sigma \varphi \mathrm{sin}\psi \mathrm{cos}\psi _\tau \theta )`$
Now let us first consider the spherical D2-brane characterized by (3). In the boundary state approach the spherical D2-brane is described by the gluing condition
$$J^a=\overline{J}^a$$
(24)
at the boundary $`\mathrm{\Sigma }`$. Here we have turned the gluing condition for the spherical D2-brane in the boundary state approach, which uses the closed string picture, into the open string picture<sup>3</sup><sup>3</sup>3There should be a minus sign difference between open and closed string picture .. Comparing $`J^a`$ and $`\overline{J}^a`$ we find that consistency of the boundary conditions (3) with the gluing codition (24) in the boundary state approach, requires<sup>4</sup><sup>4</sup>4For example, let us consider $`J^1=\overline{J}^1`$, the first line of $`J^1`$ is equal to that of $`\overline{J}^1`$ with the help of the first equation in (3), but the second line differs a minus sign. To get $`J^1=\overline{J}^1`$ at the boundary $`\mathrm{\Sigma }`$, we must demand $`(\mathrm{sin}^2\psi \mathrm{sin}\theta _\sigma \varphi \mathrm{sin}\psi \mathrm{cos}\psi _\tau \theta )|_\mathrm{\Sigma }=0`$. When we exploit the third equation in (3), we obtain (25).
$$\psi _0\frac{\mathrm{sin}2\psi _0}{2}+f=\mathrm{sin}\psi _0\mathrm{cos}\psi _0,$$
(25)
which results in
$$f=\psi _0.$$
(26)
In it was shown that the D-brane configurations in the WZW model associated with the gluing condition $`J^a=\overline{J}^a`$ (in the closed string picture) are the conjugacy classes, and in the case of $`SU(2)`$ group the D-brane configurations are spherical D2-branes, which are described by the boundary conditions (3) in the sigma model approach. Imposing the quantization of the $`U(1)`$ worldvolume flux $`F/2\pi =f\kappa /\pi `$ that follows from the definition of the action we thus recover the quantization $`\psi _0^{(n)}=\frac{n\pi }{\kappa }`$ of the brane positions.
For D0- and D1-brane configurations the gluing condition can be written as
$$J^a+R_b^a\overline{J}^b=0$$
(27)
with
$$R=\overline{e}ye^1,$$
(28)
where the vielbein matrices $`e`$ and $`\overline{e}`$ are defined in (10, 11) and the matrix $`y`$ is defined by
$$_+X^\mu =y_\nu ^\mu _{}X^\nu .$$
(29)
For the D0-branes $`y=\text{diag}(1,1,1)`$, so that $`R`$ corresponds to the inner automorphism that translates the brane to the unit at $`\psi =0`$ . For D1-branes at constant $`\psi `$ and $`\theta `$, on the other hand, we find $`y=\text{diag}(1,1,1)`$.
Since there is no place to put the magnetic field strength that could balance the tension on a D1-brane worldvolume, the D1-brane configurations are believed to be unstable. Except for the case of spherical D2-branes, and trivially for the D0 branes, the gluing matrices $`R_b^a`$ depend on the target space position, which indicates that the chiral Kac-Moody symmetry is broken. For the $`SU(2)`$ group manifold, the energy-momentum tensor is $`T(z)=\frac{1}{\kappa +2}J^aJ^a`$. Since $`R^TR=1`$, we have $`T(z)=\overline{T}(\overline{z})`$ at the boundary, so that conformal invariance is preserved even though the chiral Kac-Moody symmetry is broken .
## 5 Summary and discussion
We have investgated possible D-brane configurations from the sigma model point of view. In order to see what the counterparts of these D-branes are in the boundary state approach, we turned the boundary conditions of the spacetime fields into gluing conditions of the chiral currents at the boundary. We have shown that except for spherical D2-brane configurations the gluing matrices for all other D-brane configurations depend on the spacetime fields. For the spherical D2-branes we have seen that the configurations derived from the sigma model do not match those from the boundary state approach automatically. If we demand that they coincide with each other, the $`U(1)`$ worldvolume flux $`F`$ has to be quantized as has been conjectured in , and as it indeed follows from the ambiguity in the definition of the action.
Since the group manifold is $`O(4)`$ symmetric, which manifests itself in the global symmetry of the action $`g\lambda g\rho `$ under left- and right-multiplication with constant group elements, it is clear that there should also be (stable) spherical D2-branes that are not centered around the unit element. Our coordinates are, of course, not very convenient for the discussion of these objects, but it is obvious that our results carry over to that situation and that they are related to the (inner) automorphisms of the current algebra that were discussed in . Indeed, since $`(rgr^1)hg^1=r\lambda (hr^1)\lambda ^1=\rho (hr^1)\rho ^1r`$ with $`\lambda =g\rho ^1`$ and $`\rho =rgr^1`$ the twisted conjugay class defined by $`h`$ and the inner antomorphism corresponding to $`r`$ is just the sphere through $`hr^1`$ centered around $`r`$. In the exact CFT treatment it turns out that, at small levels , the brane positions are somewhat smeared out. It would be interesting to find out what the fate of the apparently unstable D1 branes is after quantum corrections are taken into account.
Eq.(3) shows that the D2-brane sphere should be a fuzzy sphere. Indeed, there have been some discussions of noncommutative geometry on the spherical D2-branes with B-fields . Especially in the low-energy effective action on the fuzzy $`S^2`$ was proposed and it would be interesting to see whether there exists a similar Seiberg-Witten map on the fuzzy sphere, and if so, how the nonlinear $`\mathrm{\Lambda }`$-symmetry in noncommutative geometry is realized as in . As we know, among all examples in AdS/CFT correspondence, the boundary theory of $`AdS_2\times S^2`$ is most poorly understood, see for references. In it was argued that besides the fuzzy $`S^2`$ there is also a fuzzy $`AdS_2`$. It would be interesting to see whether there is a way to study the fuzzy $`AdS_2`$ in the context of WZW models.
Acknowledgement
We would like to thank A.Y. Alekseev, N. Ishibashi, C. Schweigert, V. Schomerus and S. Stanciu for helpful discussions. This work is supported in part by the Austrian Research Funds FWF under grants Nr. P13125-TPH and Nr. M535-TPH.
|
warning/0005/astro-ph0005447.html
|
ar5iv
|
text
|
# Multi-wavelength Observations of Galaxies in the Southern Zone of Avoidance
## 1. Introduction
Understanding the origin of the peculiar velocity of the Local Group and the dipole in the Cosmic Microwave Background is one of the major goals of the study of large-scale structures. Reconstructions of large-scale structures, for instance, still suffer from large interpolation uncertainties across the Zone of Avoidance (ZOA), which extends over about 25% of the optically visible extragalactic sky. Dynamically important structures might still lie hidden in this zone, such as the recently discovered nearby galaxy Dwingeloo 1 (Kraan-Korteweg et al. 1994) and the rich massive cluster Abell 3627 (Kraan-Korteweg et al. 1996). Important large-scale structures, e.g., the Supergalactic Plane and other filaments and wall-like structures, seem to continue across this zone. A more complete knowledge of the distribution of galaxies in redshift space, as well as in distance space, will improve the reconstructed galaxy density fields and help to explain the origin of the peculiar velocity of the Local Group and the dipole in the Cosmic Microwave Background.
Various approaches are presently being employed to uncover the galaxy distribution in the ZOA (cf. Kraan-Korteweg & Lahav 2000): deep optical searches, NIR surveys (DENIS and 2MASS), far-infrared surveys (e.g., IRAS), and blind H I searches. All methods produce new results, but all suffer from limitations and selection effects. The combination of data from an optical galaxy search, a NIR survey and a systematic blind H I survey will allow us to examine the large-scale structures behind the southern Milky Way and the peculiar velocity field associated with them. Redshift independent distance estimates can be obtained via the NIR Tully – Fisher relation. Bouché & Schneider (these proceedings) have shown that the $`K_s`$-band is ideal for this approach.
Here we use (i) data from the diameter-limited, deep $`B`$-band galaxy survey by Kraan-Korteweg and collaborators for cross-identifications at intermediate extinctions, (ii) spiral galaxies detected with the systematic blind H I survey of the southern ZOA ($`|b|<5^{}`$) with the Parkes Multibeam (MB) receiver, plus pointed H I observations of partially obscured galaxies at intermediate latitudes $`5{}_{}{}^{}<|b|<10^{}`$ (Kraan-Korteweg et al. 2000), and (iii) the DENIS survey:
(i) The deep optical survey in the southern ZOA is being conducted by one of us (cf. Kraan-Korteweg & Woudt 1994, Kraan-Korteweg 2000, Woudt et al., these proceedings). In this region ($`265{}_{}{}^{}<\mathrm{}<340^{}`$, $`|b|<10^{}`$), over 11 000 previously unknown galaxies above a diameter limit of $`D=0.^{}2`$ have been identified, next to the previously known $`300`$ Lauberts galaxies with $`D1^{}`$ (Lauberts 1982). As shown by Kraan-Korteweg (2000), this diameter limited survey is complete to $`A_B3^\mathrm{m}`$, although galaxies can be identified to $`A_B<5^\mathrm{m}`$ (or $`|b|5^{}`$ on average).
(ii) The Multibeam (MB) ZOA survey is a systematic H I survey of the southern ZOA within $`|b|<5^{}`$ (see Henning et al., Staveley-Smith et al., these proceedings). It will trace gas-rich spirals out to redshifts of 12 000 km s<sup>-1</sup> with no hindrance from the Galactic dust. The survey is being conducted with the Multibeam receiver (13 beams) at the 64 m Parkes telescope and should detect of the order of 1500 galaxies above the $`5\sigma `$ detection limit of 10 mJy. Only few of these predicted galaxies will have an optical counterpart, but many might be visible in the NIR. Their identifications become feasible with the NIR surveys such as DENIS (Epchtein 1997, Epchtein et al. 1997) and 2MASS (Skrutskie et al. 1997).
(iii) The DENIS survey has currently imaged about 80% of the southern sky in the $`I(0.8\mu )`$, $`J(1.25\mu )`$ and $`K_s(2.15\mu )`$ passbands with a resolution of $`1^{\prime \prime }`$ in $`I`$ and $`3^{\prime \prime }`$ in $`J`$ and $`K_s`$. In a pilot study, we have assessed the performance of the DENIS survey at low Galactic latitudes (Schröder et al. 1997, hereafter Paper I; Kraan-Korteweg et al. 1998, hereafter Paper II; Schröder et al. 1999, hereafter Paper III). After giving some details about the DENIS survey, we present improved results on the photometry of galaxies in the cluster Abell 3627 (Sect. 3), and on cross-identifications with galaxies detected in the H I MB survey (Sect. 4).
## 2. The DENIS Survey
Observations in the NIR have several advantages over other ZOA surveys, and they provide important complementary data. Compared to the optical, the NIR is less affected by the foreground extinction (the extinction in $`K_s`$ is about 10% of the extinction in $`B`$). The NIR is sensitive to early-type galaxies, tracers of massive groups and clusters which are neither uncovered in far infrared surveys nor in the 21 cm radiation. The NIR shows little confusion with Galactic objects such as young stellar objects and cool cirrus sources. Moreover, the NIR allows a good estimation of the stellar mass content of galaxies because recent star formation contributes only little to the flux at this wavelength. It is hence ideally suited for the application of the Tully – Fisher relation.
Number counts of galaxies decrease in the ZOA due to the increasing foreground extinction. This effect depends, however, on wavelength. Interpolating from Cardelli et al. (1989), the extinctions in the NIR passbands are $`A_{I_c}=0.^\mathrm{m}45`$, $`A_J=0.^\mathrm{m}21`$, and $`A_{K_s}=0.^\mathrm{m}09`$ for $`A_B=1.^\mathrm{m}0`$. Thus the decrease in number counts as a function of extinction is considerably slower in the NIR than in the optical.
Figure 1 shows the predicted surface number density of galaxies as a function of Galactic foreground extinction for the DENIS completeness limits of $`I_{\mathrm{lim}}=16.^\mathrm{m}5`$, $`J_{\mathrm{lim}}=14.^\mathrm{m}8`$, $`K_{\mathrm{lim}}=12.^\mathrm{m}0`$ (Mamon 1998) and for $`B_{\mathrm{lim}}=19.^\mathrm{m}0`$ (Gardner et al. 1996). The figure suggests that the NIR becomes notably more efficient at $`A_B>2^\mathrm{m}`$, that the $`J`$-band is the most efficient passband to find galaxies at intermediate extinctions, and that $`K_s`$ becomes superior to $`J`$ at $`A_B12^\mathrm{m}`$. While the diameter-limited samples of the optical searches become incomplete at $`A_B3^\mathrm{m}`$, the $`J`$\- and $`K_s`$-bands will easily detect galaxies up to $`A_B=10^\mathrm{m}`$, or even higher extinctions. These are very rough predictions and do not take into account any dependence on morphological type, surface brightness, orientation and crowding, which may lower the number of actually detectable galaxies (Mamon 1994).
## 3. NIR Photometry in the Norma Cluster
We investigated currently available DENIS data at the core of the Great Attractor, i.e., in the low-latitude ($`\mathrm{}=325^{}`$, $`b=7^{}`$), rich cluster Abell 3627 (cf. Woudt et al., these proceedings), where the Galactic extinction is well determined (Woudt et al. 1998). Five high-quality DENIS strips cross the cluster Abell 3627. The inspected 110 images cover about one-fifth of the cluster area within its Abell-radius of $`R_A=1.^{}75`$ (each DENIS image is $`12^{}`$x$`12^{}`$, offset by $`10^{}`$ in declination and right ascension). The extinction over the cluster area varies as $`0.^\mathrm{m}6A_B2.^\mathrm{m}2`$.
We cross-identified the galaxies found in the optical survey with the DENIS $`I`$, $`J`$, and $`K_s`$ images. On the 110 images, 234 galaxies had been identified in the optical. We have recovered 198 (85%) galaxies in the $`I`$ band, 183 (78%) in the $`J`$ band, and 123 (53%) in the $`K_s`$ band (not including galaxies visible on more than one image). At these extinction levels, the optical survey does remain the most efficient in identifying obscured galaxies. In the NIR, the $`I`$\- and $`J`$-band are equally efficient, though the severe star crowding makes identification of faint galaxies difficult in $`I`$.
We have obtained preliminary $`I`$, $`J`$ and $`K_s`$ Kron photometry using the automated galaxy extraction pipeline (Mamon et al. 1997b) on the galaxies visually identified by us. Although many of the galaxies have a considerable number of stars superimposed on their images, comparison of the magnitudes derived from this fairly automated algorithm agree well with the few known, independent measurements.
The NIR magnitudes have been used to study the colour – colour diagram $`IJ`$ versus $`JK`$ (Fig. 2). In the left hand panel, observed colours (from a $`7^{\prime \prime }`$ aperture) are displayed; in the right hand panel the colours are corrected for foreground extinction using Mg<sub>2</sub>-indices values and interpolations according to the Galactic H I distribution. As a comparison, the range in colours of galaxies at high latitudes (Mamon et al. 1998) is indicated by the box. The displacement of the points agrees well with the path of extinction (arrow) based on the mean extinction in these five strips of $`A_B=1.^\mathrm{m}3`$ (Woudt et al. 1998), suggesting that our preliminary photometry is reasonably accurate. Moreover, the shift in colour can be fully explained by the foreground extinction or, more interestingly, the NIR colours of obscured galaxies provide, in principle, an independent way of mapping the extinction in the ZOA (see also Mamon et al. 1997a).
## 4. Cross-identification on DENIS Images of H I-detected Galaxies
Figure 3 displays the distribution of galaxies detected in the shallow MB-ZOA survey (Henning et al. 2000). Contours indicate extinction levels determined from the DIRBE maps (Schlegel et al. 1998). The outer contour corresponds to $`A_B=3.^\mathrm{m}0`$, the completeness limit for galaxies with an extinction-corrected diameter of $`D^o=1.^{}3`$ in deep optical ZOA galaxy catalogues (see Kraan-Korteweg, these proceedings). The inner contour indicates $`A_B=10^\mathrm{m}`$ (the Milky Way becomes opaque in the optical at $`A_B<5^\mathrm{m}`$).
For 100 of the 110 galaxies detected in the shallow H I survey, DENIS images ($`12{}_{}{}^{}\times 12^{}`$) covering the full positional uncertainty region ($`4{}_{}{}^{}\times 4^{}`$) were currently available. 77 galaxies can be detected in the NIR: 69, 67 and 58 in $`I`$ (crosses), $`J`$, and $`K_s`$ respectively, while only 48 have been detected in the $`B`$-band (large circles in Fig. 3). Triangles indicate the 8 galaxies visible in $`J`$ and/or $`K_s`$ but not in the $`I`$-band. They are clearly at higher extinction levels than the galaxies seen in $`I`$ or $`B`$. This is clear also in the histogram in Fig. 4 which shows the wavelength-dependence of detection rate (shaded versus unshaded regions) as a function of foreground extinction. While the detection rate in the $`B`$-band decreases rapidly with increasing extinction, the decrease in $`I`$ and particularly in $`K_s`$ is much slower. The $`I`$-band (together with the $`J`$-band) seems to be the best passband to find galaxies at extinction levels between $`2^\mathrm{m}`$ and $`10^\mathrm{m}`$ (cf. previous section), and the $`K_s`$-band becomes superior at $`A_B>10^\mathrm{m}`$. Keeping in mind (a) the low number statistics, (b) the fact that MB galaxies are gas-rich spiral and irregular galaxies, and (c) that the optical searches are diameter limited rather than magnitude limited, Fig. 4 and Fig. 1 compare very well.
For 23 galaxies no counterpart could be found. These galaxies either lie behind a very thick extinction layer (e.g., one galaxy at $`b0^{}`$ has $`A_B70^\mathrm{m}`$ according to the DIRBE maps), or they are late-type galaxies of very low surface brightness, hence below the magnitude limits of the DENIS survey. The H I survey (unshaded region in Fig. 4), is, however, not affected by the foreground extinction, therefore superior to other passbands in uncovering spiral galaxies at low Galactic latitudes and high foreground extinction levels. The low number rate at high extinctions is partly due to confusion with Galactic continuum sources at lowest latitudes, as well as the Local Void (e.g., Henning et al., these proceedings).
Figure 5 shows the dependence of the observed $`JK`$ colour of these galaxies (from a $`7^{\prime \prime }`$ aperture) on foreground extinction $`A_B`$, including data from the low-latitude cluster Abell 3627 (stars; see previous section). The broader scatter for the shallow survey galaxies (and some of their companions) can be explained by the larger error in the photometry due to the increase in star crowding. However, the systematic offset towards the red with increasing extinction (the reddening path is indicated by the arrow) is clearly evident.
## 5. Conclusion
We have demonstrated the potential of a multi-wavelength approach to penetrate the extinction layers of the Milky Way for studying extragalactic large-scale structures. The detection rate of the three surveys (a deep optical, a systematic blind H I and a NIR survey) depends on galaxy type and foreground extinction. The three surveys together find galaxies at all extinction levels and Galactic latitudes. Furthermore, with the available photometry, the NIR Tully – Fisher relation can be applied to most of these galaxies. The latter allows the mapping of the peculiar velocity field across the whole ZOA.
Optical surveys are superior for identifying galaxies at intermediate latitudes and extinctions ($`|b|>5^{}`$, $`A_B<3^\mathrm{m}`$). The additional NIR luminosities and colours will prove invaluable in analysing the optical survey data as well as the distribution of the galaxies in redshift space, and in the final merging of these data with existing sky surveys.
With the DENIS NIR survey we can trace galaxies down to $`A_B<15^\mathrm{m}`$, i.e. very low latitudes. Though the $`I`$-band is strongly affected by star crowding (which mainly depends on the limiting magnitude of the survey), it is best suited for identifying galaxies up to $`A_B<4^\mathrm{m}`$ because of the higher spatial resolution ($`1^{\prime \prime }`$). At higher foreground extinctions, the $`J`$-band and the $`K_s`$-band (for $`A_B>10^\mathrm{m}`$) become important. NIR surveys thus further reduce the width of the ZOA. In particular, they provide the only tool with which to identify early-type galaxies at high extinction. Despite the star crowding at these latitudes, $`I`$, $`J`$ and $`K_s`$ photometry from the survey data can be successfully performed and the colours can be used to calibrate the DIRBE extinction maps locally.
In addition, we can complement the NIR data-set using the $`H`$-band and the $`K_s`$-band (with a fainter magnitude limit) of 2MASS (see also Huchra et al., these proceedings) to obtain a wider range in colours.
The blind H I survey uncovers spiral galaxies independent of foreground extinction. For a significant fraction of these detections, a DENIS counterpart has been found. These MB-ZOA data cover the Galactic latitude range $`|b|<5^{}`$. We will complement this area with pointed H I observations of optically identified spiral galaxies for intermediate latitudes ($`5{}_{}{}^{}<|b|<10^{}`$). About 300 spiral galaxies have already been detected (Kraan-Korteweg et al. 1997).
## References
Cardelli, J.A., Clayton G.C., & Mathis, J.S. 1989, ApJ, 345, 245
Epchtein, N. 1997, in 2nd Euroconference, The Impact of Large Scale Near-Infrared Surveys, eds. F. Garzón et al., (Dordrecht: Kluwer), 15
Epchtein, N., Batz, B. de, Capoani, L., et al. 1997, Messenger, 87, 27
Gardner, J.P., Sharples, R.M., Carrasco, B.E., & Frenk, C.S. 1996, MNRAS, 282, L1
Henning, P.A., Staveley-Smith, L., Ekers, R.D., Green, A.J., Haynes, R.F., Juraszek, S., Kesteven, M.J., Koribalski, B., Kraan-Korteweg, R.C., Price, R.M., Sadler, E.M., & Schröder, A. 2000, AJ, in press
Kraan-Korteweg, R.C., & Woudt, P.A. 1994, in ASP Conf. Ser. Vol. 67, Unveiling Large-Scale Structures Behind the Milky Way, eds. C. Balkowski & R.C. Kraan-Korteweg, (San Francisco: ASP), 89
Kraan-Korteweg, R.C., Loan, A.J., Burton, W.B., Lahav, O., Ferguson, H.C., Henning, P.A., & Lynden-Bell, D. 1994, Nature, 372, 77
Kraan-Korteweg, R.C., Woudt, P.A., Cayatte, V., Fairall, A.P., Balkowski, C., & Henning, P.A. 1996, Nature, 379, 519
Kraan-Korteweg, R.C., Woudt, P.A., & Henning, P.A. 1997, PASA, 14, 15
Kraan-Korteweg, R.C., Schröder, A., Mamon, G.A., & Ruphy, S. 1998, in 3rd Euroconference, The Impact of Near-Infrared Surveys on Galactic and Extragalactic Astronomy, ed. N. Epchtein (Dordrecht: Kluwer), 209 (Paper II)
Kraan-Korteweg, R.C., & Lahav, O. 2000, A&ARv, in press
Kraan-Korteweg, R.C. 2000, A&AS, 141, 123
Kraan-Korteweg, R.C., Henning, P.A., & Schröder, A. 2000, in prep.
Lauberts, A. 1982, The ESO/Uppsala Survey of the ESO (B) Atlas (Garching: ESO)
Mamon G.A. 1994, in ASP Conf. Ser. 67, Unveiling Large-Scale Structures Behind the Milky Way, eds. C. Balkowski & R.C. Kraan-Korteweg, (San Francisco: ASP), 53
Mamon, G.A., Banchet, V., Tricottet, M., & Katz, D. 1997a, in 2nd Euroconference, The Impact of Large-Scale Near-Infrared Surveys, eds. F. Garzón et al., (Dordrecht: Kluwer), 239
Mamon, G.A., Tricottet, M., Bonin, W., & Banchet, V. 1997b, in XVIIth Moriond Astrophysics Meeting, Extragalactic Astronomy in the Infrared, eds. G. A. Mamon, Trinh Xuân Thuân, & J. Trân Thanh Vân, (Gif-sur-Yvette: Editions Frontières), 369
Mamon, G.A., Borsenberger, J., Tricottet, M., & Banchet, V. 1998, in 3rd Euroconference, The Impact of Near-Infrared Surveys on Galactic and Extragalactic Astronomy, ed. N. Epchtein, (Dordrecht: Kluwer), 177
Mamon, G. A. 1998, in XIVth IAP Astrophysics Meeting, Wide Field Surveys in Cosmology, eds. S. Colombi, Y. Mellier, & B. Raban, (Gif-sur-Yvette: Editions Frontières), 323
Schlegel, D.J., Finkbeiner, D.P., & Davis, M. 1998, ApJ, 500, 525
Schröder, A., Kraan-Korteweg, R.C., Mamon, G.A., & Ruphy, S. 1997, in XVIIth Moriond Astrophysics Meeting, Extragalactic Astronomy in the Infrared, eds. G. A. Mamon, Trinh Xuân Thuân, & J. Trân Thanh Vân, (Gif-sur-Yvette: Editions Frontières), 381 (Paper I)
Schröder, A., Kraan-Korteweg, R.C., & Mamon, G.A. 1999, PASA, 16, 42 (Paper III)
Skrutskie, M.F., Schneider, S.E., Stiening, R., et al. 1997, in 2nd Euroconference, The Impact of Large-Scale Near-Infrared Surveys, eds. F. Garzón et al., (Dordrecht: Kluwer), 25
Woudt, P.A., Kraan-Korteweg, R.C., Fairall, A.P., Böhringer, H., Cayatte, V., & Glass, I.S. 1998, A&A, 338, 8
|
warning/0005/hep-th0005031.html
|
ar5iv
|
text
|
# Untitled Document
hep-th/0005031 EFI-2000-15
D-branes and Strings as Non-commutative Solitons
Jeffrey A. Harvey, Per Kraus, Finn Larsen and Emil J. Martinec
Enrico Fermi Institute and Department of Physics
University of Chicago, Chicago, IL 60637, USA
The non-commutative geometry of a large auxiliary $`B`$-field simplifies the construction of D-branes as solitons in open string field theory. Similarly, fundamental strings are constructed as localized flux tubes in the string field theory. Tensions are determined exactly using general properties of non-BPS branes, and the non-Abelian structure of gauge fields on coincident D-branes is recovered.
May, 2000
1. Introduction
General arguments \[1,,2,,3\], explicit calculations in truncated open string field theory \[4,,5,,6,,7\], and renormalization group analysis of relevant boundary perturbations all suggest that D-branes can be constructed as solitons or lumps in open string field theory. Analogies between tachyon condensation in open string theory and confinement of electric charge have also motivated suggestions that macroscopic closed strings can be described by flux tubes in open string field theory \[4,,10,,11\]. It is known that fundamental strings ending on D-branes can be viewed as flux tubes \[12,,13,,14\], the new element is the idea that such a description should be valid even in the absence of D-branes.
One problem in trying to construct these solutions explicitly is that they are string scale objects, and the string field theory action contains an infinite number of higher derivatives with coefficients set by the string scale. As a result it is difficult to obtain an accurate description of $`Dp`$-branes for small $`p`$, the non-Abelian structure on multiple D-branes is far from obvious, and it has not been possible to obtain a concrete understanding of the flux tube solution.
In this paper we will show that these difficulties can be overcome by the introduction of non-commutative geometry via a background $`B`$-field \[15,,16\]. We consider solitons on the world-volume of an unstable D-brane in the presence of a large $`B`$-field. The solutions we study are the non-commutative solitons recently constructed by Gopakumar, Minwalla and Strominger (GMS) . The solitons, though string scale with respect to the original space-time metric, are much larger than the string scale as measured by the effective open string metric of Seiberg and Witten ; it is this latter fact which facilitates the analysis. Far from the solitons the tachyon on the original unstable D-brane will have condensed to its local minimum, so that the soliton field configuration is asymptotic to the closed string vacuum without D-branes. The $`B`$-field is thus pure gauge far from the solitons, and the solutions represent D-branes and strings in the standard closed string vacuum.
A remarkable feature of the GMS solitons is that many of their properties are insensitive to details of the field theory to which they are solutions — in our case open string field theory on the unstable D-brane. Thus, for the most part we will consider only the low-energy dynamics of the light modes of the open string field. For most of our considerations we do not need to know the detailed shape of the tachyon potential, we require only the height of its local maximum. Happily, this is one of the few properties of unstable D-branes that we understand precisely, from a conjecture by Sen \[1,,18\]. Combining these observations therefore allows us to verify that the solitons have exactly the right tensions to be identified as D-branes. We also obtain the correct content of world-volume fields on the D-branes, and find the expected non-Abelian gauge structure for multiple D-branes.
Classical open string field theory does not contain closed strings, they appear only as poles in loop diagrams. The arguments referred to earlier suggest however that after tachyon condensation the open string degrees of freedom are frozen out and one should find macroscopic closed strings in the classical theory. In the semi-classical limit these should be interpretable as solitons. Indeed, we find that the magic of non-commutativity facilitates a concrete construction where the fundamental string is identified with a flux tube. The fluctuations of the flux tube can be analyzed explicitly and, in a suitable approximation, are described by the Nambu-Goto action.
The paper is organized as follows. In section 2 we first review a few properties of non-commutative field theories and their solitons, we show how to embed this discussion in string theory via unstable D-branes, and then proceed to compute the tension of these solitons in string theory and identify them with D-branes. In section 3 we consider the massless gauge fields in the string field theory and see how these descend to the soliton world-volume. The gauge fields and their interplay with the tachyon field play an important role in obtaining the correct D-brane collective dynamics. In section 4 we extend the discussion to D-branes in type II string theory. Finally, in section 5, we use similar methods to construct a flux tube in the open string field theory. We compute its tension and find its excitations, and are led to conclude that it can be identified with the closed fundamental string expected in the vacuum after tachyon condensation.
Notation: we will denote the complete set of space-time coordinates by $`x^\mu `$, non-commutative directions by $`x^i`$, and the remaining directions by $`x^a`$.
2. Soliton solutions and their tensions
2.1. The Non-commutative Limit
We will be considering Minkowski space-time with closed string metric $`g_{\mu \nu }=\eta _{\mu \nu }`$ in the presence of a constant $`B_{\mu \nu }`$ field, the latter taking non-vanishing values only in purely spatial directions. As explained in , one should distinguish between the closed string metric $`g_{\mu \nu }`$ and the open string metric $`G_{\mu \nu }`$ which are related by
$$G_{\mu \nu }=g_{\mu \nu }(2\pi \alpha ^{})^2(Bg^1B)_{\mu \nu }.$$
The open string field theory action is related to the one without a $`B`$-field by using the metric $`G_{\mu \nu }`$, replacing ordinary products of fields by $``$ products,
$$A(x)B(x)AB=e^{\frac{i}{2}\theta ^{\mu \nu }_\mu _\nu ^{}}A(x)B(x^{})|_{x=x^{}},$$
and replacing the open string coupling $`g_s`$ by
$$G_s=g_s\left(\frac{detG}{det(g+2\pi \alpha ^{}B)}\right)^{1/2}.$$
Here
$$\theta ^{\mu \nu }=(2\pi \alpha ^{})^2\left(\frac{1}{g+2\pi \alpha ^{}B}B\frac{1}{g2\pi \alpha ^{}B}\right)^{\mu \nu }.$$
Now there is a new dimensionless parameter $`\alpha ^{}B_{\mu \nu }`$, or equivalently $`\theta ^{\mu \nu }/\alpha ^{}`$, that can be varied in order to simplify the analysis. Denoting the directions in which $`B_{\mu \nu }`$ is non-vanishing as $`x^i`$, we will be interested in the limit of large non-commutativity, $`\alpha ^{}B_{ij}\mathrm{}`$ with $`g_{ij}`$ held fixed. In this limit the solitons will become much larger than the string scale when measured in the open string metric and this will lead to many simplifications. To avoid confusion, we note that there is an equivalent, but perhaps more familiar, form of the limit: $`\alpha ^{}B_{ij}0`$, $`g_{ij}0`$, and $`G_{ij}`$ is fixed. In this form one has $`\theta ^{ij}/\alpha ^{}\mathrm{}`$. The two versions of the limit are simply related by a coordinate transformation, $`x^i2\pi \alpha ^{}B_{ij}x^j`$. In either form of the limit,
$$\theta ^{ij}=\left(\frac{1}{B}\right)^{ij}.$$
2.2. Solitons in String Field Theory
Let us review some aspects of D-branes as solitons in open string field theory. Our primary focus will be solitons on the world-volume of a bosonic D25-brane, although we will also discuss type II D-branes.
The bosonic D25-brane has on its world-volume a tachyon, $`m_t^2=1/\alpha ^{}`$, with a potential indicated schematically in fig. 1. We keep an explicit factor of the D25-brane tension, $`T_{25}`$, in front of the action, so that the physical tachyon potential is $`T_{25}V(t)`$, and we have shifted the tachyon so that the local minimum is at $`t=0`$. The unstable local maximum $`t=t_{}`$ represents the space filling D25-brane, with $`T_{25}V(t_{})=T_{25}`$. The local minimum $`t=0`$ is the closed string vacuum without D-branes, $`V(0)=0`$.
Fig. 1: The bosonic open string tachyon potential.
The tachyon action supports unstable soliton solutions which are asymptotic to the closed string vacuum at $`t=0`$, and it has been proposed to identify these with bosonic Dp-branes with $`p<25`$. Such solitons have been constructed numerically in level truncated open string field theory in \[4,,5\], and for sufficiently large $`p`$ good agreement was found between the tension and low lying spectra of the soliton and those of bosonic Dp-branes. However, the presence of higher derivatives in the string field theory action greatly complicates the analysis, and it seems very challenging to recover such fundamental properties as enhanced gauge symmetry for coincident D-branes.
In the present work we instead apply some powerful simplifications following from non-commutative geometry. Our starting point is an effective action for the tachyon obtained by integrating out (classically) all fields in the string field theory action which are “sourced” by the tachyon,
$$S=\frac{C}{g_s}d^{26}x\sqrt{g}\left(\frac{1}{2}f(t)g^{\mu \nu }_\mu t_\nu t+\mathrm{}V(t)\right),$$
where $`\mathrm{}`$ indicate higher derivative terms which will not be written explicitly, and we have explicitly displayed the string coupling by defining a $`g_s`$-independent constant
$$C=g_sT_{25}.$$
According to the conjecture of the entire action vanishes at the local minimum: $`f(0)=V(0)=0`$ (with corresponding equations for the higher derivative terms).
Now we turn on the $`B`$-field, which changes the action to
$$S=\frac{C}{G_s}d^{26}x\sqrt{G}\left(\frac{1}{2}f(t)G^{\mu \nu }_\mu t_\nu t+\mathrm{}V(t)\right),$$
where $``$ products are now implied. Due to the non-commutativity one needs to specify an ordering of fields to define (2.1), but this level of precision will not be needed for the present analysis.
Soliton solutions in theories of this kind were constructed in in the limit of large non-commutativity. A simple scaling computation shows that the potential term dominates over the derivative terms in this limit, so that the equation of motion for static solitons is
$$\frac{dV}{dt}=0.$$
Localized soliton solutions to this equation exist due to the presence of the $``$ product. The construction of GMS relies on the existence of functions $`\varphi `$ satisfying<sup>1</sup> Such functions have also made an appearance in the work of .
$$\varphi \varphi =\varphi ,$$
since then
$$F(\lambda \varphi )=F(\lambda )\varphi ,$$
for any function $`F`$ of the form $`F(x)=_{n=1}^{\mathrm{}}a_nx^n`$. In particular,
$$\frac{dV}{dt}|_{t=\lambda \varphi }=\left(\frac{dV}{dt}|_{t=\lambda }\right)\varphi ,$$
and (2.1) is solved by choosing $`\lambda `$ to be an extremum of $`V`$. Turning on $`B_{ij}`$ in two directions, say $`x_{1,2}`$, the simplest function satisfying (2.1) is the Gaussian
$$\varphi _0(r)=2e^{r^2/\theta },r^2=x_1^2+x_2^2,$$
with $`B=B_{12}`$, $`\theta =1/B`$. For the potential indicated in fig. 1 the solution will thus be
$$t=t_{}\varphi _0(r).$$
Note that for this solution the tachyon asymptotically approaches its value $`t=0`$ in the closed string vacuum. The resulting object is a 23+1 dimensional soliton that we will identify with the D23-brane. The coordinate size of the soliton is $`\mathrm{\Delta }x\sqrt{\theta }=1/\sqrt{B}`$, which goes to zero in the large B limit. However, for determining the importance of $`\alpha ^{}`$ corrections the relevant quantity is $`\mathrm{\Delta }x_{\mathrm{open}}=\sqrt{G_{ij}\mathrm{\Delta }x^i\mathrm{\Delta }x^j}\alpha ^{}\sqrt{B}`$. In the limit $`\alpha ^{}B\mathrm{}`$ this is much larger than $`\sqrt{\alpha ^{}}`$, so $`\alpha ^{}`$ corrections, in the form of the derivative terms in (2.1), are suppressed.
The above construction easily generalizes to arbitrary even codimension solitons, for example by turning on equal $`B`$-fields in the $`(12),(34)\mathrm{}(2q1,2q)`$ planes and by replacing $`r^2`$ in (2.1) by $`r^2=x_1^2+x_2^2+\mathrm{}+x_{2q}^2`$. The resulting soliton is to be identified with a D($`252q`$)-brane.
2.3. Tension of solitons
We now show that our solutions have the same tension as bosonic Dp-branes,
$$T_p=(2\pi )^{25p}(\alpha ^{})^{(25p)/2}T_{25}.$$
We first consider the D23-brane soliton. At large non-commutativity we neglect the explicit transverse derivatives in (2.1), and the action for translationally invariant configurations along the D23-brane is
$$S=\frac{C}{G_s}d^{26}x\sqrt{G}V(t).$$
Now we insert the soliton solution, use $`V(t)=V(t_{})\varphi _0(r)`$, and integrate over $`x_1,x_2`$:
$$S=\frac{CV(t_{})}{G_s}d^{24}xd^2x\sqrt{G}\varphi _0(r)=\frac{2\pi \theta CV(t_{})}{G_s}d^{24}x\sqrt{G}.$$
Next we use the relation (2.1) between $`G_s`$ and $`g_s`$, which for large $`B`$-field is
$$G_s=\frac{g_s\sqrt{G}}{2\pi \alpha ^{}B\sqrt{g}}.$$
In our conventions $`V(t_{})=1`$ so, inserting this into (2.1) and using $`\theta =1/B`$, we find
$$S=(2\pi )^2\alpha ^{}\frac{C}{g_s}d^{24}x\sqrt{g}.$$
Finally, recall $`C=T_{25}g_s`$. This identifies the tension of the soliton as
$$T_{\mathrm{sol}}=(2\pi )^2\alpha ^{}\frac{C}{g_s}=(2\pi )^2\alpha ^{}T_{25}=T_{23}.$$
Remarkably, in this limit we get precisely the correct answer without knowing the detailed form of the tachyon potential. We only need its value at the unstable extremum which follows from the conjecture of Sen, as substantiated in the work of \[6,,3,,20,,8\]. It is straightforward to generalize this result to arbitrary even codimension solitons, and to reproduce the formula (2.1) for odd $`p`$.
2.4. Multiple D-branes
There is a one-to-one correspondence between functions on the non-commutative $`R^2`$ transverse to the D23-brane, thought of as the phase space of a particle in one dimension, and operators acting on the Hilbert space of one-dimensional particle quantum mechanics. Multiplication by the $``$ product goes over to operator multiplication, and integration over $`R^2`$ corresponds to tracing over the Hilbert space,
$$AB\widehat{A}\widehat{B},\frac{1}{2\pi \theta }d^2x_i\mathrm{Tr}.$$
Under this correspondence, the equation (2.1) becomes the equation for a projection operator. This correspondence was utilized in to construct more general soliton solutions. The soliton solution $`t=t_{}\varphi _0`$ (2.1) corresponds to the projection operator onto the ground state of a one-dimensional harmonic oscillator, $`\varphi _0|00|`$ . Other solutions are obtained by choosing other projection operators, $`\varphi _n|nn|`$, or we can choose a superposition (a level $`k`$ solution in the terminology of GMS)
$$t_k=t_{}(\varphi _0+\varphi _1+\mathrm{}+\varphi _{k1}).$$
Since the projection operators are orthogonal, the energies just add
$$V(t_k)=kV(t_1).$$
Thus this configuration corresponds to $`k`$ coincident D-branes; further evidence for this claim will appear in succeeding sections.
Since the $`\varphi _m=|mm|`$ are a complete set of projection operators, the limit $`k\mathrm{}`$ of the level $`k`$ solution is $`t_{\mathrm{}}=t_{}1\mathrm{l}`$. This solution can be identified with the D25-brane with no tachyon condensate; indeed, the tachyon takes the value $`t=t_{}`$ everywhere, and the energy density is that of the D25-brane. At large $`k`$, the level $`k`$ solution (2.1) represents, in the basis constructed in , a lump of size $`r_k\sqrt{k}`$, which approximates the string field configuration of an unstable D25-brane for smaller radius, and the closed string vacuum state for larger radius. An amusing case is the projection operator complementary to (2.1), namely $`t=t_{}(1\varphi _0)`$. Evaluating the energy of this configuration, one formally finds the energy of a D25-brane ‘minus’ that of a D23.<sup>2</sup> Of course, the energy of the D25 is infinitely larger than that of the D23; to make this statement precise, one must go to finite volume, e.g. by compactifying the system on a large torus.
In the limit of infinite non-commutativity the action (2.1) can be written in operator form as
$$S=T_{23}d^{24}x_a\mathrm{Tr}\left(\frac{1}{2}f(\widehat{t})G^{ab}_a\widehat{t}_b\widehat{t}V(\widehat{t})\right),$$
where $`x^a`$ are the commutative directions. The action in operator form has a manifest $`U(\mathrm{})`$ symmetry
$$\widehat{t}U\widehat{t}U^{}.$$
These group operations are familiar from the construction of the Matrix theory membrane \[21,,22\]: They correspond to area preserving diffeomorphisms. This is no accident. The Matrix theory membrane in non-compact space is essentially a D2-brane that has been bound to an infinite number of D0-branes such that the D0 charge density is finite (represented by a magnetic flux $`F`$). This charge density is equivalent to the $`B`$-flux of non-commutative geometry, since $`B`$ and $`F`$ are indistinguishable on the brane. Thus the two constructions are identical, and it is useful to keep this relationship in mind. The main difference between the two situations is the presence of the tachyon field.
It is clear that, if we neglect the standard kinetic term, then acting with an area preserving diffeomorphism on a given configuration is a symmetry of (2.1) since it preserves the $``$ product (which is defined in terms of the volume two-form) and the integration measure. On the other hand, the standard kinetic term involves also the metric $`g_{ij}`$. Only the elements of $`U(\mathrm{})`$ corresponding to translations and rotations preserve this metric and so are exact symmetries of the action. Thus, for purely scalar actions, as considered by GMS, the $`U(\mathrm{})`$ is an approximate global symmetry that becomes exact in the limit of infinite non-commutativity.
Solutions of the form (2.1) preserve a $`U(k)`$ subgroup of the $`U(\mathrm{})`$ global symmetry of the action at infinite $`\theta `$ (times an irrelevant group acting in the orthogonal space), and all excitations on the branes will transform in multiplets of $`U(k)`$. The breaking of $`U(\mathrm{})`$ to $`U(k)`$ leads to an infinite number of Nambu-Goldstone bosons in the spectrum of the soliton. Again, for a purely scalar theory, these become pseudo-Nambu-Goldstone bosons at finite $`\theta `$. As we discuss in section 3, the story changes after coupling to the gauge fields on the D-brane.
2.5. Tachyon on the soliton
The D25-brane tachyon reflects the fact that $`t=t_{}`$ is an unstable point of the potential:
$$V(t)=V(t_{})+\frac{1}{2}V^{\prime \prime }(t_{})(tt_{})^2+\mathrm{},V^{\prime \prime }(t_{})<0,$$
and from (2.1) the mass of the open string tachyon is
$$m_t^2=\frac{V^{\prime \prime }(t_{})}{f(t_{})}=\frac{1}{\alpha ^{}}.$$
The lower D-branes also have a tachyon with this mass, and we should recover it by studying fluctuations around the soliton.
As usual, we first consider the simplest case of two non-commutative directions $`x^i`$. Using the operator correspondence, a complete set of functions on this space is given by
$$\varphi _{mn}(x^i)|mn|,$$
so the general fluctuation is
$$t+\delta t(x^\mu )=t_{}\varphi _{00}(x^i)+\underset{m,n=0}{\overset{\mathrm{}}{}}\delta t_{mn}(x^a)\varphi _{mn}(x^i).$$
Reality of $`t`$ requires $`\delta t_{mn}`$ to be a Hermitian matrix. As we have discussed, the fluctuations include Nambu-Goldstone bosons from the spontaneous breaking of $`U(\mathrm{})`$ by the soliton. As in GMS, the generators of $`U(\mathrm{})`$ are $`R_{mn}=|mn|+|nm|`$ and $`S_{mn}=i(|mn||nm|)`$, and the Nambu-Goldstone bosons are the nonzero components of $`\delta t=[R_{mn},t],[S_{mn},t]`$. We will see in the next section that in the D-brane application of interest $`U(\mathrm{})`$ is a gauge symmetry and the Nambu-Goldstone bosons are eaten by the Higgs mechanism. Thus we use the $`U(\mathrm{})`$ symmetry to set
$$\delta t_{0m}=\delta t_{m0}=0,m1.$$
Substituting into the action and using the fact that $`\varphi _0\varphi _{00}`$ is orthogonal to $`\varphi _{mn}`$ for $`m,n>1`$ we are left with the single tachyon fluctuation $`\delta t_{00}`$ which we rename $`\delta t`$. We now use
$$V(t+\delta t\varphi _0)=V(t_{}+\delta t)\varphi _0=\left(V(t_{})+\frac{1}{2}V^{\prime \prime }(t_{})(\delta t)^2+\mathrm{}\right)\varphi _0.$$
After integrating over the soliton the quadratic effective action for $`\delta t`$ becomes
$$S=T_{23}d^{24}x_a\left(\frac{1}{2}f(t_{})^a\delta t_a\delta t\frac{1}{2}V^{\prime \prime }(t_{})(\delta t)^2\right),$$
so the D23-brane correctly inherits the tachyon of the D25-brane,
$$m_t^2=\frac{V^{\prime \prime }(t_{})}{f(t_{})}=\frac{1}{\alpha ^{}}.$$
This analysis can be easily repeated to obtain a tachyon on the more general solution (2.1); then acting with $`U(k)`$ on this tachyon produces the expected $`k^2`$ tachyons. A notable point is that the general tachyon will contain operators of the form $`|mn|`$, $`mn`$. These correspond to non-spherically symmetric tachyon fluctuations in the transverse space. So the full $`k^2`$ of tachyons comes from including both spherically symmetric and non-symmetric tachyon configurations.
3. Coupling to Gauge Fields
A characteristic property of D-branes is the presence of gauge fields on their world-volume. In this section, we demonstrate that there exists a single massless gauge field on the non-commutative soliton, therefore providing further evidence that the soliton can be identified as a D-brane. Furthermore, when we generalize to level $`k`$ solutions such as (2.1), the gauge symmetry is enhanced to $`U(k)`$ in the appropriate way, with the components of the gauge fields transverse to the soliton behaving as the standard adjoint Higgs fields on a D-brane. Finally, the gauge symmetry removes from the spectrum the unwanted (pseudo) Nambu-Goldstone bosons describing soft deformations (2.1) of the soliton (2.1); they are eaten by the Higgs mechanism.
This last point brings out an intriguing aspect of our situation. Ordinarily, when a global symmetry is explicitly broken, no matter how softly, it cannot be gauged. One might then wonder how the $`U(\mathrm{})`$ symmetry (2.1) can be gauged, since it is broken by the tachyon kinetic term at finite $`\theta `$. The resolution of this puzzle is strikingly reminiscent of general relativity. There, the potential energy term in the action of a scalar field is invariant under volume-preserving diffeomorphisms; but the kinetic energy term is not. Rather, the coordinate transformations are gauged by coupling to a metric, without there being a corresponding global symmetry in the absence of gauging. We will see that, once we include the gauge fields, the approximate $`U(\mathrm{})`$ global symmetry is realized as an exact gauge symmetry. In the present context, the non-commutative D-brane gauge field takes over the role of the metric in covariantizing area-preserving diffeomorphisms!
A note on conventions: In this section we often set $`2\pi \alpha ^{}=1`$ to avoid clutter.
3.1. The Gauge Theory and its Symmetries
First consider the action for the D25-brane. Imagine integrating out all fields except for the tachyon and the gauge field. The result will be some gauge invariant expression involving an infinite number of derivatives and an infinite number of higher powers of the fields. Working in terms of the $``$ product implies that the gauge transformation law of the non-commutative $`U(1)`$ gauge field is
$$\delta _\lambda A_\mu =_\mu \lambda i[A_\mu ,\lambda ],$$
where
$$[A_\mu ,\lambda ]=A_\mu \lambda \lambda A_\mu .$$
The corresponding field strength is
$$F_{\mu \nu }=_\mu A_\nu _\nu A_\mu i[A_\mu ,A_\nu ].$$
We also need to know the gauge transformation of the tachyon. String theory disk diagrams reveal that the tachyon transforms in the adjoint of the non-commutative $`U(1)`$ , i.e.
$$\delta _\lambda t=i[t,\lambda ],$$
with corresponding covariant derivative
$$D_\mu t=_\mu ti[A_\mu ,t].$$
We will restrict attention to terms quadratic in the field strength. The full action contains higher derivative terms in the commutative directions, but for clarity’s sake these will not be displayed explicitly. The action is then,
$$S=\frac{C}{G_s}d^{26}x\sqrt{G}\left(\frac{1}{2}f(t)D^\mu tD_\mu tV(t)\frac{1}{4}h(t)F^{\mu \nu }F_{\mu \nu }\right).$$
As with the tachyon kinetic term, the gauge kinetic term $`h(t)`$ vanishes at the local minimum $`t=0`$ according to the conjecture of Sen. A simple RG flow argument for this result is given in .
The derivatives in non-commutative directions are all suppressed by $`\alpha ^{}`$, and so can be dropped in the large $`B`$ limit, both in the action and in the gauge transformation laws. We again take the $`B`$-field to be non-vanishing in two directions $`x^i`$ and denote the remaining coordinates by $`x^a`$. Expanding out the action in the large $`B`$ limit yields
$$\begin{array}{cc}\hfill S=\frac{C}{G_s}d^{26}x\sqrt{G}(& \frac{1}{2}f(t)D^atD_at\frac{1}{2}h(t)D^aA_iD_aA^iV(t)\frac{1}{2}f(t)[A_i,t][A^i,t]\hfill \\ & +\frac{1}{4}h(t)[A_i,A_j][A^i,A^j]\frac{1}{4}h(t)F^{ab}F_{ab}).\hfill \end{array}$$
The $`A_i`$ now appear as scalar fields. The action in this limit is invariant under the gauge transformations
$$\begin{array}{cc}\hfill \delta _\lambda t& =i[t,\lambda ],\hfill \\ \hfill \delta _\lambda A_i& =i[A_i,\lambda ],\hfill \\ \hfill \delta _\lambda A_a& =_a\lambda i[A_a,\lambda ].\hfill \end{array}$$
By passing to the operator description in which $`\frac{1}{2\pi \theta }d^2x\mathrm{Tr}`$ and fields are replaced by matrix representations of Hilbert space operators, one sees that (3.1) is a 23+1 dimensional $`U(\mathrm{})`$ gauge theory coupled to the adjoint scalars $`t,A_i`$. This 23+1 gauge theory emerges even before considering the soliton background. Note that the transformation law of the tachyon is just the infinitesimal form of (2.1); the $`U(\mathrm{})`$ symmetry is a gauge symmetry, as advertised previously. At finite non-commutativity the term $`_i\lambda `$ in $`\delta _\lambda A_i`$ should be restored; the gauge symmetry then remains exact.
Before studying the soliton solution and its fluctuations, we briefly digress to explain the emergence of the 23+1 dimensional $`U(\mathrm{})`$ gauge theory from another viewpoint (besides the relation to Matrix theory given above), essentially repeating comments in . The open string theory effective action at lowest order is given by the path integral on the disk $`\mathrm{\Sigma }`$ with boundary conditions
$$g_{ij}_nx^j+B_{ij}_tx^j|_\mathrm{\Sigma }=0.$$
For $`B=0`$ one has Neumann boundary conditions, $`_nx^i=0`$, while for $`\alpha ^{}B\mathrm{}`$ one finds Dirichlet boundary conditions, $`_tx^i=0`$. In the large $`B`$ limit one thus finds D23-branes, though located at arbitrary transverse positions. In fact, this can be thought of as a continuous distribution of D23-branes with density proportional to $`B`$ (see for example \[24,,25\]). In the large $`B`$ limit, the theory governing such a system would be a 23+1 dimensional $`U(\mathrm{})`$ gauge theory, which is in accord with the result above.
3.2. Tachyon - gauge field fluctuations about the soliton
We now consider fluctuations of the action (3.1) around the soliton solution
$$t=t_{}\varphi _{00}.$$
The soliton breaks the $`U(\mathrm{})`$ gauge symmetry down to $`U(1)`$ (times the group “$`U(\mathrm{}1)`$” which will play no role in the discussion.) Working in unitary gauge we can take the tachyon fluctuations as in (2.1), (2.1). The fluctuations of the gauge field are
$$\begin{array}{cc}\hfill A_a(x^\mu )& =\underset{m,n=0}{\overset{\mathrm{}}{}}A_a^{mn}(x^b)\varphi _{mn}(x^i),\hfill \\ \hfill A_i(x^\mu )& =\underset{m,n=0}{\overset{\mathrm{}}{}}A_i^{mn}(x^a)\varphi _{mn}(x^j),\hfill \end{array}$$
where $`A_a^{mn}`$ and $`A_i^{mn}`$ are Hermitian as matrices with indices $`mn`$. Inserting the fluctuations into the action (3.1) we find that all modes with $`m,n>0`$ are projected out by the soliton background, as was seen previously for the pure tachyon fluctuations. Remaining are $`\delta t_{00}`$, $`A_i^{00}`$, $`A_a^{00}`$, as well as the $`0m`$ and $`m0`$ components of $`A_i`$, $`A_a`$. For convenience we drop the explicit $`00`$ indices on the first three fields; after integrating over the transverse space their action is found to be
$$S=T_{23}d^{24}x\sqrt{g}\left(\frac{1}{2}f(t_{})^a\delta t_a\delta tV(t_{}+\delta t)+\frac{1}{2}h(t_{})^aA_i_aA_i\frac{1}{4}h(t_{})F^{ab}F_{ab}\right),$$
where $`F_{ab}=_aA_b_bA_a`$ is the standard field strength. This is precisely the right action to describe the tachyon, gauge field, and transverse scalars on a D23-brane. The transverse scalars $`A_i`$ play the role of translational collective coordinate, and are guaranteed to be massless by the original non-commutative $`U(1)`$ gauge invariance.
The corresponding story for the $`0m`$ and $`m0`$ components of $`A_i`$, $`A_a`$ involves some additional subtleties. For $`A_i`$ we find the action,
$$S=T_{23}\underset{m=1}{\overset{\mathrm{}}{}}d^{24}x\sqrt{g}\left(\frac{1}{2}h(t_{})^aA_i^{0m}_aA_i^{m0}\frac{1}{2}t_{}^2f(t_{})A_i^{0m}A_i^{m0}\right).$$
A very similar result holds for $`A_a`$. The second term appears to be the standard mass term for W-bosons after gauge symmetry breaking, giving a mass
$$m_A^2=\left(\frac{t_{}}{2\pi \alpha ^{}}\right)^2.$$
We have inserted for definiteness the perturbative values $`h(t_{})=(2\pi \alpha ^{})^2,f(t_{})=1`$. The appearance of the tachyon VEV $`t_{}`$ is interesting because it is the only point in this work where we need a quantity that is not known exactly. It is possible to estimate $`t_{}`$ in the level-truncation scheme. This yields an expression $`t_{}^2\alpha ^{}`$ with a coefficient that rapidly converges to some value of $`𝒪(1)`$ \[6,,3,,20\].
Although these states have mass of order the string scale and are thus removed from the low energy spectrum, we are naively left with infinitely many massive modes on the soliton, degenerate as $`\theta \mathrm{}`$ and with a spacing proportional to $`1/\sqrt{\theta }`$ at finite $`\theta `$, in clear contradiction with the known spectrum of D-branes.
We believe the resolution of this puzzle involves the higher derivative terms in the commutative directions and the freezing out of the open string degrees of freedom in the closed string vacuum. For example, Kostelecky and Samuel showed in that non-local terms in the string field theory action which appear due to the substitution
$$\varphi \stackrel{~}{\varphi }e^{\alpha ^{}\mathrm{ln}3\sqrt{3}/4_\mu ^\mu }\varphi $$
in cubic interaction terms, have the effect of modifying the tachyon propagator in the presence of tachyon condensation so that there is no physical pole.
It is reasonable to expect that, in our context, the apparent infinite tower of massive gauge fields on the soliton is similarly removed. It would be nice to justify this expectation with an explicit computation, but we have not yet done so.
We should also point out that the higher derivative terms do not affect our previous calculations in any substantial way. By writing out the higher derivative terms explicitly and repeating our analysis one sees that the D23-brane precisely inherits the higher derivative terms with the same coefficients as on the D25-brane. So for the modes considered previously, whatever the higher derivative terms do to the spectrum on the D25-brane, they do the same thing on the D23-brane.
3.3. Multiple D-branes
Solitons with the tachyon profile given in (2.1) are interpreted as $`k`$ coincident D23-branes, and we expect to see the action (3.1) replaced by an action with $`U(k)`$ gauge invariance. First consider the field strength $`F_{ab}`$. Inserting the expansion (3.1) into (3.1) we find
$$F_{ab}=F_{ab}^{mn}\varphi _{mn},$$
with
$$F_{ab}^{mn}=_aA_b^{mn}_bA_a^{mn}i[A_a,A_b]^{mn}.$$
In the latter equation $`A_a`$ are being multiplied as matrices. We can now work out the final term in (3.1) for the background (2.1) as
$$\frac{C}{G_s}d^{26}x\sqrt{G}\left(\frac{1}{4}h(t_k)F^{ab}F_{ab}\right)=T_{23}d^{24}x\sqrt{g}\left(\frac{1}{4}\underset{m=0}{\overset{k1}{}}\underset{n=0}{\overset{\mathrm{}}{}}F^{abmn}F_{ab}^{nm}\right).$$
Components of the gauge field $`A_a^{mn}`$ with $`m,n>k1`$ have been projected out by the soliton background. There still remain in the action (3.1) components with $`m<k`$, $`nk`$ or $`mk`$, $`n<k`$, but we conjecture that these states are removed from the spectrum by the mechanism described in the previous subsection. What remains then is precisely the gauge kinetic term for a $`U(k)`$ gauge theory.
Similarly including $`t`$ and $`A_i`$ is straightforward. Ordinary derivatives are replaced by $`U(k)`$ covariant derivatives, and altogether we recover the $`U(k)`$ gauge invariant version of (3.1). This enhancement of the gauge symmetry for coincident solitons, though following rather easily from the present formalism, is highly nontrivial from a broader field theory vantage point, and provides strong evidence for the identification of the solitons with D-branes.
3.4. Massive modes
Though we have so far focussed on the tachyon and gauge field fluctuations, it is not hard to generalize our considerations to show that the solitons correctly inherit from the D25-brane the entire tower of open string states. Consider for illustration the quadratic terms for some massive field $`\psi `$,
$$S=\frac{C}{G_s}d^{26}x\sqrt{G}\left(\frac{1}{2}f(\mathrm{\Phi })_\mu \psi ^\mu \psi \frac{1}{2}m^2(\mathrm{\Phi })\psi ^2\right),$$
where $`\mathrm{\Phi }`$ denotes all fields in the theory. The mass of $`\psi `$ on the D25-brane is determined by evaluating $`f(\mathrm{\Phi })`$ and $`m(\mathrm{\Phi })`$ at the local maximum $`\mathrm{\Phi }=\mathrm{\Phi }_{}`$. The soliton background corresponds to $`\mathrm{\Phi }=\mathrm{\Phi }_{}\varphi _0`$, and we consider the fluctuations $`\psi =\psi _{mn}\varphi _{mn}`$. As before, $`m,n0`$ fluctuations are projected out; $`m=0`$, $`n>0`$ and $`m>0`$, $`n=0`$ fluctuations are removed (we conjecture) by higher derivative terms; so what remains, after integrating over the soliton, is
$$S=T_{23}d^{24}x\left(\frac{1}{2}f(\mathrm{\Phi }_{})_\mu \psi _{00}^\mu \psi _{00}\frac{1}{2}m^2(\mathrm{\Phi }_{})\psi _{00}^2\right).$$
Since what appears in the action are the functions $`f(\mathrm{\Phi })`$ and $`m(\mathrm{\Phi })`$ evaluated at the local maximum of the potential, $`\psi `$ has the same mass on the soliton as on the original D25-brane. This shows that the spectrum of the soliton is inherited from the D25-brane, and so can be identified with the spectrum of a D23-brane including all massive string states.
4. Non-commutative solitons on Type II D-branes
The construction of D-branes as non-commutative solitons in the bosonic string has an obvious extension to Type II superstring theory. Type IIA theory contains non-BPS Dp-branes for p odd and IIB theory has non-BPS Dp-branes for p even \[18,,26,,27,,28\] which have a natural interpretation as sphalerons . Solitons in level truncated open superstring field theory have been studied in .
To be concrete consider the non-BPS space-filling $`D9`$-brane of IIA theory. As before, we turn on a $`B`$-field in two dimensions and consider the limit of large $`\alpha ^{}B`$. The analog of equation (2.1) is
$$\frac{C}{G_s}d^{10}x\sqrt{G}\left(\frac{1}{2}f(t)G^{\mu \nu }_\mu t_\nu t+\mathrm{}V(t)\right),$$
with
$$C=g_sT_{9A}.$$
$`T_{9A}`$ is the tension of the Type IIA D9-brane.
As before, the tachyon potential is not known exactly, but it is known to have a reflection symmetry and the height of the potential follows from the identification of the global minimum with the closed string vacuum. We choose the global minimum to be at $`t=0`$ for easier comparison with our previous results. The reflection symmetry then acts by $`(tt_{})(tt_{})`$. A schematic potential with these properties is given in fig. 2.
Fig. 2: The open superstring tachyon potential.
Repeating the analysis done in the bosonic string shows that the solution
$$t=t_{}\varphi _0$$
has the same tension as a D7-brane. By turning on constant $`B`$-fields in an even number of dimensions we construct the rest of the non-BPS branes of IIA. Note that in contrast to the bosonic string, here we obtain all non-BPS D-branes through this construction since in type II theory the codimension of these branes differs by an even integer.
This solution is unstable as in the previous analysis with the instability reflecting the presence of a tachyon of the correct mass on the D7-brane. Multiple D-branes can be incorporated as before, and the analysis of gauge field couplings is essentially the same as the discussion in the previous section.
We can apply the reflection symmetry to obtain the reflected solution <sup>3</sup> Note that this solution is not of the canonical form described in of a critical point of the potential times a projection operator. If $`V^{}(t)=t_i(t\lambda _i)`$ then $`t=\lambda _k(P+1)`$ with $`P^2=P`$ is a solution if $`\lambda _k=\lambda _i/2`$ for some $`i`$.
$$t=t_{}(2\varphi _0).$$
This reflected solution represents a physically distinct configuration; it is asymptotic to a solution whose ten-form RR field strength differs by one unit from (4.1) .
We can also construct an unexpected solution (and its $`Z_2`$ image) since the potential has an additional stationary point at $`t=2t_{}`$:
$$t=2t_{}\varphi _0.$$
Since $`V(2t_{})`$ vanishes, this soliton has vanishing tension in the limit of infinite non-commutativity! It is easy to see from the considerations in that this object is stable. If we apply this construction in $`10`$ Euclidean dimensions we would seem to obtain a zero action instanton. The presence of such an object would clearly have dramatic consequences.
One way to understand the solution (4.1) is to consider a level $`k`$ solution
$$t_k=2t_{}(\varphi _0+\varphi _1+\mathrm{}+\varphi _{k1})$$
in the limit $`k\mathrm{}`$ as in the discussion in section 2.4. This gives the solution $`t_{\mathrm{}}=2t_{}`$ which is simply the $`Z_2`$ image of the closed string vacuum at $`t=0`$. In the same way that adding up an infinite number of D-23 branes gave a D25-brane, here adding up an infinite number of these tensionless 7-branes gives the other closed string vacuum.
It may well be that at finite $`\alpha ^{}B`$ this solution develops a non-zero tension representing the fact that once derivatives are included it costs non-zero action to move from one vacuum to the other. A subleading tension of order $`1/(g_s\alpha ^{}B)`$ would also have interesting consequences since there would then be a one-parameter family of non-perturbative objects with variable tension. Since asymptotically one is in the closed string vacuum and $`B`$ can be gauged to zero, the value of $`B`$ is not a closed string modulus, but rather some modulus of the solution. In principle the tension could be non-zero only at one string loop, but this seems unlikely since it would modify the perturbative structure of string theory. More work is clearly needed to understand this mysterious object.
The other unstable D-brane system of interest in type II string theory is a $`Dp\overline{Dp}`$ system of BPS $`Dp`$-branes. These objects have a complex tachyon with a “Mexican hat” potential given by rotating fig.2 about a vertical axis at $`t=t_{}`$. There are also two $`U(1)`$ gauge fields, $`A^+`$ and $`A^{}`$, coming from the D-brane and anti D-brane respectively, and the tachyon carries charge one under the relative $`U(1)`$ with gauge field $`A^+A^{}`$. Both of the $`U(1)`$’s become non-commutative in the presence of a non-zero $`B`$-field with the tachyon transforming in the bi-fundamental representation.
For topologically trivial solutions we can use the relative $`U(1)`$ to make the tachyon field $`t`$ real. The above solutions are then also solutions to this system. Since $`t`$ is real, the tachyon configuration does not act as a source for the relative $`U(1)`$ gauge field and so the solutions carry no net lower D-brane charge. The coefficient in front of the action is now $`2T_p`$, so we can interpret the analogue of the first solution (4.1) as an unstable $`D(p2)\overline{D(p2)}`$ brane system with energy $`2T_{p2}`$. The analogue of the second solution (4.1) is apparently a tensionless bound state of the $`D(p2)\overline{D(p2)}`$ system.
BPS $`D(p2)`$-branes can be constructed as vortices in the complex tachyon field . One can find an exact vortex solution using a two-dimensional effective action which includes the tachyon field and the $`U(1)\times U(1)`$ gauge fields<sup>4</sup> The previous version of this paper had a crucial minus sign error in the equations of motion, the vortex solution presented in was an important guide in correcting this error.
The action is
$$S=𝑑z𝑑\overline{z}\left(\overline{D_\mu t}D^\mu t\frac{1}{4}F_{\mu \nu }^+F^{+\mu \nu }\frac{1}{4}F_{\mu \nu }^{}F^{\mu \nu }V(t,\overline{t})\right).$$
The gauge field strengths have the canonical form and the covariant derivatives are given by
$$D_\mu t=_\mu t+i(A_\mu ^+ttA_\mu ^{})$$
and
$$\overline{D_\mu T}=_\mu \overline{t}i(\overline{t}A_\mu ^+A_\mu ^{}\overline{t}).$$
In the limit of large non-commutativity, where we can ignore derivative terms, the equations of motion are
$$\begin{array}{cc}\hfill [A^{+\nu },[A_\nu ^+,A_\mu ^+]]& =A_\mu ^+t\overline{t}tA_\mu ^{}\overline{t}+t\overline{t}A_\mu ^+tA_\mu ^{}\overline{t}\hfill \\ \hfill [A^\nu ,[A_\nu ^{},A_\mu ^{}]]& =A_\mu ^{}\overline{t}t\overline{t}A_\mu ^+t+\overline{t}tA_\mu ^{}\overline{t}A_\mu ^+t\hfill \\ \hfill A_\mu ^+A^{+\mu }t+2A^{\mu +}tA_\mu ^{}tA^\mu A_\mu ^{}& =\alpha tV^{}\beta V^{}t.\hfill \end{array}$$
The values of $`\alpha ,\beta `$ depend on the ordering prescription for $`V`$.
These equations are solved by
$$\begin{array}{cc}\hfill t& =t_{}\underset{i=0}{\overset{\mathrm{}}{}}\varphi _{i,i+1}\hfill \\ \hfill A_z^+& =A_{\overline{z}}^+=t\overline{t}=1\hfill \\ \hfill A_z^{}& =A_{\overline{z}}^{}=\overline{t}t=1\varphi _0.\hfill \end{array}$$
Note that the field strengths of both gauge fields vanish since $`A_z^\pm `$ are Hermitian and so $`F_{z\overline{z}}[A_z,A_{\overline{z}}]=0`$. Similarly, $`D_\mu D^\mu t`$ vanishes, so to get a solution we can choose the ordering in $`V`$ to be $`V(\overline{t}t1)`$. This gives $`\alpha =1,\beta =0`$ and we have a solution using $`tV^{}t\varphi _0=0`$.
In the commutative case the vortex carries a non-zero D7 charge which arises from the coupling of the RR potential to the relative $`U(1)`$ gauge field strength, $`C_8(F^+F^{})`$. In the large non-commutativity limit we have found $`F^+F^{}=0`$, so one would naively think that there is no induced $`D7`$ charge. This is incorrect. The couplings of RR fields to gauge and tachyon fields take an elegant form worked out in . The relevant contribution is
$$Cd\mathrm{Tr}\left(t\overline{Dt}\right).$$
Substituting (4.1) one finds that the last term gives the correct induced $`D7`$ charge. To show this, it is important to keep the spatial derivative term in $`D_\mu t`$.
5. Fundamental Strings as Electric Flux Tubes
After tachyon condensation the open string degrees of freedom are confined and excitations of the theory are closed strings. We would like to describe these excitations as solitons. To this end we construct in the following an electric flux tube with the tension of the fundamental string. The classical confinement of electric flux found here is similar in spirit to that discussed in , but distinct from that discussed in \[11,,10\].
Naively, a gauge theory action of the type (3.1) gives, in the strong coupling limit $`h(t)G_s^10`$, very heavy electric flux tubes — one expects a tension of order $`h^1G_s`$. However, with the nonlinearity inherent in the Born-Infeld action, we will see that the energy cost of a flux quantum saturates, and the flux tube remains light in the limit.
5.1. The flux tube and its tension
We need the effective action for the tachyon and the gauge fields for configurations carrying electric flux in a particular direction, say $`x^1`$. The task is simplified with the introduction of a large $`B`$-field in all $`24`$ transverse directions $`x^i`$, $`i=2,\mathrm{},25`$. Then derivatives along these directions are negligible. Furthermore, it is sufficient (for now) to consider gauge field configurations that are constant in time and along the flux tube. The tachyon potential in the presence of constant background fields is known from a theorem of Sen \[33,,2\]: The potential is universal up to tachyon independent deformations of the overall metric. Thus, in our context, the action is of the Born-Infeld type
$$S=d^{26}xV(t)\sqrt{det[\eta _{\mu \nu }+2\pi \alpha ^{}F_{\mu \nu }]}.$$
(Other recent discussions of the action include \[34,,35\]). Coordinates were chosen so $`G_{\mu \nu }=\eta _{\mu \nu }`$, and $``$ products are implied. In this section we absorb the overall tension of the D25-brane in the potential $`V(t_{})`$. It is sufficient to retain just one component of the gauge field, i.e.
$$S=d^{26}xV(t)\sqrt{1(2\pi \alpha ^{}\dot{A}_1)^2}.$$
The analysis will be simplest in the canonical formalism. We therefore compute the momentum conjugate to the gauge field
$$=\frac{V(t)(2\pi \alpha ^{})^2\dot{A}_1}{\sqrt{1(2\pi \alpha ^{}\dot{A}_1)^2}}.$$
Quantization of the electric flux $``$ plays an important role in our discussion. In the large $`B`$ limit $``$ is the electric flux of a $`1+1`$ dimensional $`U(\mathrm{})`$ gauge theory. Flux quantization in $`1+1`$ Yang-Mills theory is analyzed in , and we can apply a similar analysis here. Electric flux receives contributions from two sources: from matter and gauge field charge fluctuations, and from charges at infinity. In the present context these charges have values corresponding to the endpoints of open strings. It follows that the electric flux in each $`U(1)`$ subgroup of $`U(\mathrm{})`$ is quantized, and that it can be thought of roughly as the number of open strings (signed depending on orientation) ending on the corresponding D23-brane. More precisely, in a diagonal basis the eigenstates of $``$ are
$$=\frac{1}{(2\pi \theta )^{12}}\underset{k=0}{\overset{\mathrm{}}{}}n_k\varphi _k,n_k𝐙.$$
From now on we will drop the overall factors of $`(2\pi \theta )^{12}`$; they can be absorbed into field redefinitions. The overall $`U(1)`$ flux,
$$d^{24}x=\underset{k=0}{\overset{\mathrm{}}{}}n_kN$$
is a conserved quantity and can be identified with the total number of fundamental strings in the state. The individual eigenvalues $`n_k`$ are not conserved, and the general quantum state will correspond to a superposition of different $`n_k`$. In a complete quantum mechanical treatment it would be important to include the effect of transitions among different flux states,<sup>5</sup> One might expect such fluctuations to be rather large, since the effective coupling $`1/V(t)`$ in (5.1) (we have absorbed a factor of $`1/g_s`$ in the tachyon potential) is quite large as the tachyon field approaches the closed string vacuum $`V(t)0`$. It would be interesting if the fluctuations could be related to the usual divergent vacuum fluctuations in the spatial position of a perturbative closed string. We thank S. Shenker and A. Lawrence for discussions on this point. but we will be less ambitious and consider the properties of a single flux state of the form (5.1).
We consider the Hamiltonian relevant to finding the configuration which minimizes the energy in a given total flux sector,
$$H=d^{25}x\left[\sqrt{V(t)^2+^2/(2\pi \alpha ^{})^2}+\lambda _1\right]\lambda N,$$
where $`\lambda `$ is a Lagrange multiplier enforcing the condition that we consider configurations with $`N`$ flux quanta. The corresponding equations of motion are
$$\frac{\delta H}{\delta t}=\frac{V(t)V^{}(t)}{\sqrt{V(t)^2+^2/(2\pi \alpha ^{})^2}}=0,$$
$$\frac{\delta H}{\delta }=\left[\frac{}{(2\pi \alpha ^{})^2\sqrt{V(t)^2+^2/(2\pi \alpha ^{})^2}}+\lambda \right]=0.$$
We want solutions that exist in the vacuum after tachyon condensation, so $`V(t)`$ should vanish asymptotically. We therefore try solutions of the form
$$t=t_0\varphi ,$$
where $`\varphi `$ is one of the functions satisfying $`\varphi \varphi =\varphi `$. (5.1) is solved by taking $`t_0`$ to be an extremum of $`V`$. Taking $`t=t_{}`$, (5.1) is solved by
$$=e\varphi ,$$
and the flux quantization condition (5.1) determines $`e=N/k`$ for $`\varphi `$ at level $`k`$. (5.1) also determines the Lagrange multiplier $`\lambda `$, but this is not needed in the subsequent discussion. Inserting the solution back into the Hamiltonian (5.1) the tension is found to be
$$T=\sqrt{k^2V(t_{})^2+N^2/(2\pi \alpha ^{})^2}=\frac{1}{2\pi \alpha ^{}}\sqrt{k^2/g_s^2+N^2}.$$
This is as expected for an $`(N,k)`$ type string.
An even simpler solution is found by taking $`t=0`$. In this case $`V(0)=0`$ so (5.1) places no constraint on the form of $``$, although it does determine $`\lambda `$. $``$ can take any form (5.1) consistent with the flux quantization condition (5.1). The tension of such a solution is
$$T=\sqrt{V(0)^2+e^2/(2\pi \alpha ^{})^2}=\frac{N}{2\pi \alpha ^{}},$$
which is the result expected for $`N`$ fundamental strings.
The way we derived it, the tension was essentially guaranteed to come out right. The nontrivial part was the magic of non-commutativity, which allows one to find a localized solution to equations which would otherwise be difficult to analyze.
The action that we started with (5.1) is proportional to $`1/g_s`$ (absorbed in $`V(t)`$). As mentioned above, it is therefore surprising that the flux tube solution (5.1) has energy of $`𝒪(g_s^0)`$. This is possible because, in the nonlinear Born-Infeld Hamiltonian (5.1), the coefficient of the electric field term is independent of $`V(t)`$, whereas it would have a coefficient $`1/V`$ in a Yang-Mills type action. Alternatively, the usual expansion of the Born-Infeld action in powers of the field strength is inappropriate; one is in a ‘relativistic’ limit of the field dynamics, with $`V`$ playing the role of mass.
5.2. Fluctuations
We would like to consider also excitations of the fundamental string. This is more demanding because generally the action (5.1) receives corrections with unsuppressed derivatives in the longitudinal direction $`x^1`$. However, the action is still applicable for fluctuations with large wavelength $`\sqrt{\alpha ^{}}F_{\mu \nu }^{}F_{\mu \nu }`$ (prime denotes derivatives along $`x^1`$). Note that this condition still allows for amplitudes of order string scale $`\alpha ^{}F_{\mu \nu }1`$ so the nonlinear terms in the Born-Infeld action do contain valid information. As mentioned earlier, a true quantum state corresponding to a fundamental consists of a superposition of solutions with a given total electric flux. However, we will consider the simpler exercise of analyzing fluctuations around a single solution with fixed $`n_k`$, and leave the more complete treatment for future work.
Again, our considerations will be simplest in the canonical formalism. To find the Hamiltonian, we expand the determinant in (5.1) and write the action
$$S=d^{26}xV(t)\sqrt{det(1\mathrm{l}+𝐅)(1F_{0\alpha }M^{\alpha \beta }F_{0\beta })},$$
where
$$M^{\alpha \beta }=\left(\frac{1\mathrm{l}}{1\mathrm{l}+𝐅}\right)_{\mathrm{sym}}^{\alpha \beta }.$$
Here $`\alpha ,\beta ,\mathrm{}`$ are purely spatial indices and the matrices $`1\mathrm{l},𝐅`$ have components $`\delta _{\alpha \beta },F_{\alpha \beta }`$. In this subsection we take $`2\pi \alpha ^{}=1`$ to simplify the formulae; these factors are easily restored by comparing with the previous computation. The canonical momenta are
$$^\alpha =\frac{V(t)M^{\alpha \beta }F_{0\beta }}{\sqrt{1F_{0\alpha }M^{\alpha \beta }F_{0\beta }}},$$
and the Hamiltonian becomes
$$H=d^{25}x\left(^\alpha \dot{A}_\alpha \right)=d^{25}x\left[\sqrt{^\alpha M_{\alpha \beta }^\beta +V(t)^2det(1\mathrm{l}+𝐅)}+A_0_\alpha ^\alpha \right],$$
where
$$M_{\alpha \beta }=\delta _{\alpha \beta }F_{\alpha \gamma }F_\beta ^\gamma .$$
The equation of motion for $`A_0`$ is the Gauss law constraint
$$_\alpha ^\alpha =0.$$
Hereafter it is convenient to choose the gauge $`A_0=0`$.
Derivatives in the transverse directions are negligible, so transverse field strengths simplify: $`F_{ij}=0`$ and $`F_{1i}=A_i^{}`$. The Hamiltonian can therefore be written
$$H=d^{25}x\sqrt{(^1)^2(1+(\stackrel{}{A}^{})^2)+\stackrel{}{}^2+(\stackrel{}{}\stackrel{}{A}^{})^2+V(t)^2(1+(\stackrel{}{A}^{})^2)}.$$
The vector notation applies to the transverse directions $`x^i`$, $`i=2,\mathrm{},25`$. Considering just the fundamental string we can take $`V(t)=0`$ from here onwards.
Fluctuating strings are described as flux tube solutions based on functions satisfying $`\varphi \varphi =\varphi `$, as before, but with the origin displaced in a manner depending on the coordinates along the string
$$\varphi =\varphi (x^if^i(x^0,x^1)).$$
A suitable ansatz is
$$_1=e_1\varphi ,\stackrel{}{}=\stackrel{}{e}\varphi ,\stackrel{}{A}^{}=\stackrel{}{a}^{}\varphi .$$
The coefficient $`e_1`$ will generally be $`N/k`$, but in this section we consider a single string, so $`e_1=1`$. The functions $`\stackrel{}{e},\stackrel{}{a}`$ depend on the string coordinates $`x^a=x^0,x^1`$ but not on the transverse coordinates $`x^i`$.
We want to find the effective dynamics of the variables $`f^i`$. It is not sufficient to insert our ansatz in the Hamiltonian (5.1) because that would not ensure that the equations of motion for other fields are satisfied. The correct procedure is known as Hamiltonian reduction (it is described in detail in a closely related context in ). By solving the constraints, the Hamiltonian is expressed in terms of the reduced variables $`\stackrel{}{f}`$ and their conjugate momenta. In the present context the important constraint is Gauss’ law (5.1). Inserting our ansatz we find
$$e_1\frac{\varphi }{x^i}\frac{f^i}{x^1}+e^i\frac{\varphi }{x^i}=0.$$
Recalling that $`e_1=1`$, a simple solution is
$$\stackrel{}{f}^{}=\stackrel{}{e}.$$
We need to find the momentum conjugate to $`\stackrel{}{f}^{}`$. The fields $`\stackrel{}{}`$ and $`\stackrel{}{A}`$ are canonically conjugate in the original theory. This implies that $`\stackrel{}{e}`$ and $`\stackrel{}{a}`$ are canonically conjugate in the reduced theory. From (5.1) we therefore find that the momenta conjugate to $`\stackrel{}{f}^{}`$ are $`\stackrel{}{\pi }\stackrel{}{a}^{}`$.<sup>6</sup> Up to a factor of the wavenumber, when we expand in the usual oscillator basis. We conclude that the reduced Hamiltonian
$$H=𝑑x^1\sqrt{1+\stackrel{}{\pi }^2+(\stackrel{}{f}^{})^2+(\stackrel{}{\pi }\stackrel{}{f}^{})^2},$$
describes the fluctuations of the flux tube. The dynamics of transverse modes was “integrated out” in the Hamiltonian reduction.
The significance of this Hamiltonian is best exhibited in the Lagrangian formalism. We therefore compute the time derivative using Hamilton’s equations
$$\dot{\stackrel{}{f}}=\frac{\delta H}{\delta \stackrel{}{e}}=\frac{\stackrel{}{\pi }+\stackrel{}{f}^{}(\stackrel{}{f}^{}\stackrel{}{\pi })}{\sqrt{1+\stackrel{}{\pi }^2+(\stackrel{}{f}^{})^2+(\stackrel{}{\pi }\stackrel{}{f}^{})^2}},$$
and find
$$L=d^2x\stackrel{}{\pi }\dot{\stackrel{}{f}}𝑑x^0H=d^2x\sqrt{(1(\dot{\stackrel{}{f}})^2)(1+(\stackrel{}{f}^{})^2)+(\dot{\stackrel{}{f}}\stackrel{}{f}^{})^2}.$$
This should be compared with the standard Nambu-Goto action
$$L_{NG}=d^2x\sqrt{\mathrm{det}[_aX_\mu ^aX^\mu ]}=d^2x\sqrt{(\dot{\stackrel{}{X}})^2(\stackrel{}{X}^{})^2(\dot{\stackrel{}{X}}\stackrel{}{X}^{})^2}.$$
In the static gauge $`X^\mu =(x^0,x^1,f^i)`$ used here this reduces precisely to (5.1). The effective action for the fluctuations of the flux tube is therefore the Nambu-Goto action!
The next step is to quantize the fluctuations of the flux tube. Their action is the Nambu-Goto action so the result is that of a closed fundamental string. We have justified the Lagrangian (5.1) only for large semi-classical waves so it is in this regime that the spectrum of the flux tube should be compared with that of fundamental strings. Our computation demonstrates a perfect agreement in the allowed energies and their degeneracies.
It is tantalizing that, if we extend this reasoning beyond its apparent regime of validity, we find the entire fundamental string spectrum as simple excitations in the open string field theory. This would include very light modes — such as the massless graviton, and even the closed string tachyon. Usually this kind of identification would be impossible in principle: the quantum fluctuations of a soliton are collective excitations rather than fundamental objects, because the soliton itself is made out of more basic constituents. We are better off here because, after tachyon condensation, the flux tubes are the lightest objects in the theory and therefore subject to quantization. Although this removes the objection of principle, we presently have no justification to trust our result for light modes.
5.3. Multiple strings and interactions
One can also consider electric fluxes lying in $`U(k)`$ subgroups of the $`U(\mathrm{})`$ gauge group, leading to additional solutions. These parallel closely the construction of Matrix string theory . With $`x^1`$ compactified on a large circle of radius $`R`$, one can describe multiply wrapped strings via the holonomy of the gauge field that twists the eigenvalues of the $`U(k)`$ electric field into a ‘long string’ or ‘slinky’. One can follow the arguments given in leading to the identification of the interactions of such a string with the effective twist operator that arises when $`SU(2)`$ symmetry is restored by two strands of the Matrix string coming together in the transverse space. Thus we see that the flux tubes at least qualitatively interact in the proper way; it would be interesting to see if at least some perturbative string amplitudes can be reproduced within the non-commutative framework (for instance, tree level amplitudes that just exchange winding among these macroscopic strings). A potential difficulty in this exercise is that the original string coupling constant $`g_s`$ appears only through the tachyon potential $`V(t)`$ in (5.1); after tachyon condensation $`V0`$, thus it seems that the electric flux lines do not know about the original string coupling. One does not a priori know why the interaction strength of the flux lines is $`𝒪(g_s)`$ and not $`𝒪(1)`$.
One might also be concerned that there appear to be two separate descriptions of a $`k`$ times wound fundamental string: Namely $`k`$ units of flux in a $`U(1)`$ solution, as well as $`k`$ units of flux wound into a ‘slinky’ in $`U(k)`$. In Matrix string theory, these two carry different quantum numbers; the rank of the gauge group corresponds to the number of bits of discrete light cone momentum of the Matrix string, while the electric flux represents D0-brane charge that the string is bound to. In the present context, both of these objects are embedded in the same $`U(\mathrm{})`$ gauge theory, and simply represent different spatial distributions of the flux lines; there is no apparent reason why the the path integral over the configuration space of the flux lines in the strongly coupled gauge theory will not smoothly evolve these two flux configurations into one another.
To summarize, we find macroscopic fundamental strings appearing as light electric flux excitations of the open string field theory, after the tachyon is condensed to form the closed string vacuum. The tension of the flux tubes is just right, and the fluxes join and split in the manner familiar from string perturbation theory (although it is not at present understood whether their interaction strength is related to the string coupling $`g_s`$).
6. Comments and questions
It is remarkable that turning on a large $`B`$-field simplifies the structure of open string field theory in such a dramatic way that one can construct exact D-brane and fundamental string solutions. The full force of this construction relies heavily on Sen’s conjecture that tachyon condensation represents the closed string vacuum. We use this conjecture to determine the height of the tachyon potential, but also to argue that at the end of the construction one can gauge $`B`$ to zero far from the solution so that one is discussing D-branes and fundamental string in the usual closed string vacuum with $`B=0`$.
Similarities to Matrix theory have been a persistent thread running through our discussion. In both cases a large dimensionless parameter is introduced — $`\alpha ^{}B`$ for the non-commutative limit, and the boost rapidity in Matrix theory — upon which quantities of interest do not depend, yet whose introduction facilitates calculations. The large $`B`$-field of non-commutative geometry induces a macroscopic density of lower-dimensional D-branes, while tachyon condensation removes the higher-dimensional D-brane and makes the effective gauge coupling large in regions described by the closed string vacuum. Thus the ingredients of the Matrix theory limit are present; furthermore, the scaling limit of is the analogue of the Maldacena scaling limit ($`\alpha ^{}0`$, with the energies of stretched open strings held fixed). It would be interesting to make the relation between these two circles of ideas more precise.
Our work raises a number of other interesting questions:
1. Is it possible to also construct the NS 5-brane (or 21-brane in the bosonic string) using these methods<sup>7</sup> The analogy with Matrix theory suggests that the problem is similar in nature to the construction of the transverse Matrix fivebrane. In particular, the energy density of the object scales as $`g_s^2`$, whereas classical open string field configurations have energy densities scaling as $`g_s^1`$.?
2. How does one understand the freezing out of the open string degrees of freedom and the ability to gauge $`B`$ to zero in the closed string vacuum, directly in string field theory?
3. What is the interpretation of the very light solitons (massless to leading order in $`1/\alpha ^{}B`$) found in the superstring?
4. What are the leading $`1/\alpha ^{}B`$ corrections to the results presented here?
5. What is the coupling strength of the fundamental string constructed in section 4?
We hope to address some of these questions in future work.
Acknowledgements: We would like to thank Rajesh Gopakumar, Kentaro Hori, Shamit Kachru, Albion Lawrence, Shiraz Minwalla, Steve Shenker, Eva Silverstein, and Andy Strominger for helpful conversations. This work was supported by DOE grant DE-FG02-90ER-40560 and NSF grant PHY-9901194. F.L. was supported in part by a Robert R. McCormick fellowship. P.K. thanks the Harvard and Stanford theory groups for hospitality; J.H. would like to thank the Institute for Advanced Study for hospitality; and E.M. thanks the Rutgers University theory group for hospitality.
Note Added: As this work was being typed we became aware of work by K. Dasgupta, S. Mukhi, and G. Rajesh which overlaps our discussion of non-commutative solitons in type II string theory. We thank them for bringing this work to our attention.
References
relax A. Sen, “Descent Relations Among Bosonic D-branes,” Int.J. Mod. Phys. A14 (1999) 4061, hep-th/9902105. relax A. Sen, “Universality of the tachyon potential,” JHEP 9912 (1999) 027, hep-th/9911116. relax A. Sen and B. Zwiebach, “Tachyon Condensation in String Field Theory,” hep-th/9912249. relax J. A. Harvey and P. Kraus, “D-Branes as Lumps in Bosonic Open String Field Theory,” JHEP 0004:012,2000, hep-th/0002117. relax R. de Mello Koch, A. Jevicki, M. Mihailescu and R.Tatar, “Lumps and P-Branes in Open String Field Theory,” hep-th/0003031. relax V. A. Kostelecky and S.Samuel, “On a Nonperturbative Vacuum for the Open Bosonic String,” Nucl.Phys. B336 (1990) 263. relax Nathan Berkovits, Ashoke Sen and Barton Zwiebach, “Tachyon Condensation in Superstring Field Theory”, hep-th/0002211. relax J. A. Harvey, D. Kutasov and E. J. Martinec, “On the relevance of tachyons”, hep-th/0003101. relax A. Sen, “Supersymmetric World-volume Action for Non-BPS D-branes,” JHEP 9910:008, 1999; hep-th/9909062. relax O.Bergman, K.Hori and P.Yi, “Confinement on the Brane,” hep-th/0002223. relax P. Yi, “Membranes from Five-Branes and Fundamental Strings from D$`p`$-Branes,” Nucl. Phys. B550 (1999) 214; hep-th/9901159. relax C. G. Callan and J. Maldacena, “Brane Death and Dynamics from the Born-Infeld Action,” Nucl.Phys. B513 (1998) 198, hep-th/9708147. relax P. S.Howe, N.D.Lambert and P.C. West, “The Selfdual String Soliton,” Nucl.Phys. B515 (1998) 203, hep-th/9709014. relax G. W. Gibbons, “Born-Infeld Particles and Dirichlet P-Branes” Nucl.Phys. B514 (1998) 603, hep-th/9709027. relax A.Connes, M. R. Douglas and A.Schwarz, “Noncommutative Geometry and Matrix Theory:Compactification on Tori,” JHEP 9802, 003 (1998), hep-th/9711162. relax N. Seiberg and E. Witten, “String Theory and Noncommutative Geometry,” JHEP 9909:032, 1999; hep-th/9908142. relax R. Gopakumar, S.Minwalla and A.Strominger, “Noncommutative Solitons,” hep-th/0003160. relax A. Sen, “Stable non-BPS bound states of BPS D-branes,” JHEP 9808 (1998) 010, hep-th/9805019; “SO(32) spinors of type I and other solitons on brane-antibrane pair,” JHEP 9809 (1998) 023, hep-th/9808141; “Type I D-particle and its interactions,” JHEP 9810 (1998) 021, hep-th/9809111; “Non-BPS states and branes in string theory,” hep-th/9904207, and references therein. relax A. Dhar, G. Mandal and S. R. Wadia, “String field theory of two-dimensional QCD: A Realization of $`W_{\mathrm{}}`$ algebra,” Phys. Lett. B329 (1994) 15, hep-th/9403050; “Nonrelativistic fermions, coadjoint orbits of $`W_{\mathrm{}}`$ and string field theory at $`c=1`$,” Mod. Phys. Lett. A7 (1992) 3129, hep-th/9207011. relax N. Moeller and W. Taylor, “Level Truncation and the Tachyon in Open Bosonic String Field Theory,” hep-th/0002237. relax T. Banks, W. Fischler, S.H. Shenker and L. Susskind, “M Theory As A Matrix Model: A Conjecture”, Phys.Rev.D55 (1997) 5112, hep-th/9610043. relax B. de Wit, J. Hoppe and H. Nicolai, “On the quantum mechanics of supermembranes”, Nucl. Phys. B305 (1988) 545. relax M. R. Garousi, “Tachyon couplings on non-BPS D-branes and Dirac-Born-Infeld action,” hep-th/0003122. relax N. Ishibashi, “A Relation Between Commutative and Noncommutative Descriptions of D-Branes,” hep-th/9909176. relax J.X. Lu and S. Roy, “(p + 1)-Dimensional Noncommutative Yang-Mills and D($`p2`$) Branes”, hep-th/9912165. relax O. Bergman and M. R. Gaberdiel, “Stable non-BPS D-particles,” Phys. Lett. B441 (1998) 133, hep-th/9806155. relax A. Sen, “BPS D-branes on non-supersymmetric cycles,” JHEP 9812 (1998) 021, hep-th/9812031. relax P. Hořava, “Type IIA D-Branes, K-Theory, and Matrix Theory,” Adv. Theor. Math. Phys. 2 (1999) 1373, hep-th/9812135. relax J. A. Harvey, P. Kraus and P. Hořava, “D Sphalerons and the Topology of String Configuration Space,” JHEP 0003:021,2000, hep-th/0001143. relax E.Witten, “Noncommutative Tachyons and String Field Theory,” hep-th/0006071. relax C. Kennedy and A. Wilkins,“Ramond-Ramond Couplings on Brane-Antibrane Systems,” Phys. Lett. B464 (1999) 206; hep-th/9905195. relax J. Kogut and L. Susskind “Vacuum Polarization and the Absence of Free Quarks in Four-Dimensions,” Phys. Rev. D9 (1974) 3501. relax A. Sen, “Supersymmetric world-volume action for non-BPS D-branes,” JHEP 9910 (1999) 008, hep-th/9909062. relax E. A. Bergshoeff, M. de Roo, T. C. de Wit, E. Eyras and S. Panda, “T-duality and actions for non-BPS D-branes,” hep-th/0003221. relax J. Kluson, “Proposal for non-BPS D-brane action,” hep-th/0004106. relax E. Witten, “Theta Vacua In Two-Dimensional Quantum Chromodynamics,” Nuovo Cim. A51, 325 (1979). relax F. Larsen and F. Wilczek, “Classical Hair in String Theory I: General Formulation,” Nucl. Phys. B475 (1996) 627. hep-th/9604134. relax L. Motl, “Proposals on nonperturbative superstring interactions” hep-th/9701025; T. Banks and N. Seiberg, “Strings from matrices,” Nucl. Phys. B497 (1997) 41, hep-th/9702187; R. Dijkgraaf, E. Verlinde and H. Verlinde, “Matrix String Theory”, Nucl.Phys. B500 (1997) 43-61, hep-th/9703030. relax J. Maldacena, “The large N limit of superconformal field theories and supergravity,” Adv. Theor. Math. Phys. 2 (1998) 231, hep-th/9711200. relax K. Dasgupta, S. Mukhi, and G. Rajesh, “Noncommutative Tachyons”, hep-th/0005006.
|
warning/0005/hep-th0005154.html
|
ar5iv
|
text
|
# Connecting Green’s Functions in an Arbitrary Pair of Gauges and an Application to Planar Gauges
## 1 INTRODUCTION
### The importance of Standard Model \[SM\] calculations in particle Physics cannot be overestimated.The Standard Model calculations require the choice of a gauge.There are a variety of these which have been used in various different context. Some of these are the Lorentz-type gauges, axial-type gauges,light-cone gauge,planar gauges,radial gauges,nonlinear gauges , the R<sub>ξ</sub>-gauges,Coulomb gauge etc.Different gauges have been found useful and convenient in different calculational context .A priori, we expect from gauge invariance,that the values for physical observables calculated in different gauges are identical.Formal proofs of such equivalence for S-matrix elements has been given in a given class of gauges, say the Lorentz-type gauges with a variable gauge parameter $`\lambda `$ .Some isolated attempts to connect *S-matrix elements* in *singular* \[rather than a class of them\] gauges also have been done.For example, formal equivalence of S-matrix elements in the Coulomb and the Landau gauges \[both singular gauges\] has also been established .Similar formal attempts to connect the \[singular\] temporal gauge with Feynman gauge in the canonical formalism has also been done . It is important to note that, however, the Green’s functions in the gauges such as the Coulomb ,the axial,the planar and the light-cone in the path integral formulation are ambiguous on account of the unphysical singularities in their propagators. Hence, it becomes important to know how to define the Green’s functions in such gauges in such a manner that they are compatible to those in a well-defined covariant gauge such as the Lorentz gauge. A general procedure that connects Green’s functions in the path integral formulation in two classes of gauges, say the Lorentz and the axial,has been lacking until recently. Such comparisons are important not just in a formal sense but also in practice.Precisely because of this, the proper treatment of the 1/$`\eta .q`$ type poles in axial and light-cone gauges (and also similar questions in the Coulomb gauge ) has occupied a lot of attention and the criterion used for their validation has, in fact, been the comparison with the calculational results in the Lorentz gauges.Such comparative calculations, where possible \[8,9 10\]have to be done by brute force and have been done toO\[g$`{}_{}{}^{4}]`$ generally; thus limiting the scope of their confirmation. At a time, a physically observable anomalous dimension was reported to differ in Lorentz and axial gauges . Such questions motivate us to develop a general path integral formalism that can address all these questions in a wide class of gauges in a single framework. In a purely Feynman diagrammatic approach,we ,of course, have the attempt of Cheng and Tsai .
### Recently,we developed a general path integral formalism for connecting pairs of Yang-Mills effective actions and applied it, in particular, to connecting the Lorentz and the axial type gauges.\[In reference 11, we have also applied the procedure to connecting to the general BRS-anti-BRS invariant effective action of Baulieu and Thierry-Mieg\] . Our formalism is based on the Finite Field-dependent BRS \[FFBRS\]transformations that connect the two path-integrals.These transformations are of a \[field-dependent\] BRS-type and are *evaluated in a closed form* and leave the vacuum expectation value of a gauge-invariant observable *explicitly invariant* as one transforms from say, a Lorentz-type to an axial-type gauge. Our procedure, in fact, gives a way of *defining* carefully Green’s functions in the axial-type gauges by a path-integral that explicitly takes care of the ill-defined nature of the propagator and in a manner compatible with those in Lorentz gauges.We found an effective treatment of the axial propagator using this procedure and applied this formalism also to show the preservation of the Wilson loop in axial gauges .To summarize, the output of the works \[11-14\] has been (i) an explicit closed field transformation in \[A,c,$`\overline{c}`$\] space to connect the path-integrals in the two gauges that preserves gauge-invariant observables, (ii)A relation that allows one to calculate the Green’s functions in axial gauges , *compatible with those in Lorentz gauges by the very construction*,(iii)way of dealing with axial poles that is compatible with the Lorentz gauges. A simplified proof of (ii) above was also given using the BRS WT-identity .
### Now that the above techniques have matured,we propose in this work to generalize our previous works in several directions in this one.In Section 3, we prove the existence of the FFBRS transformation the connects Faddeev-Popov effective actions \[FPEA\] with any two arbitrary gauge functions F and F’.This applies to all the gauges mentioned above in the first a paragraph \[ expect the Planar gauge for which the action in not in the manifest FPEA form\].In section 4, we generalize the identity that connects the Green’s functions in gauge F’ to the Green’s functions in the gauge F to the case of an arbitrary pair.In section 5, we develop the results for connecting Green’s functions in planar gauges to those in Landau gauge. In section 6, we give a heuristic treatment that uses the results of section 3 and which connects the vacuum expectation of gauge-invariant observable in planar gauges to those in Lorentz gauges.This result can also be verified by explicit calculations that run parallel to those in the section 3;and is not limited,however, by the heuristic treatment we have given for compactness.
## 2 PRELIMINARY
### In this section,we shall review the results in earlier works \[11-13\] and introduce notations.We consider two arbitrary gauge fixing functions F\[A\] and F’\[A\] which could be nonlinear or non-covariant and an interpolating gauge function F$`{}_{}{}^{M}[A]`$= $`\kappa `$F’\[A\]+(1-$`\kappa `$)F\[A\]; 0$`\kappa 1`$. The Faddeev-Popov effective action\[FPEA\] in each case is given by
#### *S<sub>eff</sub>* = S<sub>0</sub> \+ S<sub>gf</sub> +*S<sub>gh</sub>* (2.1)
with
S<sub>gf</sub> = -$`\frac{1}{2\lambda }`$ d$`{}_{}{}^{4}x`$ $`\widehat{F^\gamma }`$\[A\]<sup>2</sup> (2.2a)
and
*S<sub>gh</sub>= -*$``$d$`{}_{}{}^{4}x`$ $`\overline{c}`$<sup>α</sup>$`\widehat{M^{\alpha \beta }}c^\beta `$ (2.2b)
with
$`\widehat{M^{\alpha \beta }}`$=$`\frac{\delta }{\delta }\frac{\widehat{F^\alpha [A]}}{A_\mu ^\gamma }`$D<sup>γβ</sup><sub>μ</sub> \[A\] (2.3)
D<sup>αβ</sup><sub>μ</sub> \[A\]=$`\delta ^{\alpha \beta }_\mu +g_0f^{\alpha \beta \gamma }A_\mu ^\gamma `$ (2.3a)
### We denote the FPEA for the three cases $`\widehat{F}`$ =F,F’,F<sup>M</sup> by *S<sub>eff</sub>, S $`{}_{eff}{}^{}{}_{}{}^{}`$ and S<sup>M</sup><sub>eff</sub>* respectively.BRS transformations for the three effective actions are:
#### $`\varphi `$<sub>i</sub> =$`\varphi `$<sub>i</sub> +$`\delta _{iBRS}[\varphi ]\delta \mathrm{\Lambda }`$ (2.4)
#### with $`\delta _{iBRS}[\varphi ]`$ equal to D<sup>αβ</sup><sub>μ</sub>c<sup>β</sup>,-1/2 g$`{}_{0}{}^{}f_{}^{\alpha \beta \gamma }`$c<sup>β</sup>c<sup>γ</sup>,and $`\stackrel{~}{F}/\lambda `$ respectively for A,c and $`\overline{c}`$.In the case of the mixed gauge condition, $`\delta _{iBRS}[\varphi ]`$ for $`\overline{c}`$ is $`\kappa `$-dependent and in this case we show this explicitly by expressing (2.4) as
#### $`\varphi `$<sub>i</sub> =$`\varphi `$<sub>i</sub> +$`\stackrel{~}{\delta }_{iBRS}[\varphi ,\kappa ]\delta \mathrm{\Lambda }`$$`\varphi `$<sub>i</sub> +$`\{\stackrel{~}{\delta _1}_{iBRS}[\varphi ]+\kappa `$$`\stackrel{~}{\delta _2}_{iBRS}[\varphi ]\}\delta \mathrm{\Lambda }`$ (2.4a)
### Following observations in and , we guess and later prove the field transformations that takes one from the gauge F to gauge F’.It is given by the finite field-dependent BRS transformation \[FFBRS\]
#### $`\varphi `$<sub>i</sub> =$`\varphi `$<sub>i</sub> +$`\delta _{iBRS}[\varphi ]\mathrm{\Theta }`$ \[$`\varphi ]`$ (2.5)
#### where $`\mathrm{\Theta }[\varphi `$\] has been constructed by the integration of the infinitesimal field-dependent BRS \[IFBRS\] transformation
#### $`\frac{d\varphi _i}{d\kappa }=`$$`\delta _{iBRS}[\varphi (\kappa )]\mathrm{\Theta }`$’\[$`\varphi `$($`\kappa `$)\] (2.6)
(where $`\delta _{iBRS}[\varphi (\kappa )]`$ refers to the BRS variations of the gauge F) with the boundary condition $`\varphi `$\[$`\kappa `$=1\]=$`\varphi `$’and $`\varphi `$\[$`\kappa `$=0\]=$`\varphi `$ and is given in a closed form by
$`\mathrm{\Theta }`$\[$`\varphi `$\] = $`\mathrm{\Theta }`$’\[$`\varphi `$\]\[exp{f\[$`\varphi `$\]}-1\]/f\[$`\varphi `$\] (2.6a)
and f is given by
f =$`_i`$ $`\frac{\delta \mathrm{\Theta }^{}}{\delta \varphi _i}`$$`\delta _{iBRS}[\varphi ]`$ (2.6b)
We wish to develop ,in this work, an FFBRS for connecting two arbitrary gauges F and F’.We shall show that in this case it is given by an FFBRS of the form (2.5) with $`\mathrm{\Theta }`$’\[$`\varphi `$($`\kappa `$)\] given by
$`\mathrm{\Theta }`$’\[$`\varphi `$($`\kappa `$)\] =$`id^4y`$ $`\overline{c}`$<sup>γ</sup>(y) (F<sup>γ</sup>\[A($`\kappa )]`$ -F’ <sup>γ</sup>\[A($`\kappa )]`$) (2.7)
The Jacobian for the IFBRS transformation is defined as
D$`\varphi `$\[$`\kappa `$=0\] = D$`\varphi `$\[$`\kappa `$\]J($`\kappa `$) = D$`\varphi `$\[$`\kappa `$+d$`\kappa `$\]J($`\kappa `$+d$`\kappa )`$ (2.8)
The change in the Jacobian for the IFBRS of (2.6) is given by
-$`\frac{1}{J}`$$`\frac{dJ}{d\kappa }d\kappa `$=$``$ d$`{}_{}{}^{4}x`$ $``$<sub>i</sub>($`\pm )`$$`\frac{\delta \varphi __i^{}(x,\kappa )}{\delta \varphi _i(y,\kappa )}_{_{x=y}}`$ (2.9)
and is evaluated easily as
-$`\frac{1}{J}`$$`\frac{dJ}{d\kappa }d\kappa `$=-i$``$d$`{}_{}{}^{4}x`$$`\{\overline{c}`$(M-M’)c +$`\frac{1}{\lambda }F[FF^{}]`$} (2.10)
Further we define
*S<sub>eff</sub>$`{}_{}{}^{M}[`$*$`\varphi `$($`\kappa `$),$`\kappa ]`$$``$*S<sub>eff</sub>*\[$`\varphi `$($`\kappa `$)*$`]`$+S<sub>1</sub>*\[$`\varphi `$($`\kappa `$),$`\kappa ]`$ (2.11)
We then have,
*S<sub>1</sub>*\[$`\varphi `$($`\kappa `$),$`\kappa ]`$=$``$d$`{}_{}{}^{4}x`${-$`\frac{1}{2\lambda }`$\[$`\kappa ^2F^{}[A(\kappa `$)\]<sup>2</sup>\+ 2$`\kappa (1\kappa )F[A(\kappa `$)\]$`F^{}[A(\kappa )]`$
$`+\kappa `$($`\kappa 2)F[A(\kappa )]^2`$$`\kappa `$$`\overline{c}(\kappa )\{`$M\[A($`\kappa )]M^{}[A(\kappa )]\}c(\kappa )`$} (2.12)
We introduce the following notation
\<\<f\[$`\varphi `$($`\kappa `$)\]\>\><sub>κ</sub>$``$ $``$ D$`\varphi `$($`\kappa `$)f\[$`\varphi `$($`\kappa `$)\] exp{ i *S<sub>eff</sub>$`{}_{}{}^{M}[`$*$`\varphi `$($`\kappa `$),$`\kappa ]`$} (2.13)
In references and it was established that the expectation value of a gauge-invariant observable \<\<O\[A($`\kappa `$)\]\>\><sub>κ</sub> is independent of $`\kappa `$ iff the Jacobian J($`\kappa `$) and the effective action *S<sub>eff</sub>$`{}_{}{}^{M}[`$*$`\varphi `$($`\kappa `$),$`\kappa ]`$ satisfy
\<\<$`\frac{i}{J}`$$`\frac{dJ}{d\kappa }`$+$`\frac{dS_1[\varphi (\kappa ),\kappa ]}{d\kappa }`$\>\><sub>κ</sub>$``$0 (2.14)
In Section 3,we shall verify (2.14) for the IFBRS of (2.6).
## 3 FFBRS TRANSFORMATIONS CONNECTING ANY PAIR \[F,F’\] OF GAUGES
### In the references explicit field transformations of the FFBRS type that connected various pairs of equivalent effective actions for the Yang-Mills theory were constructed.In this work,we wish to give a result that generalizes it for FPEA in a pair of arbitrary gauges F and F’ that includes for example those mentioned at the beginning of the Introduction \[except the planar one\].These field transformations are such that they preserve, by an explicit construction, the expectation values of gauge-invariant observables.
### Consider the expectation value of a gauge invariant observable O\[A\] in the mixed gauge:<sup>1</sup><sup>1</sup>1 It is understood that like the Lorentz gauges,an appropriate O($`ϵ)`$ term is necesary in (3.1) to make it well-defined.
#### \<\<O\[A\]\>\>$`{}_{\kappa }{}^{}`$ $``$ D$`\varphi `$($`\kappa `$)O\[A($`\kappa `$)\] exp{ i *S<sub>eff</sub>$`{}_{}{}^{M}[`$*$`\varphi `$($`\kappa `$),$`\kappa ]`$} (3.1)
where $`\varphi (\kappa )`$ represent fields as defined in (2.6);with the specific $`\mathrm{\Theta }`$’ given by (2.7).
We show that
$`\frac{d}{d\kappa }`$\<\<O\[A\]\>\>$`{}_{\kappa }{}^{}`$0 (3.2)
As shown in , (3.2) is valid iff
0 $``$$``$ D$`\varphi `$($`\kappa `$){$`\frac{1}{J}`$$`\frac{dJ}{d\kappa }`$-i$`\frac{dS_1[\varphi (\kappa ),\kappa ]}{d\kappa }`$}O\[A($`\kappa `$)\]exp{ i *S<sub>eff</sub>$`{}_{}{}^{M}[`$*$`\varphi `$($`\kappa `$),$`\kappa ]`$} (3.3)
where we recall from (2.11) \[noting the definitions below (2.3a)\]
*S<sub>1</sub>*\[$`\varphi `$($`\kappa `$),$`\kappa ]=`$*S<sub>eff</sub>$`{}_{}{}^{M}[`$*$`\varphi `$($`\kappa `$),$`\kappa ]`$*-S<sub>eff</sub>*\[$`\varphi `$($`\kappa `$)\] *(3.4)*
We verify the result (3.3) explicitly below.The proof proceeds much as in and and hence we shall give it briefly.The change in the Jacobian under the IFBRS of (2.6)
$`\delta `$$`\varphi `$<sub>i</sub>($`\kappa `$) =$`\delta _{iBRS}[\varphi (\kappa )]\mathrm{\Theta }`$’ \[$`\varphi (\kappa )]`$d$`\kappa `$
=$`\delta _{iBRS}[\varphi (\kappa )]`$ $`id^4y`$ $`\overline{c}`$<sup>γ</sup>(y) (F<sup>γ</sup>\[A($`\kappa )]`$ -F’ <sup>γ</sup>\[A($`\kappa )]`$)d$`\kappa `$ (3.5)
viz.$`\frac{1}{J}`$$`\frac{dJ}{d\kappa }d\kappa `$ is given by (2.10); viz
$`\frac{1}{J}`$$`\frac{dJ}{d\kappa }d\kappa `$=i$``$d$`{}_{}{}^{4}x`$ {$`\overline{c}`$(M-M’)c +$`\frac{1}{\lambda }F[FF^{}]\}`$ (3.6)
Further,using the expression for *S<sub>1</sub>*\[$`\varphi `$($`\kappa `$),$`\kappa ]`$ of (2.12),we find
$`\frac{dS_1[\varphi (\kappa ),\kappa ]}{d\kappa }`$=$`\underset{}{A}`$$`\mathrm{\Theta }`$’+ $`\underset{}{B}`$ (3.7)
with
$`\underset{}{A}`$=$``$d$`{}_{}{}^{4}x`$ $`\frac{1}{\lambda }\{`$$`\kappa `$$`{}_{}{}^{2}F_{}^{\gamma }(M^{}c)^\gamma +\kappa `$(1-$`\kappa )`$$`F^\gamma (Mc)^\gamma +\kappa (1`$$`\kappa )`$$`F^\gamma (M^{}c)^\gamma +\kappa (`$$`\kappa `$-2)$`F^\gamma (Mc)^\gamma `$
$`+\kappa `$$`F^\gamma [(MM^{})c]^\gamma +\kappa `$$``$d$`{}_{}{}^{4}y`$d$`{}_{}{}^{4}z`$$`\overline{c}`$$`{}_{}{}^{\gamma }(x)\frac{\delta [F^\gamma (x)F^\gamma (x)]}{\delta A_\mu ^\beta (y)\delta A_\sigma ^\eta (z)}`$(Dc)<sup>η</sup>(z)(Dc)<sup>β</sup>(y)} (3.7a)
and
$`\underset{}{B}`$=$``$d$`{}_{}{}^{4}x`$$`\{\frac{1}{\lambda }[\kappa F^2+(12\kappa )FF^{}+(\kappa 1)F^2]+`$ $`\overline{c}`$$`[(MM^{})c]`$} (3.7b)
We note that the last term in $`\underset{}{A}`$vanishes by Bose symmetry;and $`\underset{}{A}`$ can be reorganized further as,
$`\underset{}{A}`$=$``$d$`{}_{}{}^{4}x`$ $`\frac{1}{\lambda }\{`$$`\kappa [F^{}F][(1\kappa )Mc+\kappa M^{}c]\}`$ (3.7c)
The \[generalized\] antighost equation of motion is given by
0 $``$$``$ D$`\varphi `$($`\kappa `$)O\[A($`\kappa `$)\]exp{ i *S<sub>eff</sub>$`{}_{}{}^{M}[`$*$`\varphi `$($`\kappa `$),$`\kappa ]`$}f(A($`\kappa `$),c($`\kappa `$),$`\kappa )`$
$``${M\[A($`\kappa )](1\kappa )+\kappa M^{}[A(\kappa )]\}c(\kappa )`$ (3.8)
where f\[A,c,$`\kappa `$\] is an arbitrary functional of A and c but not of $`\overline{c}`$.
Using (3.8) above,we can convert $`\underset{}{A}`$$`\mathrm{\Theta }`$’ term as
$``$ D$`\varphi `$($`\kappa `$)O\[A($`\kappa `$)\] $`\underset{}{A}`$$`\mathrm{\Theta }`$’exp{ i *S<sub>eff</sub>$`{}_{}{}^{M}[`$*$`\varphi `$($`\kappa `$),$`\kappa ]`$}
=-$``$ D$`\varphi `$($`\kappa `$)O\[A($`\kappa `$)\]$`\mathrm{\Theta }`$$``$d$`{}_{}{}^{4}x`$$`\frac{\kappa }{\lambda }[F^{}F]^\gamma `$$`\frac{1}{i}`$$`\frac{\delta }{\delta \overline{c^\gamma (x)}}`$ exp{ i *S<sub>eff</sub>$`{}_{}{}^{M}[`$*$`\varphi `$($`\kappa `$),$`\kappa ]`$}
=$``$d$`{}_{}{}^{4}x`$ $`\frac{\kappa }{\lambda }<<[FF^{}]^2`$O\[A\]\>\><sub>κ</sub> (3.9)
Using (3.9),(3.7b) and (3.6), we find
\<\<$`\frac{i}{J}`$$`\frac{dJ}{d\kappa }`$+$`\frac{dS_1[\varphi (\kappa ),\kappa ]}{d\kappa }`$\>\><sub>κ</sub>$``$0 (3.10)
This implies \[ 11,13\]that \<\<O\[$`\varphi `$\]\>\><sub>κ</sub> is independent of $`\kappa `$.Hence,
$``$ D$`\varphi `$O\[A\] exp{ i *S<sub>eff</sub>*\[$`\varphi `$\]}=$``$ D$`\varphi ^{}`$O\[A’\] exp{ i *S’<sub>eff</sub>*\[$`\varphi ^{}`$\]} (3.11)
Thus the FFBRS transformation of (2.5) takes one from gauge F to gauge F’ in the sense already pointed out at the beginning of the section. Such transformations can be used in establishing a relation between Green’s functions in the two gauges along the lines of that correctly tackle the inherent problems in many of the gauges.
## 4 A RELATION BETWEEN ARBITRARY GREEN’S FUNCTIONS IN TWO GAUGES
### In the work of reference , we had obtained a relation between arbitrary Green’s functions in axial gauges and Green’s functions in the Lorentz gauges and the former were “compatible” with those in the well-defined Lorentz gauges.This arbitrary Green’s function in axial gauges could be expressed either as \[i\] a series of Green’s functions in Lorentz gauges that also involve insertions of the BRS variations in the Lorentz gauges OR \[ii\]as an integral over parameter $`\kappa `$ involving the Green’s functions evaluated in the Mixed gauges. It was found from a practical view-point that the latter result is indeed more amenable to calculations. An axial Green’s function, to a given order, can be evaluated by an integral over $`\kappa `$ of a sum of a finite number of Feynman diagrams in the mixed gauge. This form has been employed in obtaining an Axial gauge prescription compatible with the Lorentz gauges . A simpler derivation , based on the BRS, was also given for this latter result (only).In this section,we generalize, the procedure of the reference to connecting arbitrary Green’s functions in an arbitrary pair of gauges. We emphasize that our treatment in applies equally well to operator Green’s functions \[needed say in perturbative QCD applications\] as to the usual Green’s functions.
### In this section,we shall show that the method of Ref., based on BRS WT-identity, can be generalized to a any pair of two gauges F and F’.
### We define<sup>2</sup><sup>2</sup>2 As pointed out in section 3,a proper $`\epsilon term`$ is required to be added to *S<sub>eff</sub>$`{}_{}{}^{M}[`$*$`\varphi `$,$`\kappa ]`$ to make ¡¡O\[$`\varphi `$\]¿¿<sub>κ</sub> well defined.This term depends on F,F’ and has to be done seperately in each specific context.See e.g. the applications in section 5 to planar gauges and also in . , for any operator O\[$`\varphi `$\], not necessarily local,
#### \<\<O\[$`\varphi `$\]\>\><sub>κ</sub>$`=`$ D$`\varphi `$O\[$`\varphi `$\]exp{ i *S<sub>eff</sub>$`{}_{}{}^{M}[`$*$`\varphi `$,$`\kappa ]`$} (4.1)
\[Note:Unlike in section 3 \[see (3.1)\],we do not now allow the *integration variable* $`\varphi `$ to depend on $`\kappa .`$This suits us here.\] Then
$`\frac{d}{d\kappa }`$\<\<O\[$`\varphi `$\]\>\><sub>κ</sub> =i\<\<$``$d$`{}_{}{}^{4}x`$ O\[$`\varphi `$\]{-$`\frac{1}{\lambda }[F+\kappa (F^{}F)][F^{}F]+`$$`\overline{c}`$(M-M’)c\>\><sub>κ</sub> (4.2)
The WT-identities for *S<sub>eff</sub>$`{}_{}{}^{M}[`$*$`\varphi `$,$`\kappa ]`$ are
\<\<$``$d$`{}_{}{}^{4}x`${J<sup>α</sup><sub>μ</sub>(x)D<sup>αβ</sup><sub>μ</sub>c<sup>β</sup>+$`\overline{\xi ^\alpha }`$$`(x)[1/2g_0f^{\alpha \beta \gamma }c^\beta c^\gamma ](x)`$
-$`\xi ^\alpha (x)`$\[$`\frac{F+\kappa (F^{}F)}{\lambda }`$\]<sup>α</sup>(x)}\>\> = 0 (4.3)
Recalling the definition (2.7) of $`\mathrm{\Theta }`$’;we operate by
$`\mathrm{\Theta }`$’\[-i$`\frac{\delta }{\delta J(y)},`$$`i\frac{\delta }{\delta \overline{\xi (y)}}`$$`,i\frac{\delta }{\delta \xi (y)}]d^4y\{F^\gamma `$\[-i$`\frac{\delta }{\delta J(y)}]F^\gamma [`$-i$`\frac{\delta }{\delta J(y)}]\}\frac{\delta }{\delta \xi ^\gamma (y)}`$ (4.4)
We then obtain,
\<\<$``$d$`{}_{}{}^{4}x`${J<sup>α</sup><sub>μ</sub>(x)D<sup>αβ</sup><sub>μ</sub>c<sup>β</sup>+$`\overline{\xi ^\alpha }`$$`(x)[1/2g_0f^{\alpha \beta \gamma }c^\beta c^\gamma ](x)\xi ^\alpha (x)`$\[$`\frac{F+\kappa (F^{}F)}{\lambda }`$\]<sup>α</sup>(x)}(-i$`\mathrm{\Theta }^{})`$
-i$``$d$`{}_{}{}^{4}x`$ (M-M’)c(x) $`\overline{c(x)}`$ +i$``$d$`{}_{}{}^{4}x`$\[$`\frac{F+\kappa (F^{}F)}{\lambda }`$\]<sup>α</sup>\[F-F’\]<sup>α</sup>\>\> $``$0 (4.5)
Finally, we operate on both sides by -iO\[-i$`\frac{\delta }{\delta J(y)},`$$`i\frac{\delta }{\delta \overline{\xi (y)}}`$$`,i\frac{\delta }{\delta \xi (y)}]`$ and set J=$`\xi `$$`=\overline{\xi }`$ =0 in the end to obtain
0 = $``$D$`\varphi `$ exp{ i *S<sub>eff</sub>$`{}_{}{}^{M}[`$*$`\varphi `$,$`\kappa ]`$}{$``$\[$`\frac{\delta O}{\delta A_\mu }`$D$`{}_{\mu }{}^{}c`$ +$`\frac{\delta ^RO}{\delta c^\alpha }`$$`[1/2g_0f^{\alpha \beta \gamma }c^\beta c^\gamma ]+`$$`\frac{\delta ^RO}{\delta \overline{c^\alpha }}`$\[$`\frac{F+\kappa (F^{}F)}{\lambda }`$\]<sup>α</sup>\](-i$`\mathrm{\Theta }^{})`$
+O\[A,c,$`\overline{c}`$\]{$``$d$`{}_{}{}^{4}x`$$`\{\overline{c}`$(M-M’)c +$`\frac{F+\kappa (F^{}F)}{\lambda }[FF^{}]`$}} (4.6)
Using (4.2 ) and (2.4a ) ; we obtain
$`\frac{d}{d\kappa }`$\<\<O\[$`\varphi `$\]\>\><sub>κ</sub>= i$``$D$`\varphi `$ exp{ i *S<sub>eff</sub>$`{}_{}{}^{M}[`$*$`\varphi `$,$`\kappa ]`$}
$``${$`\stackrel{~}{\delta }_{iBRS}[\varphi ,\kappa ]`$$`d^4y`$ $`\overline{c}`$<sup>γ</sup>(y) (F<sup>γ</sup>\[A\] -F’<sup>γ</sup>\[A\](y))$`\frac{\delta ^LO}{\delta \varphi _i}\}`$ (4.7)
Integrating from $`\kappa `$ = 0 to 1 we obtain
\<\<O\[$`\varphi `$\]\>\><sub>F</sub>’= \<\<O\[$`\varphi `$\]\>\><sub>F</sub> +i$``$$`__0`$<sup>1</sup>d$`\kappa `$ D$`\varphi `$ exp{ i *S<sub>eff</sub>$`{}_{}{}^{M}[`$*$`\varphi `$,$`\kappa ]`$}
$``$$`_i(\stackrel{~}{\delta _{1i}[\varphi }`$\]+$`\kappa `$$`(\stackrel{~}{\delta _{2i}[\varphi }`$\])(-i$`\mathrm{\Theta }^{})`$$`\frac{\delta ^LO}{\delta \varphi _i}`$ (4.8)
This gives us the expression for the Green’s function \[which depending on the choice of O\[$`\varphi `$\] could be a primary one or of an operator\] in one gauge F’ related to that in F ,with finite number of additional terms that can be evaluated by Feynman diagram techniques as mentioned in the beginning of this section.
## 5 Planer Gauges
### Planar gauges have been found very useful in the perturbative QCD calculations on account of a simpler propagator structure \[that avoids the double pole in ($`\eta `$.k)\] and other attractive features . Gauges similar to the planer gauges have also been used in the renormalization of higher derivative gravity theories.
### Planar gauges are defined by<sup>3</sup><sup>3</sup>3 For mathematical rigor,we may replace $`^2`$by $`^2iϵ`$ in the equations (5.1),(5.2) and (5.5) and in the definition of $`\overline{c}`$” in (5.3).
#### S<sup>P</sup><sub>gf</sub> = -$`\frac{1}{2\lambda \eta ^2}`$ d$`{}_{}{}^{4}x`$ $`\eta `$.A $`^2`$$`\eta `$.A ; $`\eta `$<sup>2</sup>$``$0 ,$`\lambda =1.`$ (5.1)
#### and the accompanying ghost term
#### *S<sup>P</sup><sub>gh</sub>* =*-*$``$d$`{}_{}{}^{4}x`$ $`\overline{c}`$<sup>α</sup>$`^2`$$`\eta `$$`.Dc^\alpha `$ (5.2)
or equivalently
*S<sup>P</sup><sub>gh</sub>* *$``$-*$``$d$`{}_{}{}^{4}x`$ $`\overline{c}`$<sup>α</sup>$`\eta `$$`.Dc^\alpha `$ (5.3)
with $`\overline{c}`$<sup>α</sup> defined as $`^2`$$`\overline{c}`$<sup>α</sup>.The net FPEA for the planar gauges,S<sup>P</sup><sub>eff</sub>,has a BRS invariance under
$`\delta \overline{c}`$ =$`\frac{\eta .A}{\eta ^2\lambda }\delta \mathrm{\Lambda }`$ (5.4)
and $`\delta A`$ and $`\delta c`$ as in (2.4 )
### We note the the gauge fixing term is not manifestly of the form of (2.2) viz. ~ $``$d$`{}_{}{}^{4}x`$ $`\widehat{F^\gamma }`$\[A\]<sup>2</sup>.Hence,we cannot apply the results established in the last two sections directly to this case <sup>4</sup><sup>4</sup>4 See, however, an alternate way presented in Section 6. .We can establish a route to connect this gauge to the Lorentz gauge as follows:
#### S<sup>P</sup><sub>gf</sub> –\> S$`^{\stackrel{~}{L}}`$<sub>gf</sub> $``$-$`\frac{1}{2\lambda \eta ^2}`$ d$`{}_{}{}^{4}x`$ $``$.A $`^2`$$``$.A –\>S<sup>L</sup><sub>gf</sub> (5.5a)
#### or as,
#### S<sup>P</sup><sub>gf</sub> –\> S<sup>A</sup><sub>gf</sub> –\>S<sup>L</sup><sub>gf</sub> (5.5b)
### We shall call, for the sake of nomenclature,the gauge $`\stackrel{~}{L}`$ the “pseudo-Lorentz” gauge.The formal connection between the pseudo-Lorentz and the Lorentz gauges as well as that between planar and the axial gauge can be established easily from the work of Lee and Zinn-Justin itself.We note from ,
### \<\<O\[A\]\>\><sub>f</sub>$``$$``$ D$`\varphi `$ exp{ i *S*$`{}_{0}{}^{}+iS_G^A`$}$`\begin{array}{c}\mathrm{\Pi }\\ \alpha ,x\end{array}\delta `$ ($`\eta `$.A$`{}_{}{}^{\alpha }f^\alpha `$ )O\[A\] (5.6)
### is independent of ’f’ for any gauge-invariant observable O\[A\].Now, the vacuum expectation values of gauge-invariant observable O\[A\] in the planar and the axial gauges are related to the above quantity \<\<O\[A\]\>\><sub>f</sub> by the relations:
#### \<\<O\[A\]\>\><sub>A</sub>$`=`$$``$ Df exp{$`\frac{i}{2\lambda }`$$``$ d$`{}_{}{}^{4}x`$f<sup>2</sup>} \<\<O\[A\]\>\><sub>f</sub> (5.7)
#### and<sup>5</sup><sup>5</sup>5 For mathematical rigor,we may replace $`^2`$by $`^2iϵ`$ ,as earlier, in the equation below and everywhere else necessary.
#### \<\<O\[A\]\>\><sub>P</sub>=$`\frac{1}{N_P}`$$``$ Df exp{$`\frac{i}{2\lambda \eta ^2}`$$``$ d$`{}_{}{}^{4}x`$f $`^2`$f}\<\<O\[A\]\>\><sub>f</sub> (5.8)
#### \[Equation (5.8) is compensated by appropriate normalization factor $`\frac{1}{N_P}`$*relative to* (5.7)\].Therefore, in view of the f-independence of \<\<O\[A\]\>\><sub>f</sub>, the two quantities above are equal:
#### \<\<O\[A\]\>\>$`{}_{A}{}^{}=`$\<\<O\[A\]\>\><sub>P</sub> (5.9)
#### Thus, as for as the gauge-invariant observables are concerned, once the Lorentz -axial route is established, Lorentz–planar gauge connection also becomes available.We shall establish this explicit connection later in section 6using the results of Sections 3. But first, we would like to derive a result similar to that in Section 4 for the the planar–Lorentz Green’s functions connection applicable to the *arbitrary* Green’s functions.
We define
W$`{}_{}{}^{P}[J,\xi ,\overline{\xi }]=`$$`\frac{1}{N_P}`$$``$ D$`\varphi `$ exp{ i *S*$`{}_{0}{}^{}+iS_G^A\frac{i}{2\lambda \eta ^2}`$$``$ d$`{}_{}{}^{4}x`$ $`\eta `$.A $`^2`$$`\eta `$.A+ Source Terms} (5.10)
As for the ghost term,we can use an identical one both for the planar and the axial gauges ,obtained from (5.2) by a field redefinition as in (5.3).
S$`{}_{G}{}^{A}=`$*-*$``$d$`{}_{}{}^{4}x`$ $`\overline{c}`$<sup>α</sup>$`\eta `$$`.Dc^\alpha `$ (5.11)
We now define, for the singular axial gauge ($`\eta `$.A$`{}_{}{}^{\alpha }f^\alpha `$ )=0,
W<sub>f</sub>$`[J,\xi ,\overline{\xi }]`$$``$$``$ D$`\varphi `$ exp{ i *S*$`{}_{0}{}^{}+iS_G^A`$\+ Source Terms}$`\begin{array}{c}\mathrm{\Pi }\\ \alpha ,x\end{array}\delta `$ ($`\eta `$.A$`{}_{}{}^{\alpha }f^\alpha `$ ) (5.12)
We then have
W$`{}_{}{}^{P}[J,\xi ,\overline{\xi }]=`$$`\frac{1}{N_P}`$$``$ Df exp{$`\frac{i}{2\lambda \eta ^2}`$$``$ d$`{}_{}{}^{4}x`$f $`^2`$f}W<sub>f</sub> $`[J,\xi ,\overline{\xi }]`$ (5.13)
W$`{}_{}{}^{A}[J,\xi ,\overline{\xi }]=`$$``$ Df exp{$`\frac{i}{2\lambda }`$$``$ d$`{}_{}{}^{4}x`$f<sup>2</sup>}W<sub>f</sub>$`[J,\xi ,\overline{\xi }]`$ (5.14)
We already know how the Green’s functions in Axial gauges are linked to those in Lorentz gauges.To relate the Green’s functions in the planar gauge to those in Lorentz gauges, we proceed as follows:
Consider the expectation value of an arbitrary operator O\[$`\varphi `$\], possibly multi-local,in the Lorentz gauge, the singular axial gauge ($`\eta `$.A$`{}_{}{}^{\alpha }f^\alpha `$ )=0 and in the planar gauge:
\<\<O\[$`\varphi `$\]\>\>$`{}_{L}{}^{}=`$$``$D$`\varphi `$O\[$`\varphi `$\]exp{ i *S<sub>eff</sub>$`{}_{}{}^{L}[`$*$`\varphi `$\]+ $`\epsilon O_L`$} (5.15)
where $`\epsilon O_L`$ are the $`\epsilon `$-terms in the Lorentz gauges:
$`\epsilon O_L`$ = $``$ d$`{}_{}{}^{4}x`$ \[$`\frac{1}{2}`$A<sub>μ</sub>A<sup>μ</sup> -$`\overline{c}`$c \] (5.16)
\<\<O\[$`\varphi `$\]\>\>$`{}_{f}{}^{}=`$$``$D$`\varphi `$O\[$`\varphi `$\]exp{ i *S<sub>0</sub>*\[$`\varphi `$\]+iS$`{}_{G}{}^{_A}[\varphi `$\]+ $`\epsilon O_A(f)`$}$`\begin{array}{c}\mathrm{\Pi }\\ \alpha ,x\end{array}\delta `$ ($`\eta `$.A$`{}_{}{}^{\alpha }f^\alpha `$ ) (5.17)
where $`\epsilon O_A(f)`$ terms are the appropriate O\[$`\epsilon ]`$ terms arrived at as in that can depend on f;and
\<\<O\[$`\varphi `$\]\>\><sub>P</sub>=$``$D$`\varphi `$O\[$`\varphi `$\]exp{ i *S<sub>eff</sub>$`{}_{}{}^{P}[`$*$`\varphi `$\]+ $`\epsilon O_P`$} (5.18)
From (5.18 ) and (5.17),
\<\<O\[$`\varphi `$\]\>\><sub>P</sub> $``$$`\frac{1}{N_P}`$$``$ Df exp{$`\frac{i}{2\lambda \eta ^2}`$$``$ d$`{}_{}{}^{4}x`$f $`^2`$f}\<\<O\[$`\varphi `$\]\>\><sub>f</sub> (5.19)
\[and the above relation in fact determines what $`\epsilon O_P`$ should be\].We note<sup>6</sup><sup>6</sup>6 For mathematical rigor,we may replace $`\sigma `$ $``$$`\frac{\sigma }{1iϵ^{}}`$ with $`ϵ^{}`$¿ 0 in what follows.We have used:$`\begin{array}{c}lim\\ \sigma 0\end{array}`$$`\sqrt{\frac{i+ϵ^{}}{2\pi \sigma ^2}}`$exp \[$`\frac{ix^2[1iϵ^{}]}{\sigma ^2}]`$ = $`\delta `$ (x). ,
\<\<O\[$`\varphi `$\]\>\><sub>f</sub> =$`\begin{array}{c}lim\\ \sigma 0\end{array}`$$``$D$`\varphi `$O\[$`\varphi `$\]exp{ i *S<sub>0</sub>*\[$`\varphi `$\]+iS$`{}_{G}{}^{_A}[\varphi `$\]-$`\frac{i}{2\sigma }`$ d$`{}_{}{}^{4}x`$ ($`\eta `$.A-f)<sup>2</sup> \+ $`\epsilon O_A(f,\sigma )`$}
=$`\begin{array}{c}lim\\ \sigma 0\end{array}`$\<\<O\[$`\varphi `$\]\>\><sub>f</sub> (5.20)
Now, \<\<O\[$`\varphi `$\]\>\><sub>f</sub> can be related to \<\<O\[$`\varphi `$\]\>\><sub>L,</sub> by the result (4.8).We thus have,
\<\<O\[$`\varphi `$\]\>\><sub>f</sub> =\<\<O\[$`\varphi `$\]\>\><sub>L,</sub><sub>σ</sub>
+i$`\frac{1}{N_P}`$$``$$`__0`$<sup>1</sup>d$`\kappa `$$``$D$`\varphi `$ exp{iS$`{}_{0}{}^{}+`$ i *S<sub>G</sub>$`{}_{}{}^{M}[`$*$`\varphi `$,$`\kappa ]`$ -$`\frac{i}{2\sigma }`$ d$`{}_{}{}^{4}x`$ F$`{}_{}{}^{M}{}_{1}{}^{}`$\[A\]<sup>2</sup>+$`\epsilon O_L`$}
$``$$`_i(\stackrel{~}{\delta _{1i}[\varphi }`$\]+$`\kappa `$$`(\stackrel{~}{\delta _{2i}[\varphi }`$\])(-i$`\mathrm{\Theta }__1^{})`$$`\frac{\delta ^LO}{\delta \varphi _i}`$ (5.21)
Here,on account of the results in Section 4,we have
F$`{}_{}{}^{M}{}_{1}{}^{}`$\[A\]=$`.A(1\kappa )+\kappa (\eta .Af)`$ (5.22)
and
$`\mathrm{\Theta }__1^{}=`$$`id^4y`$ $`\overline{c}`$<sup>γ</sup>(y) ($``$.A<sup>γ</sup>-$`\eta `$.A<sup>′γ</sup>+f<sup>γ</sup>) (5.23)
*S<sub>G</sub>$`{}_{}{}^{M}[`$*$`\varphi `$,$`\kappa ]`$=-$``$ d$`{}_{}{}^{4}x`$$`\overline{c}`$ \[M$`(1\kappa )+\kappa M^{}]c`$ (5.24)
We substitute (5.21) in (5.19) to obtain,
\<\<O\[$`\varphi `$\]\>\><sub>P</sub> $``$$`\begin{array}{c}lim\\ \sigma 0\end{array}`$$`\frac{1}{N_P}`$$``$ Df exp{$`\frac{i}{2\lambda \eta ^2}`$$``$ f $`^2`$f}{\<\<O\[$`\varphi `$\]\>\><sub>L,</sub> +i$``$$`__0`$<sup>1</sup>d$`\kappa `$ $``$D$`\varphi `$ exp{iS$`{}_{0}{}^{}+`$ i *S<sub>G</sub>$`{}_{}{}^{M}[`$*$`\varphi `$,$`\kappa ]`$ -$`\frac{i}{2\sigma }`$ d$`{}_{}{}^{4}x`$ F$`{}_{}{}^{M}{}_{1}{}^{}`$\[A\]<sup>2</sup>+$`\epsilon O_L`$$``$}$`_i(\stackrel{~}{\delta _{1i}[\varphi }`$\]+$`\kappa `$$`(\stackrel{~}{\delta _{2i}[\varphi }`$\])(-i$`\mathrm{\Theta }__1^{})`$$`\frac{\delta ^LO}{\delta \varphi _i}`$ (5.25)
As \<\<O\[$`\varphi `$\]\>\><sub>L,</sub> is independent of ’f’ and $``$ Df exp{$`\frac{i}{2\lambda \eta ^2}`$$``$ d$`{}_{}{}^{4}x`$f $`^2`$f} is absorbed in normalization,we arrive at
\<\<O\[$`\varphi `$\]\>\><sub>P</sub> = \<\<O\[$`\varphi `$\]\>\><sub>L,</sub><sub>σ</sub> <sub>0</sub> +$`\begin{array}{c}lim\\ \sigma 0\end{array}`$$``$$`__0`$<sup>1</sup>d$`\kappa `$ I($`\sigma ,\kappa )d\kappa `$ (5.26)
where
I($`\sigma ,\kappa )=i`$$`\frac{1}{N_P}`$$``$Df exp{$`\frac{i}{2\lambda \eta ^2}`$$``$d$`{}_{}{}^{4}x`$f$`^2`$f}$``$D$`\varphi `$ exp{iS$`{}_{0}{}^{}+`$i*S<sub>G</sub>$`{}_{}{}^{M}[`$*$`\varphi `$,$`\kappa ]`$-$`\frac{i}{2\sigma }`$d$`{}_{}{}^{4}x`$ \[$`.A(1\kappa )+\kappa (\eta .Af)`$\]<sup>2</sup>+$`\epsilon O_L`$$``$}$`_i(\stackrel{~}{\delta _{1i}[\varphi }`$\]+$`\kappa `$$`(\stackrel{~}{\delta _{2i}[\varphi }`$\])(-i$`\mathrm{\Theta }__1^{})`$$`\frac{\delta ^LO}{\delta \varphi _i}`$ (5.27)
Now, we take the limit $`\sigma `$0 inside to obtain
exp{-$`\frac{i}{2\sigma }`$ d$`{}_{}{}^{4}x`$ F$`{}_{}{}^{M}{}_{1}{}^{}`$\[A\]<sup>2</sup>}$``$ $`\begin{array}{c}\mathrm{\Pi }\\ \alpha ,x\end{array}`$$`\delta `${$`.A(1\kappa )+\kappa (\eta .Af)`$} (5.28)
We further note that
D$`\overline{c}`$$`\begin{array}{c}\mathrm{\Pi }\\ \alpha ,x\end{array}`$$`\delta `${$`.A(1\kappa )+\kappa (\eta .Af)`$}$``$$`\begin{array}{c}\mathrm{\Pi }\\ \alpha ,x\end{array}`$d$`\overline{c}`$<sup>α</sup>(x)$`\delta `${$`.A(1\kappa )+\kappa (\eta .Af)`$}
= $`\begin{array}{c}\mathrm{\Pi }\\ \alpha ,x\end{array}`$d$`\underset{}{\overline{c}^\alpha }`$(x)$`\delta `${$`.A(1/\kappa 1)+(\eta .Af)`$}
$``$D$`\overline{\underset{}{c}}`$$`\begin{array}{c}\mathrm{\Pi }\\ \alpha ,x\end{array}`$$`\delta `${$`.A(1/\kappa 1)+(\eta .Af)`$} (5.29)
with $`\overline{\underset{}{c}}`$=$`\overline{c}`$$`\kappa `$. We use this $`\delta `$-function to simplify
$`\mathrm{\Theta }__1^{}=`$$`id^4y`$ $`\overline{c}`$<sup>γ</sup>(y) ($``$$`_{^{..A^\gamma }}`$)/$`\kappa `$ = $`id^4y`$ $`\overline{\underset{}{c}}`$<sup>γ</sup>(y) ($``$.A<sup>γ</sup>)/$`\kappa `$<sup>2</sup> (5.30)
and
exp{$`\frac{i}{2\lambda \eta ^2}`$$``$d$`{}_{}{}^{4}x`$f$`^2`$f}–\>exp{$`\frac{i}{2\lambda \eta ^2}`$$``$d$`{}_{}{}^{4}x`$\[$`.A(1/\kappa 1)+\eta .A`$\]$`^2`$\[$`.A(1/\kappa 1)+\eta .A`$\]
$``$exp{i$`\underset{}{S}`$<sub>gf</sub>($`\kappa )\}`$ (5.31)
We further re-express *S<sub>G</sub><sup>M</sup> as*
*S<sub>G</sub><sup>M</sup>=*-$``$ d$`{}_{}{}^{4}x`$ $`\overline{\underset{}{c}}`$\[M(1/$`\kappa `$-1)$`+M^{}]c`$ $``$ $`\underset{}{S}`$*<sub>G</sub><sup>M</sup>* (5.32)
Thus, we obtain the result that connects arbitrary Green’s functions in planar gauges to those in Landau gauge:
\<\<O\[$`\varphi `$\]\>\><sub>P</sub> =\<\<O\[$`\varphi `$\]\>\><sub>L,</sub><sub>σ</sub><sub>0</sub>+
i$`\frac{1}{N_P}`$$``$$`__0`$<sup>1</sup>d$`\kappa `$$``$D$`\underset{}{\varphi }`$exp{iS$`{}_{0}{}^{}+`$i$`\underset{}{S}`$*<sub>G</sub>$`{}_{}{}^{M}[`$*$`\varphi `$,$`\kappa ]`$+i$`\underset{}{S}`$<sub>gf</sub>($`\kappa )\}`$+$`\epsilon O_L`$}
$``$$`_i(\stackrel{~}{\delta _{1i}[\varphi }`$\]+$`\kappa `$$`(\stackrel{~}{\delta _{2i}[\varphi }`$\])(-i$`\mathrm{\Theta }__1^{})`$$`\frac{\delta ^LO}{\delta \varphi _i}`$ (5.33)
We may change the notation for the integration variable $`\overline{\underset{}{c}}`$ $``$ $`\overline{c}`$.Noting that
$`_i(\stackrel{~}{\delta _{1i}[\varphi }`$\]+$`\kappa `$$`(\stackrel{~}{\delta _{2i}[\varphi }`$\] does not depend on this variable,we can express this as
\<\<O\[$`\varphi `$\]\>\><sub>P</sub> =\<\<O\[$`\varphi `$\]\>\><sub>landau</sub>+i$`\frac{1}{N_P}`$$``$$`__0`$<sup>1</sup>$`\frac{d\kappa }{\kappa ^2}`$ $``$D$`\varphi `$ exp{i$`\underset{}{S}`$$`{}_{eff}{}^{}{}_{}{}^{M}`$+$`\epsilon [\frac{1}{2}`$A<sub>μ</sub>A<sup>μ</sup>\- i
$`\overline{c}`$c \]}$`_i(\stackrel{~}{\delta _{1i}[\varphi }`$\]+$`\kappa `$$`(\stackrel{~}{\delta _{2i}[\varphi }`$\])\[$`d^4y`$ $`\overline{c}`$<sup>γ</sup>(y) ($``$.A<sup>γ</sup>)\]$`\frac{\delta ^LO[A,c,\kappa \overline{c}]}{\delta \underset{}{\varphi _i}}`$ (5.34)
with $`\underset{}{\varphi _i}`$=A,c,$`\kappa \overline{c}`$;and here
$`\underset{}{S}`$$`{}_{eff}{}^{}{}_{}{}^{M}`$ = S<sub>0</sub>$`\frac{1}{2\lambda \eta ^2}`$$``$ d$`{}_{}{}^{4}x`$\[$`.A(1/\kappa 1)+\eta .A`$\]$`^2`$\[$`.A(1/\kappa 1)+\eta .A`$\]-$``$ d$`{}_{}{}^{4}x`$ $`\overline{\underset{}{c}}`$\[M(1/$`\kappa `$-1)$`+M^{}]c`$ (5.35)
is the effective action for a mixed planar gauge but with a mixed gauge fixing function: \[$`.A(1/\kappa 1)+\eta .A`$\] in both the gauge-fixing and the ghost term.We put O\[$`\varphi `$\]=I, the identity operator, in (5.34) to obtain,
\<\< I \>\><sub>P</sub> =\<\< I \>\><sub>landau</sub> (5.36)
We then obtain for the connected part \<O\[$`\varphi `$\]\><sub>P</sub> =\<\<O\[$`\varphi `$\]\>\><sub>P</sub>/\<\< I \>\><sub>P</sub> ,etc the final result<sup>7</sup><sup>7</sup>7 We need not be alarmed by $`\kappa ^2`$ in the denominator : near $`\kappa `$=0, the ghost propagator and the longitudinal gauge propagator yield enough factors of $`\kappa `$. :
\<O\[$`\varphi `$\]\><sub>P</sub> =\<O\[$`\varphi `$\]\><sub>landau</sub>+i$``$$`__0`$<sup>1</sup>$`\frac{d\kappa }{\kappa ^2}`$ $``$D$`\varphi `$ exp{i$`\underset{}{S}`$$`{}_{eff}{}^{}{}_{}{}^{M}`$+$`\epsilon [\frac{1}{2}`$A<sub>μ</sub>A<sup>μ</sup>\- i$`\overline{c}`$c \]}
$``$$`_i(\stackrel{~}{\delta _{1i}[\varphi }`$\]+$`\kappa `$$`(\stackrel{~}{\delta _{2i}[\varphi }`$\])\[$`d^4y`$ $`\overline{c}`$<sup>γ</sup>(y) ($``$.A<sup>γ</sup>)\]$`\frac{\delta ^LO[A,c,\kappa \overline{c}]}{\delta \underset{}{\varphi _i}}`$$``$$`_{_{conn}}`$
= \<O\[$`\varphi `$\]\><sub>landau</sub>
+i$``$$`__0`$<sup>1</sup>$`\frac{d\kappa }{\kappa ^2}`$\<$`_i(\stackrel{~}{\delta _{1i}[\varphi }`$\]+$`\kappa `$$`(\stackrel{~}{\delta _{2i}[\varphi }`$\])\[$`d^4y`$$`\overline{c}`$<sup>γ</sup>(y) ($``$.A<sup>γ</sup>)\]$`\frac{\delta ^LO[A,c,\kappa \overline{c}]}{\delta \underset{}{\varphi _i}}`$\>$`_{mixed}`$$`_{_{,conn}}`$ (5.37)
We note here that in the last term, we have the connected Green’s functions in the mixed gauge with the appropriate $`ϵ`$-term.
## 6 An Alternate and procedure for Planar Gauges
### In this section,we shall present an alternate ,though somewhat heuristic, procedure that uses directly the familiar results of the section 3. This procedure applies directly to the vacuum expectation values of gauge-invariant observables, if not for their arbitrary Green’s functions. The conclusions drawn here are however easily *directly verified* *by an algebra similar to that in Section 3*; and without any need for the lack of rigor in the following derivation.We present the former way here.
### To use the results of section 3, we first establish an interpolating route. We define
#### $`\stackrel{~}{F}`$<sup>M</sup>\[A\]=$`.A(1\kappa )+\kappa \eta .A`$ (6.1)
#### Then with
S<sup>M</sup><sub>gf</sub> = -$`\frac{1}{2\lambda \eta ^2}`$ d$`{}_{}{}^{4}x`$ $`\stackrel{~}{F}`$<sup>M</sup>\[A\]$`^2`$ $`\stackrel{~}{F}`$<sup>M</sup>\[A\] (6.2)
and
*S<sup>M</sup><sub>gh</sub>= -*$``$d$`{}_{}{}^{4}x`$ $`\overline{c}`$ $`^2`$\[M$`(1\kappa )+\kappa M^{}]c`$ (6.3)
and the BRS transformation
$`\delta \overline{c}`$= {$`\stackrel{~}{F^M/}`$$`\lambda \eta ^2\}`$ $`\delta \mathrm{\Lambda }`$ (6.4)
we have the interpolating route from the pseudo-Lorentz gauges ($`\kappa =0)`$ to the planar gauges$`(\kappa =1)`$.Now, for the purpose of treating the extra singularity introduced in the propagator on account of the $`^2`$ in (6.2), we interpret it as $`^2`$-i$`ϵ`$ both there and in (6.3)\[as pointed out earlier in (5.1)\].In the momentum space,we write the equation (6.2) as,
S<sup>M</sup><sub>gf</sub> = $`\frac{1}{2\lambda \eta ^2}`$ d$`{}_{}{}^{4}k`$ $`\stackrel{~}{F}`$<sup>M</sup>\[k\]\[k<sup>2</sup>+i$`ϵ]`$ $`\stackrel{~}{F}`$$`{}_{}{}^{M}[k]`$
= - $`\frac{1}{2\lambda }`$ d$`{}_{}{}^{4}k`$ {$`\stackrel{~}{F}`$<sup>M</sup>\[k\]$`\sqrt{\frac{k^2+iϵ}{\eta ^2}}`$}<sup>2</sup> (6.5)
where $`\stackrel{~}{F}^M`$\[k\] is the Fourier transform of $`\stackrel{~}{F}`$<sup>M</sup>\[A\].We define
F.T.{$`\stackrel{~}{F}^M`$\[k\]$`\sqrt{\frac{k^2+iϵ}{\eta ^2}}`$}$``$$`\sqrt{\frac{^2iϵ}{\eta ^2}}`$$`\stackrel{~}{F}`$<sup>M</sup>\[A\]$``$F$`{}_{M}{}^{}[A]`$ (6.6)
We then have,
S<sup>M</sup><sub>gf</sub> = -$`\frac{1}{2\lambda }`$F$`{}_{M}{}^{}[A]`$<sup>2</sup>d$`{}_{}{}^{4}x`$ (6.7)
and the ghost term as
*S<sup>M</sup><sub>gh</sub>= -*$``$d$`{}_{}{}^{4}x`$ $`\overline{c}`$ $`\sqrt{}`$$`\{[^2`$-i$`ϵ`$ $`]\eta ^2\}`$$`\frac{\delta F_M}{\delta A_\mu }D_\mu `$c
*-*$``$d$`{}_{}{}^{4}x`$ $`\sqrt{}`$$`\{[^2`$-i$`ϵ`$ $`]\eta ^2\}`$$`\overline{c}`$ $`\frac{\delta F_M}{\delta A_\mu }D_\mu `$c (6.8)
The last step is seen by going to the Fourier space and comparing the expressions.We now make a linear change of variables, leading to a constant Jacobian that can be ignored,
$`\overline{c}`$$`{}_{}{}^{}=\sqrt{}`$$`\{[^2`$-i$`ϵ`$ $`]\eta ^2\}`$$`\overline{c}`$ (6.9)
Then
*S<sup>M</sup><sub>gh</sub>= -*$``$d$`{}_{}{}^{4}x`$ $`\overline{c}`$$`\frac{\delta F_M}{\delta A_\mu }D_\mu `$c (6.10)
Now the system of gauge fixing term (6.7) and the ghost term (6.10) has been cast in the standard FP form.We can now apply the results of the Section 3 to interpolate between the planar and the pseudo-Lorentz gauges with
$`\mathrm{\Theta }`$’\[$`\varphi `$\] =-$`id^4y`$ $`\overline{c}`$<sup>γ</sup>(y) \[F<sub>M</sub>($`\kappa =1)`$\- F<sub>M</sub>($`\kappa =0)`$\]<sup>γ</sup>
=-$`id^4y`$ $`\overline{c}`$<sup>γ</sup>(y)$`\sqrt{\frac{^2iϵ}{\eta ^2}}`$\[$`\eta .A.A]`$<sup>γ</sup>
= -$`id^4y`$ $`\overline{c}`$<sup>γ</sup>(y)$`[^2iϵ]`$\[$`\eta .A.A]`$<sup>γ</sup> (6.11)
Thus, under the field transformation
$`\varphi `$<sub>i</sub> =$`\varphi `$<sub>i</sub> +$`\delta _{iBRS}^L[\varphi ]\mathrm{\Theta }`$ \[$`\varphi ]`$ (6.12)
with $`\mathrm{\Theta }`$ given in terms of $`\mathrm{\Theta }`$’\[$`\varphi `$\] of (6.11)through relation in section 2, relates the planar gauge to the pseudo-Lorentz gauge. Further as far as the expectation values of gauge-invariant observables are concerned,they have been shown to have the same value in the pseudo-Lorentz and the Lorentz gauges in Section 5.
Of course,as mentioned at the beginning of this of this section,the above heuristic argument that allows one to make a direct use of the results of Section 3, can be avoided and we can, with additional labor, verify the result for the field transformation given by (6.12) directly from the Jacobian condition (2.14).We have in fact carried out this verification.
|
warning/0005/cond-mat0005482.html
|
ar5iv
|
text
|
# Complex orbital state in manganites
## I Introduction
The role of the orbital degrees of freedom has recently attracted considerable interests as one of the key to understand the colossal magneto-resistance (CMR) observed in doped manganites. The orbital state of the conduction electrons is described as a linear combination of two wavefunctions, $`|x^2y^2`$ and $`|3z^2r^2`$, of the degenerate $`e_g`$ orbitals. In previous studies, the linear combination with only real coefficients (real orbital state) has been considered. This is because theories of the orbital ordering have been developed mainly to describe the parent compounds of CMR materials, in which the static Jahn-Teller deformation is observed. Such a deformation stabilizes the real orbital state and it was reasonable to exclude the linear combination with complex coefficients (complex orbital state). Recently the orbital state in doped compounds is studied concerning the properties of CMR materials. Because the static Jahn-Teller distortion disappears in doped compounds, there is no reason to exclude the complex orbital state. Actually such a complex orbital state has been recently studied. Khomskii pointed out that the complex orbital state, $`\left(|x^2y^2\pm i|3z^2r^2\right)\mathrm{/}\sqrt{2}`$, provides locally isotropic hopping intensities with the same bandwidth as the real orbital state, and might explain the isotropic properties observed in CMR compounds. Such a local isotropy cannot be realized with the uniform real orbital state. A staggered ordering is therefore needed to explain the observed isotropic properties within the extent of the real orbital ordering (Another proposal is the orbital liquid state, where the local isotropy is recovered by a quantum resonance between anisotropic orbital configurations, $`|x^2y^2`$, $`|y^2z^2`$, and $`|z^2x^2`$ ). Based on the analogy to the Nagaoka ferromagnetism ($`F`$), Khomskii proposed that the uniform ordering (orbital $`F`$) of the complex orbital state is more stable than the staggered one (orbital $`AF`$) with real orbitals. Takahashi et al. investigated the possible complex orbital ordering, motivated by the analogy to the octapole ordering in heavy fermion systems with odd time reversal symmetry. They found that the staggered ordering of the complex orbital is stable, being contrary to Khomskii’s uniform one.
In this paper, we study the competitions among the uniform complex, staggered complex, and real orbital states by using a model of CMR compounds taking the strong on-site repulsion and the orbital degeneracy into account. The complex orbital state is taken as, $`\mathrm{cos}\left(\theta /2\right)|x^2y^2+i\mathrm{sin}\left(\theta /2\right)|3z^2r^2`$, and the whole possibility with the continuous parameter $`\theta `$ is examined. With realistic parameters, the complex orbital state is more stable than the real one in the moderately doped region ($`0.25<x<0.45`$). The complex ordering changes from the canted one ($`0.25<x<0.35`$) into the staggered ($`0.35<x<0.45`$) one due to the competition between the orbital superexchange $`AF`$ and the orbital Nagaoka $`F`$. The local isotropy is also realized in this complex staggered phase, where the band gap due to the doubled period brings about the energy gain exceeding the energy loss due to the narrower bandwidth than that of the uniform ordering with isotropic hopping. With increasing $`U/t`$ toward the strong correlation limit, the former gain decreases whereas the latter loss increases. The staggered ordering becomes unstable in this limit, where the uniform orbital ordering wins. In this case, however, the obtained uniform ordering is not the complex one but the real one with $`|x^2y^2`$. Though the uniform complex ordering becomes more stable than the staggered complex one, it has higher energy than that of the real one. In the weak correlation limit, on the other hand, Takahashi $`etal.`$ found that the normal metallic state becomes unstable toward the the staggered complex ordering near the quarter filling, with increasing $`U`$. When the Jahn-Teller coupling is further taken into account, however, it is likely that the real orbital state is stabilized, because the energy scale of the Jahn-Teller coupling becomes dominating compared with the weak $`U`$, prefering the real state.
The Jahn-Teller deformation which couples with the orbital degrees of freedom decreases in the complex canted phase and vanishes in the complex staggered phase, being consistent with the observed disappearance of the deformation. This complex state can be expressed as a resonating state among planer orbitals, as in the orbital liquid picture. When the resonance occurs with coherent correlations in time and space, the complex orbital ordering is obtained, meanwhile that with incoherent one corresponds to the orbital liquid state. These can be distinguished by experiments detecting the spatial correlation of the orbital symmetry, such as the anomalous X-ray scattering experiments. Possibilities of the phase separation and broken time-reversal symmetry are also discussed.
## II Results and discussions
We employ the same model as that in the previous report ,
$`H`$ $`=`$ $`{\displaystyle \underset{\sigma \gamma \gamma ^{}ij}{}}t_{ij}^{\gamma \gamma ^{}}d_{i\sigma \gamma }^{}d_{j\sigma \gamma ^{}}`$ (1)
$``$ $`J_H{\displaystyle \underset{i}{}}\stackrel{}{S}_{t_{2g}i}\stackrel{}{S}_{e_gi}`$ (2)
$`+`$ $`J_S{\displaystyle \underset{ij}{}}\stackrel{}{S}_{t_{2g}i}\stackrel{}{S}_{t_{2g}j}+H_{\mathrm{on}\mathrm{site}},`$ (3)
where $`\gamma `$ \[$`=a(d_{x^2y^2}),b(d_{3z^2r^2})`$\] specifies the orbital and the other notations are standard. The transfer integral $`t_{ij}^{\gamma \gamma ^{}}`$ depends on the pair of orbitals $`(\gamma ,\gamma ^{})`$ and the direction of the bond $`(i,j)`$. The spin operator for the $`e_g`$ electron is defined as $`\stackrel{}{S}_{e_gi}=\frac{1}{2}\underset{\gamma \alpha \beta }{}d_{i\gamma \alpha }^{}\stackrel{}{\sigma }_{\alpha \beta }d_{i\gamma \beta }`$ with the Pauli matrices $`\stackrel{}{\sigma }`$, while the orbital isospin operator is defined as $`\stackrel{}{T}_i=\frac{1}{2}\underset{\gamma \gamma ^{}\sigma }{}d_{i\gamma \sigma }^{}\stackrel{}{\sigma }_{\gamma \gamma ^{}}d_{i\gamma ^{}\sigma }`$. $`J_H`$ is the Hund’s coupling between $`e_g`$ and $`t_{2g}`$ spins, and $`J_S`$ is the $`AF`$ coupling between nearest neighboring $`t_{2g}`$ spins. $`H_{\mathrm{on}\mathrm{site}}`$ represents the on-site Coulomb interactions between $`e_g`$ electrons. Coulomb interactions induce both the spin and orbital isospin moments, and actually $`H_{\mathrm{on}\mathrm{site}}`$ can be written as
$`H_{\mathrm{on}\mathrm{site}}`$ $`=`$ $`{\displaystyle \underset{i}{}}\left(\stackrel{~}{\beta }\stackrel{}{T}_i^2+\stackrel{~}{\alpha }\stackrel{}{S}_{e_gi}^2\right).`$ (4)
A parameter set with $`t_0=t_{i,i+\widehat{z}}^{bb}=0.72`$ eV, $`\stackrel{~}{\alpha }=8.1t_0`$, and $`\stackrel{~}{\beta }=6.7t_0`$ corresponds to the realistic one being relevant to the actual manganese oxides. In the path-integral quantization, we introduce the Stratonovich-Hubbard fields $`\stackrel{}{\phi }_S`$ and $`\stackrel{}{\phi }_T`$, representing the spin and orbital fluctuations, respectively. With the large values of the electron-electron interactions above, both $`\stackrel{}{\phi }_S`$ and $`\stackrel{}{\phi }_T`$ are almost fully polarized. The meanfield theory corresponds to the saddle point configuration of $`\stackrel{}{\phi }_S`$ and $`\stackrel{}{\phi }_T`$. We only consider the possibility of the complex orbital state within a $`F`$-type spin alignment in the cubic cell.
We assume the two sublattices, $`I`$ and $`II`$, with $`F`$-, $`A`$-, $`C`$\- and $`G`$-type alignment. On each site, the orbital is specified as a linear combination of the two degenerate orbital bases, $`|x^2y^2`$ and $`|3z^2r^2`$, as
$$|\theta ,\phi =\mathrm{cos}\frac{\theta }{2}|x^2y^2+e^{i\phi }\mathrm{sin}\frac{\theta }{2}|3z^2r^2.$$
(5)
$`(\theta ,\phi )`$ is the polar angle of the corresponding isospin $`\stackrel{}{T}`$. In the limit of the infinite orbital polarization, $`\stackrel{~}{\beta }\mathrm{}`$, the uniform orbital ordering with any $`|\theta ,\phi `$ takes the same bandwidth, $`\left(3\mathrm{/}2\right)t_0`$. The polar angle $`(\theta ,\phi )`$ therefore controls only the dimensionality of the band structure to optimize the kinetic energy gain, leaving the bandwidth unchanged.
Previous studies have focused on the states with $`\stackrel{}{T}//\widehat{e}_y`$ as the complex orbital states. In this paper, we extend the possibility to $`\stackrel{}{T}`$ lying within $`yz`$ plane ($`|\theta ,\phi =\pi /2`$, real and pure imaginary coefficients) for the complex orbital state, whereas $`\stackrel{}{T}`$ within $`zx`$ plane ($`|\theta ,\phi =0`$) corresponds to the real orbital state. This choice includes $`|x^2y^2`$ and $`|3z^2r^2`$ as the both ends. With finite $`\stackrel{~}{\beta }`$, the generalized orbital canted structure on two sublattices is examined.
Fig. 1 shows the energy values in spin $`F`$ phase, optimized within the real and the complex orbital states, plotted as a function of the hole concentration $`x`$ (with $`J_S=0`$).
The orbital shape specified by $`\theta `$ is optimized at each $`x`$. The complex orbital state is realized in the moderately doped region ($`0.25<x<0.45`$). The phase diagram as a function of $`x`$ and $`J_S`$ ($`AF`$ interaction between $`t_{2g}`$ spins) is shown in Fig. 2.
In the shaded and hatched regions of the spin $`F`$ phase is realized the complex orbital state. The phase boundary depicted with a broken line is that for the real orbital state, reported previously.
Figure 3 shows the orbital phase diagram assuming the spin $`F`$ phase as a function of $`x`$. The orbital ordering changes from the real staggered, the complex canted, the complex staggered, and to the real uniform one. Figure 4 shows the $`x`$-dependence of the orbital canting angle, $`\left|\theta _{II}\theta _I\right|`$, for the real and the complex orbital states.
With increasing $`x`$, the canting angle once tends to take the orbital $`F`$ ($`III`$), but get back to $`AF`$ again ($`IIIII`$). The canted complex state with orbital $`C`$ is stable at $`x=0.25`$ and $`0.35`$. With increasing $`x`$, the canted state changes into the staggered one for $`0.35<x<0.45`$ with $`\theta _I=\theta _{II}=\pi /2`$ (orbital $`G`$) as found in ref. 22. We note that this $`staggered`$ state also gives the locally isotropic hopping integrals, $`t^x=\frac{1}{2}e^{i\frac{2\pi }{3}}`$, $`t^y=\frac{1}{2}e^{i\frac{2\pi }{3}}`$, and $`t^z=\frac{1}{2}`$. The uniform complex state with $`\theta _I=\theta _{II}=\pi /2`$ and $`t^{x,y,z}=1/2`$ has higher energy. With further doping ($`IV`$ with $`x>0.5`$), the orbital $`F`$ becomes stable again, but with real coefficients.
These results can be understood as follows. The orbital superexchange $`AF`$ interaction $`J_{\mathrm{AF}}`$ is represented by the shift in the center of mass of the occupied density of states (DOS), as represented by the Hamiltonian,
$$\left(\begin{array}{cc}\epsilon _k& \beta _{\mathrm{eff}}\\ \beta _{\mathrm{eff}}& \epsilon _{k+Q}\end{array}\right),$$
(6)
with $`Q=(\pi ,\pi ,\pi )`$ being the staggered orbital momentum. Therefore $`J_{\mathrm{AF}}`$ is estimated as
$$J_{\mathrm{AF}}\frac{t^2}{\beta _{\mathrm{eff}}}\frac{t^2}{\stackrel{~}{\beta }\left(1x\right)},$$
(7)
which increases as $`x`$ increases because $`\beta _{\mathrm{eff}}`$ is the constant $`\stackrel{~}{\beta }`$ times the number of electrons $`\left(1x\right)`$. The ferromagnetic double exchange interaction $`J_\mathrm{F}`$ for the orbital moments is represented by the energy of the doped holes at the top of the occupied band, which depends on the bandwidth. The bandwidth is $`t`$ for the uniform ordering whereas $`t^2/\beta _{\mathrm{eff}}`$ for the staggered one for $`t\beta _{\mathrm{eff}}`$ and small $`x`$. $`J_\mathrm{F}`$ is therefore given as,
$$J_\mathrm{F}\left(t\frac{t^2}{\beta _{\mathrm{eff}}}\right)x,$$
(8)
which represents the relative kinetic energy gain of the orbital $`F`$ state measuring from that of the staggered state. It should be noted here that the notation $`J_\mathrm{F}`$ is rather symbolic, and the Hamiltonian is not written as $`J_\mathrm{F}_{ij}\stackrel{}{T}_i\stackrel{}{T}_j`$. Based on these considerations, the results in Fig. 2 and 3 are interpreted as follows. Here we assume the ferromagnetic spin alignment. At $`x=0`$, $`J_\mathrm{F}`$ vanishes meanwhile $`J_{\mathrm{AF}}`$ is finite, leading to the orbital $`AF`$. With small doping, $`J_\mathrm{F}`$ becomes finite, leading to the tendency toward the orbital $`F`$ seen in the region $`II`$ in Fig. 4. (This corresponds to the crossover from the orbital superexchange $`AF`$ to the orbital double exchange (Nagaoka) $`F`$ with the doping.) To understand the reentrant of the orbital $`AF`$ in the region $`III`$, we note that $`t^2/\beta _{\mathrm{eff}}=t^2/\stackrel{~}{\beta }\left(1x\right)`$ increases as $`x`$, which enhances $`J_{\mathrm{AF}}\left(x\right)`$ and suppresses $`J_\mathrm{F}`$. Actually, the difference in the bandwidth of DOS between the uniform (orb. $`F`$) and the staggered (orb. $`AF`$) structures is hardly seen at $`x=0.3`$ in Fig. 5, corresponding to $`J_\mathrm{F}0`$.
The staggered ordering is therefore stabilized with increasing $`J_{\mathrm{AF}}\left(x\right)`$ in the moderately doped region. In the heavily doped region ($`IV`$), the expression of $`J_{\mathrm{F},\mathrm{AF}}`$ does not hold any more because $`t\beta _{\mathrm{eff}}`$. There the staggered ordering is unstable due to the lower bandwidth than that of the uniform one, leading to the orbital $`F`$ ordering.
The competition between the real and the complex orbital states is understood as follows. The complex state is stabilized only in the moderately doped region with the advantage of the isotropic band structure. In the other region, some other mechanism is rather important than the isotropy: In the small doping region, the real state realized in Fig. 1 and 3 ($`x<0.25`$) is found to be stabilized mainly due to the hybridization between the occupied and the unoccupied bands via the off-diagonal hopping integrals. In the heavily doped region, on the other hand, the low dimensional band structure, $`\theta _I=\theta _{II}=0`$ (two dimensional) or $`\pi `$ (quasi-one dimensional), is prefered where the isospin moment is along the $`z`$ axis (real orbital state). This is due to the relative location between the van Hove singularity of DOS and the fermi level. The fermi level with small electron concentration ($`x1`$, heavily doped region) is located near the band edge. The low dimensional band structure with singularities at the top and the bottom of the band can therefore lower the kinetic energy effectively with large accomodation at the singularity near the fermi level.
With increasing $`\stackrel{~}{\beta }/t`$, the staggered state becomes unstable because $`J_{\mathrm{AF}}`$ goes to zero whereas $`J_\mathrm{F}`$ remains to be finite. This corresponds to the recovery of the orbital Nagaoka $`F`$. One can therefore expect the uniform complex state in the small doped region with the strong correlation limit. The obtained ordering is however the real uniform one with $`|x^2y^2`$, not the complex one. The schematic phase diagram in this limit is given in Fig. 6.
This can be understood as follows. In this limit, only the DOS near the band edge matters because the bandwidth of the uniform state does not depend on the orbital shape. The DOS arises at the edge most sharply with $`|x^2y^2`$, giving the largest kinetic energy gain. Orbital Nagaoka $`F`$ is therefore realized with $`|x^2y^2`$ in the strong correlation limit. (It should not be confused with the real state in the small doping region with realistic parameters (Fig. 3), where the ordering is orbital $`AF`$ stabilized due to the inter-band hybridization.)
Curves obtained in Fig. 1 are non-monotonic with a common tangential line contacting at two different $`x`$, where the curve is convex upwards. This means that the phase coexistence with two different concentrations has higher energy than the single phase. Therefore a spontaneous phase separation does not occur in our results.
The $`e_g`$ state specified with the isospin orientation $`\stackrel{}{T}`$ stabilizes the Jahn-Teller (JT) deformation expressed as $`\left[\stackrel{}{T}_zQ_u+\stackrel{}{T}_xQ_v\right]`$, where $`Q_u`$ and $`Q_v`$ denote the normal coordinates of the displacement of the oxygen ions $`\mathrm{\Delta }_\alpha `$ ($`\alpha =x,y,z`$):
$`Q_u={\displaystyle \frac{2\mathrm{\Delta }_z\mathrm{\Delta }_x\mathrm{\Delta }_y}{\sqrt{6}}},Q_v={\displaystyle \frac{\mathrm{\Delta }_x\mathrm{\Delta }_y}{\sqrt{2}}}.`$ (9)
The complex state realized with $`0.35<x<0.45`$ corresponds to $`\stackrel{}{T}//\widehat{e}_y`$. With this ordering, therefore, the JT distortion does not occur. The observed disappearance of the JT distortion in the spin $`F`$ metallic region might be explained by this type of the orbital ordering. Another theoretical proposal is the orbital liquid state where the planer orbitals, $`|x^2y^2`$, $`|y^2z^2`$, and $`|z^2x^2`$ are resonating to form a quantum liquid state. With this resonance, the local isotropy is recovered, and the JT distortion disappears on average. The complex state can actually be expressed in the form of such a resonance as,
$$\frac{1}{\sqrt{2}}|x^2y^2\pm i\frac{1}{\sqrt{2}}|3z^2r^2=\frac{\sqrt{2}}{3}\left[|x^2y^2+e^{\pm i2\pi \mathrm{/}3}|z^2x^2+e^{i2\pi \mathrm{/}3}|y^2z^2\right].$$
(10)
This can be regarded as a formation of the $`T_y`$-component by the resonance via the transverse components $`T^\pm =T_x+iT_y`$. From Eq. (10), the complex orbital ordering corresponds to the coherent (in time and space) resonance among the planer orbitals, with the relative phases $`e^{\pm 2\pi /3}`$ being fixed. The orbital liquid state, on the other hand, corresponds to the resonance without phase coherence. In this sense, the complex orbital state obtained here is the meanfield state akin to the orbital liquid state, and may provide a rough estimation of the energy of the latter state. Because both states give no JT distortion on average, $`direct`$ observations of the orbital state is needed to distinguish them, not via the lattice deformation, but via the difference of the spatial orbital correlations. Several probes are available, such as the anomalous X-ray scattering, the X-ray charge density study by using of the maximum entropy method (MEM), the magnetic Compton scattering, and the polarized neutron scattering .
In summary, we studied the competitions among the uniform complex, staggered complex, and real orbital states in CMR compounds. In the moderately doped region, the complex orbital state is stabilized, where the ordering changes from the canted one ($`x=0.25`$, $`0.3`$) to the staggered one ($`0.35<x<0.45`$). This can be understood in terms of the competition among the band narrowing and the gap associating with the staggered structure, and the hybridization with the unoccupied band. The staggered complex ordering is not accompanied with the Jahn-Teller deformation. The obtained complex orbital is a coherent resonance among the planer orbitals with constant phase angles. If the coherency is lost, the state reduces to the orbital liquid state which can be distinguished by the observation of the spatial orbital correlations. The phase separation does not occur with the complex orbital state obtained here.
The authors would like to thank D. Khomskii, Y. Tokura for their valuable discussions. This work was supported by Priority Areas Grants from the Ministry of Education, Science and Culture of Japan. R.M. is supported by Research Fellowship of the Japan Society for the Promotion of Science (JSPS) for Young Scientists.
|
warning/0005/math0005040.html
|
ar5iv
|
text
|
# Tangent scrolls in prime Fano threefolds
## Abstract
In this paper we prove that any smooth prime Fano threefold, different from the Mukai-Umemura threefold $`X_{22}^{}`$, contains a 1-dimensional family of intersecting lines. Combined with a result in \[Sch\] this implies that any morphism from a smooth Fano threefold of index 2 to a smooth Fano threefold of index 1 must be constant, which gives an answer in dimension 3 to a question stated by Peternell.
§ 1. Introduction
(1.1) A smooth projective variety $`X`$ is called a Fano variety if the anticanonical bundle $`K_X`$ is ample. Then the index of $`X`$ is the largest positive integer $`r=r(X)`$ such that $`K_X=rH`$ for some line bundle $`H`$ on $`X`$.
The smooth Fano threefold $`X=X_d𝐏^{g+1}`$ ($`d=degX`$) is called prime if $`\rho (X)=rank\mathrm{𝐏𝐢𝐜}(X)=1`$, $`r(X)=1`$, and $`K_X`$ is the hyperplane bundle on $`X`$. By the classification of Fano threefolds smooth prime Fano threefolds exist iff $`3g12(g11)`$, and then $`d=2g2`$ (see \[I1\]).
(1.2) (see §4.2, §4.4 in \[IP\], or §1 in \[I2\]). Let $`l`$ be a line on the smooth prime Fano threefold $`X`$, and let $`N_{l/X}`$ be the normal bundle of $`lX`$. Then
(1). either (a). $`N_{l/X}=𝒪𝒪(1)`$; or (b). $`N_{l/X}=𝒪(1)𝒪(2)`$.
(2). The Hilbert scheme $`_X`$ of lines on $`X`$ is non-empty, any irreducible component $`_o`$ of $`_X`$ is one-dimensional, and either $`_o`$ is non-exotic, i.e. $`N_{l/X}`$ is of type (1)(a) for the general $`l_o`$; or $`_o`$ is exotic, i.e. $`N_{l/X}`$ is of type (1)(b) for any $`l_o`$.
(3). The component $`_o`$ is exotic if either the elements $`l_o`$ sweep out the tangent scroll $`R_oX`$ to an irreducible curve $`CX`$; or $`g=3`$ (i.e. $`X=X_4`$ is a quartic threefold), and then the lines $`l_o`$ sweep out a hyperplane section $`R_oX_4`$ which is a cone over a plane quartic curve, centered at some point $`xX_4`$.
(1.3) For example, the scheme $`_X`$ of the Fermat quartic $`X=X_4=(x_0^4+\mathrm{}+x_4^4=0)`$, which is a prime Fano threefold of $`g=3`$, is a union of $`40`$ double components each of which is of type (1.2)(1)(b) (see Rem. (3.5)(ii) in \[I1\]).
The only known example of a prime Fano threefold $`X`$ of $`g4`$ such that $`_X`$ has an exotic component $`_o`$, is the Mukai-Umemura threefold $`X_{22}^{}`$. The scheme $`_{X_{22}^{}}=2_o`$ and the surface $`R_o`$ is the hyperplane section of $`X_{22}^{}`$ swept out by the tangent lines to a rational normal curve $`C_{12}X_{22}^{}`$ of degree $`12`$ (see \[MU\]).
(1.4) By a theorem of Kobayashi and Ochiai the index $`r=r(Y)`$ of a smooth Fano $`n`$-fold $`Y`$ can’t be greater than $`n+1`$; and the only smooth Fano $`n`$-folds of $`rn`$ are $`𝐏^n`$ for which $`r=n+1`$ and the $`n`$-dimensional quadric $`Q_2^n`$ for which $`r=n`$ (see e.g. \[Pe\],p. 106). In particular, except $`𝐏^3`$ and $`Q_2^3`$, any smooth Fano $`3`$-fold must have index $`r2`$.
It is shown by Remmert and Van de Ven (for $`n=2`$) and later by Lazarsfeld (for any $`n`$) that the projective space $`𝐏^n`$ does not admit surjective morphisms to a smooth projective $`n`$-fold $`X𝐏^n`$ (see \[RV\], \[L\]). The same is true for morphisms $`f:Q_2^nX𝐏^n,Q_2^n`$ (see \[PS\]). In particular, $`𝐏^3`$ and $`Q_2^3`$ do not admit surjective moprphisms to smooth Fano threefolds $`X`$ of smaller index $`r(X)`$.
Let $`f:YX`$ be a non-constant morphism between smooth Fano $`3`$-folds of $`\rho =1`$. By Kor. 1.5 in \[RV\], $`\rho (Y)=1`$ implies that $`f`$ must be surjective, and by the preceding $`r(Y)`$ can’t be $`3`$. Therefore $`r(Y)=2`$, $`r(X)=1`$. This gives rise to the following question stated originally by Peternell (see (2.12)(2) in \[Pe\]).
(Pe) Question. Are there non-constant (hence surjective) morphisms $`f:YX`$ from a smooth Fano $`3`$-fold $`Y`$ of $`\rho (Y)=1`$ and $`r(Y)=2`$ to a smooth Fano $`3`$-fold $`X`$ of $`\rho (X)=1`$ and $`r(X)=1`$ ?
In this paper we give the expected negative answer to (Pe).
Let $`f:YX`$ be as above, and assume that $`f`$ is non-constant. Then $`f`$ must be surjective and finite since $`\rho (Y)=1`$ (see Kor. 1.5 in \[RV\]). Therefore $`f^{}:H^3(X,𝐂)H^3(Y,𝐂)`$ will be an embedding, in particular $`h^3(X)h^3(Y)`$ (see also \[Sch\]). For any Fano threefold $`h^{3,0}=0`$ and $`h^3=2h^{2,1}`$ since the anticanonical class is ample. Therefore $`h^{2,1}(X)h^{2,1}(Y)`$. Since $`h^{2,1}(Y)21`$ for any Fano 3-fold $`Y`$ of $`r=2`$ (see \[I1\]), then the existence of a non-constant morphism $`f:XY`$ as in (Pe) implies that $`h^{2,1}(X)21`$. This gives a negative answer to (Pe) whenever $`h^{2,1}(X)>21`$.
The only smooth non-prime Fano threefolds of $`\rho =1`$ and $`r=1`$ are the sextic double solid $`X_2^{}`$ and the double quadric $`X_4^{}`$ for which the answer to (Pe) is negative since $`h^{2,1}(X_2^{})=52>21`$ and $`h^{2,1}(X_4^{})=30>21`$. By the same argument the answer to (Pe) is negative also for the quartic threefold $`X_4`$ since $`h^{2,1}(X_4)=30>21`$. Any other smooth prime Fano threefold $`X=X_{2g2}𝐏^{g+1},4g12,g11`$ has $`h^{2,1}(X)20`$ (see \[I1\]).
In \[Sch\] is given a negative answer to (Pe) provided $`X`$ contains a conic of rank 2 (a pair of intersecting lines). The only known Fano threefold $`X`$ of $`\rho (X)=1`$ and $`r(X)=1`$ without intersecting lines is the Mukai-Umemura threefold $`X=X_{22}^{}`$, and a negative answer to (Pe) in case $`X=X_{22}^{}`$ is given by E. Amerik (see \[Sch\]).
Therefore, in order to give a negative answer to (Pe), it is enough to prove the following
(B) Proposition. Any smooth prime Fano threefold $`X_{2g2}𝐏^{g+1}`$, $`4g12,g11`$, different from the Mukai-Umemura threefold $`X_{22}^{}`$, contains a 1-dimensional family of conics of rank $`2`$.
In Section 2 we prove Proposition (B) for $`4g9`$ on the base of the following technical
(A) Lemma. A smooth prime Fano threefold $`X=X_{2g2}𝐏^{g+1},3g9`$ can’t contain the tangent scroll $`S_{2g2}`$ to a rational normal curve $`C_g`$ of degree $`g`$.
By a result of Yu. Prokhorov, the only smooth prime $`X=X_{2g2}`$, $`g=10,12`$ such that the scheme of lines $`_X`$ has an exotic component is the Mukai-Umemura threefold $`X_{22}^{}`$ (see \[Pr\]). This implies Proposition (B) for $`g=10,12`$. Indeed, let $`X=X_{2g2}`$ be a smooth prime Fano threefold such that the scheme of lines $`_X`$ on $`X`$ has a non-exotic component $`_o`$. Then, by Lemma (3.7) in \[I1\], the general element of $`_o`$ will represent a line $`lX`$ which intersects at least one other line on $`X`$.
This completes the proof of Proposition (B), which yields a negative answer to (Pe).
In Section 3 we prove Lemma (A) for any particular value of $`g`$, $`3g9`$.
For the prime Fano threefolds $`X_{2g2}𝐏^{g+1}`$ ($`g=3,4,5,6,8`$) we prove Lemma (A) by using the Mukai’s representation (see (3.1)) of the smooth $`X_{2g2}`$, $`3g10`$ as a complete intersection in a homogeneous or almost-homogeneous variety $`\mathrm{\Sigma }(g)`$. More concretely we see that if the threefold $`X\mathrm{\Sigma }(g)`$ ($`g=3,4,5,6,8`$) is a complete intersection in $`\mathrm{\Sigma }(g)`$ of the same type as the smooth prime $`X_{2g2}`$, and if $`X`$ contains the tangent scroll $`S_{2g2}`$ to the rational normal curve $`C_g`$ of degree $`g`$, then $`X`$ must be singular – see (3.4), (3.5), (3.6), (3.13), (3.15), (3.19).
For $`g=7`$ we use the properties of the projection from a special line $`lX_{2g2}`$, $`g7`$ to reduce the proof of Lemma (A) for $`g=7`$ to the already proved Lemma (A) for $`g=5`$ – see (3.20) - (3.24). To prove Lemma (A) in case $`g=9`$ we can use the same approach as for $`g=7`$. But a more elegant proof, based on the description of the double projection from a line, had been suggested by the Referee – see (3.25).
§ 2. Lemma (A) $``$ Proposition (B) for $`4g9`$.
(2.0) Assume that the smooth prime Fano threefold $`X=X_{2g2}`$ ($`4g9`$) does not contain a 1-dimensional family of conics of rank 2.
(2.1) Lemma. Under the assumption (2.0):
(i). The Hilbert scheme $`_X`$ of lines on $`X`$ has a unique irreducible component $`_o`$;
(ii). $`_o=_X`$ is exotic, and the lines $`l_o`$ sweep out a tangent scroll $`R_o|𝒪_X(d)|,d2`$.
Proof of (2.1).
(i). Let $`_o`$ and $`_{\mathrm{}}`$ be two different irreducible components of $`_X`$, and let $`R_o`$ and $`R_{\mathrm{}}`$ be the surfaces swept out by the lines $`l_o`$ and $`l_{\mathrm{}}`$. Since $`\mathrm{𝐏𝐢𝐜}(X)=𝐙.H`$, where $`H`$ is the hyperplane section, any effective divisor on $`X`$ must be ample. In particular the general line $`l_o`$ intersects the surface $`R_{\mathrm{}}X`$. If moreover $`lR_{\mathrm{}}`$ for the general $`l_o`$, then $`R_o=R_{\mathrm{}}`$ and this surface contains two 1-dimensional families of lines. Therefore $`R_o=R_{\mathrm{}}`$ is a quadric surface on $`X`$, which contradicts $`\mathrm{𝐏𝐢𝐜}(X)=𝐙.H`$. Therefore the general $`l_o`$ intersects $`R_{\mathrm{}}`$ and does not lie in $`R_{\mathrm{}}`$; and since $`R_{\mathrm{}}`$ is swept out by lines then there exists a line $`l^{}R_{\mathrm{}}`$ which intersects $`l`$. Since $`l_o`$ is general this produces a 1-dimensional family of intersecting lines $`l+l^{}`$ (= conics of rank 2 on $`X`$) – contradiction.
(ii). If $`_X=_o`$ is non-exotic then, by Lemma (3.7) of \[I1\], the general element of $`_o`$ will represent a line $`lX`$ which will intersect at least one other line $`mX`$, i.e. $`X`$ will contain a 1-dimensional family of intersecting lines (= conics of rank 2). Therefore $`_o`$ is exotic.
Since $`_o`$ is exotic and $`g4`$ then $`R_o`$ is the tangent scroll to a curve $`CX`$ (see (1.2)(3)), and since $`\mathrm{𝐏𝐢𝐜}(X)=𝐙.H`$ then $`R_o|𝒪_X(d)|`$ for some ineteger $`d1`$. If $`R_o`$ is a hyperplane section of $`X`$ (i.e. $`d=1`$) then, by Lemma 6 in \[Pr\], $`C=C_g`$ must be a rational normal curve of degree $`g`$. However the last is impossible since then Lemma (A) will imply that $`X`$ is singular. Therefore $`d2`$. q.e.d.
We shall show that nevertheless $`X`$ contains a 1-dimensional family of conics $`l+m`$ of rank $`2`$ where $`l,m_o`$.
Remark. Let $`C_s`$ be the (possibly empty) set of singular points of $`C`$. For any $`xCC_s`$ denote by $`l_x`$ the tangent line to $`C`$ at $`x`$. For a point $`x=x(0)C_s`$ define a tangent line to $`C`$ at $`x`$ to be any limit $`lim_{x(t)x(0)}l_{x(t)}`$ of tangent lines $`l_{x(t)}`$ to points $`x(t)CC_s`$ (see Ch.2 §4 in \[GH\]). Clearly, $`C`$ can have only a finite number of tangent lines to $`x(0)C_s`$ (see also Ch.2 §1.5 in \[Sh\]).
(2.2) By the initial assumption (2.0), $`X`$ does not contain a 1-dimensional family of conics of rank 2. In particular, $`X`$ does not contain a 1-dimensional family of pairs of intersecting tangent lines to $`C`$.
Since the surface $`R_o|𝒪_X(d)|`$, $`d2`$, and $`SpanX=𝐏^{g+1}`$ then $`SpanR_o=𝐏^{g+1}`$, $`g4`$ (see also Lemma 6 in \[Pr\]). Since $`R_o`$ is swept out by the tangent lines to $`C`$ then $`SpanC=SpanR_o=𝐏^{g+1}`$. In particular $`C`$ does not lie on a plane. Since $`R_o`$ is the tangent scroll to the non-plane curve $`C`$ then $`S_C`$ is singular along the curve $`C`$.
Let $`LC`$ be (if exists) an irreducible curve on $`R_o`$ such that $`R_o`$ is singular along $`L`$. If $`L`$ is not a tangent line to $`C`$, then the general point of $`L`$ will be an intersection point of two or more tangent lines to $`C`$ (see §4 in \[P2\]). The last is impossible since, by assumption, on $`X`$ can lie at most a finite number of pairs of intersecting lines.
Therefore any irreducible curve $`LC`$ such that $`LSingR_o`$ must be a tangent line to $`C`$. In addition, the tangent scroll $`R_o`$ to $`C`$ still can be singular along a tangent line $`L`$ to $`C`$ – for example if $`L`$ is a common tangent line to two or more branches of $`C`$ at $`x`$, or if $`C`$ has a branch with a cusp at $`x`$, or if $`xCC_s`$ but $`x`$ is an inflexion point of $`C`$ and then $`R_o`$ has a cusp along $`l_x`$, etc. (see §2, §4 in \[P2\]).
Let $`\mathrm{\Delta }R_o`$ be the union of all the irreducible curves $`L`$ on $`R_o`$ such that $`LC`$ and $`R_o`$ is singular along $`L`$. By the above argument, either $`\mathrm{\Delta }=\mathrm{}`$ or $`\mathrm{\Delta }`$ is a union of a finite number of tangent lines to $`C`$.
Let $`\nu :R_nR_o`$ be the normalization of $`R_o`$. Fix a desingularization $`\tau :\stackrel{~}{R}_nR_n`$, and let $`\sigma =\tau \nu :\stackrel{~}{R}_nR_o`$. Let $`E_1,\mathrm{},E_k`$ ($`k0`$) be all the irreducible contractable curves on $`\stackrel{~}{R}_n`$, i.e. all the irreducible curves $`E_i\stackrel{~}{R}_n`$ such that $`\sigma (E_i)R_o`$ is a point.
Denote by $``$ the linear equivalence of divisors on the smooth surface $`\stackrel{~}{R}_n`$, and let $`E`$ be a divisor on $`\stackrel{~}{R}_n`$. Call $`E`$ a zero divisor on $`\stackrel{~}{R}_n`$ if $`E0`$; call the non-zero divisor $`E`$ contractable if $`Ea_1E_1+\mathrm{}+a_kE_k`$ for some $`a_1,\mathrm{},a_k𝐙`$. Let $`C^{}\stackrel{~}{R}_n`$ be the proper $`\sigma `$-preimage of $`C`$ on $`\stackrel{~}{R}_n`$. Since $`R_o`$ is the tangent scroll to the irreducible curve $`C`$ then the curve $`C^{}`$ is irreducible and $`\sigma |_C^{}:C^{}C`$ is an isomorphism over a dense open subset of $`C`$ (see also Lemma (2.3) below).
Let $`C_1^{},\mathrm{},C_r^{}`$ be all the irreducible curves on $`\stackrel{~}{R}_n`$ such that $`\sigma (C_i^{})`$ is an irreducible component of $`\mathrm{\Delta }`$. Therefore
$`K_{\stackrel{~}{R}_n}\sigma ^{}K_{R_o}mC^{}\mathrm{\Sigma }_{i=1,\mathrm{},r}p_iC_i^{}+E`$
for some positive integers $`m,p_1,\mathrm{},p_r`$, and a contractable (or zero) divisor $`E`$ on $`\stackrel{~}{R}_n`$.
(2.3) Lemma. Let $`X`$ fulfills (2.0). Then the tangent scroll $`R_oX`$ to $`C`$ has a cusp of type $`v^2=u^3+\mathrm{}`$ along $`C`$, at a neighbourhood of the general point $`xCR_o`$; in particular $`m=mult_CR_o=2`$ (see also §5 in \[P2\] and §4 in \[P1\]).
Proof of (2.3).
(1). We shall see first that $`R_o`$ is irreducible at any neighbourhood of the general point $`xC`$, i.e. $`R_o`$ has one local branch at $`x`$.
Assume the contrary, and let $`xC`$ be general. Let $`UX`$ be a complex-analytical neighbourhood of $`x`$ such that $`R_U=R_oU`$ is reducible. Since $`R_o`$ is swept out by the tangent lines to $`C`$, the last imlies that for the general point $`yC_U=CU`$ (hence for the general $`yC`$) there exists (possibly non-unique) $`zC`$, $`zy`$ such that $`y`$ lies on a tangent line to $`C`$ at $`z`$. Since the set $`C_s=\{x_1,\mathrm{},x_s\}`$ of singular points of $`C`$ is finite (or empty), and any such $`x_i`$ has at most a finite number of tangent lines, then the general $`yC`$ doesn’t lie on a tangent line to $`zC_s`$. Therefore the general $`yC`$ lies on the tangent line $`l_z`$ to $`C`$ at some (possibly non-unique) $`zCC_s`$.
If moreover $`l_yl_z`$ (where $`l_y`$ is the tangent line to $`C`$ at $`y`$) then all such $`l_y+l_z`$ will produce a 1-dimensional family of conics of rank 2 on $`X`$, which contradicts the initial assumption (2.0) about $`X`$.
If $`l_y=l_z`$ then this will imply that the tangent line $`l_y`$ to $`C`$ at the general $`yC`$ is tangent to $`C`$ at two or more points. But then the projection $`\overline{C}`$ of $`C𝐏^{g+1}`$ from the general subspace $`𝐏^{g2}𝐏^{g+1}=SpanX`$ will be a plane curve with a 1-dimensional family of lines tangent to $`\overline{C}`$ at two or more points, which is impossible.
(2). It rests to see that the unique local branch of $`R_o`$ at the general $`xC`$ has a cusp of type $`v^2=u^3+\mathrm{}`$ along $`C`$ at a neighbourhood of $`x`$.
Since $`R_o|𝒪_X(d)|`$ and $`d2`$, then $`SpanC=SpanR_o=𝐏^{g+1}`$, $`g4`$ (see above).
Let $`x`$ be a general point of $`C`$. In order to prove that the tangent scroll $`R_oX`$ to $`C`$ has a cusp of type $`v^2=u^3+\mathrm{}`$ at a neighbourhood of $`x`$ it is enough to see that the projection of $`R_o`$ from a general $`𝐏^{g3}𝐏^{g+1}`$ has a cusp at $`x`$. This reduces the check to the case when $`R_o`$ is the tangent scroll to a curve $`C𝐏^3`$.
Since $`x`$ is a general point of $`C𝐏^3`$ then, after a possible linear change of coordinates in $`𝐏^3`$, the curve $`C`$ has (at $`x=(1:0:0:0)`$) a local parameterization, or a normal form (see §2 in \[P2\], or Ch.2 §4 in \[GH\]):
$`C_U:(x_o(z):\mathrm{}:x_n(z))=(1:z+o(z^2):z^2+o(z^3):z^3+o(z^4))`$, $`|z|<1`$,
where $`o(z^k)=\mathrm{\Sigma }_{jk}a_jz^j`$. Since the coefficient at $`z^k`$ in $`x_k(z)=z^k+o(z^{k+1})`$ is $`10`$ ($`k=2,3`$) then, after a possible linear change of $`(x_1,x_2,x_3)`$, the local parameterization of $`C`$ at $`x=(1:0:0:0)`$ can be written as
$`C_U:(x_o(z):x_1(z):x_2(z):x_3(z))=(1:z+o(z^4):z^2+o(z^4):z^3+o(z^4))`$, $`|z|<1`$,
i.e. $`C_U`$ approximates, upto $`o(z^4)`$, the twisted cubic $`C_3=\{(1:z:z^2:z^3)`$ }. Therefore, at a neighbourhood of $`x=(1:0:0:0)`$, the unique local branch (see (1)) of the tangent scroll $`R_o`$ to $`C`$ is parameterized by
$`R_U:(x_o(z,t):x_1(z,t):x_2(z,t):x_3(z,t))`$
= $`(1:z+t+o(z^4)+o(z^3)t:z^2+2zt+o(z^4)+o(z^3)t:z^3+3z^2t+o(z^4)+o(z^3)t)`$.
In affine coordinates $`(x_1,x_2,x_3)`$ the tangent line to $`C`$ at $`x=(0,0,0)`$ is spanned by the vector $`n_x`$ = $`(1,0,0)`$, and the normal space $`𝐂_o^2`$ $`𝐂^3(x_1,x_2,x_3)`$ to $`n_x`$ at $`x`$ is defined by $`x_1=0`$. In order to prove that $`R_U`$ has a cusp along $`C`$ at a neighbourhood of $`x`$ we shall see that the curve $`D_U`$ = $`R_U(x_1=0)`$ $`𝐂_o^2`$ has a cusp at $`x`$.
On $`D_U`$ = $`R_U(x_1=0)`$, one has: $`0=x_1`$ = $`z+t+o(z^4)+o(z^3)t`$, i.e. $`t`$ = $`z+o(z^4)`$. Let $`u=x_2`$, $`v=x_3/2`$. Therefore, on $`D_U𝐂_o^2`$
$`u=(z^2+2zt+o(z^4)+o(z^3)t)=z^2+o(z^4)`$
$`v=1/2(z^3+3z^2t+o(z^4)+o(z^3)t)=z^3+o(z^4))`$,
i.e. $`u^2=z^4+o(z^6)`$, $`uv=z^5+o(z^6)`$, $`v^2=z^6+o(z^7)`$, $`u^3=z^6+o(z^8)`$, $`u^2v=z^7+o(z^8)`$, …
Let $`C|_U=(f_U(u,v)=0)`$ be the local equation of $`D_U𝐂_o^2(u,v)`$ at $`x=(0,0)`$. Therefore, upto a constant non-zero factor, $`f_U(u,v)=v^2u^3+c_{2,1}u^2v+c_{1,2}uv^2+\mathrm{}`$, i.e. $`C|_U`$ has a double cusp-singularity of type $`v^2=u^3+\mathrm{}`$ at $`x=(0,0)`$ (see §5 in \[P1\], §4 in \[P2\], Ch.5 Examples 3.9.5, 3.9.1 and Ch.1 Exercise 5.14 in \[H\]).
Therefore $`R=R_o`$ has a pinch of type $`v^2=u^3+\mathrm{}`$ along $`C`$ at a neighbourhood of the general point $`xC`$, which proves Lemma (2.3).
(2.4) By the definition of $`C_1^{},\mathrm{},C_r^{}`$ any irreducible component of $`\mathrm{\Delta }`$ can be represented (possibly non-uniquely) as the image $`\sigma (C_i^{})`$ of some $`C_i^{},i=1,\mathrm{},r`$. Since $`\sigma |_C^{}:C^{}C`$ is an isomorphism over an open dense subset of $`C`$ then the general point $`xC`$ has a unique $`\sigma `$-preimage $`x^{}`$ on $`C^{}`$, and the proper preimage $`l_x^{}\stackrel{~}{R}_n`$ of the tangent line $`l_x`$ to $`C`$ at $`x`$ intersects $`C^{}`$ transversally at $`x^{}`$. Since, by assumption, on $`X`$ doesn’t lie a 1-dimensional family of pairs of intersecting lines then the tangent line $`l_x`$ to $`C`$ at the general point $`xC`$ does not intersect any other tangent line to $`C`$.
Therefore the non-singular surface $`\stackrel{~}{R}_n`$ has a structure of a possibly non-minimal ruled surface with a general fiber $`L^{}`$ := the proper $`\sigma `$-preimage of the general tangent line $`l_x`$ to $`C`$. In particular $`K_{\stackrel{~}{R}_n}.L^{}=2`$, and since the curve $`C^{}`$ is a section of $`\stackrel{~}{R}_n`$ then $`C^{}.L^{}=1`$.
By the definition of $`C_i^{}`$ the curves $`\sigma (C_i^{})R_o`$ are irreducible components of $`\mathrm{\Delta }`$; and since by (2.2) the components of $`\mathrm{\Delta }`$ can be only tangent lines to $`C`$ then $`\sigma (C_i^{})`$ is a tangent line to $`C`$. Therefore any component of $`\sigma ^1(\sigma (C_i^{}))`$, in particular $`C_i^{}`$, will not intersect the general fiber $`L^{}`$ of $`\stackrel{~}{R}_n`$, i.e. $`C_i^{}.L^{}=0`$.
Moreover a contractable curve $`E_j`$ can’t intersect the general fiber of $`\stackrel{~}{R}_n`$ since otherwise the point $`\sigma (E_j)R_o`$ will be a common point of a $`1`$-dimensional family of tangent lines to $`C`$. The last is impossible since $`g4`$ and the smooth $`X=X_{2g2}`$ can’t contain cones – see (1.2). Therefore $`E_j.L^{}=0`$ for any $`j=1,\mathrm{},k`$; and since $`E`$ is a sum of such $`E_j`$ then $`E.L^{}=0`$.
Since $`K_XH`$ and $`R_odH`$ on $`X`$ then, by adjunction, $`K_{R_o}(d1)H|_{R_o}`$. Since the hyperplane section $`H`$ intersects the general tangent line $`l`$ to $`C`$ at one point then $`\sigma ^{}(H|_{R_o})`$ is also a section of $`\stackrel{~}{R}_n`$, i.e. $`\sigma ^{}(H|_{R_o}).L^{}=1`$. Therefore
$`2=K_{\stackrel{~}{R}_n}.L^{}`$ = $`(\sigma ^{}K_{R_o}mC^{}\mathrm{\Sigma }_{i=1,\mathrm{},r}p_iC_i^{}+E).L^{}`$
= $`(d1)\sigma ^{}(H|_{R_o}).L^{}mC^{}.L^{}\mathrm{\Sigma }_{i=1,\mathrm{},r}p_iC_i^{}.L^{}+E.L^{}`$ = $`(d1)m`$, i.e. $`d=m1`$.
Since $`X=X_{2g2}`$ is smooth and $`g4`$ then, by Lemma (A), $`d>1`$. Therefore $`m=d+1>2`$, which is impossible since $`m=2`$ by Lemma (2.3).
This contradicts the initial assumption (2.0) that the smooth $`X=X_{2g2}`$, $`4g9`$ does not contain a 1-dimensional family of conics of rank 2. q.e.d.
§ 3. Proof of Lemma (A)
(3.1) By \[M1\], \[M2\] any smooth prime Fano threefold $`X_{2g2}𝐏^{g+1}`$, $`3g10`$ is a complete intersection of hypersurfaces $`F_1,F_2,\mathrm{},F_N`$ of degrees $`d_1,d_2,\mathrm{},d_N`$ in a homogeneous (for $`g=6`$ – an almost homogeneous) space $`\mathrm{\Sigma }(g)`$, and:
if $`g=3`$ then $`\mathrm{\Sigma }(3)=𝐏^4`$, $`N=1`$, $`d_1=4`$;
if $`g=4`$ then $`\mathrm{\Sigma }(4)=𝐏^5`$, $`N=2`$, $`d_1=2`$, $`d_2=3`$;
if $`g=5`$ then $`\mathrm{\Sigma }(5)=𝐏^6`$, $`N=3`$, $`d_1=d_2=d_3=2`$;
if $`g=6`$ then $`\mathrm{\Sigma }(6)=K.G(2,5)𝐏^{10}`$ is a cone over the grassmannian $`G(2,5)𝐏^9`$, $`N=4`$, $`d_1=d_2=d_3=1,d_4=2`$;
if $`7g10`$ then $`X_{2g2}=\mathrm{\Sigma }(g)𝐏^{g+1}`$, where $`\mathrm{\Sigma }(7)𝐏^{15}`$ is the spinor $`10`$-fold, $`\mathrm{\Sigma }(8)=G(2,6)𝐏^{14}`$, $`\mathrm{\Sigma }(9)𝐏^{13}`$ is the sympletic grassmann $`6`$-fold, and $`\mathrm{\Sigma }(10)𝐏^{13}`$ is the $`G_2`$-fivefold.
(3.2) To prove Lemma (A), it is enough to see that if $`X=X_{2g2}\mathrm{\Sigma }(g)`$ is a 3-fold complete intersection as in $`(\mathbf{3.1})`$ (assuming implicitly that such $`X`$ may have singularities) then $`X_{2g2}`$ can’t be smooth. We shall prove this separately for any value of $`g`$, $`3g9`$.
For $`g=3,4,5,6,8`$ we use that $`\mathrm{\Sigma }(g)`$ is either a projective space or a (cone over) grassmannian, which makes it possible to compute directly that the general such $`X_{2g2}S_{2g2}`$ must have $`12g`$ singular points on the curve $`C_g`$.
For $`g=7,9`$ we assume that $`X=X_{2g2}S_{2g2}`$ is smooth, and then project $`X`$ from a tangent line to $`C_g`$ to derive a contradiction on the base of the already known Lemma (A) for $`g=5`$.
(3.3) Notation. Let $`n1,m0`$ be integers, let $`𝐏^{n+m}(z:w)=𝐏^{n+m}(z_0:\mathrm{}:z_n:w_{n+1}:\mathrm{}:w_{n+m})`$ be the complex projective $`(n+m)`$-space, and let $`F(z:w)=F(z_0:\mathrm{}:z_n:w_{n+1}:\mathrm{}:w_{n+m})`$ be a homogeneous form. Denote by $`_zF=(F/z_0,\mathrm{},F/z_n)`$ the gradient vector of $`F`$ with respect to $`(z)=(z_0:\mathrm{}:z_n)`$.
Let $`F_1(z:w),\mathrm{},F_k(z:w)`$ ($`k1`$) be homogeneous forms. Denote by:
$`(F_1,\mathrm{},F_k)𝐂[z:w]=𝐂[z_0:\mathrm{}:z_n:w_{n+1}:\mathrm{}:w_{n+m}]`$ – the homogeneous ideal generated by $`F_1,\mathrm{},F_k`$;
$`V(F_1,\mathrm{},F_k)`$ – the projective variety defined by $`F_1,\mathrm{},F_k`$;
$`J_z|_{(a:b)}`$ = $`J_z(F_1,\mathrm{},F_k)|_{(a:b)}`$ = $`[_zF_1;\mathrm{};_zF_k]|_{(a:b)}`$ – the Jacobian matrix $`J_z`$ of partial derivatives of $`F_1,\mathrm{},F_k`$ with respect to $`(z)`$ = $`(z_0,\mathrm{},z_n)`$, computed at the point $`(a:b)𝐏^{n+m}(z:w)`$, (where $`_zF_i`$ are regarded as rows of $`J_z`$).
Let e.g. $`m=0`$, let $`X=V(F_1,\mathrm{},F_k)𝐏^n(z_0:\mathrm{}:z_n)`$, and let $`dimX=d`$. Then $`dimT_xXd`$ for any $`xX`$, where $`T_xX`$ is the tangent space to $`X`$ at $`x`$; and the point $`xX`$ is singular if $`dimT_xX>d`$ (see e.g. Ch. 2, §1.4 in \[Sh\]). Equivalently $`xX`$ is singular if $`rankJ_z|_x<nd`$. The subset $`SingX=\{xX:rankJ_z|_x<nd\}X`$ of all the singular points of $`X`$ is a proper closed subset of the projective algebraic variety $`X`$, defined on $`X`$ by vanishing of all the $`(nd)\times (nd)`$ minors of $`J_z`$.
(3.4) Proof of Lemma (A) for g = 3. The tangent scroll to the twisted cubic $`C_3:(x_0:x_1:x_2:x_3)=\stackrel{}{s}:=(s_0^3:s_0^2s_1:s_0s_1^2:s_1^3)`$ is the quartic surface $`S_4=V(f)𝐏^3(x)`$ where $`f(x)`$ = $`3x_1^2x_2^2+6x_0x_1x_2x_34x_1^3x_3x_0^2x_3^24x_0^2x_2^3`$. The surface $`S_4`$ is singular along $`C_3`$ since the gradient vector $`_xf|_\stackrel{}{s}=0`$ for any $`\stackrel{}{s}C_3`$.
Let the quartic threefold $`X_4𝐏^4(x:u)`$ be such that $`S_4=X_4𝐏^3(x)`$, and let $`X_4=V(F)𝐏^4(x:u)`$ where $`F(x:u)=\mathrm{\Sigma }_{0k4}f_i(x)u^{4i}`$. Therefore $`f_4(f)`$, i.e. $`f_4=cf`$ for some constant $`c𝐂`$.
Let $`xX_4`$. Then $`xSingX_4`$ iff $`_{x,u}F_x=0`$. Let $`s=(s_0:s_1)𝐏^1`$. Then $`(\stackrel{}{s}:0)SingX_4`$ iff $`0=_{x,u}F|_{(\stackrel{}{s}:0)}`$ = $`(_xF,F/u)|_{(\stackrel{}{s}:0)}`$ = $`(_xf|_\stackrel{}{s},f_3(\stackrel{}{s}))`$ = $`(0,f_3(\stackrel{}{s}))`$.
Therefore either $`f_3(\stackrel{}{s})0`$ (i.e. $`f_3(\stackrel{}{s})=0`$ for any $`s=(s_0:s_1)`$), and then $`X_4`$ is singular along $`C_4`$, or $`f_s(\stackrel{}{s})0`$, and then $`(\stackrel{}{s}:0)SingX_4`$ iff $`s=(s_0:s_1)`$ is a zero of the (non-vanishing) homogeneous form $`F_9(s)=f_3(\stackrel{}{s})=f_3(s_0^3:s_0^2s_1:s_0s_1^2:s_1^3)`$ of degree $`9`$.
In addition, for the general $`f_3(x)`$ all the zeros of $`F_9(s)=f_3(\stackrel{}{s})`$ are simple, i.e. different from each other. Therefore the general $`X_4S_4`$ has $`9=12g(X_4)`$ singular points on $`C_3`$. In coordinates as above, these singular points of $`X_4`$ are the images of the $`9`$ zeros of $`F_9(s)`$ under the Veronese map $`v_3:𝐏^1C_3X_4`$, $`v_3:s=(s_0:s_1)(\stackrel{}{s}:0)`$.
(3.5) Proof of Lemma (A) for g = 4. The tangent scroll to the rational normal quartic $`C_4:(x_0:x_1:\mathrm{}:x_4)=\stackrel{}{s}:=(s_0^4:s_0^3s_1:\mathrm{}:s_1^4)`$ is a complete intersection $`S_6=V(q,f)𝐏^4(x)`$ where $`q(x)=3x_2^24x_1x_3+x_0x_4`$ and $`f(x)=x_2^32x_0x_3^22x_1^2x_4+3x_0x_2x_4`$. The surface $`S_6`$ is singular along $`C_4`$ since the gradients of $`q`$ and $`f`$ are linearly dependent along $`C_4`$; more precisely $`_xf|_\stackrel{}{s}=s_0^2s_1^2_xq|_\stackrel{}{s}`$ for any $`\stackrel{}{s}C_4`$.
Let $`X_6=V(Q,F)𝐏^5(x:u)`$ be a complete intersection of the quadric $`Q(x:z)=\mathrm{\Sigma }_{0k2}q_k(x)u^{2k}=0`$ and the cubic $`F(x:z)=\mathrm{\Sigma }_{0l3}q_l(x)u^{2l}=0`$, and let $`S_6=X_6𝐏^4(x)`$. In particular, $`(q_2,f_3)(q,f)`$ as homogeneous ideals in $`𝐂[x]=𝐂[x_0:\mathrm{}:x_4]`$.
For the fixed $`X_6=V(Q,F)`$ the generators $`Q,F`$ of the homogeneous ideal $`(Q,F)`$ can be replaced by $`c^{}Q`$ and $`c^{\prime \prime }F+L(x:u)Q`$, for any pair of nonzero constants $`c^{}`$ and $`c^{\prime \prime }`$, and for any linear form $`L(x:u)`$. Now, $`(q_2,f_3)(q,f)`$ yields that one can choose $`Q`$ and $`F`$ such that $`q_2=\epsilon ^{}q`$ and $`f_3=\epsilon ^{\prime \prime }f`$, where $`\epsilon ^{},\epsilon ^{\prime \prime }`$ are either $`0`$ or $`1`$.
Consider the general case $`\epsilon ^{}=\epsilon ^{\prime \prime }=1`$; the study in the degenerate case $`\epsilon ^{}.\epsilon ^{\prime \prime }=0`$ is similar. The subscheme $`SingX_6=SingV(Q,F)`$ is defined by $`rank[_{x,u}Q;_{x,u}F]1`$. By the choice of $`Q`$ and $`F`$, $`_{x,u}Q|_{(\stackrel{}{s}:0)}=(_xq_2|_\stackrel{}{s},q_1(\stackrel{}{s}))`$ and $`_{x,u}F|_{(\stackrel{}{s}:0)}=(_xf_3|_\stackrel{}{s},f_2(\stackrel{}{s}))`$, where $`1.q_2=q`$ and $`1.f_3=f`$. Just as in case $`g=3`$, the last together with the identity $`_xf|_\stackrel{}{s}s_0^2s_1^2_xq|_\stackrel{}{s}0`$ imply that $`(\stackrel{}{s}:0)SingX_6`$ iff $`F_8(s):=f_2(\stackrel{}{s})s_0^2s_1^2.q_1(\stackrel{}{s})=0`$, where $`\stackrel{}{s}=(s_0^4:s_0^3s_1:\mathrm{}:s_1^4)`$.
The Veronese map $`v_4:𝐏^1C_4X_6`$, $`v_4:s=(s_0:s_1)(\stackrel{}{s}:0)`$ states an isomorphism between $`𝐏^1`$ and $`C_4`$. Therefore either $`F_8(s)0`$, and then $`X_6`$ is singular along $`C_4`$, or $`F_8(s)0`$, and then the singular points of $`X_6`$ on $`C_4`$ are the $`v_4`$-images of the zeros of the homogeneous form $`F_8(s)`$ of degree $`8`$. As in case $`g=3`$, for the general $`f_2(x),q_1(x)`$ the form $`F_8(s)`$ has only simple zeros. Therefore the general $`X_6S_6`$ has $`8=12g(X_6)`$ singular points on $`C_4`$.
(3.6) Proof of Lemma (A) for g = 5. The tangent scroll to the rational normal quintic $`C_5:x_i=s_0^{5i}s_1^i,0i5`$ is a complete intersection $`S_8=V(q^{},q^{\prime \prime },q^{\prime \prime \prime })𝐏^5(x)`$ where $`q^{}(x)=4x_1x_33x_2^2x_0x_4`$, $`q^{\prime \prime }(x)=3x_1x_42x_2x_3x_0x_5`$ and $`q^{\prime \prime \prime }(x)=x_1x_54x_2x_4+3x_3^2`$. The surface $`S_8`$ is singular along $`C_4`$ since the gradients of $`q^{}`$, $`q^{\prime \prime }`$ and $`q^{\prime \prime \prime }`$ are linearly dependent along $`C_5`$. More precisely
$`()`$ $`s_1^2_xq^{}|_\stackrel{}{s}s_0s_1_xq^{\prime \prime }|_\stackrel{}{s}s_0^2_xq^{\prime \prime \prime }|_\stackrel{}{s}=0`$ for any $`\stackrel{}{s}C_5`$.
Let $`X_8=V(Q^{},Q^{\prime \prime },Q^{\prime \prime \prime })𝐏^6(x:u)`$ be a complete intersection of the quadrics $`Q^i(x:z)=\mathrm{\Sigma }_{0k2}q_k^i(x)u^{2k}`$ ($`i=^{},^{\prime \prime },^{\prime \prime \prime }`$), and such that $`S_8=X_8𝐏^5(x)`$. In particular $`(q_1^{},q_1^{\prime \prime },q_1^{\prime \prime \prime })(q^{},q^{\prime \prime },q^{\prime \prime \prime })`$ as homogeneous ideals in $`𝐂[x]=𝐂[x_0:\mathrm{}:x_4]`$. Therefore $`q_1^{}`$, $`q_1^{\prime \prime }`$ and $`q_1^{\prime \prime \prime }`$ are linear combinations of $`q^{}`$, $`q^{\prime \prime }`$ and $`q^{\prime \prime \prime }`$. Since the ideal $`(Q^{},Q^{\prime \prime },Q^{\prime \prime })`$ of $`X_8`$ is generated also by any $`GL(3)`$-transform of the triple $`(Q^{},Q^{\prime \prime },Q^{\prime \prime \prime })`$, we may assume that $`q_2^{}=\epsilon ^{}q^{}`$, $`q_2^{\prime \prime }=\epsilon ^{\prime \prime }q^{\prime \prime }`$ and $`q_2^{\prime \prime \prime }=\epsilon ^{\prime \prime \prime }q^{\prime \prime \prime }`$, where $`\epsilon ^{}`$, $`\epsilon ^{\prime \prime }`$ and $`\epsilon ^{\prime \prime \prime }`$ are $`0`$ or $`1`$. The subscheme $`(SingX_8)|_{C_5}𝐏^1`$ is defined by
$`()`$ $`2rank[_{x,u}Q^{};_{x,u}Q^{\prime \prime };_{x,u}Q^{\prime \prime \prime }]|_{(\stackrel{}{s}:0)}`$
= $`rank[(_x\epsilon ^{}q^{}|_\stackrel{}{s},q_1^{}(\stackrel{}{s}));(_x\epsilon ^{\prime \prime }q^{\prime \prime }|_\stackrel{}{s},q_1^{\prime \prime }(\stackrel{}{s}));(_x\epsilon ^{\prime \prime \prime }q^{\prime \prime \prime }|_\stackrel{}{s},q_1^{\prime \prime \prime }(\stackrel{}{s})]`$.
Let $`F_7(s)=\epsilon ^{}s_1^2q_1^{}(\stackrel{}{s})\epsilon ^{\prime \prime }s_0s_1q_1^{\prime \prime }(\stackrel{}{s})\epsilon ^{\prime \prime \prime }s_0^2q_1^{\prime \prime \prime }(\stackrel{}{s})`$.
The Veronese map $`v_5:𝐏^1C_5X_8`$, $`v_5:s=(s_0:s_1)(\stackrel{}{s}:0)`$ states an isomorphism between $`𝐏^1`$ and $`C_5`$. Just as in (3.5), $`()`$ and $`()`$ imply that either $`F_7(s)0`$, and then $`X_8`$ is singular along $`C_5`$, or $`F_7(s)0`$, and then the singular points of $`X_6`$ on $`C_5`$ are the $`v_5`$-images of the zeros of the homogeneous form $`F_7(s)`$ of degree $`7`$. Moreover, for the general linear forms $`q_1^{}(x),q_1^{\prime \prime }(x),q_1^{\prime \prime \prime }(x)`$ the form $`F_7(s)`$ has only simple zeros. Therefore the general $`X_8S_8`$ has $`7=12g(X_8)`$ singular points on $`C_5`$.
Lemma (A) for g = 6,8.
(3.7) Lemma. Let $`n3`$, and let $`G=G(1:n)=G(2,n+1)𝐏^{n(n+1)/21}`$ be the grassmannian of lines in $`𝐏^n=𝐏(𝐂^{n+1})`$. Let $`CG𝐏^{n(n+1)/21}`$ be a smooth irreducible curve such that $`dimSpan(C)3`$. Let the surface $`Gr(C)𝐏^n`$ be the union of lines $`l𝐏^n`$ such that $`lCG=G(1:n)`$, and let $`S_CSpan(C)𝐏^{n(n+1)/21}`$ be the surface swept out by the tangent lines to $`C`$ (or the tangent scroll to $`C`$ – see above).
Then the tangent scroll $`S_C`$ to $`C`$ lies on $`G=G(1:n)`$ iff
either all the lines $`lC`$ have a common point, i.e. $`Gr(C)`$ is a cone,
or there exists an irreducible curve $`Z𝐏^n`$ such that all the lines $`lC`$ are tangent lines to $`Z`$, i.e. $`Gr(C)`$ is the tangent scroll to $`Z`$.
Proof. For $`n=3`$ this result can be found in \[AS\]. For $`n>3`$ one can apply induction, using the fact that projection from a point in $`𝐏^n`$ induces a projection from $`G(2,n+1)`$ onto $`G(2,n)`$.
(3.8) Lemma. Let $`C_gG(2,g/2+2)=G(1:𝐏^{g/2+1})`$ ($`g=6,8`$) be a rational normal curve such that the tangent scroll $`S_{2g2}`$ to $`C_g`$ is contained in $`G(2,g/2+2)`$. Then:
(i). If $`g=6`$ then the lines $`l𝐏^4`$, $`lC_6`$ sweep out the tangent scroll to a rational normal curve $`C_4𝐏^4`$.
(ii). If $`g=8`$, and if there exists a 3-fold linear section $`X_{14}=G(2,6)𝐏^9`$ such that $`X_{14}S_{14}`$, then the lines $`\{l𝐏^5:lC_8\}`$ sweep out the tangent scroll to a rational normal curve $`C_5𝐏^5`$.
Proof. Let $`g=6,8`$, let the lines $`lC_g`$ sweep out a cone $`Gr(C_g)𝐏^{g/2+1}`$ – see (3.7), and let $`x`$ be the vertex of $`Gr(C_g)`$. Then $`C_g`$ is contained in the Schubert $`g/2`$-space $`𝐏_x^{g/2}=\sigma _{g/2,0}(x)`$ = $`\{l𝐏^{g/2+1}:xl\}`$. Since $`C_g`$ is projectively normal it must span a $`g`$-space. Therefore $`gg/2+1`$ which contradicts $`g=6,8`$. Therefore, by (3.7), the lines $`lC_g`$ must sweep out the tangent scroll to a rational curve $`C_d𝐏^{g/2+1}`$.
Let $`g=6`$, and let $`C_d𝐏^3𝐏^4`$. Then $`C_6G(1:𝐏^3)`$ = $`\sigma _{11}(𝐏^3)G(1:𝐏^4)`$. Therefore $`6=dim(SpanC_6)dim(SpanG(1:𝐏^3))=5`$ – contradiction. Therefore $`d4`$ since $`C_d`$ must span $`𝐏^4`$, and now it is easy to see that the rational normal curve $`C_6`$ is the curve of tangent lines to $`C_d`$ iff $`d=4`$.
Let $`g=8`$, and let $`C_d𝐏^4𝐏^5`$. Then $`C_8G(1:𝐏^4)`$ = $`\sigma _{11}(𝐏^4)G(1:𝐏^5)`$. Let $`𝐏_o^9=SpanG(1:𝐏^4)`$. Then $`𝐏^8=SpanC_8𝐏_o^9`$. By condition $`C_8X_{14}=G(1:𝐏^5)𝐏^9`$. Therefore $`𝐏^8𝐏^9𝐏_o^9`$ and $`X_{14}Z_o:=G(1:𝐏^4)𝐏^8`$. Since $`𝐏^8𝐏_o^9=SpanG(1:𝐏^4)`$, $`Z_o`$ is at least a hyperplane section of the 6-dimensional grassmannian $`G(1:𝐏^4)`$. This contradicts $`X_{14}Z_o`$ and $`dimX_{14}=3`$. Therefore $`d5`$ since $`SpanC_d=𝐏^5`$, and now it is easy to see that the rational normal curve $`C_8`$ is the curve of tangent lines to $`C_d`$ iff $`d=5`$. q.e.d.
Proof of Lemma (A) for g = 6.
(3.9) By $`(\mathbf{3.1})`$ any smooth prime $`X_{10}`$ is a complete intersection of three hyperplanes and a quadric in the cone $`K.G(2,5)𝐏^{10}`$. Let $`o`$ be the vertex of the cone $`K.G(2,5)𝐏^{10}`$, and let $`XK.G(2,5)`$, be as in $`(\mathbf{3.1})`$. There are two kinds of such threefolds $`X_{10}`$ (see \[I1\], \[Gu\]):
(i). g = 6 – first kind: $`oSpan(X)`$, and then the projection $`p_o`$ from $`o`$ sends $`X`$ isomorphically to $`X_{10}=G(2,5)𝐏^7Q`$, where $`G(2,5)𝐏^9`$ by the Plücker embedding, $`𝐏^7𝐏^9`$ and $`Q`$ is a quadric.
(ii). g = 6 – second kind: $`oSpan(X)`$, and then $`\pi =p_o|_X:X=X_{10}^{}Y_5`$ is a double covering of a threefold $`Y_5=G(2,5)𝐏^6`$. In particular, if $`X_{10}^{}`$ is smooth then the intersection $`Y_5=G(2,5)𝐏^6`$ is smooth.
g = 6 (first kind)
(3.10) Let $`X_{10}=G(2,5)𝐏^7Q`$ be a (possibly singular) complete intersection as in (3.9)(i), and assume that $`X_{10}`$ contains the tangent scroll $`S_{10}`$ to a rational normal curve $`C_6`$ of degree 6. By Lemma (3.8)(i) the points of $`C_6`$ are the Plücker coordinates $`x_{ij}(s)`$ of the tangent lines to a rational normal quartic $`C_4=x_i=s^i,0i4`$, i.e.
$$(x_{ij}(s))=\left(\begin{array}{ccccc}0& 1& 2s& 3s^2& 4s^3\\ \mathrm{}& 0& s^2& 2s^3& 3s^4\\ \mathrm{}& \mathrm{}& 0& s^4& 2s^5\\ \mathrm{}& \mathrm{}& \mathrm{}& 0& s^6\\ \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}& 0\end{array}\right)$$
Therefore the subspace $`𝐏^6=Span(C_6)𝐏^9`$ is defined by $`H_0=H_1=H_2=0`$ where $`H_0=x_{03}3x_{12}`$, $`H_1=x_{04}2x_{13}`$ and $`H_2=x_{14}3x_{23}`$.
(3.11) Introduce in $`𝐏^6`$ the coordinates $`(v)=(v_0:\mathrm{}:v_6)=(x_{01}:x_{02}:x_{12}:x_{13}:x_{23}:x_{24}:x_{34})`$, and let $`I_{S_{10}}𝐂[v]:=𝐂[v_0,\mathrm{},v_6]`$ be the homogeneous ideal of the tangent scroll $`S_{10}𝐏^6`$ to $`C_6`$. Let $`Pf_k=x_{ab}x_{cd}x_{ac}x_{bd}+x_{ad}x_{bc}`$, $`0a<b<c<d4`$: $`a,b,c,dk`$ $`(k\{0,1,2,3,4\})`$ be the $`5`$ Plücker quadrics in the coordinates $`x_{jk}`$. In coordinates $`(v)`$ of $`𝐏^6=𝐏^6(v)`$ the restrictions $`q_k`$ of $`Pf_k`$ to $`𝐏^6`$ are
$`q_0(v)`$ = $`v_2v_6v_3v_5+3v_4^2`$, $`q_1(v)`$ = $`v_1v_63v_2v_5+2v_3v_4`$, $`q_2(v)`$ = $`v_0v_69v_2v_4+2v_3^2`$,
$`q_3(v)`$ = $`v_0v_53v_1v_4+2v_2v_3`$, $`q_4(v)`$ = $`v_0v_4v_1v_3+3v_2^2`$.
In the open subset $`U_0=\{v_0=1\}𝐏^6(v)`$ the curve $`C_6`$ is parameterized by $`C_6=\{(v)=\stackrel{}{s}:=(1,2s,s^2,2s^3,s^4,2s^5,s^6)\}`$.
Now, it is easy to see that the quadric $`q(v)=5v_2v_42v_1v_5+3v_0v_6`$ vanishes at the points of the tangent scroll $`S_{10}`$ to $`C_6`$, and the homogeneous ideal $`I_{S_{10}}=(q_0,\mathrm{},q_4,q)𝐂[v]`$.
Let $`J_v=J_v(q_0,\mathrm{},q_4,q)`$ be the Jacobian matrix of $`(q_0,\mathrm{},q_4,q)`$ with respect to $`v=(v_1,\mathrm{},v_6)`$. The surface $`S_{10}V_5`$ is singular along $`C_6`$ since $`rankJ_v|_\stackrel{}{s}<4=codim_{𝐏^6}S_{10}`$ for any $`\stackrel{}{s}C_6`$. For the special choice $`(v)`$ of the coordinates this can be verified directly:
$`s^2_vq_0|_\stackrel{}{s}=(0,s^4,2s^3,6s^2,2s,1)`$, $`s^1_vq_1|_\stackrel{}{s}=(s^5,6s^4,2s^3,4s^2,3s,2)`$,
$`_vq_2|_\stackrel{}{s}=(0,9s^4,8s^3,9s^2,0,1)`$, $`s_vq_3|_\stackrel{}{s}=(3s^5,4s^4,2s^3,6s^2,s,0)`$,
$`s^2_vq_4|_\stackrel{}{s}=(2s^3,6s^4,2s^3,s^2,0,0)`$, and $`_vq|_\stackrel{}{s}=(4s^5,5s^4,0,5s^2,4s,3)`$.
$`()`$. Therefore $`rankJ_v|_\stackrel{}{s}=4`$ for any $`\stackrel{}{s}C_6`$, and the linear 4-space of linear equations between the gradients $`_vq_0|_\stackrel{}{s},\mathrm{},_vq_4|_\stackrel{}{s}`$ and $`_vq|_\stackrel{}{s}`$ is spanned on the Pfaff-equation $`s^2_vq_0|_\stackrel{}{s}+s^1_vq_1|_\stackrel{}{s}+_vq_2|_\stackrel{}{s}+s_vq_3|_\stackrel{}{s}+s^2_vq_4|_\stackrel{}{s}=0`$ and the 3-space of equations $`_vq|_\stackrel{}{s}=a_0.s^2_vq_0|_\stackrel{}{s}+a_1.s^1_vq_1|_\stackrel{}{s}+a_2._vq_2|_\stackrel{}{s}+a_3.s_vq_3|_\stackrel{}{s}+a_4.s^2_vq_4|_\stackrel{}{s}=0`$ where:
$`()`$. $`a_0+4a_3+3a_4=8`$, $`a_13a_32a_4=4`$, $`a_2+2a_3+a_4=3`$.
(3.12) In the dual space $`\stackrel{ˇ}{𝐏}^9`$, let $`\stackrel{ˇ}{G}=\stackrel{ˇ}{G}(2,5)\stackrel{ˇ}{𝐏}^9`$ be the grassmannian of hyperplane equations represented by the skew-symmetric $`5\times 5`$ matrices of rank $`2`$. It is easy to see that the plane $`\mathrm{\Pi }=<H_0,H_1,H_2>\stackrel{ˇ}{𝐏}^9`$ of hyperplane equations of $`𝐏^6=SpanC_6`$ does not intersect $`\stackrel{ˇ}{G}`$.
Let $`\mathrm{\Lambda }\mathrm{\Pi }`$ be any line in $`\mathrm{\Pi }`$. In $`𝐏^9`$, the line $`\mathrm{\Lambda }`$ defines, by duality, the subspace $`𝐏^7(\mathrm{\Lambda })𝐏^6=SpanC_6`$. It is easy to see that the fourfold $`W(\mathrm{\Lambda })=G𝐏^7(\mathrm{\Lambda })`$, where $`G=G(2,5)`$, is smooth. In fact, $`W(\mathrm{\Lambda })`$ will be smooth iff the line $`\mathrm{\Lambda }`$ does not intersect $`\stackrel{ˇ}{G}`$. The last is true since $`\mathrm{\Lambda }\mathrm{\Pi }`$ and $`\mathrm{\Pi }\stackrel{ˇ}{G}=\mathrm{}`$. Therefore any $`X_{10}S_{10}`$ is a quadratic section of the smooth 4-fold $`W=W(\mathrm{\Lambda })`$.
(3.13) Let $`X_{10}=G(2,5)𝐏^7QS_{10}`$, where $`Q`$ is a quadric. We shall show that the singularities of $`X_{10}`$ on $`C_6`$ are the zeros of a homogeneous form of degree $`6`$ on $`C_6𝐏^1`$. To simplify the notation, we shall show this for one special choice of the line $`\mathrm{\Lambda }\mathrm{\Pi }`$ (see (3.12); the check for any other $`\mathrm{\Lambda }\mathrm{\Pi }`$ is similar.
Let $`\mathrm{\Lambda }=\{H_2=0\}\mathrm{\Pi }`$. Then, in coordinates $`(v)`$ and $`u=x_{14}3x_{23}`$ in $`𝐏^7(\mathrm{\Lambda })`$, the subspace $`𝐏^6(x)=(u=0)`$. Therefore any quadric $`Q𝐏^7(\mathrm{\Lambda })`$, such that $`QX_{10}=S_{10}`$, can be written in the form $`Q=Q(v,u)=cu^2+L(v)u+q(v)`$ where $`c𝐂`$ and $`L`$ is a linear form of $`(v)`$.
Let $`Q_k(v,u)`$ ($`k=0,1,\mathrm{},4`$) be the restriction of the Pfaff quadric $`Pf_k`$ on $`𝐏^7(v,u)`$, and let $`J_{v,u}=[_{v,u}Q_0;\mathrm{};_{v,u}Q_4;_{v,u}Q_0]`$ be the Jacobian matrix of $`(Q_0,\mathrm{},Q_4,Q)`$.
The singularities of $`X_{10}`$ on $`C_6`$ are the points $`(\stackrel{}{s}:0)𝐏^7`$ for which $`rankJ_{v,u}|_{(\stackrel{}{s}:0)}<4`$ = $`codim_{𝐏^7}X_{10}`$.
Let $`l_i(\stackrel{}{s})=Q_i/u|_\stackrel{}{s}`$, $`i=0,\mathrm{},4`$. The rows of $`J_{v,u}|_{(\stackrel{}{s},0)}`$ are $`_{u,v}Q_i|_{(\stackrel{}{s}:0)}=(_uq_i|_\stackrel{}{s},l_i(\stackrel{}{s}))`$, $`i=0,\mathrm{},4`$ and $`_{u,v}Q|_{(\stackrel{}{s}:0)}=(_uq|_\stackrel{}{s},L(\stackrel{}{s}))`$. For the special choice of the line $`\mathrm{\Lambda }\mathrm{\Pi }`$, the linear forms $`l_i=l_i(\stackrel{}{s})`$ are $`(l_0,l_1,l_2,l_3,l_4)=(s^4,0,3s^2,2s,0)`$.
Therefore, in view of $`()`$, $`X_{10}`$ will have a singularity at $`(\stackrel{}{s},0)C_6`$ if there exist constants $`a_0,\mathrm{},a_4`$ satisfying $`()`$ and such that $`L(\stackrel{}{s})`$ = $`(a_03a_22a_3)s^2`$ (here $`\stackrel{}{s}`$ = $`(1,2s,s^2,2s^3,s^4,2s^5,s^6)`$ – see above).
By $`()`$ $`a_03a_22a_3=1`$. Therefore, in homogeneous coordinates $`(s_0:s_1)`$, $`s=s_1/s_0`$, the variety $`X_{10}`$ will be singular at the point $`(\stackrel{}{(}s):0)C_6`$ iff $`F_6(s_0:s_1):=L(s_0^6:2s_0^5s_1:s_0^4s_1^2:2s_0^3s_1^3:s_0^2s_1^4:2s_0s_1^5:s_1^6)+s_0^4s_1^2=0`$.
The Veronese map $`v_6:𝐏^1C_6X_{10}`$, $`v_6:(s_0:s_1)(\stackrel{}{s}:0)`$ = $`(s_0^6:2s_0^5s_1:s_0^4s_1^2:2s_0^3s_1^3:s_0^2s_1^4:2s_0s_1^5:s_1^6:0)`$, states an isomorphism between $`𝐏^1`$ and $`C_6`$. Therefore either $`F_6(s_0:s_1)0`$, and then $`X_{10}`$ is singular along $`C_6`$, or $`F_6(s_0:s_1)0`$, and then the singular points of $`X_{10}`$ on $`C_6`$ are the $`v_6`$-images of the zeros of the homogeneous form $`F_6(s_0:s_1)`$ of degree $`6`$.
The choice of an arbitrary line $`\mathrm{\Lambda }\mathrm{\Pi }`$ will only change the homogeneous sextic forms defined by $`l_i(\stackrel{}{s}),i=0,\mathrm{},4`$. q.e.d.
g = 6 (second kind)
(3.14) Let $`\pi :X=X_{10}^{}Y_5`$, $`X𝐏^7`$, be a double covering of the Del Pezzo threefold $`Y_5=G(2,5)𝐏^6`$ branched along the quadratic section $`BY_5`$. Below we shall identify the branch locus $`BY_5`$ and the ramification divisor $`RX_{10}^{}`$, $`RB`$.
(3.15) Assume that $`X`$ contains the tangent scroll $`S=S_{10}`$ to the rational normal sextic $`C=C_6`$, and let $`lS`$ be a general tangent line to $`C`$. Then $`N_{l/X}𝒪(1)𝒪(2)`$ (see (1.2)), and $`\pi (l)Y_5`$ also is a line. If $`N_{\pi (l)/Y_5}𝒪𝒪`$ then $`d\pi :N_{l/X}N_{\pi (l)/Y_5}`$ has one-dimensional kernel along $`l`$. In this case $`l`$ is contained in the ramification divisor $`R`$ of $`\pi `$. The last is possible only for a finite number of $`l^{}s`$. Therefore, at least for the general tangent line $`lS`$ to $`C`$, the line $`\pi (l)\pi (S)`$ is a $`(1,1)`$-line on $`Y_5`$, i.e. $`N_{\pi (l)/Y_5}𝒪(1)𝒪(1)`$. Therefore $`\pi (S)Y_5`$ coincides with the surface $`S_{1,1}Y_5`$ swept-out by the $`(1,1)`$-lines on $`Y_5`$, which in turn is a tangent scroll to a rational normal sextic (see \[FN\]). Clearly $`\pi :SS_{1,1}`$ is an isomorphism and $`S_{1,1}`$ is the tangent scroll to $`\pi (C_6)`$.
Assume first that $`S=R`$. Then $`S_{1,1}=\pi (S)=BR`$. Therefore $`X`$ is singular along $`C_6`$ since the branch locus $`B=S_{1,1}`$ of $`\pi `$ is singular along $`C_6`$. In order to prove Lemma (A) for $`g=6`$ (second kind) it rests to see the general $`X_{10}^{}S_{10}`$ such that $`SR`$ is singular. This will imply that any $`X_{10}^{}S_{10}`$ must be singular – since the property $`X_{10}^{}S_{10}`$ to have a singularity is a closed condition.
Since $`\pi :X_{10}^{}Y_5`$ is a 2-sheeted covering with a ramification divisor $`R`$, then $`H^0(X_{10}^{},O(1))=\pi ^{}H^0(Y_5,O(1))+𝐂.R`$ where $`R`$ is the hyperplane equation of $`RX_{10}^{}`$. Therefore, since $`S\pi (S)`$ and $`SR`$, then $`\pi (S)`$ is tangent to the branch locus $`B`$ of $`\pi `$ along a (possibly singular) canonical curve $`C_{10}^6=BH=\pi (S)H`$ for some hyperplane $`H𝐏^6`$. Since $`\pi (S)=S_{1,1}`$ then the general $`X_{10}^{}S=S_{10}`$, $`SR`$ comes from a branch locus $`B`$ totally tangent to $`S_{1,1}`$ along a general hyperplane section $`HS_{1,1}`$. Since $`H`$ is general then $`H`$ intersects the rational normal sextic $`\pi (C_6)`$ at $`6`$ points $`p_1,\mathrm{},p_6`$ such that $`p_ip_j`$ for $`ij`$. By $`p_iB`$ and the identification $`B=R`$, we may cosider $`p_i`$ as points on $`RX_{10}^{}`$. We shall see that
$`()`$ Lemma. $`p_iSingX_{10}^{}`$, $`i=1,\mathrm{},6`$.
Proof of $`()`$. Let $`U_iY_5`$ be a sufficiently small neighborhood of the point $`p_i`$. Since $`Y_5`$ is smooth, we can identify $`U_i`$ with a disk in $`𝐂^3`$, and let $`(u,v,w)`$ be local coordinates in $`U_i`$ s.t. $`p_i=(0,0,0)`$. Let $`f(u,v,w)=0`$ be the local equation of $`S_{10}`$ in $`U_i`$. Since $`S_{10}`$ is the tangent scroll to the smooth rational curve $`C_6`$, and $`p_iC_6`$, one can choose the coordinates $`u,v,w`$ such that $`f=u^3v^2`$ (since the scroll $`S_{10}`$ has a double cusp-singularity along $`C_6`$ – see Lemma (2.3)). Let $`q=0`$ and $`h=0`$ be the local equations of $`B`$ and $`H`$ in $`U_i`$. Since $`B`$ and $`S_{10}`$ are singular along $`C_{10}^6`$ = $`BH`$ = $`S_{10}H`$, $`h^2`$, $`f`$ and $`q`$ are linearly dependent, i.e. $`\alpha h^2+\beta f+\gamma q=0`$ for some constants $`\alpha ,\beta ,\gamma `$. Moreover $`\alpha 0`$ and $`\gamma 0`$ since $`\pi (S)S`$ and $`\pi (S)B`$, and one may assume that $`\beta 0`$ (otherwise $`B=2H`$ and $`X_{10}^{}`$ will be singular along the surface $`R_{red}B_{red}=H`$). Since $`p_iH`$, $`h(0,0,0)=0`$ and $`h=au+bv+cw+o(2)`$, where $`o(k)`$ denotes a sum of terms of degree $`k`$. Therefore the surface $`B`$ is singular at $`p_i`$ since $`BU_i=(q=0)`$, and the series expansion
$`q(u,v,w)=\beta /\gamma f\alpha /\gamma h^2`$ = $`\beta /\gamma (v^2u^3)\alpha /\gamma (au+bv+cw+o(2))^2`$
has no linear term. Since $`B`$ is the branch locus of $`\pi :X_{10}^{}Y_5`$, the threefold $`X_{10}^{}`$ will be also singular at $`p_iRB`$, $`i=1,\mathrm{},6`$. In addition, the $`6=12g(X_{10}^{})`$ singular points $`p_1,\mathrm{},p_6`$ of $`X_{10}^{}`$ lie on the rational normal curve $`C_6X_{10}^{}`$. q.e.d.
Proof of Lemma (A) for g = 8.
(3.16) The Da Palatini construction (see \[Pu\]). Let $`𝐏^5=𝐏(V)`$ where $`V=𝐂^6`$, and let $`\widehat{V}=Hom_𝐂(V,𝐂)`$ be the dual space of $`V`$. The points $`H^2\widehat{V}=Hom_𝐂(^2V,𝐂)`$ can be regarded as skew-symmetric linear maps $`H:V\widehat{V}`$, and the hyperplanes $`(H=0)𝐏^{14}=𝐏(^2V)`$ can be regarded as points of $`\widehat{𝐏}^{14}=𝐏(^2\widehat{V})`$.
Let $`Pf=\{H^2\widehat{V}):rank(H)4\}/𝐂^{}`$. Then $`Pf`$ is the Pfaff cubic hypersurface in $`\widehat{𝐏}^{14}`$ defined, in coordinates, by vanishing of the cubic Pfaffian of the skew-symmetric $`(6\times 6)`$-matrix $`H`$. Let $`U_{10}^2V`$ be a 10-dimensional subspace, and let
$`\widehat{U}_5`$ = $`U_{10}^{}`$ = $`\{H^2\widehat{V}`$ = $`Hom_𝐂(^2V,𝐂):H|_{U_{10}}=0\}`$.
Let moreover $`U_{10}^2V`$ be such that $`rank(H)4`$ for any $`HU_{10}^{}`$, and let $`X_{14}`$ = $`G(2,6)𝐏(U_{10})`$ and $`B_3`$ = $`Pf𝐏(\widehat{U}_5)`$.
The construction “Da Palatini” of G. Fano shows that any hyperplane $`𝐏^4𝐏(V)`$ defines a birational isomorphism $`\xi :X_{14}B_3`$ which can be described as follows (see \[Pu\]):
Identify the point $`bB_3`$ and the (projective equivalence class of) the skew-symmetric $`6\times 6`$ matrix corresponding to $`b`$. Since $`rankb=4`$ for any $`bB_3`$ the projective kernel $`n_b`$ of $`b`$ will be a line in $`𝐏^5`$. Let $`W`$ := $`_{bB_3}\{n_b=𝐏(Kerb)\}𝐏^5`$.
The fourfold $`W`$ can be described by an alternative way. Identify the point $`lG(1:5)`$ and the line $`l𝐏^5`$, and let $`W^{}:=\{l𝐏^5|lX_{14}\}𝐏^5`$. Then (see \[Pu, p. 83\]):
(a). for the general $`vW^{}`$ there exists a unique $`lX_{14}`$ such that $`vl`$;
(b). for the general $`wW`$ there exists a unique $`bB`$ such that $`wn_b`$;
(c). $`W^{}=W`$.
Let $`H𝐏^5`$ be a general hyperplane. Then, by (a), (b) and (c), the maps $`\varphi :X_{14}HV`$, $`\varphi (l)=Hl`$ and $`\psi :B_3HV`$, $`\psi (b)=Hn_b`$ are birational isomorphisms. The composition $`\xi =\xi _H=\psi ^1\varphi :X_{14}B_3`$ is a birational isomorphism, depending on the choice of the hyperplane $`H𝐏^5`$ (see \[Pu, p. 85\]).
(3.17) The dual cubic fourfold of $`S_{14}`$. By Lemma (3.8)(ii) the curve $`C_8`$ is the Plücker image of the tangent scroll to a rational normal quintic $`C_5:(x_0:\mathrm{}:x_5)=\stackrel{}{s}=(1:s:s^2:\mathrm{}:s^5)`$ in $`𝐏^5(x)`$. The points of the curve $`C_8`$ are the Plücker coordinates $`x_{ij}(s)`$ of the point $`\stackrel{}{s}C_5`$:
$$(x_{ij}(s))=\left(\begin{array}{cccccc}0& 1& 2s& 3s^2& 4s^3& 5s^4\\ \mathrm{}& 0& s^2& 2s^3& 3s^4& 4s^5\\ \mathrm{}& \mathrm{}& 0& s^4& 2s^5& 3s^6\\ \mathrm{}& \mathrm{}& \mathrm{}& 0& s^6& 2s^7\\ \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}& 0& s^8\\ \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}& \mathrm{}& 0\end{array}\right)$$
Therefore $`S_{14}=G(2,6)𝐏^8`$ where $`𝐏^8=(H_0=\mathrm{}=H_5=0)𝐏^{14}`$ and:
$`H_0=x_{03}3x_{12}`$, $`H_1=x_{04}2x_{13}`$, $`H_2=3x_{05}5x_{14}`$,
$`H_3=x_{14}3x_{23}`$, $`H_4=x_{15}2x_{24}`$, $`H_5=x_{25}3x_{34}`$.
Let $`\mathrm{\Pi }^5:=<H_0,\mathrm{},H_5>\widehat{𝐏}^{14}`$ be the projective linear span of $`\{H_0,\mathrm{},H_5\}`$, and let $`𝐁_3=Pf<H_0,\mathrm{},H_5>`$. Introduce projective coordinates $`(t_0:\mathrm{}:t_5)`$ in $`\mathrm{\Pi }^5`$ such that the point $`(t_0,\mathrm{},t_5)`$ represents the vector $`t_0H_0+\mathrm{}+t_5H_5`$. Then the cubic 4-fold $`𝐁_3=Pf𝐏^5(t_0:\mathrm{}:t_5)`$ is defined by $`𝐁_3:F=32t_0t_2t_5t_0t_3t_52t_1^2t_52t_0t_4^2+3t_1t_3t_412t_1t_2t_445t_2^2t_39t_2t_3^2=0`$.
The 6-vector $`(0,\mathrm{},0)`$ is the only value of $`(t_0,\mathrm{},t_5)`$ where the $`15`$ Pfaffians $`Pf_{ij}(t_0,\mathrm{},t_5)`$ of the matrix $`H(t_0,\mathrm{},t_5)`$ = $`t_0H_0+\mathrm{}+t_5H_5`$ vanish. Therefore $`rankb=4`$ for any $`b𝐁_3`$.
The fourfold $`𝐁_3`$ = $`(F=0)\mathrm{\Pi }^5`$ is singular, and it is is not hard to check that $`Sing𝐁_3`$ = $`(_{(t_0:\mathrm{}:t_5)}F`$ = $`(0,\mathrm{},0))\mathrm{\Pi }^5`$ is the rational normal quartic curve $`C_4`$ = $`\{[r]=(1:2r:r^2/3:8r^2/3:2r^3:r^4)|r𝐂\}`$; for simplicity we let $`t_0=1`$.
(3.18) Now we are ready to prove Lemma (A) for $`g=8`$.
Let $`X_{14}=G(2,6)𝐏^9=𝐏(U_{10})`$ be such that $`X_{14}S_{14}`$, and let $`B_3=Pf𝐏(U_{10}^{})`$. Then $`S_{14}X_{14}=G(2,V)𝐏(U_{10})`$ $``$ $`𝐏^8=SpanS_{14}𝐏(U_{10})=SpanX_{14}`$ $``$ $`𝐏(U_{10}^{})\mathrm{\Pi }^5`$ $``$ $`B_3𝐁_3`$. Since $`B_3𝐁_3`$ then $`rank(H)=4`$ for any $`HB_3`$ (see (3.17)), hence the Da Palatini birationalities $`\xi :X_{14}B_3`$ (see (3.16)) are well-defined.
Assume that $`X_{14}`$ is smooth. Then $`B_3`$ must be smooth (see above or \[Pu, p. 83\]). But $`B_3`$ must be singular at any of the intersection points of the hyperplane $`𝐏(U_{10}^{})\mathrm{\Pi }^5`$ and the rational quartic curve $`C_4=Sing𝐁_3`$ – contradiction (see the end of (3.17)). Therefore any $`X_{14}S_{14}`$ must be singular. q.e.d.
(3.19) Remark. If $`𝐏(U_{10}^{})=Span(C_4)`$ then $`B_3`$ is singular along $`C_4`$, and it can be seen that then $`X_{14}`$ is singular along $`C_8`$. Let $`𝐏(U_{10}^{})Span(C_4)`$. Then the hyperplane $`𝐏(U_{10}^{})\mathrm{\Pi }^5`$ intersects the rational normal quartic $`C_4=Sing𝐁_3`$ in $`4`$ possibly coincident points $`b_1,b_2,b_3,b_4`$. Let, for simplicity $`b_i`$ be different from each other. Then one can show that the general Da Palatini birationality $`B_3X_{14}`$ sends $`\{b_1,b_2,b_3,b_4\}`$ to $`4=12g(X_{14})`$ singular points of $`X_{14}`$ which lie on $`C_8`$. Let $`H𝐏^5`$ be a hyperplane, and let $`\xi _H^1:B_3X_{14}`$ be the Da Palatini birationality defined by $`H`$. We shall see that for the general $`H`$ the rational map $`\xi _H^1=\psi _H`$ is regular at a neighborhood of any $`b_i`$, and $`\xi _H^1(b_i)C_8`$. For this, by the definition of the maps $`\varphi `$ and $`\psi `$, it is necessary to see that the kernel-map $`ker:B_3G(1:5)`$, $`bn_b`$ sends the quartic $`C_4`$ isomorphically to $`C_8`$.
Let $`b=(b_{ij})B_3`$. Then the Plücker coordinates of the line $`n_b=ker(b)`$ are $`(1)^{i+j}Pf_{ij}(b)`$, where $`Pf_{ij}(b)`$ are the 15 quadratic Pfaffians of the skew-symmetric matrix $`\widehat{b}`$; note that $`rank(b)=4`$ for any $`bB_3𝐁_3`$. Now, it rests only to replace $`b`$ by the general point $`b(t)=H_0+tH_1+t^2/12H_2+2t^2/3H_3+t^3/4H_4+t^4/16H_5C_4`$, and to see that the Plücker coordinates of $`b(t)`$ parameterize the general point $`x_{ij}(\stackrel{}{s})`$ of $`C_8`$ (where $`s=2/t`$) – see (3.17).
Proof of Lemma (A) for g = 7
(3.20) In the proof of Lemma (A) for $`g=7`$ we shall need the known by \[I2\] description of the projection from a line $`l`$ on a smooth prime Fano threefold $`X_{2g2}`$ such that $`N_{l/X}=𝒪(1)𝒪(2)`$.
(3.21) Lemma (see §1 Proposition 3 in \[I2\]). Let $`l`$ be a line on the smooth prime Fano threefold $`X=X_{2g2}𝐏^{g+1}`$ such that $`N_{l/X}=𝒪_l(2)𝒪_l(1)`$, and let $`\sigma :X^{}X`$ be the blow up of $`l`$. Let $`Z^{}=\sigma ^1(l)`$, let $`H^{}\sigma ^{}HZ^{}`$ be the proper preimage of the hyperplane section $`H`$ of $`X`$, and let $`\pi :XX^{\prime \prime }𝐏^{g1}`$ be the projection from $`l`$. Then:
(i). If $`g5`$ then the composition $`\varphi =\pi \sigma :X^{}𝐏^{g1}`$ is a birational morphism (given by the linear system $`H^{}`$), to a threefold $`X^{\prime \prime }𝐏^{g1}`$.
(ii). The restriction to $`Z^{}=𝐏(N_{l/V})=𝐅_3`$ of the linear system $`H^{}`$ is the complete linear system $`s^{}+3f^{}`$, where $`s^{}`$ and $`f^{}`$ are the classes of the exceptional section and the fiber of the rational ruled surface $`Z^{}`$.
(iii). The restriction $`\varphi _Z^{}`$ of $`\varphi `$ to $`Z^{}`$ maps $`Z^{}`$ to a cone $`Z^{\prime \prime }`$ over a twisted cubic curve, contracting the exceptional section $`s^{}`$ of $`Z^{}`$ to the vertex of $`Z^{\prime \prime }`$.
(iv). If $`g7`$ then there are only a finite number of lines $`l_iX`$ ($`i=1,\mathrm{},N`$) which intersect $`l`$. Let $`l_i^{}X^{}`$ of $`l_i`$ ($`i=1,\mathrm{},N`$) be the proper preimages $`l_i^{}X^{}`$ of $`l_i`$ ($`i=1,\mathrm{},N`$). Then the morphism $`\varphi :X^{}X^{\prime \prime }`$ is an isomorphism outside $`l_1^{}\mathrm{}l_N^{}s^{}`$, and $`\varphi `$ contracts $`s^{}`$ and any of $`l_i^{}`$ to isolated double points of $`X^{\prime \prime }`$.
(v). Let $`H^{\prime \prime }=\varphi (H^{})`$ be the hyperplane section of $`X^{\prime \prime }`$. Then if $`g7`$ then $`K_{X^{\prime \prime }}H^{\prime \prime }`$, i.e. the variety $`X^{\prime \prime }=X_{2(g2)2}^{\prime \prime }𝐏^{g1}`$ is an anticanonically embedded Fano threefold with isolated singularities as in (iv).
(3.22) Suppose that there exists a smooth prime $`X=X_{12}𝐏^8`$ which contains the tangent scroll $`S_{12}`$ to the rational normal curve $`C_7`$. Therefore the general such $`X`$ is smooth, and we may suppose that $`XS_{12}`$ is general.
Let $`lS_{12}`$ be any of the tangent lines to $`C_7`$ and let $`\pi :XX^{\prime \prime }`$ be the projection from $`l`$. By \[I2, §1\] (see also (1.2)) $`N_{l/X}=𝒪(1)𝒪(2)`$, therefore (3.21)(i)-(v) take place.
By (3.21)(v) the threefold $`X^{\prime \prime }=X_8^{\prime \prime }𝐏^6`$ is an anticanonically embedded Fano threefold of genus $`5`$. Let $`S^{\prime \prime }X^{\prime \prime }`$ be the proper image of $`S_{12}`$, and let $`C^{\prime \prime }S^{\prime \prime }X^{\prime \prime }`$ be the proper image of $`C_7`$. It is easy to see that $`C^{\prime \prime }=C_5^{\prime \prime }`$ is a rational normal quintic, and $`S^{\prime \prime }=S_8^{\prime \prime }X^{\prime \prime }`$ is the tangent scroll to $`C^{\prime \prime }`$.
In order to use the proof of Lemma (A) for $`g=5`$ we have to see whether $`X_8^{\prime \prime }𝐏^6`$ is, in fact, a complete intersection of three quadrics. If $`X^{\prime \prime }`$ were nonsingular then the classification of the smooth Fano threefolds will imply that $`X^{\prime \prime }`$ will be a complete intersection of three quadrics. But $`X^{\prime \prime }`$ is singular – see (3.21)(iv),(v).
However, especially in this case, $`X^{\prime \prime }`$ = $`X_8^{\prime \prime }𝐏^6`$ is still a complete intersection of three quadrics (see Theorem (6.1) (vii) in \[I1\]).
Denote by $`Sing(X)C`$ the set of singular points of $`X=X_{12}`$ on $`C`$, and let $`Sing(X^{\prime \prime })C^{\prime \prime }`$ be the set of singular points of $`X^{\prime \prime }`$ on $`C^{\prime \prime }`$.
(3.23) By the proof of Lemma (A) for $`g=5`$, the elements of $`Sing(X^{\prime \prime })C`$ are in a (1:1) correspondence with the different zeros of a homogeneous form $`F_7(s_o:s_1)`$ of degree $`7`$ (see (3.6)). Clearly, the vertex $`o`$ of $`Z^{\prime \prime }`$ lies on $`C^{\prime \prime }`$. Moreover, by (3.21)(iv), $`o`$ is a double singularity of $`X^{\prime \prime }`$. Since $`lSingX`$ = $`\mathrm{}`$, and since the tangent lines $`l^{}l`$ to $`C=C_7`$ do not intersect $`l`$, (3.21)(iv) yield that, set-theoretically: $`Sing(X)C`$ $``$ $`Sing(X^{\prime \prime })C^{\prime \prime }\{o\}`$.
Let $`F_7(s_0:s_1)=0`$ be as above. Since $`dimSingX^{\prime \prime }=0`$, the form $`F_7`$ does not vanish on $`C_7`$; and we can assume that o = (1:0) and $`F_7(0:1)0`$. Therefore if $`s=s_1/s_0`$ and $`f_7(s)=F(1:s)`$ then $`degf_7(s)=7`$. By the previous the elements of $`Sing(X)C`$ correspond to the different zeros of the polynomial $`s^m.f_7(s)=0`$, where $`m=mult_of_7(s)`$.
By the local definition of $`m`$, the integer $`m=m(o)=mult_of_7(s)`$ does not depend on the genus $`g7`$ of $`X_{2g2}`$ as well on the choice of the general tangent line $`l`$ to $`C_g`$. It can be seen that $`m=2`$, but for the proof it is enough to know that $`m2`$.
$`()`$ Lemma. $`m2`$.
Proof of $`().`$ By construction $`X^{\prime \prime }S^{\prime \prime }Z^{\prime \prime }`$ where $`S^{\prime \prime }=S_8^{\prime \prime }`$ is the tangent scroll to the rational normal quintic $`C^{\prime \prime }=C_5^{}`$ such that $`oC^{\prime \prime }`$, $`Z^{\prime \prime }`$ is a cone over a twisted cubic, and $`o`$ is the vertex of $`Z`$. Moreover $`Z^{\prime \prime }`$ is triple tangent to $`S^{\prime \prime }`$ at the tangent line $`F`$ to $`C^{\prime \prime }`$ at $`o`$. Indeed $`S^{\prime \prime }`$ is a hyperplane section of $`Z^{\prime \prime }`$ which passes through the vertex $`o`$ of $`Z`$. Therefore $`S^{\prime \prime }.Z=f_1+f_2+f_3`$ is a sum of 3 rulings of $`Z^{\prime \prime }`$. Since $`f_i`$ are rulings of $`Z`$, $`of_i`$ for $`i=1,2,3`$. Therefore any $`f_i`$ is a line on the tangent scroll $`S^{\prime \prime }`$ to $`C^{\prime \prime }`$ which passes through $`oC^{\prime \prime }`$. Therefore $`f_i=F`$ must be a tangent line to $`C^{\prime \prime }`$ at $`o`$, i.e. $`S^{\prime \prime }.Z^{\prime \prime }=3F`$.
By Theorem 9.9 in \[I3\], the general complete intersection $`X_8𝐏^6`$ of three quadrics, containing a cone $`Z_3`$ over a twisted cubic, is a projection of $`X_{12}`$ from a line $`l`$ such that $`N_{l/X_{12}}=𝒪(1)𝒪(2)`$. The inverse of the projection $`\pi _l`$ is defined by the linear system $`H+Z_3`$, where $`H`$ is the hyperplane section of $`X_8`$.
Let $`X_8Z_3S_8`$ be as above. Then $`X_{12}`$ will contain a tangent scroll $`S_{12}`$ to a rational normal curve $`C_7`$, and $`l`$ will be a tangent line to $`C_7`$. Therefore any $`X_8Z_3S_8`$ will be a deformation of a projection of $`X_{12}S_{12}`$ from a tangent line to $`C_7`$.
It rests to see that $`m(X_8)=mult_of_72`$ for $`f_7`$ corresponding, as above, to some particular such $`X_8`$.
Example. Let $`𝐏^5(x)=𝐏^5(x_0:\mathrm{}:x_5)`$, and let $`q_0=x_0x_4+4x_1x_33x_2^2`$, $`q_1=x_0x_5+3x_1x_42x_2x_3`$, $`q_2=x_1x_5+4x_2x_43x_3^2`$. Then $`S_8=(q_0=q_1=q_2=0)𝐏^5(x)`$ will be the tangent scroll to the rational normal quintic $`C_5:x_i=s_0^{5i}s_1^i(0i5)`$.
Let $`X_8=(Q_0=Q_1=Q_2=0)𝐏^6(x:u)=𝐏^6(x_0:x_1:x_2:x_3:x_4:x_5:u)`$, where
$`Q_0=q_0+L_o(x_4:x_5)u`$,
$`Q_1=q_1+(12x_1+L_1(x_4:x+5))u`$,
$`Q_2=q_2+(27/2x_2+L_2(x_4:x_5))u`$,
$`L_0`$, $`L_1`$ and $`L_2`$ being linear forms of $`(x_4:x_5)`$. Evidently $`X_8(u=0)=S_8`$.
Let $`𝐏^4=𝐏^4(x_0:x_1:x_2:x_3:u)𝐏^6`$, and let $`Z_3=X_8𝐏^\mathrm{𝟒}`$. Then $`Z_3=(P_0=P_1=P_2=0)𝐏^4`$, where $`P_0=x_1x_3/3x_2^2/4`$, $`P_1=x_1ux_2x_3/6`$, $`P_2=x_2ux_3^2/9`$.
Therefore $`Z_3`$ is a cone with center $`o=(1:0:\mathrm{}:0)C_5`$ over the twisted cubic curve $`C_3=Z_3(x_0=0)`$, $`C_3:(x_1:x_2:x_3:u)=(t_0^3:2t_0^2t_1:3t_0t_1^2:t_1^3)`$. Let $`s=s_1/s_0`$, and we may suppose that the point $`(0:\mathrm{}:0:1)C_5`$ is not a singular point of $`X_8`$. Then, by (3.6), the equation of $`(SingX_8)|_{C_5}`$ is
$`f_7(s)=s^2Q_0/u(1:s:\mathrm{}:s^5)sQ_1/u(1:s:\mathrm{}:s^5)+Q_2/u(1:s:\mathrm{}:s^5)`$
= $`s^2L_0(s^4,s^5)s(12s+L_1(s^4,s^5))+(27/2s^2+L_2(s^4,s^4))=3/2s^2+o(s^3)`$,
where $`o(s^3)`$ is a sum of terms of degree $`3`$. Therefore $`m(X_8)=mult_of_7(s)=2`$. q.e.d.
(3.24) Let $`X`$ = $`X_{12}S_{12}`$ be general. Since $`m2`$ then $`degs^mf_7(s)`$ $`72=5`$ = $`12g(X_{12})>0`$. In particular $`g(s):=s^mf_7(s)`$ is not a constant. Since $`g(0)0`$, and since the elements of $`Sing(X)C`$ are in a (1:1) correspondence with the different zeros of $`g(s)=s^mf_7(s)`$ (see above), then $`X_{12}`$ must be singular, which contradicts the initial assumption. This proves Lemma (A) for $`g=7`$.
Proof of Lemma (A) for g = 9.
(3.25) Let $`X_{16}𝐏^{10}`$ contains the tangent scroll $`S=S_{16}`$ to the rational normal curve $`C=C_9`$, and suppose that nevertheless $`X_{16}`$ is smooth.
Let $`LX_{16}`$ be a tangent line to $`C`$, and consider the double projection $`\pi =\pi _{2L}`$ of $`X`$ from the line $`L`$, i.e. $`\pi `$ is the rational map on $`X`$ defined by the non-complete linear system $`|𝒪_X(12L)|`$. Since $`X=X_{16}`$ is assumed to be smooth then, by §2 in \[I2\]:
$`()`$. $`\pi =\pi _{2L}`$ sends $`X`$ birationally to $`𝐏^3`$. Moreover, on $`𝐏^3`$ there exists a smooth irreducible curve $`C=C_7^3`$ of genus $`3`$ and degree $`7`$, which lies on a unique cubic surface $`S_3𝐏^3`$, and such that the inverse to $`\pi `$ birational map $`\varphi :𝐏^3X`$ is given by the non-complete linear system $`|𝒪_{𝐏^3}(72C)|`$.
By $`()`$, the proper image $`\pi (H)`$ of any hyperplane section $`HX`$ is an irreducible component of an effective divisor $`S_7|𝒪_{𝐏^3}(72C)|`$. If moreover $`H`$ contains the line $`L`$ but $`H|𝒪_X(12L)|`$ (for example if $`H=S_{16}`$) then $`\pi (H)𝐏^3`$ will be a quartic surface containing the curve $`C=C_7^3`$ (see the proof of the Main Theorem in §2 of \[I2\]), and in this case $`S_7=\pi (H)+S_3|𝒪_{𝐏^3}(72C)|`$.
Therefore $`S_4:=\pi (S_{16})`$ is a quartic surface in $`𝐏^3`$ containing the curve $`C=C_7^3`$. Moreover, the double projection $`\pi `$ sends the general tangent line $`L^{}`$ to $`C_9`$ to a tangent line $`\pi (L^{})`$ to the proper image $`\pi (C_9)`$; and since $`C_9𝐏^1`$ then $`\pi (C_9)`$ is rational. Therefore the quartic surface $`S_4=\pi (S_{16})`$ is the tangent scroll to the rational curve $`\pi (C_9)𝐏^3`$. The last is only possible if $`\pi (C_9)=C_3`$ is a twisted cubic and $`S_4`$ is the tangent scroll to $`C_3`$, and we shall see that this is impossible.
The surface $`S_4𝐏^3`$ is the tangent scroll to the twisted cubic $`C_3`$. Then, by Lemma 1.6 and p. 498 in \[MU\], the normalization of $`S_4`$ is the quadric $`𝐏^1\times 𝐏^1`$, and the map $`\nu :𝐏^1\times 𝐏^1S_4`$ is given by a linear system of bidegree $`(1,2)`$.
Let $`\mathrm{\Gamma }𝐏^1\times 𝐏^1`$ be the proper transform of $`C_7^3`$, and let $`(a,b)`$ be the bidegree of $`\mathrm{\Gamma }`$. Therefore $`7=deg(C_7^3)=2a+b`$, and $`3=g(C_7^3)=g(\mathrm{\Gamma })=(a1)(b1)`$. Obviously, these two equations for the integers $`a`$ and $`b`$ have no integral solutions – contradiction.
Therefore $`X_{16}S_{16}`$ can’t be smooth, which proves Lemma (A) in case $`g=9`$.
This completes the proof of Lemma (A).
Carmen Schuhmann University of Leiden, P.O. Box 9512
2300 RA Leiden, The Netherlands
Atanas Iliev Institute of Mathematics, Bulgarian Academy of Sciences
Acad. G. Bonchev Str. 8, 1113 Sofia, Bulgaria
|
warning/0005/hep-lat0005006.html
|
ar5iv
|
text
|
# JLAB-THY-00-10 (corrected version) On The Origin of the OZI Rule in QCD
## I Background
The phenomena which led to the formulation of the OZI rule have had a definitive impact on our understanding of strong interactions. The fact that “aces” (i.e., quarks) led to a simple interpretation of the properties of the $`\varphi `$ meson was clearly a very important clue for Zweig since it was natural for the $`\varphi `$ to be pure $`s\overline{s}`$ and for certain $`\varphi `$ production cross sections to be small so long as “hairpin graphs” were dynamically suppressed (see Fig. 1).
The dynamics behind the suppression of hairpin graphs in QCD has remained unexplained. The phenomenology of meson mixing angles in QCD-based quark models was described in the mid-1970’s in a number of papers . In such models, processes with the quark line topology of the double-hairpin graphs of Fig. 2(b) (but with arbitrary time orderings) modify the quark-antiquark transition amplitudes from the totally flavor diagonal form associated with the “scattering” quark line topology of Fig. 2(a), namely
$$𝐓=\left[\begin{array}{cccc}𝐒& \mathrm{𝟎}& \mathrm{𝟎}& \mathrm{𝟎}\\ \mathrm{𝟎}& 𝐒& \mathrm{𝟎}& \mathrm{𝟎}\\ \mathrm{𝟎}& \mathrm{𝟎}& 𝐒& \mathrm{𝟎}\\ \mathrm{𝟎}& \mathrm{𝟎}& \mathrm{𝟎}& 𝐒\end{array}\right],$$
(1)
(for illustrative purposes we have suppressed all space-time labels and specialized to the case of $`SU(2)`$ flavor where the matrix spans the basis $`u\overline{d}`$, $`d\overline{u}`$, $`u\overline{u}`$, $`d\overline{d}`$) by the addition of the annihilation amplitudes $`A`$
$$𝚫𝐓=\left[\begin{array}{cccc}0& 0& 0& 0\\ 0& 0& 0& 0\\ 0& 0& A& A\\ 0& 0& A& A\end{array}\right].$$
(2)
Using this framework , it was noted that the OZI mixing amplitude $`A`$ characterizing Fig. 2(b) was of order 10 MeV in the established meson nonets, with the sole exception of the ground state pseudoscalar meson nonet, where $`A`$ is an order of magnitude larger. These observations were consistent with the pattern one would expect for heavy quarkonia where the ground state pseudoscalar double-hairpin is larger than the vector double-hairpin by one factor of $`(\alpha _s/\pi )^1`$, and excited state double-hairpins are suppressed by having vanishing wave functions at $`\stackrel{}{r}=0`$. However, an explanation for this pattern in light quark systems was lacking.
The large size of the ground state pseudoscalar double-hairpin is a manifestation of the “$`U_A(1)`$ problem” : the equations of motion of QCD, taken naively, would imply that spontaneous chiral symmetry breaking leads to nine and not just eight Goldstone bosons , but the large mass of the $`\eta ^{}`$ seems to disqualify it from the role of the flavor singlet Goldstone boson. However, the $`U_A(1)`$ current is anomalous, and by the late 1970’s it was understood through the study of instantons that the anomaly leads to a nonconservation of the $`U_A(1)`$ charge and thereby to the evasion of Goldstone’s theorem in the flavor singlet channel when chiral symmetry is spontaneously broken. The connection between the quark model picture of double-hairpins and instantons was discussed by Witten , Veneziano , and others, who explored more generally the conflict between instantons and the large $`N_c`$ expansion .
(The reader familiar with instanton lore may be puzzled by the connection between the annihilation amplitudes $`A_{OZI}^{0^+}`$ of Eq. (2) in the pseudoscalar mesons and instanton-induced effects in the pseudoscalar mesons. The latter effects are associated with the ’t Hooft interaction which (in our illustrative $`SU(2)`$ flavor case) leads to $`u\overline{u}d\overline{d}`$ and $`d\overline{d}u\overline{u}`$ but not the diagonal entries in Eq. (2) for $`\mathrm{\Delta }T`$ corresponding to $`u\overline{u}u\overline{u}`$ or $`d\overline{d}d\overline{d}`$ transitions. Recall, however, that the ’t Hooft interaction also has $`u\overline{d}u\overline{d}`$ and $`d\overline{u}d\overline{u}`$ interactions, i.e., the $`S`$-like amplitudes of Eq. (1). Thus the instanton-induced interactions also admit the decomposition of Eqs. (1) and (2) with $`S=A`$. We will elaborate upon this point below.)
The large $`N_c`$ expansion is the only known field-theoretic basis for the general success of the valence quark model, Regge phenomenology, the observed narrowness of resonances, and the OZI rule. In particular, the OZI-violating meson mixing amplitudes of Fig. 2(b) are all of order $`1/N_c`$. Ironically, such a suppression of these amplitudes seems perfectly consistent with the effects in the pseudoscalar mesons, but not strong enough to account for the extremely small amplitudes seen in other nonets. See the second column of Table I.
The unexpected suppression of most OZI-violating amplitudes beyond a simple factor of $`1/N_c`$ is elevated from a dynamical puzzle to a paradox when the various time-orderings of Fig. 2(b) are projected into a hadronic basis. In such a basis, flavor mixing could arise through an intermediate glueball, through an instantaneous interaction, or via a hadronic loop process in which Fig. 2(b) has the time-ordering shown in Fig. 3. The paradox arises from the observation that these OZI-violating hadronic loop processes can proceed by sequential OZI-allowed vertices with known and unsuppressed strengths. These hadronic loop diagrams may be associated with contributions to meson propagators arising from second order (real and virtual) decay processes, and as such are of order $`(1/\sqrt{N_c})^2`$, as expected. This factor of $`1/N_c`$ is also perfectly consistent with the observation that the imaginary parts of these propagators give the $`1/N_c`$-suppressed meson widths which are generally of order of hundreds of MeV. Nevertheless, OZI phenomenology requires that the $`1/N_c`$-suppressed real parts (from the full meson spectrum and not just the kinematically allowed part) be an order of magnitude smaller. Explicit model calculations substantiate the generic result that individual hadronic channels of the type depicted in Fig. 3 would indeed contribute hundreds of MeV to OZI-violating meson mixing. Thus even if the other possible sources of OZI violation (from the other time-orderings) were dynamically suppressed, these hadronic loop diagrams would seem to spoil the OZI rule. This rule requires that $`A_{OZI}<<m_sm_d`$ so that the nonet and not the $`SU(3)`$ limit is realized. Thus it is a necessary (but not sufficient) condition for the OZI rule that there be some conspiracy between hadronic loop processes which suppresses them below their expected $`1/N_c`$ strength .
## II A Proposed Resolution
The authors of Ref. proposed a resolution of this paradox. They examined the OZI-violating amplitudes $`A_{OZI}`$ in non-pseudoscalar channels from the complete tower of hadronic loop processes to determine if “miraculous” cancellations between the hundred-MeV-scale real parts of individual channels could be responsible for the suppression of the sum over channels beyond a simple power of $`1/N_c`$. To make such a calculation one must have a complete model for meson trilinear vertices since, if such a conspiracy is to occur, it will have to be based on an underlying pattern of coupling strengths and signs. The nonrelativistic quark model is complete in this sense: using the standard $`{}_{}{}^{3}P_{0}^{}`$ pair creation operator for hadronic decays by flux tube breaking and valence quark model wave functions, all trilinear vertices and their associated form factors are prescribed. Since the model is a nonrelativistic one, only the time-ordering of Fig. 3 with a two meson intermediate state can be calculated in this way, but within this framework Ref. shows that in general a “miraculous” cancellation between channels does indeed occur. This cancellation occurs between groups of intermediate meson states that might have been difficult to anticipate a priori. Consider the prototypical case of $`\omega \varphi `$ mixing where $`\omega (AB)_L\varphi `$ with $`L`$ the $`AB`$ relative angular momentum. The intermediate states contributing to Fig. 3 are $`(K\overline{K})_P`$, $`(K\overline{K}^{})_P`$, $`(K^{}\overline{K}^{})_P`$, $`(K^{}\overline{K}_0^{})_S`$, $`(K\overline{K}_{a_1})_S`$, $`(K\overline{K}_{a_1})_D`$, $`(K^{}\overline{K}_{a_1})_S`$, $`(K^{}\overline{K}_{a_1})_D`$, $`(K\overline{K}_{b_1})_S`$, $`(K\overline{K}_{b_1})_D`$, $`(K^{}\overline{K}_{b_1})_S`$, $`(K^{}\overline{K}_{b_1})_D`$, $`(K\overline{K}_2^{})_D`$, $`(K^{}\overline{K}_2^{})_S`$, $`(K^{}\overline{K}_2^{})_S`$, … where $`K`$ and $`K^{}`$ are the ground state ($`\mathrm{}=0`$) pseudoscalar and vector mesons, and $`K_0^{}`$, $`K_{a_1}`$, $`K_{b_1}`$, and $`K_2^{}`$ are the first excited state ($`\mathrm{}=1`$) strange mesons with $`J^P=0^+,1^+,1^+`$, and $`2^+`$ which would be associated with the $`a_0`$, $`a_1`$, $`b_1`$, and $`a_2`$ octets in the $`SU(3)`$ limit. (Note that the ellipsis denotes more highly excited intermediate states, including ones in which each leg of the intermediate state is excited, and that charge conjugate intermediate states are implied.) As expected on the basis of the previously described arguments, a typical channel in this sum contributes of order 100 MeV to $`A_{OZI}^1^{}`$. However, intermediate states with the same total orbital angular momentum but opposite values of $`(1)^L`$ tend to cancel. Thus, for example, the $`(\mathrm{}_A=0,\mathrm{}_B=0)_P`$ channels with $`L_{total}\mathrm{}_A+\mathrm{}_B+L=1`$ all have the same sign, but they strongly cancel against the $`(\mathrm{}_A=0,\mathrm{}_B=1)_S`$+$`(\mathrm{}_A=1,\mathrm{}_B=0)_S`$ channels!
The calculation is formidable. With standard quark model parameters the form factors are quite hard and complete convergence is achieved only after summing of order 10 thousand channels, corresponding to $`L_{total}10`$. With reasonable variations of standard parameters the contribution of an individual channel waxes and wanes, as does the speed of convergence. However, the underlying mechanism of the cancellation is simple and very robust: $`A_{OZI}^1^{}`$ is much smaller than its component pieces because of an approximate “spectator plus closure limit”. This limit is illustrated in Fig. 4, which shows the standard $`{}_{}{}^{3}P_{0}^{}`$ operator with $`J^{PC}=0^{++}`$ trying to create and then annihilate quark-antiquark pairs with $`J^{PC}=1^{}`$. If a single two meson intermediate state is inserted into this diagram, it will project out pieces of this amplitude of order $`1/N_c`$ as expected, but if the original (final) $`q\overline{q}`$ pair does not distort the $`J^{PC}`$ of the produced (annihilated) pair (the spectator approximation) a complete set of intermediate states with a common energy denominator (the closure approximation) will give zero amplitude. Ref. shows that deviations from this “spectator plus closure limit” are naturally small, leading to the observed order of magnitude suppression of the loop contribution to $`A_{OZI}^1^{}`$ relative to $`1/N_c`$ expectations. See Table I. The interested reader is referred to Ref. for a detailed explanation of the resiliency of this limit. This quark model solution to the “second order paradox” associated with the OZI rule also appears to justify the conspiracies between Regge trajectories required to explain the suppression of cross sections requiring “exotic” exchanges (e.g., those with isospin 2) . Since “exotic” exchanges can occur by double Regge exchanges (analogous to the second order loop processes), only a conspiracy between exchanges (analogous to the conspiracy between loops) can give the observed suppression of such cross sections.
While the order of magnitude suppression of the loop contribution to $`A_{OZI}^1^{}`$ is robust, the contribution of individual channels and the residue after the cancellations have occurred is model sensitive, so a prediction for the actual value of this amplitude cannot be made. This is not a great loss, however, since the accuracy of the model is very suspect: its dynamics is nonrelativistic, and it has ignored the $`Z`$-graph time orderings of Fig. 3. More significantly, any such amplitude would need to be added to the unknown pure glue and instantaneous contributions to the $`q\overline{q}q^{}\overline{q}^{}`$ transition before being compared to experiment. Thus the important conclusion of Ref. is the qualitative one that the “second order paradox” can be evaded.
Ref. confirms that $`A_{OZI}`$ from meson loop diagrams is small in not only the vector mesons but in all other well-established nonets: those with $`J^{PC}=2^{++},1^{++},1^+,3^{}`$, and $`4^{++}`$. The key, of course, is that the nonet $`J^{PC}`$ must differ from that of the $`{}_{}{}^{3}P_{0}^{}`$ pair creation operator . From this simple requirement follows a rather spectacular prediction: OZI violation should be very strong in the scalar meson nonet.
The scalar mesons, and especially the isoscalar scalar mesons which would display the effects of OZI violation, have been notoriously difficult to understand experimentally. Over the last thirty years the mass of the lightest isoscalar scalar meson quoted by the Particle Data Group has varied between 400 and 1400 MeV, while the quoted width has varied between and 100 and 1000 MeV. (We have removed the $`f_0(980)`$ from this compilation under the presumption that it is a $`K\overline{K}`$ molecule, or this spread of values would be even wider.) The experimental status of the scalar meson nonet becomes even more obscure when one recalls that the lightest glueball is expected to have $`J^{PC}=0^{++}`$ and a mass around 1.5 GeV. One can only say with confidence that the experimental situation does not exclude that $`A_{OZI}^{0^{++}}`$ is large.
Fortunately, there is an alternative to checking this prediction of the quark model mechanism against experiment. We can check it against calculations from lattice QCD.
## III OZI on the Lattice
### A OZI Violation in the Quenched Approximation
Matrix elements of the type $`0|T[\overline{q}^{}(y)\mathrm{\Gamma }^{J^{PC}}q^{}(y)\overline{q}(x)\mathrm{\Gamma }^{J^{PC}}q(x)]|0`$, where $`\mathrm{\Gamma }^{J^{PC}}`$ carries space-time indices which determine the $`J^{PC}`$ of the propagator being studied and $`q^{}q`$, describe OZI violation in mesons. In leading order in $`1/N_c`$, such processes can proceed through diagrams of the type depicted in Fig. 2(b) (of which Fig. 3 is the particular time-ordering relevant to the hadronic loop diagrams), i.e., they receive leading contributions in the quenched approximation in which internal quark-antiquark loops are ignored. (Of course the accuracy of the quenched approximation can be questioned, but this is irrelevant to the main points of this paper, including checking a prediction of the quark model in which internal quark loops are also neglected.)
In the absence of OZI violation, the $`\omega `$-like $`\frac{1}{\sqrt{2}}(u\overline{u}+d\overline{d})`$ and $`\varphi `$-like $`s\overline{s}`$ sectors are segregated and each develops its own tower of meson excited states of each allowed $`J^{PC}`$. If the OZI-violating amplitudes $`A_{OZI}^{J^{PC}}`$ in that channel are small, then in leading order they simply shift the masses of each state by $`A_{OZI}^{J^{PC}}`$ and create $`\omega \varphi `$-like mixing with a mixing angle $`A_{OZI}^{J^{PC}}/\mathrm{\Delta }m`$ where $`\mathrm{\Delta }m`$ is the unperturbed mass difference between the $`\omega `$\- and $`\varphi `$-like states being mixed. In such circumstances the empirical value of $`A_{OZI}^{J^{PC}}`$ may be extracted from either the $`\omega \rho `$-like mass difference or the $`\omega \varphi `$-like mixing angle and compared directly with the quenched lattice amplitudes since the latter may be construed as correctly representing OZI-violation in the quenched approximation in lowest order perturbation theory in $`A_{OZI}^{J^{PC}}`$.
If $`A_{OZI}^{J^{PC}}`$ is strong, as in the pseudoscalar channel, the situation is more complicated. In such circumstances two new effects come into play: the masses of $`\omega `$\- and $`\varphi `$-like states can be shifted strongly, so that their mixing angle may not be determined by their unperturbed mass difference, and treating the mixed propagator from $`q\overline{q}q^{}\overline{q}^{}`$ in lowest order in $`A_{OZI}^{J^{PC}}`$ may not be valid. The former effect is straightforward, but the latter can be complex. For example, a higher order treatment of $`A_{OZI}^{J^{PC}}`$ appears to be inconsistent with the quenched approximation, as shown in Fig. 5. However, the process depicted in Fig. 5 is one of a series of processes with internal quark loops which arise from repeated iteration of the quenched amplitude. Their effect and that of the diagonal mass shifts is to create a propagator matrix with entries corresponding to the quenched approximation; when diagonalized perturbatively this matrix gives the masses and mixing angles for weak OZI violation, but for strong OZI violation it may be diagonalized exactly, thereby summing the series of sequential applications of $`A_{OZI}^{J^{PC}}`$. Another closely related possible complication is that a large $`A_{OZI}^{J^{PC}}`$ can create strong mixing with the glueball sector, requiring that the propagator matrix be enlarged yet further.
For the pseudoscalar mesons, the preceeding discussion of the effects of a large $`A_{OZI}`$ are particularly significant. In the chiral limit with $`A_{OZI}^{0^+}=0`$, the $`U_A(1)`$ meson - - - the $`\eta ^{}`$ \- - - is also massless. As a result, the quenched OZI-violating amplitude of Fig. 3 will give $`A_{OZI}^{0^+}`$ sandwiched between two massless propagators, i.e., it will give an $`\eta ^{}`$ propagator that looks nothing like that of a massive, $`SU(3)`$-flavor-mixed $`\eta ^{}`$. In this case, to even qualitatively relate the quenched amplitudes to nature one must extract $`A_{OZI}^{0^+}`$ from them and add these amplitudes to the propagator matrix (the broken $`SU(3)`$ analog of Eq. (1)) which one diagonalizes exactly. The resulting full propagator will have a massive $`SU(3)`$-flavor-mixed $`\eta ^{}`$ which sums the single particle effects of $`A_{OZI}^{0^+}`$ to all orders. Further numerical support for this interpretation of the quenched pseudoscalar double-hairpin comes from the shape of the double-hairpin propagator as a function of Euclidean time. As discussed below (see also Ref. ), this time dependence can be fit very well to the functional form $`(1+m_\pi t)\mathrm{exp}(m_\pi t)`$ expected from a mass insertion vertex surrounded by two propagators of mass $`m_\pi `$.
### B Methods
The ability to study the double-hairpin diagrams relevant to the OZI rule has been greatly improved by two recent developments in lattice QCD methodology. The global source technique (which we refer to as the “allsource” method) was introduced several years ago for the purpose of studying the $`\eta ^{}`$ mass and the $`U_A(1)`$ anomaly. In this method, the quark propagator is calculated from a sum of identical unit color-spin sources located at all space-time points on the lattice. If this allsource propagator is contracted over color indices at a given site, the result is a gauge invariant term corresponding to a closed quark loop originating from that site, plus a very large number of gauge-dependent open loops. The latter terms tend to cancel due to their random phases, allowing a determination of closed loop averages and loop-loop correlators (double-hairpins). The other recently developed technique which has greatly improved the accuracy of the results for both double-hairpin calculations and for other chiral studies with Wilson-Dirac fermions is the Modified Quenched Approximation (MQA) , which provides a practical resolution of the exceptional configuration problem that has long plagued such calculations. This method identifies the source of the exceptional configuration problem as the presence, in some gauge configurations, of exactly real eigenmodes which are displaced into the physical mass region by the artificial chiral symmetry breaking associated with the lattice Wilson-Dirac operator. By systematically identifying these real eigenmodes and calculating their contribution to the quark propagators, the corresponding propagator poles can be extracted and moved to zero quark mass (where they belong). This MQA procedure has been applied to both the allsource propagators for double-hairpin calculations as well as to valence quark propagators. The resulting MQA-improved propagators have recently been used in an extensive study of quenched chiral logs and their relation to the $`\eta ^{}`$ mass and the $`U_A(1)`$ anomaly . As a part of this study, the size and time-dependence of the pseudoscalar $`\eta ^{}`$ double-hairpin diagram was calculated, using the allsource method. Since the pole-shifting procedure has already been applied to the quark propagators, it requires very little additional effort to investigate the vector, axial-vector, and scalar double-hairpins which determine the spin-parity pattern of OZI mixing. The results we present here are from a set of 300 quenched gauge configurations on a $`12^3\times 24`$ lattice at $`\beta =5.7`$. In the study of Refs. , both naive Wilson and clover improved quark actions were studied. It was found that, at $`\beta =5.7`$ with the Wilson action, substantial lattice spacing effects suppressed the pseudoscalar double-hairpin, giving a smaller than expected value of $`𝐀_{OZI}^{0^+}=`$(0.27 GeV)<sup>2</sup> for the double-hairpin contribution to the $`\eta ^{}`$ mass versus the value (0.49 GeV)<sup>2</sup> extracted from weak-$`SU(3)`$-breaking mass formulas . A much more satisfactory result is obtained from the clover improved quark action. With a clover coefficient $`C_{sw}=1.57`$, the pseudoscalar double-hairpin gives $`𝐀_{OZI}^{0^+}=`$(0.41 GeV)<sup>2</sup>. For the calculation of OZI-violating amplitudes, we will therefore use the clover improved quark action only; we also use the physical charmonium 1S-1P splitting to set the scale ($`a^1=1.18`$ GeV) for $`\beta =5.7`$ when we quote lattice results in physical units .
### C Results
Using the method described in the previous Section, we have calculated the double-hairpin contribution to matrix elements of the form
$$\overline{q}^{}(y)\mathrm{\Gamma }^iq^{}(y)\overline{q}(x)\mathrm{\Gamma }^iq(x)$$
(3)
with Hermitian operators generated by the choices $`\mathrm{\Gamma }^i=i\gamma _5`$ (pseudoscalar), $`\mathrm{\Gamma }^i=\gamma ^\mu `$, $`\mu =1,2,3`$ (vector), $`\mathrm{\Gamma }^i=\gamma ^\mu \gamma _5`$, $`\mu =1,2,3`$ (axial vector) and $`\mathrm{\Gamma }^i=1`$ (scalar) (the antisymmetric tensor $`\sigma ^{\mu \nu }`$ does not explore new states: it also has axial vector quantum numbers). As in standard hadron spectroscopy, we Fourier transform the space-time propagator over 3-dimensional time slices at zero 3-momentum and study its time-dependence. A particular advantage of the allsource method is that the Fourier transforms can be performed over both ends of the meson propagator, unlike the usual case of a fixed local source where only one end can be transformed. This provides an improvement in statistics which is quite important for the success of the method. For the scalar double-hairpin matrix element, the expectation value of a single scalar loop is nonzero, and so a constant proportional to $`0|\overline{q}q|0^2`$ must be subtracted from the above matrix element to get the true correlator.
Even without any detailed analysis, the overall empirical OZI pattern of Table I is strikingly confirmed by the lattice results. This is easily seen from the size of the various double-hairpin correlators. In Figs. 6-9, we have plotted the double-hairpin correlators for the pseudoscalar, vector, axial vector, and scalar sources. All plots have the same scale for comparison. The calculations have been done for 9 different choices of quark mass. The data shown in the figures are from one of the lightest quark masses, for which the pion mass is about 300 MeV ($`m_\pi a=0.266\pm 0.004`$). The results quoted in Table II are chirally extrapolated to the physical pion mass. The errors in Figs. 6-9 and in Table II are statistical only. By far the largest and longest-range correlator is the pseudoscalar correlator of Fig. 6. This is expected for two reasons: the anomaly introduces a large double-hairpin vertex responsible for the large $`\eta ^{}`$ mass, and, as explained above, in the quenched approximation the external $`\overline{q}q`$ meson propagators on either side of the double-hairpin vertex are light Goldstone bosons. The results extracted from Fig. 6 have been reported in Ref .
Compared to the very strong pseudoscalar double-hairpin, the vector and axial vector double-hairpins of Figs. 7 and 8 are dramatically suppressed, consistent with the empirical observations described in Section I. Since quenched lattice QCD gives reasonable values for the three-point functions associated with the meson virtual loop processes depicted in Fig. 3, these results provide not only a first derivation of the OZI rule from QCD, but also a dramatic example of the evasion in QCD of the “second order paradox” described in Section I and a confirmation of the fact that in a complete calculation a conspiracy of the type described in Section II must occur. (Of course the results reported here include not only the meson loop contributions but also the other time orderings of the double-hairpin graphs of Fig. 2(b).) We in fact see no significant signals in the vector and axial vector channels and so report in Table II only one standard deviation upper bounds.
As described in Section II and illustrated in Fig. 4, if the conspiratorial cancellation amongst meson loops is associated with $`{}_{}{}^{3}P_{0}^{}`$ pair creation, one would expect $`A_{OZI}^{0^{++}}`$ to be very large. Fig. 9 shows this behaviour: after taking into account the heavier mass of the scalar meson (about 1.3 in lattice units ), we find that the scalar OZI amplitude is comparable in size to the pseudoscalar amplitude but of the opposite sign (see Table II). A full amplitude $`A_{OZI}`$ in general has glueball, instantaneous, and loop contributions, and in a given amplitude, any or all of these components might be important. (Recall, for example, that while the loop contribution to $`A_{OZI}^{0^+}`$ is believed to be small , the full $`A_{OZI}^{0^+}`$ is large.) That the measured $`A_{OZI}^{0^{++}}`$ is actually consistent in sign and magnitude with the hadronic loop contribution predicted by the quark model has interesting implications which we will discuss below. A large and negative $`A_{OZI}^{0^{++}}`$ has been previously reported in Ref. .
To obtain the quantitative results for the OZI mixing amplitudes quoted in Table II, we carried out an analysis similar to that used to obtain the $`\eta ^{}`$ mass from the pseudoscalar double-hairpin . For that case, the time-dependence of the pseudoscalar double-hairpin correlator corresponding to Fig. 2(b) was found to be quite well described by a “double-pole” form consisting of a $`p^2`$-independent double-hairpin insertion between a pair of meson propagators (see also Ref. ). In momentum space
$$\stackrel{~}{\mathrm{\Delta }}_h(p)=f_P\frac{1}{p^2+m_\pi ^2}𝐀_{OZI}^{0^+}\frac{1}{p^2+m_\pi ^2}f_P$$
(4)
where $`f_P`$ is the vacuum-to-one-particle matrix element
$$f_P=0|\overline{q}i\gamma ^5q|\pi (p)$$
(5)
and $`𝐀_{OZI}`$ is the (mass)<sup>2</sup> version of the $`A_{OZI}`$ defined previously (called $`m_0^2`$ in Refs. ). This gives a time-dependent double-hairpin correlator at zero 3-momentum of the form
$$\mathrm{\Delta }_h(𝐩=0;t)=\frac{f_P^2𝐀_{OZI}^{0^+}}{4m_\pi ^3}(1+m_\pi t)e^{m_\pi t}+(t(Nat))$$
(6)
to be compared to the usual valence quark (e.g., isovector) correlator corresponding to Fig. 2(a)
$$\stackrel{~}{\mathrm{\Delta }}_v(p)=f_P\frac{1}{p^2+m_\pi ^2}f_P$$
(7)
which gives
$$\mathrm{\Delta }_v(𝐩=0;t)=\frac{f_P^2}{2m_\pi }e^{m_\pi t}+(t(Nat)).$$
(8)
(The relative sign of Eqs. (6) and (8) is tricky; with our convention a positive $`𝐀_{OZI}`$ makes a positive contribution to the (mass)<sup>2</sup> of a state.) Since the values of $`f_P`$ and $`m_\pi `$ can be
separately determined from fitting Eq. (8) to the valence quark correlator, the double-hairpin vertex insertion $`𝐀_{OZI}^{0^+}`$ can be determined by a one-parameter fit of (6) to the overall size of the double-hairpin correlator. A similar analysis of the scalar double-hairpin led to the result quoted in Table II, while for the other channels such analyses provided the quoted upper bounds for the very tiny mixing amplitudes in these channels.
## IV Conclusions
The most straightforward conclusions of this work are that QCD can explain the OZI rule in channels where it is observed and that it predicts that $`A_{OZI}^{0^{++}}`$ is large and negative . This supports the quark model’s explanation of the dynamical suppression of the typical scale of hadron-loop-induced OZI violation below $`1/N_c`$ expectations, and in so doing provides further evidence for the standard $`{}_{}{}^{3}P_{0}^{}`$ pair creation amplitude, since this is the critical feature which produces this result.
While the precise consequences are unclear, the implications for phenomenology are serious. With $`A_{OZI}^{0^{++}}`$ large, the lightest scalar meson nonet (the $`1P`$ states) will be close to the $`SU(3)`$ limit. We may therefore expect an octet of scalar mesons in the 1400 MeV range with the other $`1P`$ states while the nearly singlet scalar state will be substantially lower in mass. Thus the usual assumption of phenomenological analyses that this region will contain the unmixed nonet of $`1P`$ states and the scalar glueball is incorrect. For example, this region might well contain the isoscalar state of the $`2P`$ nonet. In addition, since $`A_{OZI}^{0^{++}}`$ is comparable to the $`1P2P`$ splitting, there is no reason to assume that either the $`1P`$ or $`2P`$ singlet’s properties can be related by nonet symmetry to those of its octet. The net effect is that the definitive extraction of the glueball state from the scalar meson spectrum may be quite difficult.
Given the importance of this task, it is certainly worthwhile to study the scalar mesons more carefully in the light of this result . On the lattice it might be possible to obtain the matrix of OZI-violating amplitudes connecting the $`\omega `$-like and $`\varphi `$-like $`1P`$ and $`2P`$ states; in models the low-lying scalar meson spectrum can be studied including the effects of a strong annihilation channel.
Perhaps most critical is to use quenched lattice calculations of the mixed propagators from quarkonia to glueballs to help resolve the scalar meson OZI violation reported here into the contributions of $`q\overline{q}^{}q^{}\overline{q}`$ intermediate states, purely gluonic intermediate states associated with “true double-hairpin” graphs, and instantaneous contributions. Ultimately, quenched lattice calculations of three-point functions could directly check the predicted negative loop contributions to $`A_{OZI}^{0^{++}}`$ by measuring the vertex functions which are the “raw ingredients” of the quark model calculation. In particular, in other than the $`0^{++}`$ channel, one should see the required magnitudes and opposite signs of the virtual $`P`$-wave decays to two $`\mathrm{}=0`$ mesons and the $`S`$-wave decays to one $`\mathrm{}=1`$ and one $`\mathrm{}=0`$ meson required to build up the near cancellation that is at the heart of the quark model mechanism. In contrast, for $`0^{++}`$ mesons these channels should have the same sign.
## V Discussion
The results described here clearly have serious implications for the spectroscopy of $`0^{++}`$ states, and define the series of investigations described above required to clarify the physics behind $`A_{OZI}^{0^{++}}`$. Such investigations are not only important for their impact on phenomenology, however. They are also important because our results highlight other more fundamental questions raised long ago by Witten , on the apparent conflict between the instanton solution of the $`\eta ^{}`$ mass (i.e., $`U_A(1)`$) problem and the large $`N_c`$ limit. The quark model mechanism for the loop contributions to $`A_{OZI}^{0^{++}}`$ is based on large $`N_c`$. While our discussion of the $`{}_{}{}^{3}P_{0}^{}`$ model has focused on its prescription for the quantum numbers of the created $`q\overline{q}`$ pair, it is also an essential ingredient of the model that this pair creates $`(q\overline{q}^{})+(q^{}\overline{q})`$ and not $`(q\overline{q})+(q^{}\overline{q}^{})`$ mesons, i.e., that it respects the OZI rule at tree level. The physical picture behind this feature of the model is that pair creation (at order $`1/N_c`$) occurs by the breaking of the color flux tube connecting $`q`$ and $`\overline{q}`$. More generally, as mentioned above, this limit provides the only known field-theoretic basis for the success of not only the valence quark model, but also of Regge phenomenology, the narrow resonance approximation, and many of the systematics of hadronic spectra and matrix elements . In contrast, it is widely believed that the $`U_A(1)`$ problem is solved through instanton contributions to the axial anomaly. However, as emphasized by Witten, instantons vanish like $`e^{N_c}`$ and so do not appear in the large $`N_c`$ expansion. “Insofar as \[instantons play\] a significant role in the strong interactions, the large $`N_c`$ expansion must be bad. It is necessary to choose between the two.” Note that these arguments draw an important distinction between semiclassically calculated instanton effects, which vanish like $`e^{N_c}`$, and more general topological gauge fluctuations, which can contribute at order $`1/N_c`$ to $`m_\eta ^{}`$. The real issue is not whether there are large fluctuations of $`F\stackrel{~}{F}`$ in the QCD vacuum, but whether these fluctuations arise as local semiclassical lumps with quantized winding numbers or simply as a result of the generically large gauge fluctuations of a confining vacuum.
To place this conflict in context, recall Eqs. (1) and (2). From Section III it is apparent that the amplitude for any of $`N_f`$ massless $`q\overline{q}`$ pairs to annihilate to any other pair is the same, i.e., that $`\mathrm{\Delta }T`$ does indeed have the form of the $`N_f=2`$ matrix shown in Eq. (2). As explained earlier, this is consistent with the ’t Hooft instanton interaction since the “scattering” amplitude $`S`$ in Eq. (1) contains a contribution $`A`$ from instantons. Thus to leading order in $`A`$ the decomposition of Eqs. (1) and (2) is general and the analyses of OZI violation in Refs. \- - - including that in the pseudoscalar sector - - - are valid. It follows that from a purely phenomenological perspective it is irrelevant whether or not there is an instanton contribution to hadronic physics: a phenomenology with $`A_{OZI}^{0^+}0`$ is “legal” in any case, since the anomaly allows a resolution of the $`U_A(1)`$ problem with or without instantons. What remains unclear is the physics behind the annihilation amplitudes. Since a lattice simulation sums over all paths, it contains the instantons as tunnelling events between classical vacua, but the Feynman diagrams of QCD, which represent the quantum corrections around these vacua, are incapable of representing instanton physics. Thus if instantons are important in QCD, Feynman diagrams would have to be supplemented by effective interactions (like the ’t Hooft interaction). As noted by Witten , the foremost victim of the failure of Feynman diagrams implied if instantons are important would be the large $`N_c`$ expansion, since it assumes that all-orders properties of the QCD Feynman diagrammatic expansion are properties of QCD.
The observations reported in this paper on $`A_{OZI}^{0^{++}}`$ add one more item to a growing and closely linked set of issues where the physics of instantons and the physics of large $`N_c`$ confront each other. Assuming that confinement and the Nambu-Goldstone mechanism are properties of the all-orders Feynman diagrammatic expansion of QCD, the large $`N_c`$ expansion provides a consistent framework embracing all strong interaction phenomena. Among these phenomena are the hadron spectrum for all flavors of hadrons (including the $`1/N_c`$-suppressed hadronic widths which seem to be critical to $`A_{OZI}^{0^{++}}`$), the OZI rule (now including $`A_{OZI}^{0^{++}}`$), and the $`q\overline{q}`$ condensate. As Witten argued long ago , given the $`U_A(1)`$ anomaly and confinement, the large $`N_c`$ limit is also capable of explaining the $`\eta ^{}`$ mass at order $`1/N_c`$ without instantons.
While its limited range of applicability makes it somewhat less attractive for phenomenology (instantons offer a competing explanation only for the properties of the lightest $`SU(3)_f`$ hadrons),the instanton picture has received strong support from recent lattice results . Measurements of the topological charge of “cooled” gauge configurations show that in such circumstances this charge is quantized and localized as expected for instantons. Moreover, the zero-modes of the Dirac operator associated with the solution of the $`U_A(1)`$ problem and the near-zero-modes associated with the $`q\overline{q}`$ condensate are also localized and in “cooled” configurations can be associated with these same instantons. The lattice results on these and other hadronic properties are consistent with the instanton liquid model .
Since, as argued by Witten, confinement can replace instantons as the source of the $`U_A(1)`$ anomaly and since confinement can also produce a space-time localization of the origin of the $`\eta ^{}`$ mass and of the $`q\overline{q}`$ condensate, in our view the true origin of these effects remains unsettled. The results of this paper may help to resolve this situation since for $`A_{OZI}^{0^{++}}`$ the two competing pictures lead to mechanisms that are very distinct. Flux-tube-breaking pair creation, a prototypical large $`N_c`$ phenomenon, led to the prediction that the hadron loop contribution to $`A_{OZI}^{0^{++}}`$ is large and negative as found here. Moreover, as stated in the beginning of this paper, quark models, with their confined constituent quarks, naturally generate a large positive $`A_{OZI}^{0^+}`$ . In this case the loop contribution should be typically small , and the large positive quark model amplitude is associated with an instantaneous interaction. Instantons, through the instantaneous ’t Hooft interaction, would lead to a superficially similar pattern of OZI violation: a large positive $`A_{OZI}^{0^+}`$ and a large negative $`A_{OZI}^{0^{++}}`$. However, the origins of the large negative $`A_{OZI}^{0^{++}}`$ are very different in the two cases: the instanton $`A_{OZI}^{0^{++}}`$ is associated with an instantaneous contribution while the quantitative similarity between the quark model prediction and our measured $`A_{OZI}^{0^{++}}`$ suggests that this amplitude is associated instead with the meson loop contributions.
Our result thus favors the large $`N_c`$ and not the instanton interpretation of the solution to the $`\eta ^{}`$ mass problem. Nevertheless, while suggestive, the quark model prediction is not of sufficient quantitative accuracy for this conclusion to be reliable. Fortunately, with recent advances in lattice methods and in computing power, we believe that the results we have described here can not only be improved but also understood more deeply. In particular, through the program we described of decomposing the OZI-violating amplitudes into their component parts, it should be possible to define the mechanism driving $`A_{OZI}^{0^{++}}`$. We also believe it will be particularly fruitful to define and test confinement-based interpretations of the lattice results on such quantities as the topological susceptibility, the localization of zero modes, the correlation function of the topological charge operator, and the space-time association of the $`q\overline{q}`$ condensate with the topological charge. Through such studies, the conflict between large $`N_c`$ and instanton physics can at last be resolved.
ACKNOWLEDGEMENTS
We are grateful to Stephen Sharpe and Thomas Schaefer and to Chris Michael for alerting us to a serious sign error in the first version of this paper and to important references which had escaped our attention. This work was supported by DOE contract DE-AC05-84ER40150 under which the Southeastern Universities Research Association (SURA) operates the Thomas Jefferson National Accelerator Facility. The work of H.B. Thacker was supported in part by the Department of Energy under grant DE-FG02-97ER41027.
REFERENCES
|
warning/0005/physics0005002.html
|
ar5iv
|
text
|
# References
MAGNETIC FRACTAL DIMENSIONALITY OF THE DIELECTRIC BREAKDOWN UNDER STRONG MAGNETIC FIELDS
Y. Ben-Ezra<sup>1,2,3</sup> , Yurij V. Pershin<sup>4,2</sup>, I.D.Vagner<sup>1,2</sup> and P.Wyder<sup>2</sup>
<sup>1</sup>P.E.R.I.-Physics and Engineering Research Institute at Ruppin,
Emek Hefer, 40250 Israel.
<sup>2</sup>Grenoble High Magnetic Fields Laboratory
Max-Planck-Institute für Festkorperforschung and CNRS,
166X, F-38042, Cedex 9, France.
<sup>3</sup>School of Physics and Astronomy, Tel Aviv University,
Tel Aviv 69978, Israel.
<sup>4</sup>B.I.Verkin Institute for Low Temperature Physics and Engineering,
47 Lenin Avenue 310164 Kharkov, Ukraine.
The formation of breakdown pattern on an insulating surface under the influence of a transverse magnetic field is theoretically investigated. We have generalized the Dielectric Breakdown Model (DBM) for the case of external magnetic field. Concept of the Magnetic Fractal Dimensionality (MFD) is introduced and its universality is demonstrated. It is shown that MFD saturates with magnetic fields. The magnetic field dependence of the streamer curvature is obtained. It is conjectured that nonlinear current interaction is responsible for the experimentally observed ’spider-legs’ like streamer patterns.
PACS numbers: 52.20.Dq, 52.80.Mg
Fractal properties are common to the dielectric breakdown phenomena which range from an atmospheric lightning ,, to electric treeing in polymers and are of scientific and technical importance ,. Although the actual physical processes can be quite different in these phenomena, the global properties of the resulting discharge patterns are very similar. Filamentary gas discharges on insulating surface exhibit remarkable similarities to breakdown phenomena in long gaps , e.g., to atmospheric lightning , and thus offer the possibility to perform well-defined model experiments in laboratory. The application of sophisticated diagnostic tools, such as streak cameras, high-speed oscilloscopes, and time-resolved spectroscopy , in surface discharge experiments has improved the basic understanding of how a highly conducting phase, the filamentary ”leader” channel, advances into a nonconducting medium as the surrounding gas.
The surface discharge in compressed $`SF_6`$ gas have been studied in detail by Niemeyer and Pinnekamp. The parameters were controlled in such a way that the experiment produces, to a good approximation, an equipotential channel system growing in a plane with a radial electrode from a central point. The experiment shows that the dielectric breakdown pattern has a fractal structure.
The stochastic model containing the essential features of the fractal properties of the dielectric breakdown was introduced by Niemeyer, Pietroniero, and Wiesmann (NPW) ,. The introduced stochastic model made it possible to simulate the growth of a fractal structure which resembles a Lichtenberg figure . The basic assumption of the dielectric breakdown model (DBM) was that the growth probability depends on the local field. This model naturally leads to fractal structures. In the Haussdorf dimension and other fractal properties of the structure are determined and their close relations to the other fractal structures e.g., DLA (the diffusion limited aggregation) , is shown .
The DBM method was generalized by Wiesmann and Zeller in two ways. Firstly they introduced a critical field for growth $`F_c`$. The growth probability was assumed to be proportional to the local field $`F_{loc}`$ if $`F_{loc}>F_c`$ and zero if $`F_{loc}<F_c.`$ Secondly they have introduced an internal field in the structure $`F_s`$. The potential in the structure was no longer equal to the potential $`V_0`$ of the connecting electrode but equal to $`V_0+F_ss`$, where $`s`$ is the length of the path along the structure which connects the point to the electrode.
In the recent experiment a transverse high magnetic field was applied during the discharge evolution and thus any spatial restriction of the surface discharge was avoided in order to use a locally sensitive probe.
When a rectangular high-voltage pulse of about $`500ns`$ length, $`20kV`$ amplitude, and less than $`10ns`$ risetime is applied to a point-to- plane electrode system, the electrodes being separated by a thin dielectric film ( thickness $`100\mu m`$) , a discharge propagates in the gas just above the surface of the film. The discharge pattern was recorded directly with high resolution by using a photographic film as dielectric plate.
The experiment shows the spatial evolution of a negative surface discharge in a nitrogen atmosphere as a function of the magnetic field ( Fig.2 of Ref. ). At $`B=0`$ a very bright starlike pattern develops. At moderate magnetic field (up to $`7T`$ ), the leader channels are bent and appear to have a circular shape outside the central electrode region.
The radius of curvature is of the order of $`1cm`$ at $`7T`$. The direction of the bending corresponds to the movement of electrons in crossed electric and magnetic fields. With the increase of the magnetic field the radius of curvature decreases, the channels approach each other and branching sets in. At the highest applied magnetic field of $`12T`$, circular-shaped current filaments are only found in the outmost regions of the discharge pattern where they can develop undistributed by the fields of neighboring leader channels.
One can summarize the experimental results as follows:
(1) The observed bending effect cannot be related to the movement of a single charged particles in crossed electric and magnetic fields, which should result in the curvatures of the order of magnitude of $`15\mu m`$ at $`1T`$ for the electrons, i.e., the Larmor radius $`r_c=v_c/\omega _c`$ of the cyclotron orbit of an electron withe enough kinetic energy to ionize gas molecules. It has be assumed that the macroscopic value of the radius of curvature results from the drift movement of the Larmor centers of the gyrating electrons. The average radius of curvature follows a power law, $`RB^\alpha `$ . The exponent $`\alpha `$ depends on the nature of gas.
(2) The circular shape of the current filaments is not expected to occur for our point-to-plane electrode geometry.
(3) The complexity of the discharge pattern can be controlled by the magnetic field strength. It appears that an increase in complexity is induced on the scale of the bending observed, and not on the microscopic scale of the cyclotron radius.
In what follows we modify the model introduced by Wiesmann and Zeller for the case of external magnetic field.
Let us consider a two-dimensional square lattice in which a central point represents one of the electrodes while the other electrode is modeled as a circle at large distance, which represents the geometry of the experiment. The discharge starts at the central electrode and grows by one lattice bond per growth step. A bond connects a lattice point with one of eight adjacent lattice points as it is described in Fig.1. Once a given point is connected to the discharge structure by a bond, it becomes part of the structure. The potential of the central electrode is $`0`$, the potential of a point in the structure is $`V_{i,k}=V_{l,m}+V_Rl`$, where $`V_{l,m}`$ is the potential of a points from which growth go on, $`V_R`$ is an internal field in the structure and $`l`$ is $`1`$ for bonds parallel to the grid and is $`\sqrt{2}`$ for diagonal bonds.
The growth is computed as follows: First the Laplace equation is solved with the boundary conditions determined by the electrodes and the discharge structure. Then the local field $`F_{loc}`$ between a point which is already a part of the structure $`(i,k)`$ and a new adjacent point $`(i^{},k^{})`$ is calculated:
$$F_{loc}(i,k,i^{},k^{})=\frac{\phi _{_{i^{},k^{}}}V_{i,k}}{l}$$
(1)
where $`\phi _{_{i^{},k^{}}}`$ is the solution of Laplace equation. The breakdown can occur only if the local field is greater that the critical field of growth $`F_C`$. The probability that a new bond will form between a point which is already a part of the structure and a new adjacent point $`p(i,ki^{},k^{})`$ is calculated as a function of the local field $`F_{loc}`$ between the two points:
$$p(i,ki^{},k^{})=\{\begin{array}{ccc}0& ,& F_{loc}(i,k,i^{},k^{})<F_C\\ & & \\ \frac{F_{loc}^\eta }{{\scriptscriptstyle F_{loc}^\eta }}& ,& F_{loc}(i,k,i^{},k^{})F_C\end{array}$$
(2)
where a power-low dependence with exponent $`\eta `$ is assumed to describe adequately the relation between the local field and the probability. The sum in the denominator refers to all possible growth processes. A new bond is chosen randomly with probability distribution (2) and added to the discharge pattern. With this new discharge pattern one starts again. More detailed description of this model is done in .
Let us take into account the magnetic field. The magnetic field changes the probability distribution (2). A moving particle in the magnetic field experiences the Lorentz force $`F_L\stackrel{}{V}\times \stackrel{}{H}`$ which acts perpendicularly to its velocity. Consider each step of growth like a superposition of two processes. The first step is choosing a new bond using the probability distribution (2), and the second step is taking into account the probability of deviation of the growth due to the magnetic field, $`p_H`$.
If after first step of growth, for example, the bond from the dot $`0`$ to the dot $`4`$ was chousen (Fig.1), the growth will occurs to the dot $`3`$ with the probability $`p_H`$ and to the dot $`4`$ with the probability $`1p_H`$. The new probability of growth can be written as
$$\stackrel{~}{p}(i,ki^{},k^{})=\frac{p(i,ki^{},k^{})(1p_H(i,k,i^{},k^{}))+p(i,ki^{\prime \prime },k^{\prime \prime })p_H(i,k,i^{\prime \prime },k^{\prime \prime })}{(p(i,ki^{},k^{})(1p_H(i,k,i^{},k^{}))+p(i,ki^{\prime \prime },k^{\prime \prime })p_H(i,k,i^{\prime \prime },k^{\prime \prime }))}$$
(3)
where the point $`(i^{\prime \prime },k^{\prime \prime })`$ is the neighboring point with respect to the point $`(i^{},k^{})`$ in the clockwise direction with respect to the point $`(i,k)`$ and the sum in the denominator refers to all possible growth processes. The probability 3 was used in computer simulations in place of the probability 2 with the same algorithm.
Let us turn to the probability $`p_H`$ of deviation of the growth due to the magnetic field. In our model, during the process of growth, two constant forces act on charge cariers, $`F_L`$ and $`F_H`$ (Fig.1). When the resulting force is near the dot $`3`$, $`p_H1`$; when the resulting force is near the dot $`4`$, $`p_H0`$. It is clear that the probability $`p_H`$ is proportional to $`\frac{F_L}{F_{loc}}`$ . If $`F_{loc}(i,k,i^{},k^{})<F_C`$ then the growth does not occur and $`p_H`$ should be zero. If $`\frac{F_L}{F_{loc}}>1`$ then we let $`p_H=1`$. The Lorentz force $`F_L`$ is proportional to the velocity of charge carrier. In our model we can take into account only a local velocity, which arrises in the first step of growth process due to the acceleration in the local field $`F_{loc}`$. So, the Lorentz force should be proportional to $`\sqrt{lF_{loc}(i,k,i^{},k^{})}`$.
Based on these considerations, we choose the probability of deviation of the growth due to the magnetic field, $`p_H,`$ in the following form:
$$p_H(i,k,i^{},k^{})=\{\begin{array}{ccc}0& ,& F_{loc}(i,k,i^{},k^{})<F_C\\ & & \\ \frac{F_L(i,k,i^{},k^{})}{F_{loc}(i,k,i^{},k^{})}& ,& \frac{F_L(i,k,i^{},k^{})}{F_{loc}(i,k,i^{},k^{})}1,F_{loc}(i,k,i^{},k^{})F_C\\ & & \\ 1& ,& \frac{F_L(i,k,i^{},k^{})}{F_{loc}(i,k,i^{},k^{})}>1,F_{loc}(i,k,i^{},k^{})F_C\end{array}$$
(4)
where $`F_L(i,k,i^{},k^{})=\sqrt{lF_{loc}(i,k,i^{},k^{})}H`$ and $`H`$ is the value of the magnetic field.
In our computer simulations we consider a $`500\times 500`$ lattice. The solutions of the Laplace equation were obtained by the iteration method . Before starting of each realization of growth we performed $`20000`$ iterations and after each step of growth the number of iterations was $`40`$. This procedure gives a good convergence. The number of particles in clusters was $`9000`$.
The fractal dimension was calculated by the method described in . For every realization we plotted the $`\mathrm{log}N\left(R\right)`$ versus $`\mathrm{log}R`$ where $`N\left(R\right)`$ is the number of particles belonging to the structure and being within a circle of radius $`R`$. The fractal dimension is obtained by fitting a straight line to the data scaling region. For every set of the same parameters of the model ($`H`$, $`F_C`$, $`V_R`$, $`\eta `$) we made about $`100`$ realizations. Thus the statistical fluctuations were reduced.
Let us discuss the results of these calculations.
We start our simulations with the case of the zero magnetic field and the zero values of the parameters $`Fc`$ and $`V_R`$ in order to compare our results with the results of the different authors. In the Table.1 we present the dependence of the Hausdorff dimension $`D`$ on the exponent parameter $`\eta .`$
| $`\eta `$ | $`D`$ (our results) | $`D`$ (according to ) | $`D`$ (according to ) |
| --- | --- | --- | --- |
| $`0.5`$ | $`1.89\pm 0.02`$ | $`1.89\pm 0.01`$ | $`1.92`$ |
| $`1`$ | $`1.73\pm 0.02`$ | $`1.75\pm 0.02`$ | $`1.70`$ |
| $`2`$ | $`1.6\pm 0.03`$ | $`1.6`$ | $`1.43`$ |
Table 1. Dependence of the Hausdorff dimension $`D`$ on the exponent $`\eta `$ used in the relation between probability and local field (Eq.2).
Our results are in a good agreement with the results, obtained earlier.
The example of the computer-generated discharge pattern (Lichtenberg figure ) corresponding to the following set of parameters: $`H=0,\eta =1,F_c=0`$ and $`V_R=0`$ is shown in the Fig.2. In the Fig.3 we show the computer-generated discharge pattern in the presence of the external magnetic field. The white lines in the figures 1 and 2 correspond to the leader channels. Unlike the pattern, presented in the Fig.2, the leader channels in the magnetic field are distorted and appeare to have a circular shape outside the central electrode region. The direction of the bending corresponds to the movement of electrons in crossed electric and magnetic field. At the lower left corner of the Fig.3 we plotted the cyclotron orbit of an electron. One can note that the Larmor radius of an electron is about two orders of magnitude smaller then the radius of curvature of the leader channel.
The saturation of the fractal dimensionality, with growing magnetic fields, at the value of $`D=1.67`$ , which is one of the main results of this paper, is presented in Fig. 4. The plot starts from the value of $`D=1.65`$ , in the absence of magnetic field, which is smaller than the fractal dimensionality reported in . Such difference results from the fact that in the critical field value for the breakdown was not taken into account. We have improved their calculations by introducing the minimal value of the electric field for the breakdown between two successive points. In this case the breakdown pattern is more directionally selected, and a lower fractal dimensionality results. With the increase of the external magnetic field, the MFD growth and finally saturates at an universal for high magnetic fields values of $`1.67\pm 0.01.`$ The growth of MFD with the magnetic field could be expected, since the curved trajectories fill up the space more densely than the straight ones. The existence of an universal limit, hovewer, is far from being obvious. Following the directed percolation models, one could think that the saturation of MFD will occur at $`D=2`$ . Fig 4. shows clearly that in this system the MFD saturates due to the physics of current carrying streamers.
An unexplained feature in the experiments of is the ’spider-legs’ form of the breakdown pattern in the absence of the external magnetic field. We outline here that the streamer currents are rather strong, $`10÷100`$ $`A`$, and their influence on the streamer pattern can be very important. To describe this phenomenon we have taken into account the magnetic interaction between current carrying streamers, in the framework of the modified active walker model (MAWM) which is described in what follows. The results of our calculations, presented in Fig. 5, show that our model correctly describes the experimentally observed ’spider-legs’ effect.
Active walker models have been used to describe different pattern formation problems . In these models the walkers moovement is subject to the influences of the environment and vice versa. We describe the leader channel propagation in the magnetic field using the active walker model. The Lorentz force acting on the fast-moving electrons is particularly effective in the high-field regions in the leader tips, where the channel formation takes place.
Let us consider again a two-dimensional square lattice in which a central point represents one of the electrodes while the other electrode is modeled as a circle at large distance. The discharge starts at the central electrode, so initially several walkers are set in the vicinity of it. The walkers move in a potential which is the solution of the Laplace equation with the boundary conditions determined by electrodes and discharge structure. During a step of growth each walker moves. The solution of Laplace equation is found by iteration method after each step of growth.
When a walker moves to a point, this point starts belonging to the breakdown structure. The potential of the central electrode is $`0`$, the potential of a point in the structure is $`V_{i,k}=V_{l,m}+V_Rl`$, where $`V_{l,m}`$ is the potential of a points from which growth go on, $`V_R`$ is an internal field in the structure and $`l`$ is $`1`$ for bonds parallel to the grid and is $`\sqrt{2}`$ for diagonal bonds. In the absence of the magnetic field, the probability of a walker step is a function of the local field $`F_{loc}`$ . The breakdown occurs only if the local field is greater then the critical field of growth $`F_C`$.
The magnetic field is taken into account by the following way. We add to the local field in the direction perpendicular to the previous move of the walker the Lorentz force which is proportional to the magnetic field. The magnetic field acting on the $`i`$ walker is $`H_i=H_0+H_I^i`$, where $`H_0`$ is external magnetic field and $`H_I^i`$ is the field created by currents of the breakdown structure. We calculate $`H_I^i`$ by means of the Biot-Savart law:
$$H_I^i=\underset{ki,l}{}\frac{I}{r^3}d\stackrel{}{s}_{k,l}\times \stackrel{}{r}$$
(5)
where $`I`$ is the current in a channel, $`d\stackrel{}{s}_{k,l}`$ is $`l`$ element of the $`k`$ channel and $`\stackrel{}{r}`$ is the vector pointed from the position of $`d\stackrel{}{s}_{k,l}`$ to the position of the $`i`$ walker.
The main results of this part of studies are presented in Fig. 5.
To summarize, we have generalized and modified the existing dielectric breakdown models to explain the experimental observations of the propagation of a streamer near an insulating surface under the influence of a transverse magnetic field. We have introduced the concept of the Magnetic Fractal Dimensionality (MFD) and have obtained its saturation with growing magnetic fields. The Universal Magnetic Fractal Dimensionality (UMFD) equals $`1.670`$ which is superior to $`1.65,`$ the one in the absence of a magnetic field. Inclusion of the magnetic interaction between the current-carrying streamers results in the ’spider-legs’ like streamer patterns at lower fields, which corresponds to the experimental observations.
We acknowledge helpfull discussions with T.Maniv, and A.Zhuravlev. We acknowledge the assistance of A.Kaplunovsky in the initial stage of numerical simulations.
Figure Captions
Fig.1.
Bonds connecting a lattice point with one of eight adjacent lattice points in a two-dimensional square lattice. The geometry of the experiment is represented by a central electrode at some point while the other electrode is a circle at large distance. The discharge starts at the central electrode and grows by one lattice bond per growth step.
Fig.2.
Computer-generated discharge pattern (Lichtenberg figure) corresponding to the following set of parameters: $`H=0,\eta =1,F_c=0`$ and $`V_R=0`$ as explained in the text.The white lines correspond to the leader channels.
Fig.2.
Computer-generated discharge pattern in the presence of the external magnetic field. The leader channels in the magnetic field are distorted and appeare to have a circular shape outside the central electrode region, due to the action of the magnetic field.
Fig.4.
The saturation of the fractal dimensionality, with growing magnetic fields, at the value of $`D=1.67`$ .
Fig.5.
The ’spider-legs’ form of the breakdown pattern folowing from the magnetic interaction between the streamer currents: a) current-current interactions are not took into account, $`H=0`$; b)current-current interactions are took into account, $`H=0`$; c) $`H0`$.
|
warning/0005/nlin0005026.html
|
ar5iv
|
text
|
# Symmetry-breaking on-off intermittency under modulation: Robustness of supersensitivity, resonance and information gain
## I Introduction
Nonlinear dynamical systems possessing an invariant subspace are of great interest, particularly when the system motion within the subspace can be chaotic or stochastic. Examples include chaotic systems with symmetry or coupled chaotic systems . Bubbling and on-off intermittency are typical behaviors in system close to a threshold of transverse stability of the subspace due to the fluctuative nature of the local transverse Lyapunov exponent in different part of the subspace. In bubbling, the subspace is stable on average, but local instability can result in large bursts away from the subspace when it is perturbed. The sensitivity of this weak stability to parameter mismatch and noise has been studied by Pikovsky and Grassberger . Intermittent loss of synchronization in experiments with inevitable noise and parameter mismatch is undesirable in application of high-quality synchronization, such as to communication . In on-off intermittency, the subspace is slightly unstable on average, but local transverse attraction may keep the dynamics very close to the subspace for a long period of time. Great attention has been paid to the duration of this laminar period which exhibits universal power law distribution in a broad class of systems . Noise in the system prevents its state from approaching the subspace close beyond the noise level, thus has important effects on the laminar period distribution or the escape time . In noisy environment, bubbling and on-off intermittency are essentially the same phenomenon.
Sensitivity in nonlinear systems can be very useful for applications such as controlling global dynamics of the system by local tiny perturbations . It is interesting to ask whether the sensitivity of on-off intermittency may lead to any potential application of the phenomenon. An observation is that in a class of symmetric systems, on-off intermittency can be symmetry-breaking , namely, the bursting behavior does not possess the system symmetry when the system has two symmetric but distinct attractors. This Letter reports that a combination of the sensitivity of on-off intermittency and the symmetry-breaking of the bursting can result in remarkable features in the systems subjected to a weak modulation signal in the noisy environment.
## II The model
Let $`x(t)`$ represent the distance of the dynamics from the invariant subspace, and $`x(t)>0`$ and $`x(t)<0`$ denote the dynamics in the two symmetric components respectively. In the symmetry-breaking systems, transitions between $`x>0`$ and $`x<0`$ can only occur when $`x(t)`$ comes to the level of the weak noisy signal. For system displaying appreciable laminar state, main features of the dynamics can be described by the general linear equation close to the subspace:
$$\dot{x}(t)=[\lambda +\sigma _1\xi (t)]x(t)+\sigma _2e(t)+ps(t).$$
(1)
Here $`\lambda `$ is the transverse Lyapunov exponent of the subspace, and $`\sigma _1\xi (t)`$ with $`\xi (t)=0`$ is the fluctuation of the local Lyapunov exponent due to the chaotic or stochastic motion within the subspace. In general, chaotic system has quickly decaying correlation, and in a large enough time scale $`t`$, $`\xi (t)`$ has an asymptotic Gaussian distribution. $`e(t)`$ is the additive white noise with level $`\sigma _2\sigma _1`$ and $`s(t)`$ is a weak modulation signal. The exact form of the signal is unimportant for the phenomena reported below, provided it varies on a time scale slower than the characteristic times of the systems. Here we consider $`s(t)`$ a random binary stream ($`\pm 1`$ with probability 0.5) with a bit duration $`T`$. $`p\sigma _1`$ is the amplitude of the signal, and $`R=p/\sigma _2`$ provides a natural measure of the signal-to-noise ratio (SNR).
In Ref. , Cenys and Lustfeld studied the statistical properties of escape time of on-off intermittency subjected to noise by means of Fokker-Planck equation. It has been shown that on-off intermittency is very sensitive to noise . We employ the same approach of Fokker-Planck equation, and focus on the property of amplification of the weak external signal $`s(t)`$ in the system. Our results will demonstrate that the system is also very sensitive to the weak signal, and the amplification of the weak signal is robust to the additive noise. This sensitivity exhibits resonant behavior as the system parameters change.
The Fokker-Planck equation for Eq. (1) is
$`{\displaystyle \frac{W}{t}}`$ $`=`$ $`{\displaystyle \frac{}{x}}\left\{\left[(\lambda +{\displaystyle \frac{\sigma _1^2}{2}})x+ps(t)\right]W\right\}`$ (2)
$`+`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \frac{^2}{x^2}}[(\sigma _1^2x^2+\sigma _2^2)W].`$ (3)
In general, it is quite difficult to solve Eq. (3) exactly. Let $`T_0`$ be the relaxation time of the system after $`s(t)`$ switching from $`+1(1)`$ to $`1(+1)`$. If $`TT_0`$, the probability distribution $`W(t)`$ can establish an approximate static state during each bit of the input signal. Under the adiabatic approximation $`TT_0`$, $`W/t0`$, and the static solution can be obtained analytically as
$$W(x)=C\left(x^2+\frac{\sigma _2^2}{\sigma _1^2}\right)^{(\alpha 1)/2}\mathrm{exp}\left[\frac{2ps(t)}{\sigma _1\sigma _2}\mathrm{arctan}\frac{\sigma _1x}{\sigma _2}\right],$$
(4)
where $`\alpha =2\lambda /\sigma _1^2`$. For $`|\sigma _1x/A|1`$ ($`A=\mathrm{max}(p,\sigma _2)`$),
$$W(x)C|x|^{\alpha 1}\mathrm{exp}\left[\frac{\pi ps(t)}{\sigma _1\sigma _2}\text{sgn}x\right].$$
(5)
Now we see that the behavior of the system can be divided into two regimes. One is $`|x||\sigma _2e(t)+ps(t)|`$, where the dynamics is governed approximately by $`\dot{x}(t)=[\lambda +\sigma _1\xi (t)]x(t)`$. Let $`z=\mathrm{ln}|x|`$, then $`\dot{z}(t)=\lambda +\sigma _1\xi (t)`$ which describes a Brownian motion with a constant drift $`\lambda `$ and diffusion constant $`\sigma _1^2/2`$. The nonlinearity of the system can be modeled by an effective reflecting boundaries of the Brownian motion at $`\pm x_b`$, which is of the order of $`\sigma _1`$. The probability density has a power form $`W(x)=|x|^{\alpha 1}`$, but is asymmetric for $`x>0`$ and $`x<0`$ in the presence of $`s(t)`$. The system can rarely perform transition between $`x>0`$ and $`x<0`$ in this regime due to the symmetry-breaking property, until it comes to the other regime, where the noisy input $`\sigma _2e(t)+ps(t)`$ dominates the dynamics and the system performs transition between $`x>0`$ and $`x<0`$ frequently. The behavior of the system is determined by the competition between the diffusion and the drift of the Brownian motion. If the drift time $`t_b=\mathrm{ln}(\sigma _1/A)/|\lambda |`$ is much smaller than the diffusion time $`t_d=2\mathrm{ln}^2(\sigma _1/A)/\sigma _1^2`$, the drift is dominant over the diffusion, and the system will either come to a metastable state induced by the noisy input for $`\lambda <0`$, or approach some state away from the invariant subspace for $`\lambda >0`$. In both cases, the weak noisy input has no significant effects on the system behavior, i.e. the system is insensitive to the modulation. On the other hand, the diffusion is dominant for $`t_bt_d`$, and the system can have access to both the level of the weak input and the boundary of the nonlinearity, exhibiting typical on-off intermittency and sensitivity to the weak modulation. Variation of the parameter $`\lambda `$ or $`\sigma _1`$ affects the competition between the drift and the diffusion, and the system is expected to display optimal response to the weak modulation with resonant characterization.
## III Robustness of supersensitivity to the weak signal
Let us consider the diffusion dominant region close to the critical point of the stability of the subspace, e.g. $`|\lambda |1`$, $`|\alpha \mathrm{ln}(x_b/A)|1`$, $`\sigma _11`$, $`p10^m(m1)`$. Employing the approximation in Eq. (5), and the effective reflecting boundaries at $`\pm x_b`$, the ensemble average $`x(t)`$ is estimated as
$$x(t)s(t)\frac{x_b}{\mathrm{ln}(x_b/A)}\mathrm{tanh}\frac{\pi R}{\sigma _1},$$
(6)
which in the noise-free limit $`\sigma _20`$, assumes the form
$$x(t)s(t)\frac{x_b}{\mathrm{ln}(x_b)\mathrm{ln}p}.$$
(7)
A logarithmical dependence of $`x(t)`$ on the input level $`p`$ means an amplification of the weak signal $`ps(t)`$ with an factor $`x/p10^m/m(m1)`$, i.e. the system exhibits supersensitivity to extremely weak modulation close to the critical point. This sensitivity was also reported in an overdamped Kramers oscillator with multiplicative noise free from additive noise ($`\sigma _2=0)`$, which is a specific example in this class of systems . In the absence of $`s(t)`$, the system produces symmetric bursting pattern with $`x(t)=0`$; while the bursting pattern is reorganized to manifest the weak signal after it is fed into the system (see Fig. 1). The most interesting and practically important property is that the weak signal is manifested even buried in a relatively high level of noise, namely, the robustness of the supersensitivity. This behavior originates from the symmetry-breaking of the on-off intermittency in the system.
To demonstrate the above analysis, we employ the following system in simulations
$`\ddot{y}`$ $`=`$ $`\gamma \dot{y}+4y(1y^2)+f_0\mathrm{sin}\omega t,`$ (8)
$`\dot{x}`$ $`=`$ $`(a+by)\mathrm{sin}(x)x+\sigma _2e(t)+ps(t),`$ (9)
where $`y`$ constitutes the forced Duffing chaotic oscillator. With $`\gamma =0.05,f_0=2.3`$ and $`\omega =3.5`$, the Duffing system is chaotic and $`\sigma _10.964b`$. The nonlinearity of the variable $`x`$ is related to an experimental model of superconducting quantum interference device (SQUID) . However, we should stress that the specific form of the nonlinearity is of no importance for the phenomena. The transverse Lyapunov exponent of the invariant subspace $`x=0`$ is $`\lambda =a1`$ due to the symmetry of the Duffing chaotic attractor. Fig. 1 shows typical behavior of the system and good agreement between the analytical and the simulation results for $`x(t)`$ as a function of $`R`$. The agreement demonstrates that the general stochastic model in Eq. (1) gives good account for this type of system even though the motion in the subspace is deterministic chaos.
## IV Resonant behavior
It is difficult to perform generally a quantitative analysis of the system response to the weak signal based on the linear dynamics in Eq. (1) as parameters $`\lambda `$ or $`\sigma _1`$ changes, because the effective boundary $`x_b`$ changes with the nonlinearity and the linear dynamical model with an effective reflecting boundary is often not sufficient to capture the dynamical property if $`\alpha `$ is appreciately positive. Moreover, as the parameters change, the relaxation time $`T_0`$ may become comparable to the bit duration $`T`$, and the transient behavior plays an important role in the system response and an adiabatic approximation is not valid any more. To demonstrate the resonant properties, we rely on simulations with the system in Eq. (9), while the Brownian motion model can provide a qualitative understanding of the properties, thus showing that the properties are generic and universal for a general class of the systems.
For a system with on-off intermittent output $`x(t)`$, the ensemble average $`x(t)`$ and the correlation between $`s(t)`$ and $`x(t)`$ is relatively small even for the noise-free case $`\sigma _2=0`$, due to the power law fluctuation of $`x(t)`$. To better characterize the response of the system to the modulation, we transfer the output series $`x(t)`$ into a binary stream $`X(t)`$ by a threshold crossing process: suppose $`x(t)`$ becomes larger than a prescribed threshold $`x_{th}`$ at some moment, after that $`X(t)`$ will keep at $`X(t)=1`$ until $`x(t)`$ crosses $`x_{th}`$ at another moment; $`X(t)`$ will not switch back from $`X(t)=1`$ to $`+1`$ until $`x(t)`$ crosses $`x_{th}`$ again, and so on. This binary presentation captures the most important feature of the transition of the bursting pattern between the two symmetric attractors. $`X(t)`$ has a strong correlation with $`s(t)`$ for weak noise case $`\sigma _2<p`$ if the system is close to the critical point. The exact value of $`x_{th}`$ is not crucial for the properties described below. In the following, we fix $`p=10^7`$, $`T=2000`$ and $`x_{th}=1`$, and take the cross-correlation function $`C`$ between $`s(t)`$ and $`X(t)`$ estimated using $`10^4`$ bits of a random stream of $`s(t)`$ to demonstrate the resonant behavior in the system.
(1) With the change of $`\lambda `$. For $`\lambda `$ rather below the critical point $`\lambda =0`$, the system has a metastable state close to the level of the noisy input, and the diffusion is not strong enough to produce large bursts frequently enough, $`C`$ will be small. On the opposite, if $`\lambda `$ is rather above the critical point, the drift is also dominant so that the system can seldom access to the level of the weak modulation, and becomes insensitive to the switching of $`s(t)`$ between $`\pm 1`$, resulting in a small $`C`$ again. Close to the critical point, the system can access to the level of the weak signal and produce large bursts frequently due to strong enough diffusion. The switching of the weak signal is “sensed” and manifested by asymmetrical bursting to $`x>0`$ and $`x<0`$, giving an optimal value of $`C`$. This precess is illustrated by $`C`$ as a function of $`\lambda `$ in Fig. 2 for various $`R`$ values.
(2)With the change of $`\sigma _1`$. For a small $`\sigma _1`$ where the drift is dominant over diffusion, the system is not sensitive to the weak noisy input and $`C`$ assumes a small value. With the increase of $`\sigma _1`$, the diffusion becomes stronger, resulting in a smaller relaxation time $`T_0`$ and more frequent large bursts, and the system becomes more sensitive to the weak input. In the noise-free case $`\sigma _2=0`$, the increased sensitivity enables $`X(t)`$ to keep closer in phase to the weak signal $`s(t)`$ and $`C`$ approaches closer to $`1.0`$ if the system maintains to work in the symmetry-breaking regime, and in general a resonant behavior is not expected. The picture becomes quite different if $`\sigma _20`$. With smaller $`T_0`$ and increased sensitivity, the system can keep up with and manifest more and more noise-induced transitions in shorter time-scales, and the transition rate of $`X(t)`$ between $`\pm 1`$ may become much higher than that of $`s(t)`$, leading to a decreasing $`C`$. An optimal response is achieved when the diffusion is strong enough to become sensitive to the weak input but not too strong to manifest a lot of noise-induced transitions in short time-scales. Typical example of the system response as a function of $`\sigma _1`$ is shown in Fig. 3.
The above resonant behavior is similar to the conventional stochastic resonance where a dynamical system displays increased sensitivity to a subthreshold signal with an optimal level of additive noise, see Ref. for an extensive review. Resonance occurs when a noise-controlled time-scale in the system matches that of the signal. In our system, the underlying mechanism of the resonant behavior is quite different. The sensitivity to an extremely weak signal is induced by the multiplicative chaotic or stochastic motion in the subspace. To achieve this sensitivity, it is required that the system is in the on-off intermittency regime so that it can become susceptible to the weak signal by coming close enough to the subspace, and manifest and amplify it by quick enough large bursts away from the subspace with symmetry-breaking. As system parameters $`\lambda `$ and $`\sigma _1`$ change, a competition between these two factors leads to the resonant behavior. More interestingly, resonant behavior with respect to the change of $`\sigma _1`$, the level of the multiplicative chaos (noise), necessarily occurs only in the presence of the additive noise due to the nature of this competition. These features are generic and universal in a general class of systems displaying on-off intermittency with symmetry breaking. Multiplicative stochastic resonance has been studied by Gammaitoni et al in a multiplicatively driven bistable system with $`\lambda =1`$. In that case, the system is out of the regime of on-off intermittency, and consequently cannot display the property of (super)sensitivity, and the resonance with respect to the change of $`\lambda `$ was not resported.
## V Information gain
Now consider the system from the viewpoint of transmission and amplification of a weak signal $`ps(t)`$ contaminated with channel noise $`\sigma _2e(t)`$ through a system displaying on-off intermittency with symmetry-breaking. It is very interesting and practically important that more information about the signal may be obtained from the output $`X(t)`$ than from the noisy input $`ps(t)+\sigma _2e(t)`$ itself, besides the fact that the weak signal has been amplifyed to a level discernible with a low resolution detector. To examine the information gain, we compare $`C`$ with the correlation between the signal $`ps(t)`$ and the total noisy input $`ps(t)+\sigma _2e(t)`$, i.e. $`C_{in}=R/\sqrt{1+R^2}`$. $`C`$, $`C_{in}`$ and their difference are shown in Fig. 4. $`C`$ comes to a saturated value for $`R1`$, where the noise-induced transitions between $`x>0`$ and $`x<0`$ in short time-scales are rarely manifested by large bursts. This value does not approach $`1.0`$ as $`C_{in}`$ due to an average time-delay between $`X(t)`$ and $`s(t)`$ induced by the relaxation time which is longer for weaker signal. Clearly, appreciable information gain is obtained by the system in a broad range of $`R`$, and an optimal gain is found at a certain $`R`$ value. If taking into account the effect of the time-delay between $`X(t)`$ and $`s(t)`$, e.g. by defining $`C_\tau `$ as the maximum of the correlation between $`X(t)`$ and $`s(t\tau )`$, the information gain region can be wider (Fig. 4).
To conclude, we demonstrate interesting universal features of robustness of supersensitivity, resonance and information gain in a class of nonlinear system subjected to a weak modulation. These systems present a new mechanism of resonant behavior compared to conventional stochastic resonance. While intermittent loss of synchronization may be harmful for any applications employing high-quality synchronization , the features found in this letter are meaningful for potential applications of on-off intermittency. On-off intermittency has been demonstrated in many experimental systems and we believe that the behaviors reported in this work can be tested in physical experiments.
This work is supported in part by grants from the Hong Kong Research Grants Council (RGC) and the Hong Kong Baptist University Faculty Research Grant (FRG).
|
warning/0005/math0005095.html
|
ar5iv
|
text
|
# A generalization of Kummer’s identity
## 1 The generalization
The subject of this paper is a generalization of Kummer’s identity (see \[Kum36\], \[Bai35, 2.3\], or Cor. 3.1.2 in \[AAR99\]):
$${}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{a,b}{1+ab}|1\right)=\frac{\mathrm{\Gamma }(1+ab)\mathrm{\Gamma }(1+\frac{a}{2})}{\mathrm{\Gamma }(1+a)\mathrm{\Gamma }(1+\frac{a}{2}b)}.$$
(1)
The hypergeometric series on the left is defined if $`ab`$ is not a negative integer, and it is absolutely convergent for $`\text{Re}(b)<1/2`$. After analytic continuation of $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{a,b}{1+ab}|z\right)`$ on $`\text{C}[1,\mathrm{})`$ and after division of both sides by $`\mathrm{\Gamma }(1+ab)`$ the formula has meaning and is correct for all complex $`a,b`$. In this paper, whenever $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{A,B}{C}|z\right)`$ denotes a well-defined hypergeometric series, it also denotes its analytic continuation on $`\text{C}[1,\mathrm{})`$.
The generalization to be considered evaluates the hypergeometric series $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{A,B}{C}|1\right)`$ whenever $`CA+B`$ is any integer. In the terminology of \[AAR99\], our generalization applies to $`{}_{2}{}^{}\text{F}_{1}^{}(1)`$ series which are contiguous to a series for Kummer’s formula (1). As it is known (see \[AAR99, 2.5\]), the 15 classical Gauss contiguity relations can be iterated to produce a linear relation between any three contiguous $`{}_{2}{}^{}\text{F}_{1}^{}(z)`$ series, with coefficients being rational functions in the parameters of those series. This also applies to their analytic extensions. The generalized formula is such a relation in explicit form between contiguous $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{a+n,b}{ab}|1\right)`$, $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{a,b}{1+ab}|1\right)`$ and $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{a1,b}{ab}|1\right)`$, where $`n`$ is an integer, and the last two series are evaluated using Kummer’s identity (1). The coefficient to the first series cannot be the zero function because the quotient of the other two series is not in $`\text{C}(a,b,n)`$. In the generalized formula these coefficients are written as terminating $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series.
We write the generalization in the form
$${}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{a+n,b}{ab}|1\right)=P(n)\frac{\mathrm{\Gamma }(ab)\mathrm{\Gamma }(\frac{a+1}{2})}{\mathrm{\Gamma }(a)\mathrm{\Gamma }(\frac{a+1}{2}b)}+Q(n)\frac{\mathrm{\Gamma }(ab)\mathrm{\Gamma }(\frac{a}{2})}{\mathrm{\Gamma }(a)\mathrm{\Gamma }(\frac{a}{2}b)}.$$
(2)
Here the two $`\mathrm{\Gamma }`$-terms are equal to $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{a1,b}{ab}|1\right)`$ and $`\frac{ab}{a2b}{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{a,b}{1+ab}|1\right)`$ respectively, and $`P(n)`$, $`Q(n)`$ are rational functions in $`a,b`$ for every integer $`n`$. The most convenient expressions for $`P(n)`$ and $`Q(n)`$ are summarized in the three theorems below. In fact, expressions of $`{}_{2}{}^{}\text{F}_{1}^{}(1)`$ series in (2) in terms of terminating series and $`\mathrm{\Gamma }`$-function were known to Whipple, see \[Whi30\]. His formulas (8.3) and (8.41) would express the $`{}_{2}{}^{}\text{F}_{1}^{}(1)`$ series in (2) in terminating series for negative or positive $`n`$, respectively. Whipple’s formulas (11.5,51) form the statement of Theorem 1 below. Whipple derived them as a consequence of transformations of $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series allied to general $`{}_{2}{}^{}\text{F}_{1}^{}(1)`$ series, and from \[Fox27, (2.6,7)\] where some $`{}_{2}{}^{}\text{F}_{1}^{}(1/2)`$ series are expressed in terms of terminating series. However, Whipple’s main concern was the relations of general $`{}_{2}{}^{}\text{F}_{1}^{}(1)`$ and $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series. As we will see, his approach is not convenient when some of those series terminate.
In this paper we strive for a clear overview of possible expressions for $`P(n)`$ and $`Q(n)`$ in terms of terminating $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series, with simpler proofs. Another aim is to consider algorithmic aspects of evaluating hypergeometric series. In particular, we specialize formula (2) to two-term identities, which however seem to be beyond Zeilberger’s approach. Also a few evaluations similar to (2) are presented. Specifically, we evaluate hypergeometric series which are contiguous to the $`{}_{2}{}^{}\text{F}_{1}^{}(1/4)`$ and $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series in Gosper’s and Dixon’s identities, see (4-36).
In the following theorems we summarize the most convenient expressions for $`P(n)`$ and $`Q(n)`$. A few more such expressions are presented in (16-19).
###### Theorem 1
Suppose that $`n`$ is a non-negative integer (or $`1`$), and $`a,b`$ are complex numbers such that $`(a)_n0`$ and $`ab`$ is not zero or a negative integer. Then the coefficients $`P(n)`$ and $`Q(n)`$ in formula (2) can be written as:
$`P(n)`$ $`=`$ $`{\displaystyle \frac{1}{2^{n+1}}}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n}{2},\frac{n+1}{2},\frac{a}{2}b}{\frac{1}{2},\frac{a}{2}}}\right),`$ (3)
$`Q(n)`$ $`=`$ $`{\displaystyle \frac{n+1}{2^{n+1}}}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n1}{2},\frac{n}{2},\frac{a+1}{2}b}{\frac{3}{2},\frac{a+1}{2}}}\right).`$ (4)
###### Theorem 2
Suppose that $`n`$ is a non-negative integer, and $`a,b`$ are complex such that $`(a)_n0`$, and $`ab`$ is not zero or a negative integer. Then the coefficients $`P(n)`$ and $`Q(n)`$ in formula (2) can be written as:
$`P(n)={\displaystyle \frac{1}{2}}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n}{2},\frac{n+1}{2},b}{n,\frac{a}{2}}}\right),Q(n)={\displaystyle \frac{1}{2}}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n1}{2},\frac{n}{2},b}{n,\frac{a+1}{2}}}\right).`$ (5)
The hypergeometric sums should be interpreted as terminating series with (up to $`\pm 1`$) $`n/2`$ terms.
###### Theorem 3
Let $`P(n,a,b)`$ and $`Q(n,a,b)`$ denote the coefficients $`P(n)`$ and $`Q(n)`$ in (2) as functions of $`a,b`$ as well. If $`n`$ is a non-negative integer, and $`a,b\{0,1,\mathrm{},n\}`$ then
$`P(n1,a,b)`$ $`=`$ $`2^{2n}{\displaystyle \frac{(1\frac{a}{2})_n}{(1b)_n}}P(n1,a2n,bn),`$ (6)
$`Q(n1,a,b)`$ $`=`$ $`2^{2n}{\displaystyle \frac{(\frac{1a}{2})_n}{(1b)_n}}Q(n1,a2n,bn).`$ (7)
Because of the last theorem we do not give expressions for $`P(n)`$ and $`Q(n)`$ for a negative $`n`$, except (13-14) in the proof of Theorem 3.
These theorems are proved in section 2. There we also overview transformations between other expressions for $`P(n)`$ and $`Q(n)`$, and give a survey of Whipple’s approach in \[Whi30\]. In section 3 Theorem 2 is proved using the more universal Zeilberger’s method. The key observation is that the sequences $`P(n)`$ and $`Q(n)`$ satisfy the same recurrence relation as the left-hand side of (2). Theorem 1 can also be proved in this way. Notice that any different expressions for $`P(n)`$ and $`Q(n)`$ must represent the same rational functions in $`a`$, $`b`$ for every $`n`$, because the quotient of the $`\mathrm{\Gamma }`$-terms in (2) is not in $`\text{C}(a,b)`$. Section 4 is devoted to algorithmic aspects of evaluation of hypergeometric series, with similar generalizations of Dixon’s and Gosper’s identities.
Acknowledgments. The author would like to thank Richard Askey and Tom Koornwinder for useful suggestions, in particular for ideas of the easy proof(s) in section 2, and Dennis Stanton for the references to Whipple.
## 2 Classical proof
We assume here $`\text{Re}(a/2)>\text{Re}(b)>0`$. One can simply check that Theorems 1 and 2 must hold for the analytic continuation of the $`{}_{2}{}^{}\text{F}_{1}^{}(1)`$ series as well.
To prove Theorem 1 we recall Whipple’s identity \[Whi30, (8.41)\]
$${}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{A,B}{C}|1\right)=\frac{\mathrm{\Gamma }(C)}{2\mathrm{\Gamma }(A)}\underset{k=0}{\overset{\mathrm{}}{}}(1)^k\frac{(CA+B1)_k}{k!}\frac{\mathrm{\Gamma }(\frac{A}{2}+\frac{k}{2})}{\mathrm{\Gamma }(C\frac{A}{2}+\frac{k}{2})}.$$
(8)
As it was communicated by Askey, this identity can be easily proved using Euler’s integral representation (\[Erd53, 2.12(1)\]) for the $`{}_{2}{}^{}\text{F}_{1}^{}(z)`$ series. One has to rearrange the integrand as
$$t^{A1}(1t)^{CA1}(1+t)^B=t^{A1}(1+t)^{C+AB+1}(1t^2)^{CA1},$$
(9)
expand $`(1+t)^{C+AB+1}`$ as series, interchange integration and summation, change the variable $`t\sqrt{s}`$, and recognize the beta-integral \[Erd53, 1.5(1)\].
We apply<sup>2</sup><sup>2</sup>2The same could be done directly to $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{a+n,b}{ab}|1\right)`$, of course. We would get less-convenient formula $`{}_{2}{}^{}\text{F}_{1}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{a+n,b}{ab}}|1\right)`$ $`=`$ $`{\displaystyle \frac{1}{2}}{\displaystyle \frac{\mathrm{\Gamma }(ab)\mathrm{\Gamma }(\frac{a+n}{2})}{\mathrm{\Gamma }(a+n)\mathrm{\Gamma }(\frac{an}{2}b)}}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n}{2},\frac{n+1}{2},\frac{a+n}{2}}{\frac{1}{2},\frac{an}{2}b}}\right)`$ $`+{\displaystyle \frac{n+1}{2}}{\displaystyle \frac{\mathrm{\Gamma }(ab)\mathrm{\Gamma }(\frac{a+n+1}{2})}{\mathrm{\Gamma }(a+n)\mathrm{\Gamma }(\frac{an+1}{2}b)}}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n1}{2},\frac{n}{2},\frac{a+n+1}{2}}{\frac{3}{2},\frac{an+1}{2}b}}\right).`$ Here for each positive integer $`n`$ the two $`\mathrm{\Gamma }`$-terms are $`\text{C}(a,b)`$-multiples of the $`\mathrm{\Gamma }`$-terms in (2), so the coefficients $`P(n)`$, $`Q(n)`$ are equal to $`\text{C}(a,b)`$-multiples of the $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series in this formula. But the correspondence depends on whether $`n`$ is even or odd. formula (8) to the right-hand side of the identity \[Erd53, 2.9(2)\]:
$${}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{a+n,b}{ab}|1\right)=2^{2bn}{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{a2b,bn}{ab}|1\right).$$
(10)
After this we sum up the terms with even and odd indexes separately, transform the $`\mathrm{\Gamma }`$-factors slightly and get formula (2) with $`P(n)`$, $`Q(n)`$ defined by (3-4).
Theorem 2 follows from Theorem 1 by the following transformation of terminating $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series (see \[AAR99\], proof of Cor. 3.3.4):
$${}_{3}{}^{}\text{F}_{2}^{}\left(\genfrac{}{}{0pt}{}{m,A,B}{E,F}\right)=\frac{(EA)_m}{(E)_m}{}_{3}{}^{}\text{F}_{2}^{}\left(\genfrac{}{}{0pt}{}{m,A,FB}{1+AEm,F}\right),$$
(11)
where $`m`$ must be a non-negative integer. To make sure that the interpretation of ill-defined hypergeometric series in (5) is correct for this transformation, one may specialize $`A`$ to $`\nu /2`$ or $`(\nu \pm 1)/2`$ with complex $`\nu `$ (instead of $`n/2`$, etc.) and take the limit $`\nu n`$.
To prove Theorem 3 we use Euler’s integral again. After rearranging the integrand in (9) as $`t^{A1}(1t)^{CA+B1}(1t^2)^B`$ and expanding $`(1t)^{CA+B1}`$ we eventually get formula:
$${}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{A,B}{C}|1\right)=\frac{1}{2}\frac{\mathrm{\Gamma }(C)\mathrm{\Gamma }(1B)}{\mathrm{\Gamma }(A)\mathrm{\Gamma }(CA)}\underset{k=0}{\overset{\mathrm{}}{}}\frac{(ABC+1)_k}{k!}\frac{\mathrm{\Gamma }(\frac{A}{2}+\frac{k}{2})}{\mathrm{\Gamma }(\frac{A}{2}+\frac{k}{2}+1B)}.$$
(12)
Like in the proof of Theorem 1, we apply this formula to $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{an1,b}{ab}|1\right)`$ transformed by (10), and add the terms with even and odd indexes separately. The result is:
$`P(n1)`$ $`=`$ $`2^n{\displaystyle \frac{(1\frac{a}{2})_n}{(1b)_n}}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n}{2},\frac{n1}{2},\frac{a}{2}b}{\frac{1}{2},\frac{a}{2}n}}\right),`$ (13)
$`Q(n1)`$ $`=`$ $`n\mathrm{\hspace{0.17em}2}^n{\displaystyle \frac{(\frac{1a}{2})_n}{(1b)_n}}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n1}{2},\frac{n2}{2},\frac{a+1}{2}b}{\frac{3}{2},\frac{a+1}{2}n}}\right).`$ (14)
Comparing these expressions with (3-4) gives Theorem 3. Q.E.D.
To get more expressions for $`P(n)`$ and $`Q(n)`$ one can use standard transformations of terminating $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series. For example, one may repeatedly apply (11) or rewrite a terminating series in the reverse order. In general, a terminating $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series can be transformed to 17 other terminating $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series, see \[Whi24, sect. 8\], \[Bai35, 3.9\]. To give these transformations a group structure one has to consider transpositions of the two lower and two upper parameters as non-trivial transformations. Then one gets a group of 72 elements which acts on the set of 18 hypergeometric series, see \[RvdJR<sup>+</sup>92\]. The action of this group can be summarized as follows. Let $`y_0,\mathrm{},y_5`$ be six parameters satisfying $`y_0+y_1+y_2=y_3+y_4+y_5=1m`$. Then the expression
$$(y_0+y_4)_m(y_0+y_5)_m{}_{3}{}^{}\text{F}_{2}^{}\left(\genfrac{}{}{0pt}{}{m,y_0+y_1y_3,y_0+y_2y_3}{y_0+y_4,y_0+y_5}\right)$$
(15)
is invariant under the permutations within the sets $`\{y_0,y_1,y_2\}`$ and $`\{y_3,y_4,y_5\}`$, and gets multiplied by $`(1)^m`$ when these two sets are interchanged. For instance, formula (11) corresponds to the permutation $`y_0y_5`$, $`y_1y_4`$, $`y_2y_3`$.
Application of these transformations to the series (3-4) or (5) gives eight sets of 18 terminating $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series, one set for a choice of $`P(n)`$ or $`Q(n)`$, positive or negative and even or odd $`n`$. The number of different hypergeometric series turns out to be 96. Here we summarize a few interesting expressions for $`n0`$:
$`P(n)`$ $`=`$ $`{\displaystyle \frac{\frac{n}{2}!}{2n!}}\left(\frac{1a}{2}+b\right)_{n/2}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n}{2},\frac{a+1}{2}+\frac{n}{2},\frac{a}{2}b}{\frac{a}{2},\frac{a+1}{2}\frac{n}{2}b}}\right)`$ (16)
$`=`$ $`{\displaystyle \frac{1}{2^{n+1}}}{\displaystyle \frac{(b)_{n/2}}{(\frac{a}{2})_{n/2}}}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n}{2},1+\frac{n}{2},\frac{a}{2}b}{\frac{1}{2},\mathrm{\hspace{0.33em}1}b\frac{n}{2}}}\right),`$ (17)
$`Q(n)`$ $`=`$ $`{\displaystyle \frac{\frac{n}{2}!}{2n!}}\left(1\frac{a}{2}+b\right)_{n/2}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n}{2},\frac{a}{2}+\frac{n}{2},\frac{a+1}{2}b}{\frac{a+1}{2},\frac{a}{2}\frac{n}{2}b}}\right)`$ (18)
$`=`$ $`{\displaystyle \frac{n+1}{2^{n+1}}}{\displaystyle \frac{(b)_{n/2}}{(\frac{a+1}{2})_{n/2}}}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n}{2},1+\frac{n}{2},\frac{a+1}{2}b}{\frac{3}{2},\mathrm{\hspace{0.33em}1}b\frac{n}{2}}}\right).`$ (19)
To get expressions for negative $`n`$ one may use Theorem 3. Notice that series in (17) and (19) terminate for all $`n`$.
In the rest of this section we follow Whipple’s approach in \[Whi30\], where transformations of not necessarily terminating $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series are used to derive various identities with general $`{}_{2}{}^{}\text{F}_{1}^{}(1)`$ series. We concentrate on the $`{}_{2}{}^{}\text{F}_{1}^{}(1)`$ series which are contiguous to the series in Kummer’s formula (1). Notice that proofs of Theorems 1 and 3 are valid for any complex values of $`n`$, so that formula (2) with $`P(n)`$ and $`Q(n)`$ defined by (3-4) or (13-14) is true for any complex $`n`$. Formula (2) with $`P(n)`$, $`Q(n)`$ defined by (5) is also true for all $`n`$, see Whipple’s formulas (23-24) below. But one may check that in general these $`P(n)`$ and $`Q(n)`$ are not the same.
Transformations of general $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series were first derived by Thomae, see \[Tho79\]. Whipple introduced notation (see \[Whi24\],\[Bai35, 3.5-7\]) which gives a group-theoretical insight into those formulas. To begin with, there is an action of the symmetric group $`S_5`$ on $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$’s. Hardy described it in the notes to lecture VII in \[Har40\] by saying that the function
$$\frac{1}{\mathrm{\Gamma }(E)\mathrm{\Gamma }(F)\mathrm{\Gamma }(E+FABC)}{}_{3}{}^{}\text{F}_{2}^{}\left(\genfrac{}{}{0pt}{}{A,B,C}{E,F}\right)$$
(20)
is invariant under the permutations of $`E`$, $`F`$, $`E+FBC`$, $`E+FAC`$ and $`E+FAB`$. For example, we have (see \[AAR99\], Cor. 3.3.5):
$${}_{3}{}^{}\text{F}_{2}^{}\left(\genfrac{}{}{0pt}{}{A,B,C}{E,F}\right)=\frac{\mathrm{\Gamma }(F)\mathrm{\Gamma }(E+FABC)}{\mathrm{\Gamma }(FC)\mathrm{\Gamma }(E+FAB)}{}_{3}{}^{}\text{F}_{2}^{}\left(\genfrac{}{}{0pt}{}{EA,EB,C}{E,E+FAB}\right).$$
(21)
An orbit of general $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ consists of 10 different series. Note that the series in (20) converge when $`\text{Re}(E+FABC)>0`$, and the whole expression is well-defined and analytic for any parameters under this condition. Function (20) can be analytically continued to the region in the parameter space where at least one of the 10 series converges.
Further, a general $`S_5`$ orbit of $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$’s is associated to 11 other orbits, so that we get sets of 120 allied $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series, see \[Whi24\]. For example<sup>3</sup><sup>3</sup>3Whipple introduced for $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series six parameters $`r_0,\mathrm{},r_5`$ related by condition $`r_i=0`$ so that: all allied series can be obtained by the permutations of the six parameters and/or changing the sign of them all; an $`S_5`$-orbit is determined by fixing a parameter and an element of the set $`\{+,\}`$, and $`S_5`$ permutes the remaining five parameters. Specifically, one may choose that the $`S_5`$ action on (20) fixes $`r_0`$, and take $`E=1+r_4r_0`$., the series in (20) is allied to
$${}_{3}{}^{}\text{F}_{2}^{}\left(\genfrac{}{}{0pt}{}{A,\mathrm{\hspace{0.17em}1}+AE,\mathrm{\hspace{0.17em}1}+AF}{1+AB,\mathrm{\hspace{0.17em}1}+AC}\right)\text{and}{}_{3}{}^{}\text{F}_{2}^{}\left(\genfrac{}{}{0pt}{}{EA,EB,EC}{E,\mathrm{\hspace{0.17em}1}+FE}\right).$$
(22)
In general, two allied series are not related by a two-term identity like (21). But for any three allied series there is a linear relation between them, with coefficients being $`\mathrm{\Gamma }`$-terms. This also gives three-term relations for the 12 functions of type (20), and even defines their analytic continuation to the whole space of parameters. Indeed, if the series in (20) diverges then its ally $`{}_{3}{}^{}\text{F}_{2}^{}\left(\genfrac{}{}{0pt}{}{1A,\mathrm{\hspace{0.17em}1}B,\mathrm{\hspace{0.17em}1}C}{2D,\mathrm{\hspace{0.33em}2}E}\right)`$ converges; for the third term one can take convergent series from a similar pair of functions from other $`S_5`$-orbits. Besides, all allied series converge in a neighborhood of $`A=B=C=1/2`$, $`E=F=1`$.
In \[Whi30\] Whipple applies the relations of allied series to a general $`{}_{2}{}^{}\text{F}_{1}^{}(1)`$ series by expressing it as $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series and considering it as a member of the corresponding allied family. In particular, his formulas (3.1) and (3.51) read as:
$`{}_{2}{}^{}\text{F}_{1}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{a+\nu ,b}{ab}}|1\right)`$ $`=`$ $`{\displaystyle \frac{\mathrm{\Gamma }(ab)\mathrm{\Gamma }(\frac{a}{2})}{\mathrm{\Gamma }(a)\mathrm{\Gamma }(\frac{a}{2}b)}}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{\nu 1}{2},\frac{\nu }{2},b}{\nu ,\frac{a+1}{2}}}\right)`$ (23)
$`=`$ $`{\displaystyle \frac{\mathrm{\Gamma }(ab)\mathrm{\Gamma }(\frac{a+1}{2})}{\mathrm{\Gamma }(a)\mathrm{\Gamma }(\frac{a+1}{2}b)}}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{\nu }{2},\frac{\nu +1}{2},b}{\nu ,\frac{a}{2}}}\right).`$ (24)
If $`\nu \{0,1,2,\mathrm{},\}`$ we may relate the $`{}_{2}{}^{}\text{F}_{1}^{}(1)`$ series with the $`S_5`$-orbit of the $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series in (23-24) and get many two- and three-term relations with $`{}_{2}{}^{}\text{F}_{1}^{}(1)`$ and $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series. Some of these identities make sense and are correct even if $`\nu `$ is a non-negative integer, because singular $`\mathrm{\Gamma }`$-factors cancel. For instance, formula (2) with $`P(n)`$ and $`Q(n)`$ defined by (3-4) is a three term identity between allied series, see the last paragraph of \[Whi30\]. Similarly, (potentially) terminating series in Whipple’s formulas (8.3) and (8.41) are derived from three term identities of allied series.
On the other hand, the $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series in (23-24) cannot be identified with the terminating series in the expressions in (5). One has to compute:
$`\underset{\nu n}{lim}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{\nu }{2},\frac{\nu +1}{2},b}{\nu ,\frac{a}{2}}}\right)=2P(n){\displaystyle \frac{1}{4^{n+1}}}{\displaystyle \frac{(b)_{n+1}}{(\frac{a}{2})_{n+1}}}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n+2}{2},\frac{n+1}{2},b+n+1}{n+2,\frac{a}{2}+n+1}}\right),`$
$`\underset{\nu n}{lim}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{\nu 1}{2},\frac{\nu }{2},b}{\nu ,\frac{a+1}{2}}}\right)=2Q(n)+{\displaystyle \frac{1}{4^{n+1}}}{\displaystyle \frac{(b)_{n+1}}{(\frac{a+1}{2})_{n+1}}}{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n+3}{2},\frac{n+2}{2},b+n+1}{n+2,\frac{a+1}{2}+n+1}}\right).`$
In the sum of these two equalities the non-terminating $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series on the right-hand side cancel, since they are connected by transformation (21). In this way identities (23-24) prove Theorem 2.
Moreover, the $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series (23-24) can be transformed by $`S_5`$ to four series which are well-defined and terminate when $`\nu `$ is an (odd or even) positive integer $`n`$. Those terminating series are presented in formulas (16) and (18). However, this does not give expressions for $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{a+n,b}{ab}|1\right)`$ in terms of one terminating $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series, because the mentioned four series diverge for $`\nu >1/2`$ (except when they terminate), and we cannot use the $`S_5`$-invariance of the corresponding function in (20). Notice, for example, that (21) implies a wrong relation between the $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series in (16) and (18). As we see, Whipple’s approach in \[Whi30\] gets complicated in the case $`\nu `$ in (23-24) is an integer, and does not directly explain various expressions for our $`P(n)`$ and $`Q(n)`$.
## 3 A proof by Zeilberger’s method
Here we prove Theorem 2 only. Theorem 1 can be proved in the same way.
Let us define $`S(n)={}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{a+n,b}{ab}|1\right)`$. The contiguity relation \[Erd53, 2.8(28)\] between $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{A+1,B}{C}|z\right)`$, $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{A1,B}{C}|z\right)`$ and $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{A,B}{C}|z\right)`$ gives the following recurrence relation:
$$2(n+a)S(n+1)(3n+2a)S(n)+(n+b)S(n1)=0.$$
(25)
We claim that the sequences $`P(n)`$ and $`Q(n)`$ satisfy the same recurrence relation. Following the “creative telescoping” method of Zeilberger (\[PWZ96, Koe98\]), let
$$p(n,k)=\frac{(1)^k}{24^k}\frac{(n+1)(nk)!}{(n2k+1)!k!}\frac{(b)_k}{(\frac{a}{2})_k}$$
(26)
be the $`k`$th summand of $`P(n)`$ in (5). We set $`p(n,k)=0`$ for $`k>n/2`$. Also define
$$r_1(n,k)=\frac{2k(nk+1)(a+2k2)}{(n2k+2)(n2k+3)},R_1(n,k)=r_1(n,k)p(n,k).$$
One can check that
$`2(n+a)p(n+1,k)(3n+2a)p(n,k)+(n+b)p(n1,k)=R_1(n,k+1)R_1(n,k),`$
so
$`2(n+a)P(n+1)(3n+2a)P(n)+(n+b)P(n1)`$ $`=`$
$`{\displaystyle \underset{k=0}{\overset{n/2}{}}}\left(R_1(n,k+1)R_1(n,k)\right)R_1(n,\frac{n+1}{2})`$ $`=`$ $`0.`$ (27)
Although this looks like an artificial trick, we follow the standard Wilf-Zeilberger method of proving combinatorial identities, see \[PWZ96, Koe98\]. The expression $`r_1(n,k)`$ is the certificate of our standardized proof. Given $`p(n,k)`$ the recurrence relation for $`P(n)`$ and the certificate $`r_1(n,k)`$ can be found by Zeilberger’s algorithm. This algorithm is implemented in computer algebra packages EKHAD (see \[Zei99\], command ct) and hsum.mpl (see \[Koe99\], command sumrecursion with option certificate=true). Also check \[VK00\] for a link to a Maple worksheet for this proof. The finite sums in this proof require some attention, since they are not natural according to \[Koe98\].
In the same way:
$`2(n+a)Q(n+1)(3n+2a)Q(n)+(n+b)Q(n1)`$ $`=`$
$`{\displaystyle \underset{k=0}{\overset{(n1)/2}{}}}\left(R_2(n,k+1)R_2(n,k)\right)R_2(n,\frac{n}{2})`$ $`=`$ $`0,`$ (28)
where
$$R_2(n,k)=\frac{2k(nk+1)(a+2k1)}{(n2k+1)(n2k+2)}\frac{(1)^k}{24^k}\left(\genfrac{}{}{0pt}{}{nk}{k}\right)\frac{(b)_k}{(\frac{a+1}{2})_k}.$$
(29)
is the $`k`$th summand of $`Q(n)`$ in (5) multiplied by the corresponding certificate.
Note that the condition $`(a)_n0`$ ensures that recurrence relation (25) does not degenerate to a first order relation until we evaluate $`P(n)`$ and $`Q(n)`$. It remains to check that formula (2) holds for two initial values of $`n`$. Kummer’s identity (1) suggests $`P(1)=1`$ and $`Q(1)=0`$, which fits into the recurrence relation. We may use Gauss’ contiguity relation \[Erd53, 2.8(38)\] between $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{A,B}{C+1}|z\right)`$, $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{A,B}{C}|z\right)`$ and $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{A1,B}{C}|z\right)`$ to obtain
$$(a2b)\frac{\mathrm{\Gamma }(1+ab)\mathrm{\Gamma }(1+\frac{a}{2})}{\mathrm{\Gamma }(1+a)\mathrm{\Gamma }(1+\frac{a}{2}b)}2(ab)S(0)+(ab)S(1)=0.$$
(30)
This implies the correct $`P(0)=1/2`$ and $`Q(0)=1/2`$ and completes the proof.
Note that the Gauss contiguity relations hold for analytic extensions of hypergeometric functions on $`\text{C}[1,\mathrm{})`$. Therefore this proof does not require convergence of the $`{}_{2}{}^{}\text{F}_{1}^{}(1)`$ series. Q.E.D.
In fact, sequences $`P(n)`$ and $`Q(n)`$ satisfy recurrence relation (25) for all $`n`$. The recurrence can be directly verified for $`n=2,1,0`$. The values of $`P(n)`$ and $`Q(n)`$ for $`n=3,2,1,0,1`$ are
$$\begin{array}{cccccc}\frac{2(a2)(ab2)}{(b1)(b2)},& \frac{a2}{b1},& 1,& \frac{1}{2},& \frac{ab}{2a}& \text{and}\hfill \\ \frac{2(a1)(a3)}{(b1)(b2)},& \frac{a1}{b1},& 0,& \frac{1}{2},& \frac{1}{2}& \text{respectively.}\hfill \end{array}$$
To compute the same recurrence relation for negative $`n`$ one can use Theorem 1. Alternatively, one may choose an expression for $`P(n)`$ and $`Q(n)`$ for negative $`n`$, say (13-14), and compute the recurrence relation with Zeilberger’s algorithm.
To show equalities like in (16-18) by Zeilberger’s method one would have to compute the recurrences for odd and even integers separately. Recurrence relation (25) for any such expression and for all $`n`$ can be computed using contiguity relations for $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series. As it is known (see \[AAR99, 3.7\]), contiguous $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series satisfy three-term relations (with coefficients being rational functions in the parameters of those series), just like contiguous $`{}_{2}{}^{}\text{F}_{1}^{}(z)`$ series.
## 4 Algorithmic aspects
The generalized formula (2) can be specialized so that $`P(n)`$ or $`Q(n)`$ vanishes, giving an evaluation of $`{}_{2}{}^{}\text{F}_{1}^{}(1)`$ series with a single $`\mathrm{\Gamma }`$-term. For example,
$$Q(4)=4\frac{(a1)(a3)(2ab7)}{(b1)(b2)(b3)},$$
(31)
so if $`b=2a7`$ then $`Q(4)=0`$, which implies
$${}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{3c,\mathrm{\hspace{0.17em}7}2c}{c}|1\right)=\frac{3}{4}\frac{\mathrm{\Gamma }(c)\mathrm{\Gamma }(3\frac{c}{2})}{\mathrm{\Gamma }(5c)\mathrm{\Gamma }(\frac{3c}{2}2)}.$$
(32)
Further, $`P(5)=0`$ if $`2a^24ab+b^212a+17b+12=0`$. Parameterizing the curve given by this equation we get
$${}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{\frac{2t^27t+6}{t^22},\frac{t^2+4t8}{t^22}}{\frac{2t^2+3t8}{t^22}}|1\right)=\frac{t^2+3t6}{t(t1)}\frac{\mathrm{\Gamma }\left(\frac{3t4}{t^22}\right)\mathrm{\Gamma }\left(\frac{t^2+7t12}{2(t^22)}\right)}{\mathrm{\Gamma }\left(\frac{7t10}{t^22}\right)\mathrm{\Gamma }\left(\frac{t(t1)}{2(t^22)}\right)}.$$
(33)
It could be expected that formulas like (32) can be proved automatically by current computer algebra algorithms, say by Wilf-Zeilberger method. As it is demonstrated in \[Koo98\], this method or Zeilberger’s algorithm can be adapted to non-terminating hypergeometric series if one can justify the “creative telescoping” trick by dominated convergence, and the hypergeometric series can be evaluated in the limit $`n\mathrm{}`$, where $`n`$ is a discrete parameter. In general non-terminating hypergeometric series is given without a discrete parameter, so it must be introduced by an algorithm. For example, after substitution $`aa+2n`$ one can prove Kummer’s formula (1) with Wilf-Zeilberger method, see \[Gau99\].
In the case of equation (32) we may substitute $`cc+n`$ and apply Zeilberger’s algorithm to get the right first order difference equation. However, we cannot evaluate the hypergeometric series neither in the limit $`n\mathrm{}`$, nor for a finite value of $`n`$. What we can do is to combine explicitly Gauss’ contiguity relations in such a way that we “accidentally” get a two-term relation where one of the terms can be evaluated by Kummer’s formula. For example, the relation between contiguous $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{A,B}{C}|z\right)`$, $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{A+1,B2}{C}|z\right)`$ and, say, $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{A,B1}{C}|z\right)`$, after the specialization $`(A,B,C,z)(3c,\mathrm{\hspace{0.17em}7}2c,c1)`$ becomes
$${}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{3c,\mathrm{\hspace{0.17em}7}2c}{c}|1\right)=\frac{3}{4}{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{4c,\mathrm{\hspace{0.17em}5}2c}{c}|1\right),$$
(34)
In this way even the exotic (33) can be proved.
This shows that relations between contiguous hypergeometric series can be useful for finding new “non-standard” evaluations of $`{}_{2}{}^{}\text{F}_{1}^{}`$ series. One may take such a relation and try to find families of its two term specializations with a discrete parameter $`n`$. This would give a first order recurrence relation, and if the series can be evaluated in the limit $`n\mathrm{}`$ one gets a (perhaps) new formula! Relations between contiguous series also give a way to compute recurrence relation, alternative to Zeilberger’s algorithm.
In \[VK00\] there is a link to Maple routines which for given three integer vectors $`(k_i,l_i,m_i)`$ for $`i=1,2,3`$ derive a $`\text{C}(A,B,C,z)`$-linear relation between three contiguous functions $`{}_{2}{}^{}\text{F}_{1}^{}\left(\genfrac{}{}{0pt}{}{A+k_i,B+l_i}{C+m_i}|z\right)`$. Computer experiments found many first order recurrence relations for some values $`z=1/4`$, $`1/3`$, $`1/9`$, $`\mathrm{exp}(i\pi /3)`$, $`32\sqrt{2},\mathrm{}`$, some of them can be successfully solved. It is an interesting question which $`{}_{2}{}^{}\text{F}_{1}^{}(z)`$ series can be evaluated in terms of $`\mathrm{\Gamma }`$-function. So far produced evaluations can be obtained using quadratic or cubic transformations.
Here we generalize a few known formulas of the same type as (2). They were obtained by considering relations between three contiguous hypergeometric series where two of them can be evaluated by a known formula, and trying to express the coefficients in these relations as hypergeometric series. This was done by considering partial fraction decomposition of those coefficients empirically. The formulas can be proved by showing that all three terms in a formula satisfy the same recurrence relation by Zeilberger’s algorithm, and checking the identity for a couple of values of the discrete parameter.
We start with a generalization of Gosper’s “non-standard” evaluations of $`{}_{2}{}^{}\text{F}_{1}^{}(1/4)`$ series, see \[Gos80, 1/4.1-2\]. A generalization is
$`{}_{2}{}^{}\text{F}_{1}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{a,\frac{1}{2}}{2a+\frac{3}{2}+n}}|{\displaystyle \frac{1}{4}}\right)={\displaystyle \frac{2^{n+3/2}}{3^{n+1}}}{\displaystyle \frac{\mathrm{\Gamma }(a+\frac{5}{4}+\frac{n}{2})\mathrm{\Gamma }(a+\frac{3}{4}+\frac{n}{2})\mathrm{\Gamma }(a+\frac{1}{2})}{\mathrm{\Gamma }(a+\frac{7}{6}+\frac{n}{3})\mathrm{\Gamma }(a+\frac{5}{6}+\frac{n}{3})\mathrm{\Gamma }(a+\frac{1}{2}+\frac{n}{3})}}K(n)`$
$`(3)^{n2}\mathrm{\hspace{0.17em}2}^{3/2}{\displaystyle \frac{\mathrm{\Gamma }(a+\frac{5}{4}+\frac{n}{2})\mathrm{\Gamma }(a+\frac{3}{4}+\frac{n}{2})\mathrm{\Gamma }(a+1)}{\mathrm{\Gamma }(a+\frac{3}{2})\mathrm{\Gamma }(a+\frac{1}{2}+\frac{n}{2})\mathrm{\Gamma }(a+1+\frac{n}{2})}}L(n),`$ (35)
where
$$K(1)=L(0)=0,K(0)=L(1)=1,$$
for $`n>1`$:
$`K(n)`$ $`=`$ $`(1)^n{\displaystyle \underset{k=n/3}{\overset{n/2}{}}}{\displaystyle \frac{27^k}{4^k}}{\displaystyle \frac{n(k1)!}{(n2k)!(3kn)!}}{\displaystyle \frac{(a+\frac{1}{2})_k}{(a+1)_k}},`$
$`L(n)`$ $`=`$ $`{}_{4}{}^{}\text{F}_{3}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n1}{3},\frac{n2}{3},\frac{n3}{3},a+1}{\frac{n2}{2},\frac{n3}{2},a+\frac{3}{2}}}\right),`$
and for $`n<0`$:
$`K(n)`$ $`=`$ $`{}_{4}{}^{}\text{F}_{3}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n}{3},\frac{n1}{3},\frac{n2}{3},a}{\frac{n1}{2},\frac{n2}{2},a+\frac{1}{2}}}\right)={\displaystyle \underset{k=0}{\overset{n/3}{}}}{\displaystyle \frac{(4)^k}{27^k}}{\displaystyle \frac{n(n2k1)!}{(n3k)!k!}}{\displaystyle \frac{(a)_k}{(a+\frac{1}{2})_k}},`$
$`L(n)`$ $`=`$ $`(1)^n{\displaystyle \underset{k=(n+1)/3}{\overset{(n+1)/2}{}}}{\displaystyle \frac{27^k}{4^k}}{\displaystyle \frac{(n+1)(k1)!}{(n2k+1)!(3kn1)!}}{\displaystyle \frac{(a\frac{1}{2})_k}{(a)_k}}.`$
Gosper has found the special cases $`n=0,1`$. The $`\mathrm{\Gamma }`$-factors to $`K(n)`$ and $`L(n)`$ are $`\text{C}(a)`$-multiples of these two Gosper’s evaluations (respectively) for each $`n`$. All three terms in (4) satisfy the recurrence relation
$`2(n+2a+1)(2n+6a+3)S(n+1)+(2n+4a+3)(4n+6a+1)S(n)`$
$`3(2n+4a+1)(2n+4a+3)S(n1)`$ $`=`$ $`0.`$
Next we recall the classical Dixon’s identity which evaluates well-poised $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series, see \[Bai35, 3.1\]. We generalize it as follows:
$`{}_{3}{}^{}\text{F}_{2}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{a+n,b,c}{ab,ac}}\right)`$ $`=`$ $`{\displaystyle \frac{\stackrel{~}{P}(n)}{2}}{\displaystyle \frac{\mathrm{\Gamma }(\frac{a+1}{2})\mathrm{\Gamma }(ab)\mathrm{\Gamma }(ac)\mathrm{\Gamma }(\frac{a+1}{2}bc)}{\mathrm{\Gamma }(a)\mathrm{\Gamma }(\frac{a+1}{2}b)\mathrm{\Gamma }(\frac{a+1}{2}c)\mathrm{\Gamma }(abc)}}`$ (36)
$`+{\displaystyle \frac{\stackrel{~}{Q}(n)}{2}}{\displaystyle \frac{\mathrm{\Gamma }(\frac{a}{2})\mathrm{\Gamma }(ab)\mathrm{\Gamma }(ac)\mathrm{\Gamma }(\frac{a}{2}bc)}{\mathrm{\Gamma }(a)\mathrm{\Gamma }(\frac{a}{2}b)\mathrm{\Gamma }(\frac{a}{2}c)\mathrm{\Gamma }(abc)}},`$
where $`\stackrel{~}{P}(1)=1`$, $`\stackrel{~}{Q}(1)=0`$, then for $`n0`$:
$`\stackrel{~}{P}(n)={}_{4}{}^{}\text{F}_{3}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n}{2},\frac{n+1}{2},b,c}{n,\frac{a}{2},\frac{1a}{2}+b+c}}\right),\stackrel{~}{Q}(n)={}_{4}{}^{}\text{F}_{3}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n1}{2},\frac{n}{2},b,c}{n,\frac{1+a}{2},\mathrm{\hspace{0.17em}1}\frac{a}{2}+b+c}}\right),`$
and for $`n<0`$:
$`\stackrel{~}{P}(n1)`$ $`=`$ $`2^{2n}{\displaystyle \frac{(1\frac{a}{2})_n(\frac{1+a}{2}bc)_n}{(1b)_n(1c)_n}}{}_{4}{}^{}\text{F}_{3}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n}{2},\frac{n1}{2},bn,cn}{1n,\frac{a}{2}n,\frac{1a}{2}+b+cn}}\right),`$
$`\stackrel{~}{Q}(n1)`$ $`=`$ $`2^{2n}{\displaystyle \frac{(\frac{1a}{2})_n(\frac{a}{2}bc)_n}{(1b)_n(1c)_n}}{}_{4}{}^{}\text{F}_{3}^{}\left({\displaystyle \genfrac{}{}{0pt}{}{\frac{n1}{2},\frac{n2}{2},bn,cn}{1n,\frac{1+a}{2}n,\mathrm{\hspace{0.17em}1}\frac{a}{2}+b+cn}}\right).`$
Dixon’s identity is the special case $`n=1`$. This generalized formula is a relation between contiguous $`{}_{3}{}^{}\text{F}_{2}^{}(1)`$ series in explicit form. For positive $`n`$ it is strikingly similar to generalization (2,5) of Kummer’s identity. In fact, the generalization in Theorem 2 is the limiting case $`c\mathrm{}`$ of (36), just as Kummer’s formula is the limiting case of Dixon’s identity. The recurrence relation for the three terms in (36) is:
$`(n+a)(na+2b+2c+1)S(n+1)+(n+b)(n+c)S(n1)`$
$`(2n^2+3bn+3cn+na^2+2ab+2ac+a)S(n)`$ $`=`$ $`0.`$
More evaluations of the same type can be obtained using standard transformations of $`{}_{2}{}^{}\text{F}_{1}^{}(z)`$ series to $`{}_{2}{}^{}\text{F}_{1}^{}(z/(z1))`$ series, see \[Erd53, 2.9(3-4)\]. Applying them to the generalized Kummer’s formula (2) gives evaluations of $`{}_{2}{}^{}\text{F}_{1}^{}(1/2)`$ which generalize classical formulas of Gauss and Bailey, see \[Bai35, 2.4\]. The same transformation of (4) gives evaluation of $`{}_{2}{}^{}\text{F}_{1}^{}(1/3)`$. Similarly, one can apply (21) to identity (36) and get generalizations of Watson’s and Whipple’s formulas \[Bai35, 3.3-4\].
All these formulas evaluate hypergeometric series which are contiguous to a series which evaluation is known. In order to find these formulas automatically one needs an algorithm which would find the solutions of a recurrence relation in form of terminating hypergeometric series.
|
warning/0005/astro-ph0005315.html
|
ar5iv
|
text
|
# Dynamics of line-driven disk winds in Active Galactic Nuclei.
## 1 Introduction
Many observed spectral features of active galactic nuclei (AGN) indicate that outflows are common in these systems. The prominent broad emission lines (BEL) in ultraviolet (UV) from H I, O VI, N V, C IV, and Si IV are the defining feature of quasars (Blandford, Netzer & Woltjer 1990; Osterbrock 1989), and they may be associated with a high velocity outflow. Fast outflows can also explain narrow UV absorption lines from highly ionized species – such as C IV and N V – observed in approximately half of the HST-observed Seyfert 1 galaxies (e.g., Crenshaw 1997). These narrow absorption lines are all blueshifted relative to the systemic velocity by 0 to -1600 km s<sup>-1</sup>. Recent ASCA observations show that in a small sample observed by HST and ASCA, all of the Seyfert galaxies with warm absorbers also show intrinsic UV absorption. A high-resolution X-ray observation of the Seyfert galaxy NGC 5548 obtained by Chandra shows strong, narrow absorption lines from highly ionized species (Kaastra et al. 2000). The lines are blueshifted by a few hundred km $`\mathrm{s}^1`$ and they are reminiscent of the narrow absorption lines observed in the UV. Some QSOs also have intrinsic UV narrow-line absorbers with line-of-sight velocities as large as $`51000\mathrm{km}\mathrm{s}^1`$ (Hamann et al. 1997). Perhaps the most compelling evidence for fast outflows in QSOs are strong broad absorption lines (BALs) in the UV resonance lines of highly ionized species such as N V, C IV, and Si IV. The BALs are always blueshifted relative to the emission-line rest frame, indicating the presence of outflows from the active nucleus, with velocities as large as $`60000\mathrm{kms}^1`$.
A key constraint on any model for the origin of AGN outflows is the ionization balance. On one hand we observe very high luminosities in X-rays and the UV and on the other hand we observe spectral lines from moderately and highly ionized species. One wonders then how the gas avoids full photoionization and we see any spectral lines at all. Generally, two mechanisms have been proposed to resolve the so-called overionizaton problem: (i) the AGN outflows have filling factors less than one and consist of dense clouds and (ii) the filling factors equal one but the outflows are shielded from the powerful radiation by some material located between the central engine and the outflow (e.g., Krolik 1999, and below).
Many theoretical models have been proposed to explain outflows in AGNs. Comprehensive reviews of recent theoretical work on outflows in AGNs can be found in Arav, Shlosman & Weymann (1997). Several forces have been suggested to accelerate outflows in AGN, for example, gas pressure (e.g., Weymann et al. 1982; Begelman, de Kool & Sikora 1991), magneto-centrifugal force due to an accretion disk (e.g., Blandford & Payne 1982; Emmering, Blandford & Shlosman 1992; K$`\ddot{\mathrm{o}}`$nigl & Kartje 1994; Bottorff et al. 1997), radiation pressure due to dust (e.g., Voit, Weymann & Korista 1993; Scoville & Norman 1995), and radiation pressure due to lines (e.g., Drew & Boksenberg 1984; Shlosman, Vitello & Shaviv 1985; Arav, Li & Begelman 1994; de Kool & Begelman 1995, hereafter dKB; Murray et al. 1995, hereafter MCGV). Most of these models avoid overionization of the wind by assuming high density clouds.
One plausible scenario for the AGN outflows is that radiation pressure on spectral lines drives a wind from an accretion disk around a black hole. The presence of the BALs themselves strongly indicates that substantial momentum is transfered from a powerful radiation field to the gas. A crucial clue to the origin of AGN outflows comes from the discovery of line-locking in BAL QSO spectra (e.g., Foltz et al. 1987). Weymann et al. (1991) discovered that a composite spectrum of their BAL QSO sample shows a double trough in the C IV $`\lambda `$1549 BAL, separated in velocity space by the N V $`\lambda `$ 1240-Ly$`\alpha `$ splitting $`5,900`$ km $`\mathrm{s}^1`$ (see also Korista et al. 1993). In fact Arav & Begelman (1994, see also Arav et al. 1995 and Arav 1996) showed that an absorption hump in the C IV $`\lambda `$ BAL is “the ghost of Ly$`\alpha `$” due to modulation of the radiation force by the strong emission in Ly$`\alpha `$.
Our understanding of how line-driving produces powerful high velocity winds is based on the studies of winds in hot stars that radiate mostly in the UV (e.g., Castor, Abbott, Klein 1975, hereafter CAK; Abbott 1982). The essential concept underpinning these models is that the momentum is extracted most efficiently from the radiation field via line opacity. With the inclusion of lines, CAK showed that the effective radiation force can increase by several orders of magnitude above that due to electron-scattering alone. Thus even a star that radiates at around 0.1% of its Eddington limit, $`L_{Edd}`$ can have a strong wind. Up to now, however, it has been difficult to apply line-driven stellar wind models to QSOs because of the fundamental difference in geometry: stellar winds are to a good approximation spherically symmetric, whereas the wind in BAL QSOs likely arises from a flattened disk and is therefore axially symmetric.
Line-radiation driven disk wind models for BAL QSOs have been proposed by several authors. For example, MCGV studied a wind arising from a small distance from the central engine ($`10^{16}`$ cm). Their model relies on the local disk radiation to launch gas from the disk and on the central radiation to accelerate the gas in the radial direction to high velocities. A key ingredient to this model is that the central engine radiation is attenuated by ’hitchhiking’ gas located between the central engine and wind. Rather than showing from first principles the origin of this hitchhiking gas, they simply give plausible arguments for its existence. MCGV adopted assumptions that require the flow to be time-independent and restrict the flow geometry to a quasi 1-D radial flow. Other models of radiation-driven outflows in AGNs propose that the radiation force accelerates the wind that has been launched and kept at a low ionization state by different mechanisms. For example, the dKB model supposes that the wind is seated at a large distance ($`10^{18}`$ cm) from the central engine. Because the disk is cool at these radii, the gas is not lifted from the disk by radiation pressure but by some alternative mechanism. Once the gas is high enough above the disk, the radiation force due to the central engine accelerates the gas to high velocity. The gas is not over-ionized in this model due to a small filling factor caused by strong magnetic fields which confine the dense clouds.
Recently it has been possible to model line-driven disk winds using 2-dimensional axisymmetric numerical hydrodynamical simulations (e.g., Pereyra, Kallman & Blondin 1997; PSD I; Proga 1999; Proga, Stone & Drew 1999, hereafter PSD II). All these models have been calculated for white dwarf accretion disks. PSD I found that line-driven disk winds are produced only when the effective luminosity of the disk (i.e., the luminosity of the disk times the maximum value of the force multiplier, $`L_DM_{max}`$) exceeds the Eddington limit, $`L_{Edd}`$. If the dominant contribution to the total radiation field comes from the disk, then the outflow is intrinsically unsteady and characterized by large amplitude velocity and density fluctuations. On the other hand, if the total luminosity of the system is dominated by the central object, then the outflow is steady. In either case, PSD I and PSD II found that the structure of the wind consists of a dense, slow outflow, that is bounded on the polar side by a high-velocity, lower density stream. The flow geometry is controlled largely by the geometry of the radiation field – a brighter disk/central object produces a more polar/equatorial wind. Global properties such as the total mass loss rate and terminal velocity depend more on the system luminosity and are insensitive to geometry. The mass loss rate is a strong function of the effective Eddington luminosity and is of the same order of magnitude as that of a simple spherically-symmetric stellar wind (see also Proga 1999). Matter is fed into the fast stream from within a few central object radii. The terminal velocity of the stream is similar to that of the terminal velocity of a corresponding spherical stellar wind, i.e., $`v_{\mathrm{}}\mathrm{a}\mathrm{few}v_{esc}`$, where $`v_{esc}`$, is the escape velocity from the photosphere.
In this paper, we use the methods developed by PSD to study disk winds in AGNs. However to study AGNs winds, we need to relax some of the assumptions adopted by PSD. In particular, we can no longer assume the outflowing gas is isothermal, has a fixed ionization state, or is optically thin. Instead, we adopt a simplified treatment of photoionization, and radiative cooling and heating that allow us to compute self-consistently the ionization state, and therefore line force, in the wind.
Here we calculate a few disk wind models using our extensions to the PSD II method. We concentrate on assessing how winds can be driven from a disk in the presence of very strong ionizing radiation. We describe our calculation in Section 2. We present our results in Section 3 and discuss them together with perceived limitations in Section 4. The paper ends, in section 5, with our conclusions.
## 2 Method
### 2.1 Hydrodynamics
To calculate the structure and evolution of a wind from a disk, we solve the equations of hydrodynamics
$$\frac{D\rho }{Dt}+\rho 𝐯=0,$$
(1)
$$\rho \frac{D𝐯}{Dt}=P+\rho 𝐠+\rho 𝐅^{rad}$$
(2)
$$\rho \frac{D}{Dt}\left(\frac{e}{\rho }\right)=p𝐯+\rho ,$$
(3)
where $`\rho `$ is the mass density, $`P`$ is the gas pressure, $`𝐯`$ is the velocity, $`e`$ is the internal energy density, $``$ is the net cooling rate, $`𝐠`$ is the gravitational acceleration of the central object, $`𝐅^{rad}`$ is the total radiation force per unit mass. We adopt an adiabatic equation of state $`P=(\gamma 1)e`$, and consider models with $`\gamma =5/3`$.
Our calculations are performed in spherical polar coordinates $`(r,\theta ,\varphi )`$. We assume axial symmetry about the rotational axis of the accretion disk ($`\theta =0^o`$). However the $`\theta =90^o`$ axis is not coincident with the disk midplane. We allow this axis to be above the midplane, at the height, $`z_o`$, corresponding to the disk pressure scale height. Figure 1 shows a schematic representation of our computation domain. The offset enters our calculations in the computation of the gravity; for example the gravity has non-zero $`\theta `$ component at $`\theta =90^o`$, and in the computation of the Keplerian velocity along the $`\theta =90^o`$ axis.
This offset is introduced so that the bottom of our computational grid is located at the disk photosphere, where the wind is launched. Unlike the thin disk studied by PSD, the radiation dominated disk in AGN may have significant geometrical width. If the base of the computational grid is placed at the equatorial plane (midplane), this would require modeling the full internal structure of the disk, which is complex and turbulent (e.g., Balbus & Hawley 1998). Instead, we locate the base of the grid where the vertical radiation force due to electron scattering cancels out the vertical component of gravity.
A realistic description of the radiation field from the central engine would require detailed knowledge of the geometry of the flow near the central engine, which is beyond the scope of this investigation. Instead we model the central engine as a point source of radiation located at the origin of our computational grid (that is located a height $`z_o`$ above the disk midplane). This implies the central engine has finite width, perhaps associated with a hot corona. We expect the high column density of the disk to attenuate radiation close to the $`\theta =90^o`$ plane.
Our standard computational domain is defined to occupy the radial range $`r_i=100r_{}rr_o=1000r_{}`$, where $`r_{}`$ is the inner radius of the disk, and the angular range $`0^o\theta 90^o`$. For comparison, we also calculate some models with the radial range $`50r_{}r1000r_{}`$ and with the radial range $`200r_{}r1000r_{}`$. The $`r\theta `$ domain is discretized into zones. Our numerical resolution consists of 100 zones in each of the $`r`$ and $`\theta `$ directions, with fixed zone size ratios, $`dr_{k+1}/dr_k=1.05`$ and $`d\theta _l/d\theta _{l+1}=1.087`$. Gridding in this manner ensures good spatial resolution close to the radiating surface of the disk and the inner boundary at $`r_i`$.
The boundary conditions are specified as follows. At $`\theta =0`$, we apply an axis-of-symmetry boundary condition. For the outer radial boundary, we apply an outflow boundary condition. For the inner radial boundary $`r=r_{}`$ and for $`\theta =90^o`$, we apply reflecting boundary conditions for the density, velocity and internal energy. To represent steady conditions in the photosphere at the base of the wind, during the evolution of each model we apply the constraints that in the first zone above the $`\theta =90^o`$ plane the radial velocity $`v_r=0`$, the rotational velocity $`v_\varphi `$ remains Keplerian, and the density is fixed at $`\rho =\rho _0`$ at all times. The initial conditions are as in PSD I expect for the initial temperature profile which is described in detail in Section 2.3.
To solve eqs. 1-3 we use an extended version of the ZEUS-2D code (e.g., Stone & Norman 1992). The equation for the internal energy, eq. 3, is solved using the operator splitting method and the backward Euler scheme.
### 2.2 The radiation field and force
The geometry and assumptions needed to compute the radiation field from the disk and central object are as in PSD II (see also PSD I). The disk is flat, Keplerian, geometrically-thin and optically-thick. The disk photosphere is coincident with the $`\theta =90^o`$ axis. We specify the radiation field of the disk by assuming that the temperature follows the radial profile of the optically thick accretion disk (Shakura & Sunyaev 1973), and therefore depends on the mass accretion rate in the disk, $`\dot{M}_a`$, the mass of the black hole, $`M_{BH}`$ and the inner edge of the disk, $`r_{}=3r_S`$, where $`r_S=2GM_{BH}/c^2`$ is the Schwarzschild radius of a black hole. In particular, the disk luminosity, $`L_D=2\eta GM_{BH}\dot{M}_a/r_S`$, where $`\eta `$ is the rest mass conversion efficiency.
The geometry of the central engine in AGNs is not well known. For simplicity, we consider the central engine as the most inner part of the accretion disk plus an extended corona. We refer to the corona as the central object. The radius of the central object is comparable with the inner radius of the disk and we assume that they formally are equal.
We express the central object luminosity $`L_{}`$ in units of the disk luminosity $`L_{}=xL_D`$. In contrast to PSD I and PSD II, we allow for the situation when only some fraction of the central object luminosity takes part in driving a wind. We identify this fraction as the luminosity in the UV band, $`f_{\mathrm{UV}}L_{}`$. We refer to the fraction of the luminosity that is responsible for ionizing the wind to a very high state as the luminosity in the X-ray band, $`f_\mathrm{X}L_{}`$. For simplicity, we assume here that this fraction of the luminosity does not contribute to line driving of the wind. We call the luminosity in the remaining bands, mainly optical and infrared, as $`f_{\mathrm{O},\mathrm{IR}}L_{}`$. We assume that $`f_{\mathrm{O},\mathrm{IR}}L_{}`$ is the part of the luminosity that does not change the dynamics of the wind. We stress that the fraction $`f_i`$ is a numerical factor that we introduce here to parameterize the luminosity in each of those three domains. We set $`f_{\mathrm{OPT},\mathrm{IR}}`$ to zero in the remaining part of the paper as the central object is very hot. We take into account the irradiation of the disk by the central object, assuming that the disk re-emits all absorbed energy locally and isotropically. However we note that the contribution from irradiation is negligible for $`x1`$ and large radii (see eq. 6 below).
We approximate the radiative acceleration due to lines (line force, for short) using a modified CAK method. The line force at a point defined by the position vector $`𝐫`$ is
$$𝐅^{rad,l}(𝐫)=_\mathrm{\Omega }M(t)\left(\widehat{n}\frac{\sigma _eI(𝐫,\widehat{n})d\mathrm{\Omega }}{c}\right)$$
(4)
where $`I`$ is the frequency-integrated continuum intensity in the direction defined by the unit vector $`\widehat{n}`$, and $`\mathrm{\Omega }`$ is the solid angle subtended by the disk and central object at the point. The term in brackets is the electron-scattering radiation force, $`\sigma _e`$ is the mass-scattering coefficient for free electrons, and $`M(t)`$ is the force multiplier – the numerical factor which parameterizes by how much spectral lines increase the scattering coefficient. In the Sobolev approximation, $`M(t)`$ is a function of the optical depth parameter
$$t=\frac{\sigma _e\rho v_{th}}{\left|dv_l/dl\right|},$$
(5)
where $`v_{th}`$ is the thermal velocity, and $`\frac{dv_l}{dl}`$ is the velocity gradient along the line of sight, $`\widehat{n}`$.
We evaluate the radiation force in four steps. First, we calculate the intensity, the velocity gradient in the $`\widehat{n}`$ direction and then the optical depth parameter $`t`$. Second, we calculate the parameters of the force multiplier using a current value of the photoionization parameter, $`\xi `$ adopting results of Stevens & Kallman (1990). Then we calculate the radiation force exerted by radiation along $`\widehat{n}`$. Third, we integrate the radiation force over the solid angle subtended by the radiant surface. Finally, we correct the radiation force in the radial direction for the optical depth effects.
Our numerical algorithm for evaluating the line force for given parameters of the force multiplier, the third step, is described in PSD II. For a rotating flow, there may be an azimuthal component to the line force even in axisymmetry. However we set this component of the line force to zero because it is always less than other components (e.g., PSD II). See PSD I and PSD II for further details. Below we describe our calculations of the force multiplier for various conditions in the wind and our treatment of the optical depth effects on the radiation force.
In the disk plane at $`r=r_D`$, $`I(𝐫,\widehat{n})`$ is the local isotropic disk intensity:
$`\text{}I_D(r_D)`$ $`={\displaystyle \frac{3GM_{BH}\dot{M}_a}{8\pi ^2r_{}^3}}\{\text{}{\displaystyle \frac{r_{}^3}{r_D^3}}(1\left({\displaystyle \frac{r_{}}{r_D}}\right)^{1/2})`$ (6)
$`+{\displaystyle \frac{x}{3\pi }}(\mathrm{arcsin}{\displaystyle \frac{r_{}}{r_D}}{\displaystyle \frac{r_{}}{r_D}}(1\left({\displaystyle \frac{r_{}}{r_D}}\right)^2)^{1/2})\}.`$
We consider only the hot part of the disk, where the local disk temperature, $`T_D>`$ a few $`10^3`$ K. We then assume that all disk photons can contribute to the line force. The second term in the curly brackets corresponds to the contribution from the irradiation of a disk by a central object. For large radii and $`x1`$, the contribution from the irradiation is negligible compared to the intrinsic disk intensity (the first term in the curly brackets).
For radiation from the central object, the intensity may be written as:
$$I_{}=\frac{L_{}}{4\pi ^2r_{}^2}.$$
(7)
We account for the fact that the central radiation consists of the two distinct spectral components by using the parameter $`f_i`$, where $`i=\mathrm{X}`$ or $`\mathrm{UV}`$. We assume that the optical depth effects in the $`\theta `$ direction are negligible for the central radiation in the two bands. Note that the $`\theta `$ component of the radiation force due to the central object is negligible in our computational domain because the domain is far from the object ($`r_{}r_i`$). In the radial direction however, the column density can be high and the optical depth effects can be significant. We estimate the optical depth, $`\tau _\mathrm{i}`$ between the central source and a point in a wind from:
$$\tau _\mathrm{i}=_0^r\kappa _\mathrm{i}\rho 𝑑r,$$
(8)
where $`\kappa _\mathrm{i}`$ is the absorption coefficient representative for the $`i`$ band, and $`r`$ is the distance from the central source. We can treat the central object as a point source because our calculations are for $`r_{}/r1`$. Then we calculate the radial component of the central radiation force due to lines from
$$(F_{}^{rad,l})_r(𝐫)=f_{\mathrm{UV}}\mathrm{exp}(\tau _{\mathrm{UV}})_\mathrm{\Omega }_{}M(t)\left(n_r\frac{\sigma _eI_{}(𝐫,\widehat{n})d\mathrm{\Omega }}{c}\right),$$
(9)
where $`\mathrm{\Omega }_{}`$ is the central object solid angle. We also take into account the optical depth effects on the radial component of the electron-scattering force:
$$(F_{}^{rad,e})_r(𝐫)=\{f_{UV}\mathrm{exp}(\tau _{\mathrm{UV}})+f_\mathrm{X}\mathrm{exp}(\tau _X)\}_\mathrm{\Omega }_{}\left(n_r\frac{\sigma _eI_{}(𝐫,\widehat{n})d\mathrm{\Omega }}{c}\right).$$
(10)
The attenuation of the X-ray radiation is calculated using $`\kappa _\mathrm{X}=40\mathrm{g}^1\mathrm{cm}^2`$ for $`\xi 10^5`$ and $`\kappa _\mathrm{X}=0.4\mathrm{g}^1\mathrm{cm}^2`$ for $`\xi >10^5`$, while the attenuation of the UV radiation is calculated using $`\kappa _{\mathrm{UV}}=0.4\mathrm{g}^1\mathrm{cm}^2`$ for all $`\xi `$.
A disk wind may also be optically thick to the UV radiation emitted by the disk. In particular, the flux along the radial direction near the disk can be significantly reduced due to high column density. We approximate this effect by multiplying the radial component of the disk radiation force by the attenuation factor $`\mathrm{exp}(\tau _{\mathrm{UV}})`$.
To calculate the force multiplier, we adopt the CAK analytical expression modified by Owocki, Castor & Rybicki (1988, see also PSD I)
$$M(t)=kt^\alpha \left[\frac{(1+\tau _{max})^{(1\alpha )}1}{\tau _{max}^{(1\alpha )}}\right]$$
(11)
where $`k`$ is proportional to the total number of lines, $`\alpha `$ is the ratio of optically-thick to optically-thin lines, $`\tau _{max}=t\eta _{max}`$ and $`\eta _{max}`$ is a parameter determining the maximum value, $`M_{max}`$ achieved for the force multiplier. Equation 11 shows the following limiting behavior:
$`\underset{\tau _{max}\mathrm{}}{lim}M(t)`$ $`=`$ $`kt^\alpha `$ (12)
$`\underset{\tau _{max}0}{lim}M(t)`$ $`=`$ $`M_{max},`$ (13)
where $`M_{max}=k(1\alpha )\eta _{max}^\alpha `$. The maximum value of the force multiplier is a function of physical parameters of the wind and radiation field. Several studies have showed that $`M_{max}`$ is roughly a few thousand for gas ionized by a weak or moderate radiation field (e.g., CAK; Abbott 1982; Stevens & Kallman 1990; Gayley 1995). As the radiation field becomes stronger and the gas becomes more ionized the force multiplier decreases asymptotically to zero.
The line force, in particular the parameters of the force multiplier depend on the ionization of the wind and the spectral energy distribution (SED) of the radiation field. Self-consistent calculations of the line-force for given wind conditions and the radiation require detailed calculations of the wind photoionization structure (e.g., CAK; Abbott 1982; Puls et al. 2000). The total line force includes contributions from $`>10^5`$ lines from many species. Such calculations in connection with the 2-D, time-dependent hydrodynamic calculations are not feasible. We therefore start by adopting the analytical formulae for the force multiplier due to Stevens & Kallman (1991). They studied the effects of X-ray ionization on the the radiative force experienced by the stellar wind in a massive X-ray binary (MXRB). Wind conditions in MXRBs differ somewhat from those in QSOs. In particular, the X-ray radiation in MXRBs can be well represented by a 10 KeV bremsstrahlung spectrum while the spectrum in QSOs is better fitted by power laws with different spectral indices at different spectral bands (e.g., Zheng et al. 1997; Laor et al. 1997). Nevertheless we adopt the Stevens & Kallman results as a first order approximation because the Compton temperature and the ionization structure are similar (Kallman & McCray 1982).
Using the photoionization code XSTAR, Stevens & Kallman found that the line force due to the radiation from the primary can be parameterized in terms of the CAK force multiplier. Their results show that the line force decreases sharply with increasing photoionization parameter:
$$\xi =\frac{4\pi _\mathrm{X}}{n},$$
(14)
where $`_\mathrm{X}`$ is the local X-ray flux, $`n`$ is the number density of the gas ($`=\rho /(m_p\mu )`$ , where $`m_p`$ is the proton mass, and $`\mu `$ is the mean molecular weight). Stevens & Kallman also found some simple analytical fits to their results that allowed them to express $`k`$ and $`\eta _{max}`$ as functions of $`\xi `$:
$$k=0.03+0.385\mathrm{exp}(1.4\xi ^{0.6}),$$
(15)
and
$$\mathrm{log}_{10}\eta _{max}=\{\begin{array}{cc}6.9\mathrm{exp}(0.16\xi ^{0.4})\hfill & \mathrm{for}\mathrm{log}_{10}\xi 0.5\hfill \\ & \\ 9.1\mathrm{exp}(7.96\times 10^3\xi )\hfill & \mathrm{for}\mathrm{log}_{10}\xi >0.5\hfill \end{array}$$
(16)
The parameter $`\alpha =0.6`$ and does not change with $`\xi `$.
These expressions for the parameters of the force multiplier predict that $`M_{max}`$ increases gradually from $`2000`$ to $`5000`$ as $`\xi `$ increases from 0 to $`3`$ and then drops to $`0.1`$ at $`\xi =1000`$. The line force becomes negligible for $`\xi >100`$ because then $`M_{max}1`$.
To calculate the photoionization parameter, we first need to know the local X-ray flux $`_\mathrm{X}`$ (see eq. 14). We estimate $`_\mathrm{X}`$ assuming that the source of all X-rays is a point-like central object with the luminosity in the X-ray band, $`L_\mathrm{X}=f_\mathrm{X}L_{}`$. Making the above simplifications, we can express the local X-ray flux as
$$_\mathrm{X}=\mathrm{exp}(\tau _\mathrm{X})\frac{L_\mathrm{X}}{4\pi r^2}.$$
(17)
### 2.3 Radiation heating and cooling
We calculate the gas temperature assuming that the gas is optically thin to its own cooling radiation and that the abundances are cosmic (Withbroe 1971). The net cooling rate depends on the density, $`\rho `$, the temperature, $`T`$, the ionization parameter $`\xi `$, and the characteristic temperature of the X-ray radiation $`T_\mathrm{X}`$. It is then possible to fit analytical formulae to the heating and cooling rate obtained from detailed photoionization calculations for various $`T_X`$, $`T`$, and $`\xi `$. For example, Blondin (1994) found that for a 10 keV bremsstrahlung spectrum, his fits typical agree with a detailed calculation (Blondin et al. 1990) to within 25%. Using Blondin’s result we can express the net cooling rate $``$ in equation 3 by
$$\rho =n^2(G_{Compton}+G_\mathrm{X}L_{b,l})erg\mathrm{cm}^3\mathrm{s}^1,$$
(18)
where $`G_{Compton}`$ is the rate of Compton heating/cooling,
$$G_{Compton}=8.9\times 10^{36}\xi (T_\mathrm{X}T)\mathrm{erg}\mathrm{cm}^3\mathrm{s}^1,$$
(19)
$`G_\mathrm{X}`$ is the net rate of X-ray photoionization heating–recombination cooling
$$G_\mathrm{X}=1.5\times 10^{21}\xi ^{1/4}T^{1/2}(1T/T_\mathrm{X})\mathrm{erg}\mathrm{cm}^3\mathrm{s}^1$$
(20)
and $`L_{b,l}`$ is the rate of bremsstrahlung and line cooling
$$L_{b,l}=3.3\times 10^{27}T^{1/2}+[1.7\times 10^{18}\mathrm{exp}(1.3\times 10^5/T)\xi ^1T^{1/2}+10^{24}]\delta \mathrm{erg}\mathrm{cm}^3\mathrm{s}^1.$$
(21)
The parameter $`\delta `$ is introduced to control line cooling, $`\delta =1`$ represents optically thin cooling and $`\delta <1`$ represents the case when lines become optically thick and cooling is reduced.
As we mentioned in Section 2.1, we use the same initial and boundary conditions as in PSD I. However we here also calculate the evolution of the internal energy. We specify the boundary conditions and the initial conditions for $`e`$ as follows. For the temperature at the base of the wind, in the first grid zone above the $`\theta =90^o`$ plane, we adopt the solution for the steady state disk. Adopting the assumption that the local disk intensity is a black body and using eq. 6, the disk temperature is:
$$T_D(r_D)=(\pi I_D(r_D)/\sigma )^{1/4}.$$
(22)
Then we calculate the internal energy from
$$e_o(r_D)=\frac{\rho _okT(r_D)}{\mu m_p(\gamma 1)},$$
(23)
where $`\rho _o`$ is the density in the first grid zone above the $`\theta =90^o`$ plane. We consider models with $`\mu =1`$. We initialize the internal energy at other locations by assuming that the gas temperature at a given radius is the same as at the disk temperature (i.e., $`T(r,\theta )=T_D(r\mathrm{sin}\theta )`$ for all $`\theta `$) and using initial value of the density (see PSD I). Consistent with the boundary conditions for other gas quantities (see Sect. 2.1) we apply the constraint that in the first zone above the $`\theta =90^o`$ plane the internal energy $`e=e_o`$ at all times. We find only the initial, transient evolution of the wind is affected by this choice of initial temperature conditions.
Our treatment of radiative heating and cooling is most likely valid in a low density and high photoionization parameter regime. For a high density and low photoionization parameter regime our treatment will likely underestimate the temperature because we oversimplify line cooling by not properly taking into account optical depth effects. In dense regions, our scheme therefore may yield a very low gas temperature (e.g., $`T1000`$ K). In such cases we set the lower limit for the temperature assuming that the gas is in local thermodynamical equilibrium, i.e., the gas temperature equals the local disk radiation temperature. Specifically, if at the ($`r,\theta `$) wind point the temperature from eq. 3 is less than $`T_D(rsin\theta )`$ then we replace it with $`T_D(rsin\theta )`$.
## 3 Results
We specify our models by several parameters. In all our calculations we assume the mass of the non-rotating black hole, $`M_{BH}=10^8\mathrm{M}_{}`$ and the rest mass conversion efficiency, $`\eta =0.06`$ typical for AGN. To determine the radiation field from the disk, we assume the mass accretion rate $`\dot{M}_a=1.8`$ M yr<sup>-1</sup>. These system parameters yield the disk Eddington number, $`\mathrm{\Gamma }_D=0.5`$, the disk inner radius, $`r_{}=8.8\times 10^{13}`$ cm and the orbital period at the inner edge of the disk, $`\tau =\sqrt{\frac{r_{}^3}{GM_{BH}}}=7.22\times 10^3`$ sec. We set the offset of the computational domain, $`z_o=H_D3\mathrm{\Gamma }_D`$ that corresponds to the half width of the radiation pressure dominated disk at large radii (e.g., Shakura & Sunyaev 1973). The radiation field from the central engine is specified by the additional parameters: $`x=1`$, $`f_{\mathrm{UV}}=0.5`$, $`f_\mathrm{X}=0.5`$. As we discussed above the SED of the ionizing radiation is not well known, our choice of values for $`f_{\mathrm{UV}}`$ and $`f_\mathrm{X}`$ is guided by the results from Zheng et al. (1997) and Laor et al. (1997). To calculate the line force we adopt the force multiplier parameter $`\alpha =0.6`$. We calculate the remaining parameters of the multiplier, i.e., $`k`$ and $`\eta _{max}`$ as functions of the photoionization parameter, $`\xi `$ using eqs 15 and 16. The force multiplier depends only formally on the thermal speed, $`v_{th}`$ which we set to 20 $`\mathrm{km}\mathrm{s}^1`$, i.e., the thermal speed of a hydrogen atom at the temperature of 25000 K (Stevens & Kallman 1990). To calculate the gas temperature, we assume the temperature of the X-ray radiation, $`T_\mathrm{X}=10`$ KeV and the line cooling parameter $`\delta =1`$.
We adopted a generalized CAK method to calculate a disk wind driven by the line force (PSD II). The CAK method has been developed for OB stars with photospheres dominated by gas pressure (the inner atmosphere is in hydrostatic equilibrium), and they radiate most of their energy in the UV band. We chose the radial extent of the computational domain (i.e., $`100r_{}r1000r_{}`$) keeping in mind that the disk atmosphere should satisfy these two conditions, however some compromises had to be done. For example, the disk temperature at $`100r_{}=8.8\times 10^{15}`$ cm is 8200 K while at $`1000r_{}=8.8\times 10^{16}\mathrm{cm}`$ is 1500 K using our disk parameters. We realize that according to the $`\alpha `$-disk model the disk emits mostly in the UV band down to the radius of $`10r_{}`$ where its effective temperature is $`50000`$ K. However the disk at this small radius is likely radiation dominated (e.g., Shakura & Sunyaev 1973; Svensson & Zdziarski 1994). On the other hand, the disk atmosphere at radius $`1000r`$ and the effective temperature of $`1500K`$, radiates most of its energy in the infrared band. Nevertheless we chose a large value for the outer edge of the computational domain to make sure that the domain includes most of the acceleration zone of the disk wind launched from the inner disk and the wind velocity at the outer edge well represents the wind terminal velocity.
Figure 2 shows the instantaneous density, temperature and photoionization parameter distributions and the poloidal velocity field of the model. After $`12.6`$ years ($`5.5\times 10^4\tau `$), the disk material fills the grid for $`\theta >70^o`$ and remains in that region for the rest of the run, i.e., over next $`47.7`$ years. Figure 2 shows results after $`14.6`$ years. Although the flow is time-dependent the gross properties of the flow (e.g., the mass loss rate and the radial velocity at the outer boundary), settle down to steady time-averages over timescales on the order of $`3`$ years. Our calculation follows (i) a hot, low density flow in the polar region (ii) a dense, warm and fast equatorial outflow from the disk, (iii) a transitional zone in which the disk outflow is hot and struggles to escape the system.
In the polar region, the density is very small and close to the lower limit that we set on the grid, i.e., $`\rho _{min}=10^{20}\mathrm{gcm}^3`$. The line force is negligible because the matter is highly ionized as indicated by a very large photoionization parameter ($`10^7`$). The gas temperature is close to the temperature of the X-ray radiation, again indicative of highly ionized gas. The matter in the polar region is pulled by the gravity from the outer boundary and it is an artifact of the boundary conditions. Overall this region of the very low density is not relevant to our analysis as it has no effect on the much denser disk flow.
The equatorial region is distinctly different. In the inner part of the disk (i.e., for $`rr_i`$), the density at the wind base is high, $`10^{13}\mathrm{g}\mathrm{cm}^3`$. Thus the photoionization parameter is low despite the strong central radiation. However as the flow from the inner part of the disk is accelerated by the line force its density decreases and the gas temperature and the photoionization parameter increase. Subsequently the gas becomes fully ionized and loses all of driving lines before it reaches the escape velocity and therefore falls on the central object/inner disk. Although this gas does not produce a wind, its primary effect is to shield the gas at larger radii. Thus the wind consists of gas accelerated by the line force at larger radii, in fact driving of the disk wind extends over all radii at which the intrinsic disk radiation is large enough to launch gas.
The poloidal velocity (Fig. 2c) shows that the gas streamlines are perpendicular to the disk over some height that increases with radius. The streamlines then bend away from the central object and converge. The region where the flow is moving almost radially outward is associated with a high-velocity, high density stream. This fast stream contributes $``$ 100% to the total mass loss rate, $`\dot{M}_W=0.5\mathrm{M}_{}\mathrm{yr}^1`$. We note that the mass loss rate can increase by a factor of a few when a knot is crossing the outer boundary.
The fast stream is variable. In the upper envelope of the disk wind there is large velocity shear between the higher density fast stream moving outward and the lower density fast gas moving inward. Our simulations show that this shear gives rise to Kelvin-Helmholtz instabilities. The instabilities generate knots that propagate along the fast stream at $`10000`$ km $`\mathrm{s}^1`$. Figure 2a shows an example of such a knot at $`r750r_{}`$ and $`z200r_{}`$, in the figure coordinates. The density contrast between the knot and the fast stream is $`2`$ orders of magnitude. The knots generator is episodic and is inherent to the fast stream. We observed the generation of 19 knots over $`60.4`$ years. In other words, a knot is produced every $`3`$ years.
Figure 3 presents a sequence of density maps showing time evolution of the outflow from Figure 2 after 13.3, 14.6 and 16.47 years, left, middle and righ panel respectively. Note that the middle panel from Figure 3 is the same as the top left panel from Figure 2. Figure 3 well illustrates variability in the outflow, in particular, generation of a knot in the fast stream at $`r450r_{}`$ and $`z100r_{}`$ (left panel), the well-formed knot at $`r750r_{}`$ and $`z200r_{}`$ (middle panel), and the time when the knot left the gird at $`r950r_{}`$ and $`z300r_{}`$, and the fast stream is fairly smooth in between episodes with knots (right panel). The time dependence in the region close to the disk, bounded by the fast stream resembles outward propagation of a wave at the begin of evolution but becomes more complex with time as elements of the flow move both upwards and downwards.
Figure 4 presents the run of the density, radial velocity, photoionization parameter and column density as a function of the polar angle, $`\theta `$ at the outer boundary, $`r_o=8.8\times 10^{16}`$ cm from Figure 2. The column density is given by:
$$N_H(\theta )=_{r_i}^{r_o}\frac{\rho (r,\theta )}{\mu m_p}𝑑r.$$
(24)
The gas density is a very strong function of angle for $`\theta `$ between $`90^o`$ and $`65^o`$. The density drops by $`8`$ orders of magnitude between $`\theta =90^o`$ and $`\theta 89^o`$, as might be expected of a density profile determined by hydrostatic equilibrium. For $`70^o<\theta <89^o`$, the wind domain, $`\rho `$ varies between $`10^{19}`$ and $`10^{17}\mathrm{g}\mathrm{cm}^3`$. For $`\theta <70^o`$, density decreases gradually to so low a value that it becomes necessary to replace it by the numerical lower limit $`\rho _{min}`$. The radial velocity at $`1000r_{}=8.8\times 10^{16}`$ cm has a broad peak for $`72^o<\theta <82^o`$ with the maximum of 15000 km $`\mathrm{s}^1`$ at $`\theta 75`$. On both side of the peak the velocity is close to zero.
The photoionization parameter is very high, $`10^7`$, for $`\theta <70^o`$ because of the very low density. However it drops by 15 orders of magnitude between $`70^o<\theta <75^o`$ and stays at a very low level for $`\theta >75^o`$. The column density changes less dramatically with angle. Between the pole and $`\theta <67^o`$, $`\mathrm{N}_\mathrm{H}`$ is less than $`10^{22}\mathrm{cm}^2`$. For $`\theta >67^o`$, $`\mathrm{N}_\mathrm{H}`$ increases gradually with $`\theta `$, it reaches value of $`10^{24}\mathrm{cm}^2`$ at $`\theta 80^o`$. For $`80^o<\theta <89^o`$, $`\mathrm{N}_\mathrm{H}`$ varies between $`10^{24}\mathrm{cm}^2`$ and $`10^{26}\mathrm{cm}^2`$.
Our model shows that (i) the intrinsic disk radiation can launch a wind and (ii) the central object radiation can accelerate the disk wind to very high velocities. We have checked how these main results are sensitive to the model parameters. For a fixed disk atmosphere and central radiation source, the most important parameter of our model is the optical depth for the X-rays from the central object. Therefore we have run several tests with various X-ray opacities for $`\xi <10^5`$: $`\kappa _X=0.4\mathrm{g}^1\mathrm{cm}^2`$ and $`\kappa _X=4\mathrm{g}^1\mathrm{cm}^2`$. The former case is most conservative because it corresponds to the case when the X-ray optical depth is reduced to the Thomson optical depth. We found that in both test runs the X-ray radiation is significantly attenuated so the line force can launch a disk outflow. However the flow velocity never exceeds the escape velocity because the X-rays fully ionize the gas close to the disk and produces a hot corona with complex velocity field but which does not escape the system. We conclude then that the disk atmosphere can ’shield’ itself at least to the extent that the local disk radiation can launch gas off the disk photosphere.
The two test runs also illustrate how robust is our second result: the central radiation accelerates the wind to very high velocities. For this to happen the column density, $`N_\mathrm{H}`$ must be large enough to reduce the X-ray radiation but too not large to reduce the UV flux in the radial direction. In other words, this requires $`\tau _\mathrm{X}>1`$ and at the same time, $`\tau _{UV}1`$.
The test run with $`\kappa _\mathrm{X}=0.4\mathrm{g}^1\mathrm{cm}^2`$ corresponds to the situation where $`\tau _\mathrm{X}=\tau _{\mathrm{UV}}`$. Thus the disk gas that is sufficiently shielded from the X-rays is also shielded from the UV photons from the central object. Such gas can only be lifted from the disk surface by the disk UV radiation but fails to gain momentum in the radial direction.
The fact that the disk wind can be launched without external shielding material implies that our solution can depend on the location of the inner computational radius, $`r_i`$. We have run several tests with $`r_i=50r_{}`$ and $`r_i=200r_{}`$ to check this (the remaining model parameters were as in the models shown in Figure 1 and 2). We found that the location of the inner edge, $`r_i`$ affects the properties of the wind but not the fact the wind is produced. For example, the wind opening angle, $`\omega `$, is $`25^o`$, $`15^o`$ and $`8^o`$ for $`r_i=50,r_{},100r_{}`$ and $`200r_{}`$, respectively. The maximum radial velocity at the outer boundary decreases from 20000 $`\mathrm{km}\mathrm{s}^1`$ through, 15000 $`\mathrm{km}\mathrm{s}^1`$ to 5000 $`\mathrm{km}\mathrm{s}^1`$ for above sequence of decreasing $`r_i`$. The mass loss rate also decreases with $`r_i`$ from 0.6 $`\mathrm{M}_{}\mathrm{yr}^1`$ through 0.5 $`\mathrm{M}_{}\mathrm{yr}^1`$ to 0.2 $`\mathrm{M}_{}\mathrm{yr}^1`$.
## 4 Discussion
Our calculations show that the UV emitting accretion disk can launch a wind that will shield itself from the strong ionizing radiation emitted from the central object. The self-shielding is quite robust because the external radiation does not penetrate the disk down to the region where the line force becomes important and pushes matter away from the photosphere. We illustrated the robustness of the self-shielding by changing relevant model parameters. We also note that a similar effect was observed by Stevens & Kallman (1990) and Stevens (1991). They studied the effects of X-ray ionization on the radiative force experienced by the stellar wind in a MXRB. They calculated the detailed photoionization structure of the line-driven wind for various X-ray luminosities without optical depth effects (Stevens & Kallman 1990) and with optical depth effects (Stevens 1991). Additionally, they considered irradiation at various directions including the normal to the mass losing photosphere. In none of the models the irradiation penetrated below the so-called wind critical point where, in the CAK type models, the wind mass loss rate is determined. They found, as we did, that the irradiation can severely decrease the wind velocity in the supersonic portion of the flow.
We model here a line-driven wind from a disk that is flat, Keplerian, geometrically-thin and optically-thick. For the temperature at the base of the wind, in the first grid zone above the $`\theta =90^o`$ plane, we adopt the solution for the steady state $`\alpha `$ disk. We calculate the local disk intensity as if it emits as a black body. In all our calculations, we treat the regions for $`\theta 90^o`$ as if they are gas pressure dominated. We find that our results depend on the inner radius of the computational domain. This dependence is likely an artifact of one or more of our assumptions. For example, for very high luminosities the inner part of the disk is dominated by the radiation pressure. For the parameters adopted here, namely $`\mathrm{\Gamma }=0.5`$, the inner radius of a gas pressure dominated disk is $`10^{16}`$ cm (e.g., Svensson & Zdziarski 1994). The structure of the radiation dominated disks is not well known. We expect, however, that incorporating the radiation dominated part of the disk into our calculation will change the condition of the base of the wind, for example, the gas density and the UV flux might be reduced. These changes might affect the solution for the wind, in particular, the wind mass flux. We plan to explore these issues in a future paper.
To calculate the disk surface temperature and intensity, we took into account the irradiation of the disk by the central object. However the contribution from the irradiation is negligible compared to the disk intrinsic temperature because we considered a flat disk at large radii. Additionally, the high column density of the disk wind, that we found in our models, implies that the irradiating flux will be significantly attenuated by the wind before it reaches the disk surface. We anticipate that the last effect will significantly reduce disk irradiation regardless of the shape of the disk surface – flat or flaring. Detailed NLTE photoionization calculations are required to determine what fraction of the central radiation will reach the surface of a mass losing disk. A complete treatment of disk irradiation is difficult also because it is not obvious a priori whether the radiation incident on the disk is completely thermalized and re-radiated isotropically or it is scattered off the disk atmosphere or both.
Our calculations show that a line-driven disk wind model offers a promise of explaining outflows in AGNs. However they also illustrate some problems with this model. For example, as dKB pointed out in discussing MCGV’s model, a very small radius at which the disk wind is launched also implies a very small size to BELR because BELR lies inside or is cospatial with BALR. The small scales in turn imply short crossing time of the BALR, of order of a month or so, and it is difficult to understand that the highly complex kinematical structures that BALs often exhibit do not appear to vary on timescales of 10 years (Barlow 1994).
Our axisymmetric 2D models show that the disk wind is unsteady and generates dense knots every 3 years. These knots will correspond to rings in 3D. However in fully 3D calculations the rings may break down to spirals or clouds or both. The breaking down of the rings will change the density contrast between them and the rest of the wind. Thus it is not clear if we would be able to see any spectral signature of density fluctuations in the wind. Detail calculations of line profiles are required to check this point. Additionally, detailed photoionization calculations are required to check if the full range of ions observed to show BAL profiles can be explained: BALs from ions with ionization potentials as low as of O III or lower and as high as of O VI.
To produce a fast wind the ratio between $`\tau _\mathrm{X}`$ to $`\tau _{\mathrm{UV}}`$ is very important. The low ratio gives a slow disk wind whereas the high ratio gives a fast wind. This result is consistent with the observed anti-correlation for QSOs between the relative strength of the soft X-ray flux and the CIV absorption equivalent width (e.g., Brandt, Laor & Wills 2000). However there is an upper limit for the X-ray attenuation, or the column density between the X-ray source and wind, namely for $`N_H>10^{24}\mathrm{cm}^2`$ ($`\tau _\mathrm{X}>>1`$) the gas is well shielded from the X-rays but at the same time it is also shielded from any other radiation from the central engine including the UV radiation. In this case, the line force from the central engine could be so much reduced that it may not accelerate the gas to high radial velocities and no strong wind will be produced.
Some of the emission-line properties of QSOs could be explained as a correlation between luminosity and the slope of the ionizing spectrum, i.e., lower luminosity objects have harder spectra (e.g., Boroson & Green 1992). Using our parameterization of the AGN radiation, this QSO’s property can be represented by increasing $`f_\mathrm{X}`$ in expense of $`f_{\mathrm{UV}}`$ when we reduce the luminosity. Such a choice of model parameters will increase the strength of X-ray ionizing radiation in comparison to the UV driving radiation. Consequently a disk wind should be weaker and slower as some of our test runs indicate.
Boroson & Green (1992) argue that the dominant source of variation in the observed properties of low redshift QSOs is not driven by external orientation but rather by the fraction of the Eddington luminosity at which the object is emitting (our parameter $`\mathrm{\Gamma }_D`$). We plan to examine the parameter space of our models to define the major trends in disk wind behavior that will help us to explain the observational trends found in various AGNs.
## 5 Conclusions
We have studied radiation driven winds from luminous accretion disks using numerical methods to solve the two-dimensional, time-dependent equations of hydrodynamics. In so doing we have accounted for the radiation force due to spectral lines using a generalized multidimensional formulation of the Sobolev approximation. Additionally we have taken into account the effects of the strong central radiation on the wind photoionization structure and thermodynamics.
We find that the local disk radiation can launch a wind from the disk despite strong ionizing radiation from the central object. The central radiation may overionize the supersonic portion of the flow and severely reduce the wind velocity. To produce a fast disk wind the wind X-ray opacity must be higher than the UV opacity by $`>`$ 2 orders of magnitude.
Our calculations of a wind from a disk accreting onto a $`10^8\mathrm{M}_{}`$ black hole at the rate of 1.8 $`\mathrm{M}_{}\mathrm{yr}^1`$ show that the radiation force from an accretion disk can launch a self-shielding wind from a radius of $`<10^{16}`$ cm while the strong UV radiation from the central object can radially accelerate the disk wind to velocities $`15000`$ km $`\mathrm{s}^1`$, for the X-ray opacity of 40 $`\mathrm{g}^1\mathrm{cm}^2`$. The disk wind domain is intrinsically unsteady and its covering factor is $`0.2`$. The wind mass loss is 0.5 $`\mathrm{M}_{}\mathrm{yr}^1`$ which is a significant fraction of the mass accretion rate. The strong X-ray radiation from a central object completely ionizes the polar region and only a thin layer above the upper envelope of the disk wind. The disk wind immediately below the upper envelope can be characterized as a fast, high-density stream which is reminiscent of the PSD disk wind solution dominated by the driving radiation from a bright central object. The column density of the fast stream is between $`10^{22}\mathrm{cm}^2`$ and $`10^{24}\mathrm{cm}^2`$ so the stream is optically thin to the UV radiation and this is precisely why it can be accelerated to high velocities. On the other hand, the part of the wind bounded by the fast stream, closer to the disk, is slow and is reminiscent of the PSD solution dominated by the driving radiation from the disk. The column density of the slow wind is $`>10^{24}\mathrm{cm}^2`$ so it is completely shielded from the central radiation, both the X-rays and the UV photons.
ACKNOWLEDGEMENTS: We would like to thank J.E. Drew for useful discussions. The work presented in this paper was initiated while DP was a PPARC Research Associate at Imperial College, London and it was performed while DP held a National Research Council Research Associateship at NASA/GSFC. Computations were supported by NASA grant NRA-97-12-055-154.
## REFERENCES
Abbott, D.C. 1982, ApJ, 259, 282
Arav, N. 1996, ApJ, 465, 617
Arav, N., Korista, T.K., Barlow, T. A., & Begelman, M. C. 1995, Nature, 376, 576
Arav N., & Begelman, M.C. 1994, ApJ, 434, 479
Arav N., Li, Z.Y., & Begelman M.C. 1994, ApJ, 432, 62
Arav N., Shlosman I., & Weymann. R. ed. 1997, in ASP Conf. Ser. 128, Mass Ejection from Active Galactic Nuclei, (San Francisco: ASP)
Barlow, T.A. 1994, PASP, 106, 548
Balbus, S.A., & Hawley, J.F. 1998, Rev. Mod. Phys., 70, 1
Begelman, M.C., de Kool, M., & Sikora, M. 1991, ApJ, 382, 416
Blandford, R.D., Netzer H., Woltjer L., Courvoisier T., & Mayor M. 1990, Active Galactic Nuclei, (Berlin: Springer)
Blandford, R.D., Payne, D.G. 1982, MNRAS, 199, 883
Blondin, J.M. 1994, ApJ, 435, 756
Blondin, J.M., Kallman, T.R., Fryxell, B.A., & Taam, R.E. 1990, ApJ, 356, 591
Boroson, T.A., & Green, R.F. 1994, ApJS, 80, 109
Bottorff, M., Korista, K.T., Shlosman, I., & Blandford, D.R. 1997, ApJ, 479, 200
Brandt, W.N., Laor, A., & Wills, B. J. 2000, ApJ, 528, 637
Castor, J.I., Abbott, D.C., Klein, R.I. 1975, ApJ, 195, 157 (CAK)
Crenshaw, D.M. 1997, in ASP Conf. Ser. 128, Mass Ejection from Active Galactic Nuclei, ed. N. Arav, I. Shlosman, & R. Weymann (San Francisco: ASP), p. 121
de Kool, M., & Begelman, M.C. 1995, APJ, 455, 448 (dKB)
Drew, J.E., & Boksenberg, A. 1984, MNRAS, 211, 813
Emmering, R.T., Blandford, R.D., & Shlosman, I. 1992, ApJ, 385, 460
Foltz, C.B., Weymann, R.J., Morris, S.L., & Turnshek, D.A. 1987, ApJ, 317, 450
Gayley, K.G. 1995, ApJ, 454, 410
Hamann, F., Barlow, T.A., Cohen, R.D., Junkkarinen, V., & Burbidge, E.M. 1997, in ASP Conf. Ser. 128, Mass Ejection from Active Galactic Nuclei, ed. N. Arav, I. Shlosman, & R. Weymann (San Francisco: ASP), p. 19
Kaastra, J.S., Mewe, R., Liedahl, D.A., Komossa, S. & Brinkman, A.C. 2000, A&A, 354, L83
Kallman, T.R., & McCray, R. 1982, ApJS, 50 263
K$`\ddot{\mathrm{o}}`$nigl, A., & Kartje, J. F. 1994, ApJ, 434, 446
Korista, K.T., Voit, G.M. Morris, S.L., & Weymann R.J. 1993, ApJS, 88, 357
Krolik, J.H. 1999, Active galactic nuclei: from the central black hole to the galactic environment, Princeton, N.J.: Princeton University Press
Laor, A., Fiore, F., Elvis, E., & Wilkes B.J., & McDowell J.C. 1997, ApJ, 477, 93
Murray, N., Chiang, J., Grossman, S.A., & Voit, G.M. 1995, ApJ, 451, 498 (MCVG)
Osterbrock, D.E. 1989, Astrophysics of Gaseous Nebulae and Active Galactic Nuclei (Mill Valley: University Science Books), chap. 11
Owocki, S.P., Castor J.I., & Rybicki, G.B. 1988, ApJ, 335, 914
Pereyra, N.A., Kallman, T.R., & Blondin, J.M. 1997, ApJ, 477, 368
Proga, D. 1999, MNRAS, 304, 938
Proga, D., Stone J.M., & Drew J.E. 1998, MNRAS, 295, 595 (PSD I)
Proga, D., Stone J.M., & Drew J.E. 1999, MNRAS, 310, 476 (PSD II)
Puls, J., Springmann, U., & Lennon, M. 2000, A&AS 141, 23
Scoville, N., & Norman, C. 1995, ApJ, 451, 510
Shakura N.I., & Sunyaev R.A. 1973 A&A, 24, 337
Shlosman, I., Vitello, P., & Shaviv, G. 1985, ApJ, 294, 96
Stevens, I.R. 1991, ApJ, 379, 310
Stevens, I.R., & Kallman T.R. 1990, ApJ, 365, 321
Stone, J.M., Norman, M.L. 1992, ApJS, 80, 753
Svensson, R., & Zdziarski, A.A. 1994, ApJ, 436, 599
Voit, G.M., Weymann, R.J., & Korista, K.T. 1993, ApJ, 413, 95
Weymann, R.J., Morris, S.L., Foltz, C.B., & Hewett, P.C. 1991, ApJ, 373, 23
Weymann, R.J., Scott, J.S., Schiano, A.V.R., & Christiansen, W.A. 1982, ApJ, 262, 497
Withbroe, G. 1971, in The Menzel Symposium on Solar Physics, Atomic Spectra, and Gaseous Nebulae, ed. K.B. Gebbie (NBS Spec. Pub. 353; Washington, D.C.: NBS), p. 127
Zheng, W., Kriss, G.A., Telfer, R.C., Grimes, J.P., & Davidsen, A.F. 1997, ApJ, 475, 469
|
warning/0005/astro-ph0005520.html
|
ar5iv
|
text
|
# The Abundance of Interstellar BoronBased on observations made with the NASA/ESA Hubble Space Telescope, obtained from the data archive at the Space Telescope Science Institute. STScI is operated bythe Association of Universities for Research in Astronomy, Inc. under the NASA contract NAS 5-26555.
## 1 Introduction
Production of the light element boron in standard big bang nucleosynthesis models is negligible (see, e.g., Pagel 1997). The chemical evolution of boron is determined by its production through cosmic ray spallation of interstellar material (Reeves, Fowler, & Hoyle 1970; Meneguzzi, Audouze, & Reeves 1971) and/or neutrino-induced spallation in type II supernovae (Woosley et al. 1990) and its destruction via astration. The evolution of the cosmic boron abundance therefore reflects the cosmic ray flux and/or supernova rate history in the Milky Way. Measurements of stellar boron abundances have revealed a trend of increasing B/H over time (as expected), from $`\mathrm{B}/\mathrm{H}2\times 10^{12}`$ in very metal-poor halo stars (Edvardsson et al. 1994; Duncan et al. 1997) to the much higher Orion association ($`\mathrm{B}/\mathrm{H}[36]\times 10^{10}`$; Cunha et al. 1997) and solar system (meteoritic) abundances ($`\mathrm{B}/\mathrm{H}=[7.6\pm 0.7]\times 10^{10}`$; Anders & Grevesse 1989). Because there is evidence that the solar system may be enhanced in metals such as oxygen relative to the local interstellar medium (ISM; e.g., Meyer et al. 1998), young B stars (e.g., Gies & Lambert 1992; Kilian-Montenbruck et al. 1994) and H II regions (e.g., Peimbert, Torres-Peimbert, & Dufour 1993), it is unclear that the solar system B/H is the best fiducial point for the present-day boron abundance in studies of Galactic chemical evolution. The boron abundance of the present-day ISM may be more appropriate for such studies. The differences between the solar system and the ISM could yield important information on the recent chemical evolution of the solar neighborhood.
Here we present new high-resolution observations of interstellar B II $`\lambda 1362.461`$ and O I $`\lambda 1355.598`$ absorption towards the stars HD 104705, HD 121968, HD 177989, HD 218915, and HD 303308 taken with STIS on-board the Hubble Space Telescope (HST). We also present analyses of the B/O abundances along the $`\alpha `$ Scorpii B and $`\zeta `$ Ophiuchi sightlines using archival GHRS datasets. We have more than doubled the measurements of interstellar boron, and the available data now probe a wide range of diffuse ISM environments. After discussing our observations in §2, we will show in §3 that the present-day interstellar B/H ratio seems to be dominated by the incorporation of boron into dust grains. Thus the interstellar gas-phase boron abundance provides only a lower limit to the total (gas+dust) interstellar boron abundance.
## 2 Observations and Reductions
The STIS data used here were obtained as part of our STIS GO program (#7270), with the exception of the HD 303308 data, which were obtained from the STIS archive. In all cases the observations employed the far-ultraviolet MAMA detector and the E140H grating with the light from the star passing through the $`0\stackrel{}{\mathrm{.}}2\times 0\stackrel{}{\mathrm{.}}09`$ aperture. The resolution of these data is $`\mathrm{\Delta }v2.7`$ km s<sup>-1</sup> (FWHM). The data were extracted and the backgrounds estimated as described by Howk & Sembach (2000). For spectral regions covered in multiple orders (or observations), we have coadded the flux-calibrated data and weighted the contribution of each individual spectrum by the inverse-square of its error vector.
The archival, post-COSTAR GHRS data used in this work were reduced as discussed by Howk, Savage, & Fabian (1999). The $`\alpha `$ Sco B data were obtained through the large science aperture, while the $`\zeta `$ Oph data were obtained through the small science aperture. Both datasets employed the Ech-A grating, yielding a resolution of $`\mathrm{\Delta }v3.5`$ km s<sup>-1</sup> (FWHM).
Figure 1 shows the STIS spectra and Table 1 gives the measured equivalent widths, $`W_\lambda `$, for the new STIS and GHRS measurements. The continua were estimated using low-order Legendre polynomial fits to regions free of interstellar lines, and our $`1\sigma `$ error estimates include continuum placement uncertainties and the effects of 2% zero-point uncertainties for both spectrographs (Sembach & Savage 1992).
Federman et al. (1993) and Lambert et al. (1998) have studied the B II absorption along the $`\zeta `$ Oph sightline using the GHRS with discrepant results. Federman et al. (1993) derive $`W_\lambda =1.78\pm 0.29`$ mÅ using the G160M grating, while Lambert et al. (1998) derive $`W_\lambda =0.6\pm 0.2`$ mÅ using the Ech-A grating. Given the differences in these studies, we have reanalyzed both GHRS datasets. We derive an equivalent width using the Ech-A of $`W_\lambda =0.79\pm 0.17`$ mÅ, within $`1\sigma `$ of the Lambert et al. result. We do not find clear evidence for B II absorption in the small science aperture G160M grating data to a $`3\sigma `$ limiting equivalent width of $`W_\lambda <1.3`$ mÅ, consistent with our Ech-A results. However, the continuum placement uncertainties are large in this region of the spectrum when using the intermediate-resolution G160M grating. Alternative continuum placements could push this limiting equivalent width as high as 2.1 mÅ. We believe that the Ech-A results are the most reliable for this sightline and adopt $`W_\lambda =0.79\pm 0.17`$ mÅ for the B II absorption towards $`\zeta `$ Oph.
The sightline towards $`\alpha `$ Sco B also deserves comment. The sightline to this star passes through the stellar wind of the M1.5 Ib primary (Antares), which lies $`2\stackrel{}{\mathrm{.}}9`$ from the $`\alpha `$ Sco B sightline (van der Hucht, Bernat, & Kondo 1980; Bernat 1982; Cardelli 1984). It is therefore possible that the absorption lines in the spectra of $`\alpha `$ Sco B probe the wind of the primary star. We do not believe that the O I and B II absorption seen in the archival GHRS spectra are caused by the stellar wind. These species show two blends of material centered at $`v_{\mathrm{LSR}}=4.1`$ and +3.5 km s<sup>-1</sup>. Material associated with the wind is centered near $`v_{\mathrm{LSR}}=18`$ km s<sup>-1</sup> (van der Hucht et al. 1980). The archival GHRS dataset contains good observations of lines that trace only the outflowing stellar wind, notably Ti II, Ti II, Ti II<sup>∗∗</sup>, and S I<sup>∗∗</sup>. These lines are clearly shifted with respect to the O I and B II absorption. They are centered near $`v_{\mathrm{LSR}}=18`$ km s<sup>-1</sup>, often with wings extending towards more negative velocities. There is no evidence for wind material at the velocities of the O I and B II absorption. Therefore, we believe the absorption line measurements presented in Table 1 trace the ISM in this direction rather than the stellar wind of Antares.<sup>1</sup><sup>1</sup>1We also note that the column density of O I derived below is comparable to that of other stars in this region of sky with similar distances (Meyer et al. 1998).
## 3 Results and Discussion
Table 2 gives the derived column densities of O I and B II for the stars listed in Table 1. Also given are the total hydrogen column densities and the normalized relative gas-phase abundances \[B/O\]<sup>2</sup><sup>2</sup>2We define $`[\mathrm{B}/\mathrm{O}]\mathrm{log}N(\text{B II})/N(\text{O I})\mathrm{log}(\mathrm{B}/\mathrm{O})_{}`$ and assume a meteoritic abundance $`\mathrm{log}(\mathrm{B}/\mathrm{O})_{}=5.99`$ (Anders & Grevesse 1989). for these sightlines. We also compile all measurements of interstellar B II using HST data from the literature (with references given in the table).
The new O I and B II column densities were derived by integrating the apparent optical depth profiles (Savage & Sembach 1991) of each line. In a few cases we have deemed it necessary to apply moderate saturation corrections to the O I column densities. We have tested for saturation problems and derived the necessary corrections by applying a single-component curve of growth to the measured O I equivalent widths (Table 1), adopting $`b`$-values derived from a curve-of-growth fit to several C I lines for each sightline. The O I absorption is dominated by narrow components that are also strong in species that trace dense clouds such as C I, S I, Cl I, and CO. While C I and O I need not be coexistent, the emperical association of C I and O I absorption, particularly in the components where the saturation is likely to be greatest, gives us confidence that C I saturation effects provide a suitable means for understanding the O I saturation along these sightlines. The O I apparent column densities we obtained for four of the sightlines from Table 1 required moderate saturation corrections of +0.06 to +0.08 dex based on the curve-of-growth fits, and these have been noted in Table 2.
Figure 2 shows the sightline-integrated gas-phase abundances \[B/O\] as a function of average line of sight hydrogen densities, $`n_\mathrm{H}`$<sup>3</sup><sup>3</sup>3Defined $`n_\mathrm{H}N(\mathrm{H})/d_{}`$, where $`d_{}`$ is the distance to the star from Table 2., for the 11 sightlines from Table 2. Where no H<sub>2</sub> column density measurements exist we adopt a generous $`+0.25`$ dex uncertainty in the total hydrogen column, $`N(\mathrm{H})`$, to account for the unknown contribution from molecular material. The bottom panel of Figure 2 shows \[O/H\] versus $`n_\mathrm{H}`$ for the same sightlines. The dashed line shows the average value of \[O/H\]($`=0.34\pm 0.02`$) from the sightlines studied by Meyer et al. (1998), several of which are included in this work.<sup>4</sup><sup>4</sup>4We adopt the O I $`f`$-value suggested by Morton (2000), which implies a $`+0.03`$ dex correction to the O I column densities in Meyer et al. (1998).
Figure 2 shows a clear trend of decreasing \[B/O\] abundance with increasing $`n_\mathrm{H}`$. Such a trend is often observed for species incorporated into interstellar dust grains (cf., Jenkins 1987). The general decrease in gas-phase abundance with increasing $`n_\mathrm{H}`$ reflects the mixture of cold and warm diffuse clouds along the line of sight (Jenkins, Savage, & Spitzer 1986), where the cold clouds exhibit a greater incorporation of elements into interstellar grains.
All of the STIS targets in Table 2 have distances $`d_{}>1000`$ pc, while the GHRS targets are all at $`d_{}<500`$ pc. It is conceivable that the gas being probed by the STIS data is also significantly more distant than that being probed by the GHRS data. In this case, some of the behavior seen in Figure 2 could be caused by abundance gradients. There are several reasons to believe this is not the case. First, of the 11 stars in our sample (Table 2), 8 have galactocentric distances between 7.5 and 9.5 kpc, i.e., they lie within 1 kpc of the solar circle. Within this 2 kpc in galactocentric distance, the \[B/O\] abundance varies from $`0.31_{0.12}^{+0.10}`$ to $`1.00_{0.12}^{+0.10}`$, i.e., the abundance varies by a factor of 5. Thus, most of the trend seen in Figure 2 occurs within 1 kpc galactocentric distance of the solar circle and cannot be caused by large-scale abundance gradients.
Second, there is evidence that the absorption along all of the sightlines actually arises within the first 1-2 kpc. Most of the absorption seen in Figure 1 is at velocities that are consistent with nearby material, and given that typical cloud-to-cloud velocity dispersions are of order $`\sigma 8`$ km s<sup>-1</sup> (Sembach & Danks 1994), most of the gas is likely very local. For example, although the extended STIS sightlines towards HD 177989 and HD 218915 pass over known spiral arms with prominent absorption in other species (e.g., Mg II, Mn II, Ni II, Cu II, and Ge II), there is no evidence for any absorption in the weak O I and B II lines from those distant structures. The nearest arms probed along these sightlines are the Perseus arm seen towards HD 218915 at a distance of $`2.5`$ kpc, and the Sagittarius arm towards HD 177989, which is likely at a distance of $`1.8`$ kpc. The Perseus and Sagittarius arms are seen seen in absorption in other species at $`v_{\mathrm{LSR}}45`$ km s<sup>-1</sup> and $`v_{\mathrm{LSR}}+18`$ km s<sup>-1</sup> towards HD 218915 and HD 177989, respectively.
The relatively local origin of the gas towards HD 218915 may explain why it does not fit well the general trend with average sightline density seen in Figure 2 (point 4 in this figure). One would prefer to use a more physically-meaningful measure of physical conditions such as the fraction of hydrogen in molecular form, $`f(\mathrm{H}_2)2N(\mathrm{H}_2)/N(\mathrm{H})`$. Figure 3 shows the \[B/O\] abundances versus $`f(\mathrm{H}_2)`$ for those sightlines with $`N(\mathrm{H}_2)`$ measurements. This diagram is sparsely populated, but there seems to be a slight trend of increasing \[B/O\] abundance with decreasing $`f(\mathrm{H}_2)`$. For this diagram to be truly useful, however, more $`N(\mathrm{H}_2)`$ measurements are needed. The Far Ultraviolet Spectroscopic Explorer will soon provide molecular hydrogen column densities for a large number of sightlines towards distant stars, making it possible to fill in the missing points in Figure 3.
The trend seen in Figure 2, and the fact that the more distant stars studied by STIS show higher \[B/O\] abundances, is likely caused by the heights of these stars above the plane of the Galaxy. Because the stars studied by STIS are more distant, they generally lie at larger distances from the Galactic plane than do the stars studied by the GHRS. Thus, the sightlines probed by our STIS measurements probe lower density regions, on average, than the sightlines probed by the GHRS measurements. The well-documented trend of higher gas-phase abundances of most elements in lower density regions suggests that we should expect the segregation of of STIS and GHRS measurements observed in Figure 2.
While the average sightline \[B/O\] values show a clear trend with $`n_\mathrm{H}`$, an imperfect measure of sightline properties, there is also evidence within the observed line profiles for variation of \[B/O\] with the physical properties of the absorbing material. Several of the sightlines displayed in Figure 1 show evidence for dense clouds with lower B/O ratios than warm clouds along the same line of sight, in qualitative agreement with the Jenkins et al. (1986) model of integrated sightline properties. Figure 4 shows the apparent column density (Savage & Sembach 1991), or $`N_a(v)`$, profile of B II $`\lambda 1362`$ towards HD 104705 with the corresponding $`N_a(v)`$ profiles of O I $`\lambda 1355`$ and Ga II $`\lambda 1414`$. This sightline exhibits a narrow, cold component centered at $`v_{\mathrm{LSR}}=0`$ km s<sup>-1</sup>, which is prominent in species such as O I, S I, Cl I, and CO, as well as a blend of warmer components between $`v_{\mathrm{LSR}}=40`$ and $`10`$ km s<sup>-1</sup> (Sembach, Howk, & Savage 2000). Figure 4 shows that the B/O ratio changes between these two regions. The integrated abundances in these two components are significantly different: $`[\mathrm{B}/\mathrm{O}]=0.57_{0.16}^{+0.12}`$ for the component centered at $`v_{\mathrm{LSR}}=0`$ km s<sup>-1</sup>, and $`[\mathrm{B}/\mathrm{O}]=+0.08_{0.17}^{+0.13}`$ for the blend of warm components at negative velocities. This sightline exhibits variations in \[B/O\] that are coupled to real changes in the physical properties of the observed components.
The dependence of \[B/O\] on $`n_\mathrm{H}`$ could potentially be caused by other effects, including true abundance variations (gas+dust) and differential ionization (e.g., Sembach et al. 2000). Boron may be particularly sensitive to the latter effect since its second ionization potential is high (25.15 eV, similar to that of C II). While the effects of differential ionization may modify the component-to-component B/O ratios, they are likely not large enough (perhaps $`0.1`$ dex) to cause the 0.8 dex range of \[B/O\] seen in Figure 2.
We believe the incorporation of boron into grains is dominant among the possible effects leading to the trend seen in Figure 2. It is reasonable to expect that boron should be incorporated into dust. It has a condensation temperature ($`910964`$ K; Zhai 1995; Lauretta & Lodders 1996) similar to those of gallium and copper and is in the same group of the periodic table as aluminum and gallium, all of which are known to be significantly incorporated into interstellar dust (Hobbs et al. 1993; Savage & Sembach 1996).
The cold cloud towards HD 104705 (Figure 4) shows a solar B/Ga ratio, though the blend of warm components exhibits super-solar B/Ga ratios. If the abundance variations between these regions are caused by the destruction of dust (see Savage & Sembach 1996), then boron is more readily-stripped from grains than is gallium.
The gas-phase abundance measurements of boron in Table 2 yield no firm information on the solid-phase abundance of boron. We derive a lower limit to the present-day total (gas+dust) interstellar boron abundance of $`\mathrm{B}/\mathrm{H}(2.5\pm 0.9)\times 10^{10}`$ (using the measured \[B/O\] towards HD 121968 and assuming $`[\mathrm{O}/\mathrm{H}]=0.34\pm 0.02`$ from Meyer et al. 1998 corrected for the Morton 2000 O I $`f`$-value). The interstellar B/H is lower than the meteoritic abundances ($`\mathrm{B}/\mathrm{H}=[7.6\pm 0.7]\times 10^{10}`$; Anders & Grevesse 1989) and non-LTE abundances for stars of solar-like metallicity ($`\mathrm{B}/\mathrm{H}=5\times 10^{10}`$; see discussion in Lambert et al. 1998).
The probable incorporation of boron into interstellar dust makes it difficult to use the measured gas-phase interstellar boron abundance to study the influence of spallation on the chemical evolution of the light elements. Another probe of spallation-induced chemical evolution is the <sup>11</sup>B/<sup>10</sup>B isotope ratio (cf., Lambert et al. 1998), and some of the sightlines presented here might allow accurate measurements of this ratio with higher resolution or signal to noise. In fact, the sightline towards HD 104705 seems to show a high <sup>11</sup>B/<sup>10</sup>B ratio; the red wing of B II profile is very similar to that of the Ga II profile. However, the relative component structure seen in the $`N_a(v)`$ profiles of B II towards HD 104705, HD 177989, and HD 218915 are significantly different than all of the other ionic species covered by our observations, including Ga II, Cu II, and Ge II. The depletion and/or ionization characteristics of B II are disimilar to these species, perhaps making it inappropriate to use them as templates for B II when studying the <sup>11</sup>B/<sup>10</sup>B ratio along complicated sightlines.
## 4 Summary
We have presented new and archival observations of the gas-phase abundance of boron in the diffuse interstellar medium using STIS and GHRS. From our analysis of these high-quality absorption line data, and measurements from the literature, we have concluded the following.
1. The gas-phase abundance of \[B/O\] in the ISM is anticorrelated with the average density of hydrogen along the sightline being probed, as well as the fraction of hydrogen seen in molecular form along a sightline. Along individual sightlines, we also find a significantly higher gas-phase \[B/O\] abundance in warm than in cold diffuse clouds. The evidence strongly suggests that boron is incorporated into dust grains in the diffuse ISM.
2. The relative component-to-component strengths in the observed B II profiles are significantly different than those of any other observed species, including Ga II, Cu II, and Ge II. The depletion and/or ionization characteristics of B II are different than those of other species. Although Ga II, Cu II, and Ge II are sometimes used as templates for modeling the B II absorption, the differences seen in our data suggest it may be inappropriate to assume they trace the B II profile when deriving the <sup>11</sup>B/<sup>10</sup>B ratio along complicated sightlines.
3. We derive a lower limit to the present-day total (gas+dust) B/H abundance of $`\mathrm{B}/\mathrm{H}(2.5\pm 0.9)\times 10^{10}`$.
We thank S. Federman for suggestions on this work. This work was supported by NASA through grants GO-0720.01-96A and GO-0720.02-96A from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555.
|
warning/0005/hep-ph0005028.html
|
ar5iv
|
text
|
# 1 Introduction.
## 1 Introduction.
The studies of the gluonic content of hadrons are complicated by the fact that there exists no probes which couples directly to it. The only way for electroweak bosons to fuse with coloured gluons is via quark loops. Obviously this is an effect suppressed in the QCD coupling constant on the background of dominating tree level quark contributions. Gluon distributions mix logarithmically with quark ones and although they can affect the shape of the latter in a significant way it is still extremely complicated to disentangle them in perturbative evolution. An opportunity to measure gluons directly could appear provided there is a selection rule which forbids quarks to enter as a leading effect in the cross section.
At leading twist-two level there are three non-local light-ray operators bilinear in gluon fields. E.g. in the light-cone gauge where only transversely polarized vector fields do propagate, we have the following projections for the Lorentz structure of the composite operator $`B_\mu (x_1)B_\nu (x_2)g_{\mu \mu ^{}}^{}g_{\nu \nu ^{}}^{}`$
$`g_{\mu \mu ^{}}^{}g_{\nu \nu ^{}}^{}=\frac{1}{2}g_{\mu \nu }^{}g_{\mu ^{}\nu ^{}}^{}+\frac{1}{2}ϵ_{\mu \nu }^{}ϵ_{\mu ^{}\nu ^{}}^{}+\tau _{\mu \nu ;\rho \sigma }^{}\tau _{\mu ^{}\nu ^{};\rho \sigma }^{},`$ (1)
where $`g_{\mu \nu }^{}=g_{\mu \nu }n_\mu n_\nu ^{}n_\mu ^{}n_\nu `$, $`ϵ_{\mu \nu }^{}=ϵ_{\mu \nu \rho \sigma }n_\rho ^{}n_\sigma `$ and $`\tau _{\mu \nu ;\rho \sigma }^{}=\frac{1}{2}\left(g_{\mu \rho }^{}g_{\nu \sigma }^{}+g_{\mu \sigma }^{}g_{\nu \rho }^{}g_{\mu \nu }^{}g_{\rho \sigma }^{}\right)`$. Here $`n_\mu `$ and $`n_\nu ^{}`$ are two light-cone vectors such that $`n^2=n^{\mathrm{\hspace{0.17em}2}}=0`$ and $`nn^{}=1`$. Eq. (1) corresponds to the Clebsch-Gordon decomposition of the direct product of the two vector representations of the Lorentz group<sup>1</sup><sup>1</sup>1Recall that the representations of the Lorentz group $`L_+^{}=SO(3,1)=SO(4,𝐂)_R\left(SL(2,𝐂)SL(2,𝐂)\right)_R`$ are labeled by a pair $`(j_1,j_2)`$ which are eigenvalues $`j_i(j_i+1)`$ of the $`SL(2)`$ Casimir operators $`\widehat{𝑱}_i^2`$. $`(\frac{1}{2},\frac{1}{2})(\frac{1}{2},\frac{1}{2})=(0,0)\left((1,0)(0,1)\right)(1,1)`$. The first two tensors stand for vector and axial sectors and they contribute to the conventional deep inelastic scattering (DIS) process on spin-$`\frac{1}{2}`$ hadrons. The last one cannot appear there since it requires the photon helicity to be flipped by two units. However, it definitely can appear in scattering on spin $`J1`$ targets as was emphasized in Ref. . Since these operators belong to the spin-2 representation $`(1,1)`$ of the Lorentz group, they cannot mix with quark operators. Therefore they can serve as a clean probe of the gluonic content in hadrons not contaminated by other effects. Recently this issue was discussed in the context of current fragmentation in DIS and in the production of two pions in $`\gamma \gamma `$ fusion . Nevertheless this operator can appear in the Compton scattering on the nucleon provided it is sandwiched between states with different momenta as was pointed out and elaborated in Ref. . The corresponding kinematics of the process is known as deeply virtual Compton scattering (DVCS) . In this note we address the issue of the tensor gluon skewed parton distributions (SPD) in DVCS in great detail. We recalculate the one-loop coefficient function, and present two-loop anomalous dimensions and exclusive evolution kernels for this sector. We give an explicit prediction for weighted cross sections which can be used to extract directly the tensor gluon SPD from experimental data.
## 2 Leading twist amplitude.
The hadronic part of the deeply virtual Compton scattering amplitude is defined by the off-forward matrix element of the correlator of two electro-magnetic currents sandwiched between states with unequal momenta
$`T_{\mu \nu }(q,P_1,P_2)=i{\displaystyle 𝑑xe^{ixq}P_2|Tj_\mu (x/2)j_\nu (x/2)|P_1},`$ (2)
where $`q=(q_1+q_2)/2`$ (and the index $`\mu `$ refers to the outgoing real photon with momentum $`q_2`$) and $`P_1`$ ($`P_2`$) is the momentum of incoming (outgoing) nucleon. The leading contribution of the light-ray tensor gluon operator<sup>2</sup><sup>2</sup>2The path-ordered link factor $`\mathrm{\Phi }[x_2,x_1]`$ ensures gauge invariance.
$`{}_{}{}^{G}𝒪_{\mu \nu }^{T}(\kappa _1,\kappa _2)=G_{+\rho }(\kappa _2n)\tau _{\mu \nu ;\rho \sigma }^{}\mathrm{\Phi }[\kappa _2n,\kappa _1n]G_{\sigma +}(\kappa _1n),`$ (3)
whose off-forwards matrix element is parametrized via two SPDs
$`G_{\mu \nu }^T(t,\eta ,\mathrm{\Delta }^2)4P_+^1{\displaystyle \frac{d\kappa }{2\pi }e^{i\kappa tP_+}P_2|{}_{}{}^{G}𝒪_{\mu \nu }^{T}(\kappa ,\kappa )|P_1}=H_G^T(t,\eta ,\mathrm{\Delta }^2){\displaystyle \frac{\tau _{\mu \nu ;\alpha \beta }^{}}{2M}}{\displaystyle \frac{\mathrm{\Delta }_\alpha q_\gamma }{Pq}}\overline{U}(P_2)i\sigma _{\gamma \beta }U(P_1)`$ (4)
$`+E_G^T(t,\eta ,\mathrm{\Delta }^2){\displaystyle \frac{\tau _{\mu \nu ;\alpha \beta }^{}}{4M^2}}\mathrm{\Delta }_\alpha \overline{U}(P_2)\left({\displaystyle \frac{\mathrm{\Delta }_\beta \overline{)}q}{Pq}}\eta \gamma _\beta \right)U(P_1),`$
into the operator product expansion of currents (2) appear at one-loop order (see Fig. 1). The traceless symmetric projector $`\tau ^{}`$ in Eq. (4) possesses the properties $`\tau _{\mu \nu ;\rho \sigma }^{}\tau _{\mu \nu ;\rho ^{}\sigma ^{}}^{}=\tau _{\rho \sigma ;\rho ^{}\sigma ^{}}^{}`$, $`\tau _{\mu \nu ;\rho \sigma }^{}=\tau _{\rho \sigma ;\mu \nu }^{}`$, $`\tau _{\mu \mu ;\rho \sigma }^{}=0`$, $`\tau _{\mu \nu ;\mu \nu }^{}=2`$. The kinematical variables used here and below are introduced as $`\omega \xi ^1=Pq/q^2`$ (generalized Bjorken variable), $`\eta =\mathrm{\Delta }q/Pq`$ (skewedness), in terms of the vectors $`P=P_1+P_2`$, $`q=\frac{1}{2}(q_1+q_2)`$ and $`\mathrm{\Delta }=P_2P_1=q_1q_2`$.
A simple calculation of one-loop diagrams (see Fig. 1) gives us the following result, with restriction of the reality of the final photon being relaxed, i.e. $`q_2^20`$,
$$T_{\mu \nu }=\frac{\alpha _s}{\pi }T_F\underset{i=1}{\overset{N_f}{}}Q_i^2_1^1𝑑tG_{\mu \nu }^T(t,\eta ,\mathrm{\Delta }^2)\sigma (t,\eta )\left\{1+\frac{1\omega ^2\eta ^2}{\omega ^2(t^2\eta ^2)}\mathrm{ln}\frac{1t^2\omega ^2}{1\eta ^2\omega ^2}\right\},$$
(5)
which agrees with Ref. . The function $`\sigma (t,\eta )`$ appears from the conversion of gluon field (taken in the light-cone gauge) into strength tensor, $`G_{+\mu }=_+B_\mu `$, and reads, for the fixing of the residual gauge symmetry consistent with canonical hamiltonian formalism ,
$`\sigma (t,\eta )`$ $`=`$ $`{\displaystyle \frac{1}{(t\eta +i0)(t+\eta i0)}}.`$ (6)
Apart from (suppressed) momentum dependence of the coupling constant an additional source of scaling violation results from the renormalization of the composite operator (3) which will be discussed in the next two sections at one- and two-loop order.
## 3 Evolution: anomalous dimensions.
At leading order in the coupling constant the light-ray operator (3) obeys the light-cone position evolution equation of the form (here and below $`\overline{y}1y`$)
$$\frac{d}{d\mathrm{ln}\mu ^2}[{}_{}{}^{G}𝒪_{}^{T}(\kappa _1,\kappa _2)]=\frac{\alpha _s}{2\pi }_0^1𝑑z_0^{\overline{z}}𝑑y{}_{}{}^{GG}𝒦_{}^{T}(y,z)[{}_{}{}^{G}𝒪_{}^{T}(\overline{y}\kappa _1+y\kappa _2,z\kappa _1+\overline{z}\kappa _2)]$$
with the kernel
$${}_{}{}^{GG}𝒦_{}^{T}(y,z)=C_A\left\{y2+\left[\frac{1}{y}\right]_+\right\}\delta (z)+C_A\left\{z2+\left[\frac{1}{z}\right]_+\right\}\delta (y)\frac{\beta _0}{2}\delta (y)\delta (z),$$
(7)
where $`\beta _0=\frac{4}{3}T_FN_f\frac{11}{3}C_A`$ is the first coefficient of the QCD $`\beta `$-function. Fourier transformation of this result gives the exclusive evolution kernel known before . The tree level conformal invariance of the QCD Lagrangian allows to diagonalize this equation in the basis spanned by Gegenbauer polynomials<sup>3</sup><sup>3</sup>3We drop in what follows the Lorentz indices on the operators. Here $`=\stackrel{}{}+\stackrel{}{}`$ and $`\stackrel{}{𝒟}=\stackrel{}{𝒟}\stackrel{}{𝒟}`$.
$${}_{}{}^{G}𝒪_{jl}^{T}=G_{+\rho }(i_+)^{l1}\tau _{\mu \nu ;\rho \sigma }^{}C_{j1}^{5/2}(\underset{+}{\overset{}{𝒟}}/_+)G_{\sigma +},$$
(8)
which form an infinite dimensional representation of the conformal group in the space of bilinear operators. Therefore,
$$\frac{d}{d\mathrm{ln}\mu ^2}[{}_{}{}^{G}𝒪_{jl}^{T}]=\frac{1}{2}\underset{k=1}{\overset{j}{}}{}_{}{}^{GG}\gamma _{jk}^{T}[{}_{}{}^{G}𝒪_{jl}^{T}],\text{with}{}_{}{}^{GG}\gamma _{j}^{T(0)}=4C_A\left(\psi (j+2)\psi (1)\right)+\beta _0,$$
(9)
the first term in the expansion $`{}_{}{}^{GG}\gamma _{jk}^{T}=\left(\frac{\alpha _s}{2\pi }\right){}_{}{}^{GG}\gamma _{j}^{T(0)}\delta _{jk}+\left(\frac{\alpha _s}{2\pi }\right)^2{}_{}{}^{GG}\gamma _{jk}^{T(1)}+𝒪(\alpha _s^3)`$.
In the momentum fraction space the evolution of the SPD can be done making use of orthogonal polynomial reconstruction (in the following Gegenbauer polynomials, $`C_j^{5/2}`$) of the function from its conformal moments according to Ref.
$$G(t,\eta ,Q^2)=\underset{j=1}{\overset{N_{\mathrm{max}}}{}}\stackrel{~}{C}_{j1}^{5/2}(t)\underset{k=1}{\overset{j}{}}c_{jk}(\eta )\left(\frac{\alpha _s(Q_0^2)}{\alpha _s(Q^2)}\right)^{{}_{}{}^{GG}\gamma _{k}^{T(0)}/\beta _0}\eta ^{k1}_1^1𝑑tC_{k1}^{5/2}(t/\eta )G(t,\eta ,Q_0^2),$$
(10)
where formally $`N_{\mathrm{max}}=\mathrm{}`$. Here $`\stackrel{~}{C}_{j1}^{5/2}(t)=\frac{9}{2}\frac{(2j+3)}{(j)_4}\left(1t^2\right)^2C_{j1}^{5/2}(t)`$ are adjoint polynomials and the re-expansion coefficients $`c_{jk}(\eta )=\stackrel{~}{C}_{k1}^{5/2}(t)|C_{j1}^{5/2}(\eta t)`$ are expressed in terms of hypergeometric function $`{}_{2}{}^{}F_{1}^{}(\eta ^2)`$. The evolution is demonstrated in Fig. 2 where we have taken an $`\eta `$-independent input<sup>4</sup><sup>4</sup>4We factored out the $`\mathrm{\Delta }^2`$ dependence into the gluonic form factor $`F_G(\mathrm{\Delta }^2)`$ which for phenomenological estimations can be taken equal $`\kappa _T\left(1\mathrm{\Delta }^2/M_\mathrm{\Lambda }^2\right)^3`$ with $`M_\mathrm{\Lambda }=2.7\mathrm{GeV}`$ and unknown parameter $`\kappa _T`$ which defines the magnitude of proton matrix element of the tensor gluonic operator., $`G(t,\eta )=\frac{3}{4}(1t^2)`$, at very low $`Q_0^2=0.2\mathrm{GeV}^2`$ and evolved it up to $`Q^2=4\mathrm{GeV}^2`$ (b) and $`Q^2=100\mathrm{GeV}^2`$ (c).
As a side remark on the solution of the leading order equation for “transversity” let us note that even in the case when any arbitrary number of gluons with the same helicity, whose pair-wise interaction is described by the kernel (7), are exchanged in the $`t`$-channel, one can still diagonalize the (LO) multi-particle kernel since the problem admits a large enough number of conservation laws to be completely integrable. This can be found by noticing the equivalence of the anomalous dimensions (9) (which depend on the eigenvalues of $`SL(2)`$ Casimir operator $`\widehat{𝑱}^2`$) to the Hamiltonian of the exactly solvable one-dimensional $`XXX_{s=3/2}`$ spin chain model .
Next we compute the two-loop anomalous dimensions for the tensor gluon operator. To this end we use our machinery developed in Ref. . Let us give a brief outline of the method. The structure of anomalous dimensions of conformal operators at two-loop (and higher) order reads
$${}_{}{}^{GG}\gamma _{jk}^{T(1)}={}_{}{}^{GG}\gamma _{j}^{T(1),\mathrm{D}}\delta _{jk}+{}_{}{}^{GG}\gamma _{jk}^{T(1),\mathrm{ND}},$$
(11)
where $`{}_{}{}^{GG}\gamma _{j}^{T(1),\mathrm{D}}`$ are the next-to-leading order (NLO) forward anomalous dimensions<sup>5</sup><sup>5</sup>5We thank W. Vogelsang for providing us his result for local anomalous dimensions.
$`{}_{}{}^{GG}\gamma _{j}^{T(1),\mathrm{D}}`$ $`=`$ $`C_A^2\left\{S_1(j+1)\left({\displaystyle \frac{134}{9}}4S_2^{}\left({\displaystyle \frac{j+1}{2}}\right)\right)S_3^{}\left({\displaystyle \frac{j+1}{2}}\right)+8\stackrel{~}{S}(j+1){\displaystyle \frac{1}{j(j+3)}}{\displaystyle \frac{16}{3}}\right\}`$ (12)
$`+`$ $`C_AT_FN_f\left\{{\displaystyle \frac{8}{3}}{\displaystyle \frac{40}{9}}S_1(j+1){\displaystyle \frac{2}{j(j+3)}}\right\}+C_FT_FN_f{\displaystyle \frac{2(j+1)(j+2)}{j(j+3)}},`$
with
$`S_{\mathrm{}}(j)={\displaystyle \underset{k=1}{\overset{j}{}}}{\displaystyle \frac{1}{k^{\mathrm{}}}},S_{\mathrm{}}^{}\left({\displaystyle \frac{j}{2}}\right)=2^{\mathrm{}}{\displaystyle \underset{k=1}{\overset{j}{}}}{\displaystyle \frac{\sigma _k}{k^{\mathrm{}}}},\stackrel{~}{S}(j)={\displaystyle \underset{k=1}{\overset{j}{}}}{\displaystyle \frac{(1)^k}{k^2}}S_1(k),`$
and where $`\sigma _j=\frac{1}{2}[1+(1)^j]`$. The non-diagonal elements of anomalous dimension matrix of the conformal operators arise due to one-loop breaking of the conformal symmetry and, since the tree level conformal invariance leads to diagonal anomalous dimensions, the one-loop special conformal anomaly generates two-loop anomalous dimensions. The use of four-dimensional conformal algebra provides a relation between the anomalies of dilatation (read anomalous dimensions in question) and special conformal transformations via the commutator $`[𝒟,𝒦_{}]=i𝒦_{}`$ which is applied on the Green function of elementary fields with conformal operator insertion. To evaluate the commutator the knowledge of scale and special conformal Ward identities, with unraveled pattern of symmetry breaking for afore mentioned Green function, is indispensable. A careful analysis reveals the result
$${}_{}{}^{GG}\gamma _{jk}^{T(1),\mathrm{ND}}=\left({}_{}{}^{GG}\gamma _{j}^{T(0)}{}_{}{}^{GG}\gamma _{k}^{T(0)}\right)\left\{d_{jk}\left(\beta _0{}_{}{}^{GG}\gamma _{k}^{T(0)}\right)+{}_{}{}^{GG}g_{jk}^{T}\right\},$$
(13)
where $`{}_{}{}^{GG}\gamma _{k}^{T(0)}`$ are already known LO anomalous dimensions (9), $`d_{jk}=\sigma _{jk}(2k+3)/(jk)(j+k+3)`$ for $`j>k`$ and $`g_{jk}^T`$ appears as a counterterm required for renormalization of the product of two composite operators: integrated anomaly $`𝒪_A^{}d^4x\mathrm{\hspace{0.17em}2}x_{}𝒪_A(x)=d^4xx_{}Z_3\left(G_{\mu \nu }\right)^2`$ in the trace of energy-momentum tensor and a conformal operator. The structure of the counterterms has been established using the form of counterterms for differential vertex operator insertions and found to be
$`i[𝒪_A(x)][{}_{}{}^{G}𝒪_{jl}^{T}]=i[𝒪_A(x){}_{}{}^{G}𝒪_{jl}^{T}]`$ $``$ $`\delta ^{(d)}(x){\displaystyle \underset{k=0}{\overset{j}{}}}\left\{\widehat{Z}_A\right\}_{jk}[{}_{}{}^{G}𝒪_{kl}^{T}]{\displaystyle \frac{i}{2}}_+\delta ^{(d)}(x){\displaystyle \underset{k=0}{\overset{j}{}}}\left\{\widehat{Z}_A^{}\right\}_{jk}[{}_{}{}^{G}𝒪_{kl1}^{T}]\mathrm{}`$ (14)
$``$ $`\left(g{\displaystyle \frac{\mathrm{ln}X}{g}}2\xi {\displaystyle \frac{\mathrm{ln}X}{\xi }}\right)B_\mu ^a(x){\displaystyle \frac{\delta }{\delta B_\mu ^a(x)}}[{}_{}{}^{G}𝒪_{jl}^{T}],`$
where $`X=Z_gZ_3^{1/2}`$ is expressed in terms of the renormalization constants of the gluon wave function $`Z_3`$ and the coupling $`Z_g`$.
To determine the unknown renormalization matrix $`Z_A^{}`$ it proves convenient to work in the light-cone position formalism and calculate 1PI Green function $`[𝒪_A(x)][{}_{}{}^{G}𝒪_{}^{T}(\kappa _1,\kappa _2)]B_\mu (x_1)B_\nu (x_2)_{\mathrm{amp}}`$ shown in Fig. 3 (explicit one-loop graphs can be found in Ref. ). We write the divergent part of the operator product $`i[𝒪_A^{}][{}_{}{}^{G}𝒪_{}^{T}(\kappa _1,\kappa _2)]`$ via the following relation in leading order of the coupling constant:
$`i[𝒪_A^{}][{}_{}{}^{G}𝒪_{}^{T}(\kappa _1,\kappa _2)]`$ $`=`$ $`{\displaystyle \frac{\alpha _s}{2\pi }}{\displaystyle \frac{i}{ϵ}}{\displaystyle _0^1}dz{\displaystyle _0^{\overline{z}}}dy\{{}_{}{}^{GG}𝒦_{A}^{}(y,z)[{}_{}{}^{G}𝒪_{}^{T}(\overline{y}\kappa _1+y\kappa _2,z\kappa _1+\overline{z}\kappa _2)]`$ (15)
$`+`$ $`{}_{}{}^{GG}\stackrel{~}{𝒦}_{A}^{}(y,z){\displaystyle }d^dx\mathrm{\hspace{0.17em}2}x_{}B_\mu ^b(x){\displaystyle \frac{\delta }{\delta B_\mu ^b(x)}}[{}_{}{}^{G}𝒪_{}^{T}(\overline{y}\kappa _1+y\kappa _2,z\kappa _1+\overline{z}\kappa _2)]\}.`$
For the evolution kernel $`{}_{}{}^{GG}\stackrel{~}{𝒦}_{A}^{}`$ we find immediately from Eq. (14) that $`{}_{}{}^{GG}\stackrel{~}{𝒦}_{A}^{}(y,z)=i\frac{3}{4}\delta (y)\delta (z)`$. Explicit calculation of the amputated Green function of conformal anomaly and conformal operator gives
$${}_{}{}^{GG}𝒦_{A}^{}(y,z)=i2(\kappa _1+\kappa _2){}_{}{}^{GG}𝒦_{}^{T}(y,z)i(\kappa _1+\kappa _2)\beta _0\delta (y)\delta (z)+{}_{}{}^{GG}𝒦_{w}^{T}(y,z),$$
(16)
where $`{}_{}{}^{GG}𝒦_{}^{T}(y,z)`$ is the LO evolution kernel (7) and
$`{}_{}{}^{GG}𝒦_{w}^{T}(y,z)`$ $`=`$ $`2{\displaystyle \frac{C_A}{k_{2+}}}\left\{\left[{\displaystyle \frac{2}{z}}\right]_+\delta (y)\left[{\displaystyle \frac{1}{z^2}}\right]_+\delta (y)+(2y1)\delta (z)\right\}`$ (17)
$`+`$ $`2{\displaystyle \frac{C_A}{k_{1+}}}\left\{\left[{\displaystyle \frac{2}{y}}\right]_+\delta (z)\left[{\displaystyle \frac{1}{y^2}}\right]_+\delta (z)+(2z1)\delta (y)\right\}.`$
The evaluation of the conformal moments of (17) \[multiplied by $`1/2(jk)(j+k+3)`$\] gives us
$`{}_{}{}^{GG}g_{jk}^{T}`$ $`=`$ $`2C_A\sigma _{jk}\theta _{j2,k}{\displaystyle \frac{(3+2k)}{(jk)(j+k+3)}}`$
$`\times `$ $`\left\{2A_{jk}+(A_{jk}\psi (j+2)+\psi (1))\left[{\displaystyle \frac{\mathrm{\Gamma }(j+4)\mathrm{\Gamma }(k)}{\mathrm{\Gamma }(j)\mathrm{\Gamma }(k+4)}}1\right]\right\},`$
where we have introduced the matrix $`A`$ with its elements defined by
$$A_{jk}=\psi \left(\frac{j+k+4}{2}\right)\psi \left(\frac{jk}{2}\right)+2\psi (jk)\psi (j+2)\psi (1),$$
(19)
and $`\theta _{jk}=1`$ for $`j>k`$ and zero otherwise. Note that this result differs from our previous ones for chiral even operators only by a pure rational function
$${}_{}{}^{GG}g_{jk}^{T}={}_{}{}^{GG}g_{jk}^{V}+4C_A\sigma _{jk}(3+2k)\frac{\mathrm{\Gamma }(k)}{\mathrm{\Gamma }(k+4)}.$$
(20)
The solution of two-loop evolution equation is straightforward and can be found in Ref. .
## 4 Evolution: two-loop kernel.
In this section we give our results for the two-loop exclusive evolution kernel reconstructed from the known non-diagonal anomalous dimensions found above and two-loop splitting function of Ref. . The structure of ER-BL kernel to $`𝒪\left(\alpha _s^3\right)`$ accuracy reads
$${}_{}{}^{GG}V_{}^{T}=\frac{\alpha _s}{2\pi }\left\{\left[{}_{}{}^{GG}V_{}^{T(0)}\right]_+\frac{1}{2}{}_{}{}^{GG}\gamma _{1}^{T(0)}\delta (xy)\right\}+\left(\frac{\alpha _s}{2\pi }\right)^2\left\{\left[{}_{}{}^{GG}V_{}^{T(1)}\right]_+\frac{1}{2}{}_{}{}^{GG}\gamma _{1}^{T(1)}\delta (xy)\right\},$$
(21)
where the +-prescription is defined according to
$`\left[V(x,y)\right]_+=V(x,y)\delta (xy){\displaystyle _0^1}𝑑zV(z,y).`$ (22)
Note that this prescription is not in one-to-one correspondence with the $`+`$-prescription in the forward case. The LO kernel reads
$`{}_{}{}^{GG}V_{}^{T(0)}=C_A\theta (yx){}_{}{}^{GG}f_{}^{T}(x,y)+\left\{{\displaystyle \genfrac{}{}{0pt}{}{x\overline{x}}{y\overline{y}}}\right\}\text{with}{}_{}{}^{GG}f_{}^{T}={\displaystyle \frac{x^2}{y^2}}{\displaystyle \frac{1}{yx}}.`$ (23)
To obtain the NLO correction we use the approach described in . Since the tensor gluon sector is almost analogous to handle as the vector and axial ones, we only mention the differences to these cases. The structure of two-loop anomalous dimensions (13) implies the following form of the kernel
$`{}_{}{}^{GG}V_{}^{T(1)}={}_{}{}^{GG}\dot{V}_{}^{T}\left({}_{}{}^{GG}V_{}^{T(0)}+{\displaystyle \frac{\beta _0}{2}}\right)\left[{}_{}{}^{GG}g_{}^{T}\underset{\text{}}{}{}_{}{}^{GG}V_{}^{T(0)}\right]{\displaystyle \frac{C_A^2}{2}}{}_{}{}^{GG}G_{}^{T}+{}_{}{}^{GG}D_{}^{T},`$ (24)
where the commutator stands for $`[A\underset{\text{}}{}B](x,y)=_0^1𝑑z\left\{A(x,z)B(z,y)B(x,z)A(z,y)\right\}`$. The off-diagonal part is contained in the first two convolutions on the r.h.s. of this equation and they are known exactly. The conformal moments of the dotted kernel,
$${}_{}{}^{GG}\dot{V}_{}^{T(0)}=C_A\theta (yx)\frac{x^2}{y^2}\frac{1}{yx}\mathrm{ln}\frac{x}{y}+\left\{\genfrac{}{}{0pt}{}{x\overline{x}}{y\overline{y}}\right\},$$
(25)
are proportional to the commutator of the $`d`$-matrix with the LO anomalous dimensions. The $`g`$-matrix corresponds to the kernel
$`{}_{}{}^{GG}g_{}^{T}=C_A\theta (yx)\left[2{\displaystyle \frac{x}{y}}{\displaystyle \frac{\mathrm{ln}\left(1\frac{x}{y}\right)}{yx}}\right]+\left\{{\displaystyle \genfrac{}{}{0pt}{}{x\overline{x}}{y\overline{y}}}\right\},`$ (26)
which contains a part (second term) of the chiral even (odd) case. The difference is a rational function (first term) which can be restored from the $`GQ`$-channel result by applying appropriate convolutions in order to trade the denominator $`1/(j+1)(j+2)`$ of the $`GQ`$ conformal moments of $`x/y`$ to go over into the $`GG`$ sector<sup>6</sup><sup>6</sup>6This can be achieved by convolution with the so-called $`c`$-kernel in $`QQ`$ channels and differentiation with respect to $`y`$, i.e. $`d^3y^2\overline{y}^2/dy^3`$. Finally, we removed a pure diagonal piece., i.e. $`1/j(j+1)(j+2)(j+3)`$.
Now the problem is reduced to the reconstruction of diagonal parts. The third term reads
$`{}_{}{}^{GG}G_{}^{T}`$ $`=`$ $`\left\{\theta (yx)\left[{}_{}{}^{GG}h_{}^{T}(x,y)+\mathrm{\Delta }{}_{}{}^{GG}h_{}^{T}(x,y)\right]+\theta (y\overline{x})\left[{}_{}{}^{GG}\overline{h}_{}^{T}(x,y)+\mathrm{\Delta }{}_{}{}^{GG}h_{}^{T}(\overline{x},y)\right]\right\}`$ (27)
$`+`$ $`\left\{{\displaystyle \genfrac{}{}{0pt}{}{x\overline{x}}{y\overline{y}}}\right\},`$
and arises from the crossed-ladder diagram and can be obtained by means of $`𝒩=1`$ supersymmetry from the one in the quark sector. The particular contributions are
$`{}_{}{}^{GG}h_{}^{T}`$ $`=`$ $`2{}_{}{}^{GG}\overline{f}_{}^{T}\mathrm{ln}\overline{x}\mathrm{ln}y2{}_{}{}^{GG}f_{}^{T}\left[\mathrm{Li}_2(x)+\mathrm{Li}_2(\overline{y})\right],\mathrm{\Delta }{}_{}{}^{GG}h_{}^{T}={\displaystyle \frac{2x}{y^2\overline{y}}}{\displaystyle \frac{2\overline{x}}{y^2\overline{y}}}\mathrm{ln}\overline{x}{\displaystyle \frac{2x}{y\overline{y}^2}}\mathrm{ln}y,`$
$`{}_{}{}^{GG}\overline{h}_{}^{T}`$ $`=`$ $`\left({}_{}{}^{GG}f_{}^{T}{}_{}{}^{GG}\overline{f}_{}^{T}\right)\left[2\mathrm{L}\mathrm{i}_2\left(1{\displaystyle \frac{x}{y}}\right)+\mathrm{ln}^2y\right]+2{}_{}{}^{GG}f_{}^{T}\left[\mathrm{Li}_2(\overline{y})\mathrm{ln}x\mathrm{ln}y\right]+2{}_{}{}^{GG}\overline{f}_{}^{T}\mathrm{Li}_2(\overline{x}).`$
Here $`\mathrm{Li}_2(x)=_0^x\frac{dt}{t}\mathrm{ln}(1t)`$ is the Euler dilogarithm and we used the following shorthand notation $`{}_{}{}^{GG}\overline{f}_{}^{T}={}_{}{}^{GG}f_{}^{T}(\overline{x},\overline{y})`$. The remaining diagonal part $`D`$ has a simple representation in terms of LO kernels and can be obtained by taking the forward limit and comparison with the known DGLAP kernel . Consequent restoration of diagonal ER-BL kernels from the splitting functions is straightforward and gives
$`{}_{}{}^{GG}D_{}^{T}=`$ $``$ $`C_FT_FN_f\left[{}_{}{}^{GG}v_{}^{a}+{\displaystyle \frac{2}{3}}{}_{}{}^{GG}v_{}^{c}\right]+\beta _0C_A\left[{\displaystyle \frac{3}{8}}{}_{}{}^{GG}v_{}^{a}{\displaystyle \frac{5}{6}}{}_{}{}^{GG}v_{}^{b}+{\displaystyle \frac{1}{4}}{}_{}{}^{GG}v_{}^{c}\right]`$ (29)
$`+`$ $`C_A^2\left[{\displaystyle \frac{13}{8}}{}_{}{}^{GG}v_{}^{a}{\displaystyle \frac{11}{6}}{}_{}{}^{GG}v_{}^{b}+{\displaystyle \frac{13}{12}}{}_{}{}^{GG}v_{}^{c}\right].`$
Here the $`b`$ kernel coincides (with colour factor being dropped) with the LO kernel (23) and the $`a`$ and $`c`$ kernels having the structure $`v^i(x,y)=f^i\theta (yx)+\overline{f}^i\theta (xy)`$ are defined by the functions $`{}_{}{}^{GG}f_{}^{a}=x^2/y^2`$ and $`{}_{}{}^{GG}f_{}^{c}=x^2(2\overline{x}y+yx)/y^2`$. Eqs. (24-29) is our final result for the NLO corrections to the gluon “transversity” evolution kernel.
## 5 Cross sections.
Now we are in a position to study the cross section for electroproduction of real photon where the gluonic SPD, whose perturbative properties we studied in detail in the previous sections, can be measured. For DVCS the skewedness $`\eta `$ and generalized Bjorken variable $`\xi `$ are proportional $`\eta =\xi \left(1+\frac{\mathrm{\Delta }^2}{2𝒬^2}\right)^1`$. The differential cross section with unpolarized lepton beam and unpolarized target, in variables $`y=P_1q_1/P_1k`$, $`𝒬^2q_1^2`$ and $`x𝒬^2/(2P_1q_1)`$ with $`\xi =x\left(1+\frac{\mathrm{\Delta }^2}{2𝒬^2}\right)\left(2x+x\frac{\mathrm{\Delta }^2}{𝒬^2}\right)^1`$, reads
$`{\displaystyle \frac{d\sigma }{dxdyd|\mathrm{\Delta }^2|d\varphi _r}}={\displaystyle \frac{\alpha ^3xy}{8\pi 𝒬^2}}\left(1+{\displaystyle \frac{4M^2x}{𝒬^2}}\right)^{1/2}\left|{\displaystyle \frac{𝒯}{e^3}}\right|^2.`$ (30)
Here $`\varphi _r`$ is the azimuthal angle between the lepton and proton scattering planes in the rest frame of the target. In the following we will be interested only in the interference term $`||^2𝒯_{\mathrm{BH}}𝒯_{\mathrm{DVCS}}^{}+𝒯_{\mathrm{DVCS}}𝒯_{\mathrm{BH}}^{}`$ between the DVCS, $`𝒯_{\mathrm{DVCS}}`$, and Bethe-Heitler amplitude $`𝒯_{\mathrm{BH}}`$ since it provides a unique opportunity to extract the real/imaginary part of the gluonic amplitudes
$$\left\{\begin{array}{c}_G^T(\xi ,\mathrm{\Delta }^2)\\ _G^T(\xi ,\mathrm{\Delta }^2)\end{array}\right\}=\frac{\alpha _s}{\pi }T_F\underset{i=1}{\overset{N_f}{}}Q_i^2_1^1𝑑t\sigma (t,\xi )\left\{\begin{array}{c}H_G^T(t,\xi ,\mathrm{\Delta }^2)\\ E_G^T(t,\xi ,\mathrm{\Delta }^2)\end{array}\right\}.$$
(31)
An explicit calculation gives for unpolarized settings
$`\left|{\displaystyle \frac{}{e^3}}\right|^2`$ $`=`$ $`{\displaystyle \frac{(\pm 1)}{2𝒬^2\mathrm{\Delta }^2}}\mathrm{Sp}\left\{\overline{)}k\left[\gamma _\gamma (\overline{)}k\overline{)}\mathrm{\Delta })^1\gamma _\mu +\gamma _\mu (\overline{)}k^{}+\overline{)}\mathrm{\Delta })^1\gamma _\gamma \right]\overline{)}k^{}\gamma _\nu \right\}`$ (32)
$`\times \tau _{\mu \nu ;\alpha \beta }^{}\mathrm{\Delta }_\alpha \left\{\left(\xi g_{\beta \gamma }+{\displaystyle \frac{\mathrm{\Delta }_\beta q_\gamma }{Pq}}\right)𝒯_1+P_\gamma \mathrm{\Delta }_\beta 𝒯_2\right\},`$
where
$$𝒯_1=2(F_1+F_2)\mathrm{Re}\left(_G^T+\frac{\mathrm{\Delta }^2}{4M^2}_G^T\right),𝒯_2=\frac{1}{2M^2}\mathrm{Re}\left(F_1_G^TF_2_G^T\right).$$
(33)
and $`+()`$ sign stands for electron (positron) beam. In consequent evaluation of this expression we perform an expansion in $`1/𝒬^2`$ and keep only the first non-vanishing contribution. We form the charge asymmetry in order to extract the interference term from the total cross section. Since the double helicity flip amplitude presently considered is uniquely proportional to $`\mathrm{cos}\left(3\varphi _r\right)`$ (while all other contributions to the non-polarized cross section $`||^2`$ enter with $`\mathrm{cos}(n\varphi _r)`$ and $`n=1,2`$ one can isolate this purely gluonic contributions from the effects of quark SPD by forming an appropriate weighted cross section $`𝑑\varphi _rw(\varphi _r)\sigma (\varphi _r)`$ . Namely, choosing $`w(\varphi _r)=\mathrm{cos}\left(3\varphi _r\right)`$ we extract tensor gluon SPD from the unpolarized cross section,
$`{\displaystyle \frac{1}{2\pi }}{\displaystyle _0^{2\pi }}𝑑\varphi _r\mathrm{cos}\left(3\varphi _r\right){\displaystyle \frac{d^+\sigma d^{}\sigma }{d\varphi _r}}=16\sqrt{{\displaystyle \frac{\mathrm{\Delta }^2}{𝒬^2}}}{\displaystyle \frac{\sqrt{(1x)^3(1y)}}{xy(2x)^2}}\left(1{\displaystyle \frac{\mathrm{\Delta }_{\mathrm{min}}^2}{\mathrm{\Delta }^2}}\right)^{3/2}𝒯_2d,`$ (34)
where $`d=\frac{\alpha ^3xy}{8\pi 𝒬^2}\left(1+\frac{4M^2x}{𝒬^2}\right)^{1/2}dxdyd|\mathrm{\Delta }^2|`$ and $`\mathrm{\Delta }_{\mathrm{min}}^2=M^2x^2/(1x+xM^2/𝒬^2)`$. Note that the correction to the structure function $`𝒯_2`$ is suppressed by $`𝒪\left(\sqrt{\mathrm{\Delta }^2/𝒬^2}\right)`$, while the combination $`𝒯_1`$ is down by $`𝒪\left(M^2/\sqrt{\mathrm{\Delta }^2𝒬^2}\right)`$ relative to $`M^2𝒯_2`$.
The imaginary part of the amplitudes $`_G^T`$ and $`_G^T`$ can be accessed by means of single spin asymmetry. The polarization of the lepton beam does not induce contribution of gluon “transversity” distributions into the cross section in agreement with . However, once the proton beam is longitudinally polarized we can access this asymmetry due to the interference term since the latter has genuine $`\mathrm{sin}\left(3\varphi _r\right)`$ azimuthal angle dependence. Thus the SPD can be extracted via the following weighted cross section
$`{\displaystyle \frac{1}{2\pi }}{\displaystyle _0^{2\pi }}𝑑\varphi _r\mathrm{sin}\left(3\varphi _r\right){\displaystyle \frac{d^+\sigma _{}d^+\sigma _{}}{d\varphi _r}}=16\sqrt{{\displaystyle \frac{\mathrm{\Delta }^2}{𝒬^2}}}{\displaystyle \frac{\sqrt{(1x)^3(1y)}}{xy(2x)^2}}\left(1{\displaystyle \frac{\mathrm{\Delta }_{\mathrm{min}}^2}{\mathrm{\Delta }^2}}\right)^{3/2}\stackrel{~}{𝒯}d,`$ (35)
where in $`d^+\sigma _{}`$ the proton is polarized along the positron beam and
$$\stackrel{~}{𝒯}=\frac{1}{2M^2}\mathrm{Im}F_2\left(_G^T+\frac{x}{2}_G^T\right).$$
(36)
Let us stress that the results (34) and (35) are valid to $`1/\sqrt{𝒬^2}`$ accuracy and further expansion terms from other structure functions can mimic the $`\mathrm{cos}\left(3\varphi \right)/\mathrm{sin}\left(3\varphi \right)`$ behaviour and contaminate the double helicity flip cross sections.
## 6 Conclusion.
In this paper we have presented weighted real photon electroproduction cross sections which can be used as a probe for the magnitude of gluon content in the nucleon. The opportunity to extract the real and imaginary parts of these amplitudes is offered by the scattering of the unpolarized lepton beam on the unpolarized and longitudinally polarized nucleon targets, respectively. We have given as well the formulae for NLO coefficient function as well as two-loop anomalous dimensions of the conformal operators and momentum fraction exclusive kernels thus completing the set of results required for study of the scaling violation for all twist-two SPDs at NLO.
This work was supported by DFG and BMBF (D.M.).
|
warning/0005/hep-th0005221.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
Instantons are self-dual solutions of the pure Yang-Mills equations . For the classical groups the complete set of instanton solutions on $`\text{}^4`$ (and via stereographic projection $`S^4`$) have been known for over twenty years. Although even now some important details remain obscure. For example, what is the metric on the $`k`$-instanton moduli space for $`\text{}^4`$ instantons? This is an important ingredient in the instanton-theoretic checks of the Seiberg-Witten results in $`𝒩=2`$ supersymmetric Yang Mills theory. For other four manifolds even less is known. A particularly important manifold is the four torus $`\text{𝕋}^4`$. Firstly, it is compact, thereby removing from the outset, any infrared divergences. Unlike other compact four manifolds (e.g. $`S^4`$ or $`K3`$) the four torus retains translational invariance, and is flat. However, while $`\text{𝕋}^4`$ has all these attractive features the only known explicit $`\text{𝕋}^4`$ instanton solutions are some reducible constant curvature solutions due to ’t Hooft . These exist only for special values of the periods and can only represent singular points in the moduli space of a given instanton sector. The possibility that these constant curvature solutions are the only instantons on $`\text{𝕋}^4`$ was ruled out a long time ago by Taubes . However, using the Nahm transformation, it can be shown that there exist no untwisted instantons with unit topological charge on $`\text{𝕋}^4`$ . The work of Taubes established the existence of instantons in all higher topological charge sectors. A similar pattern is followed by the $`O(3)`$ sigma model instantons on $`\text{𝕋}^2`$ . Here the one instanton sector is empty, and this corresponds to the statement that there are no elliptic functions with a single simple pole in the fundamental torus.
How should one start to look for instanton solutions on $`\text{𝕋}^4`$? An obvious approach would be to adapt to the torus the techniques developed in the late 1970’s for the $`\text{}^4`$ problem. Loosely speaking, we seek periodic versions of these ansätze, since instantons on $`\text{𝕋}^4`$ can be viewed as periodic solutions<sup>1</sup><sup>1</sup>1 They can only be periodic in a singular gauge. on $`\text{}^4`$. The general solution to the instanton problem on $`\text{}^4`$ was provided by Atiyah, Drinfeld, Hitchin and Manin (ADHM) . This work reduces the problem of constructing instantons on $`\text{}^4`$ or $`\text{S}^4`$ to an exercise in algebra. To construct an instanton with topological charge $`k`$ one must find a quaternionic $`(k+1)\times k`$ matrix, $`M`$, obeying certain non-linear reality conditions. However, while this construction is purely algebraic, its structure is very much tied to the manifold $`\text{}^4`$ or $`S^4`$, and it appears difficult to ‘make it periodic’ in a simple way. An important subclass of solutions is provided by the ’t Hooft ansatz . This converts a (singular) positive solution of the Laplace equation into an $`SU(2)`$ instanton. Since this is a linear equation, it seems that we simply have to find a periodic solution of the Laplace equation to construct an instanton on the torus. However, it is not too difficult to show that it is impossible to construct a positive solution of the Laplace equation on $`\text{𝕋}^4`$ with acceptable singularities (i.e. singularities which do not show up in the Yang-Mills action density).
In this paper we render the ADHM construction periodic by ‘brute force’, in that we regard instantons on the torus as a periodic lattice of instantons on $`\text{}^4`$. We start with ADHM data corresponding to an infinite array of instantons embedded in $`\text{}^4`$. While our initial objective was to extract the $`\text{𝕋}^4`$ instantons, we will see that the less ambitious target to have periodicity in fewer than four directions offers considerable technical simplification. To that end we consider the application of the ADHM method to $`SU(2)`$ Yang-Mills on $`\text{𝕋}^n\times \text{}^{4n}`$ for $`n=1,2,3,4`$. Although $`\text{𝕋}^4`$ has no one instanton solution, $`S^1\times \text{}^3`$, $`\text{𝕋}^2\times \text{}^2`$ and $`\text{𝕋}^3\times \text{}`$ should have . Again the $`O(3)`$-sigma model provides a useful hint, since while there are no one-instantons on $`\text{𝕋}^2`$, one-instanton solutions have been constructed on $`S^1\times \text{}`$ . As the $`\text{}^4`$ topological charge of a $`\text{𝕋}^n\times \text{}^{4n}`$ instanton is infinite we have to deal with an infinite dimensional $`M`$ matrix. For the $`k`$-instanton problem on $`\text{𝕋}^n\times \text{}^{4n}`$, $`M`$ can be related to a $`U(k)`$ Weyl operator on $`\stackrel{~}{\text{𝕋}}^n`$, $`\stackrel{~}{\text{𝕋}}^n`$ being the torus dual to $`\text{𝕋}^n`$. This is a manifestation of the Nahm transformation .
Recently this programme has been implemented by Kraan and van Baal in the one-instanton sector of $`SU(N)`$ gauge theory on $`S^1\times \text{}^3`$ . Equivalent results were derived independently by Lee and Lu . These works revealed a vivid ‘monopole constituent’ picture of calorons (see also ). There is however an important pitfall in this whole approach; even if one has constructed a Weyl operator on $`\stackrel{~}{\text{𝕋}}^n`$ via the ADHM method one must check that it actually leads to a well defined gauge potential on $`\text{𝕋}^n\times \text{}^{4n}`$.<sup>2</sup><sup>2</sup>2 For $`n=1`$ the procedure always leads to a well defined instanton. Here we solve the ADHM constraints for the one instanton problem on $`\text{𝕋}^n\times \text{}^{4n}`$ and give particular solutions for the two instanton case. However, we are only able to explicitly check that these sometimes lead to a well defined gauge potential for $`n=2`$. This is because the technical task of solving the Weyl equation on $`\stackrel{~}{\text{𝕋}}^n`$ becomes more involved for higher $`n`$. We will see that the $`n=2`$ case (i.e. $`\text{𝕋}^2\times \text{}^2`$) boils down to a specific Aharonov Bohm problem <sup>3</sup><sup>3</sup>3 To our knowledge the extensive literature on the AB problem (see for example ) does not explicitly tackle this specific case. on $`\stackrel{~}{\text{𝕋}}^2`$. A stringy interpretation of $`\text{𝕋}^2\times \text{}^2`$ instantons can be found in . Our gauge potential on $`\text{𝕋}^2\times \text{}^2`$ is well defined only if we apply certain constraints on the ADHM parameters. In the one instanton sector there is an upper limit on the scale parameter. For our subclass of two instantons further constraints emerge. The two ‘component’ instantons must share a common scale parameter which itself is bounded from above. Furthermore, the relative group orientation of the two instantons is constrained.
The outline of this paper is as follows. In chapter 2 we briefly recall the standard ADHM construction on $`\text{}^4`$ and then explain in a general way how it can be ‘made periodic’ in one or more directions. In chapter 3 we solve the ADHM constraints for the one-instanton problem on $`\text{𝕋}^n\times \text{}^{4n}`$. The associated Weyl operator on $`\stackrel{~}{\text{𝕋}}^n`$ is given explicitly in terms of a specific Green’s function for the Laplace operator on $`\stackrel{~}{\text{𝕋}}^n`$. Then we specialise to $`\text{𝕋}^2\times \text{}^2`$, where the Weyl equations seem to be more manageable than in the general case. Finally in chapter 4 we discuss the two instanton problem. Some technical results are given in the appendices.
During the writing up of this paper we became aware of some related work by Jardim. In a series of papers a mathematically sophisticated analysis of the Nahm transformation on $`\text{𝕋}^2\times \text{}^2`$ has been given. A somewhat more physical account can be found in where the Jardim formalism is applied to periodic monopoles, i.e. instantons on $`S^1\times \text{}^2`$ so that the dual torus is $`\stackrel{~}{S}^1\times \text{}`$ instead of $`\stackrel{~}{\text{𝕋}}^2`$.
## 2 ADHM construction
In this chapter we review the standard ADHM construction on $`R^4`$. We then explain how the formalism can be extended to $`\text{𝕋}^n\times \text{}^{4n}`$. This is a straightforward extension of the $`S^1\times \text{}^3`$ formalism.
### 2.1 ADHM on $`\text{}^4`$
Closely following the presentation of Christ Weinberg and Stanton (see also ) we briefly recall the ADHM construction. For simplicity we specialise to the gauge group $`SU(2)`$. We wish to construct a self-dual $`SU(2)`$ Yang-Mills field $`A_\mu (x)`$ on $`\text{}^4`$ with topological charge or instanton number
$`\begin{array}{ccc}\hfill k={\displaystyle \frac{1}{16\pi ^2}}{\displaystyle _\text{}^4}d^4x\text{tr}\left(F_{\mu \nu }F_{\mu \nu }\right).& & \end{array}`$ (2.2)
Here the Yang-Mills field strength is
$`\begin{array}{ccc}\hfill F_{\mu \nu }=_\mu A_\nu _\nu A_\mu +[A_\mu ,A_\nu ],& & \end{array}`$ (2.4)
and the gauge field $`A_\mu `$ can be viewed as a $`2\times 2`$ anti-Hermitian traceless matrix. However, one can equally regard $`A_\mu `$ as being a purely imaginary quaternion. Recall that the space of quaternions has four generators $`\text{i}_\mu =(1,\widehat{i},\widehat{j},\widehat{k})`$ where the $`\widehat{i}`$, $`\widehat{j}`$, $`\widehat{k}`$ anticommute and satisfy
$`\begin{array}{ccc}\hfill \widehat{i}^2=\widehat{j}^2=\widehat{k}^2=1,\widehat{i}\widehat{j}\widehat{k}=1.& & \end{array}`$ (2.6)
The transition back to the standard Pauli matrix language can be made via the identifications $`\widehat{i}i\sigma _1`$, $`\widehat{j}i\sigma _2`$, $`\widehat{k}i\sigma _3`$. We will use $``$ to denote quaternionic conjugation (i.e. $`1^{}=1`$, $`\widehat{i}^{}=\widehat{i}`$, $`\widehat{j}^{}=\widehat{j}`$, $`\widehat{k}^{}=\widehat{k}`$). In the following $``$ should be understood as the transpose of the quaternionic conjugate.
The recipe for constructing a self-dual $`A_\mu `$ with instanton number $`k`$ is as follows. One simply has to construct a $`k+1\times k`$ quaternionic matrix $`M`$ with the following properties:
i) the $`k\times k`$ matrix $`M^{}M`$ is real.
ii) $`M`$ is linear in the quaternion $`xx_0+x_1\widehat{i}+x_2\widehat{j}+x_3\widehat{k}`$ formed from the four Euclidean coordinates.
The corresponding anti-hermitian self-dual gauge potential is given by
$`\begin{array}{ccc}\hfill A_\mu (x)=N^{}(x)_\mu N(x),& & \end{array}`$ (2.8)
where $`N(x)`$ is a $`k+1`$ component column vector satisfying
$`\begin{array}{ccc}\hfill M^{}N=0,\text{ and }N^{}N=1.& & \end{array}`$ (2.10)
Without loss of generality one may assume $`M`$ has the following form
$`\begin{array}{ccc}\hfill M=\left(\begin{array}{c}v\\ \widehat{M}\end{array}\right),& & \end{array}`$ (2.14)
where $`v`$ is a $`k`$-component row vector $`v`$ made up of $`k`$ constant quaternions
$`\begin{array}{ccc}\hfill v=(q_1q_2\mathrm{}q_k).& & \end{array}`$ (2.16)
These quaternions encode the scales and $`SU(2)`$ group orientation of the $`k`$ ‘component’ instantons. $`\widehat{M}`$ is a $`k\times k`$ matrix with the following ‘canonical’ form
$`\begin{array}{ccc}\hfill \widehat{M}_{ij}(x)=\delta _{ij}(y_ix)+b_{ij}.& & \end{array}`$ (2.18)
$`b_{ij}`$ is independent of $`x`$, symmetric and has no diagonal entries ($`b_{ij}=0`$ for $`i=j`$). The reality of $`M^{}M`$ translates into the following non-linear requirement on $`b_{ij}`$
$`\begin{array}{ccc}\hfill {\displaystyle \frac{1}{2}}(q_i^{}q_jq_j^{}q_i)+(y_iy_j)^{}b_{ij}+{\displaystyle \frac{1}{2}}{\displaystyle \underset{l=1}{\overset{k}{}}}\left(b_{li}^{}b_{lj}b_{lj}^{}b_{li}\right)=r_{ij},& & \end{array}`$ (2.20)
for some real $`k\times k`$ matrix $`r`$. The $`y_i`$ can be interpreted as the quaternionic positions of the instantons. One can immediately write down a column vector $`N`$ satisfying (2.10)
$`\begin{array}{ccc}\hfill N=\left(\begin{array}{c}{\displaystyle \frac{u}{\sqrt{\rho }}}\\ {\displaystyle \frac{1}{\sqrt{\rho }}}\left(\widehat{M}^{}\right)^1v^{}u\end{array}\right),& & \end{array}`$ (2.24)
and
$`\begin{array}{ccc}\hfill \rho =1+v\widehat{M}^1\left(\widehat{M}^{}\right)^1v^{}.& & \end{array}`$ (2.26)
Here $`u`$ is an arbitrary, possibly $`x`$-dependent unit quaternion; different choices for $`u`$ yield gauge equivalent Yang-Mills fields. Observe that it is necessary to invert the canonical form $`\widehat{M}`$ to extract the final gauge potential. In the singular gauge $`u(x)=1`$, the potential can be written,
$`\begin{array}{ccc}\hfill A_\mu ={\displaystyle \frac{1}{2\rho }}v(\widehat{M}^1_\mu \widehat{M}^{}{}_{}{}^{1}_\mu (\widehat{M}^1)\widehat{M}^{}{}_{}{}^{1})v^{}.& & \end{array}`$ (2.28)
The corresponding field strength reads
$`\begin{array}{ccc}\hfill F_{\mu \nu }={\displaystyle \frac{1}{\rho }}v\widehat{M}^1\text{i}_\mu f\text{i}_\nu ^{}(\widehat{M}^{})^1v^{}[\mu \nu ],& & \end{array}`$ (2.30)
where $`f`$ is the real $`k\times k`$ matrix
$`\begin{array}{ccc}\hfill f=(M^{}M)^1=\widehat{M}^1(\widehat{M}^{})^1{\displaystyle \frac{1}{\rho }}\widehat{M}^1(\widehat{M}^{})^1v^{}v\widehat{M}^1(\widehat{M}^{})^1.& & \end{array}`$ (2.32)
The reality of $`f`$ ensures that $`F_{\mu \nu }`$ is self-dual.
One immediately sees that $`A_\mu (x)`$ is unaffected by the following transformation on the ADHM data
$`\begin{array}{ccc}\hfill \widehat{M}O^1\widehat{M}O,vvO,& & \end{array}`$ (2.34)
where $`O`$ is a $`k\times k`$ real orthogonal matrix. Invoking this freedom one may argue that $`r_{ij}`$ can be set to zero . With this choice $`b_{ij}`$ is fully determined by the $`8k`$ parameters encoded in the $`q_i`$ and $`y_i`$. Three of these parameters correspond to the global gauge symmetry. This freedom can be fixed by taking $`q_1`$ to be real, leaving $`8k3`$ genuine moduli parameters. A trivial but useful consequence of the ‘symmetry’ (2.34) is that the $`q_i`$ are determined only up to a sign. If we flip the sign of one of the $`q_i`$, say $`q_3q_3`$, then this corresponds to the orthogonal transformation $`O=\text{diag}(1,1,1,1,1,\mathrm{}.)`$.
### 2.2 ADHM on $`\text{𝕋}^n\times \text{}^{4n}`$
We view $`\text{𝕋}^n`$ as $`\text{}^n`$ modulo a $`n`$ dimensional lattice $`\mathrm{\Lambda }`$ generated by $`n`$ quaternions $`e_0`$, $`e_1`$, … ,$`e_{n1}`$ corresponding to $`n`$ orthogonal vectors. The periods or equivalently the Euclidean lengths of the $`e_i`$ are denoted by $`L_i,i=0,1,\mathrm{},n1`$. First we will show how (in principle) one can produce instantons which in the singular gauge (i.e. $`u(x)=1`$ as in eqn. (2.28)) are periodic with respect to shifts by the lattice generators,
$`\begin{array}{ccc}\hfill A_\mu (x+e_i)=A_\mu (x),i=0,1,..,n1.& & \end{array}`$ (2.36)
Later we will consider a more general periodicity property which proved crucial in obtaining new 1-instanton solutions on $`S^1\times \text{}^3`$. To construct a k-instanton on $`\text{𝕋}^n\times \text{}^{4n}\text{}^4/\mathrm{\Lambda }`$ consider the following set up. For every $`\alpha \mathrm{\Lambda }`$ we have instantons at the positions $`y_i+\alpha `$ with respective scale/orientation quaternions $`q_i`$ where $`i=1,2,\mathrm{},k`$ enumerates the instantons in the fundamental cell. The quaternions $`y_i`$ give the instanton positions in the fundamental cell. Thus, our $`\widehat{M}`$ and $`v`$ now have the following structure
$`\begin{array}{ccc}\hfill v_i^\alpha =q_i,\widehat{M}_{ij}^{\alpha \beta }=\delta _{ij}\delta ^{\alpha \beta }(y_i+\alpha x)+b_{ij}^{\alpha \beta },i,j=1,2,\mathrm{},k,\alpha ,\beta \mathrm{\Lambda }.& & \end{array}`$ (2.38)
The matrix $`b_{ij}^{\alpha \beta }`$ has the properties
$`\begin{array}{ccc}\hfill b_{ij}^{\alpha \beta }=b_{ji}^{\beta \alpha },b_{ii}^{\alpha \alpha }=0\text{ (no sum)},& & \end{array}`$ (2.40)
and
$`\begin{array}{ccc}\hfill {\displaystyle \frac{1}{2}}(v_i^\alpha {}_{}{}^{}v_{j}^{\beta }v_j^\beta {}_{}{}^{}v_{i}^{\alpha })+(y_iy_j+\alpha \beta )^{}b_{ij}^{\alpha \beta }+{\displaystyle \frac{1}{2}}{\displaystyle \underset{l=1}{\overset{k}{}}}{\displaystyle \underset{\gamma \mathrm{\Lambda }}{}}\left(b_{li}^{\gamma \alpha }{}_{}{}^{}b_{lj}^{\gamma \beta }b_{lj}^{\gamma \beta }{}_{}{}^{}b_{li}^{\gamma \alpha }\right)=0.& & \end{array}`$ (2.42)
Now that $`\widehat{M}`$ is an infinite dimensional matrix the non-linear constraint appears much more formidable than its $`\text{}^4`$ counterpart (2.20). Moreover, even if we can solve the constraint we still face the problem of inverting $`\widehat{M}`$. We see that the constraint implies $`b_{ij}^{\alpha \beta }`$ has the following property
$`\begin{array}{ccc}\hfill \widehat{b}_{ij}^{\alpha \beta }=b_{ij}^{\alpha \beta \mathrm{\hspace{0.17em}\hspace{0.17em}\hspace{0.17em}0}},\alpha ,\beta \mathrm{\Lambda }.& & \end{array}`$ (2.44)
At this point it is useful to perform a Fourier transform ;
$`\begin{array}{ccc}\hfill v_i(z)={\displaystyle \underset{\alpha \mathrm{\Lambda }}{}}v_i^\alpha e^{i\alpha z},\widehat{M}_{ij}(z)\delta ^n(zz^{})={\displaystyle \underset{\alpha ,\beta \mathrm{\Lambda }}{}}\widehat{M}_{ij}^{\alpha \beta }e^{i\alpha zi\beta z^{}},& & \end{array}`$ (2.46)
where $`\delta ^n(zz^{})`$ is a $`n`$-dimensional delta function which is periodic with respect to the dual lattice
$`\begin{array}{ccc}\hfill \stackrel{~}{\mathrm{\Lambda }}=\{z\text{}^n|(2\pi )^1z\alpha \text{}\text{ for all }\alpha \mathrm{\Lambda }\}.& & \end{array}`$ (2.48)
Here $`\alpha z`$ denotes the usual scalar product in $`\text{}^n`$, i.e. $`\alpha z=_{j=0}^{n1}\alpha _jz_j`$. The delta function has the Fourier representation
$`\begin{array}{ccc}\hfill \delta ^n(z)={\displaystyle \frac{1}{\stackrel{~}{𝒱}}}{\displaystyle \underset{\alpha \mathrm{\Lambda }}{}}e^{i\alpha z},& & \end{array}`$ (2.50)
where
$`\begin{array}{ccc}\hfill \stackrel{~}{𝒱}=(2\pi )^n/L_0L_1\mathrm{}L_{n1},& & \end{array}`$ (2.52)
is the volume of the dual torus $`\stackrel{~}{\text{𝕋}}^n:=\text{}^n/\stackrel{~}{\mathrm{\Lambda }}`$. Using (2.38) $`\widehat{M}_{ij}`$ can be written as follows
$`\begin{array}{c}\stackrel{~}{𝒱}^1\widehat{M}_{ij}(z)=\delta _{ij}\left(id_zx+{\displaystyle \frac{1}{k}}{\displaystyle \underset{l=1}{\overset{k}{}}}y_l\right)i\widehat{A}_{ij}(z),d_z={\displaystyle \underset{i=0}{\overset{n1}{}}}\text{i}_i_{z_i},\end{array}`$ (2.54)
and
$`\begin{array}{ccc}\hfill i\widehat{A}_{ij}(z)=\delta _{ij}\left(y_i{\displaystyle \frac{1}{k}}{\displaystyle \underset{l=1}{\overset{k}{}}}y_l\right)+{\displaystyle \underset{\alpha \mathrm{\Lambda }}{}}b_{ij}^{\alpha 0}e^{i\alpha z},& & \end{array}`$ (2.56)
can be regarded as a $`SU(k)`$ ($`U(1)`$ for $`k=1`$) potential on the dual torus $`\stackrel{~}{\text{𝕋}}^n`$. From now on we will assume (without loss of generality) that
$`\begin{array}{ccc}\hfill {\displaystyle \underset{l=1}{\overset{k}{}}}y_l=0,& & \end{array}`$ (2.58)
so that $`\stackrel{~}{𝒱}^1\widehat{M}_{ij}(z)=\delta _{ij}(id_zx)i\widehat{A}_{ij}(z)`$. The $`z`$-space analogue of $`M`$ can be written as
$`\begin{array}{ccc}\hfill M=\left(\begin{array}{c}v_i(z^{})\\ \widehat{M}_{ij}(z)\delta ^n(zz^{})\end{array}\right).& & \end{array}`$ (2.62)
We also require $`M^{}`$
$`\begin{array}{ccc}\hfill M^{}=\left(\begin{array}{cc}(v^{})_i(z)& (\widehat{M}^{})_{ij}(z)\delta ^n(zz^{})\end{array}\right),& & \end{array}`$ (2.65)
where
$`\begin{array}{ccc}\hfill (v^{})_i(z)={\displaystyle \underset{\alpha \mathrm{\Lambda }}{}}\left(v_i^\alpha \right)^{}e^{i\alpha z},(\widehat{M}^{})_{ij}(z)\delta ^n(zz^{})={\displaystyle \underset{\alpha ,\beta \mathrm{\Lambda }}{}}\left(M_{ji}^{\beta \alpha }\right)^{}e^{i\alpha zi\beta z^{}},& & \end{array}`$ (2.67)
so that $`\stackrel{~}{𝒱}^1\widehat{M}_{ij}^{}(z)=\delta _{ij}(id_z^{}x^{})i\widehat{A}_{ij}^{}(z)`$. We now consider the product $`M^{}M`$
$`(M^{}M)_{ij}(z,z^{})`$ $`=`$ $`(v^{})_i(z)v_j(z^{})+\stackrel{~}{𝒱}^1{\displaystyle _{\stackrel{~}{\text{𝕋}}^n}}d^nw(\widehat{M}^{})_{ik}(z)\delta ^n(zw)\widehat{M}_{kj}(w)\delta ^n(wz^{})`$ (2.68)
$`=`$ $`(v^{})_i(z)v_j(z^{})`$
$`+\stackrel{~}{𝒱}^2\left(\delta _{ik}(id_z^{}x^{})i\widehat{A}_{ik}^{}(z)\right)\left(\delta _{kj}(id_zx)i\widehat{A}_{kj}(z)\right)\delta (zz^{}).`$
In $`z`$-space the constraint that $`M^{}M`$ is real reduces to the self-duality equation for the $`SU(k)`$ $`(`$ or $`U(1)`$ $`)`$ potential $`\widehat{A}_{ij}(z)`$, but with delta function sources. These sources come from the $`(v^{})_i(z)v_j(z^{})`$ term; with the choice (2.38) we have $`v_i(z)=\stackrel{~}{𝒱}q_i\delta ^n(z)`$.
It is also possible to arrange so that in the singular gauge $`u(x)=1`$, $`A_\mu (x)`$ is periodic modulo global gauge transformations. This is achieved by replacing $`v_i^\alpha =q_i`$ with
$`\begin{array}{ccc}\hfill v_i^\alpha =e^{(\alpha \omega )\widehat{l}}q_i,& & \end{array}`$ (2.70)
where $`\omega `$ is an element of the dual torus and $`\widehat{l}`$ is a purely imaginary unit quaternion. In the $`u(x)=1`$ gauge, the instanton potential has the following periodicity properties
$`\begin{array}{ccc}\hfill A_\mu (x+e_i)=e^{(e_i\omega )\widehat{l}}A_\mu (x)e^{(e_i\omega )\widehat{l}}.& & \end{array}`$ (2.72)
This choice of $`v_i^\alpha `$ still entails delta function sources on the dual torus
$`\begin{array}{ccc}\hfill v_i(z)=\frac{1}{2}\stackrel{~}{𝒱}\left[\left(1i\widehat{l}\right)\delta ^n(z\omega )+\left(1+i\widehat{l}\right)\delta ^n(z+\omega )\right]q_i.& & \end{array}`$ (2.74)
$`\left(1+i\widehat{l}\right)`$ and $`\left(1i\widehat{l}\right)`$ are projectors in the sense that
$`\begin{array}{ccc}\hfill \left(1\pm i\widehat{l}\right)^2=2\left(1\pm i\widehat{l}\right),\left(1+i\widehat{l}\right)\left(1i\widehat{l}\right)=0.& & \end{array}`$ (2.76)
Looking at the expression (2.28) for the $`\text{}^4`$ gauge potential we see that it suffices to compute the $`k`$-component row vector $`n:=v\widehat{M}^1`$. The $`\text{𝕋}^n\times \text{}^{4n}`$ analogue of this object is the $`z`$-dependent $`k`$-component row vector, $`n(z)`$, with components
$`\begin{array}{ccc}\hfill n_j(z)=\stackrel{~}{𝒱}^1{\displaystyle \underset{i}{}}{\displaystyle _{\stackrel{~}{\text{𝕋}}^n}}d^nz^{}v_i(z^{})\widehat{M}_{ij}^1(z^{},z),& & \end{array}`$ (2.78)
and similarly the $`k`$-component column vector $`n^{}(z)`$ has components
$`(n^{})_i(z)=\stackrel{~}{𝒱}^1_j_{\stackrel{~}{\text{𝕋}}^n}d^nz^{}(\widehat{M}^{})_{ij}^1(z,z^{})(v^{})_j(z^{})`$. Here
$`\widehat{M}_{ij}^1(z,z^{})=_{\alpha ,\beta }\left(\widehat{M}^1\right)_{ij}^{\alpha \beta }e^{i\alpha zi\beta z^{}}`$, so that
$$\widehat{M}(z)\widehat{M}^1(z,z^{})=\stackrel{~}{𝒱}^2\delta ^n(zz^{}).$$
(2.79)
Using (2.74) we have
$`\begin{array}{ccc}\hfill n_j(z)=\frac{1}{2}\left(1i\widehat{l}\right)q_i\widehat{M}_{ij}^1(\omega ,z)+\frac{1}{2}\left(1+i\widehat{l}\right)q_i\widehat{M}_{ij}^1(\omega ,z),& & \end{array}`$ (2.81)
which reduces to $`n_j(z)=q_i\widehat{M}_{ij}^1(0,z)`$ in the periodic case ($`\omega =0`$). The $`\text{𝕋}^n\times \text{}^{4n}`$ gauge potential can be written
$`\begin{array}{c}A_\mu ={\displaystyle \frac{\stackrel{~}{𝒱}^1}{2\rho }}{\displaystyle _{\stackrel{~}{\text{𝕋}}^n}}d^nz\left[n(z)_\mu n^{}(z)_\mu (n(z))n^{}(z)\right],\end{array}`$ (2.83)
where $`\rho `$ is now
$`\begin{array}{ccc}\hfill \rho =1+\stackrel{~}{𝒱}^1{\displaystyle _{\stackrel{~}{\text{𝕋}}^n}}d^nzn(z)n^{}(z).& & \end{array}`$ (2.85)
Note that the integrand, $`n(z)n^{}(z)`$ in (2.85) is not necessarily real, although the integral itself, $`d^nzn(z)n^{}(z)`$, is real and positive (see section 3.2).
The corresponding field strength is
$`\begin{array}{ccc}\hfill F_{\mu \nu }={\displaystyle \frac{𝒱^2}{\rho }}{\displaystyle _{\stackrel{~}{\text{𝕋}}^n}}d^nz{\displaystyle _{\stackrel{~}{\text{𝕋}}^n}}d^nz^{}n(z)\text{i}_\mu f(z,z^{})\text{i}_\nu ^{}n^{}(z^{})[\mu \nu ],& & \end{array}`$ (2.87)
where the Green’s function $`f(z,z^{})`$ is
$`f(z,z^{})`$ $`=`$ $`(M^{}M)^1(z,z^{})`$
$`=`$ $`\stackrel{~}{𝒱}^1{\displaystyle _{\stackrel{~}{\text{𝕋}}^n}}d^ny\widehat{M}^1(z,y)(\widehat{M}^{})^1(y,z^{})`$
$`{\displaystyle \frac{\stackrel{~}{𝒱}^2}{\rho }}{\displaystyle _{\stackrel{~}{\text{𝕋}}^n}}d^ny\widehat{M}^1(z,y)n^{}(y){\displaystyle _{\stackrel{~}{\text{𝕋}}^n}}d^ny^{}n(y^{})(\widehat{M}^{})^1(y^{},z^{}).`$
As we shall see, all the formulae in this section require particularly careful handling for $`n>1`$.
## 3 One-instantons
In this chapter we consider in some detail the one instanton problem on $`\text{𝕋}^n\times \text{}^{4n}`$. In particular we explicitly determine the ADHM matrix $`M`$. Under the Fourier transform this becomes a Weyl operator associated with an Abelian self-dual potential $`\widehat{A}(z)`$ on the dual torus $`\stackrel{~}{\text{𝕋}}^n`$. Unfortunately we do not have a general approach to the solution of such Weyl equations. In section 3.2 we concentrate our attention on the $`\stackrel{~}{\text{𝕋}}^2`$ Weyl equation (corresponding to one instantons on $`\text{𝕋}^2\times \text{}^2`$) where $`\widehat{A}(z)`$ is an Aharonov Bohm potential on $`\stackrel{~}{\text{𝕋}}^2`$. The ADHM construction of the instanton potential $`A_\mu (x)`$ and $`F_{\mu \nu }(x)`$ is considered. For values of $`x`$ restricted to a two dimensional subspace of $`\text{𝕋}^2\times \text{}^2`$ closed forms for $`A_\mu (x)`$ and $`F_{\mu \nu }`$ are given. From a mathematical standpoint the calculation is not completely satisfactory; a formal limiting procedure is employed to obtain the gauge potential. However, we are able to check that the field strength is self-dual and that $`\mathrm{tr}(F_{\mu \nu })^2`$ is non-zero and smooth. Moreover, in section 3.3 we see that our potential can be interpreted as the Nahm transform of the AB potential $`\widehat{A}(z)`$. More specifically, we identify the two Nahm zero modes associated with $`\widehat{A}(z)`$.
### 3.1 ADHM constraints for $`\text{𝕋}^n\times \text{}^{4n}`$
Let us start by considering $`1`$-instanton solutions on $`\text{𝕋}^n\times \text{}^{4n}`$. If we seek instantons which are strictly periodic in the $`u(x)=1`$ gauge we are immediately restricted to $`S^1\times \text{}^3`$. This is because all the instantons in our lattice will, by construction, have the same scale/group orientation $`q_1`$ and hence be of the ’t Hooft type. Since the ’t Hooft instantons on $`S^1\times \text{}^3`$ are well known we will examine the more general instanton array (2.70).
Without loss of generality we can assume that $`q_1`$ is a real quaternion which we identify as the ‘scale’ $`\lambda `$, so that
$`\begin{array}{ccc}\hfill v^\alpha =e^{(\alpha \omega )\widehat{l}}\lambda ,& & \end{array}`$ (3.2)
where we have dropped the redundant $`1`$ subscript on $`v^\alpha `$. The $`\widehat{M}`$ matrix has the form
$`\begin{array}{ccc}\hfill \widehat{M}^{\alpha \beta }=\delta ^{\alpha \beta }(\alpha x)+b^{\alpha \beta }.& & \end{array}`$ (3.4)
We now have to determine the $`b`$ matrix via (2.42). Under the Fourier transformation this is a self-duality equation on the dual torus $`\stackrel{~}{\text{𝕋}}^n`$. However, it is instructive to examine the constraint equation in the original (matrix) variables. In Appendix A we will argue that for $`k=1`$ the quadratic term in (2.42) is zero, i.e. the $`b`$ matrix is simply
$`\begin{array}{ccc}\hfill b^{\alpha \beta }={\displaystyle \frac{1}{2(\alpha \beta )^{}}}\left(v_{}^{\alpha }{}_{}{}^{}v^\beta v_{}^{\beta }{}_{}{}^{}v^\alpha \right)={\displaystyle \frac{\lambda ^2}{(\alpha \beta )^{}}}\widehat{l}\mathrm{sin}\left[(\alpha \beta )\omega \right],\alpha \beta .& & \end{array}`$ (3.6)
In order to construct the potential we must now invert the $`\widehat{M}`$ matrix. To facilitate this we perform the Fourier transform elaborated in section 2.2,
$`\begin{array}{ccc}\hfill \stackrel{~}{𝒱}^1\widehat{M}(z)=id_zxi\widehat{A}(z),& & \end{array}`$ (3.8)
where $`\widehat{A}(z)`$ is the $`U(1)`$ potential
$`\begin{array}{ccc}\hfill \widehat{A}(z)=i\lambda ^2d_z\varphi (z)\widehat{l},& & \end{array}`$ (3.10)
and $`\varphi `$ is the real function
$`\begin{array}{ccc}\hfill \varphi (z)={\displaystyle \frac{1}{2}}{\displaystyle \underset{\alpha \mathrm{\Lambda }0}{}}{\displaystyle \frac{\mathrm{exp}[i\alpha (z+\omega )]\mathrm{exp}[i\alpha (z\omega )]}{|\alpha |^2}},& & \end{array}`$ (3.12)
which is a Green’s function for the Laplace operator on $`\stackrel{~}{\text{𝕋}}^n`$
$`\begin{array}{ccc}\hfill d_zd_z^{}\varphi (z)={\displaystyle \frac{\stackrel{~}{𝒱}}{2}}\left[\delta ^n(z+\omega )\delta ^n(z\omega )\right].& & \end{array}`$ (3.14)
Clearly $`\varphi (z)`$ is an odd function
$$\varphi (z)=\varphi (z).$$
(3.15)
Writing $`\widehat{A}(z)=_{l=0}^{n1}\text{i}_l\widehat{A}_l(z)`$, one can check that the Abelian field strength $`\widehat{F}_{ij}(z)=_i\widehat{A}_j_j\widehat{A}_i`$ is self-dual, except at the singularities $`z=\pm \omega `$.
### 3.2 One-instantons on $`\text{𝕋}^2\times \text{}^2`$
Since our lattice is two dimensional we may take $`e_0`$ to be real and $`e_1`$ to be proportional to the purely imaginary unit quaternion $`\widehat{l}`$ <sup>4</sup><sup>4</sup>4 We can always perform an $`O(4)`$ Lorentz transformation to arrange this.. Now rewrite the quaternion $`z`$ as follows
$$z=z_0+\widehat{l}z_1=\frac{1}{2}\left(1i\widehat{l}\right)\text{z}+\frac{1}{2}\left(1+i\widehat{l}\right)\overline{\text{z}},$$
(3.16)
where $`\text{z}=z_0+iz_1`$, $`\overline{\text{z}}=z_0iz_1`$ denote standard complex coordinates. We can write the Fourier transformed $`\widehat{M}`$ as follows
$`\begin{array}{ccc}\hfill \stackrel{~}{𝒱}^1\widehat{M}(z)=id_zxi\widehat{A}_0(z)i\widehat{l}\widehat{A}_1(z),& & \end{array}`$ (3.18)
where
$`\begin{array}{ccc}\hfill \widehat{A}_0=i\lambda ^2_{z_1}\varphi ,\widehat{A}_1=i\lambda ^2_{z_0}\varphi ,& & \end{array}`$ (3.20)
and $`\varphi `$ is the Green’s function defined by (3.12). Since we are on $`\stackrel{~}{\text{𝕋}}^2`$ we can write $`\varphi `$ directly in terms of Jacobi theta functions<sup>5</sup><sup>5</sup>5 We follow the notation of Mumford ; $`\theta (\text{z},\tau )=_{n=\mathrm{}}^{\mathrm{}}e^{\pi in^2\tau +2\pi in\text{z}}`$. In the fundamental torus $`\theta (z,\tau )`$ has a single zero at $`z=\frac{1}{2}+\frac{1}{2}\tau `$, and has the periodicity properties $`\theta (\text{z}+1,\tau )=\theta (\text{z},\tau ),\theta (\text{z}+\tau ,\tau )=e^{\pi i\tau 2\pi i\text{z}}\theta (\text{z},\tau )`$.
$`\begin{array}{ccc}\hfill \varphi (\text{z})& =& {\displaystyle \frac{\stackrel{~}{𝒱}}{8\pi }}\mathrm{log}{\displaystyle \frac{\left|\theta (\frac{L_0}{2\pi }(\text{z}+\text{w})+\frac{1}{2}+\frac{iL_0}{2L_1},\frac{iL_0}{L_1})\right|^2}{\left|\theta (\frac{L_0}{2\pi }(\text{z}\text{w})+\frac{1}{2}+\frac{iL_0}{2L_1},\frac{iL_0}{L_1})\right|^2}}+{\displaystyle \frac{(\text{z}\overline{\text{z}})(\text{w}\overline{\text{w}})}{4}}i{\displaystyle \frac{\mathrm{w}\overline{\mathrm{w}}}{4L_1}},\hfill \end{array}`$ (3.22)
where $`\text{w}=\omega _0+i\omega _1`$, $`\overline{\text{w}}=\omega _0i\omega _1`$. The associated field strength is given by $`\widehat{F}_{01}=i\lambda ^2\mathrm{}\varphi `$, which is zero except at $`z=\pm \omega `$. At the points $`\omega +\stackrel{~}{\alpha },\stackrel{~}{\alpha }\stackrel{~}{\mathrm{\Lambda }}`$ we have a ‘flux tube’ of strength $`\frac{1}{2}\lambda ^2\stackrel{~}{𝒱}`$, and at the points $`\omega +\stackrel{~}{\alpha },\stackrel{~}{\alpha }\stackrel{~}{\mathrm{\Lambda }}`$ we have flux tubes of strength $`\frac{1}{2}\lambda ^2\stackrel{~}{𝒱}`$.
What about the $`x`$ term in (3.18)? It will prove convenient to decompose $`x`$ into two pieces
$`\begin{array}{ccc}\hfill x=x_{||}+x_{},& & \end{array}`$ (3.24)
where $`x_{||}`$ and $`x_{}`$ respectively commute and anticommute with $`\widehat{l}`$. Therefore the $`x_{||}`$ contribution just amounts to shifting $`\widehat{A}_0`$ and $`\widehat{A}_1`$ by constants, while $`x_{}`$ is akin to a mass term.
We can write $`\widehat{M}(z)`$ as follows
$`\begin{array}{ccc}\hfill \stackrel{~}{𝒱}^1\widehat{M}(z)=e^{i\widehat{l}\lambda ^2\varphi (z)}\left(id_zx_{||}\right)e^{i\widehat{l}\lambda ^2\varphi (z)}x_{}.& & \end{array}`$ (3.26)
This is not a pure gauge decomposition since the argument of the exponential is not a pure phase. If $`x_{}=0`$, one can immediately write down a formal inverse for $`\widehat{M}`$
$`\begin{array}{ccc}\hfill \widehat{M}^1(z,z^{})=\stackrel{~}{𝒱}e^{i\widehat{l}\lambda ^2\varphi (z)}G(zz^{})e^{i\widehat{l}\lambda ^2\varphi (z^{})},& & \end{array}`$ (3.28)
where $`G(zz^{})`$ is the periodic free Green’s function defined by<sup>6</sup><sup>6</sup>6 This Green’s function exists for $`x_{||}\mathrm{\Lambda }`$.
$`\begin{array}{ccc}\hfill \left(id_zx_{||}\right)G(zz^{})=\delta ^2(zz^{}),& & \end{array}`$ (3.30)
and has the Fourier series representation
$`\begin{array}{ccc}\hfill G(zz^{})=\stackrel{~}{𝒱}^1{\displaystyle \underset{\alpha \mathrm{\Lambda }}{}}{\displaystyle \frac{e^{i\alpha (zz^{})}}{\alpha x_{||}}}.& & \end{array}`$ (3.32)
The inverse (3.28) obviously satisfies $`\widehat{M}(z)\widehat{M}^1(z,z^{})=\stackrel{~}{𝒱}^2\delta ^2(zz^{})`$ for $`z\pm \omega `$. However, due to the singularities at $`z=\pm \omega `$ some caution is called for when interpreting (3.28) as the inverse of $`\widehat{M}`$. We will return to this point in the next section. For now we will stick with (3.28). $`G(z)`$ can be decomposed as follows
$`\begin{array}{ccc}\hfill G(z)=\frac{1}{2}\left(1i\widehat{l}\right)G_{}(z)+\frac{1}{2}\left(1+i\widehat{l}\right)G_+(z),& & \end{array}`$ (3.34)
where $`G_\pm (z)`$ are the following standard (i.e. complex rather than quaternionic) free Green’s functions
$`\begin{array}{ccc}\hfill \left(i_\text{z}\frac{1}{2}\overline{\text{x}}_{||}\right)G_+(z)=\frac{1}{2}\delta ^2(z),\left(i_{\overline{\text{z}}}\frac{1}{2}\text{x}_{||}\right)G_{}(z)=\frac{1}{2}\delta ^2(z).& & \end{array}`$ (3.36)
Here $`_\text{z}=\frac{1}{2}(_{z_0}i_{z_1})`$, $`\text{x}_{||}=(x_{||})_0+i(x_{||})_1`$ and the bar denotes complex conjugation. Evidently
$`\begin{array}{ccc}\hfill G_+(z)=\overline{G_{}(z)}.& & \end{array}`$ (3.38)
Now that we have the inverse of $`\widehat{M}`$ (at least for $`x_{}=0`$) let us start the computation of the gauge potential $`A_\mu (x)`$. As was emphasized in the introduction it is not guaranteed that $`A_\mu (x)`$ actually exists. We begin by considering $`\rho (x)`$ for our putative one-instanton. Inserting (3.28) into (2.81) yields
$`n(z)`$ $`=`$ $`{\displaystyle \frac{\lambda \stackrel{~}{𝒱}}{2}}[(1i\widehat{l})e^{\lambda ^2\left(\varphi (\omega )\varphi (z)\right)}G_{}(\omega z)`$
$`+(1+i\widehat{l})e^{\lambda ^2\left(\varphi (\omega )\varphi (z)\right)}G_+(\omega z)].`$
We now appear to be in trouble; $`\varphi (z)\pm \mathrm{}`$ as $`z\pm \omega `$, and so $`n(z)`$ is proportional to the ‘infinite’ constant $`e^{\lambda ^2\varphi (\omega )}`$. Thus it appears that our use of the inverse (3.28) was indeed unwarranted. Note that this problem is absent on $`S^1\times \text{}^3`$; while the derivative of $`\varphi (z)`$ is discontinuous at $`z=\pm \omega `$, $`\varphi (\pm \omega )`$ is well defined. For now we will proceed formally and treat $`\varphi (\omega )=\varphi (\omega )`$ as if it were a finite constant. The integrand in (2.85) is
$`n(z)n^{}(z)`$ $`=`$ $`{\displaystyle \frac{\lambda ^2\stackrel{~}{𝒱}^2e^{2\lambda ^2\varphi (\omega )}}{2}}[(1i\widehat{l})e^{2\lambda ^2\varphi (z)}|G_{}(\omega z)|^2`$
$`+(1+i\widehat{l})e^{2\lambda ^2\varphi (z)}|G_+(\omega z)|^2].`$
Here $`n^{}(z)=n^{}(z).`$ Clearly the integrand (3.2) has singularities over and above the questionable $`e^{2\lambda ^2\varphi (\omega )}`$ factor. We also note that $`n(z)n^{}(z)`$ is not real. Now we will argue that these singularities are integrable provided
$`\begin{array}{ccc}\hfill 0<\lambda ^2\stackrel{~}{𝒱}<4\pi .& & \end{array}`$ (3.42)
In the neighbourhood of $`z=\omega `$ we have the following singularity profile
$`\begin{array}{ccc}\hfill |G_{}(\omega z)|^2{\displaystyle \frac{1}{|\text{z}\text{w}|^2}},|G_+(\omega z)|^2\text{non-singular}.& & \end{array}`$ (3.44)
$`|G_{}(\omega z)|^2`$ has a non-integrable singularity at $`z=\omega `$. However, we must also consider the behaviour of $`\varphi (z)`$ at $`z=\omega `$
$`\begin{array}{ccc}\hfill \varphi (z){\displaystyle \frac{\stackrel{~}{𝒱}}{4\pi }}\mathrm{log}|\text{z}\text{w}|.& & \end{array}`$ (3.46)
Near $`z=\omega `$ we have
$`\begin{array}{ccc}\hfill |G_{}(\omega z)|^2e^{2\lambda ^2\varphi (z)}|\text{z}\text{w}|^{2+\lambda ^2\stackrel{~}{𝒱}/(2\pi )}.& & \end{array}`$ (3.48)
This singularity is integrable for $`\lambda ^2>0`$. In fact if we take $`\lambda ^2\stackrel{~}{𝒱}4\pi `$ the singularity disappears. However, then $`|G_{}(\omega z)|^2e^{2\lambda ^2\varphi (z)}`$ will not be integrable at $`z=\omega `$. Accordingly, for integrability at both $`z=\omega `$ and $`z=\omega `$ we must impose (3.42).
The bound (3.42) is nothing but the statement that $`\lambda ^2`$, the square of the ADHM size parameter, should not exceed the volume of the two-torus $`\text{𝕋}^2`$. Looking at the Abelian $`U(1)`$ potential $`\widehat{A}(z)`$ the bound is quite natural. Given that its associated field strength is zero away from the fluxes one can formally write it as a pure gauge, i.e. $`\widehat{A}_i(z)=_{z_i}\chi (z)`$. $`\chi (z)`$ is of course singular at the fluxes, but for $`0<\lambda ^2\stackrel{~}{𝒱}<4\pi `$ has a branch cut joining the two fluxes. At the critical value $`\lambda ^2\stackrel{~}{𝒱}=4\pi `$ the branch cut disappears, i.e. $`\chi `$ is single-valued on $`\stackrel{~}{\text{𝕋}}^2`$. Then $`\widehat{A}(z)`$ is truly a pure gauge and hence physically indistinguishable from the $`\lambda ^2\stackrel{~}{𝒱}=0`$ case.
Let us now return to the problem of the infinite constant $`e^{\lambda ^2\varphi (\omega )}`$ which seems to render our instanton meaningless. Define a ‘finite’ $`n`$ as follows
$`\begin{array}{ccc}\hfill \lambda \stackrel{~}{𝒱}n_f(z):=e^{\lambda ^2\varphi (\omega )}n(z).& & \end{array}`$ (3.50)
For $`x_{}=0`$ we have $`n_f(z)=\frac{1}{2}\left(1i\widehat{l}\right)e^{\lambda ^2\varphi (z)}G_{}(\omega z)+\frac{1}{2}\left(1+i\widehat{l}\right)e^{\lambda ^2\varphi (z)}G_+(\omega z)`$, which is finite except at the fluxes $`z=\pm \omega `$. The gauge potential can be written
$$A_\mu (x)=\frac{_{\stackrel{~}{\text{𝕋}}^2}d^2z\left[n_f(z)_\mu n_f^{}(z)_\mu \left(n_f(z)\right)n_f^{}(z)\right]}{2\left(e^{2\lambda ^2\varphi (\omega )}\lambda ^2\stackrel{~}{𝒱}^1+_{\stackrel{~}{\text{𝕋}}^2}d^2zn_f(z)n_f^{}(z)\right)},$$
(3.51)
where the $`_\mu `$ derivative is with respect to $`x_\mu `$. The only remnant of the infinite constant is the $`e^{2\lambda ^2\varphi (\omega )}`$ term in the denominator of (3.51); this exponential can be interpreted as ‘zero’, i.e. for our final potential we should take
$$A_\mu (x)=\frac{_{\stackrel{~}{\text{𝕋}}^2}d^2z\left[n_f(z)_\mu n_f^{}(z)_\mu \left(n_f(z)\right)n_f^{}(z)\right]}{2\rho _f(x)},$$
(3.52)
where
$`\begin{array}{ccc}\hfill \rho _f(x)={\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2zn_f(z)n_f^{}(z).& & \end{array}`$ (3.54)
Although $`n_f(z)n_f^{}(z)`$ is not real a short calculation suffices to express $`\rho _f`$ in a manifestly real and positive form (here we use that $`\varphi (z)`$ is an odd function, i.e. equation (3.15))
$`\begin{array}{ccc}\hfill \rho _f(x_{||})={\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2ze^{2\lambda ^2\varphi (z)}|G_{}(\omega z)|^2.& & \end{array}`$ (3.56)
So finally, the role of the infinite constant is simply to expunge the $`1`$ from the definition of $`\rho `$. Without the $`1`$ the infinite constant simply drops out of the final potential $`A_\mu (x)`$. This is in sharp contrast to the situation on $`S^1\times \text{}^3`$, where the 1 term must be kept since $`\varphi (\omega )`$ is a finite constant.
While (3.52) represents the final gauge potential we have only given $`n_f(z)`$ and $`\rho _f`$ explicitly for the special case $`x_{}=0`$. To construct $`n_f(z)`$ for $`x_{}0`$ is non-trivial. If we try to bring the $`x_{}`$ inside the bracket of equation (3.26) we get
$`\begin{array}{ccc}\hfill \stackrel{~}{𝒱}^1\widehat{M}(z)=e^{i\widehat{l}\lambda ^2\varphi (z)}\left(id_zx_{||}x_{}e^{2i\widehat{l}\lambda ^2\varphi (z)}\right)e^{i\widehat{l}\lambda ^2\varphi (z)}.& & \end{array}`$ (3.58)
Proceeding as in the $`x_{}=0`$ case we can write the inverse as follows
$`\begin{array}{ccc}\hfill \widehat{M}^1(z,z^{})=\stackrel{~}{𝒱}e^{i\widehat{l}\lambda ^2\varphi (z)}\stackrel{~}{G}(z,z^{})e^{i\widehat{l}\lambda ^2\varphi (z^{})},& & \end{array}`$ (3.60)
where $`\stackrel{~}{G}(z,z^{})`$ is no longer a free Green’s function
$`\begin{array}{ccc}\hfill \left(id_zx_{||}x_{}e^{2i\widehat{l}\lambda ^2\varphi (z)}\right)\stackrel{~}{G}(z,z^{})=\delta ^2(zz^{}).& & \end{array}`$ (3.62)
Inserting (3.60) into (3.50) yields
$`\begin{array}{ccc}\hfill n_f(z)={\displaystyle \frac{1}{2}}\left[\left(1i\widehat{l}\right)\stackrel{~}{G}(\omega ,z)+\left(1+i\widehat{l}\right)\stackrel{~}{G}(\omega ,z)\right]e^{i\widehat{l}\lambda ^2\varphi (z)}.& & \end{array}`$ (3.64)
A more detailed discussion of the properties of $`n_f`$ for $`x_{}0`$ will be given elsewhere.
The field strength derived from (3.52) is
$`\begin{array}{ccc}\hfill F_{\mu \nu }={\displaystyle \frac{\stackrel{~}{𝒱}^1}{\rho _f(x)}}{\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2z{\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2z^{}n_f(z)\text{i}_\mu f(z,z^{})\text{i}_\nu ^{}n_f^{}(z^{})[\mu \nu ],& & \end{array}`$ (3.66)
where $`f(z,z^{})`$ is
$`f(z,z^{})`$ $`=`$ $`\stackrel{~}{𝒱}^1{\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2y\widehat{M}^1(z,y)(\widehat{M}^{})^1(y,z^{})`$
$`{\displaystyle \frac{\stackrel{~}{𝒱}^1}{\rho _f(x)}}{\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2y\widehat{M}^1(z,y)n_f^{}(y){\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2y^{}n_f(y^{})(\widehat{M}^{})^1(y^{},z^{}).`$
Equations (3.66) and (3.2) are ‘finite’ forms of (2.87) and (2.2), respectively; as with the gauge potential the $`n(z)`$ vector is replaced with its finite form, $`n_f(z)`$, and the $`1`$ in $`\rho `$ is removed.
Since on the plane $`x_{}=0`$ the explicit form of $`n_f(z)`$ and $`\widehat{M}^1(z,z^{})`$ are at hand we can also give a closed form for $`f(z,z^{})`$:
$$f(z,z^{})=\frac{1}{2}\left(1i\widehat{l}\right)f_{}(z,z^{})+\frac{1}{2}\left(1+i\widehat{l}\right)f_+(z,z^{}),$$
(3.68)
where
$$f_\pm (z,z^{})=\stackrel{~}{𝒱}e^{\lambda ^2\varphi (z)}g_\pm (z,z^{})e^{\lambda ^2\varphi (z^{})},$$
(3.69)
and
$`g_\pm (z,z^{})`$ $`=`$ $`{\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2yG_\pm (zy)e^{\pm 2\lambda ^2\varphi (y)}G_{}(yz^{})`$
$`{\displaystyle \frac{1}{\rho _f}}{\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2yG_\pm (zy)e^{\pm 2\lambda ^2\varphi (y)}G_{}(\pm \omega +y)`$
$`\times {\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2y^{}G_\pm (\omega y^{})e^{\pm 2\lambda ^2\varphi (y^{})}G_{}(y^{}z^{}).`$
A sufficient condition for the self-duality of $`F_{\mu \nu }(x)`$ is that $`f(z,z^{})`$ commutes with the quaternions. This condition is equivalent to
$$g_+(z,z^{})=e^{2\lambda ^2\varphi (z)}g_{}(z,z^{})e^{2\lambda ^2\varphi (z^{})}.$$
(3.71)
A (somewhat roundabout) proof of (3.71) is given in Appendix B.
To sum up, the gauge potential, $`A_\mu (x)`$, and hence the field strength, $`F_{\mu \nu }(x)`$, can be written in terms of the ‘renormalised’ $`n_f(z)`$. We have explicitly determined $`n_f(z)`$ on the plane $`x_{}=0`$. At the point $`x=0`$ (i.e. $`x_{||}=x_{}=0`$) $`n_f`$ and hence $`A_\mu `$ is ill defined. This is no surprise since we are working in the singular gauge $`u(x)=1`$. The singularity has its origins in the zero mode structure of the $`G_\pm (z)`$; we can write
$$G_+(z)=\frac{1}{\stackrel{~}{𝒱}\overline{\text{x}}_{||}}+G_+^{}(z),G_{}(z)=\frac{1}{\stackrel{~}{𝒱}\text{x}_{||}}+G_{}^{}(z),$$
(3.72)
where the $`G_\pm ^{}(z)`$ have no zero modes and are thus well defined for $`x_{||}=0`$. Although $`A_\mu `$ diverges at $`x=0`$, local gauge invariants such as $`\text{tr}(F_{\mu \nu })^2`$ (no sum) should be smooth (presumably $`C^{\mathrm{}}`$). As for the field strength itself, $`F_{\mu \nu }(x)`$, this is not smooth at $`x=0`$, but its components must be bounded. Let us consider $`F_{\mu \nu }`$ at $`x_{}=0`$ with $`x_{||}0`$. For $`x_{||}0`$ the zero modes in (3.72) dominate and so we have<sup>7</sup><sup>7</sup>7 Strictly speaking (3.73) is only good away from $`z=\pm \omega `$. But as we are always dealing with integrable singularities we may safely employ (3.73) under the integral sign.
$$n_f(z)\frac{e^{\lambda ^2\varphi (z)}}{2\text{x}_{||}\stackrel{~}{𝒱}}\left(1i\widehat{l}\right)\frac{e^{\lambda ^2\varphi (z)}}{2\overline{\text{x}}_{||}\stackrel{~}{𝒱}}\left(1+i\widehat{l}\right),$$
(3.73)
thus
$$\rho _f\frac{c}{|\text{x}_{||}|^2\stackrel{~}{𝒱}^2},$$
(3.74)
where
$$c=_{\stackrel{~}{\text{𝕋}}^2}d^2ze^{2\lambda ^2\varphi (z)}.$$
(3.75)
Plugging (3.73) and (3.74) into the field strength formula (3.66) we see that in order to have a bounded $`F_{\mu \nu }`$ in the vicinity of $`x=0`$, $`f(z,z^{})`$ must be well behaved for $`x_{||}0`$. To see this consider, $`F_{01}=F_{23}`$, which for $`x_{}=0`$ and $`x_{||}0`$ has the form
$$F_{01}\frac{2\text{i}_1\stackrel{~}{𝒱}^1}{c}_{\stackrel{~}{\text{𝕋}}^2}d^2z_{\stackrel{~}{\text{𝕋}}^2}d^2z^{}e^{\lambda ^2\varphi (z)}e^{\lambda ^2\varphi (z^{})}f(z,z^{}).$$
(3.76)
$`F_{02}`$ and $`F_{03}`$ are a bit more complicated; here one finds phases of the form $`\overline{\text{x}}_{||}/\text{x}_{||}`$ which do not have definite values at $`x_{||}=0`$. These phases are an artifact of the singular gauge; $`\text{tr}(F_{02})^2`$ and $`\text{tr}(F_{03})^2`$ are well behaved at $`x_{||}=0`$. We now show that $`f(z,z^{})`$ is smooth in the vicinity of $`x_{||}0`$. Since the exponentials in (3.69) are $`x_{||}`$-independent it suffices to show that $`g_+(z,z^{})`$ has a well defined $`x_{||}0`$ limit. Glancing at (3.2) one sees that the first term in $`g_+(z,z^{})`$ has double and single poles in $`\text{x}_{||}`$ and $`\overline{\text{x}}_{||}`$. These poles are cancelled by the second term. After some algebra one finds that
$`g_+(z,z^{})`$ $`=`$ $`{\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2y\left(G_+^{}(zy)G_+^{}(\omega y)\right)e^{2\lambda ^2\varphi (y)}\left(G_{}^{}(yz^{})G_{}^{}(y+\omega )\right)`$ (3.77)
$`{\displaystyle \frac{1}{c}}{\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2ye^{2\lambda ^2\varphi (y)}\left(G_+^{}(zy)G_+^{}(\omega y)\right)`$
$`\times {\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2y^{}e^{2\lambda ^2\varphi (y^{})}(G_{}^{}(y^{}z^{})G_{}^{}(y^{}+\omega ))`$
$`+𝒪(x_{||}),`$
which is well defined at $`x_{||}=0`$. A similar expression can be obtained for $`g_{}(z,z^{})`$. From (3.69) the integrand in (3.76) is simply $`g_+(z,z^{})`$ and so all we have to do is to integrate the right hand side of (3.77) over $`z`$ and $`z^{}`$. Since the $`G_\pm ^{}(z)`$ integrate to zero this is trivial. Putting all this together yields
$`F_{01}`$ $`=`$ $`{\displaystyle \frac{2\text{i}_1\stackrel{~}{𝒱}^2}{c}}[{\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2ye^{2\lambda ^2\varphi (y)}|G_+^{}(\omega y)|^2`$
$`{\displaystyle \frac{1}{c}}\left|{\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2ye^{2\lambda ^2\varphi (y)}G_+^{}(\omega y)|^2\right]+𝒪(x_{||}).`$
The content of the brackets is strictly positive, i.e. we have not simply determined the field strength at a point where it is zero.
### 3.3 Nahm transform interpretation
In the previous section we implemented the ADHM construction in the one-instanton sector for $`\text{𝕋}^2\times \text{}^2`$. However, in contrast to the caloron problem $`n(z)`$ appears not to exist. This was circumvented by formally extracting an infinite factor to obtain the ‘finite’ $`n_f(z)`$. Here we will explain precisely how the gauge potential (3.52) can be interpreted as the Nahm transform of the AB potential (3.20). We would like to stress that this does not entail the kind of formal manipulations we used to derive (3.52) in the first place via the ADHM construction.
The Weyl operator on $`\stackrel{~}{\text{𝕋}}^2`$ associated with $`\widehat{A}(z)`$ has two square integrable zero modes <sup>8</sup><sup>8</sup>8In ref where the dual torus was take to be $`\stackrel{~}{S}^1\times \text{}`$ a limiting case of $`\stackrel{~}{\text{𝕋}}^2`$, $`\text{dim}(\text{ker}\widehat{D}^{})=2`$ was also obtained.. These modes can be identified with the columns of $`n_f^{}(z)`$ when the quaternionic object $`n_f(z)`$ is recast as a $`2\times 2`$ matrix with complex entries. To set the scene let us briefly recall how the Nahm transformation is formulated on $`\text{𝕋}^4`$. Consider a self-dual $`SU(N)`$ potential $`A_\mu (x)`$ on $`\text{𝕋}^4`$ with instanton number $`k`$. Then one studies the Weyl operator associated with the $`U(N)`$ potential obtained by adding a constant abelian potential $`iz_\mu `$ to $`A_\mu `$
$$D_z(A)=i_\mu D_z^\mu (A),D_z^\mu =^\mu +A^\mu (x)iz^\mu .$$
(3.79)
Provided certain mathematical technicalities are met $`D^{}=i_\mu ^{}D_z^\mu (A)`$ has $`k`$ square integrable zero modes $`\psi _z^i(x)`$ with $`i=1,2,\mathrm{},k`$. For convenience we take them to be normalised to unity. The $`U(k)`$ potential
$$\widehat{A}_\mu ^{ij}(z)=_{\text{𝕋}^4}d^4x\psi _{z}^{i}{}_{}{}^{}(x)\frac{}{z^\mu }\psi _z^j(x),$$
(3.80)
is a self-dual potential on the dual torus $`\stackrel{~}{\text{𝕋}}^4`$ with instanton number $`N`$. On $`\text{𝕋}^4`$ this procedure is involutive and (in a suitable gauge) free of singularities.
Let us write the Weyl operator associated with the AB potential (3.20) as a $`2\times 2`$ matrix:
$$\frac{i}{2}D_x^{}(\widehat{A})=S\left(\begin{array}{cc}i_{\overline{\text{z}}}+\frac{1}{2}\text{x}_{||}i_{\overline{\text{z}}}\varphi & \frac{1}{2}\text{x}_{}\\ \frac{1}{2}\overline{\text{x}}_{}& i_\text{z}+\frac{1}{2}\overline{\text{x}}_{||}+i_\text{z}\varphi \end{array}\right)S^1,$$
(3.81)
where <sup>9</sup><sup>9</sup>9 $`S`$ is a unitary transformation with the property $`S^1\sigma _1S=\sigma _3`$, $`S^1\sigma _2S=\sigma _2`$ and $`S^1\sigma _3S=\sigma _1`$. $`S=(1\text{l}i\sigma _2)/\sqrt{2}`$ and $`\text{x}_{}=x_2+ix_3`$. For $`\text{x}_{}=0`$ one can write down two square-integrable zero modes for $`D_x^{}(\widehat{A})`$
$$\psi _x^1(z)=\frac{1}{\sqrt{\rho _f}}S\left(\begin{array}{c}e^{\lambda ^2\varphi (z)}G_{}(z+\omega )\\ 0\end{array}\right),\psi _x^2(z)=\frac{1}{\sqrt{\rho _f}}S\left(\begin{array}{c}0\\ e^{\lambda ^2\varphi (z)}G_+(z\omega )\end{array}\right).$$
(3.82)
Both zero modes are singular at $`z=\pm \omega `$. Inserting these (normalised) zero modes into (3.80) yields exactly the same potential (discarding the $`U(1)`$ part of the U(2) connection) as constructed in the previous section. If one writes $`n_f^{}`$ as a $`2\times 2`$ matrix the columns are (upto a normalisation factor) the Nahm zero modes. As should be clear from the considerations of the previous section it is non-trivial to obtain the zero modes for $`\text{x}_{}0`$. The crucial feature of these zero modes is that although they are singular at the fluxes $`z=\pm \omega `$ the Weyl equation does not have sources, i.e. $`D_x^{}(\widehat{A})\psi _x^i(z)`$ is exactly zero. Basically, the damping exponentials soften the singularities of the Green’s functions $`G_{}(z+\omega )`$ and $`G_+(z\omega )`$ so that no delta function sources occur on the right hand side of the Weyl equation.
It is also instructive to compare the situation on $`\text{𝕋}^2\times \text{}^2`$ with the caloron case ($`S^1\times \text{}^3`$). It is easy to write down the corresponding zero modes on $`\stackrel{~}{S}^1`$ for the caloron problem. One simply replaces the $`\stackrel{~}{\text{𝕋}}^2`$ Green’s functions $`\varphi `$, $`G_+`$ and $`G_{}`$ with their $`\stackrel{~}{S}^1`$ counterparts. However, in this case the Weyl equations do have sources. The $`e^{\pm \lambda ^2\varphi (z)}`$, being finite at $`z=\pm \omega `$, have no damping effect on the $`G_\pm `$. Because of these sources, direct insertion of the $`\stackrel{~}{S}^1`$ ‘zero modes’ into (3.80) does not yield a self-dual potential on $`S^1\times \text{}^3`$. Rather, one has to change the normalisation of the zero modes to compensate for the sources. This amounts to including $`1`$ in the definition of $`\rho `$.
Given that the $`\stackrel{~}{\text{𝕋}}^2`$ Weyl operator has perfect zero modes what exactly is the status of the inverse of $`\widehat{M}`$ introduced in the previous section? What is clear is that our $`\widehat{M}^1(z,z^{})`$ is not the inverse of $`\widehat{M}`$ on the space of square integrable spinors; no such inverse exists. Our $`\widehat{M}^1(z,z^{})`$ can be viewed as the inverse of $`\widehat{M}`$ on a space of functions on $`\stackrel{~}{\text{𝕋}}^2`$ having softer singularities at the fluxes than the zero modes. In any case $`\widehat{M}^1(z,z^{})`$ only enters at intermediate stages of the calculation. What is important is $`n_f(z)`$, which, as we have shown here, encodes two perfect zero modes of our Weyl operator.
Thus it seems there are three types of Nahm transformation. First and foremost is the $`\text{𝕋}^4`$ transformation where all potentials and attendant zero modes are smooth. For $`\text{𝕋}^n\times \text{}^{4n},n<4`$ the self-duality equations on $`\stackrel{~}{\text{𝕋}}^n`$ have source terms. The Weyl zero modes on $`\stackrel{~}{\text{𝕋}}^n`$ are also singular but for $`n=2`$ (and presumably $`n=3`$) there are no source terms in the Weyl equation and so (3.80) can be applied without modification. For $`n=1`$ (and $`n=0`$ for that matter) the Weyl equation has source terms which are finessed by altering the normalisation of the zero modes.
## 4 Two-instantons
The two-instanton problem on the torus presents new challenges. In particular, the Nahm potential, $`\widehat{A}(z)`$, on $`\stackrel{~}{\text{𝕋}}^n`$ is non-Abelian; for $`k=2`$ instantons $`\widehat{A}(z)`$ is an $`SU(2)`$ potential. In contrast to the one-instanton case the determination of $`\widehat{A}(z)`$ is itself a non-trivial exercise. For $`\text{𝕋}^2\times \text{}^2`$ and $`S^1\times \text{}^3`$ the field strength associated with the Nahm potentials is zero, except at the singularities. But even here we do not have closed forms for $`\widehat{A}(z)`$. In section 4.1 we give some particular solutions to the $`k=2`$ ADHM constraints. The associated Weyl equations for the $`\text{𝕋}^2\times \text{}^2`$ problem are investigated in section 4.2. This analysis is very similar to that of section 3.2 for the one instantons. Indeed, the resulting two-instantons can be viewed as twisted one instantons when the torus is cut in half.
### 4.1 ADHM constraints on $`\text{𝕋}^n\times \text{}^{4n}`$
In the previous chapter we considered the general one-instanton which (apart for $`S^1\times \text{}^3`$) is non-periodic. For $`k=2`$ the ADHM constraint (2.42) is obviously more complicated. In particular, the quadratic term in (2.42) is, in general, non-zero. There is however one simplification at the two-instanton level; there exist non trivial solutions of the ADHM constraints which correspond to periodic gauge potentials on $`\text{𝕋}^n\times \text{}^{4n}`$. This is because we can choose the two ‘component’ instantons to have a different orientation in group space.
For simplicity, let us restrict ourselves to the periodic case. Then for $`k=2`$ we can write $`v`$ and $`\widehat{M}`$ as follows
$`\begin{array}{ccc}\hfill v=(v_1^\alpha v_2^\alpha ),\widehat{M}=\left(\begin{array}{cc}\widehat{M}_{11}^{\alpha \beta }& \widehat{M}_{12}^{\alpha \beta }\\ \widehat{M}_{21}^{\alpha \beta }& \widehat{M}_{22}^{\alpha \beta }\end{array}\right),& & \end{array}`$ (4.4)
where $`v_1^\alpha =q_1`$, $`v_2^\alpha =q_2`$, and
$`\begin{array}{cc}\widehat{M}_{11}^{\alpha \beta }=\delta ^{\alpha \beta }(\alpha +y_1x)+b_{11}^{\alpha \beta },\hfill & \widehat{M}_{12}^{\alpha \beta }=\widehat{M}_{21}^{\beta \alpha }=b_{12}^{\alpha \beta }\hfill \\ \widehat{M}_{22}^{\alpha \beta }=\delta ^{\alpha \beta }(\alpha +y_2x)+b_{22}^{\alpha \beta }.\hfill & \end{array}`$ (4.7)
We now have to determine the $`b`$ matrices via (2.42). In the one instanton calculation we relied on the vanishing of the quadratic term in (2.42). While this will not hold, in general, for the two instanton case there may be particular solutions where the quadratic term is zero. Indeed on $`\text{}^4`$, the $`k=2`$ problem is expedited by the vanishing of the quadratic term in (2.20) . If the quadratic term in (2.42) is zero, the $`b`$ matrices read
$`\begin{array}{ccc}\hfill b_{11}^{\alpha \beta }=b_{22}^{\alpha \beta }=0,b_{12}^{\alpha \beta }={\displaystyle \frac{1}{2(\alpha \beta +y_1y_2)^{}}}Q,& & \end{array}`$ (4.9)
where
$`\begin{array}{ccc}\hfill Q=q_1^{}q_2q_2^{}q_1.& & \end{array}`$ (4.11)
In Appendix A we will prove that if $`2(y_1y_2)\mathrm{\Lambda }`$ and $`y_1y_2\mathrm{\Lambda }`$ then the quadratic term does indeed vanish. For example this happens for $`y_1y_2=\frac{1}{2}(e_0+e_1+\mathrm{}+e_{n1})`$. This means that the lattice points of the second ‘species’ of instanton lie exactly at the midpoints (see figure 2) of the lattice points of the first.
In the special case $`n=1`$ (i.e. the caloron problem) one only needs $`y_1y_2`$ to be parallel to $`e_0`$ for the quadratic term to vanish. This is a consequence of the fact that for $`S^1\times \text{}^3`$ one may take $`e_0`$ and hence the elements of $`\mathrm{\Lambda }`$ to be real. For $`n>1`$, $`2(y_1y_2)\mathrm{\Lambda }`$ is a necessary condition for the vanishing of the quadratic term. Thus for $`2(y_1y_2)\mathrm{\Lambda }`$ (4.9) is an approximation; (4.9) is then the first term of a power series expansion in the scale parameters.
Let us concentrate on the cases where the quadratic terms does vanish. Fourier transformation yields $`\stackrel{~}{𝒱}^1\widehat{M}=id_zx+\widehat{A}(z)`$, where $`\widehat{A}(z)`$ is the $`SU(2)`$ potential
$`\begin{array}{ccc}\hfill i\widehat{A}(z)=\left(\begin{array}{cc}\frac{1}{2}(y_1y_2)& \frac{1}{2}ie^{i(y_1y_2)z}d_z\psi (z)Q\\ \frac{1}{2}ie^{i(y_1y_2)z}d_z\psi (z)Q& \frac{1}{2}(y_2y_1)\end{array}\right),& & \end{array}`$ (4.15)
and
$`\begin{array}{ccc}\hfill \psi (z)={\displaystyle \underset{\alpha \mathrm{\Lambda }}{}}{\displaystyle \frac{e^{i(\alpha +y_1y_2)z}}{|\alpha +y_1y_2|^2}}.& & \end{array}`$ (4.17)
$`\psi (z)`$ is a Green’s function for the Laplace operator on $`\stackrel{~}{\text{𝕋}}^n`$
$`\begin{array}{ccc}\hfill d_zd_z^{}\psi (z)=\stackrel{~}{𝒱}e^{i(y_1y_2)z}\delta ^n(z).& & \end{array}`$ (4.19)
Observe that $`\psi `$ is non-periodic
$`\begin{array}{ccc}\hfill \psi (z+\stackrel{~}{e}_i)=e^{i(y_1y_2)\stackrel{~}{e}_i}\psi (z),& & \end{array}`$ (4.21)
where $`\stackrel{~}{e}_i`$ refers to the dual basis; $`\stackrel{~}{e}_ie_j=2\pi \delta _{ij}`$. Now if $`2(y_1y_2)\mathrm{\Lambda }`$ and $`(y_1y_2)\mathrm{\Lambda }`$, $`\psi (z)`$ will be antiperiodic in at least one direction, and periodic in the remaining directions. One can also see that for these special values of $`y_1y_2`$, $`\psi (z)`$ is real. The reality of $`\psi `$ is a sufficient condition for the potential (4.15) to be self-dual.
We now appear to have to deal with a non-Abelian Weyl operator. In what follows the inversion problem is reduced to an Abelian problem much like that for the one instanton case. Of course, in the light of the previous chapter due care regarding the meaning of the inverse is in order. $`\widehat{M}`$ can be rewritten as follows
$`\begin{array}{ccc}\hfill \stackrel{~}{𝒱}^1\widehat{M}=e^{\frac{i}{2}(y_1y_2)z\sigma _3}P^1\left(\begin{array}{cc}D^+& 0\\ 0& D^{}\end{array}\right)Pe^{\frac{i}{2}(y_1y_2)z\sigma _3},& & \end{array}`$ (4.25)
where $`D^\pm `$ are the (Abelian) Weyl operators
$`\begin{array}{ccc}D^\pm =id_zx\pm \frac{1}{2}d_z\psi Q,& & P={\displaystyle \frac{1}{\sqrt{2}}}(1\text{l}+i\sigma _1).\end{array}`$ (4.27)
The inverse of $`\widehat{M}`$ is simply
$`\begin{array}{ccc}\hfill \widehat{M}^1(z,z^{})=\stackrel{~}{𝒱}e^{\frac{i}{2}(y_1y_2)z\sigma _3}P^1\mathrm{\Delta }(z,z^{})Pe^{\frac{i}{2}(y_1y_2)z^{}\sigma _3},& & \end{array}`$ (4.29)
where $`\mathrm{\Delta }(z,z^{})`$ is a Green’s function for the diagonal operator $`\text{diag}(D^+,D^{})`$. Note that the exponentials in the decomposition of $`\widehat{M}^1(z,z^{})`$ are not periodic. To ensure a periodic $`\widehat{M}^1(z,z^{})`$ we must impose certain non-periodic boundary conditions on $`\mathrm{\Delta }(z,z^{})`$. Since we require $`\widehat{M}(z)\widehat{M}^1(z,z^{})=\stackrel{~}{𝒱}^2\delta ^n(zz^{})`$, then it follows that
$`\begin{array}{ccc}\hfill \left(\begin{array}{cc}D^+& 0\\ 0& D^{}\end{array}\right)\mathrm{\Delta }(z,z^{})=Pe^{\frac{i}{2}(zz^{})(y_1y_2)\sigma _3}P^1\delta ^n(zz^{}).& & \end{array}`$ (4.33)
It is convenient to absorb the exponential factor into the delta function. That is, consider the following (non-periodic) delta functions
$`\begin{array}{ccc}\hfill \delta _1^n(z)=e^{\frac{i}{2}z(y_1y_2)}\delta ^n(z),\delta _2^n(z)=e^{\frac{i}{2}z(y_1y_2)}\delta ^n(z).& & \end{array}`$ (4.35)
Using the following four (Abelian) Green’s functions, $`\mathrm{\Delta }_i^\pm (z,z^{}),i=1,2`$, where
$`\begin{array}{ccc}\hfill D_z^\pm \mathrm{\Delta }_i^\pm (z,z^{})=\delta _i^n(zz^{}).& & \end{array}`$ (4.37)
$`\mathrm{\Delta }`$ can be written as
$`\begin{array}{ccc}\hfill \mathrm{\Delta }(z,z^{})=\frac{1}{2}\left(\begin{array}{cc}\mathrm{\Delta }_1^++\mathrm{\Delta }_2^+& i\left(\mathrm{\Delta }_1^+\mathrm{\Delta }_2^+\right)\\ i\left(\mathrm{\Delta }_1^{}\mathrm{\Delta }_2^{}\right)& \mathrm{\Delta }_1^{}+\mathrm{\Delta }_2^{}\end{array}\right)(z,z^{}).& & \end{array}`$ (4.41)
Accordingly
$`\begin{array}{ccc}\hfill \widehat{M}^1(z,z^{})={\displaystyle \frac{\stackrel{~}{𝒱}}{2}}e^{\frac{i}{2}z(y_1y_2)\sigma _3}\left(\begin{array}{cc}\mathrm{\Delta }_1^++\mathrm{\Delta }_1^{}& i\left(\mathrm{\Delta }_2^{}\mathrm{\Delta }_2^+\right)\\ i\left(\mathrm{\Delta }_1^{}\mathrm{\Delta }_1^+\right)& \mathrm{\Delta }_2^++\mathrm{\Delta }_2^{}\end{array}\right)(z,z^{})e^{\frac{i}{2}z^{}(y_1y_2)\sigma _3}.& & \end{array}`$ (4.45)
### 4.2 Two-instanton on $`\text{𝕋}^2\times \text{}^2`$
Much as in section 3.2 we may take $`e_0`$ to be real and $`e_1`$ to be proportional to $`Q`$. Thus $`\widehat{Q}=Q/|Q|`$ plays the same role as $`\widehat{l}`$ did in the previous section. Indeed, the analogue of (3.16) is just $`z=\frac{1}{2}\left(1i\widehat{Q}\right)\text{z}+\frac{1}{2}\left(1+i\widehat{Q}\right)\overline{\text{z}}`$. We can write the Abelian Dirac operators $`D^\pm `$ defined in (4.27) as follows
$`\begin{array}{ccc}\hfill D^\pm =e^{{\scriptscriptstyle \frac{1}{2}}iQ\psi (z)}\left(id_zx_{||}\right)e^{\pm {\scriptscriptstyle \frac{1}{2}}iQ\psi (z)}x_{}.& & \end{array}`$ (4.47)
For the case $`y_2y_1=\frac{1}{2}(e_0+e_1)`$, we have
$`\begin{array}{ccc}\hfill \psi (z)={\displaystyle \frac{\stackrel{~}{𝒱}}{4\pi }}\mathrm{log}{\displaystyle \frac{\left|\theta (\frac{L_0}{4\pi }\text{z}+\frac{iL_0}{4L_1},\frac{iL_0}{2L_1})\right|^2}{\left|\theta (\frac{L_0}{4\pi }\text{z}+\frac{1}{2},\frac{iL_0}{2L_1})\right|^2}},& & \end{array}`$ (4.49)
which is antiperiodic in both directions.
When $`x_{}=0`$, the four Green’s functions $`\mathrm{\Delta }_i^\pm `$ read<sup>10</sup><sup>10</sup>10Note that $`\mathrm{\Delta }_i^\pm (z,z^{})=e^{iQ\psi (z)}G_i(zz^{})e^{\pm iQ\psi (z^{})}`$ is not correct, since one has to take into account the non-periodicity of the exponentials $`e^{\pm iQ\psi }=\mathrm{cosh}\left(|Q|\psi \right)\pm i\widehat{Q}\mathrm{sinh}\left(|Q|\psi \right)`$.
$`\mathrm{\Delta }_1^\pm (z,z^{})`$ $`=`$ $`e^{{\scriptscriptstyle \frac{1}{2}}iQ\psi (z)}\left[G_1(zz^{})\mathrm{cosh}\left(\frac{1}{2}|Q|\psi (z^{})\right)\pm G_2(zz^{})i\widehat{Q}\mathrm{sinh}\left(\frac{1}{2}|Q|\psi (z^{})\right)\right]`$
$`\mathrm{\Delta }_2^\pm (z,z^{})`$ $`=`$ $`e^{{\scriptscriptstyle \frac{1}{2}}iQ\psi (z)}\left[G_2(zz^{})\mathrm{cosh}\left(\frac{1}{2}|Q|\psi (z^{})\right)\pm G_1(zz^{})i\widehat{Q}\mathrm{sinh}\left(\frac{1}{2}|Q|\psi (z^{})\right)\right],`$
where the $`G_i(zz^{})`$ are (non-periodic) free Green’s functions defined as
$`\begin{array}{ccc}\hfill \left(id_zx_{||}\right)G_i(zz^{})=\delta _i^2(zz^{}),i=1,2.& & \end{array}`$ (4.52)
Inserting (4.2) into (4.45) yields
$`\begin{array}{ccc}\hfill \widehat{M}^1(z,z^{})=\stackrel{~}{𝒱}\mathrm{\Psi }(z)\left(\begin{array}{cc}G_1(zz^{})& 0\\ 0& G_2(zz^{})\end{array}\right)\mathrm{\Psi }^1(z^{}),& & \end{array}`$ (4.56)
where $`\mathrm{\Psi }(z)`$ is the $`2\times 2`$ matrix
$`\begin{array}{ccc}\hfill \mathrm{\Psi }(z)=\left(\begin{array}{cc}e^{{\scriptscriptstyle \frac{1}{2}}i(y_1y_2)z}\mathrm{cosh}\left(\frac{1}{2}|Q|\psi (z)\right)& \widehat{Q}e^{{\scriptscriptstyle \frac{1}{2}}i(y_1y_2)z}\mathrm{sinh}\left(\frac{1}{2}|Q|\psi (z)\right)\\ \widehat{Q}e^{{\scriptscriptstyle \frac{1}{2}}i(y_1y_2)z}\mathrm{sinh}\left(\frac{1}{2}|Q|\psi (z)\right)& e^{{\scriptscriptstyle \frac{1}{2}}i(y_1y_2)z}\mathrm{cosh}\left(\frac{1}{2}|Q|\psi (z)\right)\end{array}\right).& & \end{array}`$ (4.60)
The two component row vector $`n(z)`$ is
$`\begin{array}{ccc}\hfill n(z)=\stackrel{~}{𝒱}(q_1,q_2)\mathrm{\Psi }(0)\left(\begin{array}{cc}G_1(z)& 0\\ 0& G_2(z)\end{array}\right)\mathrm{\Psi }^1(z).& & \end{array}`$ (4.64)
Again we encounter infinite constants; $`\psi (z)\mathrm{}`$ as $`z0`$ and so all entries of the matrix $`\mathrm{\Psi }(0)`$ are ‘infinite’. As in section 3.2 we will temporarily treat $`\mathrm{\Psi }(0)`$ as a finite object. In the light of our one instanton calculation we expect some constraints on $`q_1`$ and $`q_2`$. We can choose $`q_1`$ to be real. In appendix B we show that for $`n(z)n^{}(z)`$ to be integrable requires that
$`\begin{array}{ccc}\hfill (q_1,q_2)=\lambda (1,\widehat{Q}),& & \end{array}`$ (4.66)
where $`\lambda `$ is a common scale parameter since $`|q_1|=|q_2|=\lambda `$. Observe that the relative group orientation of the two instantons is fixed. If the orientation of the first instanton lies at the ‘North pole’ of $`S^3SU(2)`$, then the orientation of the second instanton sits on the equator. Much as in the one instanton case the absence of non-integrable singularities leads to an upper bound on the scale parameter
$`\begin{array}{ccc}\hfill 0<\lambda ^2\stackrel{~}{𝒱}<2\pi .& & \end{array}`$ (4.68)
Another consequence of (4.66) is that $`(q_1,q_2)`$ is an eigenvector of the infinite matrix $`\mathrm{\Psi }(0)`$, i.e. $`(q_1,q_2)\mathrm{\Psi }(0)=e^{{\scriptscriptstyle \frac{1}{2}}|Q|\psi (0)}(q_1,q_2)`$. As in the one instanton calculation we define a ‘finite’ row vector $`\lambda \stackrel{~}{𝒱}n_f(z)=e^{{\scriptscriptstyle \frac{1}{2}}|Q|\psi (0)}n(z)`$. The final gauge potential is obtained by replacing $`n(z)`$ with $`n_f(z)`$ in (2.83) and replacing (2.85) with $`\rho =\stackrel{~}{𝒱}^1\rho _f=\stackrel{~}{𝒱}^1_{\stackrel{~}{\text{𝕋}}^2}n_f(z)n_f^{}(z).`$
In the course of the construction a number of constraints have been put on the ADHM data. It is helpful to divide these constraints into two. The first constraints are simply those imposed by hand to achieve technical simplification, i.e. we imposed periodicity and the midpoint condition in order that we could exactly determine the Weyl operator. In addition to these constraints we were forced to impose the additional constraints (4.66) and (4.68). By virtue of the midpoint prescription and (4.66) our two instantons begin to resemble one instantons if we cut $`\text{𝕋}^2`$ in half. In fact if we had chosen $`y_1y_2=\frac{1}{2}e_0`$ or $`y_1y_2=\frac{1}{2}e_1`$ instead of $`y_1y_2=\frac{1}{2}(e_0+e_1)`$ then our ‘two instanton’ would be nothing more than a ‘doubled’ one instanton. That is one can always produce a two-instanton on $`\text{𝕋}^n\times \text{}^{4n}`$ by taking a one instanton and doubling one of the periods. To show this equivalence one simply compares the ‘two instanton’ with $`y_1y_2=\frac{1}{2}e_0`$ or $`y_1y_2=\frac{1}{2}e_1`$ with the one instanton with $`\omega =\frac{1}{4}\stackrel{~}{e}_0`$ or $`\omega =\frac{1}{4}\stackrel{~}{e}_1`$. Then using the $`q_iq_i`$ symmetry mentioned at the end of section 2.2 one can show that the two sets of ADHM data correspond to the same instanton. The two instanton corresponding to $`y_1y_2=\frac{1}{2}(e_0+e_1)`$ appears to be ‘genuine’ in the sense it is not equivalent to some one-instanton solution. However it seems plausible that the $`y_1y_2=\frac{1}{2}(e_0+e_1)`$ case corresponds to a twisted one instanton (the twisted Nahm transformation is discussed in ).
## 5 Discussion
In this paper we have described in a general way how to implement the ADHM construction of $`SU(2)`$ instantons on $`\text{𝕋}^n\times \text{}^{4n}`$. The first step (which corresponds to solving the quadratic ADHM constraint) is to construct a self-dual $`SU(k)`$ ($`U(1)`$ for $`k=1`$) potential, $`\widehat{A}(z)`$, on the dual torus $`\stackrel{~}{\text{𝕋}}^n`$ (here $`k`$ is the topological charge). $`\widehat{A}(z)`$ has singularities which are determined by the ADHM data (i.e. the scales, positions and group orientation of the ‘component’ instantons). We have constructed the Weyl operators corresponding to the general one-instanton and some two instantons on $`\text{𝕋}^n\times \text{}^{4n}`$. However, the problem of solving the Weyl equations poses a considerable technical challenge. One is therefore motivated to start with lower values of $`n`$. We have considered the $`n=2`$ problem in some detail.
The solutions here are not deformations of ’t Hooft instantons; the ’t Hooft ansatz fails to provide solutions on $`\text{𝕋}^2\times \text{}^2`$. Unlike for $`S^1\times \text{}^3`$ we are forced to impose constraints on the ADHM parameters in order to guarantee a well defined potential on $`\text{𝕋}^2\times \text{}^2`$. In particular, we find an upper bound on the scale parameters; for the one-instanton, $`\lambda ^2\stackrel{~}{𝒱}<4\pi `$ and for our restricted two-instanton we found that $`\lambda ^2\stackrel{~}{𝒱}<2\pi `$ (here we were forced to give the two component instantons a common scale parameter).
For $`n>2`$, i.e. $`\text{𝕋}^3\times \text{}`$ and $`\text{𝕋}^4`$, the Weyl equations seem more problematic. While the $`\text{𝕋}^2\times \text{}^2`$ Weyl operator corresponds to an Aharonov-Bohm problem on $`\stackrel{~}{\text{𝕋}}^2`$, on $`\text{𝕋}^3\times \text{}`$ we have to solve the Weyl equation on $`\stackrel{~}{\text{𝕋}}^3`$ in the (self-dual) background of an electric and magnetic dipole field . For $`\text{𝕋}^4`$ the one instanton calculation should fail. Presumably there is no way to avoid non-integrable singularities. For our restricted two instantons the prospects seem a little brighter. This is because these seemingly correspond to twisted one instantons (or even $`\frac{1}{2}`$ instantons in the presence of non-orthogonal twists). There is no known obstacle to the existence of such objects on $`\text{𝕋}^4`$.
Although the $`\text{𝕋}^3\times \text{}`$ and $`\text{𝕋}^4`$ problems certainly merit more attention the $`\text{𝕋}^2\times \text{}^2`$ case requires further development. Even in the 1-instanton sector we were only able to provide closed forms for $`A_\mu (x)`$ and $`F_{\mu \nu }(x)`$ in a 2-dimensional subspace ($`x_{}=0`$) of $`\text{𝕋}^2\times \text{}^2`$. To obtain analytic results for $`x_{}0`$ requires progress in dealing with massive Aharonov-Bohm type Dirac equations on $`\stackrel{~}{\text{𝕋}}^2`$. Furthermore, we have said nothing about the geometry of the moduli space or the constituent monopoles of our instantons. One could numerically plot the action density of the one instantons in the plane $`x_{}=0`$ to see if there are two peaks associated with the two expected monopole constituents.
## Acknowledgements
C. F. is grateful to C. J. Biebl for helpful discussions. We thank P. van Baal for his comments on a preliminary version of the manuscript. Part of the research of T. T. was performed during his stay at the Institute of Theoretical Physics in Jena, and in the latter stages of the work he was supported by the Deutsche Forschungsgemeinschaft (grant DFG-Re 856/4-1). In the early stages of this work C. F. was supported by the DFG (grant DFG-Wi 777/3-2).
## Appendix A The quadratic term in (2.42)
In this appendix we show that the quadratic term in (2.42) vanishes for the one instanton and particular two instanton described in chapter 4.
Let us start with the one instanton. The quadratic term in question is
$`\begin{array}{ccc}\hfill ^{\alpha \beta }={\displaystyle \underset{\gamma \mathrm{\Lambda }}{}}\left(b^{\gamma \alpha }{}_{}{}^{}b_{}^{\gamma \beta }b^{\gamma \beta }{}_{}{}^{}b_{}^{\gamma \alpha }\right).& & \end{array}`$ (A.2)
Assuming $`^{\alpha \beta }=0`$ leads to (3.6). Inserting this into (A.2) gives
$`\begin{array}{ccc}\hfill ^{\alpha \beta }& =& \lambda ^4{\displaystyle \underset{\gamma \mathrm{\Lambda }\{\alpha ,\beta \}}{}}\widehat{l}\left({\displaystyle \frac{1}{(\gamma \alpha )^{}}}{\displaystyle \frac{1}{\gamma \beta }}{\displaystyle \frac{1}{(\gamma \beta )^{}}}{\displaystyle \frac{1}{\gamma \alpha }}\right)\widehat{l}\hfill \\ & & \times \mathrm{sin}\left[(\alpha \gamma )\omega \right]\mathrm{sin}\left[(\beta \gamma )\omega \right].\hfill \end{array}`$ (A.5)
It is clear that each summand in (A.5) does not separately vanish. Rather there is a pairwise cancellation; for each $`\gamma \mathrm{\Lambda }\{\alpha ,\beta \}`$ there is exactly one other lattice point $`\gamma ^{}\mathrm{\Lambda }\{\alpha ,\beta \}`$ so that the two summands add up to zero. It is apparent that the appropriate choice for $`\gamma ^{}`$ is $`\gamma ^{}=\gamma +\alpha +\beta .`$ If $`2\gamma =\alpha +\beta `$, i.e. $`\gamma ^{}=\gamma `$, then the summand itself vanishes.
The argument is similar for the two instanton of section 4. Here the quadratic term is
$`\begin{array}{ccc}\hfill _{ij}^{\alpha \beta }={\displaystyle \underset{\gamma \mathrm{\Lambda }}{}}\left(b_{1i}^{\gamma \alpha }{}_{}{}^{}b_{1j}^{\gamma \beta }b_{1j}^{\gamma \beta }{}_{}{}^{}b_{1i}^{\gamma \alpha }+b_{2i}^{\gamma \alpha }{}_{}{}^{}b_{2j}^{\gamma \beta }b_{2j}^{\gamma \beta }{}_{}{}^{}b_{2i}^{\gamma \alpha }\right).& & \end{array}`$ (A.7)
Inserting (4.9) gives $`_{12}^{\alpha \beta }=_{21}^{\alpha \beta }=0`$, and
$`_{22}^{\alpha \beta }`$ $`=`$ $`{\displaystyle \underset{\gamma \mathrm{\Lambda }}{}}\left(b_{12}^{\gamma \alpha }{}_{}{}^{}b_{12}^{\gamma \beta }b_{12}^{\gamma \beta }{}_{}{}^{}b_{12}^{\gamma \alpha }\right)`$
$`=`$ $`{\displaystyle \frac{1}{4}}{\displaystyle \underset{\gamma \mathrm{\Lambda }}{}}(Q{\displaystyle \frac{1}{\gamma \alpha +y_1y_2}}{\displaystyle \frac{1}{(\gamma \beta +y_1y_2)^{}}}Q`$
$`Q{\displaystyle \frac{1}{\gamma \beta +y_1y_2}}{\displaystyle \frac{1}{(\gamma \alpha +y_1y_2)^{}}}Q).`$
Now we will show that $`_{22}`$ is zero for $`2(y_1y_2)\mathrm{\Lambda }`$. As in the one instanton case each summand in (A) does not separately vanish. For each $`\gamma \mathrm{\Lambda }`$ there is one other lattice point $`\gamma ^{}\mathrm{\Lambda }`$ so that the two summands add up to zero
$`\begin{array}{ccc}\hfill \gamma ^{}=\gamma +\alpha +\beta 2(y_1y_2).& & \end{array}`$ (A.10)
Since $`\gamma ^{}\mathrm{\Lambda }`$ we require $`2(y_1y_2)\mathrm{\Lambda }`$. If $`2\gamma =\beta +\alpha 2(y_1y_2)`$ then $`\gamma ^{}=\gamma `$ so that we do not have two counterbalancing summands. However, in this case the summand itself vanishes.
## Appendix B Equation (3.71)
In this appendix we outline a proof of (3.71) which, for $`x_{}=0`$, is equivalent to the statement that $`f(z,z^{})`$ commutes with the quaternions. In the caloron problem one simply notes that $`f`$ is the inverse of $`M^{}M`$ which by construction commutes with the quaternions. We could also explicitly check that our $`f`$ is the inverse of $`M^{}M`$. However, we would face the thorny problem of coincident fluxes and sources . Therefore, we will adopt a more pedestrian approach. Before we embark on this we note that for $`z+z^{}=0`$ a trivial change of variables in the integrals defining $`g_{}(z,z^{})`$ suffices to verify (3.71). For $`z+z^{}0`$ we have a more indirect argument. When $`z\omega `$ it is easy to check that
$$\left(i_{\overline{\text{z}}}\frac{1}{2}\text{x}_{||}\right)e^{2\lambda ^2\varphi (z)}\left(i_\text{z}\frac{1}{2}\overline{\text{x}}_{||}\right)\left(g_+(z,z^{})e^{2\lambda ^2\varphi (z)}g_{}(z,z^{})e^{2\lambda ^2\varphi (z^{})}\right)=0.$$
(B.1)
This shows that the left and right hand sides of (3.71) satisfy the same differential equations. To complete the argument we must show that they obey the same boundary conditions. Clearly both are periodic on $`\stackrel{~}{\text{𝕋}}^2`$, but we also need to show that $`g_+(z,z^{})`$ and $`e^{2\lambda ^2\varphi (z)}g_{}(z,z^{})e^{2\lambda ^2\varphi (z^{})}`$ have the same asymptotics at the fluxes. Let us examine $`g_\pm (z,z^{})`$ in the neighbourhood of $`z=\omega `$. One can see that $`g_+(\omega ,z^{})`$ is well defined for $`\lambda ^2\stackrel{~}{𝒱}<2\pi `$, while $`g_{}(\omega ,z^{})=0`$. This does not contradict (3.71) since the exponential $`e^{2\lambda ^2\varphi (z)}`$ diverges as $`\kappa |\text{z}\text{w}|^{\lambda ^2\stackrel{~}{𝒱}/(2\pi )}`$ for $`z\omega `$ where $`\kappa `$ is a constant. Consistency requires that $`g_{}(z,z^{})\kappa ^1|\text{z}\text{w}|^{\lambda ^2\stackrel{~}{𝒱}/(2\pi )}g_+(\omega ,z^{})e^{2\lambda ^2\varphi (z^{})}`$ for $`z\omega `$. One can show that $`g_{}(z,z^{})`$ decays as it should in the limit $`z\omega `$ by considering the derivative of $`g_{}(z,z^{})`$:
$`\left(i_{\overline{\text{z}}}\frac{1}{2}\text{x}_{||}\right)g_{}(z,z^{})`$ $`=`$ $`\frac{1}{2}e^{2\lambda ^2\varphi (z)}G_+(zz^{})`$
$`{\displaystyle \frac{e^{2\lambda ^2\varphi (z)}}{2\rho _f}}G_+(\omega +z){\displaystyle _{\stackrel{~}{\text{𝕋}}^2}}d^2y^{}G_{}(\omega y^{})e^{2\lambda ^2\varphi (y^{})}G_+(y^{}z^{}).`$
In the neighbourhood of $`z=\omega `$, $`2\pi G_+(\omega +z)i/(\overline{\text{z}}\overline{\text{w}})`$, and so the second term in (B) dominates (provided $`z^{}\pm \omega `$). Integrating yields
$$g_{}(z,z^{})\frac{1}{\lambda ^2\stackrel{~}{𝒱}\kappa \rho _f}|\text{z}\text{w}|^{\lambda ^2\stackrel{~}{𝒱}/(2\pi )}_{\stackrel{~}{\text{𝕋}}^2}d^2y^{}G_{}(\omega y^{})e^{2\lambda ^2\varphi (y^{})}G_+(y^{}z^{}),$$
(B.3)
which indeed decays correctly. Full agreement with (3.71) requires
$$g_+(\omega ,z^{})=\frac{e^{2\lambda ^2\varphi (z^{})}}{\lambda ^2\stackrel{~}{𝒱}\rho _f}_{\stackrel{~}{\text{𝕋}}^2}d^2y^{}G_{}(\omega y^{})e^{2\lambda ^2\varphi (y^{})}G_+(y^{}z^{}).$$
(B.4)
To check this one simply notes that away from $`z^{}=\pm \omega `$ the left and right hand sides are annihilated by the same differential operator, $`\left(i_\text{z}^{}\frac{1}{2}\overline{\text{x}}_{||}2i\lambda ^2_\text{z}^{}\varphi (z^{})\right)\left(i_{\overline{\text{z}}^{}}\frac{1}{2}\text{x}_{||}\right)`$. It is simple to also check that they agree in the neighbourhoods of $`z^{}=\pm \omega `$ which completes the proof.
## Appendix C Two instanton singularities
Consider the 2-component row vectors $`v_\pm =(1,\pm \widehat{Q})`$ which are (formally) eigenvectors of $`\mathrm{\Psi }(0)`$ in that $`v_\pm \mathrm{\Psi }(0)=e^{\pm {\scriptscriptstyle \frac{1}{2}}|Q|\psi (0)}v_\pm `$. We now make the decomposition $`(q_1,q_2)=\alpha _+v_++\alpha _{}v_{}`$ where the quaternions $`\alpha _\pm `$ are not completely free since $`q_1^{}q_2q_2^{}q_1=Q`$. The integrand in the definition of $`\rho `$ is
$`\stackrel{~}{𝒱}^1n(z)n^{}(z)`$ $`=`$ $`|\alpha _+|^2e^{|Q|\psi (0)}\left[𝒢_+(z)𝒢_+^{}(z)e^{|Q|\psi (z)}+𝒢_{}(z)𝒢_{}^{}(z)e^{|Q|\psi (z)}\right]`$
$`+|\alpha _{}|^2e^{|Q|\psi (0)}\left[𝒢_+(z)𝒢_+^{}(z)e^{|Q|\psi (z)}+𝒢_{}(z)𝒢_{}^{}(z)e^{|Q|\psi (z)}\right]`$
$`+\text{terms linear in }\alpha _+\alpha _{}^{}\text{ and }\alpha _{}\alpha _+^{}\text{,}`$
where we have employed the notation
$$𝒢_\pm (z)=G_1(z)\pm G_2(z),$$
(C.2)
not to be confused with the $`G_\pm (z)`$ introduced in section 3.2! First, let us consider the singularity structure of the free Green’s functions $`𝒢_\pm (z)`$ which satisfy $`(id_zx)𝒢_\pm (z)=\delta _1(z)\pm \delta _2(z).`$ Now $`\delta _1^2(z)`$ and $`\delta _2^2(z)`$ are zero except for all dual lattice points ($`z\stackrel{~}{\mathrm{\Lambda }}`$). However $`\delta _1^2(z)+\delta _2^2(z)`$ is only singular at half of the lattice points, while $`\delta _1^2(z)\delta _2^2(z)`$ is singular at the remaining dual lattice points. This can be seen from the following identities
$$\delta _1^2(z)+\delta _2^2(z)=2\mathrm{cos}\left(\frac{1}{2}(y_1y_2)z\right)\delta ^2(z),\delta _1^2(z)\delta _2^2(z)=2i\mathrm{sin}\left(\frac{1}{2}(y_1y_2)z\right)\delta ^2(z).$$
(C.3)
Now since $`2(y_1y_2)\mathrm{\Lambda }`$ it follows that $`\frac{1}{2}(y_1y_2)z=\frac{1}{2}\pi n,n\text{}`$ for $`z\stackrel{~}{\mathrm{\Lambda }}`$ which means that either the sine or the cosine must be zero for $`z\stackrel{~}{\mathrm{\Lambda }}`$. In particular, we see that unlike $`\delta _1^2(z)+\delta _2^2(z)`$, $`\delta _1^2(z)\delta _2^2(z)`$ has no singularity at $`z=0`$. Thus we conclude that $`𝒢_{}(z)`$ has no singularity at $`z=0`$. In the neighbourhood of $`z=0`$ we have
$`\begin{array}{ccc}\hfill 𝒢_+(z)𝒢_+^{}(z){\displaystyle \frac{1}{|\text{z}|^2}},𝒢_{}(z)𝒢_{}^{}(z)\text{ non-singular.}& & \end{array}`$ (C.5)
We also require the behaviour of $`\psi (z)`$ at $`z=0`$, $`\psi (z)(\stackrel{~}{𝒱}/2\pi )\mathrm{log}|\text{z}|`$. Near $`z=0`$ we have
$`\begin{array}{ccc}\hfill 𝒢_+(z)𝒢_+^{}(z)e^{|Q|\psi (z)}|\text{z}|^{2+|Q|\stackrel{~}{𝒱}/(2\pi )},𝒢_+(z)𝒢_+^{}(z)e^{|Q|\psi (z)}|\text{z}|^{2|Q|\stackrel{~}{𝒱}/(2\pi )}.& & \end{array}`$ (C.7)
The second part of (C.7), i.e. $`𝒢_+(z)𝒢_+^{}(z)e^{|Q|\psi (z)}`$ is non-integrable. However, this term is absent in the $`|\alpha _+|^2`$ contribution to (C) and so if we make the choice $`\alpha _{}=0`$ we do not encounter this singularity. The first part of (C.7) is an integrable singularity for $`|Q|>0`$. In fact if we take $`|Q|\stackrel{~}{𝒱}>4\pi `$ the singularity disappears. However, then $`𝒢_{}(z)𝒢_{}^{}(z)e^{|Q|\psi (z)}`$ will become non integrable. Accordingly, for the singularities in (2.85) to be integrable we require $`\alpha _{}=0,`$ and $`0<|Q|\stackrel{~}{𝒱}<4\pi `$ which implies (4.66) and (4.68).
|
warning/0005/nlin0005003.html
|
ar5iv
|
text
|
# Simple Driven Maps As Sensitive Devices
Abstract
Sensitive dependence of nonlinear systems on initial conditions or parameters can be useful in applications. We propose in this paper that bubbling behavior in simple driven symmetrical maps may be used as a working principle of sensitive devices. The system is stable when there is no input and displays bursting behavior when there is small input. The symmetrical property of the bursting pattern is very sensitive to the bias of the noisy inputs, which makes the system promising for detecting weak signals among noisy environment.
PACS number(s): 05.45.+b;
A common property of many nonlinear systems is their sensitive dependence on initial conditions or parameters. This effect can be useful in applications. For example, the sensitivity of a chaotic system can be used to control its state to unstable periodic orbits embedded in it , in targeting the state of the system to desired points in the state space, to control the system to follow a desired goal dynamics in order to synchronize with another system or to allow a message being encoded in a chaotic series for the purpose of secure communication, only by small modifications of the parameters or state of the chaotic system. The capability of achieving quite different behavior by applying only small perturbations improves greatly the flexibility of a system to be used in various applications.
By definition, sensitivity is referred to as the growth of small perturbations to the system. So, naively, sensitivity of nonlinear systems can be used to design sensor devices. Many systems possess a period-doubling bifurcation when some parameter is varied. Near the onset of a period-doubling bifurcation, any dynamical system can be used to amplify perturbations near half the fundamental frequency. One disadvantage associated with the application of such parameter sensitivity for sensor device purpose is that the control parameter of the system must be located extremely close to the critical value of the bifurcation.
Recently, Böhme and Schwarz proposed to use two identical chaotic systems to construct sensitive devices. In particular, they employed the following symmetrically coupled chaotic systems
$`\dot{x}`$ $`=`$ $`f(x)k(xy)+s_{in},`$ (1)
$`\dot{y}`$ $`=`$ $`f(y)+k(xy)s_{in},`$ (2)
named chaotic bridge as a sensor device. $`s_{in}`$ represents a constant input to be sensed. The coupling gain $`k`$ is chosen near the threshold $`k_c`$ of synchronization, so that for $`s_{in}=0`$, the coupled systems are in synchronization state, and the output $`s_{out}=xy=0`$; while for $`s_{in}0`$, the symmetry of the chaotic bridge is broken, and it may have a large output at some moment. Since the synchronization manifold is transversely stable for $`s_{in}=0`$, there must exist local instabilities in the system in order to obtain amplification of small perturbations. As pointed out by the authors in , in the neighborhood of the boundary $`k_c`$ of synchronization, one can expect the highest sensitivity of the system.
We would like to highlight the connection of the working principle of the above device to the phenomenon of attractor bubbling studied recently\[7-10\]. When $`k`$ is just beyond the threshold $`k_c`$, the synchronization manifold is transversely stable. However, there still exist some invariant sets, such as the unstable periodic orbits embedded in the synchronization manifold, which are transversely unstable. As a consequence, small perturbations in the systems which destroy its synchronization manifold will result in large intermittent bursts from the synchronization manifold, no matter how small the perturbations are. This is the origin of the sensitivity of the above system. The difficulties of application of the system for sensor devices lie in practical implementations. Just like additive perturbations, any parameter mismatches between the systems can also lead to intermittent bursts. Parameter mismatches are inevitable in experiment implementations. This is the reason that intermittent desynchronization was observed beyond the threshold of synchronization in many experiments of synchronization between well matched electrical circuits\[7-10\]. This effect imposes great difficulties in the experimental implementation of the above sensor devices, because inevitable parameter mismatches lead to large output even for $`s_{in}=0`$. The above devices can work only if the two systems are ideally identical, which is extremely difficult to realize. On the other hand, small external noise can also result in large bursts when $`S_{in}=0`$, which makes it very difficult to tell a signal from noise which is always present in the practical environment.
To avoid the above difficulties, we propose in the following to use simple driven systems as sensor devices.
Attractor bubbling and on-off intermittency are common behaviors that occur in coupled nonlinear systems which possess an invariant manifold. They can be achieved in very simple parametrically driven one-dimensional maps
$$y_{n+1}=z_nf(y_n),$$
(3)
where $`f(0)=0,f(y)/y|_00,`$ and $`z_n=ax_n>0`$ is random or chaotic driving signal with density function $`\rho _z`$ and $`a`$ is a parameter. For the purpose of the application of the systems as sensitive devices, we require the maps to have odd symmetry, i.e. $`f(y)=f(y)`$.
The stability of the invariant manifold $`y=0`$ is governed by the linear equation
$$y_{n+1}=z_ny_n,$$
(4)
which describes the evolution of small perturbations transverse to the invariant line $`y=0`$. Here $`f(y)/y|_0`$ is absorbed into the parameter $`a`$. The transverse Lyapunov exponent $`\lambda `$ of the invariant manifold defined as
$$\lambda =\underset{N\mathrm{}}{lim}\frac{1}{N}\underset{n=1}{\overset{N}{}}\mathrm{ln}z_n=<\mathrm{ln}z>$$
(5)
determines the stability of the invariant manifold. The critical point $`a_c`$ at which $`\lambda =0`$ is the onset point of on-off intermittency. For $`z_n`$ being a uniform random driving signal on $`(0,a)`$, $`\rho _z=1/a`$, and $`<\mathrm{ln}z>=\mathrm{ln}a1=0`$ gives $`a_c=e`$. Just above the onset point, $`aa_c`$, the random driven system displays universal features of on-off intermittency behavior, which are unaffected by the form of the confining nonlinearity. The nonlinearity of the system serves to bound or reject the dynamics back towards small values of $`y`$ after bursts. For $`a`$ below $`a_c`$, the invariant manifold $`y=0`$ is stable, but the stability is quite weak if $`a`$ is near $`a_c`$. Attractor bubbling occurs in the system when there are inputs of perturbations such as noise.
For a sensitive device, it should be stable when there is no input, and is expected to produce large outputs when there are small perturbations. In this paper, we are not considering the on-off intermittency behavior with $`aa_c`$. For the purpose of sensor purpose, we employ the bubbling behavior with $`a<a_c`$. The sensor system reads
$$y_{n+1}=z_nf(y_n)+s_{in},$$
(6)
and $`s_{out}=y`$. When there is no input, i.e. $`s_{in}=0`$, $`y=0`$ is a stable solution, and the output $`s_{out}=0`$. Since the critical parameter $`a_c`$ and the evolution of small perturbations are independent of the form of the nonlinearity, one can choose a map which is simple for implementation. For example, we employ a piecewise linear map
$$f(y)=\{\begin{array}{cc}\frac{c_1}{c_2}(c_1c_2y),\hfill & y<c_1,\hfill \\ y,\hfill & |y|c_1,\hfill \\ \frac{c_1}{c_2}(c_1+c_2y),\hfill & y>c_1,\hfill \end{array}$$
(7)
where $`c_1`$ and $`c_2`$ are two parameters, as shown in Fig. 1.
If there is no input, $`s_{in}=0`$, staring from a small initial condition $`y_0=p`$, we have $`y_n=z_{n1}z_{n2}\mathrm{}z_0p=k_np`$. The average order of the factor $`k_n`$, which can be defined as $`<\mathrm{ln}k_n>`$, decreases linearly with $`n`$ as
$$<\mathrm{ln}k_n>=n<\mathrm{ln}z>=n(\mathrm{ln}a1).$$
(8)
Now, suppose there is a small constant input $`s_{in}=p`$, starting from $`y_0=0`$, the evolution of the output reads
$$\begin{array}{c}y_1=(z_0+1)p=k_1p,\\ y_2=(z_1z_0+z_1+1)p=k_2p,\\ \mathrm{}\\ y_n=(z_{n1}\mathrm{}z_0+\mathrm{}+z_{n2}z_{n1}+z_{n1}+1)p=k_np,\end{array}$$
(9)
if $`\mathrm{max}(|y_n|)=|p|(1+a+\mathrm{}a^n)=|p|\frac{a^{n+1}1}{a1}c_1`$, or, $`nn_c=\text{int}[\mathrm{ln}(c_1(a1)/|p|+1)/\mathrm{ln}a]1`$ , where $`\text{int}[x]`$ is the interger part of a real number $`x`$. The average order of $`k_n`$, $`<\mathrm{ln}k_n>`$, cannot be obtained analytically as that for $`s_{in}=0`$ in Eq. 8. A numerical estimation of $`<\mathrm{ln}k_n>`$ is carried out with $`10^6`$ samples of $`k_n`$. Unlike the case $`s_{in}=0`$, it is an increasing function of $`n`$, as shown in Fig. 2.
When $`n>n_c`$, $`y_n`$ has nonvanishing probability to exceed $`y_n>c_1`$. Following Böhme and Schwarz, a measure $`s`$ for the sensitivity of the system to constant input can be defined as
$$s=\frac{\underset{n}{\mathrm{max}}|y_n|}{|s_{in}|}.$$
(10)
Since $`\mathrm{max}(k_n)=\frac{a^{n+1}1}{a1}`$ increases exponentially with $`n`$, an infinitely small input value can create a finite output. The largest output $`\underset{n}{\mathrm{max}}|y_n|=ac_1`$ is due to the confinement of the nonlinearity of the map. The sensitivity $`s`$ of the system thus goes to infinity and is referred to as supersensitivity.
An important question concerning the system is the time needed for small input to produce a large output. We examine the time $`N`$ for a small input $`p`$ to produce for the first time an output $`s_{out}c_1`$. The distribution $`P(N)`$ of $`N`$ is the following probability
$$P(N)=\text{Prob}(\underset{i=1}{\overset{N1}{}}y_i<c_1y_Nc_1).$$
(11)
By defining the event $`E_N=\underset{i=1}{\overset{N}{}}y_i<c_1`$ and the corresponding probability $`\mathrm{\Lambda }_N=\text{Prob}(E_N)`$, it follows that
$$P(N)=\mathrm{\Lambda }_N\mathrm{\Lambda }_{N1},$$
which is a function of both $`a`$ and $`p`$. In principle, $`\mathrm{\Lambda }_N`$ can be evaluated by the joint density $`\mathrm{\Phi }(K)`$ of $`k_1,k_2,\mathrm{},k_N`$, namely,
$$\mathrm{\Phi }(K)=\frac{\rho (z_0)\rho (z_1)\mathrm{}\rho (z_{N1})}{|J|}=\frac{\rho (k_11)\rho (\frac{k_21}{k_1})\mathrm{}\rho (\frac{k_N1}{k_{N1}})}{k_1k_2\mathrm{}k_{N1}},$$
where $`J`$ is the Jacobian of the transformation between $`k_i`$’s and $`z_i`$’s defined in Eq. (9). Specifically, one has
$$\mathrm{\Lambda }_N=_1^b𝑑k_N_{(k_N1)/a}^b\rho (\frac{k_N1}{k_{N1}})\frac{dk_{N1}}{k_{N1}}\mathrm{}_{(k_{i+1}1)/a}^b\rho (\frac{k_{i+1}1}{k_i})\frac{dk_i}{k_i}\mathrm{}$$
$$\times _{(k_21)/a}^b\rho (\frac{k_21}{k_1})\rho (k_11)\frac{dk_1}{k_1}$$
where $`b=c_1/p`$ if $`i>n_c`$ and $`b=\frac{a^{i+1}1}{a1}`$ if $`in_c`$. However, to the best of our knowledge, a closed-form solution of the above integral for any $`N`$ is not available.
In the following, we are going to carry out some simulations. We specify $`c_1=1`$ and $`c_2=2`$ in these simulations. In Fig. 3, as an example, the output sequence is shown for $`s_{in}=1.0\times 10^4`$ at $`a=2.6`$. The dashed lines indicate the switch on and off of the constant input. The output in the presence of input is a intermittent process, similar to the result of the chaotic bridge in .
In the next simulation, we estimate $`P(N)`$ for different values of $`a`$ and $`p`$, as shown in Fig. 4. It is seen that the distributions peak at rather small $`N`$ values, and after the peak, they decrease exponentially. The average time $`<N>`$ for first putting out $`y_Nc_1`$ is also evaluated as a function of $`a`$ and $`p`$ in Fig. 5(a) and (b), respectively. So, on average, the closer the $`a`$ to $`a_c`$ and the larger the input $`p`$, the quicker the system reaches a large output.
The simulation results show that this system may be used as a detector for weak signal. However, the above discussion is only valid in a noise-free environment. In the practical application of the system as a detector, external noise is unavoidable. The system now reads
$$y_{n+1}=z_nf(y_n)+s_{in}+e_n,$$
(12)
where $`e_n`$ denotes external noise. It is plausible to assume that $`e_n`$ has vanishing mean value and a Gaussian distribution $`\sigma N(0,1)`$, with a standard deviation $`\sigma `$. The behavior of the system in the presence of noise is quite different from the noise-free case, because bubbling occurs even without a signal $`s_{in}`$. Very small external noise can also lead to large output of the system, as illustrated in Fig. 6 (a) with $`a=2.6`$ and $`\sigma =1\times 10^4`$. In this system with $`a<a_c`$, bursting behavior always means that there are some inputs to the system, and the largest output will be independent of the inputs. The problem now becomes whether we can distinguish that the input is a meanful signal or just noise, and moreover whether we can detect any significant signal among the white noise environment. As will be shown in the following, this system is quite promising for this task, because the symmetrical property of the bursting behavior is very sensitive to the bias of the inputs.
This sensitivity is due to the odd symmetry of the map. An inspection of Fig. 6(a) reveals that the number of large bursts to positive and negative values is quite symmetric for inputs of white noise. When a small positive constant input $`s_{in}=p=0.3\times 10^4`$ is present along with the noise, the symmetry is clearly broken, as seen in Fig. 6(b). Note that the constant input $`p`$ is much smaller than the noise level $`\sigma `$ in this example. This result indicates that the symmetry of burst is quite sensitive to the bias of the total inputs $`s_{in}+e_n`$ of the system. To characterize the symmetry breaking property quantitatively, we introduce the degree of asymmetry $`D_{asy}`$ as
$$D_{asy}=\frac{N_+N_{}}{N_++N_{}},$$
(13)
where $`N_+`$ ($`N_{}`$) is the number of large burst ($`|y|1`$) to positive (negative) values during a period of observation time $`T`$. Because of the symmetry of the map and that of the white noise, one can expect that $`D_{asy}0`$ for white noise inputs. For positive (negative) constant inputs in the noise-free case, it is clear that $`D_{asy}=1(1)`$.
A comparison between Fig. 6(a) and (b) also shows that large bursts occur more frequently when constant input is present with the noise. We define the bursting frequency $`F`$ as
$$F=\frac{N_++N_{}}{T}.$$
(14)
We expect that $`F`$ increases with larger constant input $`p`$ and larger noise level $`\sigma `$.
A measure of the significance of a constant signal among the noise can be the “signal to noise ratio” $`R=p/\sigma `$. In the following, simulations are carried out to examine the dependence of $`D_{asy}`$ and $`F`$ on $`R`$ for different noise level $`\sigma `$ and system parameter $`a`$. The results shown in Fig. 7 are obtained with $`T=2\times 10^6`$. The result of $`D_{asy}`$ is very interesting: as a function of $`R`$, $`D_{asy}`$ is independent of the noise level $`\sigma `$ and is not sensitive to parameter $`a`$. Whether a constant signal embedded in the noise environment can be detected depends only on its significance with respect to the noise level. However, $`a`$ and $`\sigma `$ have effect on the bursting frequency $`F`$, as seen from Fig. 7(b). A combination of Fig. 7(a) and (b) makes it possible to determine the noise level and $`R`$, and thus to estimate the amplitude of the constant signal in a observation.
To get a good estimation of $`D_{asy}`$ and $`F`$, $`T`$ should be large enough. In practice, one may not expect a signal that stays constant for so long a time. The term constant signal is a concept relative to the time scale of the detector, and the time scale can be controlled in the implementation. In numerical experiment of this dimensionless system, we can simulate a shorter signal (or a “slower” detector) with smaller $`T`$, e.g., $`T=2000`$. In this case, $`D_{asy}`$ has large fluctuations, as shown in Fig. 8(a) where $`D_{asy}`$ of 20 realizations of the driving $`z_n`$ and white noise $`e_n`$ of the system are plotted for each $`R`$ value. When $`R`$ is getting larger, more points coincide at $`D_{asy}=1`$. A good way to examine the fluctuation behavior is to construct a histogram of $`D_{asy}`$, as shown in Fig. 8(b) for $`a=2.6,\sigma =10^4`$. The results show that even for quite short signal and low $`R`$ value, $`D_{asy}`$ has very high probability near $`D_s=1`$. An implication of the results is that several detectors can be used at the same time to detect and confirm a short and weak signal embedded in white noise.
As an example of a little more realistic input signal, we present the response of the system to the noisy input $`A\mathrm{sin}(0.003n)+e_n`$, where the noise level is $`\sigma =1\times 10^4`$ and the amplitude of the sine wave is $`A=0.3\times 10^4`$. Both the total input and the output of the system are displayed in Fig. 9. The bursting feature reflects the weak wave among the noise quite clearly.
In summary, we have shown that a kind of very simple driven symmetrical maps below the onset point of on-off intermittency have two distinguishing features of (i) being stable at the invariant state $`y=0`$ and (ii) being sensitive to small input. In practice, the environment cannot be noise-free, and the systems exhibit bubbling behavior in the presence of noise. Another interesting and useful property of the systems is that the bursting pattern is symmetrical for white noise input, and the symmetry is broken when there is signal among the noise environment. The significance of the signal is manifested by the degree of asymmetry in the bursting pattern. These features make them promising candidates for designing sensitive devices.
Although our study is based on numerical simulations of a mapping model, it should be noted that system response to small inputs is governed by its linearized equation, and the nonlinearity only serves to keep the system bounded. Many long time properties shown above thus are universal in a class of driven systems possessing odd symmetry. The following can be advantages for such systems when considering applications in sensitive devices:
1) The sensitivity is maintained in a large range of parameter below the critical point. This avoids the difficulty of locating parameter in a very small neighborhood of a bifurcation point in a period-doubling system.
2) The sensitivity of the system is, in principle, infinite. In a noise-free environment, an infinitesimal input signal can produce a finite output. When noise is present, weak signal can also be manifested by the asymmetry in the bursting pattern. The degree of asymmetry depends on the significance of the signal with respect to the noise. The sensitive behavior is universal for different forms of nonlinearity of the systems, as well as for different form of driving signals. This is very useful because one can thus choose a system that is simple and easy to implement in practice.
It could be meaningful to consider implementation of such simple systems and explore their application in small signal detection. Since in applications, pivotal role is played by the symmetry properties of the system, one should take care to maintain such properties. In order to avoid perturbations which may make the system appreciably asymmetric, one should avoid using different parameters for the two symmetrical parts of the systems. Also, one should note that it takes longer for the system to produce static output states for lower level of inputs. There seems to be a frequency cutoff associated with the input levels and the relaxation time of the systems. Above the cutoff, the small signal in the noise can no longer be manifested by clear asymmetry in the bursting pattern. Such limits should be taken into consideration in applications.
Acknowledgements:
This work was supported in part by research grant RP960689 at the National University of Singapore. CZ was supported by NSTB.
References
E. Ott, C. Grebogi, and J. Yorke, Phys. Rev. Lett. 64, 1196(1990).
T. Shinbrot, E. Ott, C. Grebogi, and J. Yorke, Phys. Rev. Lett. 65, 3215(1990); T. Shinbrot, et al, Phys. Rev. Lett. 68, 2863 (1992).
Y. C. Lai, C. Grebogi, Phys. Rev. E 47, 2357 (1993).
S. Hayes, C. Grebogi, and E. Ott, Phys. Rev. Lett. 70, 3031(1993); S. Hayes, C. Grebogi, E. Ott, and A. Mark, Phys. Rev. Lett. 73, 1781(1994); E. M. Bollt and M. Dolnik, Phys. Rev. E 55, 6404 (1997).
K. Wiesenfeld and B. McNamara, Phys. Rev. Lett. 55, 13(1985).
F. Böhme and W. Schwarz, in Nonlinear Dynamics of Electronic Systems, eds by A.C. Davies and W. Schwarz (World Scientific, Singapore, 1994), pp 281.
P. Ashwin, J. Buescu, and I. Stewart, Phys. Lett. A 193, 126(1994).
J. F. Heagy, T. L. Carroll, and L. M. Pecora, Phys. Rev. E 52, R1253(1995).
D. J. Gauthier and J. C. Bienfang, Phys. Rev. Lett. 77, 1751(1996).
S. C. Venkataramani, B. R. Hunt, E. Ott, D. J. Gauthier, and J. C. Bienfang, Phys. Rev. Lett. 77, 5361(1996).
N. Platt, E. A. Spiegel, and C. Tresser, Phys. Rev. Lett. 70, 279 (1993).
J. F. Heagy, N. Platt, and S. M. Hammel, Phys. Rev. E 49, 1140 (1994).
Figure Captions
Fig. 1 The piecewise linear map $`f(y)`$. Here $`c_1=1`$ and $`c_2=2`$.
Fig. 2 $`<\mathrm{ln}k_n>`$, the average order of $`k_n`$, as a function of $`n`$. It decreases linearly for the case without input (plot a) and increases for the case with input (plot b).
Fig. 3 An illustration of the output process of the sensitive system at $`a=2.6`$. The constant input is $`p=10^4`$, and is switched on and off alternately for every 500 iterations, as shown by the dashed lines.
Fig. 4 Numerically evaluated distribution of $`N`$. The parameters of the plots are: (a) $`a=2.5,p=10^3`$, (b) $`a=2.5,p=10^4`$ and (c) $`a=2.6,p=10^4`$.
Fig. 5 (a) Average value of $`N`$ as a function of $`a`$ with $`p=10^3`$, and (b) as a function of $`p`$ with $`a=2.6`$.
Fig. 6 (a) Output of the system when only white noise is present as input. (b) Output of the system when constant signal $`p=0.3\sigma `$ is present along with the noise. The parameters are $`a=2.6,\sigma =1\times 10^4`$. Noting the change of symmetrical property of the bursting pattern of the system.
Fig. 7 (a) Degree of asymmetry $`D_{asy}`$ as a function of “signal to noise ratio” $`R`$ for different $`a`$ and $`\sigma `$. (b) Bursting frequency $`F`$ as a function of $`R`$ for different $`a`$ and $`\sigma `$. (b) shares the same legends of (a). The results are obtained from observation during a period of time $`T=2\times 10^6`$.
Fig. 8 (a) $`D_{asy}`$ of short time observation, $`T=2000`$. (b) Normalized histograms of $`D_{asy}`$ for different $`R`$ value. The histograms are constructed with 50000 observations for each $`R`$ value. The parameters are $`a=2.6,\sigma =1\times 10^4`$.
Fig. 9 (a) A weak sine wave embedded in the noise. (b) The response of the detector to the inputs of (a).
|
warning/0005/hep-ex0005007.html
|
ar5iv
|
text
|
# Mighty MURINEs: Neutrino Physics at Very High Energy Muon Colliders To appear in Proc. HEMC’99 Workshop – Studies on Colliders and Collider Physics at the Highest Energies: Muon Colliders at 10 TeV to 100 TeV; Montauk, NY, September 27-October 1, 1999, http://pubweb.bnl.gov/people/bking/heshop/ . This work was performed under the auspices of the U.S. Department of Energy under contract no. DE-AC02-98CH10886.
## I Introduction: the Role of Mighty MURINES
The dominant motivation for high energy muon colliders (HEMCs) is unquestionably to explore elementary particle physics at many-TeV energy scales. For the sake of completeness, however, this paper instead discusses what would be the most promising subsidiary fixed target physics program, namely, the parasitic use of the free and profuse neutrino beams at HEMCs to provide complementary precision studies of high energy physics (HEP) at lower energies. Perhaps, this might complement collider studies in fostering new and helpful insights into the properties of elementary particles.
Neutrino interactions have unique potential for precision HEP studies because they only participate in the weak interaction. Today’s neutrino beams from pion decays lack the intensity to fully exploit this potential but future MUon RIng Neutrino Experiments (MURINEs), using neutrino beams from the decays of muons in a muon collider or other storage ring, hold the promise of neutrino beams that are several orders of magnitude more intense than today’s beams bjkthesispaper . The first MURINEs may well be muon storage rings dedicated to neutrino production geer (“neutrino factories”) while the collider rings of any first generation muon colliders will also make excellent MURINEs. The topic of this paper is MURINEs at very high energies and these will be referred to as “Mighty MURINEs” <sup>1</sup><sup>1</sup>1this is a play on words alluding to the venerable cartoon character Mighty Mouse through the dictionary definition of “murine” as “to do with mice”..
At a minimum, Mighty MURINEs will improve on previous MURINEs in providing much useful bread-and-butter precision HEP to feed the hungry masses of HEP experimentalists with aversions to collider mega-experiments. More interestingly, there are a couple of plausible scenarios under which they might do much more; namely (i) if quark mixing and/or B physics offer more than predicted by our naive prejudices as parameterized in the standard model (SM) of elementary particles, and (ii) if leptoquarks or other exotica begin to emerge at or below the 100 GeV energy scale.
The following section surveys the experimental conditions and parameters that might be found at Mighty MURINEs. This is followed by an overview of the potential physics analyses and by three sections going into more detail on the most interesting topics: one each on exploiting Mighty MURINEs as B factories, on the possibilities for quark mixing studies and on the potential for heavy particle production up to the 100 GeV scale.
## II Experimental Overview
### II.1 The Neutrino Beams
Neutrinos are emitted from the decay of muons in the collider ring:
$`\mu ^{}`$ $``$ $`\nu _\mu +\overline{\nu _\mathrm{e}}+\mathrm{e}^{},`$
$`\mu ^+`$ $``$ $`\overline{\nu _\mu }+\nu _\mathrm{e}+\mathrm{e}^+.`$ (1)
As is illustrated in figure 1, the thin pencil beams of neutrinos for experiments will be produced from the most suitable long straight sections in the collider ring or, possibly, in the accelerating rings. These will be referred to as the production straight sections. The divergence of the neutrino beam is typically dominated by the decay opening angles of the neutrinos rather than the divergence of the parent muon beam. Relativistic kinematics boosts the forward hemisphere in the muon rest frame into a narrow cone in the laboratory frame with a characteristic opening half-angle, $`\theta _\nu `$, given in obvious notation by
$$\theta _\nu \mathrm{sin}\theta _\nu =1/\gamma _\mu =\frac{m_\mu c^2}{E_\mu }\frac{10^4}{E_\mu [\mathrm{TeV}]}.$$
(2)
For the example of 5 TeV muons, the neutrino beam will have an opening half-angle of approximately 0.02 mrad.
The large muon currents and tight collimation of the neutrinos results in such intense neutrino beams that potential radiation hazards bjkthesispaper ; hemc99nurad are a serious design issue for the neutrino beam-line and even for the less intense neutrino fluxes emanating from the rest of the collider ring.
### II.2 Luminosities at Neutrino Experiments
For a cylindrical experimental target extending out from the beam center to an angle $`\theta _\mu =1/\gamma _\mu `$, the luminosity, $``$, is proportional to the product of the mass depth of the target, $`l`$, and the number of muon decays per second in the beam production straight section, according to:
$$[\mathrm{cm}^2.\mathrm{s}^1]=\mathrm{N}_{\mathrm{Avo}}\times \mathrm{f}_{\mathrm{ss}}\times \mathrm{n}_\mu [\mathrm{s}^1]\times l[\mathrm{g}.\mathrm{cm}^2],$$
(3)
where $`\mathrm{f}_{\mathrm{ss}}`$ is the fraction of the collider ring circumference occupied by the production straight section, $`n_\mu `$ is the rate at which each sign of muons is injected into the collider ring (assuming they all circulate until decay rather than being eventually extracted and dumped) and the appropriate units are given in square brackets in this equation and all later equations in this paper. The proportionality constant is Avagadro’s number, $`\mathrm{N}_{\mathrm{Avo}}=6.022\times 10^{23}`$, since exactly one neutrino per muon is emitted on average into the boosted forward hemisphere, i.e. each muon decay produces two neutrinos and half of them travel forwards in the muon rest frame.
Because Avagadro’s number is so large, the luminosities at Mighty MURINEs will be enormous compared to those at collider experiments. The luminosities for a reasonable scenario using the workshop’s straw-man parameter sets hemc99intro are given in table 1 along with, for comparison, the final design goal luminosity for the HERA ep collider. It can be seen that, roughly speaking, Mighty MURINEs might achieve of order a million times the luminosity of HERA. An “accelerator year’s” running – $`10^7`$ seconds – at the 50 TeV MURINE’s luminosity of $`2\times 10^{37}\mathrm{cm}^2.\mathrm{s}^1`$ would amount to an impressive integrated luminosity of 200 inverse attobarns per year while the even bigger straw-man luminosity at 5 TeV, $`1\times 10^{39}\mathrm{cm}^2.\mathrm{s}^1`$, requires a luminosity prefix that is even less familiar to the HEP community: 10 inverse zeptobarns per year.
### II.3 Center-of-Mass Energies for the Neutrino-Nucleon Interactions
It will be seen in section VI that HERA provides a useful comparison for some of the physics capabilities of Mighty MURINEs, particularly since the maximum center-of-mass energies, $`\mathrm{E}_{\mathrm{CoM}}`$, at the highest energy HEMCs might even be comparable to those at the HERA collider. The electron-proton $`\mathrm{E}_{\mathrm{CoM}}`$ at the collider is given by relativistic kinematics as
$$\mathrm{E}_{\mathrm{CoM}}^{\mathrm{HERA}}=2\sqrt{\mathrm{E}_\mathrm{p}\mathrm{E}_\mathrm{e}},$$
(4)
which is 314 GeV for the proton and electron energies of the year 2000 upgrade to HERA, $`\mathrm{E}_\mathrm{p}=820`$ GeV and $`\mathrm{E}_\mathrm{e}=30`$ GeV. For comparison, the MURINE’s $`\mathrm{E}_{\mathrm{CoM}}`$ is
$$\mathrm{E}_{\mathrm{CoM}}^{\mathrm{MURINE}}=\sqrt{2\mathrm{E}_\nu \mathrm{M}_\mathrm{p}\mathrm{c}^2+(\mathrm{M}_\mathrm{p}\mathrm{c}^2)^2},$$
(5)
where the proton mass corresponds to $`\mathrm{M}_\mathrm{p}\mathrm{c}^2=0.938`$ GeV. The neutrino energy can range right up to the muon beam energy, $`\mathrm{E}_\nu ^{\mathrm{max}}=\mathrm{E}_\mu `$, and the energy spectrum seen by the detector is relatively hard numcbook , with an average neutrino energy within the $`1/\gamma _\mu `$ cone that is 49% of the muon beam energy. The comparative center-of-mass energies for Mighty MURINEs and HERA are summarized in table 1.
### II.4 Cross Sections and Event Rates
The event rate in the neutrino detector is a product of the luminosity given in table 1 and the neutrino-nucleon scattering cross-section, which we now discuss.
The predominant interactions of neutrinos and anti-neutrinos at all energies above a few GeV are charged current (CC) and neutral current (NC) deep inelastic scattering (DIS) off nucleons ($`N`$, i.e. protons and neutrons) with the production of several hadrons ($`X`$):
$`\nu (\overline{\nu })+N`$ $``$ $`\nu (\overline{\nu })+X(NC)`$
$`\nu +N`$ $``$ $`l^{}+X(\nu CC)`$
$`\overline{\nu }+N`$ $``$ $`l^++X(\overline{\nu }CC),`$ (6)
where the charged lepton, $`l`$, is an electron if the neutrino is an electron neutrino and a muon for muon neutrinos. The cross-sections for these processes are approximately proportional to the neutrino energy, $`E_\nu `$, with numerical values of quigg :
$$\sigma _{\nu \mathrm{N}}\mathrm{for}\left(\begin{array}{c}\nu CC\\ \nu NC\\ \overline{\nu }CC\\ \overline{\nu }NC\end{array}\right)\left(\begin{array}{c}0.72\\ 0.23\\ 0.38\\ 0.13\end{array}\right)\times 10^{35}\mathrm{cm}^2\times \mathrm{E}_\nu [\mathrm{TeV}].$$
(7)
The number of events in the detector is easily seen to be given by:
$$N_{events}=[\mathrm{cm}^2.\mathrm{s}^1]\times 0.73\times 10^{35}\times 0.49\times \mathrm{E}_\mu [\mathrm{TeV}]\times \mathrm{T}[\mathrm{s}],$$
(8)
where T is the running time, $`0.49\times \mathrm{E}_\mu `$ is the average neutrino beam energy into the detector numcbook and $`0.73\times 10^{35}`$ is the total cross-section-divided-by-energy that is obtained from equation 7 after summing over the NC and CC interactions and averaging over neutrinos and anti-neutrinos.
The final column of table 1 shows the impressive event sample sizes predicted from equation 8: up to of order $`10^{11}`$ events per year in a reasonably sized neutrino target.
### II.5 High Performance Neutrino Detectors for Mighty MURINEs
The unprecedented event samples in small targets at MURINEs will undoubtedly also spark a revolution in neutrino detector design and performance, both to cope with event rates and to fully exploit the physics potential of the beams. An example of a novel general purpose neutrino detector that has been proposed previously nufnal97 for MURINEs is shown in figure 2. The neutrino target is the cylinder at mid-height on the left hand side of the figure. It comprises a stack of equally-spaced CCD tracking planes, oriented perpendicular to the beam and with spacings of order 1 mm, that provides vertex tagging for events with hadrons containing charm or bottom quarks.
The general detector design of figure 2 should remain appropriate for Mighty MURINEs although it would likely be elongated to cope with the more boosted events, including perhaps lengthening the target to several meters to increase the target mass-per-unit-area and, correspondingly, the event rate.
At these higher energies, the target mass-per-unit-area could be increased still further by interspersing thin tungsten disks with the CCD planes. There are two reasons why such a dense, high-Z target becomes more practical than at lower energies: (i) multiple coulomb scattering becomes less important at higher energies, so the tracking resolution is degraded less, and (ii) the narrower pencil beam for Mighty MURINEs allows the disks to have smaller radii than at lower energies – smaller than the characteristic Moliere radius for electromagnetic showers – so it is speculated that electromagnetic showers will not develop to excessively pollute the events, despite the large number of radiation lengths along the axis of the target. (This assumption needs to be checked in more detailed follow-up studies.)
A specific scenario for the neutrino target that gives the 1000 $`\mathrm{g}.\mathrm{cm}^2`$ target mass assumed in table 1 is as follows: a 4 meter long target containing 4000 millimeter-long tracking subunits, where each tracking subunit contains a thin tungsten disk of thickness 118 microns (0.227 $`\mathrm{g}.\mathrm{cm}^2`$) in front of a 100 micron thick CCD pixel detector (0.023 $`\mathrm{g}.\mathrm{cm}^2`$). Each tungsten disk could have a radius of 2 cm to match the beam radius at approximately 1 km from production for a 5 TeV muon beam (as predicted from equation 2). The CCD detectors can be wider than the beam radius to also track particles moving outside the radial extent of the neutrino beam.
The vertex tagging performance of the target in figure 2 is expected nufnal97 to be better than any other existing or planned high-rate detector for heavy quark physics, and should continue to improve with beam energy. Mighty MURINEs should attain close to 100 percent efficiency for both c and b (easier) vertex tagging in the target (excepting all-neutral decay modes, of course) since the average boosted lifetime for TeV-scale charm and beauty hadrons – of order 10 cm – would span many planes of CCD’s. The extremely favorable geometry for vertexing is illustrated in figure 3, where it is also compared with the vertexing geometry at a collider detector.
As with lower energy MURINEs, the detector backing the neutrino target should faithfully reconstruct both CC and NC event kinematics. The lower energy MURINEs provide essentially full particle identification through the muon toroids and dE/dx plus cherenkov radiation in the TPC. The particle ID for long-lived charged hadrons would become more difficult at higher energies although, speculatively, the cherenkov radiation in an elongated TPC might still give effective PID for particle energies up to a couple of hundred GeV – this requires further study. To somewhat compensate, the trajectories of most photons should be very well measured when they convert in the stack of tungsten disks that comprise most of the target mass. This will be particularly helpful for the reconstruction of neutral pions.
## III MURINEs: A New Realm for Neutrino Physics
This subsection gives a brief non-technical overview of the high rate neutrino physics topics expected for MURINEs in general. Topics that might be expanded on with the higher energies of Mighty MURINEs are pointed out in preparation for more detailed discussion in the following sections.
### III.1 Neutrino Interactions with Quarks
Put simply, the DIS interactions of equation 6 involve a simple projectile (the neutrino), interesting interactions (the CC and NC weak interactions) and a complicated target (the nucleon). The bulk of the physics interest lies in the interactions with the quark constituents rather than in the properties of the neutrinos themselves. The complementary analyses that study the potential oscillations of neutrino flavors tend to be more the domain of lower energy MURINEs, at muon energies of order 100 GeV or below, and won’t be discussed further here.
By the TeV-energy scale and above, the CC and NC interactions of equation 6 have become very well described as being the quasi-elastic (elastic) scattering of neutrinos off one of the many quarks (and anti-quarks), q, inside the nucleon:
$`\nu (\overline{\nu })+q`$ $``$ $`\nu (\overline{\nu })+q(NC)`$ (9)
$`\nu +q^{()}`$ $``$ $`l^{}+q^{(+)}(\nu CC)`$ (10)
$`\overline{\nu }+q^{(+)}`$ $``$ $`l^++q^{()}(\overline{\nu }CC).`$ (11)
The CC and NC interactions are mediated through the exchange of a virtual W or Z boson, respectively. All quarks – up quarks (u), down quarks (d) and the smaller “seas” of the progressively heavier strange (s), charm (c) and even beauty (b) quarks – participate in NC scattering interactions of both neutrinos and anti-neutrinos. In contrast, charge conservation specifies the charge sign of the quarks participating in the CC processes as indicated by the labels: $`q^{()}d,s,b,\overline{u},\overline{c}`$ and $`q^{(+)}u,c,\overline{d},\overline{s},\overline{b}`$. The hadrons seen in the detector are produced by the “hadronization” of the final state struck quark at the nuclear distance scale.
### III.2 The Intrinsic Richness of Neutrino Physics
Experimentally, the interaction type of almost all events can be distinguished with little ambiguity by the charge of the final state lepton (and its flavor: i.e. electron or muon): neutral (an unseen neutrino), negative or positive for the 3 respective processes in equation 6.
It is seen that equations 9 through 11 probe 3 different weightings of the quark flavors inside a nucleon, through weak interactions involving both the W and Z. For comparison, only a single and complementary weighting is probed by the best competitive process – the photon exchange interactions of charged lepton scattering experiments at HERA and fixed target facilities. Much of the uniqueness and richness of neutrino scattering physics derives from this variety of interaction processes.
### III.3 Physics Topics at MURINEs
Mighty MURINEs will extend and improve on the already considerable range of unique topics that will have been explored at lower energy MURINEs. The interested reader is referred to reference numcbook for more details. Here we list those topics that will have already been well addressed in earlier MURINEs and comment on any added potential that might be available using the higher energies at Mighty MURINEs.
#### Probing Nucleon Structure
The redundant probes of the proton’s and neutron’s internal structure should provide some of the most precise measurements and tests of perturbative QCD - the theory of the strong interaction - and will also be invaluable input for many analyses at pp colliders. Neutrinos are also intrinsically 100% longitudinally polarized, so experiments with polarized targets could additionally map out the spin structure of the nucleon. Some of these analyses might obtain modest benefits from the higher energies and statistics at Mighty MURINEs.
#### Precision Electroweak Measurements
Besides using W and Z exchange as nuclear probes, the interactions themselves provide important precision tests of the standard model of elementary particles. Two measurements of total interaction cross sections will provide determinations of the fundamental weak mixing angle parameter of the electroweak theory, $`\mathrm{sin}^2\theta _W`$, from (i) the ratio of NC to CC total cross sections and (ii) the absolute cross section for the rarer process of neutrino-electron scattering, which is 3 orders of magnitude less common than neutrino-nucleon scattering. In both cases, the fractional uncertainties in $`\mathrm{sin}^2\theta _W`$ might approach the $`10^4`$ level, which would be complementary and competitive to the best related measurements in collider experiments.
The first of the two measurements will already be systematically limited at MURINEs numcbook so large gains should not be expected at Mighty MURINEs. The situation is not so clear for the electron scattering process, where the higher event statistics could still be beneficial. Speculatively, these higher statistics might also allow the use of liquid hydrogen targets with improved experimental capabilities.
#### Charm and Beauty Factories
MURINEs of all energies will be excellent charm factories, with of order 1% to 10% of the events containing a charmed hadron, depending on the beam energy. TeV-scale MURINEs and above will also produce enough B hadrons to be considered as beauty factories and Mighty MURINES might be very impressive B factories, as will be discussed in section IV.
#### Quark Mixing Studies
There is much additional interest in experimentally partitioning the CC event sample to obtain the partial cross sections for the various possible quark flavor transition combinations represented by the $`q^{()}`$ and $`q^{(+)}`$ symbols in equations 10 and 11. MURINEs, in general, should have the quark-tagging capability to separate the various quark flavor contributions, as was discussed in the preceding section. Mighty MURINES, with their extra capability for producing heavy final-state quarks, could make great strides beyond previous MURINEs for these “quark mixing” studies, as will be expanded on in section V.
#### Rare and Exotic Processes
The higher statistics and, particularly, energies available at Mighty MURINEs would clearly expand the scope for studies of rare processes and searches for exotic processes. This will be covered in section VI.
## IV Mighty MURINEs as B Factories
The charged current production of b quarks off the light quarks in the nucleon is heavily suppressed due to small off-diagonal CKM matrix elements. However, the fraction of neutrino-induced events containing B hadrons rises rapidly with energy numcbook due to the decreasing threshold suppression for two higher-order processes involving gluons in the initial state:
1. $`\mathrm{b}\overline{\mathrm{b}}`$ pair production in neutral current interactions:
$$\nu N\nu b\overline{b}X.$$
(12)
2. charged current production of $`\mathrm{c}\overline{\mathrm{b}}`$ and $`\mathrm{b}\overline{\mathrm{c}}`$ from the charged current interactions of neutrinos or anti-neutrinos:
$$\nu Nl^{}\overline{b}cX$$
(13)
and
$$\overline{\nu }Nl^+b\overline{c}X,$$
(14)
respectively.
Preliminary estimates Timcomm ; numcbook for the fraction of events from each of these processes are tabulated versus neutrino energy in table 2. The second of the two processes is seen to be less common than the first. To compensate, it provides an extremely pure and efficient tag to distinguish between b and anti-b quark production: b production is always accompanied by a positive primary lepton (from anti-neutrino interactions) and anti-b production by a negative primary lepton (from neutrino interactions). This will be very helpful for studies of oscillations of $`\mathrm{B}_0`$’s and $`\mathrm{B}_\mathrm{S}`$’s.
Combining the numbers in tables 1 and 2 predicts event rates of perhaps $`10^8`$ to $`10^9`$ B’s per year at Mighty MURINEs. This is intermediate between the expectations of the $`\mathrm{e}^+\mathrm{e}^{}`$ B factory experiments ($`10^7`$ events/year) and the hadron B factories, HERA-B, BTeV and LHC-B (up to $`10^{11}`$ events/year, with up to a few times $`10^9`$ events tagged for analysis). As already mentioned, however,the vertexing capabilities and other experimental conditions at Mighty MURINES should be superior in some aspects to those at the $`\mathrm{e}^+\mathrm{e}^{}`$ B factories and vastly superior to the very difficult experimental conditions at the hadronic B factories.
Three speculative examples of B analyses that would benefit from the unique experimental conditions at Mighty MURINEs are:
* the superior vertexing capabilities should be ideal for studying the expected fast oscillations of $`\mathrm{B}_\mathrm{s}`$’s, perhaps following up on previous B factories with more precise measurements of the oscillation frequency and greater sensitivity to any asymmetry in the $`\mathrm{B}_\mathrm{s}`$ and $`\overline{\mathrm{B}_\mathrm{s}}`$ decay rates
* some studies of the B baryons, $`\mathrm{\Lambda }_b`$, $`\mathrm{\Xi }_b^{}`$ and $`\mathrm{\Xi }_b^0`$, which are not produced in $`e^+e^{}`$ B factories, may also plausibly be best performed at a Mighty MURINE
* it might have a chance Bigi to measure the branching ratio for the all-neutral rare decay $`B_d\pi ^0\pi ^0`$, which is expected to be of order $`10^6`$. This would provide an estimate for the otherwise problematic “penguin-diagram pollution” in the analogous charged pion decay $`B_d\pi ^+\pi ^{}`$, and this could go some way to resurrecting the charged decay mode as one of the central CKM processes at B factories. However, observing the neutral decay mode does not look feasible at any future B factories other than Mighty MURINEs.
The final process deserves some further explanation since the decay itself doesn’t provide a vertex. However, the close to 100% vertex reconstruction efficiency could instead act as a veto to reduce the backgrounds from the pair-produced B’s in neutral current interactions. The signature for the signal process would be a neutral current event with (i) a single vertex from the other B, (ii) 4 converted gammas reconstructing to 2 high energy $`\pi ^0`$’s that, in turn, reconstruct to the $`B_d`$ mass and (iii) no suspicion of another B or charm vertex. Hence, the analysis – although admittedly still exceedingly difficult – would benefit from both the exceptional vertexing and neutral pion reconstruction at Mighty MURINEs.
Therefore, to summarize this section, the initial expectation is that Mighty MURINEs should be able to do follow-up precision studies in at least some of the most difficult areas of B physics, even after the other B factories have run.
## V Quark Mixing: Measurements Beyond the B Factories
This section enlarges on the theoretical interest in measurements of quark mixing at MURINEs and also provides detail on the central role that Mighty MURINEs could assume if they reached sufficient energies to begin producing top quarks.
### V.1 Theoretical Interest in the CKM Matrix
Quark mixing is one of the least understood and most intriguing parts of elementary particle physics, and the confinement of quarks inside hadrons also makes it one of the hardest areas to study. The CC weak interaction for quarks differs from this interaction for leptons by mixing quarks from different families, i.e. any positively charged quark, $`q^{(+)}u,c,t`$, has some probability of being converted into any of its negatively charged counterparts, $`q^{()}d,s,b`$, and vice versa, rather than being uniquely associated with its same-family counterpart (i.e. $`du`$, $`sc`$ and $`bt`$). This feature is accommodated in the standard model of elementary particle physics through the unitary 3-by-3 Cabbibo-Kobayashi-Maskawa (CKM) matrix, $`\mathrm{V}_{ij}`$, that connects the positively charged $`q_i^{(+)}`$’s with the three $`q_j^{()}`$’s. (The corresponding matrix for leptons is trivially the 3-by-3 identity matrix.)
Apart from verifying that the SM description for quark mixing is indeed correct, the CKM matrix has additional interest through its hypothesized association with CP violation: the puzzling phenomenon that some particle properties, such as decay rates, have been found to differ slightly from those of the corresponding anti-particles. CP violation could also have cosmological implications; it has been invoked as one possible explanation for the comparative scarcity of anti-matter in the universe. The presence of a complex phase in the CKM matrix is the largely untested standard model explanation/parameterization for CP violation.
It is a testament to the perceived importance of the CKM matrix and CP violation that much of today’s experimental HEP effort is devoted such studies, including B and phi factory colliders, LHC-B, HERA-B, B-TeV, K-TeV and many others. Neutrino-nucleon scattering has impressive potential to augment these studies but, until the arrival of MURINEs, it will be held back by inadequate beam intensities.
### V.2 Quark Mixing Studies at MURINEs
The new neutrino studies at MURINEs will be much cleaner theoretically than most of the other experimental processes and will offer much complementary information.
Figure 4 is the Feynman diagram for the basic scattering process, showing that the CKM matrix element $`V_{qq^{}}`$ participates as an amplitude in the W-quark coupling. As a fundamental difference between $`\nu `$N DIS and all other types of CKM measurements, the scattering process involves the interaction of an external W boson probing the quarks inside a nucleon rather than an internal W interaction inside a hadron that, e.g., initiates a B decay. (In principle, the HERA ep collider could also do measurements involving an external W exchange, but these turn out not to be feasible in practice HERA\_CKM .) This is a substantial theoretical advantage for neutrino scattering because the “asymptotic freedom” property of QCD predicts quasi-free quarks with reduced influence from their hadronic environment at the higher 4-momentum-transfer (Q) scales available with an external W exchange.
The CKM measurements at MURINEs will be complementary to, say, the CKM measurements at B factories in that the measurements are of the magnitudes of individual CKM matrix elements rather than of interference terms involving pairs of elements. As can be inferred from figure 4, this arises because the differential cross-sections, $`\frac{d\sigma }{dx}(q_iq_j)`$, for the quark transitions are proportional to the absolute squares of the CKM elements:
$`{\displaystyle \frac{d\sigma }{dx}}(dc)xd(x)|V_{cd}|^2\times T(m_c,x)`$ (15)
$`{\displaystyle \frac{d\sigma }{dx}}(sc)xs(x)|V_{cs}|^2\times T(m_c,x)`$ (16)
$`{\displaystyle \frac{d\sigma }{dx}}(ub)xu(x)|V_{ub}|^2\times T(m_b,x)`$ (17)
$`{\displaystyle \frac{d\sigma }{dx}}(cb)xc(x)|V_{cb}|^2\times T(m_b,x)`$ (18)
$`{\displaystyle \frac{d\sigma }{dx}}(dt)xd(x)|V_{td}|^2\times T(m_t,x)`$ (19)
$`{\displaystyle \frac{d\sigma }{dx}}(st)xs(x)|V_{ts}|^2\times T(m_t,x)`$ (20)
$`{\displaystyle \frac{d\sigma }{dx}}(bt)xb(x)|V_{tb}|^2\times T(m_t,x),`$ (21)
where the Bjorken scaling variable, $`x`$, with $`0<x<1`$, is a relativistically invariant quantity that can be reconstructed for each event and, roughly speaking, measures the fraction of the nucleon’s 4-momentum carried by the struck quark. The respective initial-state quark densities as functions of Bjorken $`x`$ have been labeled $`d(x)`$, $`s(x)`$, $`u(x)`$, $`c(x)`$ and $`b(x)`$, and the $`T(m_q,x)`$’s are threshold suppression factors due to the masses, $`m_q`$, of the final-state quarks.
The $`T(m_q,x)`$ mass suppression factors are zero or much less than unity for all $`x`$ below neutrino energies that can readily supply enough CoM energy to produce the massive final state quarks. From equation 5, the $`T(m_q,x)`$’s will asymptotically approach unity only for muon beam energies such that:
$$m_q^22M_pE_\mu /c^2+M_p^2.$$
(22)
This places the following lower bounds on beam energies for the efficient production of charm, beauty and top quarks, respectively:
$`\mathrm{m}_\mathrm{c}1.31.7\mathrm{GeV}/\mathrm{c}^2`$ $`\mathrm{E}_\mu 1\mathrm{GeV}`$ (23)
$`\mathrm{m}_\mathrm{b}5\mathrm{GeV}/\mathrm{c}^2`$ $`\mathrm{E}_\mu 13\mathrm{GeV}`$
$`\mathrm{m}_\mathrm{t}175\mathrm{GeV}/\mathrm{c}^2`$ $`\mathrm{E}_\mu 16\mathrm{TeV}.`$
The extraction of the CKM matrix elements from the MURINEs’ experimental data will be analogous to, but vastly superior to, current neutrino measurements of $`|\mathrm{V}_{\mathrm{cd}}|`$, the only CKM matrix element that is currently best measured in neutrino-nucleon scattering Vcdmeas . The experimentally determined event counts and kinematic distributions of the quark-tagged event samples provide measurements of the differential distributions for each of the final state quarks. The differential cross-sections, $`\frac{d\sigma }{dx}(q_iq_j)`$, and CKM matrix elements, $`|\mathrm{V}_{\mathrm{ij}}|`$, are derived from equations 15 through 21 using some auxiliary knowledge of the quark x-distributions within the nucleons and also a model for the mass threshold suppression terms, $`T(m_q,x)`$. In practice, this information should be obtainable largely from the data samples themselves: from CC and NC structure function measurements and the observed kinematic dependences in the heavy quark event sample.
### V.3 Expected Measurement Precisions at MURINEs
Today’s measurements of $`|\mathrm{V}_{\mathrm{cd}}|`$ in $`\nu `$N scattering are already the most precise in any process, despite the coarse instrumentation of the neutrino detectors and the consequent requirement to use the semi-muonic subsample of charm decays for final state charm tagging. Even the lowest energy MURINEs under consideration (dedicated neutrino factories with $`E_\mu 10`$ GeV and up) will provide an opportunity numcbook to extend to unique and precise measurements of the elements $`|\mathrm{V}_{\mathrm{cd}}|`$ and probably $`|\mathrm{V}_{\mathrm{cs}}|`$, now using vertex tagging of charm and with much improved knowledge of the quark distributions.
Further measurements of the more theoretically interesting elements $`|\mathrm{V}_{\mathrm{ub}}|`$ and $`|\mathrm{V}_{\mathrm{cb}}|`$ will become available at MURINES with muon energies of around 100 GeV or above, which can provide high enough neutrino energies for B production. The B-production analyses at these higher energy MURINEs should be experimentally rather similar to the charm analyses but would have vastly greater theoretical interest.
Both $`|\mathrm{V}_{\mathrm{ub}}|`$ and $`|\mathrm{V}_{\mathrm{cb}}|`$ determine the lengths of sides of the “unitarity triangle” that is predicted to exist if the CKM matrix is indeed unitary PDG\_CKM . The main goal of today’s B factories is to measure the interior angles of this triangle to confirm that it is indeed a triangle, and the complementary input from a MURINE will be an enormous help in this verification process. In particular, the predicted nufnal97 ; numcbook 1-2 % accuracy in $`|\mathrm{V}_{\mathrm{ub}}|^2`$ is several times better than predicted accuracies in any future measurements of other processes, and will obviously provide a very strong constraint on the unitarity triangle.
Predicted experimental accuracies for a 500 GeV MURINE are summarized in table 3. These measurements would likely be improved still further with the higher event statistics and cleaner theoretical analysis available at a Mighty MURINE.
### V.4 Possible Measurements of $`\mathrm{V}_{\mathrm{td}}`$ in Flavor Changing Neutral Current Interactions
The increased event statistics and neutrino energies at Mighty MURINEs might even allow the measurement of the further matrix element $`|\mathrm{V}_{\mathrm{td}}|`$ through the flavor changing neutral current (FCNC) interaction of figure 5.
This process is analogous to the predicted rare B decay $`\mathrm{B}\mathrm{X}_\mathrm{d}\nu \overline{\nu }`$, shown in figure 6, where $`\mathrm{X}_\mathrm{d}`$ represents inclusive production of hadrons containing a down quark. As related measurements, the very rare kaon decay processes $`K^{}\pi ^{}\nu \overline{\nu }`$ and $`K_L^0\pi ^0\nu \overline{\nu }`$ proceed through diagrams equivalent to 6 except with the incoming b quark replaced by an s quark and, correspondingly, $`\mathrm{V}_{\mathrm{tb}}`$ replaced by $`\mathrm{V}_{\mathrm{ts}}`$. Therefore, the charged kaon decay has the potential to measure $`|\mathrm{V}_{\mathrm{ts}}^{}\mathrm{V}_{\mathrm{td}}|`$ and its neutral counterpart actually measures the imaginary part of this quantity, $`\mathrm{Im}(\mathrm{V}_{\mathrm{ts}}^{}\mathrm{V}_{\mathrm{td}})`$, due to $`\mathrm{K}^0\overline{\mathrm{K}^0}`$ interference. One event of the decay $`K^{}\pi ^{}\nu \overline{\nu }`$ has been seen so far E787 , consistent with its predicted tiny branching ratio of $`(8.2\pm 3.2)\times 10^{11}`$, and the even rarer neutral decay process has yet to be observed.
The search for the B decay signal looks to be extremely challenging even at future B factories, so it is unlikely to yield an accurate measurement of $`|\mathrm{V}_{\mathrm{td}}|`$. Therefore, a neutrino measurement of this quantity, at or below the 10% level, might still be valuable even after the B factories have run, augmenting the complementary measurements, perhaps eventually with comparable accuracy Littenberg , of $`|\mathrm{V}_{\mathrm{ts}}^{}\mathrm{V}_{\mathrm{td}}|`$ and $`\mathrm{Im}(\mathrm{V}_{\mathrm{ts}}^{}\mathrm{V}_{\mathrm{td}})`$ expected in future generations of rare kaon decay experiments and the precise measurement of the ratio $`|\mathrm{V}_{\mathrm{ts}}|/|\mathrm{V}_{\mathrm{td}}|`$ that is to be eventually expected from $`\mathrm{B}_\mathrm{d}`$ and $`\mathrm{B}_\mathrm{s}`$ oscillations.
For neutrino energies well above the B production threshold, the process of figure 5 will occur at the level numcbook of order $`10^8`$ of the total neutrino-induced event sample unless some exotic physics process intervenes to increase the production rate. The signature is production of single $`\mathrm{B}^{}`$ mesons from the valence d quarks at high x, whose rate should be directly proportional to the product of $`|\mathrm{V}_{\mathrm{td}}|`$ and the known valence quark density in nuclons.
The main background will come from b–anti-b production where the partner B meson containing the b anti-quark has escaped detection from either its primary decay vertex or through the decay vertex of its daughter charmed meson. The background process is easily separable from the signal on a statistical basis because it is symmetric in $`\mathrm{B}^{}`$ versus $`\mathrm{B}^+`$ mesons. However, the raw production rate numcbook is roughly five orders of magnitude above the signal so the statistical viability of the analysis would require the raw background to be reduced by perhaps 4 to 5 orders of magnitude. This can be contemplated only because of (1) the very different event kinematics, with almost all of the background events at low Bjorken $`x`$ and the signal mostly at high $`x`$, and (2) the unprecedented veto power for B decays that is expected in the vertexing detectors at MURINEs. Even so, a raw signal event sample of thousands of events might be needed for a 10% measurement of $`|\mathrm{V}_{\mathrm{td}}|`$ given the statistical dilution that background processes might entail. This would require several years running for the experimental parameters of table 1.
### V.5 CKM Measurements from the Production of Top Quarks at the Highest Energy Mighty MURINEs
Aside from top production in loops, a daunting leap of 3 orders of magnitude in beam energy would be required to move from the CKM elements involving B production to those involving top production, as is seen from comparing the second and third rows of equation V.2. Uniquely precise direct measurements of $`|\mathrm{V}_{\mathrm{td}}|^2`$ and $`|\mathrm{V}_{\mathrm{ts}}|^2`$ and, possibly, $`|\mathrm{V}_{\mathrm{tb}}|^2`$ from the production of top quarks will become available if and when muon colliders eventually reach the 100 TeV CoM energy scale. (Note that muon collider energies even up to 1000 TeV, i.e. 1 PeV, have been speculated, using muon acceleration in linacs Zimmermann .)
Such impressive machines are prospects for the far distant future, and would be intended to zero in on a coherent understanding of the elementary building blocks of our universe. It should be stated that a major sea change from current theoretical prejudices would be required if the CKM matrix and its information on CP violation was to become central to the construction or verification of such a “theory of everything”. Disregarding the current prejudices, top production at these highest energy Mighty MURINEs would move the experimental probing of the CKM matrix to a level of accuracy that appears to be inaccessible to any other type of experiment.
As will be explained further in the following section, a simple scaling from top production estimates calculated for HERA HERAref predicts of order $`10^5`$ top quark events for 1 inverse zeptobarn of integrated luminosity at muon energies slightly above 50 TeV. (A more accurate and detailed calculation is obviously required!)
Experimentally, top production should be relatively easy to tag with very high efficiency and purity. The two signatures are:
$`\nu _\mu N\mu ^{}(2\mathrm{jets})(\mathrm{b}\mathrm{jet})`$ (24)
$`\nu _\mu N\mu ^{}l^+\nu (\mathrm{b}\mathrm{jet}),`$ (25)
with 68% and 32% BR’s, respectively. Because of the large top mass, the final state jets can each have large acoplanarities, and the rarity of backgrounds with b quarks makes both signatures particularly distinctive. Additionally, in the first case the 2 other jets will reconstruct to the W mass while the presence of a second high-$`\mathrm{p}_\mathrm{t}`$, high energy lepton and large missing $`\mathrm{p}_\mathrm{t}`$ from the neutrino will make the second signature even more striking.
No attempt will be made to even guess at the measurement accuracy. As a general comment, the beam energy will never be very far above the threshold for top quark production, so the feasibility and accuracy of the measurements would depend more strongly on the muon beam energy than the beam intensity. In almost all cases, the measurements of the CKM matrix elements involving top should be statistically limited because of the relatively small statistics (except at PeV-scale colliders!), their distinctive experimental signature and the accurately predictable threshold behavior. The sequentially decreasing populations at high $`x`$ of the progressively heavier initial state quarks – d, s and b – should compensate or over-compensate the trend for the higher couplings to the top quark in the respective measurements of $`|\mathrm{V}_{\mathrm{td}}|`$, $`|\mathrm{V}_{\mathrm{ts}}|`$ and $`|\mathrm{V}_{\mathrm{tb}}|`$. Whether the first, the first two or all three matrix could be measured would presumably also depend strongly on the beam energy.
CKM measurements involving top would extend CKM studies beyond the paradigm of the unitarity triangle that is connected to B factory studies. For example, the conventional unitarity triangle is formed from the dot product of columns 1 and 3 of the CKM matrix PDG\_CKM . Measurements of the CKM elements involving the top quark would also provide enough experimental input to test the corresponding triangle involving rows 1 and 3 of the matrix, which is of comparable theoretical value in exploring CP violation. The analysis of experimental results would more likely be couched in more general theoretical terms, including unitarity tests and global fits to the 4 parameters – three magnitudes and a phase – that characterize the unitary 3-by-3 CKM matrix. Consistency of these fits would probe the SM hypothesis at a level that would not be possible without Mighty MURINEs.
## VI Heavy Particle Production – Mighty MURINEs Versus HERA
The HERA ep collider is a convenient reference point for assessing the physics potential of Mighty MURINEs at the highest energy scales. As indicated in figure 7, MURINEs have the same weak interactions as HERA while avoiding the predominantly soft electromagnetic interactions that dominate the HERA event trigger rates but are less interesting because lower energy transfers probe physics at relatively lower energy scales. The event samples at MURINEs will correspond to the weak interactions in the “hard scattering tail” of the HERA event sample.
The HERA event samples involving weak interactions can be compared with MURINEs at energies where the high energy tail of the neutrino beam is comparable to the 314 GeV HERA center-of-mass energy. The maximum neutrino energy is the muon beam energy, which, according to equation 5, equals the HERA CoM energy for $`E_\mu =53`$ TeV. At this energy or slightly above, very rough estimates of the event rates of similar or identical processes can be simply transcribed from HERA calculations after scaling by the $`10^5`$ luminosity ratio shown in table 1. Such a comparison is shown in table 4. A range of MURINE energies has been given, in deference to the very approximate nature of the comparison. At the low end ($`E_\mu =50`$ TeV), the MURINE event rates will probably be lower than the estimate, and the rates will normally be higher at the high end ($`E_\mu =100`$ TeV).
The standard model physics processes involving weak interactions are the same in all cases except for the production of W and Z bosons, where HERA has the advantage due to processes involving photon exchange. The SM Higgs has not been found at the time of writing, so the 120 GeV mass is an example only.
The first three processes in table 4 have already been discussed in the preceding section. To be realistic, at the event rates shown it is very doubtful that W, Z and SM Higgs production could contribute anything useful beyond collider studies, despite the the astounding neutrino beam parameters and superior event reconstruction. Beyond this, possible exotic processes at the 100 GeV scale or below provide the only substantial potential for exciting discoveries. This motivation could become much stronger in the near future if, for example, one of the current leptoquark searches at HERA returned a discovery. It is noted that the leptoquarks produced at a Mighty MURINE might well be different – coupling to neutrinos rather than electrons – and so studies at MURINEs could potentially be complementary to those at a future ep collider with a higher $`\mathrm{E}_{\mathrm{CoM}}`$ than HERA.
## VII Summary
The Mighty MURINE neutrino experiments that would come almost for free at any future many-TeV muon collider could improve on the pioneering advances from the previous MURINEs that would have existed at lower energy muon colliders. The most important improvements might well be on the unique and important measurements from previous lower energy MURINEs of $`|\mathrm{V}_{\mathrm{ub}}|`$ and $`|\mathrm{V}_{\mathrm{cb}}|`$, perhaps pushing the accuracy of both measurements below 1%. With total event statistics of a few times $`10^{11}`$ events, the rare production of B’s through flavor changing neutral current interactions off valence d quarks might provide one of the best indirect determinations of $`|\mathrm{V}_{\mathrm{td}}|`$. More common channels for B production, particularly through neutral current interactions, might also provide some capabilities as a B factory with novel experimental strengths.
Upon crossing the threshold for top production, the even more interesting elements $`|\mathrm{V}_{\mathrm{td}}|`$, $`|\mathrm{V}_{\mathrm{ts}}|`$ and $`|\mathrm{V}_{\mathrm{tb}}|`$ could become successively available to uniquely precise measurements at the highest energy Mighty MURINEs. The addition of any or all of these three precise measurements would clearly advance our knowledge of the CKM matrix to a level where small perturbations from the Standard Model scenario could be searched for and, if found, could be studied. MURINEs would then truly play the central role in determining the CKM matrix parameters, with the best measurements of the magnitudes of perhaps seven of the nine elements (all but the two elements that are currently best measured: $`|\mathrm{V}_{\mathrm{ud}}|`$ and $`|\mathrm{V}_{\mathrm{us}}|`$) to add to the phase information from various other experimental processes.
If muon colliders ever reach the 100 TeV center of mass energy scale then their neutrino experiments will attain a center of mass energy reach comparable to the existing HERA ep collider, but at a luminosity that might be perhaps 5 orders of magnitude higher. HERA then becomes a convenient reference point for assessing their physics capabilities. Despite the promise of impressive luminosities, none of the standard model processes other than the CKM matrix appear to offer the chance of competitive physics potential to studies of the same processes at colliders. Therefore, only i) an enlarged theoretical importance for the CKM matrix or ii) the discovery, then or beforehand, of an exotic process that is accessible to Mighty MURINEs, would give Mighty MURINEs a chance for physics analyses of a comparable importance to those at the colliders. Leptoquarks that couple to neutrinos are the obvious candidate for such a new process.
|
warning/0005/hep-ph0005286.html
|
ar5iv
|
text
|
# Higgs Mechanism and Bulk Gauge Boson Masses in the Randall-Sundrum Model
\[
## Abstract
Assuming the breaking of gauge symmetries by the Higgs mechanism, we consider the associated bulk gauge boson masses in the Randall-Sundrum background. With the Higgs field confined on the TeV-brane, the W and Z boson masses can naturally be an order of magnitude smaller than their Kaluza-Klein excitation masses. Current electroweak precision data requires the lowest excited state to lie above about 30 TeV, with fermions on the TeV-brane. This bound is reduced to about 10 TeV if the fermions reside sufficiently close to the Planck-brane. Thus, some tuning of parameters is needed. We also discuss the bulk Higgs case, where the bounds are an order of magnitude smaller.
\]
It has recently been realized that the large hierarchy between the Planck scale and the electroweak scale could be related to the presence of extra dimensions . An interesting realization of this concept is the Randall-Sundrum model . It relies on the 5-dimensional non-factorizable geometry
$$ds^2=e^{2\sigma (y)}\eta _{\mu \nu }dx^\mu dx^\nu +dy^2,$$
(1)
where $`\sigma (y)=k|y|`$. The 4-dimensional metric is $`\eta _{\mu \nu }=\mathrm{diag}(1,1,1,1)`$, $`k`$ is the AdS curvature, and $`y`$ denotes the fifth dimension. This metric results from a suitable adjustment of the bulk cosmological constant and the tensions of the two 3-branes which reside at the $`S_1/Z_2`$ orbifold fixed points $`y=0`$, $`y=\pi R`$. Because of the exponential (“warp”) factor, the effective mass scale on the brane located at $`y=\pi R`$ is $`M_Pe^{\pi kR}`$. If $`kR11`$ this scale will be $`𝒪`$(TeV), and the brane is referred to as the ‘TeV-brane’. Hence the model can generate an exponential hierarchy of scales from a small extra dimension.
In the setting of ref. only gravity propagates in the 5d bulk, while the Standard Model (SM) fields are confined to the TeV-brane. However, since a microscopic derivation is still missing, it is interesting to study other possibilities. Bulk scalar fields were first discussed in ref. . The consequences of SM gauge bosons in the bulk were studied in ref. . In ref. the behavior of fermions in the bulk was investigated, and in ref. the complete SM was put in the bulk. Finally, bulk supersymmetry was considered in ref. .
Bulk gauge fields are necessary if the SM fermions live in the bulk. By localizing the fermions at different positions in the fifth dimension it seems possible to address the questions of fermion mass hierarchy, non-renormalizable operators and proton decay . New possibilities for baryogenesis may open up if the fermion separation is reduced by thermal correction in the hot early universe .
Bulk vector bosons with bulk masses have been considered to some extent in refs. . It was found that the ‘zero’ mode acquires a mass comparable to the mass of the ‘first’ Kaluza-Klein (KK) excitation, unless the bulk gauge boson mass is extremely fine-tuned . Since gauge boson KK excitations should have masses in the TeV range , the W-boson mass could only be generated by reintroducing the original hierarchy problem. This suggests that the Higgs should be confined to the TeV-brane, i.e. the gauge boson mass arises from the boundary.
In this letter we will investigate this scenario in more detail. We will study the properties of bulk gauge bosons which are related to broken gauge symmetries, i.e. bulk W and Z bosons. We will show that in the case of a TeV-brane Higgs field, the W and Z boson masses are naturally an order of magnitude smaller than the mass of their first KK excitations. We will demonstrate that the W and Z boson mass ratio and $`\mathrm{sin}^2\theta _W`$ can be successfully reproduced by a moderate tuning of the brane mass parameter. We also discuss constraints from universality of the coupling of the gauge bosons to fermions. In the phenomenologically viable parameter range we recover the 4d relationship between gauge and Higgs boson masses. Contraints arising in the bulk Higgs case are also briefly discussed.
Let us consider the following equation of motion for a U(1) gauge boson $`A_N`$ of 5-dimensional mass $`M`$
$$\frac{1}{\sqrt{g}}_M(\sqrt{g}g^{MN}g^{RS}F_{NS})M^2g^{RS}A_S=0,$$
(2)
where $`g_{MN}`$ denotes the 5-dimensional metric. In general, $`M`$ arises from some Higgs mechanism and consists of bulk and boundary contributions
$$M^2(y)=b^2k^2+a^2k\delta (y\pi R)+\stackrel{~}{a}^2k\delta (y),$$
(3)
depending on whether the Higgs fields live in the bulk and/or on the branes. The gauge boson masses can be expressed in terms of the parameters of the Higgs potential. For the TeV-brane Higgs, for instance, we have
$$a^2=\frac{g_5^2\mu ^2}{2\lambda k}=\frac{g_5^2}{2k}v^2,$$
(4)
where $`\mu `$ denotes the Higgs mass parameter and $`\lambda `$ the quartic coupling, both understood as 4d quantities, and $`g_5`$ is the 5d gauge coupling. $`v=\mu /\sqrt{\lambda }`$ is the vev of the Higgs field.
Using the metric (1) and decomposing the 5d field as
$$A_\mu (x^\mu ,y)=\frac{1}{\sqrt{2\pi R}}\underset{n}{}A_\mu ^{(n)}(x^\mu )f_n(y),$$
(5)
one obtains
$$(_y^2+2\sigma ^{}_y+M^2)f_n=e^{2\sigma }m_n^2f_n,$$
(6)
where $`m_n`$ are the masses of the Kaluza-Klein excitations $`A_\mu ^{(n)}`$, and $`\sigma ^{}=_y\sigma `$. (We work in the gauge $`A_5=0`$ and $`_\mu A^\mu =0`$.)
Requiring the gauge boson wave function to be even under the $`Z_2`$ orbifold transformation, $`f_n(y)=f_n(y)`$, one finds
$$f_n=\frac{e^\sigma }{N_n}\left[J_\alpha (\frac{m_n}{k}e^\sigma )+\beta _\alpha (m_n)Y_\alpha (\frac{m_n}{k}e^\sigma )\right],$$
(7)
where the order of the Bessel functions is $`\alpha =\sqrt{1+b^2}`$. The coefficients $`\beta _\alpha `$ obey
$`\beta _\alpha (x_n,\stackrel{~}{a}^2)`$ $`=`$ $`{\displaystyle \frac{(\frac{\stackrel{~}{a}^2}{2}+1\alpha )J_\alpha (x_n)+x_nJ_{\alpha 1}(x_n)}{(\frac{\stackrel{~}{a}^2}{2}+1\alpha )Y_\alpha (x_n)+x_nY_{\alpha 1}(x_n)}},`$ (8)
$`\beta _\alpha (x_n,\stackrel{~}{a}^2)`$ $`=`$ $`\beta _\alpha (\mathrm{\Omega }x_n,a^2),`$ (9)
where we defined the warp factor $`\mathrm{\Omega }=e^{\pi kR}`$, and $`x_n=\frac{m_n}{k}`$. Note that for non-vanishing boundary mass terms the derivative of $`f_n`$ becomes discontinuous on the boundaries. The normalization constants $`N_n`$ are defined such that $`(1/\pi R)_0^{\pi R}𝑑yf_n^2=1`$.
An analogous discussion also holds in the non-abelian case.
Eqs. 8, 9 encode the masses of the different KK states. In the limit $`m_nk`$ and $`m_n\mathrm{\Omega }k`$, one finds
$$m_n(n+\frac{\alpha }{2}\frac{3}{4})\pi k\mathrm{\Omega }^1.$$
(10)
In this regime the masses of the excited KK states are independent of the boundary mass terms. The bulk mass term enters via $`\alpha `$, but its contribution is also suppressed by the warp factor. This is because the excited states are localized at the TeV-brane as a result of the exponential in their wave functions. If the SM fermions live on the TeV-brane, it was found that the masses of the gauge boson KK states should be in the multi TeV range in order to be in agreement with the electroweak precision data . In the case of bulk fermions the corresponding constraints becoming weaker , reducing to $`m_1>0.5`$ TeV for fermions on the Planck-brane .
Let us now consider $`m_0`$, the mass of the lowest lying state. In the case with neither bulk nor boundary mass term, one finds $`m_0=0`$, and the corresponding (zero mode) wave function is not localized in the extra dimension, $`f_0(y)1`$. If a bulk or boundary mass term is added, the ‘zero’ mode picks up a mass, and its wave function displays a y-dependence.
In the case of a bulk mass term $`b1`$, one finds $`m_0\pi k\mathrm{\Omega }^1`$, i.e. approximately of the same value as the first excited state in the massless case . Although the bulk mass term is of order $`M_p`$, the gauge boson mass does not become Planck-sized, because $`f_0`$ is localized at the TeV-brane, where the effective mass scale is small. At first sight this seems encouraging, but it was also shown in ref. that extreme fine tuning $`b\mathrm{\Omega }^1`$ is necessary in order to bring down the W-boson mass, i.e. $`m_0`$, from its natural TeV size range to the experimental value. One therefore would have to start with a weak scale bulk mass term, which is nothing but the original fine tuning problem. These results follow from expanding eqs. 8, 9 in the regime $`x_n1`$ and $`x_n\mathrm{\Omega }1`$. Along the same lines we find that a gauge boson mass term at the Planck-brane has the same implications,
$$x_0^2\frac{\stackrel{~}{a}^2}{2\mathrm{ln}\mathrm{\Omega }}=\frac{\stackrel{~}{a}^2}{2\pi kR}.$$
(11)
Since their is no warp factor suppression, only for $`\stackrel{~}{a}\mathrm{\Omega }^1`$ is a W-mass below the Kaluza-Klein scale possible.
If the Higgs is on the TeV-brane however, we arrive at a different conclusion. Expanding eqs. 8, 9 for $`x_01`$ and $`\mathrm{\Omega }x_01`$ we find
$$\mathrm{\Omega }^2x_0^2\frac{a^2}{2\mathrm{ln}\mathrm{\Omega }}\text{ }\frac{1}{1\frac{a^2}{4\mathrm{ln}\mathrm{\Omega }}(\gamma +\mathrm{ln}\frac{x_0}{2})},$$
(12)
where $`\gamma 0.5772`$, which reduces to
$`\mathrm{\Omega }^2x_0^2`$ $``$ $`{\displaystyle \frac{a^2}{2\mathrm{ln}\mathrm{\Omega }}}={\displaystyle \frac{a^2}{2\pi kR}},a1`$ (13)
$`\mathrm{\Omega }^2x_0^2`$ $``$ $`{\displaystyle \frac{2}{\mathrm{ln}\mathrm{\Omega }}}={\displaystyle \frac{2}{\pi kR}},a1.`$ (14)
Similar to the case of a bulk or Planck-brane mass term, we find a linear relationship between $`a`$ and $`x_0`$ for small values of $`a`$. But in contrast to the former, this behavior remains valid up to $`a\text{ }\stackrel{<}{}\text{ }1`$, because of the appearance of the warp factor. For $`a\text{ }\stackrel{>}{}\text{ }1`$, $`x_0`$ saturates at a value typically an order of magnitude smaller than $`x_1`$, which corresponds to the mass of the first excited state. This demonstrates that a Higgs boson at the TeV-brane can, in principle, explain weak gauge boson masses of order 100 GeV, while keeping the KK states in the TeV range . The saturation results from the drop of the wave function near the TeV-brane for large $`a`$ which diminishes the overlap with the brane mass term.
In fig. 1 we show $`\mathrm{\Omega }x_0`$ as a function of $`a`$ for $`\mathrm{\Omega }=10^{14}`$, i.e. $`kR=10.26`$. For $`a1`$ we obtain $`\mathrm{\Omega }x_00.24`$. (In the evaluation we numerically solved eqs. 8, 9, but eq. 12 would also reproduce the results at a percent accuracy level.) The mass of the first excited KK state depends very weakly on $`a`$. In our example we find it rises from $`\mathrm{\Omega }x_1(a=0)=2.45`$ to $`\mathrm{\Omega }x_1(\mathrm{})=3.88`$. In fig. 2 we display the resulting ratio between the mass of the ground state and the first excited state. For small enough $`a`$ we find $`x_1/x_0m_1/m_01/a`$, while for large $`a`$ the ratio approaches $`x_1/x15.4`$. For different values of $`\mathrm{\Omega }`$ the results hardly change since the warp factor only enters logarithmically in (13, 14).
Since the ground state mass scale, i.e. the W and Z boson masses, is experimentally known to be $`100`$ GeV, we conclude that in the brane Higgs scenario $`m_1\text{ }\stackrel{>}{}\text{ }1.5`$ TeV is necessary. This bound does not rely on the electroweak precision data and is independent of the position of the fermions in the fifth dimension. It could only be weakened if the warp factor in (14) is substantially reduced, which would reintroduce the hierarchy problem.
We next discuss constraints on KK excitations arising from the electroweak precision data. From fig. 1 we deduce that the relationship between the boundary mass term $`a`$ and the ground state mass becomes highly non-linear in the regime $`a\text{ }\stackrel{>}{}\text{ }1`$. As a result the very successful SM prediction that the gauge boson masses are proportional to their couplings to the Higgs could be spoiled. We measure the deviation from the linear behavior of $`x(a)`$ by
$$\delta _1\frac{x_0(ar)}{x_0(a)}r,$$
(15)
and take $`r=M_W/M_Z=0.88`$. For $`\mathrm{\Omega }=10^{14}`$ the results are shown in fig. 3. They are well approximated by $`\delta _10.025a^2`$, i.e. the non-linearity increases quadratically with $`a`$. Since $`r`$ is measured to an accuracy of about $`10^3`$ and no deviations from the SM prediction have been found , we require $`\delta _1\text{ }\stackrel{<}{}\text{ }10^3`$. This leads to the modest constraint $`a\text{ }\stackrel{<}{}\text{ }0.2`$. From fig. 2 we deduce that $`x_1/x_0\text{ }\stackrel{>}{}\text{ }100`$. As a result the mass of the first KK excitation has to obey $`m_1\text{ }\stackrel{>}{}\text{ }10`$ TeV. The constraint on $`m_1`$ is proportional to $`1/\sqrt{\delta _1}`$. We stress again that it does not depend on where the fermions live. The warp factor only enters logarithmically.
Once the ‘zero’ mode acquires a mass its wave function (7) depends on the $`y`$ coordinate. In contrast to excited states the wave function tries to avoid the TeV-brane where its mass arises, as shown in fig. 4. As a result the successful SM predictions of the gauge couplings to fermions of the W and Z bosons can be affected. The resulting constraints depend on the position of the fermions in the fifth dimension.
For example, the coupling of the W boson to a fermion on the TeV-brane is given by $`g_0f_0(\pi R)`$, where $`g_0`$ denotes the coupling if the boson were massless. Since $`f_0(\pi R)<1`$ in the brane Higgs scenario, the resulting gauge coupling is somewhat reduced. In fig. 3 we present the resulting deviation from the SM prediction,
$$\delta _21f_0(\pi R),$$
(16)
as a function of the brane mass parameter $`a`$ ($`\mathrm{\Omega }=10^{14}`$). For $`\delta _2\text{ }\stackrel{<}{}\text{ }10^3`$, we find $`a\text{ }\stackrel{<}{}\text{ }0.06`$, a constraint more stringent than the one from the mass ratio $`r=M_W/M_Z`$. From fig. 2 we learn that $`x_1/x_0\text{ }\stackrel{>}{}\text{ }310`$, i.e. $`m_1\text{ }\stackrel{>}{}\text{ }30`$ TeV, a bound which is proportional to $`1/\sqrt{\delta _2}`$. With this restriction the effects of the KK states are automatically in agreement with the electroweak precision data, which only requires $`m_1\text{ }\stackrel{>}{}\text{ }23`$ TeV .
If the massive gauge boson is coupled to fermions on the Planck-brane the effective gauge couplings hardly deviate from the SM prediction, since $`f_0(0)1`$ (see fig. 4). The resulting constraint, $`m_1\text{ }\stackrel{>}{}\text{ }4`$ TeV, is weaker than that arising from eq. 15.
Bulk fermions interpolate between the TeV- and the Planck-brane scenarios. As discussed in refs. , depending on the bulk mass term $`m_\psi =c\sigma ^{}`$ for the fermion, the zero mode of the fermion is localized at the TeV-brane ($`c<1/2`$) or at the Planck-brane ($`c>1/2`$). For $`c=1/2`$ the fermionic zero mode is delocalized in the fifth dimension. Since the W-boson wave function has a nontrivial $`y`$-dependence, it then couples non-universally to fermions localized at different positions in the fifth dimension. This is completely analogous to the $`c`$-dependent coupling of the excited gauge boson states discussed in ref. . We have repeated the analysis for the ground state of the massive gauge boson. In fig. 5 we display $`g/g_01`$ as function of $`c`$ for $`a=0.14`$ and $`\mathrm{\Omega }=10^{14}`$, where $`g_0`$ would be the coupling of a massless gauge boson. The shape of the gauge coupling of the massive ground state is similar to those of the excites KK states . However, the amplitude of the variation is much smaller. In the limit $`c\mathrm{}`$, $`g`$ approaches the result of the TeV-brane fermions (16). In the regime $`c\text{ }\stackrel{>}{}\text{ }1/2`$ the deviation of the SM prediction for $`g`$ becomes small. In this case $`a`$ is only constrained by the W,Z boson mass ratio (15).
If the SM fermions reside on the TeV-brane, non-renormalizable operators are typically suppressed only by a few TeV instead of the large 4d Planck mass . This may induce rapid proton decay, large flavor changing neutral currents, and large neutrino masses. Global symmetries like baryon and and number may be imposed to forbid the corresponding operators. If the SM fermions are bulk fields, the non-renormalizable interactions can be suppressed to some extent by localizing the first and second generation fermions near the Planck-brane . We will address this topic in a future publication .
Taking into account the warp factor, the Higgs mass on the TeV-brane is given by
$$M_H=\mu \mathrm{\Omega }^1=\sqrt{\lambda }v\mathrm{\Omega }^1.$$
(17)
In 4d the gauge and Higgs boson masses are related by $`M_W^2=\frac{g^2}{2\lambda }M_H^2`$. In the brane Higgs scenario this relationship is certainly violated in the regime $`a\text{ }\stackrel{>}{}\text{ }1`$ due to the non-linearity in $`m_0(a)`$ (see fig. 1). However, in the phenomenologically viable parameter range $`a\text{ }\stackrel{<}{}\text{ }a_{\mathrm{max}}0.14`$, where $`m_0^2(g_5^2/4\pi R)v^2`$ (13, 4), and $`gg_5/\sqrt{2\pi R}`$, the 4d relation is recovered, up to small correction of order $`10^3`$. Using (4), the parameters of the Higgs potential have to obey
$$\frac{\mu ^2}{k^2}<\frac{a_{\mathrm{max}}^2}{2\pi kR}\frac{2\lambda }{g^2}.$$
(18)
Assuming $`\lambda 1`$ we find that a moderate tuning $`\mu \text{ }\stackrel{<}{}\text{ }0.04k`$ is required to reproduce the measured W and Z boson masses in the brane Higgs scenario.
Finally, let us briefly summarize our results for the bulk Higgs case, which may be especially interesting if SM fermions reside on the Planck-brane in order to eliminate unwanted higher dimensional operators. A TeV-brane Higgs cannot provide masses for these fermions. To solve the hierarchy problem, one has to rely on some additional mechanism, for instance supersymmetry. The W,Z boson mass ratio (15) leads to $`m_1\text{ }\stackrel{>}{}\text{ }250`$ GeV, which is rather weak compared to the brane Higgs case ($`m_1\text{ }\stackrel{>}{}\text{ }30`$ TeV). Stronger restrictions arise from the modification of the gauge couplings (16). For Planck-brane fermions we find $`m_1\text{ }\stackrel{>}{}\text{ }600`$ GeV, while for TeV-brane fermions this increases to $`m_1\text{ }\stackrel{>}{}\text{ }3.5`$ TeV, bounds which are again weaker than for the TeV-brane Higgs case. The wave function of the ‘zero’ mode increases near the TeV-brane. This results also apply in the Planck-brane Higgs scenario.
We would like to thank G. Dvali and G. Gabadadze for very useful discussions. S. H. is supported in part by the Alexander von Humboldt foundation. This work was also supported by DOE under contract DE-FG02-91ER40626.
email address: shuber@bartol.udel.edu
email address: shafi@bartol.udel.edu
|
warning/0005/cs0005008.html
|
ar5iv
|
text
|
# A Denotational Semantics for First-Order Logic
## 1 Introduction
### Background
To explain properly the motivation for the work here discussed we need to go back to the roots of logic programming and constraint logic programming. Logic programming grew out of the seminal work of Robinson \[Rob65\] on the resolution method and the unification method. First, Kowalski and Kuehner \[KK71\] introduced a limited form of resolution, called linear resolution. Then Kowalski \[Kow74\] proposed what we now call SLD-resolution. The SLD-resolution is both a restriction and an extension of the resolution method. Namely, the clauses are restricted to Horn clauses. However, in the course of the resolution process a substitution is generated that can be viewed as a result of a computation. Right from the outset the SLD-resolution became then a crucial example of the computation as deduction paradigm according to which the computation process is identified with a constructive proof of a formula (a query) from a set of axioms (a program) with the computation process yielding the witness (a substitution).
This lineage of logic programming explains two of its relevant characteristics:
1. the queries and clause bodies are limited to the conjunctions of atoms,
2. the computation takes place (implicitly) over the domain of all ground terms of a given first-order language.
The restriction in item 1. was gradually lifted and through the works of Clark \[Cla78\] and Lloyd and Topor \[LT84\] one eventually arrived at the possibility of using as queries and clause bodies arbitrary first-order formulas. This general syntax is for example available in the language Gödel of Lloyd and Hill \[HL94\].
A way to overcome the restriction in item 2. was proposed in 1987 by Jaffar and Lassez in their influential CLP(X) scheme that led to constraint logic programming. In this proposal the computation takes place over an arbitrary interpretation and the queries and clause bodies can contain constraints, i.e., atomic formulas interpreted over the chosen interpretation. The unification mechanism is replaced by a more general process of constraint solving and the outcome of a computation is a sequence of constraints to which the original query reduces.
This powerful idea was embodied since then in many constraint logic programming languages, starting with the CLP$`()`$ language of Jaffar, Michaylov, Stuckey, and Yap \[JMSY92\] in which linear constraints over reals were allowed, and the CHIP language of Dincbas et al. \[DVS<sup>+</sup>88\] in which linear constraints over finite domains, combined with constraint propagation, were introduced. A theoretical framework for CHIP was provided in van Hentenryck \[Van89\].
This transition from logic programming to constraint logic programming introduced a new element. In the CLP(X) scheme the test for satisfiability of a sequence of constraints was needed, while a proper account of the CHIP computing process required an introduction of constraint propagation into the framework. On some interpretations these procedures can be undecidable (the satisfiability test) or computationally expensive (the “ideal” constraint propagation). This explains why in the realized implementations some approximation of the former or limited instances of the latter were chosen for.
So in both approaches the computation (i.e., the deduction) process needs to be parametrized by external procedures that for each specific interpretation have to be provided and implemented separately. In short, in both cases the computation process, while parametrized by the considered interpretation, also depends on the external procedures used. In conclusion: constraint logic programming did not provide a satisfactory answer to the question of how to lift the computation process of logic programming from the domain of all ground terms to an arbitrary interpretation without losing the property that this process is effective.
Arbitrary interpretations are important since they represent a declarative counterpart of data types. In practical situations the selected interpretations would admit sorts that would correspond to the data types chosen by the user for the application at hand, say terms, integers, reals and/or lists, each with the usual operations available. It is useful to contrast this view with the one taken in typed versions of logic programming languages. For example, in the case of the Gödel language (polymorphic) types are provided and are modeled by (polymorphic) sorts in the underlying theoretic model. However, in this model the computation still implicitly takes place over one fixed domain, that of all ground terms partitioned into sorts. This domain properly captures the built-in types but does not provide an account of user defined types. Moreover, in this approach different (i.e., not uniform) interpretation of equality for different types is needed, a feature present in the language but not accounted for in the theoretical model.
### Formulas as Programs
The above considerations motivated our work on a computational interpretation of first-order formulas over arbitrary interpretations reported in Apt and Bezem \[AB99\]. This allowed us to view first-order formulas as executable programs. That is why we called this approach formulas as programs. In our approach the computation process is a search of a satisfying valuation for the formula in question. Because the problem of finding such a valuation is in general undecidable, we had to introduce the possibility of partial answers, modeled by an existence of run-time errors.
This ability to compute over arbitrary interpretations allowed us to extend the computation as deduction paradigm to arbitrary interpretations. We noted already that the SLD-resolution is both a restriction and an extension of the resolution method. In turn, the formulas as programs approach is both a restriction and an extension of the logic programming. Namely, the unification process is limited to an extremely simple form of matching involving variables and ground terms only. However, the computation process now takes place over an arbitrary structure and full-first order syntax is adopted.
The formulas as programs approach to programming has been realized in the programming language Alma-0 \[ABPS98\] that extends imperative programming by features that support declarative programming. In fact, the work reported in Apt and Bezem \[AB99\] provided logical underpinnings for a fragment of Alma-0 that does not include destructive assignment or recursive procedures and allowed us to reason about non-trivial programs written in this fragment.
### Rationale for This Paper
The computational interpretation provided in Apt and Bezem \[AB99\] can be viewed as an operational semantics of first-order logic. The history of semantics of programming languages has taught us that to better understand the underlying principles it is beneficial to abstract from the details of the operational semantics. This view was put forward by Scott and Strachey \[SS71\] in their proposal of denotational semantics of programming languages according to which, given a programming language, the meaning of each program is a mathematical function of the meanings of its direct constituents.
The aim of this paper is to complement the work of \[AB99\] by providing a denotational semantics of first-order formulas. This semantics combines a number of ideas realized in the areas of (nondeterministic) imperative programming languages and the field of logic programming. It formalizes a view according to which conjunction can be seen as sequential composition, disjunction as “don’t know” nondeterminism, existential quantification as declaration of a local variable, and it relates negation to the “negation as finite failure” rule.
The main result is that the denotational semantics is sound with respect to the truth definition. The proof is reminiscent in some aspects of the proof of the soundness of the SLDNF-resolution of Clarke \[Cla78\]. The semantics of equations allows matching involving variables and non-ground terms, a feature not present in \[AB99\] and in Alma-0. This facility introduces logical variables in this framework but also creates a number of difficulties in the soundness proof because bindings to local variables can now be created.
First-order logic is obviously a too limited formalism for programming. In \[AB99\] we discussed a number of extensions that are convenient for programming purposes, to wit sorts (i.e., types), arrays, bounded quantification and non-recursive procedures. This leads to a very expressive and easy to program in subset of Alma-0. We do not envisage any problems in incorporating these features into the denotational semantics here provided. A major problem is how to deal with recursion.
The plan of the paper is as follows. In the next section we discuss the difficulties encountered when solving arbitrary equations over algebras. Then, in Section 3 we provide a semantics of equations and in Section 4 we extend it to the case of first-order formulas interpreted over an arbitrary interpretation. The resulting semantics is denotational in style. In Section 5 we relate this semantics to the notion of truth by establishing a soundness result. In Section 6 we draw conclusions and suggest some directions for future work.
## 2 Solving Equations over Algebras
Consider some fixed, but arbitrary, language of terms $`L`$ and a fixed, but arbitrary algebra $`𝒥`$ for it (sometimes called a pre-interpretation). A typical example is the language defining arithmetic expressions and its standard interpretation over the domain of integers.
We are interested in solving equations of the form $`s=t`$ over an algebra, that is, we seek an instantiation of the variables occurring in $`s`$ and $`t`$ that makes this equation true when interpreted over $`𝒥`$. By varying $`L`$ and $`𝒥`$ we obtain a whole array of specific decision problems that sometimes can be solved efficiently, like the unification problem or the problem of solving linear equations over reals, and sometimes are undecidable, like the problem of solving Diophantine equations.
Our intention is to use equations as a means to assign values to variables. Consequently, we wish to find a natural, general, situation for which the problem of determining whether an equation $`s=t`$ has a solution in a given algebra is decidable, and to exhibit a “most general solution”, if one exists. By using most general solutions we do not lose any specific solution.
This problem cannot be properly dealt with in full generality. Take for example the polynomial equations over integers. Then the equation $`x^23x+2=0`$ has two solutions, $`\{x/1\}`$ and $`\{x/2\}`$, and none is “more general” than the other under any reasonable definition of a solution being more general than another.
In fact, given an arbitrary interpretation, the only case that seems to be of any use is that of comparing a variable and an arbitrary term. This brings us to equations of the form $`x=t`$, where $`x`$ does not occur in $`t`$. Such an equation has obviously a most general solution, namely the instantiation $`\{x/t\}`$.
A dual problem is that of finding when an equation $`s=t`$ has no solution in a given algebra. Of course, non-unifiability is not a rescue here: just consider the already mentioned equation $`x^23x+2=0`$ the sides of which do not unify.
Again, the only realistic situation seems to be when both terms are ground and their values in the considered algebra are different. This brings us to equations $`s=t`$ both sides of which are ground terms.
## 3 Semantics of Equations
After these preliminary considerations we introduce specific “hybrid” objects in which we mix the syntax and semantics.
###### Definition 1
Consider a language of terms $`L`$ and an algebra $`𝒥`$ for it. Given a function symbol $`f`$ we denote by $`f_𝒥`$ the interpretation of $`f`$ in $`𝒥`$.
* Consider a term of $`L`$ in which we replace some of the variables by the elements of the domain $`D`$. We call the resulting object a generalized term.
* Given a generalized term $`t`$ we define its $`𝒥`$-evaluation as follows:
+ replace each constant occuring in $`t`$ by its value in $`𝒥`$,
+ repeatedly replace each sub-object of the form $`f(d_1,\mathrm{},d_n)`$ where $`f`$ is a function symbol and $`d_1,\mathrm{},d_n`$ are the elements of the domain $`D`$ by the element $`f_𝒥(d_1,\mathrm{},d_n)`$ of $`D`$.
We call the resulting generalized term a $`𝒥`$-term and denote it by $`[[t]]_𝒥`$. Note that if $`t`$ is ground, then $`[[t]]_𝒥`$ is an element of the domain of $`𝒥`$.
* By a $`𝒥`$-substitution we mean a finite mapping from variables to $`𝒥`$-terms which assigns to each variable $`x`$ in its domain a $`𝒥`$-term different from $`x`$. We write it as $`\{x_1/h_1,\mathrm{},x_n/h_n\}`$. $`\mathrm{}`$
The $`𝒥`$-substitutions generalize both the usual substitutions and the valuations, which assign domain values to variables. By adding to the language $`L`$ constants for each domain element and for each ground term we can reduce the $`𝒥`$-substitutions to the substitutions. We preferred not to do this to keep the notation simple.
In what follows we denote the empty $`𝒥`$-substitution by $`\epsilon `$ and arbitrary $`𝒥`$-substitutions by $`\theta ,\eta ,\gamma `$ with possible subscripts.
A more intuitive way of introducing $`𝒥`$-terms is as follows. Each ground term of $`s`$ of $`L`$ evaluates to a unique value in $`𝒥`$. Given a generalized term $`t`$ replace each maximal ground subterm of $`t`$ by its value in $`𝒥`$. The outcome is the $`𝒥`$-term $`[[t]]_𝒥`$.
We define the notion of an application of a $`𝒥`$-substitution $`\theta `$ to a generalized term $`t`$ in the standard way and denote it by $`t\theta `$. If $`t`$ is a term, then $`t\theta `$ does not have to be a term, though it is a generalized term.
###### Definition 2
* A composition of two $`𝒥`$-substitutions $`\theta `$ and $`\eta `$, written as $`\theta \eta `$, is defined as the unique $`𝒥`$-substitution $`\gamma `$ such that for each variable $`x`$
$$x\gamma =[[(x\theta )\eta ]]_𝒥.$$
$`\mathrm{}`$
Let us illustrate the introduced concepts by means of two examples.
###### Example 1
Take an arbitrary language of terms $`L`$. The Herbrand algebra $`Her`$ for $`L`$ is defined as follows:
* its domain is the set $`HU_L`$ of all ground terms of $`L`$ (usually called the Herbrand universe),
* if $`f`$ is an $`n`$-ary function symbol in $`L`$, then its interpretation is the mapping from $`(HU_L)^n`$ to $`HU_L`$ which maps the sequence $`t_1,\mathrm{},t_n`$ of ground terms to the ground term $`f(t_1,\mathrm{},t_n)`$.
Consider now a term $`s`$. Then $`[[s]]_{Her}`$ equals $`s`$ because in $`Her`$ every ground term evaluates to itself. So the notions of a term, a generalized term and a $`Her`$-term coincide. Consequently, the notions of substitutions and $`Her`$-substitutions coincide. $`\mathrm{}`$
###### Example 2
Take as the language of terms the language $`AE`$ of arithmetic expressions. Its binary function symbols are the usual $``$ (“times”), $`+`$ (“plus”) and $``$ (“minus”), and its unique binary symbol is $``$ (“unary minus”). Further, for each integer $`𝐤`$ there is a constant $`k`$.
As the algebra for $`AE`$ we choose the standard algebra $`Int`$ that consists of the set of integers with the function symbols interpreted in the standard way. In what follows we write the binary function symbols in the usual infix notation.
Consider the term $`sx+(((3+2)4)y)`$. Then $`[[s]]_{AE}`$ equals $`x+(\mathrm{𝟐𝟎}y)`$. Further, given the $`AE`$-substitution $`\theta :=\{x/\mathrm{𝟔}z,y/\mathrm{𝟑}\}`$ we have $`s\theta (\mathrm{𝟔}z)+(((3+2)4)\mathrm{𝟑})`$ and consequently, $`[[s\theta ]]_{AE}=(\mathrm{𝟔}z)+\mathrm{𝟏𝟕}`$. Further, given $`\eta :=\{z/\mathrm{𝟒}\}`$, we have $`\theta \eta =\{x/\mathrm{𝟐},y/\mathrm{𝟑},z/\mathrm{𝟒}\}`$. $`\mathrm{}`$
To define the meaning of an equation over an algebra $`𝒥`$ we view $`𝒥`$-substitutions as states and use a special state
* error, to indicate that it is not possible to determine effectively whether a solution to the equation $`s\theta =t\theta `$ in $`𝒥`$ exists.
We now define the semantics $`[[]]`$ of an equation between two generalized terms as follows:
$`\begin{array}{ccc}[[s=t]](\theta )\hfill & :=\hfill & \{\begin{array}{cc}\{\theta \{s\theta /[[t\theta ]]_𝒥\}\}\hfill & \text{if }s\theta \text{ is a variable that does not occur in }t\theta \text{,}\hfill \\ \{\theta \{t\theta /[[s\theta ]]_𝒥\}\}\hfill & \text{if }t\theta \text{ is a variable that does not occur in }s\theta \hfill \\ & \text{and }s\theta \text{ is not a variable,}\hfill \\ \{\theta \}\hfill & \text{if }[[s\theta ]]_𝒥\text{ and }[[t\theta ]]_𝒥\text{ are identical,}\hfill \\ \mathrm{}\hfill & \text{if }s\theta \text{ and }t\theta \text{ are ground and }[[s\theta ]]_𝒥[[t\theta ]]_𝒥\text{,}\hfill \\ \{error\}\hfill & \text{otherwise.}\hfill \end{array}\hfill \end{array}`$
It will become clear in the next section why we collect here the unique outcome into a set and why we “carry” $`\theta `$ in the answers.
Note that according to the above definition we have $`[[s=t]](\theta )=\{error\}`$ for the non-ground generalized terms $`s\theta `$ and $`t\theta `$ such that the $`𝒥`$-terms $`[[s\theta ]]_𝒥`$ and $`[[t\theta ]]_𝒥`$ are different. In some situations we could safely assert then that $`[[s=t]](\theta )=\{\theta \}`$ or that $`[[s=t]](\theta )=\mathrm{}`$. For example, for the standard algebra $`Int`$ for the language of arithmetic expressions we could safely assert that $`[[x+x=2x]](\theta )=\{\theta \}`$ and $`[[x+1=x]](\theta )=\mathrm{}`$ for any $`AE`$-substitution $`\theta `$.
The reason we did not do this was that we wanted to ensure that the semantics is uniform and decidable so that it can be implemented.
## 4 A Denotational Semantics for First-Order Logic
Consider now a first-order language with equality $``$. In this section we extend the semantics $`[[]]`$ to arbitrary first-order formulas from $``$ interpreted over an arbitrary interpretation. $`[[]]`$ depends on the considered interpretation but to keep the notation simple we do not indicate this dependence. This semantics is denotational in the sense that meaning of each formula is a mathematical function of the meanings of its direct constituents.
Fix an interpretation $``$. $``$ is based on some algebra $`𝒥`$. We define the notion of an application of a $`𝒥`$-substitution $`\theta `$ to a formula $`\varphi `$ of $``$, written as $`\varphi \theta `$, in the usual way.
Consider an atomic formula $`p(t_1,\mathrm{},t_n)`$ and a $`𝒥`$-substitution $`\theta `$. We denote by $`p_{}`$ the interpretation of $`p`$ in $``$.
We say that
* $`p(t_1,\mathrm{},t_n)\theta `$ is true if $`p(t_1,\mathrm{},t_n)\theta `$ is ground and $`([[t_1\theta ]]_𝒥,\mathrm{},[[t_n\theta ]]_𝒥)p_{}`$,
* $`p(t_1,\mathrm{},t_n)\theta `$ is false if $`p(t_1,\mathrm{},t_n)\theta `$ is ground and $`([[t_1\theta ]]_𝒥,\mathrm{},[[t_n\theta ]]_𝒥)p_{}`$.
In what follows we denote by $`Subs`$ the set of $`𝒥`$-substitutions and by $`𝒫(A)`$, for a set $`A`$, the set of all subsets of $`A`$.
For a given formula $`\varphi `$ its semantics $`[[\varphi ]]`$ is a mapping
$$[[\varphi ]]:Subs𝒫(Subs\{error\}).$$
The fact that the outcome of $`[[\varphi ]](\theta )`$ is a set reflects the possibility of a nondeterminism here modeled by the disjunction.
To simplify the definition we extend $`[[]]`$ to deal with subsets of $`Subs\{error\}`$ by putting
$$[[\varphi ]](error):=\{error\},$$
and for a set $`XSubs\{error\}`$
$$[[\varphi ]](X):=\underset{eX}{}[[\varphi ]](e).$$
Further, to deal with the existential quantifier, we introduce an operation $`DROP_x`$, where $`x`$ is a variable. First we define $`DROP_x`$ on the elements of $`Subs\{error\}`$ by putting for a $`𝒥`$-substitution $`\theta `$
$`\begin{array}{ccc}DROP_x(\theta )\hfill & :=\hfill & \{\begin{array}{cc}\theta \hfill & \text{if }x\text{ is not in the domain of }\theta \text{,}\hfill \\ \eta \hfill & \text{if }\theta \text{ is of the form }\eta \{x/s\}\text{,}\hfill \end{array}\hfill \end{array}`$
and
$$DROP_x(error):=error.$$
Then we extend it element-wise to subsets of $`Subs\{error\}`$, that is, by putting for a set $`XSubs\{error\}`$
$$DROP_x(X):=\{DROP_x(e)eX\}.$$
$`[[]]`$ is defined by structural induction as follows, where $`A`$ is an atomic formula different from $`s=t`$:
* $`[[A]](\theta ):=\{\begin{array}{cc}\{\theta \}\hfill & \text{if }A\theta \text{ is true,}\hfill \\ \mathrm{}\hfill & \text{if }A\theta \text{ is false,}\hfill \\ \{error\}\hfill & \text{otherwise, that is if }A\theta \text{ is not ground,}\hfill \end{array}`$
* $`[[\varphi _1\varphi _2]](\theta ):=[[\varphi _2]]([[\varphi _1]](\theta ))`$,
* $`[[\varphi _1\varphi _2]](\theta ):=[[\varphi _1]](\theta )[[\varphi _2]](\theta )`$,
* $`[[\neg \varphi ]](\theta ):=\{\begin{array}{cc}\{\theta \}\hfill & \text{if }[[\varphi ]](\theta )=\mathrm{}\text{,}\hfill \\ \mathrm{}\hfill & \text{if }\theta [[\varphi ]](\theta )\text{,}\hfill \\ \{error\}\hfill & \text{otherwise,}\hfill \end{array}`$
* $`[[x\varphi ]](\theta ):=DROP_y([[\varphi \{x/y\}]](\theta ))`$, where $`y`$ is a fresh variable.
To better understand this definition let us consider some simple examples that refer to the algebras discussed in Examples 1 and 2.
###### Example 3
Take an interpretation $``$ based on the Herbrand algebra $`Her`$. Then
$$[[f(x)=zg(z)=g(f(x))]](\{x/g(y)\})=[[g(z)=g(f(x))]](\theta )=\{\theta \},$$
where $`\theta :=\{x/g(y),z/f(g(y))\}`$. On the other hand
$$[[g(f(x))=g(z)]](\{x/g(y)\})=\{error\}.$$
$`\mathrm{}`$
###### Example 4
Take an interpretation $``$ based on the standard algebra $`AE`$ for the language of arithmetic expressions. Then
$$[[y=z1z=x+2]](\{x/\mathrm{𝟏}\})=[[z=x+2]](\{x/\mathrm{𝟏},y/z\mathrm{𝟏}\})=\{x/\mathrm{𝟏},y/\mathrm{𝟐},z/\mathrm{𝟑}\}.$$
Further,
$$[[y+1=z1]](\{y/\mathrm{𝟏},z/\mathrm{𝟑}\})=\{y/\mathrm{𝟏},z/\mathrm{𝟑}\}$$
and even
$$[[x(y+1)=(v+1)(z1)]](\{x/v+\mathrm{𝟏},y/\mathrm{𝟏},z/\mathrm{𝟑}\})=\{x/v+\mathrm{𝟏},y/\mathrm{𝟏},z/\mathrm{𝟑}\}.$$
On the other hand
$$[[y1=z1]](\epsilon )=\{error\}.$$
$`\mathrm{}`$
The first example shows that the semantics given here is weaker than the one provided by the logic programming. In turn, the second example shows that our treatment of arithmetic expressions is more general than the one provided by Prolog.
This definition of denotational semantics of first-order formulas combines a number of ideas put forward in the area of semantics of imperative programming languages and the field of logic programming.
First, for an atomic formula $`A`$, when $`A\theta `$ is ground, its meaning coincides with the meaning of a Boolean expression given in de Bakker \[dB80, page 270\]. In turn, the meaning of the conjunction and of the disjunction follows \[dB80, page 270\] in the sense that the conjunction corresponds to the sequential composition operation “;” and the disjunction corresponds to the “don’t know” nondeterministic choice, denoted there by $``$.
Next, the meaning of the negation is inspired by its treatment in logic programming. To be more precise we need the following observations the proofs of which easily follow by structural induction.
###### Note 1
1. If $`\eta [[\varphi ]](\theta )`$, then $`\eta =\theta \gamma `$ for some $`𝒥`$-substitution $`\gamma `$.
2. If $`\varphi \theta `$ is ground, then $`[[\varphi ]](\theta )\{\theta \}`$. $`\mathrm{}`$
First, we interpret $`[[\varphi ]](\theta )Subs\mathrm{}`$ as the statement “the query $`\varphi \theta `$ succeeds”. More specifically, if $`\eta [[\varphi ]](\theta )`$, then by Note 1(i) for some $`\gamma `$ we have $`\eta =\theta \gamma `$.
In general, $`\gamma `$ is of course not unique: take for example $`\theta :=\{x/0\}`$ and $`\eta =\theta `$. Then both $`\eta =\theta \epsilon `$ and $`\eta =\theta \theta `$. However, it is easy to show that if $`\eta `$ is less general than $`\theta `$, then in the set $`\{\gamma \eta =\theta \gamma \}`$ the $`𝒥`$-substitution with the smallest domain is uniquely defined. In what follows given $`𝒥`$-substitutions $`\eta `$ and $`\theta `$ such that $`\eta `$ is less general than $`\theta `$, when writing $`\eta =\theta \gamma `$ we always refer to this uniquely defined $`\gamma `$.
Now we interpret $`\theta \gamma [[\varphi ]](\theta )`$ as the statement “$`\gamma `$ is the computed answer substitution for the query $`\varphi \theta `$”. In turn, we interpret $`[[\varphi ]](\theta )=\mathrm{}`$ as the statement “the query $`\varphi \theta `$ finitely fails”.
Suppose now that $`[[\varphi ]](\theta )Subs\mathrm{}`$, which means that the query $`\varphi \theta `$ succeeds. Assume additionally that $`\varphi \theta `$ is ground. Then by Note 1(ii) $`\theta [[\varphi ]](\theta )`$ and consequently by the definition of the meaning of negation $`[[\neg \varphi ]](\theta )=\mathrm{}`$, which means that the query $`\neg \varphi \theta `$ finitely fails.
In turn, suppose that $`[[\varphi ]](\theta )=\mathrm{}`$, which means that the query $`\varphi \theta `$ finitely fails. By the definition of the meaning of negation $`[[\neg \varphi ]](\theta )=\{\theta \}`$, which means that the query $`\neg \varphi \theta `$ succeeds with the empty computed answer substitution.
This explains the relation with the “negation as finite failure” rule according to which for a ground query $`Q`$:
* if $`Q`$ succeeds, then $`\neg Q`$ finitely fails,
* if $`Q`$ finitely fails, then $`\neg Q`$ succeeds with the empty computed answer substitution.
In fact, our definition of the meaning of negation corresponds to a generalization of the negation as finite failure rule already mentioned in Clark \[Cla78\], according to which the requirement that $`Q`$ is ground is dropped and the first item is replaced by:
* if $`Q`$ succeeds with the empty computed answer substitution, then $`\neg Q`$ finitely fails.
Finally, the meaning of the existential quantification corresponds to the meaning of the block statement in imperative languages, see, e.g., de Bakker \[dB80, page 226\], with the important difference that the local variable is not initialized. From this viewpoint the existential quantifier $`x`$ corresponds to the declaration of the local variable $`x`$. The $`DROP_x`$ operation was introduced in Clarke \[Cla79\] to deal with the declarations of local variables.
We do not want to make the meaning of the formula $`x\varphi `$ dependent on the choice of $`y`$. Therefore we postulate that for any fresh variable $`y`$ the set $`DROP_y([[\varphi \{x/y\}]](\theta ))`$ is a meaning of $`x\varphi `$ given a $`𝒥`$-substitution $`\theta `$. Consequently, the semantics of $`x\varphi `$ has many outcomes, one for each choice of $`y`$. This “multiplicity” of meanings then extends to all formulas containing the existential quantifier. So for example for any variable $`y`$ different from $`x`$ and $`z`$ the $`𝒥`$-substitution $`\{z/f(y)\}`$ is the meaning of $`x(z=f(x))`$ given the empty $`𝒥`$-substitution $`\epsilon `$.
## 5 Soundness
To relate the introduced semantics to the notion of truth we first formalize the latter using the notion of a $`𝒥`$-substitution instead of the customary notion of a valuation.
Consider a first-order language $``$ with equality and an interpretation $``$ for it based on some algebra $`𝒥`$. Let $`\theta `$ be a $`𝒥`$-substitution. We define the relation $`_\theta \varphi `$ for a formula $`\varphi `$ by structural induction. First we assume that $`\theta `$ is defined on all free variables of $`\varphi `$ and put
* $`_\theta s=t`$ iff $`[[s\theta ]]_𝒥`$ and $`[[t\theta ]]_𝒥`$ coincide,
* $`_\theta p(t_1,\mathrm{},t_n)`$ iff $`p(t_1,\mathrm{},t_n)\theta `$ is ground and $`([[t_1\theta ]]_𝒥,\mathrm{},[[t_n\theta ]]_𝒥)p_{}`$.
In other words, $`_\theta p(t_1,\mathrm{},t_n)`$ iff $`p(t_1,\mathrm{},t_n)\theta `$ is true. The definition extends to non-atomic formulas in the standard way.
Now assume that $`\theta `$ is not defined on all free variables of $`\varphi `$. We put
* $`_\theta \varphi `$ iff $`_\theta x_1,\mathrm{},x_n\varphi `$ where $`x_1,\mathrm{},x_n`$ is the list of the free variables of $`\varphi `$ that do not occur in the domain of $`\theta `$.
Finally,
* $`\varphi `$ iff $`_\theta \varphi `$ for all $`𝒥`$-substitutions $`\theta `$.
To prove the main theorem we need the following notation. Given a $`𝒥`$-substitution $`\eta :=\{x_1/h_1,\mathrm{},x_n/h_n\}`$ we define $`\eta :=x_1=h_1\mathrm{}x_n=h_n`$.
In the discussion that follows the following simple observation will be useful.
###### Note 2
For all $`𝒥`$-substitutions $`\theta `$ and formulas $`\varphi `$
$$_\theta \varphi \text{ iff }\theta \varphi .$$
$`\mathrm{}`$
The following theorem now shows correctness of the introduced semantics with respect to the notion of truth.
###### Theorem 5.1 (Soundness)
Consider a first-order language $``$ with equality and an interpretation $``$ for it based on some algebra $`𝒥`$. Let $`\varphi `$ be a formula of $``$ and $`\theta `$ a $`𝒥`$-substitution.
1. For each $`𝒥`$-substitution $`\eta [[\varphi ]](\theta )`$
$$_\eta \varphi .$$
2. If $`error[[\varphi ]](\theta )`$, then
$$\varphi \theta \underset{i=1}{\overset{k}{}}𝐲_i\eta _i,$$
where $`[[\varphi ]](\theta )=\{\theta \eta _1,\mathrm{},\theta \eta _k\}`$, and for $`i[1..k]`$ $`𝐲_i`$ is a sequence of variables that appear in the range of $`\eta _i`$.
Note that by $`(ii)`$ if $`[[\varphi ]](\theta )=\mathrm{}`$, then
$$_\theta \neg \varphi .$$
In particular, if $`[[\varphi ]](\epsilon )=\mathrm{}`$, then
$$\neg \varphi .$$
Proof. The proof proceeds by simultaneous induction on the structure of the formulas.
$`\varphi `$ is $`s=t`$.
If $`\eta [[\varphi ]](\theta )`$, then three possibilities arise.
1. $`s\theta `$ is a variable that does not occur in $`t\theta `$.
Then $`[[s=t]](\theta )=\{\theta \{s\theta /[[t\theta ]]_𝒥\}\}`$ and consequently $`\eta =\theta \{s\theta /[[t\theta ]]_𝒥\}`$. So $`_\eta (s=t)`$ holds since $`s\eta =[[t\theta ]]_𝒥`$ and $`t\eta =t\theta `$.
2. $`t\theta `$ is a variable that does not occur in $`s\theta `$ and $`s\theta `$ is not a variable.
Then $`[[s=t]](\theta )=\{\theta \{t\theta /[[s\theta ]]_𝒥\}\}`$. This case is symmetric to 1.
3. $`[[s\theta ]]_𝒥`$ and $`[[t\theta ]]_𝒥`$ are identical.
Then $`\eta =\theta `$, so $`_\eta (s=t)`$ holds.
If $`error[[\varphi ]](\theta )`$, then four possibilities arise.
1. $`s\theta `$ is a variable that does not occur in $`t\theta `$.
Then $`[[s=t]](\theta )=\{\theta \{s\theta /[[t\theta ]]_𝒥\}\}`$. We have $`(s=t)\theta s\theta =[[t\theta ]]_𝒥`$.
2. $`t\theta `$ is a variable that does not occur in $`s\theta `$ and $`s\theta `$ is not a variable.
Then $`[[s=t]](\theta )=\{\theta \{t\theta /[[s\theta ]]_𝒥\}\}`$. This case is symmetric to 1.
3. $`[[s\theta ]]_𝒥`$ and $`[[t\theta ]]_𝒥`$ are identical.
Then $`[[s=t]](\theta )=\{\theta \}`$. We have $`[[s=t]](\theta )=\{\theta \epsilon \}`$ and $`_\theta s=t`$, so $`(s=t)\theta \epsilon `$, since $`\epsilon `$ is vacuously true.
4. $`s\theta `$ and $`t\theta `$ are ground $`𝒥`$-terms and $`[[s\theta ]]_𝒥[[t\theta ]]_𝒥`$.
Then $`[[s=t]](\theta )=\mathrm{}`$ and $`_\theta \neg (s=t)`$, so $`(s=t)\theta falsum`$, where $`falsum`$ denotes the empty disjunction.
$`\varphi `$ is an atomic formula different from $`s=t`$.
If $`\eta [[\varphi ]](\theta )`$, then $`\eta =\theta `$ and $`\varphi \theta `$ is true. So $`_\theta \varphi `$, i.e., $`_\eta \varphi `$.
If $`error[[\varphi ]](\theta )`$, then either $`[[\varphi ]](\theta )=\{\theta \}`$ or $`[[\varphi ]](\theta )=\mathrm{}`$. In both cases the argument is the same as in case 3. and 4. for the equality $`s=t`$.
Note that in both cases we established a stronger form of $`(ii)`$ in which each list $`𝐲_i`$ is empty, i.e., no quantification over the variables in $`𝐲_i`$ appears.
$`\varphi `$ is $`\varphi _1\varphi _2`$. This is the most elaborate case.
If $`\eta [[\varphi ]](\theta )`$, then for some $`𝒥`$-substitution $`\gamma `$ both $`\gamma [[\varphi _1]](\theta )`$ and $`\eta [[\varphi _2]](\gamma )`$. By induction hypothesis both $`_\gamma \varphi _1`$ and $`_\eta \varphi _2`$. But by Note 1(i) $`\eta `$ is less general than $`\gamma `$, so $`_\eta \varphi _1`$ and consequently $`_\eta \varphi _1\varphi _2`$.
If $`error[[\varphi ]](\theta )`$, then for some $`XSubs`$ both $`[[\varphi _1]](\theta )=X`$ and $`error[[\varphi _2]](\eta )`$ for all $`\eta X`$.
By induction hypothesis
$$\varphi _1\theta \underset{i=1}{\overset{k}{}}𝐲_i\eta _i,$$
where $`X=\{\theta \eta _1,\mathrm{},\theta \eta _k\}`$ and for $`i[1..k]`$ $`𝐲_i`$ is a sequence of variables that appear in the range of $`\eta _i`$. Hence
$$(\varphi _1\varphi _2)\theta \underset{i=1}{\overset{k}{}}(𝐲_i\eta _i\varphi _2\theta ),$$
so by appropriate renaming of the variables in the sequences $`𝐲_i`$
$$(\varphi _1\varphi _2)\theta \underset{i=1}{\overset{k}{}}𝐲_i(\eta _i\varphi _2\theta ).$$
But for any $`𝒥`$-substitution $`\delta `$ and a formula $`\psi `$
$$\delta \psi \delta \psi \delta ,$$
so
$$(\varphi _1\varphi _2)\theta (\underset{i=1}{\overset{k}{}}𝐲_i(\eta _i\varphi _2\theta \eta _i).$$
(5)
Further, we have for $`i[1..k]`$
$$[[\varphi _2]](\theta \eta _i)=\{\theta \eta _i\gamma _{i,j}j[1..\mathrm{}_i]\}$$
for some $`𝒥`$-substitutions $`\gamma _{i,1},\mathrm{},\gamma _{i,\mathrm{}_i}`$. So
$$[[\varphi _1\varphi _2]](\theta )=\{\theta \eta _i\gamma _{i,j}i[1..k],j[1..\mathrm{}_i]\}.$$
By induction hypothesis we have for $`i[1..k]`$
$$\varphi _2\theta \eta _i\underset{j=1}{\overset{\mathrm{}_i}{}}𝐯_{i,j}\gamma _{i,j},$$
where for $`i[1..k]`$ and $`j[1..\mathrm{}_i]`$ $`𝐯_{i,j}`$ is a sequence of variables that appear in the range of $`\gamma _{i,j}`$.
Using (5) by appropriate renaming of the variables in the sequences $`𝐯_{i,j}`$ we now conclude that
$$(\varphi _1\varphi _2)\theta \underset{i=1}{\overset{k}{}}\underset{j=1}{\overset{\mathrm{}_i}{}}𝐲_i𝐯_{i,j}(\eta _i\gamma _{i,j}),$$
so
$$(\varphi _1\varphi _2)\theta \underset{i=1}{\overset{k}{}}\underset{j=1}{\overset{\mathrm{}_i}{}}𝐲_i𝐯_{i,j}\eta _i\gamma _{i,j},$$
since the domains of $`\eta _i`$ and $`\gamma _{i,j}`$ are disjoint and for any $`𝒥`$-substitutions $`\gamma `$ and $`\delta `$ with disjoint domains we have
$$\gamma \delta \gamma \delta .$$
$`\varphi `$ is $`\varphi _1\varphi _2`$.
If $`\eta [[\varphi ]](\theta )`$, then either $`\eta [[\varphi _1]](\theta )`$ or $`\eta [[\varphi _2]](\theta )`$, so by induction hypothesis either $`_\eta \varphi _1`$ or $`_\eta \varphi _2`$. In both cases $`_\eta \varphi _1\varphi _2`$ holds.
If $`error[[\varphi ]](\theta )`$, then for some $`𝒥`$-substitutions $`\eta _1,\mathrm{},\eta _k`$
$$[[\varphi _1]](\theta )=\{\theta \eta _1,\mathrm{},\theta \eta _k\},$$
where $`k0`$, for some $`𝒥`$-substitutions $`\eta _{k+1},\mathrm{},\eta _{k+\mathrm{}}`$,
$$[[\varphi _2]](\theta )=\{\theta \eta _{k+1},\mathrm{},\theta \eta _{k+\mathrm{}}\},$$
where $`\mathrm{}0`$, and
$$[[\varphi _1\varphi _2]](\theta )=\{\theta \eta _1,\mathrm{},\theta \eta _{k+\mathrm{}}\}.$$
By induction hypothesis both
$$\varphi _1\theta \underset{i=1}{\overset{k}{}}𝐲_i\eta _i$$
and
$$\varphi _2\theta \underset{i=k+1}{\overset{k+\mathrm{}}{}}𝐲_i\eta _i$$
for appropriate sequences of variables $`𝐲_i`$. So
$$(\varphi _1\varphi _2)\theta \underset{i=1}{\overset{k+\mathrm{}}{}}𝐲_i\eta _i.$$
$`\varphi `$ is $`\neg \varphi _1`$.
If $`\eta [[\varphi ]](\theta )`$, then $`\eta =\theta `$ and $`[[\varphi _1]](\theta )=\mathrm{}`$. By induction hypothesis $`_\theta \neg \varphi _1`$, i.e., $`_\eta \neg \varphi _1`$.
If $`error[[\varphi ]](\theta )`$, then either $`[[\varphi ]](\theta )=\{\theta \}`$ or $`[[\varphi ]](\theta )=\mathrm{}`$. In the former case $`[[\varphi ]](\theta )=\{\theta \epsilon \}`$, so $`[[\varphi _1]](\theta )=\mathrm{}`$. By induction hypothesis $`_\theta \neg \varphi _1`$, i.e., $`(\neg \varphi _1)\theta \epsilon `$, since $`\epsilon `$ is vacuously true. In the latter case $`\theta [[\varphi _1]](\theta )`$, so by induction hypothesis $`_\theta \varphi _1`$, i.e., $`(\neg \varphi _1)\theta falsum`$.
$`\varphi `$ is $`x\varphi _1`$.
If $`\eta [[\varphi ]](\theta )`$, then $`\eta DROP_y([[\varphi _1\{x/y\}]](\theta ))`$ for some fresh variable $`y`$. So either (if $`y`$ is not in the domain of $`\eta `$) $`\eta [[\varphi _1\{x/y\}]](\theta )`$ or for some $`𝒥`$-term $`s`$ we have $`\eta \{y/s\}[[\varphi _1\{x/y\}]](\theta )`$. By induction hypothesis in the former case $`_\eta \varphi _1\{x/y\}`$ and in the latter case $`_{\eta \{y/s\}}\varphi _1\{x/y\}`$. In both cases $`y(\varphi _1\{x/y\}\eta )`$, so, since $`y`$ is fresh, $`(y\varphi _1\{x/y\})\eta `$ and consequently $`(x\varphi _1)\eta `$, i.e., $`_\eta x\varphi _1`$.
If $`error[[\varphi ]](\theta )`$, then $`error[[\varphi _1\{x/y\}]](\theta )`$, as well, where $`y`$ is a fresh variable. By induction hypothesis
$$\varphi _1\{x/y\}\theta \underset{i=1}{\overset{k}{}}𝐲_i\eta _i,$$
(6)
where
$$[[\varphi _1\{x/y\}]](\theta )=\{\theta \eta _1,\mathrm{},\theta \eta _k\}$$
(7)
and for $`i[1..k]`$ $`𝐲_i`$ is a sequence of variables that appear in the range of $`\eta _i`$.
Since $`y`$ is fresh, we have $`y(\varphi _1\{x/y\}\theta )(y\varphi _1\{x/y\})\theta `$ and $`(y\varphi _1\{x/y\})\theta (x\varphi _1)\theta `$. So (6) implies
$$(x\varphi _1)\theta \underset{i=1}{\overset{k}{}}y𝐲_i\eta _i.$$
But for $`i[1..k]`$
$$y\eta _iyDROP_y(\eta _i),$$
since if $`y/s\eta _i`$, then the variable $`y`$ does not appear in $`s`$. So
$$(x\varphi _1)\theta \underset{i=1}{\overset{k}{}}𝐲_iyDROP_y(\eta _i).$$
(8)
Now, by (7)
$$[[x\varphi _1]](\theta )=\{DROP_y(\theta \eta _1),\mathrm{},DROP_y(\theta \eta _k)\}.$$
But $`y`$ does not occur in $`\theta `$, so we have for $`i[1..k]`$
$$DROP_y(\theta \eta _i)=\theta DROP_y(\eta _i)$$
and consequently
$$[[x\varphi _1]](\theta )=\{\theta DROP_y(\eta _1),\mathrm{},\theta DROP_y(\eta _k)\}.$$
This by virtue of (8) concludes the proof. $`\mathrm{}`$
Informally, $`(i)`$ states that every computed answer substitution of $`\varphi \theta `$ validates it. It is useful to point out that $`(ii)`$ is a counterpart of Theorem 3 in Clark \[Cla78\]. Intuitively, it states that a query is equivalent to the disjunction of its computed answer substitutions written out in an equational form (using the $`\eta `$ notation). In our case this property holds only if $`error`$ is not a possible outcome. Indeed, if $`[[s=t]](\theta )=\{error\}`$, then nothing can be stated about the status of the statement $`(s=t)\theta `$.
Note that in case $`error[[\varphi ]](\theta )`$, $`(ii)`$ implies $`(i)`$ by virtue of Note 2. On the other hand, if $`error[[\varphi ]](\theta )`$, then $`(i)`$ can still be applicable while $`(ii)`$ not.
Additionally existential quantifiers have to be used in an appropriate way. The formulas of the form $`𝐲\eta `$ also appear in Maher \[Mah88\] in connection with a study of the decision procedures for the algebras of trees. In fact, there are some interesting connections between this paper and ours that could be investigated in a closer detail.
## 6 Conclusions and Future Work
In this paper we provided a denotational semantics to first-order logic formulas. This semantics is a counterpart of the operational semantics introduced in Apt and Bezem \[AB99\]. The important difference is that we provide here a more general treatment of equality according to which a non-ground term can be assigned to a variable. This realizes logical variables in the framework of Apt and Bezem \[AB99\]. This feature led to a number of complications in the proof of the Soundness Theorem 5.1.
One of the advantages of this theorem is that it allows us to reason about the considered program simply by comparing it to the formula representing its specification. In the case of operational semantics this was exemplified in Apt and Bezem \[AB99\] by showing how to verify non-trivial Alma-0 programs that do not include destructive assignment.
Note that it is straightforward to extend the semantics here provided to other well-known programming constructs, such as destructive assignment, while construct and recursion. However, as soon as a destructive assignment is introduced, the relation with the definition of truth in the sense of Soundness Theorem 5.1 is lost and the just mentioned approach to program verification cannot be anymore applied. In fact, the right approach to the verification of the resulting programs is an appropriately designed Hoare’s logic or the weakest precondition semantics.
The work here reported can be extended in several directions. First of all, it would be useful to prove equivalence between the operational and denotational semantics. Also, it would interesting to specialize the introduced semantics to specific interpretations for which the semantics could generate less often an error. Examples are Herbrand interpretations for an arbitrary first-order language in which the meaning of equalities could be rendered using most general unifiers, and the standard interpretation over reals for the language defining linear equations; these equations can be handled by means of the usual elimination procedure. In both cases the equality could be dealt with without introducing the error state at all.
Other possible research directions were already mentioned in Apt and Bezem \[AB99\]. These involved addition of recursive procedures, of constraints, and provision of a support for automated verification of programs written in Alma-0. The last item there mentioned, relation to dynamic predicate logic, was in the meantime extensively studied in the work of van Eijck \[vE98\] who, starting with Apt and Bezem \[AB99\], defined a number of semantics for dynamic predicate logic in which the existential quantifier has a different, dynamic scope. This work was motivated by applications in natural language processing.
## Acknowledgments
Many thanks to Marc Bezem for helpful discussions on the subject of this paper.
|
warning/0005/astro-ph0005229.html
|
ar5iv
|
text
|
# The Case for an Accelerating Universe from Supernovae To Appear as an Invited Review for PASP
## 1 Introduction
Two teams have presented observational evidence from high-redshift type Ia supernovae (SNe Ia) that the expansion of the Universe is accelerating, propelled by vacuum energy (Riess et al. 1998; Perlmutter et al. 1999). The primary evidence for this hypothesis is the faintness of distant SNe Ia relative to their expected brightness in a decelerating Universe. The question we propose to answer in this review is whether the observations of distant supernovae compel us to conclude that the expansion is accelerating.
## 2 Supernova Measurements
### 2.1 Past Work
Supernovae (SNe) have a long history of employment in the quest to measure Hubble’s constant (see Branch 1998 for review) and currently provide a column of support for a strong consensus that $`H_0`$=60 to 70 km s<sup>-1</sup> Mpc<sup>-1</sup>. The history of utilizing supernovae to measure the time evolution of the expansion rate is far briefer. All initial proposals for using high-redshift SNe to constrain global deceleration recognized the necessity of an optical space telescope to collect the data. Wagoner (1977) envisioned application of Baade’s method or the the Expanding Photosphere method (Kirshner & Kwan 1974) to measure the angular diameter distance of Type I (hydrogen-deficient) and Type II (hydrogen-rich) supernovae at $`z=0.3`$. Colgate (1979) demonstrated even greater prescience, suggesting that SNe I at $`z=1`$ could be used as standard candles for “determining the cosmological constant with greater accuracy than other standard candles.” Further thoughts by Tammann (1979) included the necessity to account for the redshift of the observed light using spectrophotometry of SNe in the ultraviolet (i.e., $`K`$-corrections), host galaxy extinction, and time dilation of the light curves.
Yet even before the launch of the Hubble Space Telescope (HST), a persistent two-year ground-based effort by a Danish group was rewarded by the discovery of their first (and only reported) high-redshift SN Ia, SN 1988U at $`z=0.31`$ (Norgarrd-Nielsen et al. 1989) as well as modest bounds on the deceleration parameter, $`0.6<q_0<2.5`$. This team employed modern image processing techniques to scale the brightness and resolution of images of high-redshift clusters to match previous images and looked for supernovae in the difference frames. Unfortunately, the project’s low discovery rate coupled with the dispersion of SNe Ia when treated as perfect standard candles ($``$0.5 mag) suggested that the determination of the deceleration parameter would require a scientific lifetime.
However, great progress was made by the Supernova Cosmology Project (SCP) in the detection rate of high-redshift SNe Ia by employing large-format CCDs, large-aperture telescopes, and more sophisticated image-analysis techniques (Perlmutter et al. 1995). These advances led to the detection of seven SNe Ia at $`z0.4`$ between 1992 and 1994, yielding a confidence region that suggested of a flat, $`\mathrm{\Lambda }=0`$ universe but with a large range of uncertainty (Perlmutter et al. 1997).
The High-$`z`$ Supernova Search Team (HZT) joined the hunt for high-redshift SNe Ia with their discovery of SN 1995K at $`z=0.48`$ (Schmidt et al. 1998). Both teams made rapid improvements in their ability to discover ever greater numbers of SNe Ia at still larger redshifts. One-time redshift record holders included SN 1997ap at $`z=0.83`$ (SCP; Perlmutter et al. 1998), SN 1997ck at $`z=0.97`$ (HZT; Garnavich et al. 1998a), SN 1998eq at $`z=1.20`$ (SCP; Aldering et al. 1998) and SN 1999fv at z=1.23 (HZT; Tonry et al. 1999). Before either teams’ claimed samples of high-redshift SNe Ia were large enough to detect the acceleration signal, both teams found the data to be inconsistent with a Universe closed by matter (Garnavich et al. 1998a; Perlmutter et al. 1998).
These observational feats were preceded by increased understanding and ability to make use of SNe Ia observations to constrain the cosmological parameters. Empirical correlations between SN Ia light curve shapes and peak luminosity improved the precision of distance estimates beyond the standard candle model (Phillips 1993; Hamuy et al. 1995; Riess, Press, Kirshner 1995; Perlmutter et al. 1995; Tripp 1997; Saha et al. 1999; Parodi et al. 2000). Studies of SN Ia colors provided the means to distinguish supernovae which were reddened by dust from those which were intrinsically red (Riess, Press, Kirshner 1996; Riess et al. 1998; Phillips et al. 1998; Tripp & Branch 1999). Goobar & Perlmutter (1995) showed that measurements of SNe Ia at different redshift intervals could break degeneracies between $`\mathrm{\Omega }_M`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$. Additional work on cross-filter $`K`$-corrections provided the ability to accurately transform the observations of high-redshift SNe Ia to the rest-frame (Kim, Goobar, & Perlmutter 1996).
### 2.2 Observations
In order to resolve whether the results from high-redshift SNe Ia are compelling it is important to review how the measurements were obtained.
Both the SCP and the HZT detected their samples of high-redshift SNe Ia by using large-format CCDs at telescopes with large apertures (most commonly the CTIO 4-m Blanco telescope). Using two sets of deep images in the $`R`$ or $`I`$-band spaced across a lunation, the “template” images are subtracted from the second-epoch images and automated software searches for sources in the difference images whose intensity surpasses a specific threshold. The observations taken in pairs are spaced over a small time interval of a few minutes to eliminate moving transients. Human inspectors “filter” the automated results in an effort to maximize the likelihood that candidates are supernovae (Schmidt et al. 1998; Perlmutter et al. 1999). Because the resources available for collecting spectral identifications for all the candidates are insufficient, candidates which appear most likely to be SNe Ia are given priority. The factors which are favored in the human selection criteria include candidates which are separated from the host galaxy nuclei and those that show good contrast with the host galaxy (especially those with little or no apparent host). The signal-to-noise ratios of the identifying spectra vary greatly (Perlmutter et al. 1995, 1998; Riess et al. 1998) but have improved with the availability of the Keck Telescope and an increased emphasis on the search for clues of SN Ia evolution. Most of the spectral identifications were made by visual comparison to template spectra of nearby SNe Ia. More recently automated cross-correlation techniques have been employed (Riess et al. 1997). Approximately half of the SN Ia redshifts were determined from narrow emission lines or Ca H and K absorption; the rest were derived from the broad supernova features. Although formal statistics are not currently available, we are aware that the SCP candidates have yielded a greater fraction of SNe Ia than the HZT candidates, a significant number of which turn out to be SNe II. Taken together, about half of the two teams’ candidates are revealed to be SNe Ia, with the rest classified as SNe II, AGNs, flare stars, or unclassified objects. SNe Ia in the desired redshift range are monitored photometrically in two colors by observatories around the world using red-sensitive passbands; HST has monitored the light curves of about two dozen of these objects.
The initial search template images are eventually replaced by deeper template images obtained with good seeing and taken one or more years after discovery when the SNe Ia have faded by five to eight magnitudes. More recently, repeated searching of the same fields has provided deep template images before the SN explosion. After the SN images are bias corrected and flat-fielded they are geometrically aligned, and the resolution and intensity are scaled to match those of the templates. After subtracting the host galaxy light the SN magnitudes are measured. The SCP uses aperture photometry; the HZT fits point-spread-functions. Uncertainties are determined synthetically by the injection of artificial SNe of known brightness. The SCP uses standard passbands and Landolt (1982) standards of comparison while the HZT uses a custom passband system which is transformed to the Landolt scale (Schmidt et al. 1998). Custom cross-band $`K`$-corrections are calculated using spectrophotometry of nearby SNe Ia whose colors are reddened to match the high-redshift objects.
The distances are measured by fitting empirical families of light curves to the flux observations of individual supernovae. The measured distances are derived from the luminosity distance,
$$D_L=\left(\frac{}{4\pi }\right)^{\frac{1}{2}}$$
(1)
where $``$ and $``$ are the SN’s intrinsic luminosity and observed flux, respectively and $`D_L`$ is in Mpc. Alternately, a logarithmic luminosity distance (i.e., the distance modulus) is used:
$$.\mu =mM=5\mathrm{log}D_L+25,$$
(2)
where $`M`$ is the SN’s absolute magnitude and $`m`$ is the observed magnitude in a given passband.
Three different light-curve fitting methods have been used to measure the distances, each of which determines the shape of the best-fitting light curve to identify the individual luminosity of the SN Ia. The HZT has used both the multi-color light-curve shape method (MLCS; Riess, Press, & Kirshner 1996; Riess et al. 1998) and a template fitting approach based on the parameter $`\mathrm{\Delta }m_{15}(B)`$ (Phillips 1993; Phillips et al. 1999), while the SCP uses the “stretch method” (Perlmutter et al. 1995; 1999).
Nearby SNe Ia provide both the measure of the Hubble flow and the means to calibrate the relationship between light-curve shape and luminosity. The SCP uses $``$20 nearby SNe Ia in the Hubble flow from the Calán/Tololo Survey (Hamuy et al. 1996) while the HZT adds to this set an equal number of SNe from the CfA Sample (Riess et al. 1999).
Dust extinction is handled somewhat differently by the two teams. The HZT measures the extinction from the $`BV`$ reddening (i.e., the color excess) and then combines this measurement in a Bayesian formalism using a prior host galaxy extinction distribution calculated by Hatano, Branch, & Deaton (1998). This treatment assumes that extinction makes supernovae appear dimmer (farther), never brighter (closer). The SCP uses a number of different approaches (including the HZT approach), but favors making no individual extinction corrections and discarding outliers. Figure 1 shows a single Hubble diagram made with the data from both teams (Riess et al. 1998; Perlmutter et al. 1999).
The measured distances are then compared to those expected for their redshifts as a function of the cosmological parameters $`\mathrm{\Omega }_M`$, $`\mathrm{\Omega }_\mathrm{\Lambda }`$, and $`H_0`$:
$$D_L=cH_0^1(1+z)\left|\mathrm{\Omega }_k\right|^{1/2}sinn\{\left|\mathrm{\Omega }_k\right|^{1/2}_0^z𝑑z[(1+z)^2(1+\mathrm{\Omega }_Mz)z(2+z)\mathrm{\Omega }_\mathrm{\Lambda }]^{1/2}\},$$
(3)
where $`\mathrm{\Omega }_k=1\mathrm{\Omega }_M\mathrm{\Omega }_\mathrm{\Lambda }`$, and $`sinn`$ is $`\mathrm{sinh}`$ for $`\mathrm{\Omega }_k0`$ and $`\mathrm{sin}`$ for $`\mathrm{\Omega }_k0`$ (Carroll, Press, & Turner 1992). The likelihoods for cosmological parameters are determined by minimizing the $`\chi ^2`$ statistic between the measured and predicted distances (Riess et al. 1998; Perlmutter et al. 1999). Determination of $`\mathrm{\Omega }_M`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$ are independent of the value of $`H_0`$ or the absolute magnitude calibration of SNe Ia. The confidence intervals in the $`\mathrm{\Omega }_M`$-$`\mathrm{\Omega }_\mathrm{\Lambda }`$ plane determined by both teams are shown in Figure 2. In this plane, acceleration is defined by the region where the deceleration parameter, $`q_0`$, is negative,
$$q_0=\frac{\mathrm{\Omega }_M}{2}\mathrm{\Omega }_\mathrm{\Lambda }<0.$$
(4)
### 2.3 Results
The visual impression from Figure 2 is that the supernova observations favor the parameter space containing a positive cosmological constant and an accelerating Universe with high statistical confidence. The confidence regions determined by the two teams are in remarkably good agreement. Both teams claim that these results do not result from chance with more than 99% confidence (Riess et al. 1998; Perlmutter et al. 1999). However, the specific confidence of these cosmological conclusions depends on which parameter space intervals one considers to be equally likely, a priori. This point is addressed by Drell, Loredo, & Wasserman (2000) who suggest that models with $`\mathrm{\Omega }_\mathrm{\Lambda }`$=0 and $`\mathrm{\Omega }_\mathrm{\Lambda }0`$ could be considered equally probable, a priori. While Riess et al. (1998) and Perlmutter et al. (1999) considered the probability that $`\mathrm{\Omega }_\mathrm{\Lambda }>0`$ with a flat prior in a linear $`\mathrm{\Omega }_\mathrm{\Lambda }`$ space, an alternative is to use a flat prior in a logarithmic $`\mathrm{\Omega }_\mathrm{\Lambda }`$ space.
Another useful way to quantify the SN Ia constraints has been given by Perlmutter et al. (1999) as $`0.8\mathrm{\Omega }_M0.6\mathrm{\Omega }_\mathrm{\Lambda }=0.2\pm 0.1`$, a result which applies equally well to the Riess et al. (1998) data.
A more illuminating way to quantify the evidence for an accelerating Universe is to consider how the SN Ia distances depart from decelerating or “coasting” models. The average high-redshift SN Ia is 0.19 mag dimmer or $``$10% farther than expected for a Universe with no cosmological constant and negligible matter ($`\mathrm{\Omega }_\mathrm{\Lambda }=0`$, $`\mathrm{\Omega }_M=0`$). Of course it’s apparent that the Universe has more than negligible matter and the current consensus from the mass, light, X-ray emission, numbers, and motions of clusters of galaxies is that $`\mathrm{\Omega }_M0.3`$ (Carlberg et al. 1996; Bahcall, Fan, & Cen 1997; Lin et al. 1996; Strauss & Willick 1995). The high-redshift SNe Ia from both teams are 0.28 mag dimmer or 14% farther than expected in a Universe with this much matter and no cosmological constant. The statistical uncertainty of these values is 0.08 and 0.06 mag (or 4% and 3% in distance) for the HZT and SCP, respectively. The observed dispersion of the high-redshift SNe Ia around the best-fit cosmology is 0.21 mag for the HZT and 0.36 mag for the SCP. A frequentist would consider the accelerating Universe to be statistically likely at the 3$`\sigma `$ to 4$`\sigma `$ level (i.e., $`>`$ 99%), while the Bayseian likelihood would depend on a statement of the natural space and scale for $`\mathrm{\Omega }_\mathrm{\Lambda }`$.
Simply put, high-redshift SNe Ia are $``$0.25 mag fainter than expected in our Universe with its presumed mass density but without a cosmological constant (or which is not accelerating). The statistical confidence that SNe Ia are fainter than expected is high enough to accept that it does not result from chance and additional SNe Ia continue to support this conclusion (Schmidt 2000). Rather, this result is only challenged by systematic uncertainties not reflected in the variance of high-redshift SN Ia distance measurements. In the following sections we review the challenges to the cosmological interpretation of the SN Ia observations and consider whether the evidence compels us to believe that the Universe is accelerating.
## 3 Challenges and Tests of the Accelerating Universe
### 3.1 Evolution
Could SNe Ia at $`z=0.5`$, a look-back time of $``$ 5 Gyr, be intrinsically fainter than nearby SNe Ia by 25%? For the purpose of using SNe Ia as distance indicators near and far, we are concerned only with an evolution which changes the luminosity of a SN Ia for a fixed light-curve shape. Evolution is a major obstacle to the measurement of cosmological parameters, having plagued workers who tried to infer the global deceleration rate from brightest cluster galaxies in the 1970s (Sandage & Hardy 1973). We will consider both the theoretical and empirical indications for SN Ia evolution.
Theoretical understanding of SNe Ia provides reasons to believe that evolution is not a challenge to the accelerating Universe. SNe Ia are events which occur on stellar scales, not galaxian scales, and therefore should be less subject to the known evolution of stellar populations. However, our inability to conclusively identify the progenitor systems (see Livio 2000 for a review) and our lack of a complete theoretical model (see Leibundgut 2000 for a review) means we cannot rely exclusively on theory to rule out the critical degree of evolution.
Nevertheless, theoretical calculations can provide some insight into this question. Höflich, Wheeler, & Thielemann (1998) have calculated models of spectra of SNe Ia with solar and one-third solar metallicities and have found little difference between the spectral energy distribution over the wavelengths where the SNe Ia have been observed (see Figure 3). In principle, changes in the age and hence initial mass of the progenitor star at high redshift could yield white dwarfs of varying carbon-to-oxygen (C/O) ratio. It is currently difficult to assess if such a variation could produce significant evolution as these calculations lack the necessary precision (Dominguez et al. 1999; von Hippel, Bothun, & Schommer 1997). Umeda et al. (1999) suggest that a change in the C/O ratio is the source of the inhomogeneity in SN Ia luminosity, but they conclude that calibration of the luminosity via light-curve shape relations effectively inoculates the cosmological measurements to an evolution in the C/O ratio.
To date, answers to the question of whether SNe Ia evolve have been sought from empirical evidence. In the nearby sample, SNe Ia are observed in a wide range of host-galaxy morphologies including ellipticals, post-starburst galaxies (e.g., SN 1972E in NGC 5253), low surface brightness galaxies (e.g., SN 1995ak in IC 1844), irregulars (e.g., SN 1937C in IC 4182), S0s (e.g., SN 1995D in NGC 2962), and early to late-type spirals (van den Bergh 1994; Cappellaro et al. 1997; Hamuy et al. 1996; Riess et al. 1999). The range of metallicity, stellar age, and interstellar environments probed by the nearby hosts is much greater than the mean evolution in these properties for individual galaxies between $`z=0`$ and $`z=0.5`$. Some variation in the observed characteristics of SNe Ia with host morphology has been seen in the nearby sample (Hamuy et al. 1996; Branch, Romanishin, & Baron 1996) Yet after correction for the light-curve shape/luminosity relationship and extinction, the observed residuals from the Hubble flow do not correlate with host galaxy morphology (see Figure 4). In addition, Hubble flow residuals show no correlation with the projected distance from the host center (see §3.6). This evidence suggests that SN Ia distance estimates are insensitive to variations in the supernova progenitor environment and is the strongest argument against significant evolution to $`z=0.5`$ (Schmidt et al. 1998). However, this evidence is still circumstantial as we cannot be sure that the local environments of the SN Ia progenitors are similar to the average environments of the hosts. Future studies which probe the local regions of nearby SNe Ia should be able to explore their variance.
The other empirical test of evolution has been to compare the observed characteristics of low-redshift and high-redshift SNe Ia. The assumption of this test is that a luminosity evolution of $``$25% would be accompanied by other visibly altered characteristics of the explosion. Here too our lack of firm theoretical footing makes it difficult to gauge the correspondence between any evolution in distance-independent quantities and luminosity. Therefore we must conservatively demand that observations of all observables of distant SNe Ia be statistically consistent with the nearby sample.
#### 3.1.1 Spectra
Comparisons of high-quality spectra between nearby and high-redshift SN Ia, such as those seen in Figure 5, have revealed remarkable similarity (Riess et al. 1998; Perlmutter et al. 1998, 1999; Filippenko et al. 2000). The spectral energy distribution is sensitive to the atmospheric conditions of the supernova (i.e., temperature, abundances, and ejecta velocities). Even primitive modeling indicates that it would be difficult to retain the primary features of the SN Ia spectrum while altering the luminosity by about 20% to 30%. Further, comparisons of temporal sequences of spectra reveal no apparent differences as the photosphere recedes in mass (Filippenko et al. 2000), indicating that the superficial similarities persist at deeper layers. However, among the variety of nearby SNe Ia are objects which are both $``$25% fainter than the average and also display very typical spectral features (e.g., SN 1992A, see Figure 5). Therefore, the existing spectral test alone is not sufficient to check for this degree of evolution.
While spectral similarity between nearby and distant SNe Ia provides no indication of evolution, a lack of any spectral peculiarities among high-redshift SNe Ia could signal some changes at high redshift. Li et al. (2000) find from the most unbiased survey of nearby SNe Ia to date that $``$ 20% of SNe Ia are spectroscopically similar to the overluminous SN 1991T. SN 1991T-like objects show weak Ca II, Si II, and S II, but prominent features of Fe III (Filippenko 1997), and close cousins, such as SN 1999aa, have similar characteristics with the exception that Ca II absorption is more normal. Monte Carlo simulations of the search criteria used by the SCP and the HZT team performed by Li, Filippenko, & Riess (2000) indicate that such overluminous objects should comprise approximately 25% of high-redshift SNe Ia (with some uncertainty due to a possible link between such objects and circumstellar dust). To date, neither team has reported the existence of a single SN 1991T-like object among $``$ 100 high-redshift objects.
It is certain that the low signal-to-noise ratio of the spectra of high-redshift SNe Ia, coupled with the redshifting of spectral features out of the observable window makes it more difficult to identify individuals from this peculiar class. In addition, the spectroscopic peculiarities of SN 1991T-like objects are only apparent close to maximum light (or earlier), and some high-redshift SNe Ia may not have been observed early enough to identify their spectral peculiarities. This same effect may also explain why the Calán/Tololo survey of 29 SNe Ia yielded no SN 1991T-like objects (Hamuy et al. 1996). If, however, these observational biases are not to blame, the absence of SN 1991T-like SNe Ia at high redshift could result from an evolution of the population of progenitor systems (see Livio 2000 for a review) or a subtle difference in selection criteria (see §3.6). If true, this type of evolution may yield important clues which help identify the progenitor systems (Ruiz-Lapuente & Canal 1998), but it is unlikely to affect the measurement of the cosmological parameters since spectroscopically normal SNe Ia at low and high redshift have been used to derive the cosmological constraints.
#### 3.1.2 Broad-band
The distributions of light-curve shapes for nearby and distant SNe Ia are statistically consistent (Riess et al. 1998; Perlmutter et al. 1999). Such consistency appears to extend to infrared light curves of high-redshift SNe Ia which show the characteristic second maximum of typical, low-redshift SNe Ia (Riess et al. 2000).
An analysis by Drell et al. (2000) indicates that different light-curve fitting methods may not be statistically consistent and that the apparent differences may be a function of the light curve shape. However, these conclusions are highly sensitive to estimates of the correlated distance uncertainties between different fitting methods and these correlated uncertainties are difficult to estimate.
Evolutionary changes in the model temperature and hence the thermal output of the explosion could be detected from the colors of pre-nebular supernovae. The most significant analysis of $`BV`$ colors, performed by Perlmutter et al. (1999), demonstrated consistency between low and high-redshift SNe Ia at maximum light. Likewise, Riess et al. (2000) found that the $`BI`$ colors of a SN Ia at $`z=0.5`$ were consistent with those of nearby SNe Ia. However, Falco et al. 1999 (see also McLeod et al. 1999) suggested that the $`BV`$ colors of high-redshift SNe Ia from the HZT may be excessively blue, a conclusion which cannot be rejected by the $`BI`$ color measurements by Riess et al. (2000). More data are needed to confirm or refute this possibility. If true, this could indicate either evolution or the existence of a halo of Milky Way dust which would redden the observed wavelengths of nearby SNe Ia more than redshifted objects. This latter possibility has been suggested by recent Milky Way dust maps of Schlegel, Finkbeiner, & Davis (1999) in contrast to the previous maps of Burstein & Heiles (1982), but it would augment rather than fully explain the faintness of distant SNe Ia.
The risetime (i.e., the time interval between explosion and maximum light) is sensitive to the ejecta opacity and the distribution of <sup>56</sup>Ni . The risetime of nearby SNe Ia (Riess et al. 1999b) and the high-redshift SNe Ia of the SCP (Goldhaber 1998; Groom 1998) were initially strongly discrepant (Riess et al. 1999c). However, a reanalysis of the SCP high-redshift data by Aldering, Knop, & Nugent (2000) finds the high-redshift risetime to be longer and much more uncertain than indicated by Groom (1998). The remaining difference could be no more than a $``$2.0$`\sigma `$ chance occurrence. More early photometry of distant SNe Ia is needed to increase the significance of this test of evolution.
Evolution is arguably the most serious challenge to the cosmological interpretation of high-redshift SNe Ia. Further studies, currently underway, seek to compare the host galaxy morphologies and luminosity versus light-curve shape relations for nearby and distant SNe Ia. The results reviewed in this section do not appear to provide any clear evidence of evolution. However, absence of evidence is not necessarily evidence of absence. The paucity of high signal-to-noise ratio observations of high-redshift SNe Ia and the current lack of a comprehensive theoretical model or a well-understood progenitor system keeps the embers of skepticism aglow.
### 3.2 Dust
Consideration of a non-cosmological explanation for the dimming of distant supernova light must invariably turn to a famous pitfall of optical astronomy: extinction. Trouble has often followed dust in astronomy, a point first appreciated by Trumpler (1930) when analyzing the spatial distribution of Galactic stars.
#### 3.2.1 Ordinary Dust
An additional $``$25% opacity of visual light by dust in the light paths of distant supernovae would be sufficient to nullify the measurement of the accelerating Universe. Both teams currently measure SN Ia colors to correct for the ordinary kind of interstellar extinction which reddens light. Galactic extinction maps from Burstein & Heiles (1982) and Schlegel, Finkbeiner, & Davis (1998) were used by the HZT and the SCP, respectively, to correct individual SNe Ia for Milky Way extinction. Such corrections were typically less than 0.1 mag due to the high galactic latitudes of the SNe Ia. Even a previously unknown halo of Galactic dust would dim the restframe light of nearby SNe Ia more than highly redshifted SNe Ia and would therefore not explain the cosmological indications.
Measurements of $`BV`$ colors have been used by both teams to test for and remove host galaxy extinction (Riess et al. 1998; Perlmutter et al. 1999; see Figure 6). Totani & Kobayashi (1999) have suggested that the remaining uncertainty in the mean measured $`BV`$ color excess ($`\sigma =0.02`$ mag; Perlmutter et al. 1999), when multipled by reddening ratios of 3 to 4 to determine the optical opacity, may be too large to discriminate between open and lambda-dominated cosmologies with high confidence. However, such concern seems unwarranted as this uncertainty remains 3 to 4 times smaller than the size of the cosmological effect of an accelerating Universe.
Another measurable effect of the critical amount of mean interstellar extinction is that it would introduce more dispersion in the distance measurements than is currently observed. A random line of sight into a host galaxy will intersect a nonuniform amount of extinction. Hatano, Branch, & Deaton (1998) have calculated the expected distribution of extinction along random lines of sight into host galaxies. A mean, uncorrected extinction of 0.25 mag would induce twice the distance dispersion observed by the HZT (Riess et al. 1998). In addition, high-redshift surveys are biased towards finding SNe Ia which have even less extinction than would be expected from the distributions of Hatano et al. (1998).
A more powerful way to search for reddening by dust is to observe high-redshift SNe Ia over a large wavelength span: from the optical to the infrared. Infrared color excesses would be more than twice as large as $`E_{BV}`$ for ordinary dust. A set of such observations for SN 1999Q ($`z=0.46`$) disfavor $`A_V=0.25`$ mag of dust with Galactic-type reddening at high confidence (Riess et al. 2000), but more SNe Ia need to be observed in the near-infrared to strengthen this conclusion.
#### 3.2.2 Grey Dust
More pernicious than ordinary dust is “grey” dust which could leave little or no imprint on the spectral energy distribution of SNe Ia. Perfectly grey dust is only a theoretical construct, but dust which is greyer than Galactic-type dust (i.e., larger reddening ratios) does exist (Mathis 1990) and could challenge the cosmological interpretation of high-redshift SNe Ia.
Grey dust can be made with large spherical dust grains or elongated “whiskers.” Past studies of whiskers (Aguirre 1999a; Rana 1979, 1980) indicate that they would distort the cosmic microwave background (CMB), an effect which has not been seen. Like non-grey extinction, grey interstellar extinction does not provide an acceptable explanation for the dimness of SNe because the inherent variations in the opacity along random lines of sight would induce more distance dispersion than is observed (Riess et al. 1998).
Grey intergalactic extinction could affect measurements of the deceleration parameter (Eigenson 1949) without tell-tale dispersion or reddening. Indeed, observations of neither SNe Ia nor other astrophysical objects rule out a 30% opacity by large semi-spherical dust grains (Aguirre 1999b). Aguirre (1999a,b) has shown that a uniformly distributed component of intergalactic grey dust with a mass density of $`\mathrm{\Omega }_{dust}5\times 10^5`$ and graphite grains of size $`>0.1\mu `$m could explain the faintness of high-$`z`$ SNe Ia without detectable reddening and without overproducing the currently unresolved portion of the far-infrared (far-IR) background. However, this physical model of dust would provide some reddening which can readily be detected with observations in the optical and infrared. Measurements by Riess et al. (2000; see Figure 7) of $`E_{BI}`$ for a single high-redshift SN Ia disfavor a 30% visual opcaity of grey dust at the $`2.5\sigma `$ confidence level, but more observations are needed to strengthen this conclusion. Additional studies of the faint far-IR sources seen with SCUBA may soon provide definitive constraints on the unresolved component of the far-IR background and the viability of extragalactic grey dust.
### 3.3 Gravitational Lensing
The inhomogeneous distribution of matter in the Universe typically deamplifies and very rarely amplifies the observed brightness of distant SNe Ia compared to the average. (Note that the mean observed brightness must equal the unamplified value expected in a perfectly smooth Universe.)
The size of the typical deamplification is a function of the SN Ia redshift, the mass density of the Universe and the fraction of dark matter locked into compact objects. This effect has been quantified by a wide range of techniques (Kantowski, Vaughan, & Branch 1995; Frieman 1997; Wambsganss et al. 1997; Holz & Wald 1998; Kantowski 1998; Metcalf 1999; Barber 2000). The effect of weak lensing on the observed distribution of luminosities of SNe Ia at $`z=1`$ and $`z=0.5`$ can be seen in Figure 8. In the most relevant regime for the current SNe Ia at $`z=0.5`$ (i.e., $`\mathrm{\Omega }_M0.3`$, and mostly diffuse dark matter) the typical deamplification is $``$2%, much smaller than the cosmological effect. An extreme case (i.e., $`\mathrm{\Omega }_M0.5`$, all matter in point masses) could deamplify the median SN Ia at $`z=0.5`$ by 5% (Holz 1998), but this model is unlikely to be correct and the effect is still not large enough to negate the cosmological interpretation of high-redshift SNe Ia. Perlmutter et al. (1999) considered lensing by up to $`\mathrm{\Omega }_M=0.25`$ in compact material in the determination of their confidence intervals. They found little impact on the likelihood of a positive cosmological constant (see Figure 9). Wang (2000) suggests that by flux-averaging (i.e., binning the SNe Ia distances by redshift) one can reduce the bias due to weak lensing. In the future, any bias due to weak lensing will naturally vanish as the sample sizes become larger and the mean observed luminosity more robust. It is interesting and potentially useful to note that the observed distribution of SN Ia distances (see Figure 8) can in principle be used to determine the fraction of gravitating matter contained in compact objects (Seljak & Holz 1999; Metcalf & Silk 1999).
### 3.4 Measurement Biases
In this section we consider if biases in the measurement process of high-redshift SNe Ia could mimic the evidence for an accelerating Universe. An exhaustive list of such biases has been considered by Hogg (2000) and Hogg & Turner (1998). Here we discuss how these biases may apply to the supernova measurements.
The observational challenge is to measure the distance to high-redshift SNe Ia which are 6 to 7 magnitudes fainter and have lower signal-to-noise ratio than those which delineate the Hubble flow. Differences in the way low and high-redshift SNe Ia are observed must not introduce biases in their distance measures at more than the few percent level. Indeed, Hogg (2000) has noted that the proximity of high-redshift SNe Ia to any reasonable world model is a testament to the feasibility of measuring distances across such a large range. However, because the goal of these observations is precision cosmology (and not simply to demonstrate the dynamic range of useful photometry) our scrutiny must be greater.
Charge transfer inefficiency (CTI) and detector non-linearities can cause faint objects to appear fainter. However, ground-based observations of high-redshift SN Ia are limited by the bright sky, a regime in which these effects are widely found to be negligible. For space-based observatories such as the Hubble Space Telescope, CTI is far more troublesome but quite correctable (Whitmore, Heyer, & Casertano 1999). In addition, only a subset of the high-redshift SNe Ia have been measured with HST and the cosmological conclusions do not depend on the inclusion of these objects.
High precision, flux-conserving algorithms have been developed to properly subtract images of the host galaxy from images with SN light (Alard & Lupton 1998). Correctly employed, these methods reduce any biases in the measurement of the SN brightness to less than a few percent. Tests for measurement biases and estimates of uncertainty are performed by both teams by the injection of artificial SNe into the observed images.
Hogg & Turner (1998) discuss a bias towards higher observed fluxes which naturally occurs when measuring the brightness at discovery of low signal-to-noise ratio sources. This bias results from the preferential selection of faint sources on the bright side of the Poisson distribution of photon statistics. Follow-up observations of the source would not incur this bias. This effect would have little impact on the supernova distances measured by the HZT and SCP because the light curves are dominated by observations made after discovery. In addition, the direction of this effect is opposite to the signal of an accelerating Universe. However, this effect may become more important for SNe Ia found at $`z>1`$ for which the discovery observation may provide one of the most significant measurements.
### 3.5 Selection Biases
Do the HZT and SCP preferentially select faint SNe Ia at high redshift? Because we have already considered evolution in §3.1, here we are only concerned with the characteristics of a high-redshift sample which is drawn from the same population as the nearby sample. In so doing, we must also consider if the nearby sample is a fair representation of that population.
As an example, consider the set of nearby SNe Ia which appear fainter than expected for their redshift in the bottom panel of Figure 1. Presumably these objects appear dim due to the intrinsic random scatter of SNe Ia. If, however, these SNe Ia had a characteristic in common which, in addition, favored their discovery at high-redshift, a bias would result. To date, no such characteristic has been identified and the observed dispersion of nearby SNe Ia is consistent with their measurement errors.
Howell, Wang, & Wheeler (2000) found a difference between the projected distances from the hosts’ centers for the nearby and distant SNe Ia (see Figure 10). Many of the nearby SNe Ia were found in the photographic Calán/Tololo survey in which saturated galaxy cores masked SNe near their hosts’ centers (Shaw 1979; Hamuy & Pinto 1999). The result is that distant SNe Ia are more centrally located than the nearby sample. However, in an analysis of 44 nearby SNe Ia, Riess et al. (1999) found no dependence of the distance measurement on the projected distance from the host center, so this selection effect appears to have no bearing on the cosmological use of SNe Ia (see Figure 11).
Malmquist bias (Malmquist 1924; 1936) can shift the mean distance too close in a magnitude-limited survey of SNe Ia. This effect seems to contrast with the cosmological dimming perceived in an accelerating Universe. However, if the nearby sample were more afflicted by Malmquist bias than the distant sample, this bias could mimic an accelerating Universe. Because the intrinsic scatter of SN Ia distances is low ($`0.15`$ mag), Malmquist bias, which scales with the square of the dispersion (see Mihalas & Binney 1981 for a derivation), is small for SNe Ia. Perlmutter et al. (1999) made analytic calculations of Malmquist bias arising from the intrinsic dispersion of SNe Ia (assumed to be $`0.17`$ mag) and the SCP search incompleteness (determined empirically) to estimate that the net bias between the samples is no more than 0.03 mag. \[Perlmutter et al. (1999) notes that the net bias may actually be closer to zero due to a compensating bias against the selection of light curves which are “fast” for their luminosity and therefore spend less time above the detection limit.\] Riess et al. (1998) used a Monte Carlo exercise to simulate the selection of SNe Ia near and far. Inputs to this exercise included the time interval between successive search epochs, limiting magnitudes, observed light-curve shapes, and the distribution of SN Ia luminosities. They report a net bias of less than $`0.01`$ mag. These results indicate that the net Malmquist bias has negligible impact on the cosmological conclusions.
### 3.6 Alternative Cosmological Models
The conclusions drawn from high-redshift SNe Ia are predicated on a model with two free parameters, $`\mathrm{\Omega }_M`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$, and a Friedmann-Robertson-Walker cosmology. In the absence of a sound fundamental motivation for $`\mathrm{\Omega }_\mathrm{\Lambda }\mathrm{\Omega }_M`$, alternate and more general descriptions of an energy density with negative pressure have been suggested (Caldwell, Dave, & Steinhardt 1998). These phenomenological or “quintessence” models invoke a decaying scalar field rolling down a potential as the source of today’s acceleration (Wang et al. 2000). A distinction of these models from a cosmological constant is that $`w`$, the ratio of pressure to energy density, is between -1 and 0, whereas $`w`$ is exactly -1 for a cosmological constant. For feasible quintessence models, $`w`$, the equation-of-state parameter, varies slowly with time and can be approximated today by a constant equal to
$$\stackrel{~}{w}𝑑a\mathrm{\Omega }_Q(a)w(a)/𝑑a\mathrm{\Omega }_Q(a),$$
(5)
where $`a`$ is the scale factor and $`\mathrm{\Omega }_Q`$ is the energy density of the vacuum component. The current acceleration for these models (assuming only two significant energy components today, $`\mathrm{\Omega }_M`$ and $`\mathrm{\Omega }_Q`$) is
$$q_0=\ddot{a}(t_0)a(t_0)/\dot{a}^2(t_0)=\frac{1}{2}(\mathrm{\Omega }_M+\mathrm{\Omega }_Q(1+3w)),$$
(6)
and is generally less than for a cosmological constant (all other parameters fixed). Inspection of equation (2) reveals that the Universe is accelerating if $`q_0`$ is negative ($`\ddot{a}(t_0)`$ is positive), requiring that $`w<\frac{1}{3}`$, independent of the value of $`\mathrm{\Omega }_M`$.
Can we determine if the expansion is accelerating in a quintessence model? The SN Ia data from Perlmutter et al. (1999) and Riess et al. (1998) already provide meaningful constraints on $`w`$ (Garnavich et al. 1998b; Perlmutter et al. 1999). Increasing the value of $`w`$ from -1 (i.e., for a cosmological constant) reduces the acceleration provided by a fixed value of $`\mathrm{\Omega }_Q`$, but larger values of $`\mathrm{\Omega }_Q`$ are needed to retain an acceptable fit to the data. Graphically, increasing $`w`$ from -1 rotates the error ellipses in Figure 2 to favor lower values of $`\mathrm{\Omega }_M`$ and greater values of $`\mathrm{\Omega }_Q`$. As seen from equation (6), the line separating an accelerating and decelerating Universe rotates in the same direction (always anchored at $`\mathrm{\Omega }_M=0`$ and $`\mathrm{\Omega }_Q=0`$), providing no gain on an acceptable region of parameter space which is not accelerating. The nearest intersection between a non-accelerating region of parameter space and one which is preferred by the data remains when $`\mathrm{\Omega }_M<<1`$ and $`\mathrm{\Omega }_Q<<1`$. However, values near $`\mathrm{\Omega }_M=0`$ and $`\mathrm{\Omega }_Q=0`$ are poor fits to the data independent of the value of $`w`$.
Perhaps the simplest way to understand why the SN Ia data favor an accelerating Universe is to consider an FRW cosmology with $`\mathrm{\Omega }_M=0`$ and no vacuum energy. This empty Universe must be neither accelerating nor decelerating but simply coasting. The fact that the high-redshift SNe Ia are systematically farther for their redshift than expected in this cosmology means that the distance between low and high-redshift SNe Ia (where redshift is a surrogate for time) grew faster than expected for a Universe which has been coasting on today’s Hubble expansion. This implies that the Universe has been accelerating.
An alternate cosmological explanation to acceleration has been posited by Goodwin et al. (1999) and Tomita (2000). They suggest that the supernova data are also consistent with a decrease in the Hubble expansion by 10-20% beyond $`z=0.1`$ ($`300h^1`$ Mpc). The distance at which the Hubble expansion dips would correspond to the approximate radius of the “local” underdensity in which we live. Although a few peculiar flow surveys support bulk motions on scales up to half this size (Lauer & Postman 1994; Hudson et al. 1999), most recent surveys do not (Dale et al. 1999; Courteau et al. 2000; Colless et al. 1999; Riess et al. 1999; see Willick 2000 for a review). However, the biggest problem with such a commodious, local underdensity is its great improbability. Power spectra demonstrate (Watkins & Feldman 1995; Feldman & Watkins 1998) that the density of the Universe is extremely homogeneous on this scale, and finding ourselves in the midst of such a vacuous location would be virtually anti-Copernican. Using cold dark matter power spectra constrained by CMB observations and large scale structure, Shi and Turner (1998) and Wang, Spergel, and Turner (1998) expect 0.5% to 1.5% variations in the Hubble constant on $`300h^1`$ Mpc scales, a factor of 20 times smaller than required in the local void model. By filling in the Hubble diagram of SNe Ia at $`0.1<z<0.2`$ it would be possible to test this model.
Outside the FRW cosmologies the SN Ia data can have significantly different interpretations. For example, in steady-state cosmologies, SN redshifts do not come from expansion, but rather through “tired-light” processes. However, the SN Ia data exhibit the time dilation effect expected in an expanding Universe, implying that the tired-light hypothesis is incorrect (Leibundgut et al. 1996; Goldhaber et al. 1996; Riess et al. 1997; but see Narlikar & Arp 1997). In the quasi-steady state cosmology, the SN Ia data lead to modifications of the model, such as matter creation during periodic expansion phases (Hoyle, Burbidge, & Narlikar 2000). A detailed consideration of how to interpret the SN Ia data in non-FRW cosmologies is beyond the scope of this review, but is thoroughly addressed by Hoyle, Burbidge, & Narlikar (2000). Alternative theories of gravity such as modified Newtonian dynamics models (MOND; Milgrom 1983, 1998; McGaugh & de Blok 1998a,b) could also modify the interpretation of the observations of high-redshift SNe Ia.
## 4 Conclusion
After reviewing the cosmological interpretation of SN Ia observations and the current challenges to the analysis of the data, we can now offer an answer to the question initially posed: do the observations of distant supernovae compel us to conclude that the expansion of the Universe is accelerating?
With full consideration of the evidence, we conclude that an accelerating Universe remains the most likely interpretation of the data because the alternatives, individually, appear less likely. However, the quantity and quality of the SN Ia evidence alone is not yet sufficient to compel belief in an accelerating Universe. The primary sources of reasonable doubt are evolution and extinction, as discussed above. Although the types of studies also described above could potentially yield evidence that either of these non-cosmological contaminants is significant, the current absence of such evidence does not suffice as definitive evidence of their absence. Our current inability to identify the progenitors of SNe Ia and to formulate a self-consistent model of their explosions exacerbates such doubts. Even optimists would acknowledge that neither of these theoretical challenges is likely to be met in the near future.
Fortunately there are at least two routes to obtain compelling evidence to accept (or refute) the accelerating Universe, one of which employs the use of SNe Ia at even greater redshifts.
## 5 Epilogue
### 5.1 The Era of Deceleration
If the Universe is accelerating, it is a rather recent phenomenon likely commencing between $`z0.4`$ and 1. Before this time the Universe was more compact and the pull of matter dominated the push of vacuum energy in the equation of motion. As a result, the Universe at $`z1`$ must be decelerating. This cosmological signature should be readily apparent by extending the Hubble diagram of SNe Ia to $`z1.2`$. By this redshift SNe Ia in an accelerating Universe will cease to diverge in distance from an equally massive Universe without vacuum energy. Alternatively, if a monotonically increasing, systematic effect is the source of the excessive faintness of high-redshift SNe Ia, the measured distances of SNe Ia at $`z1`$ will continue to diverge from a cosmology without vacuum energy and in addition would diverge from the cosmological model inferred from SNe Ia at $`z=0.5`$ (see Figure 12). Complex parameterizations of evolution or extinction selected to match both the accelerating and decelerating epochs of expansion would require a near conspiracy of fine tuning and are highly doubtful.
Efforts are already underway to find and measure SNe Ia at $`z>1`$. Gilliland, Nugent, & Phillips (1999) used a subsequent epoch of the Hubble Deep Field to detect two SNe, one (SN 1997ff) with a photometric redshift of $`z=1.32`$. The elliptical host of SN 1997ff suggests that this object is of Type Ia, but the observations are insufficient to provide a useful distance estimate. The SCP reported the discovery of SN 1998ef at $`z=1.20`$ (Aldering et al. 1998) and follow-up observations with the HST will provide a useful distance estimate (Aldering 2000). The HZT recently reported the discovery of four SNe Ia at $`z>1`$ including SN 1999fv at $`z=1.23`$ (Tonry et al. 1999). From this growing sample will likely come the means to search for the epoch of deceleration.
### 5.2 Cosmic Complements
We previously sought to determine if the observations of SNe Ia alone require an accelerating Universe. Now we will briefly consider the cosmological constraints provided by other astrophysical phenomena. A thorough discussion of these constraints is beyond the scope of this review but can be found elsewhere (Tyson & Turner 1999; Roos & Harun-or-Rashid 2000)
Current measurements of the cosmic microwave background (CMB) power spectrum indicate that the sum total of energy densities is within 10% of unity. This result is seen from the BOOMERANG (Melchiorri et al. 1999; Lange et al. 2000; de Bernardis et al. 2000), the TACO (Miller et al. 1999) and MAXIMA experiments (Hanany et al. 2000; Balbi et al. 2000) and from a compilation of all other CMB measurements (Tegmark & Zaldarriaga 2000). In addition, estimates of $`\mathrm{\Omega }_M`$ from the mass, light, X-ray emission, numbers, and motions of clusters of galaxies converge around 0.2 to 0.3 (Carlberg et al. 1996; Bahcall, Fan, & Cen 1997; Lin et al. 1996; Strauss & Willick 1995). These two pieces of information alone indicate a significant contribution by vacuum energy, sufficient to produce an accelerating Universe (see Figure 13). Additional constraints from observations of the Lyman-alpha forest, cluster evolution, double radio galaxies, and statistics of gravitational lenses have been used to tighten these conclusions (Roos & Harun-or-Rashid 2000; Turner 1999; Eisenstein, Hu, & Tegmark 1999; Lineweaver 1998).
Although no single cosmological observation yields a conclusive census of the energy densities in the Universe, the combined constraints from multiple experiments is providing strong bounds on the cosmological parameters. Each individual experiment has unique sources of systematic uncertainty. By combining the results of many experiments it should be possible to negate their impact on the determination of the cosmological parameters.
I wish to thank Mario Livio, Alex Filippenko, and Robert Kirshner for helpful discussions and comments.
References
Aguirre, A. 1999a, ApJ, 512, 19
Aguirre, A. 1999b, ApJ, 525, 583
Alard, C. & Lupton, R. H., 1998, ApJ, 503, 325
Aldering, G., Knop, R., & Nugent, P. E., 2000, accepted ApJ (astro-ph/0001049)
Aldering, G., et al., 1998, IAU Circ. 7046
Aldering, G., et al., 2000, private communication
Bahcall, J. N. et al. 1996, ApJ, 457,19
Bahcall, N. A., Fan, X., & Cen, R., ApJ, 1997, 485, 53
Balbi, A., et al., 2000, astro-ph/0005124
Barber, A. J., 2000, astro-ph/0003304
Branch, D., ARAA, 1998, 36, 17
Branch, D., Romanishan, W., & Baron, E., 1996, ApJ, 465, 73
Burstein, D., & Heiles, C. 1982, AJ, 87, 1165
Cappellaro, E., Turatto, M., Tsvetkov, D. Yu., et al., 1997, A&A, 322,431
Carlberg, R. G., Yee, H. K. C., Ellingson, E., Abraham, R., Gravel, P., Morris, S. & Pritchet, C. J., 1996, ApJ, 462, 32
Caldwell, R. R., Dave, R., & Steinhardt, P. J., 1998, Ap&SS, 261, 303
Carroll, S. M., Press, W. H., & Turner, E. L., 1992, ARAA, 30, 499
Colgate, S. 1979, ApJ, 232, 404
Colless, M., Saglia, R. P., Burstein, D., Davies, R., McMahan, R. K., & Wegner, G., 1999, in Cosmic Flows Workshop astro-ph/9909062
Courteau, S., Willick, J. A., Strauss, M. A., Schlegel, D., & Postman, M., 2000, ApJ, submitted (astro-ph/0002420)
Dale, D. A., Giovanelli, R., Haynes, M. P., Campusano, L. E., & Hardy, E., 1999, AJ, 118, 1489
de Bernardis, P., et al., 2000, Nature, 404, 955
Dominguez, I., Höflich, P., Straniero, O., & Wheeler, C., 1999, in Future Directions of Supernovae Research: Progenitors to Remnants (Assergi, Italy)
Drell, P. S., Loredo, T. J., & Wasserman, I., 2000, ApJ, 530, 593
Eigenson, M. 1949, Astron. Zh., 26, 278
Eisenstein, D., Hu, W., & Tegmark, M., 1999, ApJ, 518, 2
Falco, E. et al. 1999, ApJ, 523, 617
Feldman, H. A., & Watkins, R., 1998, ApJ, 494, 129
Filippenko, A. V. et al. 2000, in preparation
Filippenko, A.V. 1997, ARA&A, 35, 309
Frieman, J. A., 1996, astro-ph/9608068
Garnavich, P., et al. 1998a, ApJ, 493, 53
Garnavich, P., et al. 1998b, ApJ, 509, 74
Gilliland, R. L., Nugent, P. E., & Phillips, M. M., 1999, ApJ, 521, 30
Goldhaber, G. 1998, B.A.A.S., 193, 4713
Goldhaber, G., et al. 1997, in Thermonuclear Supernovae, eds. P. Ruiz-Lapuente, R. Canal, & J. Isern (Dordrecht: Kluwer), p. 777
Goobar, A. & Perlmutter, S., 1995, ApJ, 450, 14
Goodwin, S. P., Thomas, P. A., Barber, A. J., Gribbin, J., & Onuora, L. I., 1999, astro-ph/9906187
Groom, D. E. 1998, B.A.A.S., 193, 11102
Hamuy, M., Phillips, M. M., Maza, J. Suntzeff, N. B., Schommer, R.A. & Avilés, R. 1995, AJ, 109, 1
Hamuy, M., et al. 1996a, AJ, 112, 2408
Hamuy, M., et al. 1996b, AJ, 112, 2438
Hamuy, M. & Pinto, P. A., 1999, AJ, 117, 1185
Hanany, S., et al. 2000, astro-ph/0005123
Hatano, K., Branch, D., & Deaton, J., 1998, ApJ, 502, 177
Höflich, P., Wheeler, J. C., & Thielemann, F. K. 1998, ApJ, 495, 617
Hogg, D., & Turner, E. L.,1998, PASP, 110, 727
Hogg, D., 2000, astro-ph/0001419
Holz, D. E., 1998, ApJ, 506, 1
Holz, D. E. & Wald, R. M., 1998, Phys.Rev., D58, 063501
Howell, D. A., Wang, L., & Wheeler, J. C., 2000, ApJ, 530, 166
Hoyle, F., Burbidge, G., & Narlikar, J. V., 2000, in A Different Approach to Cosmology, Cambridge University Press (Cambridge: England)
Hudson, M. J., Smith, R. J., Lucey, J. R., Schlegel, D. J., & Davies, R. L., 1999, ApJ, 512, 79
Jha, S., et al. 1999, ApJS, 125, 73
Kantowski, R., Vaughan, T., & Branch, D. 1995, ApJ, 447, 35
Kantowski, R., 1998, ApJ, 507, 483
Kim, A., Goobar, A., & Perlmutter, S. 1996, PASP, 108, 190
Kirshner, R. P., & Kwan, J., 1974, ApJ, 193, 27
Landolt, A. U. 1992, AJ, 104, 340
Lange, A. E., et al., 2000, astro-ph/0005004
Lauer, T., & Postman, M. 1994, Ap.J. 425, 418
Leibundgut, B., 2000, A&A, in press
Leibundgut, B., 1996, ApJ, 466, 21
Li, W., Filippenko, A. V., Riess, A. G., Treffers, R. R., Hu., J., & Qiu, Y., 2000, ApJ, submitted
Li, W., Filippenko, A. V., & Riess, A. G., 2000, ApJ, submitted
Lin, H. et al. 1996, ApJ, 471, 617
Lineweaver, C. H., 1998, ApJ, 505, 69
Livio, M., 2000, in Type Ia Supernovae: Theory and Cosmology (Cambridge University Press)
Malmquist, K. G., Medd. Lund Astron. Obs. Ser. II, 32, 64
Malmquist, K. G., 1936, Stockholm Obs. Medd., no. 26
Mathis, J. S., 1990, ARA&A, 28, 37
McGaugh, S. & de Blok, W. J. G., 1998a, ApJ, 499, 66
McGaugh, S. & de Blok, W. J. G., 1998b, ApJ, 499, 41
McLeod, B. A., et al. 1999, in Gravitational Lensing: Recent Progress and Future Goals, ed. Brainerd T. & Kochanek C. S, ASP (astro-ph/9910065)
Melchiorri, A., et al., 1999, astro-ph/9911445
Metcalf, R. B., 1999, MNRAS, 305, 746
Metcalf, R. B. & Silk, J., 1999, ApJ, 519, 1
Mihalas, D., & Binney, J. 1981, Galactic Astronomy: Structure and Kinematics (2d ed.; San Francisco: W. H. Freeman)
Milgrom, M., 1983, ApJ, 270, 371
Milgrom, M., 1988, ApJ, 333, 689
Miller, A. D., et al., 1999, ApJ, 524, 1
Narlikar, J. V. & Arp, H. C., 1997, ApJ, 482, 119
Norgaard-Nielsen, H., et al. 1989, Nature, 339, 523
Parodi, B. R., Saha, A., Sandage, A., & Tammann, G. A., 2000, astro-ph/0004063
Perlmutter, S. et al., 1995, ApJ, 440, 41
Perlmutter, S. et al., 1997, ApJ, 483, 565
Perlmutter, S., et al. 1998, Nature, 391, 51
Perlmutter, S., et al. 1999, ApJ, 517, 565
Phillips, M. M. 1993, ApJ, L105, 413
Phillips, M. M., Lira, P., Suntzeff, N. B., Schommer, R. A., Hamuy, M., Maza, J., 1999, AJ, 118, 1766
Rana, N. C., 1980, Ap&SS, 71, 123
Rana, N. C., 1979, Ap&SS, 66, 173
Riess, A. G., Press W.H., Kirshner, R.P., 1995, ApJ, 438 L17
Riess, A. G., Press, W.H., & Kirshner, R.P. 1996, ApJ, 473, 88
Riess, A. G., et al. 1997, AJ, 114, 722
Riess, A. G., et al. 1998, AJ, 116, 1009
Riess, A. G., et al. 1999b, AJ, 118, 2675
Riess, A. G., et al. 1999a, AJ, 117, 707
Riess, A. G., Filippenko, A. V., Li, W., & Schmidt, B. P. 1999c, AJ, 118, 2668
Riess, A. G., 1999, in Cosmic Flows Workshop, Victoria, Canada, July 1999, ed. S. Courteau, M. Strauss & J. Willick, ASP, astro-ph/9908237
Riess, A. G., et al. 2000, ApJ, accepted (astro-ph/0001384)
Roos, M., & Harun-or-Rashid, S. M., 2000, astro-ph/0003040
Ruiz-Lapuente, P., & Canal, R., 1998, ApJ, 497, 57
Saha, A., et al. 1999, ApJ, 522, 802
Sandage, A., & Hardy, E., 1973, ApJ, 183, 743
Schlegel, D. J., Finkbeiner, D. P., & Davis, M. 1998, ApJ, 500, 525
Schmidt, B. P., et al. 1998, ApJ, 507, 46
Schmidt, B. P., 2000, private communication
Seljak, U., & Holz, D., 1999, A&A, 351, 10
Shaw, R. L., 1979, A&A, 76, 188
Shi, X. & Turner, M. S., 1998, 493, 519
Strauss, M.A., & Willick, J.A., 1995, PhR, 1995, 261, 271
Tammann, G. A. 1979, in ESA/ESO Workshop on Astronomical Uses of the Space Telescope, ed. F. Macchetto, F. Pacini, and M. Tarenghi (Geneva:ESO), 329
Tegmark, M., & Zaldarriaga, M., 2000, astro-ph/0002091
Tonry, J., et al., 1999, IAU Circ. 7312
Tomita, K., 2000, astro-ph/ 0005031
Totani, T., & Kobahashi, C., ApJ, 526, 65
Tripp, R. 1997, A&A, 325, 871
Tripp, R., & Branch, D., 1999, ApJ, 525, 209
Trumpler, R., 1930, PASP, 42, 214
Turner, M. S., & Tyson, J. A., 1999, Rev. Mod. Phys., 71, 145
Turner, M. S., 1999, PASP, 111, 264
Umeda, H. et al. 1999, ApJ, 522, 43
van den Bergh, S., 1994, ApJS, 92, 219
von Hippel, T., Bothun, G. D., & Schommer, R. A., 1997, AJ, 114, 1154
Wagoner, R. V., 1977, ApJ, 214, L5
Wambsganss, J., Cen, R., Guohong, X., & Ostriker, J. 1997, ApJ, 475, L81
Wang, Y., 2000, accepted to ApJ, astro-ph/9907405
Wang, Y., Spergel, D. N., & Turner, E. L., 1998, 498, 1
Watkins, R. & Feldman, H. A., 1995, ApJ, 453, 73
Whitmore, B., Heyer, I., & Casertano, S., 1999, PASP, 111, 1559
Willick, J. A., 2000, in Proceedings of the XXXVth Rencontres de Moriond: Energy Densities in the Universe astro-ph/0003232
Figure Captions:
Figure 1: Hubble diagrams of SNe Ia from Perlmutter et al. (1999; SCP) and Riess et al. (1998; HZT). Overplotted are three cosmologies: “low” and “high” $`\mathrm{\Omega }_M`$ with $`\mathrm{\Omega }_\mathrm{\Lambda }=0`$ and the best fit for a flat cosmology, $`\mathrm{\Omega }_M=0.3`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.7`$. The bottom panel shows the difference between data and models from the $`\mathrm{\Omega }_M=0.3`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0`$ prediction. The average difference between the data and the $`\mathrm{\Omega }_M=0.3`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0`$ prediction is 0.28 mag.
Figure 2: SNe Ia joint confidence intervals for ($`\mathrm{\Omega }_M`$,$`\mathrm{\Omega }_\mathrm{\Lambda }`$) from Perlmutter et al. (1999; SCP) and Riess et al. (1998; HZT). Regions representing specific cosmological scenarios are indicated.
Figure 3: Type Ia supernova model spectral energy distributions for solar and $`\frac{1}{3}`$ metallicities. Superimposed are the transmission functions for standard passbands; from left to right is $`U,B,V`$, and $`R`$. From Höflich, Wheeler & Thielemann (1998)
Figure 4: The nearby Hubble diagram of SNe Ia in different host galaxy morphologies. In the top panel the SNe Ia are treated as standard candles and in the bottom panel distances are determined with the MLCS method (Riess, Press, & Kirshner 1996; Riess et al. 1998). After MLCS corrections are made, the distance Hubble flow residuals are independent of host galaxy morphology.
Figure 5: Spectra of 4 nearby and 1 high-redshift SN Ia at the same phase (Riess et al. 1998). Within the observed variance of SN Ia spectral features, the spectra of high-redshift SNe Ia are indistinguishable from the low-redshift SNe Ia. The spectra of the low-redshift SNe Ia were resampled and convolved with Gaussian noise to match the quality of the spectrum of SN 1998ai.
Figure 6: Comparison of color excess, $`E_{BV}`$, for nearby SNe Ia from the Calan/Tololo survey (Hamuy et al. 1996a,b) and high-redshift SNe Ia from Perlmutter et al. (SCP; 1999). The color excesses at high and low redshift are consistent indicating negligible extinction and/or no evidence for color evolution.
Figure 7: The color evolution, $`BI`$, and the color excess, $`E_{BI}`$, of a h igh-redshift SN Ia, SN 1999Q ($`Z=0.46`$), compared to the custom MLCS template curve with no dust and enough dust (of either Galactic-type or greyer) to nullify the cosmological constant. The smaller error bars are from photometry noise; the larger error bars include all sources of uncertainty such as intrinsic dispersion of SN Ia $`BI`$ color, $`K`$-corrections, and photometry zeropoints. The data for SN 1999Q are consistent with no reddening by dust, moderately inconsistent with $`A_V`$=0.3 mag of grey dust (i.e., graphite dust with minimum size $`>0.1\mu `$m; Aguirre 1999a,b) and $`A_V`$=0.3 mag of Galactic-type dust. From Riess et al. (2000).
Figure 8: Probability distribution $`P(\mu )`$ for supernova apparent brightness $`\mu `$ normalized to $`\mu =1`$ for a filled beam (i.e., a homogeneous Universe). The vertical lines are at the empty-beam value. “Galaxies” are treated as isothermal spheres and truncated at a radius of 380 kpc; “compact objects” are point masses. From Holz (1998.)
Figure 9: Cosmological constraints from Perlmutter et al. (SCP; 1999) for three weak lensing scenarios. Fit C assumes a filled beam, case K assumes an empty beam, and fit L is a model with weak lensing by up to $`\mathrm{\Omega }_M=0.25`$ in compact objects.
Figure 10: The projected distances from the host centers of nearby SNe Ia discovered photographically and with CCDs and high-redshift SNe Ia discovered with CCDs (Howell, Wang, & Wheeler 2000).
Figure 11: MLCS distance residuals from the Hubble flow for nearby SNe Ia versus their projected distance from their host centers (Riess et al. 1999a).
Figure 12: The Hubble diagram of SNe Ia (see Figure 1) and the effect of a systematic error which grows linearly with redshift (e.g., evolution or grey extinction).
Figure 13: Cosmological constraints from SNe Ia, CMB, and matter (Turner 1999).
|
warning/0005/astro-ph0005558.html
|
ar5iv
|
text
|
# Large-Scale Power Spectrum and Structures From the ENEAR galaxy Peculiar Velocity Catalog
## 1 INTRODUCTION
The canonical model of cosmology assumes that large-scale structure has grown out of small density perturbations via the process of gravitational instability. These initial fluctuations are usually assumed to satisfy the statistics of a Gaussian random field, solely characterized by its power spectrum. In the linear regime, the fluctuations grow self-similarly and retain their initial distribution and power spectrum shape. Therefore, mapping the underlying cosmological velocity field and its power spectrum on large scales, provides a direct probe to the origin of structure in the universe.
The PS, the three-dimensional distribution of luminous matter and the predicted peculiar velocity field have been derived from a variety of data sets, especially from all-sky redshift surveys (e.g. Strauss & Willick 1995 for a review of earlier work; Sutherland et al. 1999; Branchini et al. 1999). Unfortunately, however, the distribution of galaxies in these catalogs is not necessarily an unbiased tracer of the underlying mass distribution, and suffer from the infamous “galaxy biasing” problem. Furthermore, in estimates from redshift surveys, uncertainties arise from the complicated relation between the real space and the redshift space distributions, known as redshift distortions (e.g., Kaiser 1987, Zaroubi and Hoffman 1996). In order to avoid these problems altogether it is advantageous to appeal to dynamical data, in particular catalogs of galaxy peculiar velocities on large scales.
Peculiar velocities enable a direct and reliable determination of the mass PS and distribution, under the natural assumption that the galaxies are unbiased tracers of the large-scale, gravitationally-induced, velocity field. Furthermore, since peculiar velocities are non-local and have contributions from different scales, analysis of the peculiar velocity field provides information on scales somewhat larger than the sampled region (e.g. Hoffman et al. 2000). For the same reason peculiar velocities are adequately described by linear theory even when densities become quasi-linear (e.g., Freudling et al. 1999). Consequently, the dynamics and the distribution of peculiar velocities are well described by the linear regime of gravitational instability and by a Gaussian probability distribution function (PDF), respectively.
Assuming that both the underlying velocity field and the errors are drawn from independent random Gaussian fields, the observed peculiar velocities constitute a multi-variant Gaussian data set, albeit the sparse and inhomogeneous sampling. The corresponding posterior PDF is a multivariate Gaussian that is completely determined by the assumed PS and covariance matrix of errors. Under these conditions one can write the joint PDF of the model PS and the underlying velocity or density field.
The purpose of the present study is to calculate, from the joint PDF, the PS and 3D mass distribution, as well as the 3D peculiar velocity field, as derived from the newly completed ENEAR galaxy peculiar velocity catalog (da Costa et al. 2000a, Paper I). First, the PS model parameters are estimated by maximizing the likelihood function given the model (Zaroubi et al. 1997). An identical likelihood estimation of the power spectrum has been previously applied to the Mark III (Zaroubi et al. 1997) and the SFI (Freudling et al. 1999) data sets. In both cases the analysis yielded a high amplitude power spectrum. Although the results from those two catalogs are consistent with each other, they are marginally inconsistent with the power spectra measured from redshift catalogs (e.g., da Costa et al. 1996; Sutherland et al. 1999), inferred from the analysis of the velocity correlation function (e.g. Borgani et al. 2000a, 2000b), and from velocity-velocity comparisons (e.g., Davis et al. 1996, da Costa et al. 1998). One of our goals is to use the same methodology employed before for the Mark III and SFI to the new ENEAR catalog to directly test the reproducibility of the results with an independent sample based on a different distance indicator but probing a comparable volume.
Second, the Wiener filter (WF) solution of the field is recovered by finding the most probable field given the PS and the data (Zaroubi et al. 1995, 1999). Constrained realizations (CR) are then used to sample the statistical scatter around the WF field (Hoffman and Ribak 1991). The mass density PS is used to calculate the smoothed Wiener filtered density and 3D velocity fields given the measured radial velocities (Zaroubi et al. 1995, 1999). The WF provides an optimal estimator of the underlying field in the sense of a minimum-variance solution given the data and an assumed prior model (Wiener 1949; Press et al. 1992). The prior defines the data auto-correlation and the data-field cross-correlation matrices. In the case where the data is drawn from a random Gaussian field, the WF estimator coincides with the conditional mean field and with the most probable configuration given the data (see Zaroubi et al. 1995). It should be noted that Kaiser & Stebbins (1991) were the first to propose a Bayesian solution to the problem of reconstruction from peculiar velocity data sets.
Finally, the recovered three-dimensional velocity field is used to compute the amplitude of the bulk flow and to decompose the velocity field in terms of a divergent and tidal components which enables one to separate the contribution to the measured peculiar velocity field from mass fluctuations within and outside the volume probed by the data (Hoffman et al. 2000).
The methods adopted in this study do not involve any explicit window function, weighting or smoothing the data. In addition, they automatically underweight noisy, unreliable data. However, a few simplifying assumptions are required: 1) peculiar velocities are drawn from a Gaussian random field; 2) peculiar velocities are related to the densities through linear theory; 3) errors in the $`D_n\sigma `$ inferred distances constitute a Gaussian random field with two components, the first scales linearly with distance while the second models the nonlinear evolution of the velocities as a constant scatter. The need to assume a parametric functional form for the PS is also a limitation.
The outline of this paper is as follows. In § 2 we briefly describe the peculiar velocity data used in the present analysis. The PS analysis is carried out in § 3. The Wiener filtering is applied to the ENEAR data in § 4, where maps of the density field are presented and compared to those predicted from redshift surveys. Also shown in this section are the recovered three-dimensional velocity field and the results of its analysis. Our results are summarized and discussed in § 5.
## 2 The Data
In the present analysis, we use the ENEAR redshift-distance survey described in greater detail in Paper I of this series. Briefly, the ENEAR sample consists of roughly 1600 early-type galaxies brighter than $`m_B=14.5`$ and with $`cz7000\mathrm{km}\mathrm{s}^1`$, for which $`D_n\sigma `$distances are available for 1359 galaxies. Of these 1145 were deemed suitable for peculiar velocity analysis according to well-defined criteria (Paper I; M. Alonso et al. , in preparation). To the magnitude-limited sample we added 285 fainter and/or with redshifts $`>7000\mathrm{km}\mathrm{s}^1`$, 129 within the same volume as the magnitude-limited sample. In total, the cluster sample consists of 569 galaxies in 28 clusters, which are used to derive the distance relation (M. Bernardi et al. , in preparation). Over 80% of the galaxies in the magnitude-limited sample and roughly 60% of the cluster galaxies have new spectroscopic and R-band photometric data obtained as part of this program. Furthermore, repeated observations of several galaxies in the sample (M. Alonso et al. , in preparation; G. Wegner et al. 2000) provide overlaps between observations conducted with different telescope/instrument configurations and with data available from other authors. These overlaps are used to tie all measurements into a common system, thereby ensuring the homogeneity of the entire dataset. In contrast to other samples new observations conducted by the same group are available over the entire sky. The comparison between the sample of galaxies with distances and the parent catalog also shows that the sampling across the sky is uniform.
Individual galaxy distances were estimated from a direct $`D_n\sigma `$ template relation derived by combining all the available cluster data (M. Bernardi et al. , in preparation), corrected for incompleteness and associated diameter-bias (Lynden-Bell et al. 1988). From the observed scatter of the template relation the estimated fractional error in the inferred distance of a galaxy is $`\mathrm{\Delta }0.19`$, nearly independent of the velocity dispersion.
Since early-type galaxies are found preferentially in high-density regions, galaxies have been assigned to groups/clusters using well-defined criteria imposed on their projected separation and velocity difference relative to the center of groups and clusters. These systems were identified using objective algorithms applied to the available magnitude-limited samples, comprising all morphological types, with complete redshift information probing the same volume. For membership assignment we used group catalogs published by Geller & Huchra (1983), Maia, da Costa & Latham (1988), Ramella, Pisani & Geller (1997) as well as unpublished results (M. Ramella et al. , in preparation) covering other regions of the sky. The characteristic size and velocity dispersion of these groups/clusters were used to establish the membership of the ENEAR early-types, as described in Paper I. We find isolated galaxies, groups with only one early-type, and groups with two or more early-types. Early-type galaxies in a group/cluster are replaced by a single object having: (1) the redshift given by the group’s mean redshift, which is determined considering all morphologies; (2) the distance given by the error-weighted mean of the inferred distances, for groups with two or more early-types; and (3) the fractional distance error given by $`\mathrm{\Delta }/\sqrt{(}N)`$, where $`N`$ is the number of early-types in the group. In some cases groups were identified with Abell/ACO clusters within the same volume as the ENEAR sample and fainter cluster galaxies were added, as described in Paper I. In the analysis below we compute the dipole component of the velocity field out to $`6000\mathrm{km}\mathrm{s}^1`$ as probed by all objects, and by splitting the sample into two independent sub-samples consisting of field galaxies and groups/clusters. The latter is done to evaluate directly from data the amplitude of possible sampling errors.
The inferred distances are corrected for the homogeneous and inhomogeneous Malmquist bias (IMB). The latter was estimated using the PCSz density field (Branchini et al. 1999), corrected for the effects of peculiar velocities, in the expressions given by Willick et al. (1997). In this caculation we also include the correction for the redshift limit of the sample. A complete account of the sample used and the corrections applied will be presented in a subsequent paper of this series (M. Alonso et al. , in preparation).
## 3 Power Spectrum
The calculation of the matter PS from the peculiar velocity data by means of likelihood analysis requires a relation between the velocity correlation function and the power spectrum. Define the two-point velocity correlation ($`3\times 3`$) tensor by the average over all pairs of points $`𝐫_i`$ and $`𝐫_j`$ that are separated by $`𝐫=𝐫_j𝐫_i`$,
$$\mathrm{\Psi }_{\mu \nu }(𝐫)v_\mu (𝐫_i)v_\nu (𝐫_j),$$
(1)
where $`v_\mu (𝐫_i)`$ is the $`\mu `$ component of the peculiar velocity at $`𝐫_i`$. In linear theory, it can be expressed in terms of two scalar functions of $`r=|𝐫|`$ (Górski 1988), computed from the parallel and perpendicular components of the peculiar velocity, relative to the separation vector $`𝐫`$,
$$\mathrm{\Psi }_{\mu \nu }(𝐫)=\mathrm{\Psi }_{}(r)\delta _{\mu \nu }+[\mathrm{\Psi }_{}(r)\mathrm{\Psi }_{}(r)]\widehat{𝐫}_\mu \widehat{𝐫}_\nu .$$
(2)
The spectral representation of these radial correlation functions is
$$\mathrm{\Psi }_,(r)=\frac{H_0^2f^2(\mathrm{\Omega })}{2\pi ^2}_0^{\mathrm{}}P(k)K_,(kr)𝑑k,$$
(3)
where $`K_{}(x)=j_1(x)/x`$ and $`K_{}(x)=j_02j_1(x)/x`$, with $`j_l(x)`$ the spherical Bessel function of order l. The cosmological $`\mathrm{\Omega }`$ dependence enters as usual in linear theory via $`f(\mathrm{\Omega })\mathrm{\Omega }^{0.6}`$, and $`H_0`$ is the Hubble constant. A parametric functional form of $`P(k)`$ thus translates to a parametric form of $`\mathrm{\Psi }_{\mu \nu }`$. Note that the quantity that can be derived from peculiar-velocity data via the linear approximation is $`f^2(\mathrm{\Omega })P(k)`$, where $`P`$ is the mass density PS.
Let $`𝐦`$ be the vector of model parameters and $`𝐝`$ the vector of $`N`$ data points. Then Bayes’ theorem states that the posterior probability density of a model given the data is
$$𝒫(𝐦|𝐝)=\frac{𝒫(𝐦)𝒫(𝐝|𝐦)}{𝒫(𝐝)}.$$
(4)
The denominator is merely a normalization constant. The probability density of the model parameters, $`𝒫(𝐦)`$, is unknown, and in the absence of any other information we assume it is uniform within a certain range. The conditional probability of the data given the model, $`𝒫(𝐝|𝐦)`$, is the likelihood function, $`(𝐝|𝐦)`$. The objective in this approach, which is to find the set of parameters that maximizes the probability of the model given the data, is thus equivalent to maximizing the likelihood of the data given the model (cf. Zaroubi et al. 1997; Jaffe & Kaiser 1994). The Bayesian analysis measures the relative likelihood of different models. An absolute frequentist’s measure of goodness of fit is provided by the $`\chi ^2`$ per degree of freedom (hereafter, d.o.f.), which we use as a check of the best parameters obtained by the likelihood analysis.
Assuming that the velocities are a Gaussian random field, the two-point velocity correlation tensor $`\mathrm{\Psi }`$ fully characterizes the statistics of the velocity field. Define the radial-velocity correlation ($`N\times N`$) matrix $`U_{ij}`$ by
$$U_{ij}=\widehat{𝐫}_i\mathrm{\Psi }\widehat{𝐫}_j=\mathrm{\Psi }_{}(r)\mathrm{sin}\theta _i\mathrm{sin}\theta _j+\mathrm{\Psi }_{}(r)\mathrm{cos}\theta _i\mathrm{cos}\theta _j,$$
(5)
where $`i`$ and $`j`$ refer to the data points, $`r=|𝐫|=|𝐫_j𝐫_i|`$ and the angles are defined by $`\theta _i=\widehat{𝐫_i}\widehat{𝐫}`$ (Górski 1988; Groth, Juszkiewicz & Ostriker 1989). Let the inferred radial peculiar velocity at $`𝐫_i`$ be $`u_i^o`$, with the corresponding error $`ϵ_i`$ also assumed to be a Gaussian random variable. The observed correlation matrix is then $`\stackrel{~}{U}_{ij}=U_{ij}+ϵ_i^2\delta _{ij}`$, and the likelihood of the $`N`$ data points is
$$=[(2\pi )^Ndet(\stackrel{~}{U}_{ij})]^{1/2}\mathrm{exp}\left(\frac{1}{2}\underset{i,j}{\overset{N}{}}u_i^o\stackrel{~}{U}_{ij}^1u_j^o\right).$$
(6)
Given that the correlation matrix, $`\stackrel{~}{U}_{ij}`$, is symmetric and positive definite, we can use the Cholesky decomposition method (Press et al. 1992) for computing the likelihood function (Eq. (6)). The significant contribution of the errors to the diagonal terms makes the matrix especially well-conditioned for decomposition.
The errors are assumed to have two contributions, the first is the usual $`D_n\sigma `$ distance proportional errors (about $`19\%`$ per galaxy for ENEAR). The second is a constant error that accounts for the non-linear velocities of galaxies in the high density environment in which early-type galaxies reside. This term represents our poor understanding of the complex correlations introduced by non-linear evolution. For each power spectrum model, we have performed the likelihood analysis assuming this constant value to be either null or $`250\mathrm{km}\mathrm{s}^1`$ but, as shown below, the difference in the results are only marginal and do not affect our general conclusions.
### 3.1 COBE-Normalized CDM Models
We first restrict our attention to the generalized family of CDM cosmological models, allowing variations in the cosmological parameters $`\mathrm{\Omega }`$, $`\mathrm{\Lambda }`$ and $`h`$. Furthermore, four-year COBE normalization is imposed as an additional external constraint. The general form of the PS for these models is
$$P(k)=A_{\mathrm{COBE}}(n,\mathrm{\Omega },\mathrm{\Lambda })T^2(\mathrm{\Omega },\mathrm{\Omega }_B,h;k)k^n,$$
(7)
where the CDM transfer function proposed by Sugiyama (1995) is adopted,
$`T(k)`$ $`=`$ $`{\displaystyle \frac{\mathrm{ln}(1+2.3q))}{2.34q}}`$ (8)
$`\left[1+3.89q+(16.1q)^2+(5.46q)^3+(6.71q)^4\right]^{1/4},`$
$$q=k\left[\mathrm{\Omega }h\mathrm{exp}(\mathrm{\Omega }_bh_{50}^{1/2}\mathrm{\Omega }_b/\mathrm{\Omega })(h\mathrm{Mpc}^1)\right]^1.$$
(9)
The parameters $`\mathrm{\Omega }`$ and $`h`$ are varied such that they span the range of currently popular CDM models, including $`\mathrm{\Lambda }`$CDM ($`\mathrm{\Omega }+\mathrm{\Lambda }=1`$, $`\mathrm{\Omega }1`$) and OCDM ($`\mathrm{\Lambda }=0`$, $`\mathrm{\Omega }1`$). In all cases, the baryonic density is assumed to be $`\mathrm{\Omega }_b=0.019h^2`$, which is the value currently favored by primordial nucleosynthesis analysis (e.g., Burles & Tytler 1998). We limit our inquiry to models without tilt, namely to models with $`n=1`$. For each model, the normalization of the PS is fixed by the COBE 4-year data (Bennet et al. 1996); for more details see Zaroubi et al. (1997 and references therein).
Figure 1a shows the likelihood contour map in the $`\mathrm{\Omega }h`$ plane, for the $`\mathrm{\Lambda }`$CDM family of models with $`n=1`$ (normalization by Sugiyama 1995), the error matrix is assumed here to have an additional diagonal random contribution of $`250\mathrm{km}\mathrm{s}^1`$ that accounts for the nonlinear evolution of the galaxies. The most probable parameters in this case (in the range $`\mathrm{\Omega }1`$) are $`\mathrm{\Omega }=1`$ and $`h=0.5`$. The elongated contours clearly indicate that neither $`\mathrm{\Omega }`$ nor $`h`$ are independently well constrained. It is rather a degenerate combination of the two parameters, approximately $`\mathrm{\Omega }h^x`$ with $`x1`$ (i.e. a combination close to the $`\mathrm{\Gamma }`$ parameter) that is being determined tightly by the elongated ridge of high likelihood.
Figure 1b shows the likelihood results for the same $`\mathrm{\Lambda }`$CDM model shown in Fig. 1a but with no random contribution to the error matrix. The contours in Fig. 1b show very little changes relative to those shown in panel (a), notably they get tighter and the best values of $`\mathrm{\Omega }`$ for a given Hubble constant are somewhat higher. The addition of a reasonable random component to the error matrix does not alter the results in any significant way for any of the PS models considered in this study. For the rest of the PS models we show the calculation with the addition of a constant error of $`250\mathrm{km}\mathrm{s}^1`$.
Figure 2 shows the similar likelihood map for OCDM with $`n=1`$. The most probable values here are $`\mathrm{\Omega }=0.53`$ and $`h=1`$. The values of $`\mathrm{\Omega }`$ and $`h`$ are not independently constrained here as well.
We can thus quote stringent constraints on the conditional best value of $`\mathrm{\Omega }`$ given $`h`$ for the COBE normalized CDM models shown in Figs. 1a and 2: $`\mathrm{\Omega }(0.377\pm 0.08)h^{1.3}`$ for $`\mathrm{\Lambda }`$CDM, and $`\mathrm{\Omega }(0.517\pm 0.083)h^{0.88}`$ for OCDM.
### 3.2 The $`\mathrm{\Gamma }`$ Model
To recover the PS from the velocity data independent of the COBE normalization, we use as a parametric prior the so-called $`\mathrm{\Gamma }`$ model (e.g., Efstathiou, Bond and White 1992),
$`P(k)`$ $`=`$ $`AkT^2(k),`$
$`T(k)`$ $`=`$ $`\left(1+[ak/\mathrm{\Gamma }+(bk/\mathrm{\Gamma })^{3/2}+(ck/\mathrm{\Gamma })^2]^\nu \right)^{1/\nu },`$ (10)
with $`a=6.4h^1\mathrm{Mpc}`$, $`b=3.0h^1\mathrm{Mpc}`$, $`c=1.7h^1\mathrm{Mpc}`$ and $`\nu =1.13`$. The free parameters to be determined by the likelihood analysis are the normalization factor $`\eta _8\sigma _8\mathrm{\Omega }^{0.6}`$ and the $`\mathrm{\Gamma }`$ parameter. In the context of the CDM cosmological model, $`\mathrm{\Gamma }`$ has a specific cosmological interpretation, $`\mathrm{\Gamma }=\mathrm{\Omega }h`$. Here, however, Eq. (10) serves as a generic function with logarithmic slopes $`n=1`$ and $`3`$ on large and small scales respectively, and with a turnover at some intermediate wavenumber that is determined by the single shape parameter $`\mathrm{\Gamma }`$.
Figure 3 shows the contour map of $`\mathrm{ln}`$ in the $`\mathrm{\Gamma }`$$`\eta _8`$ plane. Although the likelihood analysis poses strong constraint on the allowed values of $`\eta _8`$ ($`=1._{0.28}^{+0.3}`$ with $`3\sigma `$ c.l.), it only weakly constrains the value of $`\mathrm{\Gamma }`$ ($`0.18`$ with $`3\sigma `$ c.l.), and $`\mathrm{\Gamma }=0.25`$ is excluded with $`2\sigma `$ c.l..
### 3.3 Results and Comparison between the Various Models
The best fit models for each CDM family have a comparable likelihood, with the most likely model is the OCDM model with $`\mathrm{\Omega }=0.53`$ and $`h=1`$. All best fit models agree within $`20\%`$ for $`k>0.1h\mathrm{Mpc}^1`$. The amplitude of the PS at $`k=0.1h\mathrm{Mpc}^1`$ for all models lies within $`P(k)\mathrm{\Omega }^{1.2}=(6.5\pm 3)\times 10^3(h^1\mathrm{Mpc})^3`$ and the values of $`\eta _8`$ are within the range $`1.1_{0.35}^{+0.2}`$.
Figure 4 shows the power spectrum of the most-likely COBE-normalized model and the $`3\sigma `$ errors about it. It also shows the PS corresponding to most likely models of $`\mathrm{\Lambda }`$CDM and $`\mathrm{\Gamma }`$ models. Within the errors, the most likely PS for each CDM family are very consistent, especially at intermediate scales ($`3050h^1\mathrm{Mpc}`$), where the data information content provide the strongest constraint. Also shown in Fig. 4 are the best-fit PS, obtained from a similar likelihood analysis, for the Mark III and SFI data sets. As can be seen the most likely PS for the three catalogs are in good agreement. This result shows that the high amplitude PS found from peculiar velocity data is unlikely to be due to possible non-uniformities of these catalogs or to the type of galaxies used. In fact, while Mark III and SFI relied predominantly on TF distances to spirals, ENEAR relies on $`D_n\sigma `$ distances to early-type galaxies. On the other hand, the reason for the discrepancy in the cosmological constraints between the maximum likelihood method and other methods (da Costa et al. 1998; Strauss & Willick 1998; Borgani et al. 2000a,2000b) remains unresolved. The former yields a systematically higher amplitude PS, as reflected by the high values of $`\eta _8`$, which is also in disagreement with the constraints derived from other analysis of LSS data. Possible explanations are given in § 5.
In all the COBE-normalized PS models considered the $`\chi ^2/d.o.f.`$ of the best fit models is of the order of 0.93. This value deviates by about $`2\sigma `$ from the $`\chi ^2/d.o.f`$ desired value of unity. This, however, does not pose any serious problem since many of the models within the likelihood most likely contours have a $`\chi ^2/d.o.f.1`$. The $`\chi ^2/d.o.f`$ for the $`\mathrm{\Gamma }`$-model is 0.99.
## 4 Wiener Filter & Constrained Realizations
### 4.1 The Method
Having determined the power spectrum, all the ingredient needed to Wiener reconstruct the density and velocity fields are ready. Details on the general application of the WF/CR method to the reconstruction of large-scale structure are described in Zaroubi et al. (1995), where the theoretical foundation is discussed in relation with other methods of estimation, such as Maximum Entropy. The specific application of the WF/CR method to peculiar velocity data sets has been presented in Zaroubi et al. (1999). Here we provide only a brief description of the WF/CR method, for more details the reader is referred to the original references references given above.
We assume that the peculiar velocity field $`𝐯(𝐫)`$ and the density fluctuation field $`\delta (𝐫)`$ are related via linear gravitational-instability theory. Under the assumption of a specific theoretical prior for the power spectrum $`P(k)`$ of the underlying density field, one can write the WF minimum-variance estimator of the fields as
$$𝐯^{\mathrm{WF}}(𝐫)=<𝐯(𝐫)u_i^o><u_i^ou_j^o>^1u_j^o$$
(11)
and
$$\delta ^{\mathrm{WF}}(𝐫)=<\delta (𝐫)u_i^o><u_i^ou_j^o>^1u_j^o.$$
(12)
A well known problem of the WF is that it attenuates the estimator to zero in regions where the noise dominates. The reconstructed mean field is thus statistically inhomogeneous. In order to recover statistical homogeneity we produce constrained realizations (CR), in which random realizations of the residual from the mean are generated such that they are statistically consistent both with the data and the prior model (Hoffman and Ribak 1991; see also Bertschinger 1987). In regions dominated by good quality data, the CRs are dominated by the data, while in the limit of no data the realizations are practically unconstrained.
The CR method is based on creating random realizations, $`\stackrel{~}{\delta }(𝐫)`$ and $`\stackrel{~}{𝐯}(𝐫)`$, of the underlying fields that obey the assumed PS and linear theory, and a proper set of random errors $`\stackrel{~}{ϵ}_i`$. The velocity random realization is then “observed” like the actual data to yield a mock velocity dataset $`\stackrel{~}{u}_i^o`$. Constrained realizations of the dynamical fields are then obtained by
$$𝐯^{\mathrm{CR}}(𝐫)=\stackrel{~}{𝐯}(𝐫)+<𝐯(𝐫)u_i^o><u_i^ou_j^o>^1\left(u_j^o\stackrel{~}{u}_j^o\right)$$
(13)
and
$$\delta ^{\mathrm{CR}}(𝐫)=\stackrel{~}{\delta }(𝐫)+<\delta (𝐫)u_i^o><u_i^ou_j^o>^1\left(u_j^o\stackrel{~}{u}_j^o\right).$$
(14)
The two types of covariance matrices in the above equations are computed within the framework of linear theory as follows. The covariance matrix of the data $`<u_i^ou_j^o>`$ is the same matrix $`\stackrel{~}{U}_{ij}`$ that appears in eq. 6.
The cross-correlation matrix of the data and the underlying field enters the above equations as, e.g.,
$$<\delta (𝐫)u_j^o>=<\delta (𝐫)𝐯(𝐫_j)>\widehat{𝐫}_j.$$
(15)
The two-point cross-correlation vector between the density and velocity fields is related to the PS via
$$<\delta (𝐱)𝐯(𝐱+𝐫)>=\frac{H_0f(\mathrm{\Omega }_0)}{2\pi ^2}\widehat{𝐫}_0^{\mathrm{}}kP(k)j_1(kr)dk.$$
(16)
The assumption that linear theory is valid on all scales enables us to choose the resolution as well, and in particular to use different smoothing radii for the data and for the recovered fields. In our case no smoothing were applied to the radial velocity data while we choose to reconstruct the density field with a finite Gaussian smoothing of radius $`R`$. This alters the density-velocity correlation function by inserting the multiplicative term $`\mathrm{exp}[k^2R^2/2]`$ into the integrand of eq. (16).
A theoretical estimate of the signal-to-noise ratio ($`S/N`$) at every point in space is given by a simple expression (see Zaroubi et al. 1999) but it requires the calculation and inversion of very large matrices. Therefore, in this study we estimate the point to point error by conducting a large number of CRs. In the case of random Gaussian fields, the ensemble of CRs defined in eq. (13)and eq. (14)samples the distribution of uncertainties in the mean Wiener density and velocity fields (Hoffman & Ribak 1991).
It is worth noting that the WF represents a general minimum-variance solution under the sole assumption that the field is a random field with a known power spectrum. No assumption has to be made here regarding higher order correlations (or the full joint probability distribution functions) of the underlying field. On the other hand, the CRs are derived under the explicit assumption of a full Gaussian random field.
### 4.2 Maps of Density and Velocity Fields
Figure 5 shows the map of the density field along the Supergalactic plane obtained from the ENEAR data using a Gaussian smoothing radius of $`1200\mathrm{km}\mathrm{s}^1`$ (hereafter G12). The shaded area corresponds to the region where the error, as estimated from performing 10 CRs, in density is less than 0.3. The main features of our local universe are easily identified in the WF map, including the Great Attractor (GA) on the left and the Perseus-Pisces supercluster (PP) in the lower right. There is also a hint of the Coma cluster, which lies just outside the sample, in the upper part on the map. Even though different in details, the gross features of the density field are remarkably similar to those obtained by Zaroubi, Hoffman & Dekel (1999) from the application of the same formalism to the Mark III catalog. This is an outstanding result considering the different ways the two catalogs were constructed and the pecuiar velocities measured.
Fig. 6 compares a higher resolution map of the density field recovered from the ENEAR data (left panel) to the density field reconstructed from the PSCz redshift catalog (right panel; Branchini et al. 1999). Both maps are along the Supergalactic plane and were reconstructed using a $`900\mathrm{km}\mathrm{s}^1`$ smoothing radius. The shaded area in the left panel indicates regions where the error is less than 0.45. Even though different in detail the similarities between the density fields are striking lending further credence to the reality of the structures observed in the mass distribution. Note that with the higher resolution some structures become resolved. One can clearly see the Local supercluster at the center of the map and that both the GA and PP may consist of different sub-structures.
The velocity field along the Supergalactic plane is presented in Fig. 7, showing the existence of two convergence regions which roughly coincide with the locations of the GA and PP.
### 4.3 Bulk Velocity
The velocity field has been fitted using a monopole, dipole (i.e. bulk flow) and quadrupole (i.e. shear) expansion within spheres of radii ranging from $`1000`$ to $`6000\mathrm{km}\mathrm{s}^1`$. The three Cartesian components of the bulk velocity (in Supergalactic coordinates) and its absolute value ($`V_B`$) are shown in Fig. 8 as a function of the depth over which the fit has been done. The plots present the bulk velocity of the WF field and of an ensemble of 10 CRs. The plot of the absolute value of the bulk velocity contains also the mean and standard deviation calculated over the ensemble of the CRs. Note that the mean $`V_B`$ of the CRs is higher than its WF value. This result is expected as the WF attenuates the velocity field with the depth, as the observational errors become more dominant.
The amplitude of the bulk flow measured from the reconstructed three-dimensional velocity field ranges from $`V_B=300\pm 70\mathrm{km}\mathrm{s}^1`$ for a sphere of $`R=20h^1\mathrm{Mpc}`$ to $`160\pm 60\mathrm{km}\mathrm{s}^1`$ for $`R=60h^1\mathrm{Mpc}`$. This value is in good agreement with that obtained from a direct fit to the radial peculiar velocities (da Costa et al. 2000b). This result disagrees with the bulk flow determination from the Mark III survey where the amplitude is roughly twice that of ENEAR with a comparable scatter (Zaroubi et al. 1999) but comparable to that measured from the SFI sample.
### 4.4 Large Scale Tidal field
An alternative way of describing the velocity field is to decompose it into two components, one which is induced by the local mass distribution and a tidal component due to mass fluctuations external to the volume considered. Here we follow the procedure suggested by Hoffman (1998a,b) and more recently by Hoffman et al. (2000). The key idea is to solve for the particular solution of the Poisson equation with respect to the WF density field within a given region and zero padding outside. This yields the velocity field induced locally, hereafter the divergent field. The tidal field is then obtained by subtracting the divergent field from the full velocity field. Fig. 9 shows the results of this decomposition applied to the ENEAR survey, where the local volume is a sphere of $`80h^1\mathrm{Mpc}`$ centered on the Local Group. The plots show the full velocity field (upper left panel), the divergent (upper right panel) and the tidal (lower left panel) components. To further understand the nature of the tidal field its bulk velocity component has been subtracted and the residual is shown in the lower right panel. This residual is clearly dominated by a quadrupole component. In principle, the analysis of this residual field can shed light on the exterior mass distribution.
For the ENEAR catalog we find that the local dynamics is hardly affected by structure on scales larger than its depth. For this sample not only the bulk velocity at large radii is small but so is the rms value of the tidal field estimated to be of the order of $`60\mathrm{km}\mathrm{s}^1`$. This is in marked contrast to the the results obtaine from the analysis of the Mark III survey which yields a much stronger tidal field, pointing (in the sense of its quadrupole moment) towards the Shapley concentration. For Mark III the tidal field contributes roughly third of the total bulk velocity ($`200`$km/s).
## 5 CONCLUSION
In the first part of this paper the maximum-likelihood method (Zaroubi et al. 1997) has been used to measure the mass-density power spectrum from the newly completed ENEAR early–type redshift-distance survey. The method assumes that the galaxy peculiar velocities satisfy Gaussian random statistics and that they are linearly related to the mass-density field. The initial fluctuation power spectrum is assumed to be CDM-like, with or without COBE normalizations. In addition the measured peculiar velocities error are assumed to be proportional to the distance with some thermal component to account for the nonlinear evolution of high-density environment in which the early–type galaxies reside.
General results that are valid for all the models used in the analysis, and are independent of the detailed parameterization and normalization used in each model, can be summarized as follows. The amplitude of the power spectrum at $`k=0.1h\mathrm{Mpc}^1`$ is $`P(k)\mathrm{\Omega }^{1.2}=(6.5\pm 3)\times 10^3(h\mathrm{Mpc}^1)^3`$ yielding $`\eta _8=1.1_{0.35}^{+0.2}`$. For the family of COBE-normalized CDM models the following range of parameters was considered: $`\mathrm{\Omega }1`$; $`0.4<h<1`$; and $`n=1`$. Within this range we have obtained a constraint on a combination of the parameters $`\mathrm{\Omega }`$ and $`h`$ which can be approximated by $`\mathrm{\Omega }(0.38\pm 0.08)h^{1.3}`$ for $`\mathrm{\Lambda }`$CDM, and $`\mathrm{\Omega }(0.52\pm 0.083)h^{0.88}`$ for OCDM. For $`h=0.65`$, $`\mathrm{\Lambda }`$CDM yields $`\mathrm{\Omega }=0.50.8`$. Similar constraints are obtained from the analysis of the generic $`\mathrm{\Gamma }`$-models, independent of the COBE normalization. We find that the power spectrum amplitude and shape parameter are constrained to be $`\eta _8=1.0_{0.28}^{+0.3}`$ and $`\mathrm{\Gamma }0.18`$, with larger values of $`\mathrm{\Gamma }`$ ($`>0.4`$) being more probable. We point out that these constraints are consistent with the results obtained from a similar analysis of the Mark III and the SFI peculiar velocity catalogs. This agreement is encouraging since it shows that the results are robust and independent of the sample used.
Examination of the $`\chi ^2/d.o.f.`$ for the most likely COBE-normalized models shows that their values are of the order of 0.93. These values are about $`2\sigma `$ away from the preferred value of 1. However, this should not be too alarming as many of the models within the errors have $`\chi ^2/d.o.f1`$. The $`\chi ^2/d.o.f`$ for the best-fit $`\mathrm{\Gamma }`$-model is 0.99.
As pointed out by previous papers that have analyzed the PS derived from peculiar velocity data (Zaroubi et al. 1997, Freudling et al. 1999), the constraints on $`\eta _8`$ and $`\mathrm{\Gamma }`$ are considerably higher than those obtained from other types of analyzes including peculiar velocity data (Borgani et al. 1997, 2000), cluster abundances, and the galaxy power-spectrum (Efstathiou et al. 1992; Sutherland et al. 1999). They are also not consitent with those obtained by combining the results from high redshift supernovae type Ia (Perlmutter et al. 1999) and the CMB data (Efstathiou et al. 1999) which yields values of $`\mathrm{\Omega }0.25\pm 0.15`$ and $`\mathrm{\Lambda }0.65\pm 0.2`$. Furthermore, assuming a linear galaxy-mass relation the value of $`\eta _8`$ obtained from the present analysis would imply $`\beta =1.0`$ or a $`\beta _I1.4`$ (e.g.Willmer, da Costa & Pellegrini 1999; Sutherland et al. 1999), where the subscript refers to IRAS galaxies, at least a factor of 2 larger than those derived from a velocity-velocity comparison of the IRAS 1.2 Jy gravity field and the Mark III (Davis et al. 1996), SFI ( da Costa et al. 1998) and ENEAR (Nusser et al. 2000) all leading to $`\beta _I0.5`$. These values are also consistent with those derived from small-scale velocities (Fisher et al. 1995).
There are many possible explanations for the above discrepancies. One possibility is that all other analyses have somehow conspired to produce consistent results but yet incorrect interpretation. Even though at first glance this seems unlikely, the recent results from the CMB ballon experiments Boomerang (Bernardis et al. 2000; Lange et al. 2000) and MAXIMA (Hanany et al. 2000; Balbi et al. 2000) have shown that the height of the second peak in the CMB angular power-spectrum is consistent with higher values of $`\mathrm{\Omega }`$. From their most likely models these authors derive $`\mathrm{\Omega }=0.40.8`$
It is important to point out that the method is very sensitive to the assumed error model which can add or supress power. It also implicitly gives a high weight to nearby galaxies, likely to be slow rotators or low velocity dispersion systems, for which the measurements and the distance relations are the least reliable. However, tests have shown that these effects are not very important for the present data set. Another potential problem arises due to the rapid decrease of the weight with distance, the effective volume of the currently available catalogs is small and the shape of the power spectrum is poorly constrained, as illustrated by the case of the $`\mathrm{\Gamma }`$-model. All these factors may impact on the reliability of the constraints obtained from the PS analysis.
Finally, one or more of the theoretical model ingredients could be inaccurate, e.g., power spectrum assumed shapes, Gaussianity of the distribution; or even some inherent bias in the method itself that has eluded the extensive numerical tests carried out with the data and mock samples (e.g.Freduling et al. 1999). In fact, through an eigenmode expansion of the Mark III and SFI galaxy catalogs, Hoffman and Zaroubi (2000) have conducted a mode–by–mode goodness–of–fit analysis. They found that when the surveys are analyzed with their corresponding CDM most likely models, there is a systematic inconsistency between the data and the ‘best-fit’ models suggesting either a generic problem in the peculiar velocity data sets or the inadequacy of the theoretical or error models. Unfortunately, however, the analysis has not been able to point to the exact source of inconsistency.
Finally, in this study we have also performed, given the most probable power spectrum, a Wiener reconstruction of the density and velocity fields. The maps shown here have $`1200\mathrm{km}\mathrm{s}^1`$ and $`900\mathrm{km}\mathrm{s}^1`$ Gaussian resolution and they are limited to the Supergalactic plane. The main features shown are similar to the features in the IRAS reconstruction, corrected for perculiar velocities. The constrained realizations allow us to estimate the point-by-point uncertainties in the recovered maps. In terms of their recovered density fields ENEAR, SFI and Mark III mostly agree. However, they do differ in the velocity fields. ENEAR shows no significant tidal component which contributes about half of the Mark III local bulk velocity. This tidal field accounts for the very different bulk velocities obtained from ENEAR and Mark III, with SFI situated in between these surveys. The results suggest that volumes of $`6080h^1\mathrm{Mpc}`$ are essentially at rest relative to the CMB and that the Local Group motion is primarily due to mass fluctuations within the volume sampled by the existing catalogs of peculiar velocity data.
## 6 Acknowledgments
We thank Enzo Branchini for providing the PSCz density field. We acknowledge Avishai Dekel, Enzo Branchini, Tony Banday, Ravi Sheth, Simon White and Idit Zehavi for stimulating discussions. SZ gratefully acknowledge the hospitality of Kapteyn Astronomical Institute – Groningen.The authors would like to thank M. Maia, C. Rité and O. Chaves for their contribution over the years. MB thanks the Sternwarte München, the Technische Universität München, ESO Studentship program, and MPA Garching for their financial support during different phases of this research. MVA is partially supported by CONICET, SecyT and the Antorchas–Andes– Vitae cooperation. GW is grateful to the Alexander von Humboldt-Stiftung for making possible a year’s stay at the Ruhr-Universität in Bochum, and to ESO for support for visits to Garching which greatly aided this project. Financial support for this work has been given through, Israel Science Foundation grant 103/98 (YH), FAPERJ (CNAW, MAGM, PSSP), CNPq grants 201036/90.8, 301364/86-9 (CNAW), 301366/86-1 (MAGM); NSF AST 9529098 (CNAW); ESO Visitor grant (CNAW). PSP and MAGM thank CLAF for financial support and CNPq fellowships. The results of this paper are based on observations at Complejo Astronomico El Leoncito (CASLEO), operated under agreement between the Consejo Nacional de Investigaciones Científicas de la República Argentina and the National Universities of La Plata, Córdoba and San Juan; Cerro Tololo Interamerican Observatory (CTIO), operated by the National Optical Astronomical Observatories, under AURA; European Southern Observatory (ESO), partially under the ESO-ON agreement; Fred Lawrence Whipple Observatory (FLWO); Observatório do Pico dos Dias, operated by the Laboratório Nacional de Astrofísica (LNA); and the MDM Observatory at Kitt Peak
|
warning/0005/hep-ph0005036.html
|
ar5iv
|
text
|
# Finite Temperature and Large Gauge Invariance
## Introduction:
Gauge theories are beautiful theories which describe physical forces in a natural manner and because of their rich structure, the study of gauge theories at finite temperature is quite interesting in itself. However, to avoid getting into technicalities, we will not discuss the intricacies of such theories either at zero temperature or at finite temperature. Rather, we would study the problem of a fermion interacting with an external gauge field in odd space-time dimensions where some interesting issues arise.
To motivate, let us note that gauge invariance is realized as an internal symmetry in quantum mechanical systems. Consequently, we do not expect a macroscopic external surrounding such as a heat bath to modify gauge invariance. This is more or less what is also found by explicit computations at finite temperature, namely, that gauge invariance and Ward identities continue to hold even at finite temperature . This is certainly the case when one is talking about small gauge transformations for which the parameters of transformation vanish at infinity.
However, there is a second class of gauge invariance, commonly known as large gauge invariance where the parameters do not vanish at infinity and this brings in some new topological character to physical theories. For example, let us consider a $`2+1`$ dimensional Chern-Simons theory of the form
$``$ $`=`$ $`M_{\mathrm{CS}}+_{\mathrm{fermion}}`$ (1)
$`=`$ $`Mϵ^{\mu \nu \lambda }\mathrm{tr}A_\mu (_\nu A_\lambda {\displaystyle \frac{2}{3}}A_\nu A_\lambda )+\overline{\psi }(\gamma ^\mu (i_\mu gA_\mu )m)\psi `$
where $`M`$ is a mass parameter, $`A_\mu `$ a matrix valued non-Abelian gauge field and “tr” stands for the matrix trace. The first term, on the right hand side, is known as the Chern-Simons term which exists only in odd space-time dimensions. We can, of course, also add a Maxwell like term to the Lagrangian and, in that case, the Chern-Simons term behaves like a mass term for the gauge field. Consequently, such a term is also known as a topological mass term (topological because it does not involve the metric). For simplicity of discussion, however, we will not include a Maxwell like term to the Lagrangian.
Under a gauge transformation of the form
$`\psi `$ $``$ $`U^1\psi `$
$`A_\mu `$ $``$ $`U^1A_\mu U{\displaystyle \frac{i}{g}}U^1_\mu U`$ (2)
it is straightforward to check that the action in eq. (1) is not invariant, rather it changes as
$$S=d^3xS+\frac{4\pi M}{g^2}\mathrm{\hspace{0.17em}2}i\pi W$$
(3)
where
$$W=\frac{1}{24\pi ^2}d^3xϵ^{\mu \nu \lambda }\mathrm{tr}_\mu UU^1_\nu UU^1_\lambda UU^1$$
(4)
is known as the winding number. It is a topological quantity which is an integer (Basically, the fermion Lagrangian density is invariant under the gauge transformations, but the Chern-Simons term changes by a total divergence which does not vanish if the gauge transformations do not vanish at infinity. Consequently, the winding number counts the number of times the gauge transformations wrap around the sphere.). For small gauge transformations, the winding number vanishes since the gauge transformations vanish at infinity.
Let us note from eq. (3) that even though the action is not invariant under a large gauge transformation, if $`M`$ is quantized in units of $`\frac{g^2}{4\pi ^2}`$, the change in the action would be a multiple of $`2i\pi `$ and, consequently, the path integral would be invariant under a large gauge transformation. Thus, we have the constraint coming from the consistency of the theory that the coefficient of the Chern-Simons term must be quantized. We have derived this conclusion from an analysis of the tree level behavior of the theory and we have to worry if the quantum corrections can change the behavior of the theory. At zero temperature, an analysis of the quantum corrections shows that the theory continues to be well defined with the tree level quantization of the Chern-Simons coefficient provided the number of fermion flavors is even. The even number of fermion flavors is also necessary for a global anomaly of the theory to vanish and so, everything is well understood at zero temperature.
At finite temperature, however, the situation appears to change drastically. Namely, the fermions induce a temperature dependent Chern-Simons term effectively making
$$MM\frac{g^2}{4\pi }\frac{mN_f}{2|m|}\mathrm{tanh}\frac{\beta |m|}{2}$$
(5)
Here, $`N_f`$ is the number of fermion flavors and this shows that, at zero temperature ($`\beta \mathrm{}`$), $`M`$ changes by an integer (in units of $`g^2/4\pi `$) for an even number of flavors. However, at finite temperature, this becomes a continuous function of temperature and, consequently, it is clear that it can no longer be an integer for arbitrary values of the temperature. It seem, therefore, that temperature would lead to a breaking of large gauge invariance in such a system. This is, on the other hand, completely counter intuitive considering that temperature should have no direct influence on gauge invariance of the theory.
## C-S Theory in $`0+1`$ Dimension:
As we have noted, Chern-Simons terms can exist in odd space-time dimensions. Consequently, let us try to understand this puzzle of large gauge invariance in a simple quantum mechanical theory. Let us consider a simple theory of an interacting massive fermion with a Chern-Simons term in $`0+1`$ dimension described by
$$L=\overline{\psi }_j(i_tAm)\psi _j\kappa A$$
(6)
Here, $`j=1,2,\mathrm{},N_f`$ labels the fermion flavors. There are several things to note from this. First, we are considering an Abelian gauge field for simplicity. Second, in this simple model, the gauge field has no dynamics (in $`0+1`$ dimension the field strength is zero) and, therefore, we do not have to get into the intricacies of gauge theories. There is no Dirac matrix in $`0+1`$ dimension as well making the fermion part of the theory quite simple as well. And, finally, the Chern-Simons term, in this case, is a linear field so that we can, in fact, think of the gauge field as an auxiliary field.
In spite of the simplicity of this theory, it displays a rich structure including all the properties of the $`2+1`$ dimensional theory that we have discussed earlier. For example, let us note that under a gauge transformation
$$\psi _je^{i\lambda (t)}\psi _j,AA+_t\lambda (t)$$
(7)
the fermion part of the Lagrangian is invariant, but the Chern-Simons term changes by a total derivative giving
$$S=𝑑tLS2\pi \kappa N$$
(8)
where
$$N=\frac{1}{2\pi }𝑑t_t\lambda (t)$$
(9)
is the winding number and is an integer which vanishes for small gauge transformations. Let us note that a large gauge transformation can have a parametric form of the form, say,
$$\lambda (t)=iN\mathrm{log}\left(\frac{1+it}{1it}\right)$$
(10)
The fact that $`N`$ has to be an integer can be easily seen to arise from the requirement of single-valuedness for the fermion field. Once again, in light of our earlier discussion, it is clear from eq. (8) that the theory is meaningful only if $`\kappa `$, the coefficient of the Chern-Simons term, is an integer.
Let us assume, for simplicity, that $`m>0`$ and compute the correction to the photon one-point function arising from the fermion loop at zero temperature.
$$iI_1=(i)N_f\frac{dk}{2\pi }\frac{i(k+m)}{k^2m^2+iϵ}=\frac{iN_f}{2}$$
(11)
This shows that, as a result of the quantum correction, the coefficient of the Chern-Simons term would change as
$$\kappa \kappa \frac{N_f}{2}$$
As in $`2+1`$ dimensions, it is clear that the coefficient of the Chern-Simons term would continue to be quantized and large gauge invariance would hold if the number of fermion flavors is even. At zero temperature, we can also calculate the higher point functions due to the fermions in the theory and they all vanish. This has a simple explanation following from the small gauge invariance of the theory . Namely, suppose we had a nonzero two point function, then, it would imply a quadratic term in the effective action of the form
$$\mathrm{\Gamma }_2=\frac{1}{2}𝑑t_1𝑑t_2A(t_1)F(t_1t_2)A(t_2)$$
(12)
Furthermore, invariance under a small gauge transformation would imply
$$\delta \mathrm{\Gamma }_2=𝑑t_1𝑑t_2\lambda (t_1)_{t_1}F(t_1t_2)A(t_2)=0$$
(13)
The solution to this equation is that $`F=0`$ so that there cannot be a quadratic term in the effective action which would be local and yet be invariant under small gauge transformations. A similar analysis would show that small gauge invariance does not allow any higher point function to exist at zero temperature.
Let us also note that eq. (13) has another solution, namely,
$$F(t_1t_2)=\mathrm{constant}$$
In such a case, however, the quadratic action becomes non-extensive, namely, it is the square of an action. We do not expect such terms to arise at zero temperature and hence the constant has to vanish for vanishing temperature. As we will see next, the constant does not have to vanish at finite temperature and we can have non-vanishing higher point functions implying a non-extensive structure of the effective action.
The fermion propagator at finite temperature (in the real time formalism) has the form
$`S(p)`$ $`=`$ $`(p+m)\left({\displaystyle \frac{i}{p^2m^2+iϵ}}2\pi n_F(|p|)\delta (p^2m^2)\right)`$ (14)
$`=`$ $`{\displaystyle \frac{i}{pm+iϵ}}2\pi n_F(m)\delta (pm)`$
and the structure of the effective action can be studied in the momentum space in a straightforward manner. However, in this simple model, it is much easier to analyze the amplitudes in the coordinate space. Let us note that the coordinate space structure of the fermion propagator is quite simple, namely,
$$S(t)=\frac{dp}{2\pi }e^{ipt}\left(\frac{i}{pm+iϵ}2\pi n_F(m)\delta (pm)\right)=(\theta (t)n_F(m))e^{imt}$$
(15)
In fact, the calculation of the one point function is trivial now
$$iI_1=(i)N_fS(0)=\frac{iN_f}{2}\mathrm{tanh}\frac{\beta m}{2}$$
(16)
This shows that the behavior of this theory is completely parallel to the $`2+1`$ dimensional theory in that, it would suggest
$$\kappa \kappa \frac{N_f}{2}\mathrm{tanh}\frac{\beta m}{2}$$
and it would appear that large gauge invariance would not hold at finite temperature.
Let us next calculate the two point function at finite temperature.
$`iI_2`$ $`=`$ $`(i)^2{\displaystyle \frac{N_f}{2!}}S(t_1t_2)S(t_2t_1)`$ (17)
$`=`$ $`{\displaystyle \frac{N_f}{2}}n_F(m)(1n_F(m))`$
$`=`$ $`{\displaystyle \frac{N_f}{8}}\mathrm{sech}^2{\displaystyle \frac{\beta m}{2}}={\displaystyle \frac{1}{2}}{\displaystyle \frac{1}{2!}}{\displaystyle \frac{i}{\beta }}{\displaystyle \frac{(iI_1)}{m}}`$
This shows that the two point function is a constant as we had noted earlier implying that the quadratic term in the effective action would be non-extensive.
Similarly, we can also calculate the three point function trivially and it has the form
$$iI_3=\frac{iN_f}{24}\mathrm{tanh}\frac{\beta m}{2}\mathrm{sech}^2\frac{\beta m}{2}=\frac{1}{2}\frac{1}{3!}\left(\frac{i}{\beta }\right)^2\frac{^2(iI_1)}{m^2}$$
(18)
In fact, all the higher point functions can be worked out in a systematic manner. But, let us observe a simple method of computation for these. We note that because of the gauge invariance (Ward identity), the amplitudes cannot depend on the external time coordinates as is clear from the calculations of the lower point functions. Therefore, we can always simplify the calculation by choosing a particular time ordering convenient to us. Second, since we are evaluating a loop diagram (a fermion loop) the initial and the final time coordinates are the same and, consequently, the phase factors in the propagator (15) drop out. Therefore, let us define a simplified propagator without the phase factor as
$$\stackrel{~}{S}(t)=\theta (t)n_F(m)$$
(19)
so that we have
$$\stackrel{~}{S}(t>0)=1n_F(m),S(t<0)=n_F(m)$$
(20)
Then, it is clear that with the choice of the time ordering, $`t_1>t_2`$, we can write
$`{\displaystyle \frac{\stackrel{~}{S}(t_1t_2)}{m}}`$ $`=`$ $`\beta \stackrel{~}{S}(t_1t_3)\stackrel{~}{S}(t_3t_2)t_1>t_2>t_3`$
$`{\displaystyle \frac{\stackrel{~}{S}(t_2t_1)}{m}}`$ $`=`$ $`\beta \stackrel{~}{S}(t_2t_3)\stackrel{~}{S}(t_3t_1)t_1>t_2>t_3`$ (21)
In other words, this shows that differentiation of a fermionic propagator with respect to the mass of the fermion is equivalent to introducing an external photon vertex (and, therefore, another fermion propagator as well) up to constants. This is the analogue of the Ward identity in QED in four dimensions except that it is much simpler. From this relation, it is clear that if we take a $`n`$-point function and differentiate this with respect to the fermion mass, then, that is equivalent to adding another external photon vertex in all possible positions. Namely, it should give us the $`(n+1)`$-point function up to constants. Working out the details, we have,
$$\frac{I_n}{m}=i\beta (n+1)I_{n+1}$$
(22)
Therefore, the $`(n+1)`$-point function is related to the $`n`$-point function recursively and, consequently, all the amplitudes are related to the one point function which we have already calculated. (Incidentally, this is already reflected in eqs. (17,18)).
With this, we can now determine the full effective action of the theory at finite temperature to be
$`\mathrm{\Gamma }`$ $`=`$ $`i{\displaystyle \underset{n}{}}a^n(iI_n)`$ (23)
$`=`$ $`{\displaystyle \frac{i\beta N_f}{2}}{\displaystyle \underset{n}{}}{\displaystyle \frac{(ia/\beta )^n}{n!}}\left({\displaystyle \frac{}{m}}\right)^{n1}\mathrm{tanh}{\displaystyle \frac{\beta m}{2}}`$
$`=`$ $`iN_f\mathrm{log}\left(\mathrm{cos}{\displaystyle \frac{a}{2}}+i\mathrm{tanh}{\displaystyle \frac{\beta m}{2}}\mathrm{sin}{\displaystyle \frac{a}{2}}\right)`$
where we have defined
$$a=𝑑tA(t)$$
(24)
There are several things to note from this result. First of all, the higher point functions are no longer vanishing at finite temperature and give rise to a non-extensive structure of the effective action. More importantly, when we include all the higher point functions, the complete effective action is invariant under large gauge transformations, namely, under
$$aa+2\pi N$$
(25)
the effective action changes as
$$\mathrm{\Gamma }\mathrm{\Gamma }+NN_f\pi $$
(26)
which leaves the path integral invariant for an even number of fermion flavors. This clarifies the puzzle of large gauge invariance at finite temperature in this model. Namely, when we are talking about large changes (large gauge transformations), we cannot ignore higher order terms if they exist. This may provide a resolution to the large gauge invariance puzzle in the $`2+1`$ dimensional theory as well.
## Exact Result:
In the earlier section, we discussed a perturbative method of calculating the effective action at finite temperature which clarified the puzzle of large gauge invariance. However, this quantum mechanical model is simple enough that we can also evaluate the effective action directly and, therefore, it is worth asking how the perturbative calculations compare with the exact result.
The exact evaluation of the effective action can be done easily using the imaginary time formalism. But, first, let us note that the fermionic part of the Lagrangian in eq. (6) has the form
$$L_f=\overline{\psi }(i_tAm)\psi $$
(27)
where we have suppressed the fermion flavor index for simplicity. Let us note that if we make a field redefinition of the form
$$\psi (t)=e^{i_0^t𝑑t^{}A(t^{})}\stackrel{~}{\psi }(t)$$
(28)
then, the fermionic part of the Lagrangian becomes free, namely,
$$L_f=\overline{\stackrel{~}{\psi }}(i_tm)\stackrel{~}{\psi }$$
(29)
This is a free theory and, therefore, the path integral can be easily evaluated. However, we have to remember that the field redefinition in (28) changes the periodicity condition for the fermion fields. Since the original fermion field was expected to satisfy anti-periodicity
$$\psi (\beta )=\psi (0)$$
it follows now that the new fields must satisfy
$$\stackrel{~}{\psi }(\beta )=e^{ia}\stackrel{~}{\psi }(0)$$
(30)
Consequently, the path integral for the free theory (29) has to be evaluated subject to the periodicity condition of (30).
Although the periodicity condition (30) appears to be complicated, it is well known that the effect can be absorbed by introducing a chemical potential , in the present case, of the form
$$\mu =\frac{ia}{\beta }$$
(31)
With the addition of this chemical potential, the path integral can be evaluated subject to the usual anti-periodicity condition. The effective action can now be easily determined
$`\mathrm{\Gamma }`$ $`=`$ $`i\mathrm{log}\left({\displaystyle \frac{det(i_tm+\frac{ia}{\beta })}{(i_tm)}}\right)^{N_f}`$ (32)
$`=`$ $`iN_f\mathrm{log}\left({\displaystyle \frac{\mathrm{cosh}\frac{\beta }{2}(m\frac{ia}{\beta })}{\mathrm{cosh}\frac{\beta m}{2}}}\right)`$
$`=`$ $`iN_f\mathrm{log}\left(\mathrm{cos}{\displaystyle \frac{a}{2}}+i\mathrm{tanh}{\displaystyle \frac{\beta m}{2}}\mathrm{sin}{\displaystyle \frac{a}{2}}\right)`$
which coincides with the perturbative result of eq. (24).
## Large Gauge Ward Identity:
It is clear from the above analysis that, to see if large gauge invariance is restored, we have to look at the complete effective action. In the $`0+1`$ dimensional model, it was tedious, but we can derive the effective action in closed form which allows us to analyze the question of large gauge invariance. On the other hand, in the theory of interest, namely, the $`2+1`$ dimensional Chern-Simons theory, we do not expect to be able to evaluate the effective action in a closed form. Consequently, we must look for an alternate way to analyze the question of large gauge invariance in a more realistic model. One such possible method may be to derive a Ward identity for large gauge invariance which will relate different amplitudes much like the Ward identity for small gauge invariance does. In such a case, even if we cannot obtain the effective action in a closed form, we can at least check if the large gauge Ward identity holds perturbatively.
It turns out that the large gauge Ward identities are highly nonlinear , as we would expect. Hence, looking for them within the context of the effective action is extremely hard (although it can be done). Rather, it is much simpler to look at the large gauge Ward identities in terms of the exponential of the effective action. Let us define
$$\mathrm{\Gamma }(a)=i\mathrm{log}W(a)$$
(33)
Namely, we are interested in looking at the exponential of the effective action (i.e. up to a factor of $`i`$, $`W`$ is the basic determinant that would arise from integrating out the fermion field). We will restrict ourselves to a single flavor of massive fermions. The advantage of studying $`W(a)`$ as opposed to the effective action lies in the fact that, in order for $`\mathrm{\Gamma }(a)`$ to have the right transformation properties under a large gauge transformation, $`W(a)`$ simply has to be quasi-periodic. Consequently, from the study of harmonic oscillator (as well as Floquet theory), we see that $`W(a)`$ has to satisfy a simple equation of the form
$$\frac{^2W(a)}{a^2}+\nu ^2W(a)=g$$
(34)
where $`\nu `$ and $`g`$ are parameters to be determined from the theory. In particular, let us note that the constant $`g`$ can depend on parameters of the theory such as temperature whereas we expect the parameter $`\nu `$, also known as the characteristic exponent, to be independent of temperature and equal to an odd half integer for a fermionic mode. However, all these properties should automatically result from the structure of the theory. Let us also note here that the relation (34) is simply the equation for a forced oscillator whose solution has the general form
$$W(a)=\frac{g}{\nu ^2}+A\mathrm{cos}(\nu a+\delta )=\frac{g}{\nu ^2}+\alpha _1\mathrm{cos}\nu a+\alpha _2\mathrm{sin}\nu a$$
(35)
The constants $`\alpha _1`$ and $`\alpha _2`$ appearing in the solution can again be determined from the theory. Namely, from the relation between $`W(a)`$ and $`\mathrm{\Gamma }(a)`$, we recognize that we can identify
$`\nu ^2\alpha _1`$ $`=`$ $`{\displaystyle \frac{^2W}{a^2}}|_{a=0}=\left(\left({\displaystyle \frac{\mathrm{\Gamma }}{a}}\right)^2i{\displaystyle \frac{^2\mathrm{\Gamma }}{a^2}}\right)|_{a=0}`$
$`\nu \alpha _2`$ $`=`$ $`{\displaystyle \frac{W}{a}}|_{a=0}=i{\displaystyle \frac{\mathrm{\Gamma }}{a}}|_{a=0}`$ (36)
From the general properties of the fermion theories we have discussed, we intuitively expect $`g=0`$. However, these should really follow from the structure of the theory and they do, as we will show shortly.
The identity (34) is a linear relation as opposed to the Ward identity in terms of the effective action. In fact, rewriting this in terms of the effective action (using eq. (33)), we have
$$\frac{^2\mathrm{\Gamma }(a)}{a^2}=i\left(\nu ^2\left(\frac{\mathrm{\Gamma }(a)}{a}\right)^2\right)ige^{i\mathrm{\Gamma }(a)}$$
(37)
So, let us investigate this a little bit more in detail. We know that the fermion mass term breaks parity and, consequently, the radiative corrections would generate a Chern-Simons term, namely, in this theory, we expect the one-point function to be nonzero. Consequently, by taking derivative of eq. (37) (as well as remembering that $`\mathrm{\Gamma }(a=0)=0`$), we determine (The superscript represents the number of flavors.)
$`(\nu ^{(1)})^2`$ $`=`$ $`\left[\left({\displaystyle \frac{\mathrm{\Gamma }^{(1)}}{a}}\right)^23i{\displaystyle \frac{^2\mathrm{\Gamma }^{(1)}}{a^2}}\left({\displaystyle \frac{\mathrm{\Gamma }^{(1)}}{a}}\right)^1\left({\displaystyle \frac{^3\mathrm{\Gamma }^{(1)}}{a^3}}\right)\right]_{a=0}`$
$`g^{(1)}`$ $`=`$ $`\left[2i{\displaystyle \frac{^2\mathrm{\Gamma }^{(1)}}{a^2}}+\left({\displaystyle \frac{\mathrm{\Gamma }^{(1)}}{a}}\right)^1\left({\displaystyle \frac{^3\mathrm{\Gamma }^{(1)}}{a^3}}\right)\right]_{a=0}`$ (38)
This is quite interesting, for it says that the two parameters in eq. (34) or (37) can be determined from a perturbative calculation. Let us note here some of the perturbative results in this theory, namely,
$`{\displaystyle \frac{\mathrm{\Gamma }^{(1)}}{a}}|_{a=0}`$ $`=`$ $`{\displaystyle \frac{1}{2}}\mathrm{tanh}{\displaystyle \frac{\beta m}{2}}`$
$`{\displaystyle \frac{^2\mathrm{\Gamma }^{(1)}}{a^2}}|_{a=0}`$ $`=`$ $`{\displaystyle \frac{i}{4}}\mathrm{sech}^2{\displaystyle \frac{\beta m}{2}}`$
$`{\displaystyle \frac{^3\mathrm{\Gamma }^{(1)}}{a^3}}|_{a=0}`$ $`=`$ $`{\displaystyle \frac{1}{4}}\mathrm{tanh}{\displaystyle \frac{\beta m}{2}}\mathrm{sech}^2{\displaystyle \frac{\beta m}{2}}`$ (39)
Using these, we immediately determine from eq. (38) that
$$(\nu ^{(1)})^2=\frac{1}{4},g^{(1)}=0$$
(40)
so that the equation (37) leads to the large gauge Ward identity for a single fermion theory of the form,
$$\frac{^2\mathrm{\Gamma }^{(1)}}{a^2}=i\left(\frac{1}{4}\left(\frac{\mathrm{\Gamma }^{(1)}}{a}\right)^2\right)$$
(41)
Furthermore, we determine now from eq. (36)
$$\alpha _1^{(1)}=1,\alpha _2^{(1)}=\pm i\mathrm{tanh}\frac{\beta m}{2}$$
(42)
The two signs in of $`\alpha _2^{(1)}`$ simply corresponds to the two possible signs of $`\nu ^{(1)}`$. With this then, we can solve for $`W(a)`$ in the single flavor fermion theory and we have (independent of the sign of $`\nu ^{(1)}`$)
$$W_f^{(1)}(a)=\mathrm{cos}\frac{a}{2}+i\mathrm{tanh}\frac{\beta m}{2}\mathrm{sin}\frac{a}{2}$$
(43)
which can be compared with eq. (32). For $`N_f`$ flavors, similarly, we can determine the Ward identity to be
$$\frac{\mathrm{\Gamma }^{(N_f)}}{a^2}=iN_f\left(\frac{1}{4}\frac{1}{N_f^2}\left(\frac{\mathrm{\Gamma }^{(N_f)}}{a^2}\right)^2\right)$$
(44)
where the nonlinearity of the Ward identity is manifest.
Similarly, we can determine the large gauge Ward identity for scalar theories as well as supersymmetric theories, but we will not go into the details of this.
## Back to $`2+1`$ dimensions:
With all of this analysis in the simpler $`0+1`$ dimensional model, let us return to the $`2+1`$ dimensional model. Following the results of the $`0+1`$ dimensional model, it has been shown that, for the special choice of the gauge field backgrounds where $`A_0=A_0(t)`$ and $`\stackrel{}{A}=\stackrel{}{A}(\stackrel{}{x})`$, the parity violating part of the effective action of a fermion interacting with an Abelian gauge field takes the form
$$\mathrm{\Gamma }^{PV}=\frac{ie}{2\pi }d^2x\mathrm{arctan}\left(\mathrm{tanh}\frac{\beta m}{2}\mathrm{tan}(\frac{ea}{2})\right)B(\stackrel{}{x})$$
(45)
However, because the choice of the background is very special, it would seem that this may not represent the complete effective action in a general background. In fact, in higher dimensions, such as $`2+1`$, one has to also tackle with the question of the non-analyticity of the thermal amplitudes which leads to a nonuniqueness of the effective action .
With these issues in mind, we have studied the parity violating part of the four point function in $`2+1`$ dimensions at finite temperature. The calculations are clearly extremely difficult and we have evaluated the amplitudes at finite temperature by using the method of forward scattering amplitudes . Without going into details, let me summarize the results here . First, the parity violating part of the box diagram is nontrivial at zero temperature and comes from an effective action of the form
$$\mathrm{\Gamma }_{T=0}^4=\frac{e^4}{64\pi m^6}d^3xϵ^{\mu \nu \lambda }F_{\mu \nu }(^\tau F_{\tau \lambda })F^{\rho \sigma }F_{\rho \sigma }$$
(46)
This is Lorentz invariant and is invariant under both small and large gauge transformations and is compatible with the Coleman-Hill theorem .
At finite temperature, however, the amplitude is not manifestly Lorentz invariant (because of the heat bath) and is non-analytic. We have investigated the amplitude in two interesting limits. Namely, in the long wave limit (all spatial momenta vanishing), the leading term of the amplitude, at high temperature, can be seen to come from an effective action of the form
$$\mathrm{\Gamma }_{LW}^4=\frac{e^4}{512mT}d^3xϵ_{0ij}E_i(_t^1E_j)(_t^1E_k)(_t^1E_k)$$
(47)
where $`\stackrel{}{E}`$ represents the electric field. There are several things to note here. First, this is an extensive action, be it non-local. Second, it is manifestly large gauge invariant and finally, the leading behavior at high temperature goes as $`\frac{1}{T}`$.
In contrast, we can evaluate the amplitude in the static limit (all energies vanishing) where we find the presence of both extensive as well as non-extensive terms at high temperature. However, the extensive terms are suppressed by powers of $`T`$ and the leading term seems to come from an effective action of the form
$$\mathrm{\Gamma }_S^4=\frac{e^4}{4\pi T^2}\left(\mathrm{tanh}\frac{\beta m}{2}\mathrm{tanh}^3\frac{\beta m}{2}\right)d^3xa^3B$$
(48)
This coincides with the amplitude that will come from eq. (45) and has the leading behavior of $`\frac{1}{T^3}`$ at high temperature. Such a term is not invariant under a large gauge transformation. However, we can now derive a large gauge Ward identity for the leading part of the static action and the solution of the Ward identity coincides with the form given in eq. (45).
## Conclusion:
In this talk, we have discussed the question of large gauge invariance at finite temperature. We have discussed the resolution of the problem in a simple $`0+1`$ dimensional model. We have derived the Ward identity for large gauge invariance in this model. We have analyzed the box diagram in $`2+1`$ dimensions and have obtained the form of the effective action at zero temperature. We have also obtained the amplitude as well as the quartic effective actions in the long wave as well as static limits, at finite temperature. The LW limit has only extensive terms in the action which goes as $`\frac{1}{T}`$ at high temperature. The leading term in the static action, however, is non-extensive, goes as $`\frac{1}{T^3}`$ at high temperature and coincides with the effective action proposed earlier for a restrictive gauge background.
This work was supported in part by the U.S. Dept. of Energy Grant DE-FG 02-91ER40685 and NSF-INT-9602559.
|
warning/0005/quant-ph0005028.html
|
ar5iv
|
text
|
# Violations of local realism by two entangled qu𝑁its are stronger than for two qubits
## Abstract
Tests of local realism vs quantum mechanics based on Bell’s inequality employ two entangled qubits. We investigate the general case of two entangled qu$`N`$its, i.e. quantum systems defined in an $`N`$-dimensional Hilbert space. Via a numerical linear optimization method we show that violations of local realism are stronger for two maximally entangled qu$`N`$its ($`3N9`$), than for two qubits and that they increase with $`N`$. The two qu$`N`$it measurements can be experimentally realized using entangled photons and unbiased multiport beamsplitters.
John Bell has shown that no local realistic models can agree with all quantum mechanical predictions for the maximally entangled states of two two-state systems (qubits). After some years researchers started to ask questions about the Bell theorem for more complicated systems. The most surprising answer came from the GHZ theorem : for three or more qubits the conflict between local realism and quantum mechanics is much sharper than for two qubits. The other possible extension are entangled states of pairs of $`N`$-state systems, qu$`N`$its, with $`N3`$. First results, in 1980-82, suggested that the conflict between local realism and quantum mechanics diminishes with growing N . This was felt to be in concurrence with the old quantum wisdom of higher quantum numbers leading to a quasi-classical behavior. However, that early research was confined to Stern-Gerlach type measurements performed on pairs of entangled $`\frac{N1}{2}`$ spins . Since operation of a Stern-Gerlach device depends solely on the orientation of the quantization axis, i.e. on only two parameters, devices of this kind cannot make projections into arbitrary states of the subsystems. That is, they cannot make full use of the richness of the $`N`$-dimensional Hilbert space.
In early 1990’s Peres and Gisin considered certain dichotomic observables applied to maximally entangled pairs of qu$`N`$its. They showed that the violation of local realism, or more precisely of the CHSH inequalities, survives the limit of $`N\mathrm{}`$, but never exceeds the violation by two qubits, in agreement with Cirel’son limit , i.e. it is limited by the factor of $`\sqrt{2}`$. Therefore, the question whether the violation of local realism increases or not with growing $`N`$ for general observables was still left open.
To answer this question it is necessary first to adopt an objective measure of the magnitude of violation of local realism. To this end, consider two qu$`N`$it systems described by mixed states in the form of
$`\rho _N(F_N)=F_N\rho _{noise}+(1F_N)|\mathrm{\Psi }_{max}^N\mathrm{\Psi }_{max}^N|,`$ (1)
where the positive parameter $`F_N1`$ determines the ”noise fraction” within the full state, $`\rho _{noise}=\frac{1}{N^2}\widehat{I}`$, and $`|\mathrm{\Psi }_{max}^N`$ is a maximally entangled two qu$`N`$it state, say
$`|\mathrm{\Psi }_{max}^N={\displaystyle \frac{1}{\sqrt{N}}}{\displaystyle \underset{m=1}{\overset{N}{}}}|m_A|m_B.`$ (2)
In (2) $`|m_A`$ ($`|m_B`$) describes particle $`A`$ ($`B`$) in its mode $`m`$. One has $`{}_{x}{}^{}m|m^{}_{x}^{}=\delta _{m,m^{}}`$, with $`x=A,B`$. The threshold maximal $`F_N^{max}`$, for which the state $`\rho _N(F_N)`$ still does not allow a local realistic model, will be our value of the strength of violation of local realism. The higher $`F_N^{max}`$ the higher noise admixture will be required to hide the non-classicality of the quantum prediction. In experiments the visibility parameter $`V`$, effectively equivalent to $`1F_N`$, is the usual measure of the reduction of interferometric contrast (visibility).
We shall study the case of two observers Alice and Bob performing measurements of local non-degenerate observables, each on her/his qu$`N`$it of an entangled pair in the state $`\rho _N(F_N)`$. Let us imagine that Alice can choose between two non-degenerate observables $`A_1`$ and $`A_2`$, and that each observable is defined such that it has the full spectrum characterized by all integers from $`k=1`$ to $`N`$. Bob can choose between $`B_1`$ and $`B_2`$, both with the same spectrum as above ($`l=1,2,\mathrm{},N`$). Thus, the observers can perform $`2\times 2`$ mutually exclusive global experiments. The quantum probability for the specific pair of results, $`k`$ for Alice and $`l`$ for Bob, provided a specific pair of local observables is chosen, $`A_i`$ by Alice and $`B_j`$ by Bob, will be denoted by $`P_{F_N}^{QM}(k;l|A_i,B_j)`$. Quantum mechanics makes predictions for the complete set of $`4N^2`$ such probabilities, and nothing more.
The hypothesis of local hidden variables tries to go beyond. The basic assumption there is that each particle carries a probabilistic or deterministic set of instructions how to respond to all possible local measurements it might be subject to. Therefore local realism assumes the existence of non-negative joint probabilities involving all possible observations from which it should be possible to obtain all the quantum predictions as marginals (see, e.g. , ). Let us denote these hypothetical probabilities by $`P^{HV}(k,m;l,n|A_1,A_2,B_1,B_2)`$, where $`k`$ and $`m`$, represent the outcome values for Alice’s observables ($`l`$ and $`n`$ for Bob’s). In quantum mechanics one cannot even define such objects, since they involve mutually incompatible measurements. The local hidden variable probabilities for the experimentally observed events, $`k`$ ($`m`$) by Alice measuring $`A_1`$ ($`A_2`$), and $`l`$ ($`n`$) by Bob measuring $`B_1`$ ($`B_2`$), are the marginals
$`P^{HV}(k;l|A_1,B_1)={\displaystyle \underset{m}{}}{\displaystyle \underset{n}{}}P^{HV}(k,m;l,n),`$ (3)
$`P^{HV}(k;n|A_1,B_2)={\displaystyle \underset{m}{}}{\displaystyle \underset{l}{}}P^{HV}(k,m;l,n),`$ (4)
$`P^{HV}(m;l|A_2,B_1)={\displaystyle \underset{k}{}}{\displaystyle \underset{n}{}}P^{HV}(k,m;l,n),`$ (5)
$`P^{HV}(m;n|A_2,B_2)={\displaystyle \underset{k}{}}{\displaystyle \underset{l}{}}P^{HV}(k,m;l,n),`$ (6)
(7)
where $`P^{HV}(k,m;l,n)`$ is a short hand notation for $`P^{HV}(k,m;l,n|A_1,A_2,B_1,B_2)`$. The $`4N^2`$ equations (7) form the full set of necessary and sufficient conditions for the existence of local realistic description of the experiment, i.e., for the joint probability distribution $`P^{HV}(k,m;l,n)`$. The Bell Theorem says that certain predictions by quantum mechanics are in conflict with the local hidden variable model (7). Evidently, the conflict disappears when enough noise is added, as in the state (1), since that noise has a local realistic model. Therefore a threshold $`F_N^{max}`$ exists below which one cannot have any local realistic model with $`P^{HV}(k;l|A_i,B_j)=P_{F_N}^{QM}(k;l|A_i,B_j)`$. Our goal is to find observables for the two qu$`N`$its returning the highest possible critical $`F_N^{max}`$.
Up to date, no one has derived Bell-type inequalities that are necessary and sufficient conditions for (7) to hold, with the exception of the $`N=2`$ case (see ). However there are numerical tools, in the form of the very well developed theory and methods of linear optimization, which are perfectly suited for tackling exactly such problems .
The quantum probabilities, when the state is given by (1), have the following structure
$`P_{F_N}^{QM}(k;l|A_i,B_j)`$ (8)
$`={\displaystyle \frac{1}{N^2}}F_N+(1F_N)P^{QM}(k;l|A_i,B_j),`$ (9)
where $`P^{QM}(k;l|A_i,B_j)`$ is the probability for the given pair of events for the pure maximally entangled state. The set of conditions (7) with $`P_{F_N}^{QM}(k;l|A_i,B_j)`$ replacing $`P^{HV}(k;l|A_i,B_j)`$ imposes linear constraints on the $`N^4`$ “hidden probabilities” $`P^{HV}(k,m;l,n)`$ and on the parameter $`F_N`$, which are the nonnegative unknowns. We have more unknowns ($`N^4+1`$) than equations ($`4N^2+1`$, with the normalization condition for the hidden probabilities), and we want to find the minimal $`F_N`$ for which the set of constraints can still be satisfied. This is a typical linear optimization problem for which lots of excellent algorithms exist. We have used the state-of-the-art algorithm HOPDM 2.30. (Higher Order Primal Dual Method) . It is important to stress that for cross-checking four independently written codes were used, one of them employing a different linear optimization procedure (from the NAG Library).
We were interested in finding such observables for which the threshold $`F_N`$ acquires the highest possible value. To find optimal sets of observables we have used a numerical procedure based on the downhill simplex method (so called amoeba) . If the dimension of the domain of a function is $`D`$ (in our case $`D=4n`$, where $`n`$ is the number of parameters specifying the nondegenerate local observables belonging to a chosen family), the procedure first randomly generates $`D+1`$ points. In this way it creates the vertices of a starting simplex. Next it calculates the value of the function at the vertices and starts exploring the space by stretching and contracting the simplex. In every step, when it finds vertices where the value of the function is higher than in others, it ”goes” in this direction (see e.g. ).
Let us now move to the question of finding a family of observables, which returns critical $`F_N`$’s that are above the well known threshold for the two qubit case, $`1\frac{1}{\sqrt{2}}`$. As it was said earlier, and was confirmed by our numerical results, Stern-Gerlach type measurements are not suitable. More exotic observables are needed.
First we discuss how experiments on two entangled qu$`N`$its might be performed. In view of the unavailability of higher spin entanglement it is fortunate that qu$`N`$it entanglement can be studied exploiting momentum conservation in the many processes of two-particle generation, most notably in the parametric down conversion generation of entangled photon pairs. This results in strong correlations between the propagation directions of the particles in a pair. One can then submit $`N`$ spatial modes of each particle to a multiport beamsplitter .
Application of multiports in the context of quantum entanglement has been first discussed by Klyshko . Proposals of Bell experiments with the multiports were presented in , and further developed in . Multiport devices can reproduce all finite dimensional unitary transformations for single-photon states , therefore they are characterized by $`N^21`$ real parameters.
In order to limit computer time we restricted our analysis to unbiased multiports , more specifically to Bell multiports. Unbiased multiports have the property that a photon entering into any single input port (out of the $`N`$), has equal chances to exit from any output port. In addition, for Bell multiports the elements of their unitary transition matrix, $`𝐔^N`$, are solely powers of the N-th root of unity $`\gamma _N=exp(i2\pi /N),`$ namely $`𝐔_{ji}^N=\frac{1}{\sqrt{N}}\gamma _N^{(j1)(i1)}`$.
Let us now imagine two spatially separated experimenters who perform the experiment of FIG. 1. (described in the caption). The initial maximally entangled state (2) of the two qu$`N`$its can be prepared with the aid of parametric down conversion (see ). The two sets of phase shifters at the inputs of the multiports (one phase shifter in each beam) introduce phase factor $`e^{i(\varphi _A^m+\varphi _B^m)}`$ in front of the $`m`$-th component of the state (2), where $`\varphi _A^m`$ and $`\varphi _B^m`$ denote the local phase shifts.
Each set of local phase shifts constitutes the interferometric realizations of the ”knobs” at the disposal of the observer controlling the local measuring apparatus, which incorporates also the Bell multiport and N detectors. In this way the local observable is defined. Its eigenvalues refer simply to registration at one of the $`N`$ detectors behind the multiport. The quantum prediction for the joint probability $`P_{F_N}^{QM}(k,l)`$ to detect a photon at the $`k`$-th output of the multiport A and another one at the $`l`$-th output of the multiport B is given by :
$`P_{F_N}^{QM}(k,l;\varphi _A^1,\mathrm{}\varphi _A^N,\varphi _B^1,\mathrm{}\varphi _B^N)={\displaystyle \frac{F_N}{N^2}}`$ (11)
$`+{\displaystyle \frac{1F_N}{N}}\left|{\displaystyle \underset{m=1}{\overset{N}{}}}\mathrm{exp}[i(\varphi _A^m+\varphi _B^m)]𝐔_{mk}^N𝐔_{ml}^N\right|^2`$ (12)
$`=({\displaystyle \frac{1}{N^3}})\left(N+2(1F_N){\displaystyle \underset{m>n}{\overset{N}{}}}\mathrm{cos}(𝚽_{kl}^m𝚽_{kl}^n)\right),`$ (13)
where $`𝚽_{kl}^m\varphi _A^m+\varphi _B^m+[m(k+l2)]\frac{2\pi }{N}`$. The counts at a single detector, of course are constant, and do not depend upon the local phase settings: $`P_{F_N}^{QM}(k)=P_{F_N}^{QM}(l)=1/N.`$
The numerical values of the threshold $`F_N`$ are given in fig. 2.
It is evident, that indeed two entangled qu$`N`$its violate local realism stronger than two entangled qubits, and that the violation increases monotonically with $`N`$. It is tempting to contemplate the limit of $`N\mathrm{}`$. While obviously the values of $`F_N^{max}`$ seem to saturate, at present we cannot give a definite asymptotic value.
A few words of comment are needed. One may argue that because of the rather large number of local macroscopic parameters (the phases) defining the function to be maximized with the amoeba we could have missed the global minimum. While this argument cannot be ruled out in principle, we stress that in that case the ultimate violation would even be larger. This would only strengthen our conclusion that two entangled qu$`N`$its are in stronger conflict with local realism than two entangled qubits.
Based on the numerical results, i.e. the values of the optimal phase settings, and on the structure of the local hidden variable model for $`F_3^{max}`$, an algebraic calculation was performed showing that for the two qutrits $`(N=3)`$ experiment the exact value for $`F_3^{max}`$ is $`\frac{116\sqrt{3}}{2}`$. One should also mention that for two spin 1 particles in a singlet state observed by two Stern Gerlach apparatuses our method gives $`F_3^{SG}=0.1945`$, which is much smaller than $`1\frac{1}{\sqrt{2}}`$, confirming that such measurements are not optimal in the sense of leading to maximal possible violations of local realism.
An important question is whether unbiased Bell multiports provide us with a family of observables in maximal conflict with local realism. For a check of this question we have also calculated the threshold value of $`F_3`$ for the case where both observers apply to the incoming qutrit the most general unitary transformation belonging to a full SU(3) group (i.e. we have any trichotomic observables on each side). Again we have assumed that each observer chooses between two sets of local settings. However, in this case each set consists of 8 local settings rather than the three (effectively two) in the tritter case. The result appears to be the same as for two tritters. While this might suggest that for $`N=3`$ Bell multiports are optimal devices to test quantum mechanics against local realism, this needs to be further investigated.
It is interesting to compare our results with the limit for the non-separability of the density matrices (1). The critical minimal $`F_N`$ for which a density matrix (1) is separable is $`\frac{N}{N+1}`$ (see ). The fact that this limit is always higher than ours indicates that the requirement of having a local quantum description of the two subsystems is a much more stringent condition than the requirement of admitting any possible local realistic model.
It will be interesting to consider within our approach different families of states, generalizations to more than two particles, extensions of the families of observables and to see if more than two (e.g. $`A_1,A_2,A_3`$) experiments performed on either side can lead to even stronger violations of local realism. The questions concerning the critical $`F_N`$ are also important in the attempts to generalize Ekert’s quantum cryptographic protocol to qutrits and higher systems .
Our method is numerical, and is based on linear optimization. It is a development of the approach of . The exploding (with $`N`$) difficulty of approaching this type of problems via algebraic-analytical methods (generalized Bell inequalities, via the Farkas lemma, etc.) has been exposed in .
It will certainly be fascinating to see laboratory realizations of the experimental schemes discussed here.
We thank Jacek Gondzio (Edinburgh) for courtesy in allowing to use his most recent version of the code HOPDM. We also thank Adam Baturo and Jan-Åke Larsson for their contribution to the two qubit stage of the project . The work is supported by the Austrian-Polish program 24/2000 Quantum Communication and Quantum Information. Additional support: AZ was supported by the Austrian FWF project F1506; MZ was supported by the University of Gdansk Grant No BW/5400-5-0032-0 and The Erwin Schrödinger International Institute for Mathematical Physics, Vienna; DK was supported by Fundacja na Rzecz Nauki Polskiej and the KBN Grant 2 P03B 096 15.
The paper is dedicated to the memory of the late Professor David N. Klyshko, our Friend (Z & Ż) and a great innovator in the field of nonlinear quantum optics.
|
warning/0005/hep-ph0005273.html
|
ar5iv
|
text
|
# SCALE FACTOR IN DOUBLE PARTON COLLISIONS AND PARTON DENSITIES IN TRANSVERSE SPACE
## I Introduction
Double parton collisions are a new feature of high energy hadron interactions which becomes increasingly important at high energies. The number of partons that can undergo a hard collision is in fact a fast growing function of $`s`$ and, as a consequence, when the c.m. energy is large enough, the probability of having more than one hard partonic interaction in the same inelastic event becomes sizable. Double parton collisions, foreseen long ago by several authors, have been in fact observed recently by CDF. The non-perturbative input to the double parton collisions are the double parton distributions, which are independent on the parton distributions usually considered in large$`p_t`$ physics, since they are related directly to the two-body parton correlation of the hadron structure. In the simplified hypothesis of neglecting all correlations in fractional momenta, the inclusive double parton scattering cross section for the two parton processes $`A`$ and $`B`$ reduces nevertheless to a simplest factorized expression:
$$\sigma _D=m\frac{\sigma _A\sigma _B}{2\sigma _{eff}}$$
(1)
where $`m=1`$ when $`A`$ and $`B`$ are indistinguishable processes and $`m=2`$ when they are distinguishable. $`\sigma _A`$ and $`\sigma _B`$ are the inclusive single scattering cross sections for producing the processes $`A`$ and $`B`$ respectively and all the new information on the structure of the hadron in transverse space is summarized in the value of the scale factor $`\sigma _{eff}`$. This simplest hypothesis has not been contradicted by the experimental evidence, whose results are in fact described with a single parameter (the scale factor $`\sigma _{eff}`$). The experimental value quoted by CDF, $`\sigma _{eff}=14.5\pm 1.7_{2.3}^{+1.7}`$ mb, is however too small to be understood in a simplest uncorrelated picture of the multi-parton distribution and indicates that correlations in transverse space play an important role. A possibility which has been considered to explain the smallness of the value of $`\sigma _{eff}`$ is to correlate the population of gluons and sea quarks with the configuration of the valence in transverse space, in such a way that the average number of gluons and sea quarks is small when the valence quarks are all close in transverse space and, on the contrary, is large when they are separated by a (relatively) large transverse distance. Such a mechanism increases the dispersion in the number of multiple parton collisions and, as a consequence, $`\sigma _D`$ (which is proportional to that dispersion). The value of $`\sigma _{eff}`$ is therefore diminished with respect to the uncorrelated case. The scale factor, on the other hand, is the result of the overlap in transverse space of the matter distribution of the two interacting hadrons and a feature of the model above is that the average transverse distance of a pair of valence quarks is different as compared with the average transverse distance of a pair of gluons or sea quarks. In the model $`\sigma _{eff}`$ is therefore different for double parton scatterings involving valence quarks and for double parton scattering involving sea quarks and gluons. Although the details of the model should be understood in a qualitative rather than in a quantitative sense, it is rather natural to expect different values of the scale factor in different reactions. We think therefore that it may be interesting to have an indication, even if only qualitative, on the size of the effect and, to that purpose, we work out in the present note the values of the scale factor foreseen in the model in a few cases of interest.
## II Scale factor in a correlated model
At large energies one will be able to observe double parton collisions in various channels. In particular the production of two equal sign $`W`$ bosons, at relatively low $`p_t`$, goes almost entirely through double parton collisions and is originated, to a large extent, by interactions with valence quarks. Another case where double parton collisions are expected to play an important role is in the production of two $`b\overline{b}`$ pairs. The corresponding integrated cross section is in fact rather large, by using Eq.1 one may estimate that the double parton scattering cross section for producing two $`b\overline{b}`$ pairs is of the order of $`10\mu \mathrm{b}`$ at the energy of the LHC. The dominant mechanism of $`b\overline{b}`$ production is gluon fusion and the observation of double parton collisions, in a equal sign $`W`$ boson pair production and in the production of two $`b\overline{b}`$ pairs, allows therefore one to compare the distribution in transverse space of valence quarks with the distribution in transverse space of gluons.
To work out the scale factor in the two cases we write the inclusive double parton scattering cross section, for the parton interactions $`A`$ and $`B`$, as
$$\sigma _D(A,B)=\frac{m}{2}\underset{ijkl}{}_{p_t^{cut}}\mathrm{\Gamma }_{ij}(x_1,x_2;b)\widehat{\sigma }_{ik}^A(x_1,x_1^{})\widehat{\sigma }_{jl}^B(x_2,x_2^{})\mathrm{\Gamma }_{kl}^{}(x_1^{},x_2^{};b)𝑑x_1𝑑x_1^{}𝑑x_2𝑑x_2^{}d^2b$$
(2)
where $`\mathrm{\Gamma }_{ij}(x_1,x_2;b)`$ are the double parton distributions, depending on the two fractional momenta of the partons $`x_1`$ and $`x_2`$ and on their relative transverse distance $`b`$. The indices $`i`$ and $`j`$ refer to the different kinds of partons, $`\widehat{\sigma }_{ik}^A`$ and $`\widehat{\sigma }_{jl}^B`$ are the partonic cross sections and the QCD dependence on the scale of the interaction is implicit in all quantities. In the case we are considering the dependence of $`\mathrm{\Gamma }`$ on $`x_1`$, $`x_2`$ and $`b`$ is factorized:
$$\mathrm{\Gamma }_{ij}(x_1,x_2;b)=G_i(x_1)G_j(x_2)F_j^i(b)$$
(3)
where $`G_i(x)`$ are the usual parton distributions. If $`F_j^i(b)`$ do not depend on $`i`$ and $`j`$ one obtains Eq.1 and the scale factor $`\sigma _{eff}`$ is universal. In general the double scattering cross section is however written as
$$\sigma _D(A,B)=\frac{m}{2}\underset{ijkl}{}\mathrm{\Theta }_{kl}^{ij}\sigma _{ij}(A)\sigma _{kl}(B)$$
(4)
where $`\sigma _{ij}(A)`$ is the hadronic inclusive cross section for two partons of kind $`i`$ and $`j`$ to undergo the hard interaction $`A`$, and
$$\mathrm{\Theta }_{kl}^{ij}=d^2bF_k^i(b)F_{l}^{j}{}_{}{}^{}(b)$$
(5)
are geometrical coefficients, with dimensions of the inverse of a cross section, depending on the various parton processes.
In the model the probability to find the proton in a given configuration, with $`n`$ gluons or sea-quarks, has the following expression:
$`P(X_v,𝐁_v;x_1,𝐛_1,\mathrm{}x_n,𝐛_n)`$ $`=q_v(X_1)q_v(X_2)q_v(X_3)\phi (𝐁_D,𝐁)`$ (6)
$`\times {\displaystyle \frac{1}{n!}}[g(x_1)f(B_D,b_1)`$ $`\mathrm{}g(x_n)f(B_D,b_n)\left]\mathrm{exp}\right\{{\displaystyle \frac{B_D^2}{B_D^2}}{\displaystyle }g(x)dx\}`$ (7)
where $`q_v(X)`$ are the inclusive distributions of valence quarks, as a function of the momentum fraction $`X`$, and $`\phi (𝐁_D,𝐁)`$ is the density of valence quarks in transverse space, where the coordinates of valence quarks in transverse space $`𝐁_v`$ are given by
$`𝐁_1`$ $`=`$ $`{\displaystyle \frac{1}{2}}𝐁_D+𝐁`$ (8)
$`𝐁_2`$ $`=`$ $`{\displaystyle \frac{1}{2}}𝐁_D𝐁`$ (9)
$`𝐁_3`$ $`=`$ $`𝐁_D`$ (10)
In the model $`\phi (𝐁_D,𝐁)=𝑑Z_D𝑑Z\varphi (𝐑_D,𝐑)`$, where $`\varphi (𝐑_D,𝐑)`$ represents the distribution density of the proton in coordinate space. The explicit expression used is
$$\varphi (𝐑_D,𝐑)=\frac{\lambda _D^3\lambda ^3}{(8\pi )^2}\mathrm{exp}\left\{(\lambda _DR_D+\lambda R)\right\}$$
(11)
where
$`\lambda _D`$ $`=`$ $`{\displaystyle \frac{2\sqrt{3}}{\sqrt{r^2}}}`$ (12)
$`\lambda `$ $`=`$ $`{\displaystyle \frac{4}{\sqrt{r^2}}}`$ (13)
with $`\sqrt{r^2}=0.81`$fm, the proton charge radius.
Given a configuration of the valence, the average density of gluons and sea quarks in a point with transverse coordinate $`b`$ and momentum fraction $`x`$ is given by $`g(x)f(B_D,b)`$. The overall average number of gluons and sea quarks at a given $`x`$ (namely after integrating on $`b`$ and on the configurations of the valence) is $`g(x)`$ and is identified with the inclusive distribution of gluons and sea quarks. In the same way the inclusive distributions of the valence quarks, $`q_v(X)`$, are the result of integrating over the transverse coordinates $`𝐁_v`$ and of summing over all configurations of gluons and sea quarks.
The dependence of the average density of gluons and sea quarks on the transverse coordinate $`b`$ is expressed by
$$f(B_D,b)=\frac{3}{2\pi }\left(1\frac{b^2}{B_D^2}\right)^{1/2}\frac{1}{B_D^2}\theta (B_Db)$$
(14)
which is the projection on the transverse plane of a sphere of radius $`B_D`$, rescaled with the factor $`B_D^2/B_D^2`$, where the average $`B_D^2`$ is taken with the density $`\phi (𝐁_D,𝐁)`$. The average density of gluons and sea quarks and the corresponding average number (which grows as $`B_D^2/B_D^2`$) are therefore correlated with the configuration of the valence, while all fractional momenta are uncorrelated.
The model distinguishes two different kinds of partons, as far as the number and density in transverse space are concerned, the valence quarks and the sea quark and gluons. The indices $`i`$ and $`j`$, as a consequence, can take two different values, $`v`$ and $`s`$, and the relevant transverse densities of parton pairs to be considered are
$`F_v^v(b)`$ $`={\displaystyle \frac{1}{6}}{\displaystyle \underset{ji=1}{\overset{3}{}}}{\displaystyle d^2Bd^2B_D\phi (𝐁_D,𝐁)\delta (𝐁_i𝐁_j𝐛)}`$ (15)
$`F_s^v(b)`$ $`={\displaystyle \frac{1}{3}}{\displaystyle \underset{i=1}{\overset{3}{}}}{\displaystyle d^2Bd^2B_Dd^2b^{}\phi (𝐁_D,𝐁)f(B_D,b^{})\delta (𝐁_i𝐛^{}𝐛)}`$ (16)
$`F_s^s(b)`$ $`={\displaystyle d^2Bd^2B_Dd^2b^{}d^2b^{\prime \prime }\phi (𝐁_D,𝐁)f(B_D,b^{})f(B_D,b^{\prime \prime })\delta (𝐛^{}𝐛^{\prime \prime }𝐛)}`$ (17)
By using Eq.5 and the expression below, the matrix $`\mathrm{\Theta }_{kl}^{ij}`$ and the effective cross section are readily evaluated:
$$\underset{ijkl}{}\mathrm{\Theta }_{kl}^{ij}\sigma _{ij}(A)\sigma _{kl}(B)=\frac{\sigma (A)\sigma (B)}{\sigma _{eff}}$$
(18)
## III Results
The resulting values of $`1/\mathrm{\Theta }_{kl}^{ij}`$ are shown in the table. The scale factor $`\sigma _{eff}`$ is plotted as a function of the c.m. energy in fig.1 and in fig.2, corresponding to $`pp`$ and $`p\overline{p}`$ interactions respectively, in various processes of interest: $`i`$) production of two equal sign $`W`$ bosons, $`ii`$) production of a $`W`$ boson of either positive or negative sign together with two jets or $`iii`$) with a $`b\overline{b}`$ pair, $`iv`$) production of four jets, $`v`$) production of two $`b\overline{b}`$ pairs. In the case of jets the scale factor has been evaluated by using as a lower cutoff in $`p_t`$ the value of $`5GeV`$ (the scale factor is however rather insensitive to that choice). All the single scattering cross sections in Eq.18 have been computed at the lowest order in the coupling constant and by making use of the Martin-Roberts-Stirling 1999 (MRS99) parton distributions. The different results obtained for $`W^+W^+`$ and $`W^{}W^{}`$ production for $`pp`$ collisions is mainly due to the different content of $`d`$ and $`u`$ quarks in the proton. A consequence is in fact the different contribution of the sea quarks in the two cases, whose distribution is sizably different in the model in comparison with the distribution in transverse space of the valence. To check the dependence on the choice of the distribution functions we have repeated the calculation by using the CTEQ5 parton distributions. In all cases the scale factor is not changed by more than a few percent. In fact the results for $`\sigma _{eff}`$ are rather insensitive to the choice of parton distributions, since $`\sigma _{eff}`$ is obtained by making ration of cross sections. The main qualitative feature of the results obtained is the strong difference, at Tevatron energy, between final states with and without a $`W`$ boson, and the energy dependence of the scale factor, which is sizable when moving from Tevatron to LHC energy in the channels containing a $`W`$ boson.
While the actual values should be considered only as indicative, the qualitative features just pointed out are likely to be of more general validity, since they are originated by the different source of initial state partons in the case of $`W`$ production, which in most cases involves valence quarks, in comparison with the other channels considered, that are mainly generated by interactions of sea quarks and gluons. The overall indication is that it is plausible to expect non minor differences in the values of $`\sigma _{eff}`$ in different processes. It is also apparent that a basic element, to understand the three dimensional structure of the proton, is a quantitative information on the correlations in transverse space of the various pairs of initial state partons and that the scale factors, of the different double parton collision processes, are the physical observables which can provide that information.
Acknowledgment
This work was partially supported by the Italian Ministry of University and of Scientific and Technological Researches (MURST) by the Grant COFIN99.
|
warning/0005/hep-ph0005300.html
|
ar5iv
|
text
|
# Phase structure of the QCD vacuum in a magnetic field at low temperature
## Abstract
We study the QCD phase structure in magnetic field $`H`$ at low temperature $`T`$. The hadronic phase free energy in a constant homogeneous magnetic field is calculated in one-loop approximation of the chiral perturbation theory. The dependence of the quark and gluon condensates upon the temperature and field strength is found. It is shown that the chiral phase transition order parameter $`\overline{q}q`$ remains constant provided field strength and temperature are related via $`H=constT^2`$.
PACS:11.10.Wx,12.38.Aw,12.38.Mh
1. The investigation of the vacuum state behavior under the influence of the various external factors is known to be one of the central problems in quantum field theory. In the realm of strong interactions (QCD) the main factors are the temperature and the baryon density. At temperatures below the chiral phase transition, $`T<T_c`$, the dynamics of the system is characterized by confinement and spontaneous breaking of chiral symmetry (SBCS). At low temperatures, $`T<T_c`$, the partition function of the system is dominated by the contribution of the lightest particles in the physical spectrum. In QCD this role is played by the $`\pi `$ – meson which is the Goldstone excitation mode in chiral condensate. Therefore the low temperature physics (the hadron phase) enables an adequate description in terms of the effective chiral theory . A very important problem is the behavior of the order parameter (the quark condensate $`\overline{q}q)`$ with the increase of the temperature. In the ideal gas approximation the contribution of the thermal pions into the quark condensate is proportional to $`T^2`$ . In chiral perturbation theory (ChPT) the two – and three–loops contributions ($`T^4`$ and $`T^6`$ correspondingly) into $`\overline{q}q`$ have been found in and .
The situation with the gluon condensate $`G^2(G_{\mu \nu }^a)^2`$is very different. The gluon condensate is not an order parameter in phase transition and it does not lead to any spontaneous symmetry breaking (SSB). At the quantum level the trace anomaly leads to the breaking of the scale invariance and this in turn results in nonzero value of $`G^2`$. However, this is not a SSB phenomenon and hence does not lead to the appearance of the Goldstone particle. The mass of the lowest excitation (dilaton) is directly connected to the gluon condensate, $`m_D(G^2)^{1/4}`$. Thus the thermal excitations of glueballs are exponentially suppressed by the Boltzmann factor $`\mathrm{exp}\{m_{gl}/T\}`$ and their contribution to the shift of the gluon condensate is small ($`\mathrm{\Delta }G^2/G^2`$0.1 % at $`T=`$100 MeV) . Next we note that in the one-loop approximation ChPT pions are described as a gas of massless noninteracting particles. Such a system is obviously scale-invariant and therefore does not contribute into the trace of the energy-momentum tensor and correspondingly into $`G^2`$. As it has been demonstrated in the gluon condensate temperature dependence arises only at the ChPT three–loop level.
Another interesting problem is the study of the vacuum phase structure under the influence of the external magnetic field $`H`$. Quarks play an active role in shaping the QCD vacuum structure. Being dual carriers of both ’color’ and ’electric’ charges they also respond to externally applied electromagnetic fields. The vacuum of strong interactions influences some QED processes has been discussed in Ref. . The behavior of the quark condensate in the presence of a magnetic field was studied in Nambu-Iona-Lasinio model earlier . For QCD, the analogous investigation in the one-loop approximations was done in . It was found that the quark condensate grows with the increase of the magnetic field $`H`$ in both cases. It implies that a naive analogy with superconductivity, where the order parameter vanishes at same critical field, is not valid here. The behavior of the gluon condensate $`G^2`$ in the Abelian magnetic field is also a nontrivial effect. Gluons do not carry electric charge; nevertheless, virtual quarks produced by them interact with electromagnetic field and lead to the changes in the gluon condensate. This phenomenon was studied in , based on the low-energy theorems in QCD. The vacuum energy density, the values of $`G^2`$ and $`\overline{q}q`$ as functions of $`H`$ have been found in the two–loop approximation ChPT in .
The low-energy theorems, playing an important role in the understanding of the vacuum state properties in quantum field theory, were discovered almost at the same time as quantum field methods appeared in particle physics (see, for example Low theorems ).In QCD, these theorems were obtained in the beginning of eighties . These theorems, being derived from the very general symmetrical considerations and not depending on the details of confinement mechanism, sometimes give information which is not easy to obtain in another way. Also, they can be used as ”physically sensible” restrictions in the constructing of effective theories. An important step was made in , where low-energy theorems for gluodynamics were generalized to finite temperature case.
In the present paper the vacuum free energy in magnetic field at finite temperature is calculated in the framework of ChPT. The general relations are established which allow to obtain the dependence of the quark and gluon condensates on $`T`$ and $`H`$. A new phenomenon is displayed, namely the ”freezing” of the chiral phase transition order parameter by the magnetic field when the temperature increases. The physical meaning of this fact is discussed.
2. The QCD Euclidean partition function in external Abelian field $`A_\mu `$ has the following form ($`T=1/\beta `$)
$$Z=exp\left\{\frac{1}{4e^2}_0^\beta 𝑑x_4_Vd^3xF_{\mu \nu }^2\right\}[DB][D\overline{q}][Dq]\mathrm{exp}\left\{_0^\beta 𝑑x_4_Vd^3x\right\}.$$
(1)
Here the QCD Lagrangian in the background field is
$$=\frac{1}{4g_0^2}(G_{\mu \nu }^a)^2+\underset{q=u,d}{}\overline{q}[\gamma _\mu (_\mu iQ_qA_\mu i\frac{\lambda ^a}{2}B_\mu ^a)+m_q]q,$$
(2)
where $`Q_q`$ – is the matrix of the quark charges for the quarks $`q=(u,d)`$, and for the simplicity the ghost terms have been omitted. The free energy density is given by the relation $`\beta VF`$ $`(T,H,m_q)=\mathrm{ln}Z`$. In the chiral limit $`m_q0`$ Eq. (1) yields the following expressions for the quark and gluon condensates
$$\overline{q}q(T,H)=\frac{F(H,T,m_q)}{m_q}|{}_{m_q=0}{}^{},$$
(3)
$$G^2(T,H)=4\frac{F(H,T,m_q)}{(1/g_0^2)}|{}_{m_q=0}{}^{}.$$
(4)
The phenomenon of dimensional transmutation results in the appearance of a nonperturbative dimensional parameter
$$\mathrm{\Lambda }=M\mathrm{exp}\left\{_{\alpha _s(M)}^{\mathrm{}}\frac{d\alpha _s}{\beta (\alpha _s)}\right\},$$
(5)
where $`M`$ is the ultraviolet cutoff, $`\alpha _s=g_0^2/4\pi `$, and $`\beta (\alpha _s)=d\alpha _s(M)/dlnM`$ is the Gell-Mann-Low function. In chiral limit $`(m_q=0)`$ the system described by the partition function (1) is characterized by three dimensionful parameters $`M,T`$ and $`H`$. The free energy density is renorm-invariant quantity and hence its anomalous dimension is zero. Thus $`F`$ has only a normal (canonical) dimension equal to 4. Making use of the renorm-invariance of $`\mathrm{\Lambda }`$, one can write in the most general form
$$F=\mathrm{\Lambda }^4f(\frac{T}{\mathrm{\Lambda }},\frac{H}{\mathrm{\Lambda }^2}),$$
(6)
where the function $`f`$ is still unknown. From (5) and (6) one gets
$$\frac{F}{(1/g_0^2)}=\frac{F}{\mathrm{\Lambda }^2}\frac{\mathrm{\Lambda }^2}{(1/g_0^2)}=\frac{4\pi \alpha _s^2}{\beta (\alpha _s)}(4T\frac{}{T}2H\frac{}{H})F.$$
(7)
With the account of (4) the gluon condensate is given by
$$G^2(T,H)=\frac{16\pi \alpha _s^2}{\beta (\alpha _s)}(4T\frac{}{T}2H\frac{}{H})F(T,H)$$
(8)
At $`T=0`$, $`H=0`$ we recover the well known expression for the nonperturbative vacuum energy density in the chiral limit. In the one-loop approximation $`(\beta =b_0\alpha _s^2/2\pi ,b_0=(11N_c2N_f)/3)`$ it has the form
$$\epsilon _v=F(0,0)=\frac{b_0}{128\pi ^2}G^2$$
(9)
Let us now derive the low-energy theorems at finite temperature in the presence of magnetic field. Strictly speaking, $`\beta `$-function depends on $`H`$ and the low-energy theorems acquire electromagnetic corrections $`e^4`$. However, since the free energy is independent of the scale $`M`$ at which the ultraviolet divergences are regulated, we can choose $`M^2H,T^2,\mathrm{\Lambda }^2`$. Hence, we can take the lowest order expression for $`\beta `$-function ($`\beta (\alpha _s)=b_0\alpha _s^2/2\pi )`$ and the electromagnetic corrections vanish. Let us introduce the field $`\sigma (\tau =x_4,𝐱)`$ and operator $`\widehat{D}`$,
$$\sigma (\tau ,𝐱)=\frac{b_0}{32\pi ^2}(G_{\mu \nu }^a(\tau ,𝐱))^2,$$
(10)
$$\widehat{D}=4T\frac{}{T}2H\frac{}{H}.$$
(11)
Differentiating (4) $`n`$ times with respect to $`(1/g_0^2)`$ and taking into account (7), (10) and (11) one obtains
$$\widehat{D}^{n+1}F=\widehat{D}^n\sigma (0,\mathrm{𝟎})$$
$$=𝑑\tau _nd^3x_n\mathrm{}𝑑\tau _1d^3x_1\sigma (\tau _n,𝐱_n)\mathrm{}\sigma (\tau _1,𝐱_1)\sigma (0,\mathrm{𝟎})_c.$$
(12)
The subscript $`c`$ means that only connected diagrams are to be included. Proceeding in the same way, it is possible to obtain the similar theorems for renorm-invariant operator $`O(x)`$ which is built from quark and/or gluon fields
$$\left(T\frac{}{T}+2H\frac{}{H}d\right)^nO$$
$$=𝑑\tau _nd^3x_n\mathrm{}𝑑\tau _1d^3x_1\sigma (\tau _n,𝐱_n)\mathrm{}\sigma (\tau _1,𝐱_1)O(0,\mathrm{𝟎})_c.$$
(13)
where $`d`$ is the canonical dimension of operator $`O`$. If operator $`O`$ has also anomalous dimension, the appropriate $`\gamma `$-function should be included.
3.The above equations enable to obtain the values of the condensates as functions of $`T`$ and $`H`$ provided the free energy density is known. To get the latter the ChPT will be used. At low temperatures $`T<T_c(T_c`$ is the chiral phase transition temperature) and for weak fields $`H<\mu _{hadr}^2(4\pi F_\pi )^2`$ the characteristic momenta in the vacuum loops are small and theory is adequately described by the low-energy effective chiral Lagrangian $`L_{eff}`$ . This Lagrangian can be represented as a series expansion over the momenta (derivatives) and quark masses
$$L_{eff}=L^{(2)}+L^{(4)}+L^{(6)}+\mathrm{}$$
(14)
The leading term in (14) is similar to the Lagrangian of the non-linear sigma model in the external field
$$L^{(2)}=\frac{F_\pi ^2}{4}Tr(_\mu U^+_\mu U)+\mathrm{\Sigma }ReTr(\widehat{M}U^+),$$
$$_\mu U=_\mu Ui[U,V_\mu ].$$
(15)
Here $`U`$ is a unitary $`SU(2)`$ matrix, $`F_\pi =93`$MeV is the pion decay constant, and $`\mathrm{\Sigma }`$ has the meaning of the quark condensate $`\mathrm{\Sigma }=|\overline{u}u|=|\overline{d}d|`$. The external Abelian magnetic field $`H`$ is aligned along the $`z`$ -axis and corresponds to $`V_\mu (x)=(\tau ^3/2)A_\mu (x)`$ with the vector-potential $`A_\mu `$ chosen as $`A_\mu (x)=\delta _{\mu 2}Hx_1`$. The mass difference between the $`u`$ and $`d`$ quarks appears in the effective chiral Lagrangian only quadratically. Further, to obtain an expression for the quark condensate in the chiral limit we use only the first derivative with respect to the mass of one of the quarks. Therefore, we can neglect the mass difference between the $`u`$ and $`d`$ quarks and assume the mass matrix to be diagonal $`\widehat{M}=m\widehat{I}`$.
At $`T<T_c,H<\mu _{hadr}^2`$ the QCD partition function coincides with the partition function of the effective chiral theory
$$Z_{eff}[T,H]=e^{\beta VF_{eff}[T,H]}=Z_0[H][DU]\mathrm{exp}\{_0^\beta 𝑑x_4_Vd^3xL_{eff}[U,A]\}$$
(16)
At the one-loop level it is sufficient to restrict the expansion of $`L_{eff}`$ by the quadratic terms with respect to the pion field. Using the exponential parameterization of the matrix $`U(x)=\mathrm{exp}\{i\tau ^a\pi ^a(x)/F_\pi \}`$ one finds
$$L^{(2)}=\frac{1}{2}(_\mu \pi ^0)+\frac{1}{2}M_\pi ^2(\pi ^0)^2+(_\mu \pi ^++iA_\mu \pi ^+)(_\mu \pi ^{}iA_\mu \pi ^{})+M_\pi ^2\pi ^+\pi ^{},$$
(17)
where the charged $`\pi ^\pm `$ and neutral $`\pi ^0`$ meson fields are introduced
$$\pi ^\pm =(\pi ^1\pm i\pi ^2)/\sqrt{2},\pi ^0=\pi ^3$$
(18)
Thus (16) can be recasted into the form<sup>1</sup><sup>1</sup>1 The partition function $`Z_{eff}^R`$ describes charged $`\pi ^\pm `$ and neutral $`\pi ^0`$ ideal Bose gas in magnetic field. Relativistic charged Bose gas in magnetic field at finite temperature and density with application to Bose-Einstein condensation and Meissner effect was studied in Refs. , , , .
$$Z_{eff}^R[T,H]=Z_{p.t.}^1Z_0[H][D\pi ^0][D\pi ^+][D\pi ^{}]\mathrm{exp}\{_0^\beta 𝑑x_4_Vd^3xL^{(2)}[\pi ,A]\}$$
(19)
where partition function is normalized for the case of perturbation theory at $`T=0,H=0`$
$$Z_{p.t.}=[det(_\mu ^2+M_\pi ^2)]^{3/2}.$$
(20)
Integration of (19) over $`\pi `$-fields leads to
$$Z_{eff}^R=Z_{p.t.}^1Z_0[H][\mathrm{det}_T(_\mu ^2+M_\pi ^2)]^{1/2}[\mathrm{det}_T(|D_\mu |^2+M_\pi ^2)]^1,$$
(21)
where $`D_\mu =_\mu iA_\mu `$ is a covariant derivative and a symbol $`T`$ means that the determinant is calculated at finite temperature $`T`$ according to standard Matsubara rules. Taking (20) into account and regrouping multipliers in (21) one gets the following expression for $`Z_{eff}^R`$
$$Z_{eff}^R[T,H]=Z_0[H]\left[\frac{det_T(_\mu ^2+M_\pi ^2)}{det(_\mu ^2+M_\pi ^2)}\right]^{1/2}\left[\frac{det(|D_\mu |^2+M_\pi ^2)}{det(_\mu ^2+M_\pi ^2)}\right]^1$$
$$\times \left[\frac{det_T(|D_\mu |^2+M_\pi ^2)}{det(|D_\mu |^2+M_\pi ^2}\right]^1$$
(22)
Then the effective free energy can be written in the form
$$F_{eff}^R(T,H)=\frac{1}{\beta V}\mathrm{ln}Z_{eff}^R=\frac{H^2}{2e^2}+F_{\pi ^0}(T)+F_{\pi ^\pm }(H)+F_s(T,H).$$
(23)
Here $`F_{\pi ^0}`$ is the free energy of massive scalar boson
$$F_{\pi ^0}(T)=T\frac{d^3p}{(2\pi )^3}\mathrm{ln}(1\mathrm{exp}(\sqrt{𝐩^2+M_\pi ^2}/T)),$$
(24)
$`F_{\pi ^\pm }`$ is a Schwinger result for the vacuum energy density of charged scalar particles in the magnetic field.
$$F_{\pi ^\pm }(H)=\frac{1}{16\pi ^2}_0^{\mathrm{}}\frac{ds}{s^3}e^{M_\pi ^2s}[\frac{Hs}{\mathrm{sinh}(Hs)}1],$$
(25)
and
$$F_s(T,H)=\frac{HT}{\pi ^2}\underset{n=0}{\overset{\mathrm{}}{}}_0^{\mathrm{}}𝑑k\mathrm{ln}(1\mathrm{exp}(\omega _n/T)),$$
$$\omega _n=\sqrt{k^2+M_\pi ^2+H(2n+1)},$$
(26)
where $`\omega _n`$ are Landau levels of the $`\pi ^\pm `$ mesons in constant field $`H`$. <sup>2</sup><sup>2</sup>2Technically, a transition for the free energy $`F=\frac{1}{2}Tr\mathrm{ln}(p_4^2+\omega _0^2(𝐩))`$ from the vacuum case ($`H=0,T=0)`$ to the case of $`H0,T0`$ is straightforward. Omitting the details of the calculations, we note that, eventually, this transition reduces to the substitutions $`p_4\omega _k=2\pi kT`$ ($`k=0,\pm 1,\mathrm{}`$), $`\omega _0=\sqrt{𝐩^2+M_\pi ^2}\omega _n=\sqrt{p_z^2+M_\pi ^2+H(2n+1)}`$ and $`Tr\frac{HT}{2\pi }_{n=0}^{\mathrm{}}_{k=\mathrm{}}^+\mathrm{}_{\mathrm{}}^+\mathrm{}\frac{dp_z}{2\pi },`$ where the degeneracy multiplicity of $`H/2\pi `$ has been taken into account for the Landau levels. Performing summation over Matsubara frequencies, we obtain (26).
4.The free energy $`F_{eff}^R`$ determines the thermodynamical properties and the phase structure of the QCD vacuum state below the temperature of the chiral phase transition, i.e. in the phase of confinement.
Equations (8) and (23) describe the dependence of $`G^2`$ on $`T`$ and $`H`$. The action of the operator $`\widehat{D}`$ on $`F_{eff}^R`$ leads to $`\widehat{D}F_{\pi ^0}(T)=0`$ since $`M_\pi ^20`$ and $`F_{\pi ^0}(T)T^4`$ in chiral limit and thus (4–$`T/T)F_{\pi ^0}(T)=0`$. It can be easily shown by direct calculation that $`\widehat{D}F_s(T,H)=0`$. The nontrivial dependence of $`G^2`$ on $`H`$ arises only due to Schwinger term $`F_{\pi ^\pm }(H)`$
$$G^2(T,H)=G^2+\frac{\alpha _s^2}{3\pi \beta (\alpha _s)}H^2$$
(27)
Next we note that because of the asymptotic freedom the QCD $`\beta `$-function, $`\beta (\alpha _s)=b_0\alpha _s^2/2\pi +\mathrm{}`$ is negative and hence the gluon condensate diminishes with the $`H`$ increasing
$$G^2(T,H)=G^2\frac{2}{3b_0}H^2.$$
Thus, the temperature corrections to the gluon condensate in magnetic field vanish at the ChPT one-loop level.
In order to get the dependence of the quark condensate upon $`T`$ and $`H`$ use is made of the Gell-Mann-Oakes–Renner relation (GMOR)
$$F_\pi ^2M_\pi ^2=\frac{1}{2}(m_u+m_d)\overline{u}u+\overline{d}d=2m\mathrm{\Sigma }$$
(28)
Substituting (23) into (3), calculating the derivative over $`M_\pi ^2`$ and then taking the limit $`M_\pi ^20`$ one gets
$$\overline{q}q(T,H)=\overline{q}q(1\frac{1}{3}\frac{T^2}{8F_\pi ^2}+\frac{H}{(4\pi F_\pi )^2}\mathrm{ln}2\frac{H}{2\pi ^2F_\pi ^2}\phi (\frac{\sqrt{H}}{T}))$$
$$\phi (\lambda )=\underset{n=0}{\overset{\mathrm{}}{}}_0^{\mathrm{}}\frac{dx}{\omega _n(x)(\mathrm{exp}(\lambda \omega _n(x))1)},\omega _n(x)=\sqrt{x^2+2n+1}$$
(29)
Now we consider various limiting cases. In the strong field, $`\sqrt{H}T`$ $`(\lambda 1)`$, the lowest Landau level ($`n=0`$) gives the main contribution to the sum (29)
$$\phi (\lambda 1)=\sqrt{\frac{\pi }{2\lambda }}e^\lambda +O(e^{\sqrt{3}\lambda }).$$
(30)
In the opposite limit of weak field, $`\sqrt{H}T(\lambda 1)`$, the sum in (29) is calculated with required accuracy using the Euler-MacLaren formula. Furthermore, one gets the following result with the use of the asymptotic expansion of integral (29) at $`\lambda 1`$
$$\phi (\lambda 1)=\frac{\pi ^2}{6}\frac{1}{\lambda ^2}+\frac{7\pi }{24}\frac{1}{\lambda }+\frac{1}{4}\mathrm{ln}\lambda +C+\frac{\zeta (3)}{48\pi ^2}\lambda ^2+O(\lambda ^4),$$
(31)
here $`C=\frac{1}{4}(\gamma \mathrm{ln}4\pi \frac{1}{6}),\gamma =0.577\mathrm{}`$ is Euler’s constant and $`\zeta (3)=1.202`$ is Riemann zeta function. Thus, one obtains the following limiting expressions for the quark condensate in the chiral limit in a magnetic field at $`T0`$
$$\frac{\overline{q}q(T,H)}{\overline{q}q}=1\frac{1}{3}\frac{T^2}{8F_\pi ^2}+\frac{H}{(4\pi F_\pi )^2}\mathrm{ln}2\frac{H^{3/4}T^{1/2}}{(2\pi )^{3/2}F_\pi ^2}e^{\sqrt{H}/T},\sqrt{H}T$$
(32)
and
$$\frac{\overline{q}q(T,H)}{\overline{q}q}=1\frac{T^2}{8F_\pi ^2}+\frac{H}{(4\pi F_\pi )^2}A\frac{7\sqrt{H}T}{48\pi F_\pi ^2}\frac{H}{(4\pi F_\pi )^2}\mathrm{ln}\frac{H}{T^2},\sqrt{H}T$$
(33)
where $`A=\mathrm{ln}2+8C4.93`$.
In the framework of ChPT the quark condensate (29) at $`H0`$, $`T0`$ is determined by three dimensionless parameters $`H/(4\pi F_\pi )^2`$, $`T^2/F_\pi ^2`$ and $`\lambda =\sqrt{H}/T`$. The quantity $`\lambda `$ is a natural dimensionless parameter in this approach. The motion of a particle (massless pion) in the field is characterized by the curvature radius of it’s trajectory, and in the magnetic field this is the Larmor radius $`R_L=1/\sqrt{H}`$. On the other hand, there is another length $`l_T=1/T`$ \- ”temperature length” at $`T0`$. Therefore, charged $`\pi ^\pm `$-mesons in magnetic field effectively acquire ”mass”, $`m_{eff}=\sqrt{H}`$, determined by the lowest Landau level, when Larmor radius of a particle in the field is much less than $`l_T(\lambda 1)`$. Correspondingly, their contribution to the shift of the chiral condensate is suppressed by the Boltzman factor $`\mathrm{exp}\{m_{eff}/T\}`$. In the weak field limit $`\pi ^\pm `$-mesons give standard temperature one-loop approximation ChPT contribution to $`\overline{q}q`$. Besides, additional temperature and magnetic corrections appear. Neutral $`\pi ^0`$-meson contributes to $`\overline{q}q(T,H)`$ as usual massless scalar particle.
An interesting phenomenon reveals itself in the vacuum QCD phase structure under consideration. One can find from (29) such a function $`H(T)`$ that the chiral condensate $`\overline{q}q(T,H)`$ remains unchanged when the temperature and magnetic field change in accordance with $`H_{}=H(T)`$. Then $`H_{}`$ is found by solving the following equation (see (29) $`\overline{q}q(T,H_{})\overline{q}q=0)`$
$$1\frac{3}{2\pi ^2}\lambda ^2\mathrm{ln}2+\frac{12}{\pi ^2}\lambda ^2\phi (\lambda )=0$$
(34)
The numerical solution of (34) yields $`\lambda _{}=0.111\mathrm{}`$ Thus, quark condensate stays unchanged when $`T`$ and $`H`$ are increased according to $`H=0.013T^2`$. Hence it is possible to say that the order parameter $`\overline{q}q`$ of the chiral phase transition is ”frozen” by the magnetic field.
Note that $`H(T_c)/(4\pi F_\pi )^2210^41`$ at $`T=T_c150`$MeV and therefore the above relations remain valid up to the deconfined phase transition point. In the vicinity of $`T_c`$ the effective low energy chiral Lagrangian fails to provide an adequate description of the QCD vacuum thermodynamical properties, and strictly speaking becomes physically invalid. The following is worth noting. In deriving (29), at the first step the physical quantity as functions of $`M_\pi `$ where obtained, and only then the chiral limit $`M_\pi 0`$ was taken. Acting in the inversed sequence we would have obtained all temperature corrections to condensate identically equal to zero. This points to the fundamental difference of the two cases: the exactly massless particle and the particle with infinitesimal small mass.
5. It has been shown in the present letter that the quark condensate is ”frozen” by the magnetic field when both temperature $`T`$ and magnetic field $`H`$ are increased according to the $`H=constT^2`$ law. This points to the fact that the direct analogy between the quark condensate in QCD and the theory of superconductivity is untenable. In the BCS theory the Cooper pairs condensate is extinguished by the temperature and magnetic field. The ”freezing” phenomenon can be understood in terms of the general Le Chatelier–Braun principle <sup>3</sup><sup>3</sup>3The external action disturbing the system from the equilibrium state induces processes in this system which tend to reduce the result of this action.The external field contributes into the system an additional energy density $`H^2/2`$. The system tends to compensate this energy change and to decrease the free energy by increasing the absolute value of the quark condensate: $`\mathrm{\Delta }\epsilon _v=m|\mathrm{\Sigma }(H)\mathrm{\Sigma }(0)|<0`$. On the other hand, if the temperature of the system is increased (by bringing some heat into it) the processes with heat absorbtion by damping the condensate are switched on. The interplay of these processes is at the origin of the above ”freezing” of $`\mathrm{\Sigma }(T,H)`$. Next, since gluons do not carry electric charge, the magnetic field affects the gluon sector of the vacuum only indirectly via the quark sector and thus the Le Chatelier–Braun principle is not applicable directly to the gluon condensate. For the same reason gluon condensate decreases nonlinearly with $`H`$ increasing according to $`\mathrm{\Delta }G^2H^2`$, while for the quark condensate $`\mathrm{\Delta }\mathrm{\Sigma }H`$.
The author is grateful to B.L.Ioffe, V.A.Novikov, Yu.A.Simonov, A.V.Smilga and S.M.Fedorov for comments and discussions. The financial support of RFFI grant 00-02-17836 is gratefully acknowledged.
|
warning/0005/astro-ph0005215.html
|
ar5iv
|
text
|
# Spectroscopy of Giant Stars in the Pyxis Globular Cluster
## 1 Introduction
Evidence continues to accumulate that the outermost Milky Way globular clusters may not have originated in the same process that formed the inner globular clusters. Based on the recognition that the second parameter effect of horizontal branch morphology in globular clusters is found predominantly among outer halo ($`R_{gc}>8`$ kpc) clusters, Searle & Zinn (1978) proposed that the outermost globular clusters may have formed in chemically distinct “fragments” that later fell into the Milky Way halo. Building on the suggestion by Kunkel & Demers (1976) that several red horizontal branch (second parameter) globular clusters were potentially associated with the Magellanic Plane group of dwarf galaxies, Majewski (1994) showed that there is a planar alignment between a particular sample of second parameter globular clusters and the Milky Way dwarfs. Recently, Palma et al. (2000) reaffirmed that there may be a dynamical relationship between the second parameter globular clusters and the Milky Way dwarf satellites.
The Pyxis globular cluster (Irwin et al. (1995); Da Costa (1995)) at $`R_{gc}=41`$ kpc (Sarajedini & Geisler 1996) defines the inner edge of the prominent gap in the globular cluster radial distribution between $`40R_{gc}60`$ kpc. The presence of this gap has been used to argue that the primordial Galactic globular cluster system ends at $``$40 kpc while the distant, $`R_{gc}>60`$ kpc clusters originated in Galactic satellite dwarf galaxies (e.g., Zinn (1985)). Although Pyxis lies among the “inner group” of globular clusters (i.e., inside the gap in $`R_{gc}`$), Irwin et al. (1995) propose that Pyxis may be a captured LMC globular cluster based on the young age they infer for the globular cluster and on its proximity to the plane of the LMC orbit derived from the Jones et al. (1994) proper motion. Further support for the tidal capture hypothesis comes from Palma et al. (2000), where a statistical analysis of the likely orbital poles of the Galactic satellite galaxies and the globular clusters identifies Pyxis, NGC 6229, NGC 7006, and Pal 4 as the globular clusters most likely to share a common orbital pole with either the Magellanic Plane galaxies (the LMC, the SMC, Draco, and Ursa Minor) or the Fornax–Leo–Sculptor Stream galaxies. However, these postulations on the origin of Pyxis have been made without the benefit of any kinematical data on the cluster.
Although deep photometry of Pyxis exists (Sarajedini & Geisler 1996), no spectroscopic observations have been published. Indeed, Pyxis is one of the last few known globular clusters lacking a radial velocity (cf. Harris 1996). We report here on du Pont 2.5-m Telescope spectroscopic observations of Pyxis stars (§2). With our derived radial velocity for the cluster, we re-address the stripped LMC hypothesis for Pyxis’ origin (§3), but point out that, in the end, we can only make predictions on the proper motions expected under this scenario. Unfortunately, the proper motion is required for a definitive solution to the question of the cluster’s origin.
## 2 Observations
On the nights of 17 January and 20 January 1997, the 2.5-m du Pont Telescope at the Las Campanas Observatory was used to obtain spectra of Pyxis giant stars with magnitudes of $`R18`$. A finding chart for these stars made from the Digitized Sky Survey (Lasker et al. (1990)) is presented as Figure 1. Observations of the Pyxis stars made use of the modular spectrograph and the SITe2 detector with a 1200 lines/mm grating. The selected wavelength region with this setup was approximately 7700 – 8750 Å at $``$1.3 Å per pixel resolution. Typical exposure times were 900 seconds per observation, which provided enough signal-to-noise to measure radial velocities.
Radial velocities were measured for six Pyxis giant stars by cross-correlating their spectra against those of bright radial velocity standard stars. The cross-correlation peak yielded the adopted radial velocity difference between the Pyxis stars and the standard. The star HD80170, a K5 giant, was observed multiple times on both nights to serve as the main standard. Two other HD stars and 11 LMC globular cluster stars were also observed as radial velocity calibration objects. All standards with known radial velocities were used to remove any nightly zero point offsets in the radial velocities determined by cross-correlating against the HD80170 template. The Pyxis stars were observed multiple times, and the radial velocity for each star was determined by taking the median of the values for a particular star. In Table 1 we present the positions, magnitudes, and velocities of the stars observed. We find a mean radial velocity for the six stars of $`34.3`$ km/sec with $`\sigma =4.6`$ km/sec. If we ignore the one outlier, Pyxis D, the mean radial velocity for the five remaining stars increases slightly, to $`35.9`$ km/sec with $`\sigma =2.5`$ km/sec.
### 2.1 Ca II Triplet Metallicity Estimate
The wavelength range of our spectra includes the Ca II triplet lines at $`\lambda \lambda `$ 8498, 8542, and 8662 Å. Rutledge et al. (1997a) measured Ca II equivalent widths for a large sample of globular cluster stars and in a companion paper (Rutledge et al. 1997b) derived a conversion from Ca II reduced equivalent width, $`W^{}`$, to metallicity. To determine a spectroscopic metallicity for Pyxis, we attempted to measure the strengths of these lines in our spectra. Of the six Pyxis stars observed, only the brightest, Pyxis A, produced equivalent width measurements with errors small enough to give a reliable estimate of the metallicity. The technique used to determine the equivalent widths for the Pyxis A Ca II lines was nearly identical to that used in Rutledge et al. (1997a).
For each observation of Pyxis A, the continuum was determined by linearly interpolating the average intensity in the Rutledge et al. (1997a) continuum bandpasses. The equivalent width was then calculated by integrating the difference between the fit continuum and the line feature over the line bandpass. The line feature was fit with a Gaussian function, and the integral was performed using the fit rather than numerically integrating the data since Rutledge et al. (1997a) concluded that this technique is preferable to direct numerical integration. The three individual lines were combined into a single index, $`\mathrm{\Sigma }`$Ca, following the method of Rutledge et al. (1997a), where $`\mathrm{\Sigma }`$Ca $`=0.5\lambda _{8498}+\lambda _{8542}+0.6\lambda _{8662}`$. For Pyxis A, the result is $`\mathrm{\Sigma }`$Ca $`=4.5\pm 0.2`$ Å.
The Rutledge et al. (1997a) method for converting $`\mathrm{\Sigma }`$Ca to metallicity requires the reduction of the equivalent width to the value for giants at the level of the horizontal branch, $`W^{}`$, so that a mean value for all cluster stars can be obtained. This is done by adopting a slope $`\mathrm{\Delta }(\mathrm{\Sigma }\mathrm{Ca})/\mathrm{\Delta }(V_{HB}V)`$ and extrapolating the calculated width to the expected value at the magnitude of the horizontal branch, $`V_{HB}`$. All of the published photometry for Pyxis stars has used the $`B`$ and $`R`$ bands, so we have had to approximate $`\mathrm{\Delta }(V_{HB}V)`$ using color-color relations for giant stars. Adopting the photometry from Sarajedini & Geisler (1996), we find that Pyxis A has $`R=17.08`$ and that the horizontal branch of Pyxis is at $`R=18.75`$. Caldwell et al. (1993) find an almost linear relationship between $`(BV)`$ and $`(VR)`$ for giant stars, so we have determined rough $`V`$ magnitudes for Pyxis A and the Pyxis horizontal branch stars by estimating a $`(VR)`$ color from the $`(BR)`$ color given in Sarajedini & Geisler (1996). We estimate that for Pyxis, $`V_{HB}=19.25`$ and $`V_{PyxisA}=17.77`$, or $`\mathrm{\Delta }(V_{HB}V)=1.48`$. In the table of globular cluster properties by Harris (1996), the magnitude of the horizontal branch is also given as $`V_{HB}=19.25`$, so the adopted $`(VR)`$ colors are most likely a good approximation to the true colors.
The reduced Ca II equivalent width of Pyxis A is therefore $`W^{}=3.6`$ if we follow Rutledge et al. (1997a) and adopt a slope of $`\mathrm{\Delta }(\mathrm{\Sigma }\mathrm{Ca})/\mathrm{\Delta }(V_{HB}V)=0.62`$ Å/magnitude. This value is simply an estimate, since there is a dispersion of 0.2 Å in the values of $`\mathrm{\Sigma }`$Ca from the four individual observations and since there is some uncertainty in $`\mathrm{\Delta }(V_{HB}V)`$, probably of order 0.1 magnitudes. However, we can use this determination to get an estimate of the metallicity of Pyxis A for comparison with the photometrically determined metallicity estimates of Irwin et al. (1995) and Sarajedini & Geisler (1996). Using the Rutledge et al. (1997b) calibration of $`W^{}`$ to Zinn-West metallicity, the reduced equivalent width measured for Pyxis A implies a metallicity of $`[`$Fe/H$`]_{ZW}=1.4\pm 0.1`$. This value for Pyxis A is more metal-poor than the photometrically derived values of Sarajedini & Geisler (1996) and Irwin et al. (1995), who estimate $`1.2\pm 0.15`$ and $`1.1\pm 0.3`$ respectively, however it is consistent within the overlap of the $`1\sigma `$ error bars. Any systematic error that leads to an underestimated equivalent width for Pyxis A results in a smaller determined metallicity. An error of 10% in the equivalent width measured for Pyxis A is enough to bring the metallicity up to $`1.2`$ and into better agreement with the photometric values. If the equivalent width measurement is correct, then it is unlikely that the metallicity is much higher than $`[`$Fe/H$`]_{ZW}=1.4`$, since an unlikely error in the $`V`$ magnitude of Pyxis A of 0.7 magnitudes is required to raise the metallicity of the star to $`1.2`$.
## 3 Discussion
The observations presented here were partially motivated by the possibility that the Pyxis globular cluster was captured from the LMC by the Milky Way. This assertion was originally made by Irwin et al. (1995), who noted that Pyxis, at $`(l,b)=(261.3,7.0)^{}`$, lies within a few degrees of the orbital plane of the LMC determined from the Jones et al. (1994) proper motion. Further support for this hypothesis is provided by Palma et al. (2000), who place Pyxis in a group with Pal 4, NGC 6229, and NGC 7006 as the most likely globular clusters to share a common orbital pole with the Magellanic Plane galaxies (the LMC, the SMC, Ursa Minor, and Draco). Although full space motion information is required to verify the Irwin et al. (1995) hypothesis, a radial velocity can provide some constraints on the shapes of allowed orbits for a cluster if the magnitude of the radial component is a significant fraction of the expected magnitude of the space velocity.
The essence of the argument given in Palma et al. (2000) to support a capture origin for the Pyxis globular cluster is as follows: if it is assumed that Pyxis was captured recently from the LMC by the Milky Way, then the orbital pole of Pyxis is likely to be aligned with that of the LMC (both the LMC and Pyxis are far enough from the Galactic Center that precession will not significantly affect the positions of their orbital poles over a Hubble time). If one assumes rotation about the Galactic Center, the direction of the orbital pole of the LMC can be determined by taking the cross product of the Galactocentric radius vector to the LMC and its space motion vector. If one accounts for the space velocity vector error, the position of the orbital pole of the LMC can only be confined to a family of poles along an arc segment in Galactocentric coordinates (cf. Figure 1 in Palma et al. 2000). Since the space motion of Pyxis is currently unknown, its orbital pole is not well constrained. However, the orbital pole can be assumed to be perpendicular to its current Galactocentric position, so the direction of Pyxis’ orbital pole should lie on the great circle that contains all possible normals to its current radius vector (cf. Lynden-Bell & Lynden-Bell (1995)). Figure 2 shows an Aitoff projection of the sky in Galactocentric coordinates. The arc segment that defines the possible locations of the LMC’s orbital pole (based on the Jones et al. proper motion, as adopted by Palma et al. ) is shown as well as the great circle along which lies all possible orbital poles of Pyxis. That these two families of possible orbital poles for the LMC and for Pyxis intersect (at Galactocentric $`(l,b)=(163,22)^{}`$) indicates that it is possible for these two objects to share a coplanar orbit with a common direction of angular momentum. It now remains to be seen whether our derived radial velocity can clarify whether this is likely, i.e., is the orbital pole of Pyxis likely to be near the crossing point of the two orbital pole families?
We start our analysis by adopting a simple strawman model wherein Pyxis is following a circular orbit that is nearly coplanar with the orbit of the LMC. Simulations of tidal stripping of dwarf galaxies by the Milky Way (Johnston 1998) show that the debris is distributed around the orbit of the parent satellite with a spread in energy given by
$$\mathrm{\Delta }E=r_{\mathrm{tide}}\frac{d\mathrm{\Phi }}{dR}\left(\frac{m_{\mathrm{sat}}}{M_{\mathrm{Gal}}}\right)^{1/3}v_{\mathrm{circ}}^2fv_{\mathrm{circ}}^2$$
(1)
where $`r_{\mathrm{tide}}`$ is the tidal radius of the satellite, $`\mathrm{\Phi }`$ is the parent galaxy gravitational potential, $`v_{\mathrm{circ}}`$ is the circular velocity of the Galactic halo, $`m_{\mathrm{sat}}`$ is the satellite’s mass, $`M_{\mathrm{Gal}}`$ is the mass of the parent galaxy enclosed within the satellite’s orbit, and the last equality defines the tidal scale $`f`$. Thus, the spread in energy translates into a characteristic angular width $`f`$ (in radians) to the debris. Taking reasonable values for the mass of the LMC and the Milky Way, the value of $`f`$ for any debris pulled from the LMC corresponds to roughly $`15^{}`$. In Figure 2, an arc segment along the great circle of possible orbital poles for Pyxis is marked; this arc segment is defined by a length within $`\pm 15`$ degrees of the intersection point with the possible orbital pole of the LMC, and indicates expectations for the orbital poles of LMC debris at the position of Pyxis on the sky.
It is straightforward to derive the direction of the space motion vector required for Pyxis to follow a circular orbit and have an orbital pole along the arc segment in Figure 2, i.e. in the direction of $`\widehat{P}`$, or $`(l,b)=(163,22)^{}`$. The geometry is illustrated in Figure 3. The Galactocentric, Cartesian radius vector of Pyxis is $`(X,Y,Z)=(13.9,38.6,4.8)`$ kpc (assuming that the distance from the Sun to the Galactic Center, $`R_0=8`$ kpc, and where $`X_{\mathrm{}}8.0`$ kpc). The unit vector $`\widehat{P}`$ from the Galactic center in the direction of the orbital pole at Galactocentric $`(l,b)=(163,22)^{}`$ is $`(X,Y,Z)=(0.89,0.27,0.37)`$. In Figure 3, this vector has been translated to the location of Pyxis. The vector that is mutually perpendicular to the Galactocentric radius vector of Pyxis and to the orbital pole $`\widehat{P}`$ gives the direction of the space motion of the Pyxis globular cluster for a circular orbit around the Galactic center. The unit vector direction of this space motion is $`\widehat{V}_{circ}=(V_X,V_Y,V_Z)=(0.32,0.23,0.92)`$. Since our line of sight to Pyxis is mostly in the $`Y`$ direction and since $`\stackrel{}{V}_{circ}`$ is mostly in the $`+Z`$ direction, clearly any component of $`\stackrel{}{V}_{circ}`$ along our line of sight will be small. Adopting the basic solar motion of $`(9,12,6)`$ km/sec (Mihalas & Binney (1981)) and a rotational velocity of the local standard of rest (LSR) of 220 km/sec, then the component of the Sun’s velocity along the line of sight from the Sun to Pyxis is $`v_{\mathrm{}LOS}227`$ km/sec (the negative sign here indicates that the component of the Sun’s velocity with respect to the Galactic Standard of Rest along the line of sight to Pyxis is in the sense of receding from Pyxis; see the Appendix for a discussion of the sign conventions used in reducing the radial velocity to a $`v_{GSR}`$). Since the heliocentric radial velocity measured for Pyxis is the difference between the intrinsic radial velocity of Pyxis with respect to the position of the Sun, $`v_{GSR}`$, and the magnitude of the Sun’s velocity projected along the line of sight to Pyxis ($`v_{helio}=v_{GSR}v_{\mathrm{}LOS}`$), the globular cluster would have a large, positive heliocentric radial velocity if it were following a circular orbit in the plane defined by the pole at $`\widehat{P}`$.
It is the solar motion that dominates the radial velocity in the circular orbit case. For example, if we assume that Pyxis has a velocity that is approximately the circular velocity of the Galaxy at 40 kpc, or $``$200 km/sec, its space motion would then be $`(V_X,V_Y,V_Z)=(64,46,184)`$ km/sec. Since the Galactic Cartesian unit vector in the direction of Pyxis from the Sun (the line of sight) is $`(X,Y,Z)=(0.15,0.98,0.12)`$, the component of this space velocity along the line of sight<sup>3</sup><sup>3</sup>3Again the negative sign is in the radial system of the Sun, and it indicates that this velocity points towards the Sun. In the Galactic Cartesian system this velocity is positive in the $`X`$, $`Y`$, and $`Z`$ directions. would be $`(64\times 0.15)+(46\times 0.98)+(184\times 0.12)`$ or $`v_{GSR}=33`$ km/sec (the direction and magnitude of this component of the $`v_{GSR}`$ in the circular orbit case is shown as $`V_C`$ in Figure 3, while the direction and magnitude of the measured $`v_{GSR}`$ for Pyxis is shown as $`V_M`$). Including the component of the Sun’s velocity along the line of sight ($`227`$ km/sec), the heliocentric radial velocity ($`V_{Chelio}`$ in Figure 3) for Pyxis in this case would be $`33(227)=194`$ km/sec. Since this is much larger than the measured value (36 km/sec, $`V_{Mhelio}`$ in Figure 3), we conclude that Pyxis is not in an orbit that gives rise to a present space velocity near $`\stackrel{}{V}_{circ}`$ and, therefore, the strawman model of Pyxis being on a circular orbit and sharing the LMC orbital pole grossly fails expectations. Thus, one or both of the assumptions in the strawman model must be invalid: either Pyxis is not on a circular orbit or/and Pyxis’ orbit does not share a pole with the LMC.
Since the inferred radial velocity of the Pyxis globular cluster with respect to a stationary observer at the location of the Sun has a large magnitude, $`v_{GSR}191`$ km/sec, constraints can be placed on non-circular orbits Pyxis may follow that also share the plane and direction of rotation of the LMC’s orbit. We have calculated the orbital pole for the Pyxis globular cluster given all possible, realistic proper motions and accounting for our derived radial velocity. In Figure 4, we present the region in proper motion space that produces an orbital pole for the Pyxis globular cluster that is within $`15^{}`$ of the pole of the LMC’s orbit. We limit the possible proper motions to those that yield a space velocity less than the escape velocity from the Milky Way at the position of Pyxis ($``$415 km/sec in the Galactic model of Kochanek ) and obtain the shaded region in Figure 4. This is a prediction for the magnitude and direction of the proper motion of Pyxis with respect to the Sun assuming that the Milky Way capture from the Magellanic Clouds hypothesis is correct.
The proper motions in the shaded region of Figure 4 are those that can produce an orbital pole in the direction of $`\widehat{P}`$ given a $`v_{GSR}`$ of $`191`$ km/sec for Pyxis. We have calculated the shapes and energies of the orbits allowed for Pyxis to determine if a pole in the direction of $`\widehat{P}`$ is only likely for a very restricted range of conditions. For example, is the LMC capture origin for Pyxis only viable if Pyxis is following an extremely eccentric orbit? In fact, a range of orbits is possible for Pyxis given a proper motion in the shaded region in Figure 4. For a given elliptical orbit, the angle between the instantaneous velocity and radius vectors varies with position along the ellipse and that angle has a well defined minimum value for a given orbital eccentricity. For each space motion derived from our radial velocity and a proper motion from Figure 4, we have determined the angle between the velocity vector and the present Galactocentric radius vector for Pyxis. Assuming a closed, elliptical orbit, we can determine the lower limit for the eccentricity of the associated orbit having the given angle between the velocity and radius vectors at the present position of Pyxis. The orbits determined for Pyxis given our radial velocity and a proper motion in the shaded region in Figure 4 have eccentricities of $`e>0.70`$, with the peak of the distribution of all allowable eccentricities near $`e`$0.8.
We note here that few of the globular clusters with measured proper motions are following nearly circular orbits. Dinescu et al. (1999) has compiled all of the measured proper motions for a sample of 38 Galactic globular clusters and integrated orbits for each cluster. Figure 5 presents a histogram of the orbital eccentricities that Dinescu et al. (1999) calculated for the globular clusters in their sample. The open histogram in Figure 5 represents the data on the whole sample, while the hatched histogram represents the data on the 10 clusters with apoGalactica greater than 20 kpc. Less than half of the entire sample have eccentricities of $`e<0.5`$, and, more importantly, for the outer halo globular clusters the measured eccentricities are mostly found in the range $`0.6<e<0.8`$. Therefore, one might conclude that it is more likely than not that Pyxis is following an eccentric orbit and its space motion is not perpendicular to its current position. However, such a conclusion must be tempered with the acknowledgement of a potential selection bias for the latter subsample. The majority of the globular clusters with measured proper motions are those that are currently close to the Sun. Therefore, the clusters with large apoGalactica that have measured proper motions are, for the most part, currently near periGalacticon and thus must be following eccentric orbits. Thus, the sample of globular clusters that make up the hatched histogram in Figure 5 may be selected preferentially from the sample of outer halo globular clusters on eccentric orbits. Since few outer halo globular clusters near apoGalacticon have measured proper motions, the true distribution of eccentricities for outer halo globular clusters is unknown. However, the fact remains that a non-negligible (and perhaps dominant) fraction of the outer halo globular cluster population is orbiting the galaxy with eccentricities near 0.8, and since the majority of the outer halo globular clusters with orbits integrated by Dinescu et al. (1999) have eccentricities near $`e`$0.8, it is at least conceivable that the orbit of Pyxis has a similar eccentricity.
For our measured Pyxis radial velocity, we have integrated orbits for a grid of $`>1500`$ proper motions found in the shaded region in Figure 4 in the potential of Johnston et al. (1995) for 10 Gyr each. The orbital energy of Pyxis determined from the majority of the proper motions that produce an orbital pole at $`\widehat{P}`$ is within the $`1\sigma `$ error bars of the orbital energy of the LMC, although the error bar is large ($`E_{LMC}=2.1\pm 0.9\times 10^4`$ km<sup>2</sup>/sec<sup>2</sup>). However, Johnston (1998) found that in simulations of tidal stripping, debris was found within $`\pm 3\mathrm{\Delta }E`$ of the parent object (see eq. 1). All of the orbits produced from our measured radial velocity and a proper motion in the shaded region of Figure 4 are within $`\pm 3\mathrm{\Delta }E`$ of the orbital energy of the LMC. Since the $`\widehat{P}`$ orbits do not require extremely unlikely constraints on the eccentricity and since the orbital energies for these orbits are similar to expectations for debris from the LMC, the LMC capture origin for Pyxis remains viable.
## 4 Conclusions
It has been proposed since its discovery that the Pyxis globular cluster may have been captured by the Milky Way from the Magellanic Clouds. If the space motion for Pyxis were known, a comparison of the position of its orbital pole with respect to the LMC as well as a comparison of its angular momentum and orbital energy to that of the LMC would allow one to determine if the two objects share similar orbits. Although only one component of the space motion of Pyxis is now measured, some constraints can be placed on its possible orbit in the tidal capture scenario. A circular orbit with an orbital pole at $`(l,b)=(163,22)^{}`$ is completely ruled out by the measured radial velocity. However, we have shown here that the large radial velocity of Pyxis with respect to a stationary observer at the position of the Sun does not rule out the possibility that the cluster was captured from the LMC since a reasonable range of viable orbits with $`e`$0.8 exist for Pyxis that are also similar in energy and angular momentum to that of the LMC. No suitable first epoch plate material is known to exist for Pyxis, so an attempt to measure its proper motion to better determine the likelihood that Pyxis may be a captured LMC globular cluster will require precise observations with the HST or the Space Interferometry Mission (SIM).
Although proper motions are not available for the majority of the outer halo globular clusters, their spatial distribution has been used to argue that they are likely to have been accreted into the halo (e.g., Majewski 1994, Palma et al. 2000). Recently, Dinescu et al. (2000) have measured a proper motion for the young globular cluster Pal 12 and they find that its orbit is what one would expect if it had been captured from the Sagittarius dwarf galaxy. An accretion origin of the outer halo, second parameter horizontal branch globular clusters is often invoked to explain the possible younger age of some of these objects (where youth is inferred either from the second parameter effect itself or from relative age estimates determined from the cluster CMDs). The physical mechanism that causes the second parameter effect in globular clusters is still unknown: Although it is now generally agreed that there are indeed some globular clusters with anomalously young ages, age differences alone may not be enough to explain the second parameter effect. Whether or not the physical mechanism that causes the effect is age, the possibility that conditions somehow favor the formation of second parameter globular clusters preferentially in Milky Way satellite galaxies (which later get accreted by the Galaxy) may explain the source of the differences between second parameter and non-second parameter globular clusters.
The age measurement for the Pyxis globular cluster by Sarajedini & Geisler (1996), $`13.3\pm 1.3`$ Gyr, suggests that it is younger by $``$3 Gyr than the oldest Milky Way globular clusters when measured on the same age scale. Recently, age measurements for the oldest LMC globular clusters have been made (Olsen et al. (1998)) using a different technique than that used for Pyxis, but their average age of $`15.3\pm 1.5`$ Gyr places them similar in age to the oldest Milky Way globular clusters, when calibrated onto the same absolute age scale. Another study of a different sample of LMC clusters (Johnson, et al. (1999)) also finds the oldest LMC clusters to be as old as the old Milky Way clusters. Thus, we may conclude that typical LMC clusters are older than Pyxis. However, at least one of the clusters in the Olsen et al. (1998) sample is $``$2 Gyr younger than the others (NGC 1898), which makes it similar in age to Pyxis. Therefore, it is not impossible to place Pyxis in the “LMC family” of clusters from age arguments, though it does appear that Pyxis would be at the young end of the age range for old LMC clusters.
It may be noted, however, that the current orbital pole of the Small Magellanic Cloud (SMC) is also very near the intersection point of the poles of the LMC and Pyxis (see Figure 2). Since the SMC is more fragile due to its weaker gravitational potential, perhaps a more attractive origin for Pyxis is from stripping of the SMC rather than the LMC. Recent studies of SMC globular clusters have found that the SMC clusters show a range in ages (e.g., Shara et al. (1998), Mighell et al. (1998)) including at least one cluster with an age similar to Pyxis (NGC 121). The orbital energy of the SMC has a larger magnitude and a smaller error bar than that of the LMC, so not all of the orbits produced from a proper motion in the shaded region in Figure 4 have orbital energies similar to expectations of SMC debris. Only the orbits having proper motions found in the inner part of the shaded region, with a total magnitude of the proper motion of $``$0.75 mas/yr, have orbital energies consistent with an SMC capture origin. Since the same orbital energy and age arguments applied to support the LMC capture origin also apply to the SMC, we consider it a possibility that Pyxis may have been captured from either the LMC or the SMC.
We would like to thank Ata Sarajedini for providing us with an electronic version of his table of CCD photometry for Pyxis stars. We also appreciate a helpful conversation with Knut Olsen. We wish to thank the anonymous referee for useful comments that improved the manuscript. CP and SRM acknowledge support for this research from NSF CAREER Award grant AST-9702521, the David and Lucile Packard Foundation, and The Research Corporation.
## Appendix A Definition and Sign Conventions for $`V_{GSR}`$
The conversion of velocities among various reference frames is treated in the literature and in the standard texts, such as Mihalas & Binney (1981). Because we have found some confusing misuse of the standard terminology in the literature, we provide this detailed explanation of our sign and naming conventions.
The radial velocity that one measures for a star is the velocity of that object with respect to the Earth. Often, corrections are made to this velocity to remove the motions of the Earth and Sun, which reduces the measured radial velocity to a velocity with respect to some standard of rest. For example, the measured heliocentric radial velocity ($`v_{helio}`$) is reduced to the radial velocity with respect to the Local Standard of Rest ($`v_{LSR}`$) by removing the Sun’s peculiar velocity with respect to the LSR:
$$v_{LSR}=v_{helio}+[9\mathrm{cos}(b)\mathrm{cos}(l)+11\mathrm{cos}(b)\mathrm{sin}(l)+6\mathrm{sin}(b)]km/sec.$$
(A1)
This velocity can be further reduced to the Galactic Standard of Rest (GSR) by removing the Sun’s orbital velocity around the Galactic Center. So,
$$v_{GSR}=v_{LSR}+[220\mathrm{cos}(b)\mathrm{sin}(l)]km/sec.$$
(A2)
Referring to this velocity as “$`v_{GSR}`$” or a “Galactocentric” velocity apparently causes some confusion in the interpretation of velocity data. When reduced using the above two equations, the velocity referred to as $`v_{GSR}`$ is the velocity of the object as seen by a stationary observer at the position of the Sun. The direction of this velocity is along the line of sight between the object and the Sun and not along the line of sight between the object and the Galactic Center. The latter misinterpretation of “Galactocentric” velocity (as we have found in some articles in the literature) can lead to misleading or erroneous conclusions.
There is an additional ambiguity in the definition of $`v_{GSR}`$, and that is the sign convention. For a typical radial velocity, positive refers to a velocity that is moving away from the origin, and negative refers to a velocity that is approaching the origin. The origin for the “Galactocentric” radial velocity, or $`v_{GSR}`$, is the Sun and not the Galactic center. Therefore, the sign convention for $`v_{GSR}`$ is that a positive velocity indicates that the object is moving away from a stationary observer at the position of the Sun and a negative velocity indicates that the object is moving towards a stationary observer at the position of the Sun. The right hand sides of the two equations above are collectively the velocity of the Sun along the line of sight. This sign convention introduces additional confusion because the sign may not agree with the sign convention for the Cartesian Galactocentric $`(U,V,W)`$ system, and because the sign of the contribution of the Sun’s motion, $`v_{\mathrm{}LOS}`$ can seem counterintuitive.
For example, the Pyxis globular cluster has been measured to have a heliocentric radial velocity of $``$36 km/sec. Using the above equations, $`v_{GSR}=191`$ km/sec for Pyxis. The proper interpretation of this velocity is that Pyxis is approaching a stationary observer at the position of the Sun with a velocity magnitude of 191 km/sec. However, the Sun is located in Galactic Cartesian coordinates at $`(X,Y,Z)=(8,0,0)`$ and Pyxis is located at $`(X,Y,Z)=(13.9,38.6,4.8)`$. Therefore, the components of this velocity, $`v_{GSR}=191`$ km/sec, in Galactic Cartesian coordinates are positive in $`X`$, positive in $`Y`$, and negative in $`Z`$. Moreover, even though the Sun’s motion is increasing the separation of Pyxis from us (i.e., increasing the recessional velocity) $`v_{\mathrm{}LOS}`$ is negative.
Fig. 3. — A 3-D projection of the sky in Galactic Cartesian coordinates. The Sun is located at $`(X,Y,Z)=(8,0,0)`$ kpc and Pyxis lies at $`(13.9,38.6,4.8)`$ kpc. The shaded plane contains the line of sight connecting the Galactic Center to Pyxis and is perpendicular to the vector $`\widehat{P}`$, which is a unit vector that points in the direction of the orbital pole located at $`(l,b)=(163,22)^{}`$, a pole that Pyxis may share with the LMC. Any velocity vector for Pyxis that lies in the shaded plane will give Pyxis an orbital pole in the direction of $`\widehat{P}`$ (by definition). The circular motion orbital vector $`\widehat{v}_{circ}`$ is a unit vector that is mutually perpendicular to both $`\widehat{P}`$ and the Pyxis/Galactic Center line of sight and therefore lies in the shaded plane. If Pyxis is following a circular orbit with its pole at $`(163,22)^{}`$, then its space velocity should lie along $`\widehat{v}_{circ}`$. Assuming that the space velocity of Pyxis has a magnitude near the circular velocity of the Galaxy at 40 kpc, or $``$200 km/sec, the predicted radial velocity (as seen by a stationary observer at the location of the Sun) for Pyxis is $`V_c`$, or $`33`$ km/sec, which corresponds to a heliocentric radial velocity ($`V_{chelio}`$) of 194 km/sec. Given that the solar motion along the line of sight to Pyxis ($`V_{Sun}`$) is $`227`$ km/sec and that the measured heliocentric radial velocity of Pyxis ($`V_{mhelio}`$) is 36 km/sec, the measured radial velocity of Pyxis with respect to a stationary observer at the location of the Sun ($`V_m`$), is $`191`$ km/sec, ruling out a circular orbit for Pyxis with a pole in the direction of $`\widehat{P}`$. The unknown proper motion of Pyxis is only constrained to be perpendicular to $`V_m`$ (by definition). It is plausible that Pyxis may be following an eccentric orbit, as is the case for the majority of the Galactic globular cluster population, in which case the space velocity of Pyxis is unlikely to be perpendicular to its radius vector. There exists a set of proper motions for Pyxis (see Figure 4) that, when combined with $`V_m`$, produce a space motion with a pole at $`(163,22)^{}`$ and leave the cluster bound to the Milky Way.
|
warning/0005/cond-mat0005123.html
|
ar5iv
|
text
|
# Violation of general Friedel sum rule in mesoscopic systems
## Abstract
In the wake of a new kind of phase generally occurring in mesoscopic transport phenomena, we discuss the validity of Friedel sum rule in the presence of this phase. We find that the general Friedel sum rule may be violated.
PACS numbers: 73.23.-b; 73.23.Ad; 72.10.-d
With large scale research in Mesoscopic Physics over the last few decades, many of the well established notions of Condensed Matter Physics has been found to be violated in mesoscopic samples. Breakdown of Onsager reciprocity relation , violation of Ohms law , absence of material specific quantities like resistivity , violation of Hund’s rule etc., are a few such examples. The purpose of this work is to show the violation of Friedel sum rule in mesoscopic systems.
Friedel sum rule relates the density of states inside a fixed potential scatterer to the scattering phase shifts . A deduction of the sum rule can be found in many text books and intuitively it can be understood as follows. Consider for example a fixed spherically symmetric potential scatterer. Now we enclose it in a larger spherical volume. In an energy interval $`dE`$, the number of states depend on the number of times the specific boundary conditions can be fulfilled by the wave function of the electron. So when the energy is changed, it introduces a phase shift of the electron wave function and so changes the number of times the specific boundary condition can be satisfied. And hence the density of states $`\rho ^{}`$ inside the impurity is related to the scattering phase shift $`\eta `$ in the following fashion .
$$\frac{\eta }{E}=\pi \rho ^{}.$$
(1)
This can be extended to the partial wave analysis of scattering states and many important issues can be understood in terms of the Friedel sum rule . In case of a non-spherical scatterer or non-spherical Fermi surface, the scattering matrix is in general an NxN matrix. For any such general NxN scattering matrix $`S`$, the Friedel sum rule can be written as
$$\theta /E=\pi \rho ^{},$$
(2)
where $`\theta =\mathrm{\Sigma }_i^N\xi _i`$, $`exp[2i\xi _i]`$ being the eigenvalues of the scattering matrix $`S`$. This can be further written in a compact form as
$$\frac{1}{2i}\frac{}{E}(ln(det[S]))=\pi \rho ^{}.$$
(3)
For one-dimensional systems where the scattering matrix is 2x2, the Friedel sum rule was thought to be further simplified to give
$$\frac{arg(t)}{E}=\pi \rho ^{},$$
(4)
where $`t`$ is the transmission amplitude but this is not true.
Recently a new phase has been discussed in Ref. for scattering by a stub where the scattering matrix is 2x2, and it is believed that this phase is also observed in mesoscopic systems experimentally . This phase is a general feature of transmission zeroes that always occur in Fano resonances in Quantum Wires and Dots, the stub structure being the simplest example . It was shown in Ref. that the phase slips is a new phase associated with the violation of parity effect because it is different from Aharonov-Bohm phase, statistical phase and phase due to wave-like motion of electrons depending on their wave vector or energy. Had it not been different from the other three phases, parity effect would not have been violated . The specialty of this phase is that it is discontinuous as a function of energy, i.e., the phase of the wavefunction changes by $`\pi `$ although its energy does not change. To be more precise this phase does not originate from change in wave-vector due to change in energy. Hence in view of the discussions before Eq. 1 one can question the validity of Friedel sum rule in the presence of this phase . We shall give a pictorial description of this phase later (short-dashed and long-dashed curves in Fig. 1).
The scattering matrix for the stub is
$$S=\left(\begin{array}{cc}r& t\\ t& r\end{array}\right)$$
(5)
where $`r`$ and $`t`$ are reflection and transmission amplitudes across the stub and are
$$r=cos[kL]/(cos[kL]+2isin[kL])$$
(6)
and
$$t=(2isin[kL])/(cos[kL]2isin[kL]).$$
(7)
The eigenvalues of the S matrix are
$$(cos[kL]+2isin[kL])/(cos[kL]+2isin[kL])and1$$
(8)
Hence as defined in Eq. 2
$$\theta =\frac{1}{2}ArcTan[4cos[kL]sin[kL]/(cos[kL]^2+4sin[kL]^2)]$$
(9)
In Fig. 1 we plot $`\theta `$ (solid curve), $`arg(t)`$ (short-dashed curve) and $`arg(r)`$ (long-dashed curve) (given in Eqs. 6, 7 and 9) versus $`kL`$. It can be seen that $`arg(t)`$ and $`arg(r)`$ show discontinuous jumps and drops by $`\pi `$ but they cancel in such a way that $`\theta `$ is continuous and monotonously increasing. Hence one finds that the Friedel sum rule (Eqs. 2 and 3) is not violated although because of the discontinuous slips in $`arg(t)`$ Eq. 4 is obviously violated because density of states can never be infinite while the LHS of Eq. 4 can be infinite. And hence one can say that so far no one has found a violation of Friedel sum rule (Eqs. 2 and 3). We shall show the violation of the Friedel sum rule in the presence of this new phase.
Transport across the stub structure has acquired a lot of importance recently . All analysis so far are based on calculations with a hard wall boundary condition (an infinite step barrier potential or an infinite step well potential) at the dead end of the stub (we refer to it as the hard walled stub and for which Eqs. 6, 7, 8 and 9 are derived). An infinite potential well at the dead end of the stub reflects an incident electron with unit probability. Now a small perturbation from this would be a finite but very deep potential well at the dead end of the stub (soft walled stub). Electrons are almost entirely reflected from the end of the stub and a negligible fraction escapes. Dephasing can also give similar escape probability. The scattering problem in this case is depicted in Fig. 2 and also explained in the figure caption. It is solved using the mode matching technique or Griffith’s boundary conditions , that give the continuity of wavefunction and the conservation of currents at the junctions. In this case the transmission zero in x-direction is replaced by a minimum . We first intend to understand what happens to the discontinuous phase change that occur due to transmission zeroes in this case. So in Fig. 3 we plot transmission coefficient $`T=|t|^2`$ (solid curve) and the argument of the transmission amplitude $`t`$ (short-dashed curve) in x direction, versus $`kL`$ for an almost hard walled stub. The transmission coefficient shows very deep minima and at the same points $`arg(t)`$ show very sharp but continuous drops. For the completely hard walled stub there is an exact zero and associated with it a discontinuous slip by $`\pi `$ as shown in Fig. 1. In the same figure (Fig. 3) we also plot transmission coefficient in the x-direction (dash-dotted curve) and the corresponding argument of the transmission amplitude (long-dashed curve) versus $`kL`$ for a very soft walled stub. At the points where the solid curve show very deep minima, dash-dotted curve show shallow minima. Also the fast phase drops change over to a slower decrease.
Having understood the phase slips further we move on to the three prong scatterer (Fig. 4) that is often encountered in mesoscopic systems including the experimental set up of Ref. and many such similar experiments. The scattering problem in this case is described in the figure caption. From the continuity of wavefunctions (first Griffith’s boundary condition) we get the following equations (variables and parameters are defined in Fig. 4 and it’s caption).
$$1+r=aexp[iqL_1]+bexp[iqL_1];a+b=c+d;$$
$$a+b=f+g,cexp[iqL_2]+dexp[iqL_2]=e;$$
$$fexp[iqL_3]+gexp[iqL_3]=h.$$
(10)
And from the second Griffith’s boundary condition which is the conservation of currents at the junctions ($`\mathrm{\Sigma }_i\frac{d\psi _i}{dx_i}=0`$, that can be derived from current conservation, here $`\psi _i`$ is a wavefunction at a junction, $`x_i`$ is coordinate at that junction, and the sum over $`i`$ stands for all such wavefunctions incoming or outgoing at a junction, the convention followed is that currents flowing into the junction is positive while currents flowing out of the junction is negative) we get the following equations.
$$kkrqaexp[iqL_1]+qbexp[iqL_1]=0;$$
$$abc+df+g=0;$$
$$qcexp[iqL_2]qdexp[iqL_2]ke=0;$$
$$qfexp[iqL_3]qgexp[iqL_3]kh=0.$$
(11)
Thus we have 9 equations and exactly 9 unknown quantities ($`a,b,c,d,e,f,g,h`$ and $`r`$) and so the problem is completely defined. Once the unknowns are solved, the wavefunction is known at all points exactly and so the density of states as well as the scattering matrix can be calculated exactly. The scattering matrix in this case is
$$S=\left(\begin{array}{ccc}t_{11}& t_{12}& t_{13}\\ t_{21}& t_{22}& t_{23}\\ t_{31}& t_{32}& t_{33}\end{array}\right).$$
(12)
Here $`t_{11}`$=$`r`$=transmission amplitude to the first prong when the incident beam is from the first prong. $`t_{12}=e`$ is the transmission amplitude to the second prong when the incident beam is from the first prong. $`t_{13}=h`$ is the transmission amplitude to the third prong when the incident beam is from the first prong. The other matrix elements are to be calculated when similar incident beam of unit flux is from the other two directions in Fig. 4. For the case of Fig. 4, the partial density of states is given by the following expression
$$\rho _1=\pi \rho _1^{}=\frac{\pi }{hv}[_{L_1}^0|aexp[iqx]+bexp[iqx]|^2dx+$$
$$_0^{L_2}|cexp[iqy]+dexp[iqy]|^2𝑑y+$$
$$_0^{L_3}|fexp[iqz]+gexp[iqz]|^2dz],$$
(13)
where $`v=\mathrm{}k/m`$. $`a,b,c,d,f`$ and $`g`$ are determined from Eqs. 10 and 11. $`\rho _2`$ and $`\rho _3`$ are to be evaluated similarly when the incident beam is from the other two directions in Fig. 4, and $`\rho =\rho _1+\rho _2+\rho _3`$. For the symmetric three prong scatterer ($`L_1=L_2=L_3=L`$), the antiresonances are almost cancelled by the resonances but still a violation of Eq. 2 or 3 can be seen at low energy. This cancelling effect of resonance and antiresonance can be avoided by choosing incommensurate values of ($`L_1+L_3`$) and $`L_2`$, i.e., for asymmetric configurations. We will now go to the asymmetric configuration and demonstrate a large difference between $`\tau =\theta /E=\frac{1}{2i}\frac{}{E}[ln[Det[S]]`$ and $`\rho =\pi \rho ^{}`$ at large energies ($`EV`$). This is shown in Fig. 5. We want to emphasize that at very high energy, compared to the energy scale $`V`$ in the system, when multiple scattering and the new phase becomes insignificant, we recover Friedel sum rule perfectly. But when this new phase is present at energies ($`E<V`$), there is a large difference between $`\tau `$ and $`\rho `$ and hence a complete violation of Friedel sum rule. In Fig. 5, $`\tau `$ or $`\theta /E`$, can become substantially negative, i.e., $`\theta `$ can undergo a drop like $`arg(t)`$ in Fig. 3. The new phase need not always appear as a drop but can also appear as a rise and then the LHS of Eq. 3 can remain positive all the time while deviating from the RHS of Eq. 3. This is shown in Fig 6.
Thus our exact calculation of density of states and scattering matrix elements show the deviation of $`\frac{1}{2i}\frac{}{E}ln[Det[s]]`$ from $`\pi \rho ^{}`$ in the presence of phase slips. The phase slips are a general feature of Quantum wires with defects and Quantum Dots and these phase slips are at the origin of drops in $`\theta `$ and hence deviation or violation of Friedel sum rule. Only 2x2 S-matrix is a special case where as shown in Fig. 1 some scattering matrix elements undergo a phase jump and some undergo a phase drop in such a manner that they cancel and the phase slips do not affect $`ln[Det[S]]`$ or $`\theta `$. The general feature is that they do not cancel and Friedel sum rule gets violated. The attractive potential in the three prong scatterer offsets the symmetry between the phase jumps and the phase drops so that they do not cancel each other.
The author acknowledges useful discussions with Prof. M. Manninen.
Figure captions
Fig. 1 Arg(r) (long-dashed curve), arg(t) (short-dashed curve) and $`\theta `$ (solid curve) for the hard walled stub. Length of the stub is $`L`$ and it is taken to be the unit of length. We choose $`\mathrm{}=2m=1`$.
Fig. 2 A scattering problem with conventional notations is depicted here. $`k=\sqrt{E}`$ is the wave vector in the thin regions where the Quantum Mechanical potential is 0. $`q=\sqrt{E+V}`$ is the wave vector in the thick regions where the Quantum Mechanical potential is $`V`$. x and y are coordinates and the origin of coordinates is also depicted in the figure. t and c are transmission amplitudes in x and y directions, respectively, while r is the reflection amplitude. Distance between points P and Q is L.
Fig. 3 The solid curve is transmission coefficient T =$`|t|^2`$ across the soft walled stub described in Fig. 2. The short-dashed curve is the phase of the transmission amplitude $`t`$ across the stub. We choose $`VL^2=10^6`$ so that it is in the hard wall limit, and $`\mathrm{}=2m=1`$. Next we make $`VL^2`$=-100 and plot the transmission coefficient T in dash-dotted curve. The phase of the transmission amplitude t is given by long-dashed curve.
Fig. 4 A scattering problem with conventional notations is depicted here. $`k=\sqrt{E}`$ is the wave vector in the thin regions where the Quantum Mechanical potential is 0. $`q=\sqrt{E+V}`$ is the wave vector in the thick regions where the Quantum Mechanical potential is $`V`$. x,y,z,u,v and w are coordinates and the origin of coordinates is also depicted in the figure. e and h are transmission amplitudes in respective directions, while r is the reflection amplitude. Distance between (u=0) and (x=0,y=0,z=0) is $`L_1`$. Distance between (v=0) and (x=0,y=0,z=0) is $`L_2`$. Distance between (w=0) and (x=0,y=0,z=0) is $`L_3`$.
Fig. 5 $`\rho `$ (solid curve) and $`\tau =\frac{d\theta }{dE}`$ =LHS of Eq. 3 (dotted curve) versus $`kL`$ for the scattering problem described in Fig. 4. We choose $`VL^2=100`$, $`L_1=L_3=L`$, $`L_2`$=4$`L`$ and $`\mathrm{}=2m=1`$.
Fig. 6 $`\rho `$ (solid curve) and $`\tau =\frac{d\theta }{dE}`$ =LHS of Eq. 3 (dotted curve) versus $`kL`$ for the scattering problem described in Fig. 4. We choose $`VL^2=100`$, $`L_1=L_3=L`$, $`L_2`$=2.4$`L`$ and $`\mathrm{}=2m=1`$.
|
warning/0005/astro-ph0005526.html
|
ar5iv
|
text
|
# A large age for the pulsar B1757–24 from an upper limit on its proper motion
## Acknowledgements
We thank Vicky Kaspi and Deepto Chakrabarty for useful discussions, Namir Kassim for supplying 90 cm data on G5.4–1.2 and Andrew Lyne for providing timing data on PSR B1757–24. The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. B.M.G. acknowledges the support of NASA through a Hubble Fellowship awarded by the Space Telescope Science Institute.
Correspondence should be addressed to B.M.G. (e-mail: bmg@space.mit.edu).
|
warning/0005/nucl-th0005007.html
|
ar5iv
|
text
|
# A DYNAMICAL MODEL FOR THE RESONANT MULTIPOLES AND THE Δ STRUCTURE
## 1 Introduction
The study of excitations of the hadrons can shed light on the nonperturbative aspects of QCD. One case which has recently been under intensive study is the electromagnetic excitation of the $`\mathrm{\Delta }(1232)`$ resonance. At low four-momentum transfer squared $`Q^2`$, the interest is motivated by the possibility of observing a $`D`$ state in the $`\mathrm{\Delta }`$. The existence of a $`D`$ state in the $`\mathrm{\Delta }`$ has the consequence that the $`\mathrm{\Delta }`$ is deformed and the photon can excite a nucleon through electric $`E2`$ and Coulomb $`C2`$ quadrupole transitions. In a symmetric SU(6) quark model, the electromagnetic excitation of the $`\mathrm{\Delta }`$ could proceed only via magnetic $`M1`$ transition. In pion electroproduction, $`E2`$ and $`C2`$ excitations would give rise to nonvanishing $`E_{1+}^{(3/2)}`$ and $`S_{1+}^{(3/2)}`$ multipole amplitudes. Recent experiments give nonvanishing ratio $`R_{EM}=E_{1+}^{(3/2)}/M_{1+}^{(3/2)}0.03`$ at $`Q^2=0`$ which has been widely taken as an indication of the $`\mathrm{\Delta }`$ deformation.
At sufficiently large $`Q^2`$, the perturbative QCD (pQCD) is expected to work. It predicts that only helicity-conserving amplitudes contribute at high $`Q^2`$, leading to $`R_{EM}=E_{1+}^{(3/2)}/M_{1+}^{(3/2)}1`$ and $`R_{SM}=S_{1+}^{(3/2)}/M_{1+}^{(3/2)}const`$. This behavior in the perturbative domain is very different from that in the nonperturbative one. It is an intriguing question to find the region of $`Q^2`$ which signals the onset of the pQCD.
In a recent measurement , the electromagnetic excitation of the $`\mathrm{\Delta }`$ was studied at $`Q^2=2.8`$ and $`4.0GeV^2`$ via reaction $`p(e,e^{}p)\pi ^0`$. The extracted ratios $`R_{EM}`$ and $`R_{SM}`$ remain small and negative. This disagrees with the previous analysis of the earlier DESY data which gave small but positive $`R_{EM}`$ and $`R_{SM}`$ at $`Q^2=3.2GeV^2`$, though both analyses indicate that pQCD is still not applicable in this region of $`Q^2`$. In this talk, we want to show that the recent data of Ref. 2 can be understood from the dominance of the pion cloud contribution at low $`Q^2`$ in both $`E_{1+}^{(3/2)}`$ and $`S_{1+}^{(3/2)}`$ multipoles, as predicted by a dynamical model for electromagnetic production of pion, together with a simple scaling assumption for the bare $`\gamma ^{}N\mathrm{\Delta }`$ form factors.
## 2 Dynamical Model for $`\gamma ^{}N\pi N`$
The main feature of dynamical approach to pion photo- and electro-production is that the unitarity is built in by explicitly including the final state $`\pi N`$ interaction in the theory, namely, t-matrix is expressed as
$`t_{\gamma \pi }(E)=v_{\gamma \pi }+v_{\gamma \pi }g_0(E)t_{\pi N}(E),`$ (1)
where $`v_{\gamma \pi }`$ is the transition potential operator for $`\gamma ^{}N\pi N`$, and $`t_{\pi N}`$ and $`g_0`$ denote the $`\pi N`$ t-matrix and free propagator, respectively, with $`EW`$ the total energy in the CM frame.
Multipole decomposition of Eq. (1) gives the physical amplitude in channel $`\alpha `$
$`t_{\gamma \pi }^{(\alpha )}(q_E,k;`$ $`E`$ $`+iϵ)=\mathrm{exp}(i\delta ^{(\alpha )})\mathrm{cos}\delta ^{(\alpha )}`$ (2)
$`\times `$ $`\left[v_{\gamma \pi }^{(\alpha )}(q_E,k)+P{\displaystyle _0^{\mathrm{}}}𝑑q^{}{\displaystyle \frac{q^2R_{\pi N}^{(\alpha )}(q_E,q^{};E)v_{\gamma \pi }^{(\alpha )}(q^{},k)}{EE_{\pi N}(q^{})}}\right],`$
where $`\delta ^{(\alpha )}`$ and $`R^{(\alpha )}`$ are the $`\pi N`$ scattering phase shift and reaction matrix in channel $`\alpha `$, respectively; $`q_E`$ is the pion on-shell momentum and $`k=|𝐤|`$ is the photon momentum.
The multipole amplitude in Eq. (2) manifestly satisfies the Watson theorem and shows that $`\gamma \pi `$ multipoles depend on the half-off-shell behavior of $`\pi N`$ interaction. We remark that the use of K-matrix unitarization scheme as employed in, e.g., Ref. 7 would amount to approximating Eq. (2) with
$`t_{\gamma \pi }^{(\alpha )}(q_E,k;E+iϵ)=\mathrm{exp}(i\delta ^{(\alpha )})\mathrm{cos}\delta ^{(\alpha )}v_{\gamma \pi }^{(\alpha )}(q_E,k).`$ (3)
The difference between Eqs. (2) and (3) lies in the fact that only the on-shell rescatterings are included in the K-matrix unitarization scheme.
In the resonant (3,3) channel in which $`\mathrm{\Delta }(1232)`$ plays a dominant role, the transition potential $`v_{\gamma \pi }`$ consists of two terms
$`v_{\gamma \pi }(E)=v_{\gamma \pi }^B+v_{\gamma \pi }^\mathrm{\Delta }(E),`$ (4)
where $`v_{\gamma \pi }^B`$ is the background transition potential which includes Born terms and vector mesons exchange contributions, as described in Ref. 8. The second term of Eq. (4) corresponds to the contribution of bare $`\mathrm{\Delta }`$.
With Eq. (4), we may decompose Eq. (1) in the following way
$`t_{\gamma \pi }(E)=t_{\gamma \pi }^B+t_{\gamma \pi }^\mathrm{\Delta }(E),`$ (5)
where
$`t_{\gamma \pi }^B(E)`$ $`=`$ $`v_{\gamma \pi }^B+v_{\gamma \pi }^Bg_0(E)t_{\pi N}(E),`$ (6)
$`t_{\gamma \pi }^\mathrm{\Delta }(E)`$ $`=`$ $`v_{\gamma \pi }^\mathrm{\Delta }+v_{\gamma \pi }^\mathrm{\Delta }g_0(E)t_{\pi N}(E).`$ (7)
Here $`t_{\gamma \pi }^B`$ includes contributions from the nonresonant background and renormalization on the vertex $`\gamma ^{}N\mathrm{\Delta }`$ due to $`\pi N`$ scattering. The advantage of such a decomposition is that all the processes which start with the electromagnetic excitation of the bare $`\mathrm{\Delta }`$ are summed up in $`t_{\gamma \pi }^\mathrm{\Delta }`$.
Multipole decomposition of Eq. (6) takes the same form as Eq. (2) and is used to calculate the multipole amplitudes $`M_{1+}^B(W,Q^2),E_{1+}^B(W,Q^2)`$ and $`S_{1+}^B(W,Q^2)`$ with $`R_{\pi N}^{(\alpha )}(q_E,q^{};E)`$ obtained from a meson exchange model for $`\pi N`$ interaction. Note that to make the principal value integration in Eq. (2) associated with $`v_{\gamma \pi }^B`$ convergent, we introduce an off-shell dipole form factor with cut-off parameter $`\mathrm{\Lambda }`$=440 MeV. The gauge invariance, violated due to the off-shell rescattering effects, is restored by the substitution $`J_\mu ^BJ_\mu ^Bk_\mu kJ^B/k^2`$, where $`J_\mu ^B`$ is the electromagnetic current corresponding to the background contribution $`v_{\gamma \pi }^B`$.
## 3 $`\gamma ^{}N\mathrm{\Delta }`$ transition form factors
Now let us consider the $`\mathrm{\Delta }`$ resonance contribution $`t_{\gamma \pi }^\mathrm{\Delta }`$ in Eq. (5). In keeping with the standard way of experimental analysis and constituent quark model (CQM) calculations, we describe the resonant multipoles $`t_{\gamma \pi }^{\mathrm{\Delta },\alpha }`$ with a Breit-Wigner type of energy dependence, as was done in the isobar model of Ref. 8,
$$t_{\gamma \pi }^{\mathrm{\Delta },\alpha }(W,Q^2)=\overline{𝒜}_\alpha ^\mathrm{\Delta }(Q^2)\frac{f_{\gamma \mathrm{\Delta }}\mathrm{\Gamma }_\mathrm{\Delta }M_\mathrm{\Delta }f_{\pi \mathrm{\Delta }}}{M_\mathrm{\Delta }^2W^2iM_\mathrm{\Delta }\mathrm{\Gamma }_\mathrm{\Delta }}e^{i\varphi },$$
(8)
where $`f_{\pi \mathrm{\Delta }}(W)`$ is the usual Breit-Wigner factor describing the decay of the $`\mathrm{\Delta }`$ resonance with total width $`\mathrm{\Gamma }_\mathrm{\Delta }(W)`$ and physical mass $`M_\mathrm{\Delta }`$=1232 MeV. The expressions for $`f_{\gamma \mathrm{\Delta }},f_{\pi \mathrm{\Delta }}`$ and $`\mathrm{\Gamma }_\mathrm{\Delta }`$ are taken from Ref. 8. The phase $`\varphi (W)`$ in Eq. (8) is to adjust the phase of $`t_{\gamma \pi }^{\mathrm{\Delta },\alpha }`$ to be equal to the corresponding pion-nucleon scattering phase $`\delta ^{(33)}`$. Note that at the resonance energy $`\varphi (M_\mathrm{\Delta })=0`$.
The main parameters in the bare $`\gamma ^{}N\mathrm{\Delta }`$ vertex are the $`\overline{𝒜}^\mathrm{\Delta }(Q^2)`$’s in Eq. (8). For the magnetic dipole $`\overline{}^\mathrm{\Delta }`$ and electric quadrupole $`\overline{}^\mathrm{\Delta }`$ transitions, they are related to the conventional electromagnetic helicity amplitudes $`A_{1/2}^\mathrm{\Delta }`$ and $`A_{3/2}^\mathrm{\Delta }`$ by
$`\overline{}^\mathrm{\Delta }(Q^2)={\displaystyle \frac{1}{2}}(A_{1/2}^\mathrm{\Delta }+\sqrt{3}A_{3/2}^\mathrm{\Delta }),\overline{}^\mathrm{\Delta }(Q^2)={\displaystyle \frac{1}{2}}(A_{1/2}^\mathrm{\Delta }+{\displaystyle \frac{1}{\sqrt{3}}}A_{3/2}^\mathrm{\Delta }).`$ (9)
At the photon point $`Q^2=0`$, the bare amplitudes $`\overline{}^\mathrm{\Delta }(0)`$ and $`\overline{}^\mathrm{\Delta }(0)`$ of Eq. (8) are determined from the best fit to the results of the recent analyses of Mainz (open circles) and VPI group (full circles), as shown in Fig. 1. The dashed curves denote the contribution from $`t_{\gamma \pi }^B`$ only. The dotted curves correspond to the K-matrix approximation to $`t_{\gamma \pi }^B`$, namely, without the inclusion of principal value integral term. Solid curves are the full results of our calculation with bare $`\mathrm{\Delta }`$ excitation.
The numerical values for $`\overline{}^\mathrm{\Delta }`$ and $`\overline{}^\mathrm{\Delta }`$ and the helicity amplitudes, at $`Q^2=0`$, are given in Table 1. Here we also give ”dressed” values obtained using K-matrix approximation for $`t_{\gamma \pi }^B`$. One notices that the bare values determined above for the helicity amplitudes amount to only about $`60\%`$ of the corresponding dressed values and are close to the predictions of the CQM, as pointed out by Sato and Lee . The large reduction of the helicity amplitudes from the dressed to the bares ones results from the fact that the principal value integral part of $`t_{\gamma \pi }^B`$, which represents the effects of the off-shell pion rescattering, contributes approximately half of the $`M_{1+}`$ as indicated by the dashed curves in Fig. 1.
We now turn to the $`Q^2`$ evolution of the multipoles $`\overline{}^\mathrm{\Delta }(Q^2)`$ and $`\overline{}^\mathrm{\Delta }(Q^2)`$. In the present work, we parametrize the $`Q^2`$ dependence of the dominant $`\overline{}^\mathrm{\Delta }`$ amplitude by
$`\overline{}^\mathrm{\Delta }(Q^2)=\overline{}(0){\displaystyle \frac{𝐤}{k_\mathrm{\Delta }}}(1+\beta Q^2)e^{\gamma Q^2}G_D(Q^2),`$ (10)
where $`G_D`$ is the nucleon dipole form factor. For the small $`\overline{}^\mathrm{\Delta }`$ and $`\overline{𝒮}^\mathrm{\Delta }`$ amplitudes, we follow Ref. 8 and assume that they have the same $`Q^2`$ dependence as $`\overline{}^\mathrm{\Delta }`$ (scaling assumption). This is motivated by the scaling law which has been observed for the nucleon form factors.
We remind the reader that, in contrast to Ref. 8, amplitudes $`\overline{}_{1+}`$ and $`\overline{}_{1+}`$ and the corresponding helicity amplitudes in Eq. (9) correspond to the ”bare” $`\gamma N\mathrm{\Delta }`$ transition. For the real photon, they are equal to the standard $`M1`$ and $`E2`$ amplitudes of the $`\mathrm{\Delta }N\gamma `$ transition defined in accordance with the convention of the Particle Data Groups. At the resonance energy, they can be easily expressed in terms of the Dirac-type form factors $`g_1`$ and $`g_2`$ used in Ref. 7, or Sachs-type form factors $`G_M`$ and $`G_E`$ used in Ref. 12. The relation between $`\overline{}_{1+}`$ amplitude and the bare $`G_M`$ form factors is as follows
$$\overline{}_{1+}(Q^2)=\frac{e}{2m}\frac{𝐤}{k_\mathrm{\Delta }}\sqrt{\frac{k_\mathrm{\Delta }M_\mathrm{\Delta }}{m}}G_M(Q^2),$$
(11)
where $`k_\mathrm{\Delta }=(M_\mathrm{\Delta }^2m^2)/2M_\mathrm{\Delta }`$ with $`m`$ and $`M_\mathrm{\Delta }`$ denoting the nucleon and $`\mathrm{\Delta }`$ mass, respectively. Expression for the electric amplitude is similar, but with opposite sign. Relation between physical $`M_{1+}^{(3/2)}`$ multipole and experimentally measured $`G_M^{}`$ form factor is given by Eq. (24) of Ref. 8. Note that we employ the ”Ash” definition for the $`G_M^{}`$ which differs from the $`\mathrm{\Delta }`$ form factor used in Ref. 2 by a factor $`f=\sqrt{1+Q^2/(m+M_\mathrm{\Delta })^2}`$, i.e., $`G_M^{}(our)=G_M^{}(`$Ref. 2)$`/f`$.
## 4 Results and Discussion
Using the $`\beta `$ and $`\gamma `$ in Eq. (10) as free parameters, we fit the recent experimental data as well as old one quoted in Ref. 8 on the $`Q^2`$ dependence of the $`M_{1+}^{(3/2)}`$ or equivalently, the $`G_M^{}`$ form factor. Our result is shown in Fig. 2. The obtained values for the $`\beta `$ and $`\gamma `$ parameters are: $`\beta =0.44GeV^2`$ and $`\gamma =0.38GeV^2`$. Here the dashed curve corresponds to contribution from the bare $`\mathrm{\Delta }`$, i.e., $`t_{\gamma \pi }^\mathrm{\Delta }`$ of Eq. (7).
The results for the ratios $`R_{EM}=E_{1+}^{(3/2)}/M_{1+}^{(3/2)}`$ and $`R_{SM}=S_{1+}^{(3/2)}/M_{1+}^{(3/2)}`$ are shown in the right column of Fig. 2. It is seen that they are in good agreement with the results of the model independent analysis of Ref. 2 up to $`Q^2`$ as high as $`4.0GeV^2`$. Note that since the bare values for the E2 and C2 excitations are small, the absolute values and shape of these ratios are determined, to a large extent, by the pion rescattering contribution. The bare $`\mathrm{\Delta }`$ excitation contributes mostly to the $`M_{1+}^{(3/2)}`$ multipole.
Pion cloud has been found to play an important role in hadron structure in many studies. For example, in the cloudy bag model (CBM), a reasonably good agreement with the measured $`R_{EM}`$ can be obtained with bag radius of $`R=0.60.8fm.`$ In a recent improved CBM calculation , where the relativistic effects and CM motion were better treated, it was also found that the pion cloud gives large contribution to $`G_M^{}(Q^2)`$. It is generally concluded that the smaller the bag radius, the larger the pion cloud contribution. Similar conclusion is reached with respect to the proton EM form factors within the cloudy bag model. In fact, it has been found that the recent Jlab data on the scaling violation in the electromagnetic form factors can be explained within CBM with $`R=0.7fm.`$ One might then be tempted to interpret our result as another indication of preferring a smaller bag radius. However, at these small bag radii, the pion field is so strong that the use of the perturbative approach employed in these CBM studies is questionable. In this connection, we should mention that in a nonperturbative calculation within a chiral chromodielectric model and a linear $`\sigma `$-model, Fiolhais, Golli, and Sirca reached a similar conclusion as ours, namely, the large experimental values of $`R_{EM}`$ and $`R_{SM}`$ could be explained in terms of the pion contribution alone.
In nonrelativistic CQM it is well known that, the values of $`R_{EM}`$ and $`R_{SM}`$ obtained with a $`D`$state admixture in the $`\mathrm{\Delta }`$ generated by one-gluon-exchange hyperfine interaction are in general too small. It has recently been suggested by the Tübigen group that this problem can be fixed with the inclusion of exchange currents in the calculation. The reason is that two-body exchange currents could flip the spins of the two quarks in the nucleon to convert it into a $`\mathrm{\Delta }`$. Such a transition would just be a transition between the S-states in the nucleon and $`\mathrm{\Delta }`$, and accordingly is greatly enhanced. We would like to point out here that the diagram involving a two-body exchange current induced by the exchange of a pion between two quarks in the nucleon can also be interpreted as pion rescattering effects, as considered in the current study.
## 5 Summary
In summary, we calculate the $`Q^2`$ dependence of the ratios $`E_{1+}/M_{1+}`$ and $`S_{1+}/M_{1+}`$ in the $`\gamma ^{}N\mathrm{\Delta }`$ transition, with the use of a dynamical model for electromagnetic production of pions. We find that both ratios $`E_{1+}/M_{1+}`$ and $`S_{1+}/M_{1+}`$ remain small and negative for $`Q^24.0GeV^2`$. Our results agree well with the recent measurement of Frolov et al. , but deviate strongly from the predictions of pQCD. Our results indicate that the bare $`\mathrm{\Delta }`$ is almost spherical and hence very difficult to be directly excited via electric E2 and Coulomb C2 quadrupole excitations. The experimentally observed $`E_{1+}^{(3/2)}`$ and $`S_{1+}^{(3/2)}`$ multipoles are, to a very large extent, saturated by the contribution from pion cloud, i.e., pion rescattering effects. It remains an intriguing question, both experimentally and theoretically, to find the region of $`Q^2`$ which will signal the onset of pQCD.
## Acknowledgments
This work is supported in part by the NSC/ROC under the grant no. NSC 88-2112-M002-015.
## References
|
warning/0005/hep-th0005061.html
|
ar5iv
|
text
|
# Plane Wave Limits and T-Duality
## Appendix
Our conventions are as follows: In all $`D2`$ we use the “mostly minus” signature $`(+,,\mathrm{},)`$ and the orientation $`ϵ_{012\mathrm{}D1}=1`$. The Ricci tensor is defined as $`R_{\mu \nu }=R_{}^{\lambda }{}_{\mu \nu \lambda }{}^{}`$ and the Riemann curvature obeys $`(_\nu _\mu _\mu _\nu )T_\kappa =R_{}^{\lambda }{}_{\kappa \mu \nu }{}^{}T_\lambda `$ for an arbitrary $`T_\mu `$. The Hodge dual of a p-form $`(pD)`$ is defined by
$`(V^{\alpha _1}\mathrm{}V^{\alpha _p})={\displaystyle \frac{(1)^{(D1)}}{(Dp)!}}ϵ^{\alpha _1\mathrm{}\alpha _p\alpha _{p+1}\mathrm{}\alpha _D}V_{\alpha _{p+1}}\mathrm{}V_{\alpha _D},`$
in terms of an orthonormal basis $`\{V^\alpha \}`$.
## Acknowledgements
I thank The Abdus Salam International Centre for Theoretical Physics for the warm hospitality. The research reported in this paper has been supported in part by the Bogaziçi University Foundation (BÜVAK) and the Turkish Academy of Sciences(TÜBA).
|
warning/0005/gr-qc0005099.html
|
ar5iv
|
text
|
# Hermitian extension of the four-dimensional Hooke’s law
## 1. Introduction
It is well known that A. Einstein spent the last decade of his life in the search of a non-symmetric extension of his symmetric, Riemannian theory of 1915 ,. The very idea of such a generalisation dates back to 1925 , when Einstein first tried to unify the description of gravitation and of electromagnetism through an extension of the Riemannian geometry built from a nonsymmetric fundamental tensor and a nonsymmetric affine connection. But relaxing the symmetry of the geometrical objects that enter the equations of general relativity means an unwelcome arbitrariness in the choice of the generalisation. In Einstein’s words :
> The introduction of non-symmetric fields meets with the following difficulty. If $`\mathrm{\Gamma }_{ik}^l`$ is a displacement field, so is $`\stackrel{~}{\mathrm{\Gamma }}_{ik}^l(=\mathrm{\Gamma }_{ki}^l)`$. If $`g_{ik}`$ is a tensor, so is $`\stackrel{~}{g}_{ik}(=g_{ki})`$. This leads to a large number of covariant formations among which it is not possible to make a selection on the principle of relativity alone.
The way out from such an arbitrariness was envisaged by Einstein in the requirement of “transposition invariance” as a generalisation of the principle of symmetry. A tensorial expression built with the fundamental form $`g_{ik}`$ and with the affine connection $`\mathrm{\Gamma }_{ik}^l`$ is said to be transposition invariant with respect to the pair of indices, say, $`p`$ and $`q`$, if it is transformed into itself when one simultaneously substitutes $`\stackrel{~}{g}_{ik}`$ for $`g_{ik}`$, $`\stackrel{~}{\mathrm{\Gamma }}_{ik}^l`$ for $`\mathrm{\Gamma }_{ik}^l`$ and then interchanges the indices $`p`$ and $`q`$. The requirement of transposition invariance played a key rôle in obtaining what Einstein’s considered the logically most satisfactory solution of his problem, i.e. the field equations of the metric-affine theory :
(1.1) $`g_{ik,l}g_{nk}\mathrm{\Gamma }_{il}^ng_{in}\mathrm{\Gamma }_{lk}^n=0,`$
(1.2) $`\mathrm{\Gamma }_{[is]}^s=0,`$
(1.3) $`R_{(ik)}=0,`$
(1.4) $`R_{[ik],l}+R_{[kl],i}+R_{[li],k}=0;`$
whether a given expression be symmetric or antisymmetric with respect to a pair of indices, say $`p`$ and $`q`$, is henceforth denoted by enclosing the mentioned indices within respectively round or square brackets; $`R_{ik}`$ means the usual Ricci tensor
(1.5)
$$R_{ik}=\mathrm{\Gamma }_{ik,a}^a\mathrm{\Gamma }_{ia,k}^a\mathrm{\Gamma }_{ib}^a\mathrm{\Gamma }_{ak}^b+\mathrm{\Gamma }_{ik}^a\mathrm{\Gamma }_{ab}^b$$
built with the nonsymmetric connection $`\mathrm{\Gamma }_{ik}^l`$.
Einstein’s theory of the nonsymmetric field came in two versions, according to whether the fundamental tensor $`g_{ik}`$ was chosen to be real nonsymmetric or complex Hermitian. Much effort was done by many authors in order to grasp the physical meaning of the proposed field equations; they looked for solutions either by exact or by approximate methods that mimicked the ones used in general relativity. All these efforts, however, either implicity or explicitly allowed for singularities at the right-hand side of one or another of the equations (1.1)-(1.4), while Einstein believed that, at variance with what occurs with his theory of 1915, the new theory had to be considered complete, and only everywhere regular solutions could disclose its physical meaning. Therefore, by agreeing with Einstein’s conviction, one can assert that no conclusion has been drawn up to now about the validity of the nonsymmetric theory; this is one of the unsolved problem left as a challenge to the skillness of future mathematicians.
In the meantime one may well explore whether the very concept of invariance under transposition, that Einstein found so helpful in choosing his non Riemannian geometry, can be of heuristic value in some down to earth instance, dealing with some well known chapters of classical physics. Without attempting to solve difficult equations, one can content himself with a preliminary, modest task: seeking whether up to now unrelated physical entities, whose mathematical representation happens to require respectively symmetric and antisymmetric tensors, can be given an at least formally unified description in terms of either nonsymmetric or Hermitian tensors.
## 2. The four-dimensional Hooke’s law as starting point
By availing of Cartesian coordinates and of the three-dimensional tensor formalism, Hooke’s law can be written as:
(2.1)
$$\mathrm{\Theta }^{\lambda \mu }=\frac{1}{2}C^{\lambda \mu \rho \sigma }(\xi _{\rho ,\sigma }+\xi _{\sigma ,\rho }),$$
where $`\mathrm{\Theta }^{\lambda \mu }`$ is the three-dimensional tensor that defines the stresses arising in elastic matter due to its displacement, given by the three-vector $`\xi ^\rho `$, from a supposedly relaxed condition, and $`C^{\lambda \mu \rho \sigma }`$ is the constitutive tensor whose build depends on the material features and on the symmetry properties of the elastic medium. It has been shown that this law admits of a natural generalisation to the four-dimensions of the Riemannian spacetime, whose metric tensor be $`g_{ik}`$. From a formal standpoint, this extension is straightforward: one introduces a real four-vector field $`\xi ^i`$, that aims at representing some four-dimensional displacement, and builds the four-dimensional, symmetric deformation tensor
(2.2)
$$S_{ik}=\frac{1}{2}(\xi _{i;k}+\xi _{k;i}),$$
where the semicolon indicates covariant differentiation performed with the Christoffel symbols calculated from $`g_{ik}`$. A four-dimensional stiffness tensor density $`𝐂^{iklm}`$ is then introduced; it will be real and symmetric with respect to both the first and the second pair of indices, since it will be used for producing the real symmetric stress-momentum-energy tensor density
(2.3)
$$𝐓^{ik}=𝐂^{iklm}S_{lm},$$
through the generalisation of equation (2.1) to the four dimensions of spacetime. This generalisation happens to make physical sense, since it allows encompassing both inertia and elasticity in a sort of four-dimensional elasticity . Let us consider a coordinate system such that, at a given event, $`g_{ik}`$ reduce to the form, say:
(2.4)
$$g_{ik}=\eta _{ik}diag(1,1,1,1),$$
while the Christoffel symbols are vanishing, and the components of the four-velocity of matter are:
(2.5)
$$u^1=u^2=u^3=0,u^4=1.$$
We imagine that in such a coordinate system we are able to measure, at the chosen event, the three components of the (supposedly small) spatial displacement of matter from its relaxed condition, and that we adopt these three numbers as the values taken by $`\xi ^\rho `$ in that coordinate system, while the reading of some clock ticking the proper time and travelling with the medium will provide the value of the “temporal displacement” $`\xi ^4`$ in the same coordinate system. By applying this procedure to all the events of the manifold where matter is present and by reducing the collected data to a common, arbitrary coordinate system, we can define the vector field $`\xi ^i(x^k)`$. From such a field we shall require that, when matter is not subjected to ordinary strain and is looked at in a local rest frame belonging to the ones defined above, it will exhibit a deformation tensor $`S_{ik}`$ such that its only nonzero component will be $`S_{44}=\xi _{4,4}=1`$. This requirement is met if we define the four-velocity of matter through the equation
(2.6)
$$\xi _{;k}^iu^k=u^i.$$
The latter definition holds provided that
(2.7)
$$det(\xi _{;k}^i\delta _k^i)=0,$$
and this shall be one equation that the field $`\xi ^i`$ must satisfy; in this way the number of independent components of $`\xi ^i`$ will be reduced to three. A four-dimensional stiffness tensor $`C^{iklm}`$ endowed with physical meaning can be built as follows. We assume that in a locally Minkowskian rest frame the only nonvanishing components of $`C^{iklm}`$ are: $`C^{\lambda \nu \sigma \tau }`$, with the meaning of ordinary elastic moduli, and
(2.8)
$$C^{4444}=\rho ,$$
where $`\rho `$ measures the rest density of matter. We need defining the four-dimensional stiffness tensor in an arbitrary co-ordinate system; this task is easily accomplished if the unstrained matter is isotropic when looked at in a locally Minkowskian rest frame. Let us define the auxiliary metric
(2.9)
$$\gamma ^{ik}=g^{ik}+u^iu^k;$$
then the part of $`C^{iklm}`$ accounting for the ordinary elasticity of the isotropic medium can be written as
(2.10)
$$C_{\mathrm{el}}^{iklm}=\lambda \gamma ^{ik}\gamma ^{lm}\mu (\gamma ^{il}\gamma ^{km}+\gamma ^{im}\gamma ^{kl}),$$
where $`\lambda `$ and $`\mu `$ are assumed to be constants. The part of $`C^{iklm}`$ that accounts for the inertia of matter shall read instead
(2.11)
$$C_{\mathrm{in}}^{iklm}=\rho u^iu^ku^lu^m.$$
The elastic part $`T_{\mathrm{el}}^{ik}`$ of the energy tensor is orthogonal to the four-velocity, as it should be ; thanks to equation (2.6) it reduces to
$`T_{\mathrm{el}}^{ik}=C_{\mathrm{el}}^{iklm}S_{lm}=\lambda (g^{ik}+u^iu^k)(\xi _{;m}^m1)`$
(2.12) $`\mu [\xi ^{i;k}+\xi ^{k;i}+u_l(u^i\xi ^{l;k}+u^k\xi ^{l;i})],`$
while, again thanks to equation (2.6), the inertial part of the energy tensor proves to be effectively so, since
(2.13)
$$T_{\mathrm{in}}^{ik}=C_{\mathrm{in}}^{iklm}S_{lm}=\rho u^iu^k.$$
The energy tensor defined by summing the contributions (2) and (2.13) encompasses both the inertial and the elastic energy tensor of an isotropic medium; when the macroscopic electromagnetic field is vanishing it should represent the overall energy tensor, whose covariant divergence must vanish according to Einstein’s equations , :
(2.14)
$$T_{;k}^{ik}=0.$$
Imposing the latter condition allows one to write the equations of motion for isotropic matter subjected only to elastic strain . We show this outcome in the limiting case when the metric is everywhere flat and the four-velocity of matter is such that $`u^\rho `$ can be dealt with as a first order infinitesimal quantity, while $`u^4`$ differs from unity at most for a second order infinitesimal quantity. Also the spatial components $`\xi ^\rho `$ of the displacement vector and their derivatives are supposed to be infinitesimal at first order. An easy calculation then shows that equation (2.7) is satisfied to the required first order, and that equations (2.14) reduce to the three equations of motion:
(2.15)
$$\rho \xi _{,4,4}^\nu =\lambda \xi _{,\rho }^{\rho ,\nu }+\mu (\xi ^{\nu ,\rho }+\xi ^{\rho ,\nu })_{,\rho },$$
and to the conservation equation
(2.16)
$$\{\rho u^4u^k\}_{,k}=0,$$
i.e., to the required order, they come to coincide with the well known equations of the classical theory of elasticity for an isotropic medium.
## 3. Hermitian extension of the four-dimensional Hooke’s law
The extension of Hooke’s law outlined in the previous Section has been a fruitful move, since it has allowed a unified account of inertia and of elasticity. But there is another aspect that deserves attention: having enlarged Hooke’s law to the four dimensions of spacetime opens the way to this new question: is it useful, i.e., does it make physical sense to look for either a nonsymmetric or a Hermitian extension of the four-dimensional Hooke’s law?
Let us try the Hermitian version, and see how it can be formulated in a Riemannian spacetime whose metric, given a priori, is the real, symmetric tensor $`g_{ik}`$. We introduce a complex “displacement” four-vector
(3.1)
$$\omega ^i\xi ^i+i\phi ^i;$$
the real four-vectors $`\xi ^i`$ and $`\phi ^i`$ enter respectively the real and the imaginary part of $`\omega ^i`$. By closely following the pattern of the real case, in lieu of (2.2) we introduce the Hermitian “deformation” tensor
(3.2)
$$S_{ik}=\frac{1}{2}(\omega _{i;k}^{}+\omega _{k;i});$$
the asterisk is henceforth used to denote complex conjugation. $`S_{ik}`$ splits into a real, symmetric part $`S_{(ik)}`$, that can still be interpreted as a genuine deformation tensor:
(3.3)
$$S_{(ik)}=\frac{1}{2}(\xi _{i;k}+\xi _{k;i})$$
and into an antisymmetric, purely imaginary contribution:
(3.4)
$$S_{[ik]}=\frac{i}{2}(\phi _{k,i}\phi _{i,k})$$
that immediately reminds us of Helmholtz’ seminal attempt at producing a hydrodinamic simile of magnetism , and of the subsequent extension of the idea to the four dimensions of the theory of relativity. We assume that $`\phi _i`$ shall play the rôle of the electromagnetic four-potential and, starting from this apparently promising beginning, we try to figure out what the idea of a Hermitian generalisation of Hooke’s law can entail from a physical standpoint.
Instead of the stiffness tensor density of equation (2.3), that is symmetric in both the first and the second pair of indices, one shall confront a complex “stiffness” tensor density $`𝐂^{iklm}`$, that will be Hermitian with respect to both the first and the second pair:
(3.5)
$$𝐂^{kilm}=𝐂^{iklm}=𝐂^{ikml}.$$
This stiffness tensor density shall be contracted with the Hermitian deformation tensor $`S_{ik}`$ to produce the Hermitian tensor density
(3.6)
$$𝐓^{ik}=𝐂^{iklm}S_{lm}.$$
One calls for now $`𝐓^{ik}`$ the Hermitian stress-momentum-energy tensor density, and asks that it fulfil the generalised “conservation” equation
(3.7)
$$𝐓_{;k}^{ik}=0,$$
which of course entails severally:
(3.8)
$$𝐓_{;k}^{(ik)}=0,$$
and
(3.9)
$$𝐓_{,k}^{[ik]}=0.$$
A survey of the tensorial parts into which (3.6) can be split and expanded in keeping with the symmetry properties is now useful. The real, symmetric part of (3.6) can be written as:
(3.10)
$$𝐓^{(ik)}=𝐂^{(ik)(lm)}S_{(lm)}+𝐂^{(ik)[lm]}S_{[lm]}.$$
The interpretation of the first term at the right-hand side has been already provided, since $`S_{(lm)}`$ is just the four-dimensional, symmetric deformation tensor of Section 2; as recalled there, if $`S_{(lm)}`$ is contracted with the appropriate $`𝐂^{(ik)(lm)}`$, it can produce the material part of the symmetric energy tensor density, that accounts for both the inertia and the elasticity of matter. Let us call it $`𝐓_\mathrm{m}^{(ik)}`$.
Due to the presence of $`S_{[lm]}`$, the second term at the right-hand side of (3.10) awaits an electromagnetic interpretation. One tentatively sets
(3.11)
$$F_{ik}2iS_{[ik]},$$
where $`F_{ik}`$ is the antisymmetric tensor used in electromagnetism, according to a long established convention, to encompass both the electric field $`\stackrel{}{E}`$ and the magnetic induction $`\stackrel{}{B}`$. A symmetric energy tensor density $`𝐓_{\mathrm{em}}^{(ik)}`$ for the electromagnetic field in matter, like e.g. Abraham’s tensor density , can obviously be recast in the form<sup>1</sup><sup>1</sup>1It may be objected that this way of writing the electromagnetic energy tensor is artificial, since the four-potential $`\phi _i`$ enters not only $`S_{[lm]}`$, but also the “stiffness” factor $`𝐂^{(ik)[lm]}`$. However the previous Section showed that a similar occurrence already happens with the displacement vector $`\xi _i`$ in $`𝐓_\mathrm{m}^{(ik)}`$ as soon as one abandons the non-relativistic regime.
(3.12)
$$𝐓_{\mathrm{em}}^{(ik)}=𝐂^{(ik)[lm]}S_{[lm]};$$
therefore the overall $`𝐓^{(ik)}`$ is apt to summarise the energy tensor density of elastic matter and of the electromagnetic field. Equating to zero the four-divergence of this density, as done in (3.8), would produce four equations for the motion of possibly charged elastic matter under the influence of an electromagnetic field.
We consider now the imaginary, antisymmetric part of $`𝐓^{ik}`$, that can be written as
(3.13)
$$𝐓^{[ik]}=𝐂^{[ik](lm)}S_{(lm)}+𝐂^{[ik][lm]}S_{[lm]}.$$
The second term at the right-hand side of this equation naturally fits the physical picture already emerged from the analysis of $`𝐓^{(ik)}`$. In fact, since $`2iS_{[ik]}`$ has already been identified in (3.11) as representing the electric field $`\stackrel{}{E}`$ and the magnetic induction $`\stackrel{}{B}`$, one is led to pose
(3.14)
$$𝐇^{ik}i𝐓_\mathrm{f}^{[ik]}i𝐂^{[ik][lm]}S_{[lm]},$$
i.e. to read off the second term at the right-hand side of (3.13) the so called constitutive equation of electromagnetism . This equation defines the electric displacement $`\stackrel{}{D}`$ and of the magnetic field $`\stackrel{}{H}`$, summarised by the four-tensor density $`𝐇^{ik}`$, in terms of $`\stackrel{}{E}`$, $`\stackrel{}{B}`$, and of whatever fields may enter $`𝐂^{[ik][lm]}`$. For definiteness, let us remind an example of the constitutive equation, valid when matter is homogeneous and isotropic in its local rest frame :
(3.15)
$$\mu H^{ik}=\left[g^{il}(ϵ\mu 1)u^iu^l\right]\left[g^{km}(ϵ\mu 1)u^ku^m\right]F_{lm};$$
here $`ϵ`$ is the dielectric constant and $`\mu `$ is the magnetic permeability. This constitutive equation has just the form (3.14).
If only the second term $`𝐂^{[ik][lm]}S_{[lm]}`$ where present at the right-hand side of (3.13), the imaginary part (3.9) of the “conservation” equation would entail $`𝐇_{,k}^{ik}=0`$, i.e. the electromagnetic field would be sourceless, and our description of matter would be defective. The Hermitian extension of the four-dimensional Hooke’s law however provides a solution to this problem through the first term at the right-hand side of (3.13). Let us define:
(3.16)
$$𝐏^{ik}i𝐓_{\mathrm{ch}}^{[ik]}i𝐂^{[ik](lm)}S_{(lm)},𝐣^i𝐏_{,k}^{ik},$$
where $`𝐣^i`$ is a conserved quantity, since it is the divergence of an antisymmetric tensor density:
(3.17)
$$𝐣_{,i}^i=0.$$
Equation (3.9) can be rewritten as
(3.18)
$$𝐓_{\mathrm{ch},k}^{[ik]}+𝐓_{\mathrm{f},k}^{[ik]}=0,$$
and, due to the definitions (3.14) and (3.16), it entails:
(3.19)
$$𝐇_{,k}^{ik}=𝐣^i,$$
i.e. the validity of the inhomogeneous Maxwell’s equation.
## 4. Charge, like matter, exists since time elapses
We have completed the attribution of a tentative physical meaning to the four tensorial terms into which the Hermitian $`𝐓^{ik}`$ can be split according to the symmetry properties. Three of them, namely the material contribution $`𝐓_\mathrm{m}^{(ik)}`$, the electromagnetic energy tensor $`𝐓_{\mathrm{em}}^{(ik)}`$ and the term that provides for the constitutive equation of electromagnetism, are entities known since a long time. Their mathematical form and their physical meaning have been carefully investigated by generations of scholars. The very existence of the fourth one, $`𝐓_{\mathrm{ch}}^{[ik]}`$, is predicted by the Hermitian extension of the four-dimensional Hooke’s law. It is certainly welcome from a physical standpoint. Its very build, however, would be a surprise, had we not already met in Section 2 with an occurrence that constitutes its symmetric counterpart. We have seen there that the extension of Hooke’s law to the four dimensions of spacetime allows one to account, besides the ordinary elasticity, also for the very existence of inertial matter. This happens thanks to the completion of the displacement vector with a fourth component which, as shown in Section 2, has the meaning of displacement in time. In the same way we must expect that the term $`𝐂^{[ik](lm)}S_{(lm)}`$ shall account, besides the phenomenon of ordinary piezoelectricity, also for the very existence of the electric charge and current in unstrained matter. Through the generalised deformation tensor $`S_{(lm)}`$ the relativistic and Hermitian extension of Hooke’s law relates the presence of both matter and charge to the lapsing of time.
## 5. Perspectives
One still knows nothing about the explicit expression of $`𝐂^{[ik](lm)}`$, and it is fully premature to address here the enormous task of providing models that best suit the manifold properties displayed by electricity in macroscopic matter. Of course one shall proceed in this undertaking by adhering to the pattern already followed with the other three terms that compose $`𝐂^{iklm}`$, i.e. one shall try to build this fourth term by availing only of the two four-vectors $`\xi _i`$, $`\phi _i`$, of the scalar density $`\rho `$ and of the metric tensor $`g_{ik}`$ This is possible, since (2.6) allows expressing the four-velocity $`u^i`$ through $`\xi _i`$ and $`g_{ik}`$. The postulate of Hermitian symmetry will restrict the choice of the possible forms. In the present formulation $`g_{ik}`$ is an entity prescribed from the outside, and we know from the condition (2.7) that $`\xi _i`$ has only three independent components. Pending a detailed scrutiny of the Cauchy problem, we can notice that the complex equation $`𝐓_{;k}^{ik}=0`$ imposes eight conditions on the eight independent variables of our problem. The model of electrified, elastic matter provided by the Hermitian extension of the four-dimensional Hooke’s law has therefore a chance to stand up as a complete model, in which the values of all the quantities accounting for the physical behaviour of both matter and the electromagnetic field can be determined at least in principle by solving, with given initial conditions, the equations of motion stemming from the Hermitian “conservation equation”.
|
warning/0005/hep-th0005156.html
|
ar5iv
|
text
|
# Matrix Models: Fermion Doubling vs. Anomaly
## 1 Introduction
A tool for nonperturbative study of superstring theory (M-theory) proved to be matrix models, in special BFSS (M(atrix) theory) and IKKT (IIB matrix theory) ones, . These models are formulated in terms of $`N\times N`$ Hermitian matrices, where $`N\mathrm{}`$.
For a finite (but large) $`N`$ the last model, plays the rôle of regularised “second quantised” IIB superstring in Schild formulation , while the former one is a regularisation of $`D=11`$ quantum super-membrane.
Even for finite $`N`$ these models have problems with definiteness due to divergence of the partition function arising from integration along flat directions or vacua. Due to this also the limit $`N\mathrm{}`$ is bad defined. However, if not to try to consider the matrix model as a whole, but to consider a “perturbative sector” connected with fluctuations around a particular vacuum solution one may have a well defined system.
The models obtained this way depend on the chosen vacuum solution, in particular, various compactifications of IIB matrix model on noncommutative tori were shown to yield super Yang–Mills (sYM) models on these tori, as well as BFSS model, .
One can show that in fact Yang–Mills type models describe all regular fluctuations around a vacuum solution of IKKT matrix model, . Under regular fluctuations we understand those which have bounded value of momentum and position operators. In practice e.g. numerical computations this restriction can be implemented by additional regularisation which makes the undesirable modes decouple.
Once there is a close correspondence between matrix models and the sYM models there appears the following problem. Ten dimensional sYM model is known to be anomalous and the anomaly seems to exist also in the noncommutative case, . From the other hand there is no visible source for anomaly in the matrix model. For any finite $`N`$, the matrix model is finite manifestly gauge and Lorentz invariant as well as supersymmetric. Naively, these properties must hold also in the limit $`N\mathrm{}`$, contradicting the anomaly of $`D=10`$ sYM.
The explanation of this discrepancy may reside just in the contribution of the singular configurations. In the actual paper we show that the matrix model fails to reproduce the anomaly in the limit $`N\mathrm{}`$ due to the spectrum doubling in the fermionic sector of the matrix model.
In fact the actual situation is not new. A class of models is known to suffer from the spectrum doubling, . In particular, for lattice gauge theory, it is known that the the naive discretisation of chiral fermionic action leads to doubling of the fermionic spectrum. The last results in appearance for each mode another mode(s) carrying the opposite chirality which compensate the parity odd contribution of fermions, . In fact, there is a no-go theorem due to Nielsen and Ninomiya , which states that the doubling cannot be avoided unless the gauge, Lorentz or other relevant symmetry is destroyed in the continuum limit. While the gauge symmetry plays the crucial rôle in the consistency of the model, breaking of the Lorentz symmetry is not dangerous for the model and maybe even desirable as a mechanism of spontaneous breaking of the ten dimensional Lorentz group to a lower dimensional one, .
The plan of the paper is as follows. First we introduce the description of the fermionic sector of IIB matrix model for finite $`N`$, after that we explicitly find the doubling states for free fermionic fluctuations (quadratic approximation), and discuss the issue for the interacting case as well as possibility to eliminate the doubling.
## 2 Finite $`N`$ Matrix Model
The IIB matrix model is described by the action,
$$S=\frac{1}{g^2}\mathrm{t}r\left(\frac{1}{4}[A_\mu ,A_\nu ]^2+\frac{1}{2}\overline{\psi }\mathrm{\Gamma }^\mu [A_\mu ,\psi ]\psi \right),$$
(1)
where $`A_\mu `$ and $`\psi `$ are ten dimensional vector and Majorana–Weyl spinor Hermitian $`N\times N`$ matrices.
An important class of configurations which tend to satisfy equations of motion in the limit $`N\mathrm{}`$, are BPS ones given by $`A_\mu =p_\mu `$, where Hermitian matrices $`p_\mu =p_\mu ^{}`$ satisfy,
$$[p_\mu ,p_\nu ]=\mathrm{i}B_{\mu \nu },$$
(2)
here $`B_{\mu \nu }`$ is proportional to the unity matrix $`B_{\mu \nu }B_{\mu \nu }𝕀`$.
Although, such a set of matrices does not exist for finite $`N`$, it can be approximated by a sequence of matrices converging to (2) in the sense of operator norm on the Hilbert space of smooth finite functions (vectors), .
Before searching such an approximation, consider a Lorentz transformation which brings matrix $`B_{\mu \nu }`$ to the canonical form, having $`2\times 2`$ antisymmetric diagonal blocks with $`\pm \mathrm{}_i`$ entries. In this case the set of matrices $`p_\mu `$ is split in pairs $`(p_i,q^i)`$, $`i=1,\mathrm{},D/2`$, where $`p_i`$ and $`q^i`$ are canonically conjugate,
$`[p_i,p_j]=[q^i,q^j]=0,`$ (3)
$`[p_i,q^j]=\mathrm{i}\mathrm{}_i\delta _i^j.`$ (4)
Eqs. (3,4) give the $`D/2`$ dimensional Heisenberg algebra. It is known that the Heisenberg algebra can be represented e.g. in terms of square integrable functions defined on on the spectrum of operators $`q^i`$.
Finite $`N`$ analog of the Heisenberg algebra (4) is given by the position and the Hermitian (symmetric) lattice derivative operator on a compact rectangular periodic lattice $`\mathrm{\Gamma }`$. Let $`|n`$ be eigenvector of $`q^i`$ with eigenvalues $`L_i\mathrm{sin}(2\pi /N_i)`$, $`_iN_i=N`$,
$$q^i|n=L_i\mathrm{sin}(2\pi n^i/N_i)|n,$$
(5)
then operators $`p_i`$ act on this basis as follows,
$$p_i|n=\frac{\mathrm{i}\mathrm{}_i}{2a_i}\left(|n+e_i|ne_i\right),$$
(6)
where $`e_i`$ is the unity lattice vector along $`i`$-th link, and $`a_i=\frac{2\pi L_i}{N_i}`$. The choice of $`q`$ which we use differs from one usually used in lattice model by redefinition of $`qL\mathrm{sin}2\pi q/L`$. It is as good as the former one, but beyond this it treats $`p`$ and $`q`$ in a symmetric manner. It is not difficult to see that in the basis of $`p`$ eigenvectors eqs. (5) and (6) keep the same form with $`p`$ and $`q`$ interchanged. It also introduces explicitly the periodic boundary conditions, i.e. the identification $`|n+N_ie_i|n`$. Due to the symmetry between $`p`$ and $`q`$ the continuum limit is achieved after the “UV cutoff” removing: $`a_i0`$, combined with the “IR cutoff” removing or “decompactification”: $`L_i\mathrm{}`$.
The commutator of $`p_i`$ and $`q^j`$ looks as follows,
$$\begin{array}{c}[p_i,q^j]|n=\mathrm{i}\mathrm{}\delta _i^j\times \hfill \\ \hfill \frac{N}{4\pi }(\{\mathrm{sin}\frac{2\pi n^j}{N_j}\mathrm{sin}\frac{2\pi (n^j+1)}{N_j}\}|n+e_i\\ \hfill +\left\{\mathrm{sin}\frac{2\pi (n^j1)}{N_j}\right\}|ne_i)\\ \hfill \mathrm{i}\delta _i^j\mathrm{}_j𝕀_{(j)}(N)|n,\end{array}$$
(7)
where we have introduced the notation $`𝕀_{(j)}(N)`$ for the following matrix,
$$\begin{array}{c}𝕀_{(j)}(N)|n=\frac{N_j}{4\pi }(\{\mathrm{sin}\frac{2\pi n^j}{N_j}\mathrm{sin}\frac{2\pi (n^j+1)}{N_j}\}|n+e_i\hfill \\ \hfill +\left\{\mathrm{sin}\frac{2\pi (n^j1)}{N_j}\right\}|ne_i).\end{array}$$
(8)
As it is not difficult to see the equations of motion are not satisfied by such background. It still can be shown that the $`N=\mathrm{}`$ solution (2) *can not* be approximated by *solutions* to equations of motion at a *finite* value of $`N`$, because at finite $`N`$ the only solutions are those with zero commutator, $`[A_\mu ,A_\nu ]=0`$.
For (sequences of) vectors $`|f=_nf_n|n`$, on which $`p`$ and $`q`$ remain bounded as $`N`$ goes to infinity,
$$f|(p^2+q^2)|fC,$$
(9)
where $`C`$ does not depend neither on $`N`$ nor on $`L`$ (or $`a`$), operator $`𝕀_{(i)}(N)`$ approaches the unity one,
$$\begin{array}{c}𝕀_{(j)}(N)|f\hfill \\ \hfill \underset{n}{}\frac{1}{2}\left(f_{n+e_i}\mathrm{cos}\frac{2\pi (n^j1)}{N_j}+f_{ne_i}\mathrm{cos}\frac{2\pi (n^j+1)}{N_j}\right)|n|f,\end{array}$$
(10)
since in this case $`n^jN_j`$, and $`f_{n\pm e_i}f_n=O(N^1)`$.
The last equation means that the operators preserving the property (9) tend to commute with $`𝕀_{(j)}(N)`$ as $`N`$ approaches the infinity.
## 3 Fermionic Contribution
We are ready now to proceed to the analysis of the fermionic contribution to the partition function of the model with action (1). Integration over fermionic matrices results in the Pfaffian of the fermionic operator,
$$Z(A)=\psi \mathrm{}^{S_\mathrm{f}},$$
(11)
where, $`S_\mathrm{f}`$ stands for the fermionic part of the action (1).
For finite $`N`$ consider the bosonic background given by matrices $`p_i,q^i`$ from eqs. (5,6). An arbitrary Hermitian matrix fluctuation around the given background is $`A_\mu =p_\mu +ga_\mu `$. The fermionic part $`S_\mathrm{f}`$ of the action in this case looks as follows,
$$S_\mathrm{f}=\frac{1}{2}\mathrm{t}r\overline{\psi }\mathrm{\Gamma }^\mu [(p_\mu +ga_\mu ),\psi ],$$
(12)
where we rescaled $`\psi g\psi `$.
The free ($`a_\mu =0`$) part of the fermionic action can be written in the representation of $`|n`$. It looks as follows,
$$\begin{array}{c}S_\mathrm{f}=\underset{m,n,i}{}((\mathrm{i}\mathrm{}_i/(2a))\overline{\psi }_{n,m}\mathrm{\Gamma }^i(\psi _{m+e_i,n}\psi _{me_i,n}\psi _{m,n+e_i}+\psi _{m,ne_i})\hfill \\ \hfill +L_i\overline{\psi }_{n,m}\overline{\mathrm{\Gamma }}_i(\mathrm{sin}\frac{2\pi m^i}{N_i}\mathrm{sin}\frac{2\pi n^i}{N_i})\psi _{m,n}),\end{array}$$
(13)
where,
$`\psi _{n,m}=n|\psi |m,`$ (14)
$`\overline{\psi }_{m,n}=m|\overline{\psi }|n,`$ (15)
and $`\mathrm{\Gamma }^i`$, $`\overline{\mathrm{\Gamma }}_i`$ are Dirac matrices,
$$\mathrm{\Gamma }^ip_i+\overline{\mathrm{\Gamma }}_iq^i\mathrm{\Gamma }^\mu p_\mu .$$
(16)
Although, the action (13) differs from the naive lattice fermionic action, it shares many common features with it.
As for naive lattice fermions, they are described by the action ,
$$S_{\mathrm{naive}}=\frac{\mathrm{i}}{2}\underset{n}{}\overline{\psi }_n\mathrm{\Gamma }^\mu (\psi _{n+e_i}\psi _{ne_i}),$$
(17)
where for shortening notations we put lattice spacing $`a`$ to unity.
A well-known fact is that the actual model suffers from the fermionic spectrum doubling. The last manifests in the fact that for each chiral fermionic state there is always another one present in the spectrum with the opposite chirality but with other quantum numbers coinciding with the original state. This phenomenon can be described by introducing some discrete symmetry which relates these states. The states obtained by action of this symmetry are called doublers. It is clear that if such a symmetry exists it completely destroys the chiral asymmetry.<sup>1</sup><sup>1</sup>1The absence of an explicit symmetry of such kind, however, does not prove necessarily, the absence of doubling.
In the case of action (17), such a symmetry indeed exists and its generators in a even dimension $`D`$ look as follows ,
$$T_\alpha =\mathrm{i}\mathrm{\Gamma }_{(D+1)}\mathrm{\Gamma }_\alpha (1)^{n_\alpha },$$
(18)
where $`\mathrm{\Gamma }_{(D+1)}`$ is the $`D`$-dimensional analog of the Dirac $`\gamma _5`$-matrix ($`D`$ is even), $`\mathrm{\Gamma }_{(D+1)}=ϵ_D\mathrm{\Gamma }^1\mathrm{\Gamma }^2\mathrm{}\mathrm{\Gamma }^D`$, $`ϵ`$ is chosen to be either $`\mathrm{i}`$ or $`1`$ in order to make $`\mathrm{\Gamma }_{(D+1)}`$ Hermitian.
Finding the order of the discrete group generated by $`T_\alpha `$, one finds that the number of doubling states is $`2^D1`$.
Now, let us return back to to the matrix model given by eq. (13) and try to find a similar symmetry in this case.
One can check that the action (13) is left invariant by the following symmetry,
$$\psi _{m,n}\mathrm{i}\mathrm{\Gamma }_{11}\mathrm{\Gamma }_i(1)^{n^im^i}\psi _{m,n}.$$
(19)
or in the matrix form,
$$\psi T_i\psi =\mathrm{i}\mathrm{\Gamma }_{11}\mathrm{\Gamma }_iU_i^1\psi U_i,$$
$`(\text{19}^{})`$
where unitary matrix $`U_i`$ is given by $`U_i=(1)^{\frac{N_i}{2\pi }\mathrm{arcsin}(q^i/L_i)}`$.
Indeed, factor $`\mathrm{\Gamma }_{11}\mathrm{\Gamma }_i`$ commute with all $`\overline{\mathrm{\Gamma }}_i`$ and $`\mathrm{\Gamma }^j`$ for $`ji`$ while factors $`(1)^{n^im^i}`$ are the same for both $`\psi `$ and $`\overline{\psi }`$, and, therefore cancel. In the remaining term $`\mathrm{\Gamma }_{11}\mathrm{\Gamma }_i`$ anticommutes with $`\mathrm{\Gamma }_i`$, but the extra minus sign is compensated by the variation of the factor $`(1)^{n^im^i}`$. Thus, all terms in the action (13) remains invariant under the transformation (19).
Interchange $`pq`$ gives the remaining symmetries,
$$\psi \overline{T}_i\psi =\mathrm{i}\mathrm{\Gamma }_{11}\overline{\mathrm{\Gamma }}_i\overline{U}_i^1\psi \overline{U}_i,$$
(20)
where $`\overline{U}_i`$ is the unitary transformation, $`\overline{U}_i=(1)^{\frac{N_i}{2\pi }\mathrm{arcsin}(pa_i/\mathrm{}_i)}`$.
Summarising one has the action (1) invariant with respect to discrete symmetry generated by $`T_\mu `$,
$$T_\mu \psi =\mathrm{i}\mathrm{\Gamma }_{11}\mathrm{\Gamma }_\mu U_\mu ^1\psi U_\mu ,$$
(21)
where $`U_\mu `$ satisfy,
$$U_\mu ^1p_\nu U_\mu =(12\delta _{\mu \nu })p_\nu .$$
(22)
As in the case with naive lattice fermions these transformations act in such a way that in the continuum limit the states become $`2^D`$-fold degenerate with half of that for each chirality. In particular eq. (20) means that for each matrix state of given chirality which connects $`m`$ and $`n`$ there are $`2^{D/2}`$ states of different chiralities connecting $`\frac{N_i}{2}e_im`$ with $`\frac{N_i}{2}e_in`$, $`\frac{N_i}{2}e_i+\frac{N_j}{2}e_j+m`$ with $`\frac{N_i}{2}e_i+\frac{N_j}{2}e_j+n`$, $`ij`$, and so on. Eq. (19) have the same interpretation in the “momentum space” spanned by the eigenvectors of $`p`$.
It is clear now that if one wants to compute the gauge anomaly, one will have contributions from doublers of both chiralities which cancel each other.
So far, we considered the interaction free part of the fermionic action. Presence of interaction at least in the framework of perturbation theory does not change the situation, in this case one has in the continuum limit an interacting $`2^D`$-plet instead of a free one. Let us note, that this analysis may not remain true beyond the perturbation theory, as for strong field $`a_\mu `$ the interaction part dominates and the symmetry (21) of the free part does not play in this case such an important role. Unlike the usual lattice models the study of doubling in nonperturbative regime is too complicate, but the perturbative considerations are enough to doubt the result of naive continuum limit.
One may ask, what happens in this case to the supersymmetry of the original matrix model given by the action (1)?
In fact, in spite of the fermionic doubling, the supersymmetry still exists as it is valid for any matrix configuration and is irrelevant to the chosen representation. The apparent paradox with doubled number of fermions is solved if one see that the number of bosonic “degrees of freedom” is also doubled. The bosonic doublers are gauge equivalent configurations and are related by gauge transformations $`a_\mu U_\nu ^1a_\mu U_\nu `$.
## 4 Doubling States Removing
The results of the previous section show that matrix model (1) fails to reproduce a chiral continuum model in the limit $`N\mathrm{}`$. Thus, in particular, one can not obtain from it the noncommutative SYM model in this limit.
Such a situation can be interpreted either as a presence of finite $`N`$ artefact which does not decouple in the limit $`N\mathrm{}`$, and must be removed by additional effort like in traditional lattice models, or as an indication that the model possesses nontrivial symmetries. In traditional lattice models the doubling states should be removed since the doubling contradict the continuum “phenomenology”, in special the chirality properties of the model.
In the case of matrix models which pretend to describe M-theory the situation is different. As it is known the M-theory unites perturbative models with different field content. In particular IIA models contain states with both chiralities while in IIB models there are only ones with definite chirality. The duality symmetry which must relate these models should contain a mechanism which flips the chirality of states. As we see, such a mechanism exists on compact noncommutative spaces and is given by doubling.
In spite of this perspective for doubling states to describe the physical reality there exists, however, possibility to remove them in order to get a chiral model in the continuum limit. Consider briefly the ways one can do this. As the problem is a “lattice” one, we can look for specific lattice solutions. In the lattice case the doubling is cured by addition of a Wilson term to the naive lattice action . The problem is that there is no gauge invariant Wilson term which could be added to our “naive” fermionic action (12), as there is no gauge and Lorentz invariant fermionic mass term in the model.
One can, however, write down Wilson terms preserving either of two symmetries.
A possible gauge non-invariant Wilsonian prescription is given by addition to the naive action (12) of the following term,
$$\mathrm{\Delta }S_{\mathrm{W},\mathrm{gauge}}=\frac{1}{2}\mathrm{t}r\overline{\eta }\mathrm{\Gamma }^\mu [p_\mu ,\eta ]+[p_\mu ^{(+)},\overline{\eta }][p_\mu ^{()}\psi ]+[p_\mu ^{(+)},\overline{\psi }][p_\mu ^{()}\eta ],$$
(23)
where $`\eta `$ is $`U(N)`$-singlet Majorana-Weyl spinor matrix, and $`p_\mu ^{(\pm )}`$ are respectively forward and backward one-step scaled lattice derivatives,
$`p_i^{(\pm )}|n=\pm \mathrm{i}\sqrt{{\displaystyle \frac{\mathrm{}_i}{a_i}}}(|n\pm e_i|n)`$ (24)
$`q^{(\pm )i}|n=\mathrm{i}\sqrt{L_i}\mathrm{}^{\pm 2\pi \mathrm{i}n^i/N_i}|n.`$ (25)
Due to the term (23) the states with large phase of $`q`$ and $`p`$, (i.e. with $`n^iN_i`$ and $`k_i\frac{\mathrm{}_i}{a_i}`$) acquire large masses and decouple in the limit $`N\mathrm{}`$, as it happens in the case with usual Wilson term.
Another possibility is given by that in contrast to low-dimensional field theory models where the Lorentz invariance is obligatory, in the Matrix model it is less important. Moreover, its breaking to lower dimensional symmetries is desirable if one wants to describe a four dimensional theory in low energy limit, as it was proposed in the Ref. . It is not difficult to construct Wilson term which breaks, say Lorentz group SO(10) down to SO(9), but preserves the gauge symmetry. It looks as follows,
$$\mathrm{\Delta }S_{\mathrm{W},\mathrm{Lorentz}}=\frac{1}{2}\mathrm{t}r[p_\mu ^{(+)},\overline{\psi }]\mathrm{\Gamma }^9[p_\mu ^{()},\psi ],$$
(26)
where $`\mathrm{\Gamma }^9`$ is the 9-th ten dimensional Dirac gamma matrix.
Under this choice modes with large $`n`$ and $`k`$ also acquires large masses in the limit $`N\mathrm{}`$, but it produces terms which are not invariant with respect to rotations involving the 9-th axis. The gauge symmetry here remains intact. The last should not appear strange, because giving up a part of Lorentz invariance in an anomalous model allows one to cancel anomaly. Thus, in the simplest case of Abelian gauge anomaly in $`D=4`$ one can cancel the anomaly $`ϵ^{\mu \nu \lambda \sigma }F_{\mu \nu }F_{\lambda \sigma }`$, where $`\mu ,\nu ,\mathrm{}=0,1,2,3`$ by addition a local counterterm $`A_0ϵ^{ijk}A_i_jA_k`$, with $`i,j,\mathrm{}=1,2,3`$ which is not invariant with respect to Lorentz boosts.
## 5 Discussions
We have shown that perturbative fluctuations in IIB matrix model at finite $`N`$ exhibit a phenomenon similar to one in lattice gauge theories with fermions, consisting in doubling of the fermionic spectrum.
We considered a simple example of a background given by lattice shift/momentum shift operators. Basing on analogy with lattice model, we conjecture that this is a universal feature appearing for arbitrary choice of background configuration $`p_\mu `$ which is Hermitian, nondegenerate, etc., and can not be eliminated without breaking either Lorentz or gauge invariance. The actual results are straightforwardly translated to the case of the genuine finite $`N`$ vacua considered in .
We give prescriptions for the elimination of the doubling states in the limit $`N\mathrm{}`$, but preserving either Lorentz or gauge invariance and, respectively, breaking another one. This prescriptions should break the supersymmetry, since they do not restrict the bosonic spectrum as well.
However, in contrast to lattice gauge models, one may *not need* to eliminate such doubling states. Since there is a conjecture that IKKT matrix model nonperturbatively describes both IIB and IIA string models , which have different chirality content, it may be possible that the doubling in the language of matrix models is related to the string duality.
I am grateful to the members of our informal seminar for useful discussions and comments.
|
warning/0005/cond-mat0005455.html
|
ar5iv
|
text
|
# Limits of validity for a semiclassical mean-field two-fluid model for Bose-Einstein condensation thermodynamics
## I Introduction
The recent experimental observation of Bose-Einstein condensation (BEC) in a weakly interacting dilute gas of <sup>87</sup>Rb , <sup>23</sup>Na , <sup>7</sup>Li , and <sup>1</sup>H employing magnetic traps at ultra-low temperatures calls for a theoretical investigation on various aspects of the condensate. The condensate consists of few thousands to few millions of atoms confined by the trap potential. As the temperature is lowered below the critical temperature $`T_0`$ of BEC, the condensate starts to form and finally at 0 K all the available atoms will be condensed. In the absence of a microscopic equation, the condensate is usually described by the mean-field Gross-Pitaevskii (GP) equation . One of the primary interest on the process is to study the thermodynamic observables of the system, such as, the condensate fraction, internal energy, and specific heat.
There have been several comprehensive studies on temperature dependencies of thermodynamic observables of the condensate using semiclassical mean-field two-fluid models . The physical ingredients of these models are quite similar. One such model using the GP wave function provides satisfactory description of the temperature dependencies of the thermodynamic observables in two , and three dimensions. These studies employed a iterative solution of the system of equations involved. For the condensation of a system composed of 40000 trapped <sup>87</sup>Rb atoms with repulsive interatomic interaction the iterative scheme converged rapidly and provided a satisfactory account of the condensate fraction, internal energy, and specific heat . Similar conclusion was also reached in the study of condensation of <sup>7</sup>Li atoms . In case of <sup>7</sup>Li the attractive interatomic interaction is responsible for collapse if the number of atoms is larger than approximately 1400 .
Here we reinvestigate critically the BEC of a weakly interacting dilute gas in two and three dimensions using the two-fluid mean-field model mentioned above in order to define the domain of its applicability. We have included the two-dimensional BEC in this study because of considerable recent interest in this topic . We employ the usual iterative solution of the nonlinear two-fluid mean-field model and study the convergence of the iterative scheme. Although, the convergence is rapid for a weakly interacting dilute system, with an increase of the strength of interaction and/or density, the model breaks down and leads to physically unacceptable results for the thermodynamic observables. Specifically, below the critical temperature, the model may yield negative specific heat.
However, it is well-known that the mean-field description of the condensate via the GP equation as used in the semiclassical models above should hold under the condition of diluteness of a weakly interacting Bose gas and is expected to break down once the conditions of diluteness and weak interaction are violated See, for example, page 474 of Ref. . The breakdown should happen for a large number of condensed particles reflecting a large density as well as for a large modulus of the scattering length denoting a strong interatomic interaction. These two conditions correspond to a large nonlinearity of the system and may lead to a breakdown of the mean-field two-fluid thermodynamic model . In addition, the finite size of the system may necessitate corrections in the thermodynamic model as we are away from the real thermodynamic limit $`N\mathrm{}`$, $`V\mathrm{}`$, where $`N`$ is the number of particles and $`V`$ the volume of the system <sup>§</sup><sup>§</sup>§These limitations of the thermodynamic model are discussed in Sec. 5 of Ref.. Nevertheless exact numerical conditions for the breakdown of the mean-field thermodynamical model have never been investigated. In this work by performing numerical calculations we identify such conditions. We find that the semiclassical model may break down under possible present experimental conditions of BEC of a trapped Bose gas.
We present the semiclassical model in Sec. 2, numerical results in Sec. 3, and conclusions in Sec. 4.
## II Mean-field Model
We consider a system of $`N`$ bosons with repulsive interaction at temperature $`T`$ under the influence of a trap potential. The condensate is described by the following GP wave function in the Thomas-Fermi approximation :
$$|\mathrm{\Psi }(r)|^2=\frac{\mu V_{\text{ext}}2gn_1(r)}{g}\theta (\mu V_{\text{ext}}2gn_1(r)),$$
(1)
where $`\theta (x)`$ is the step function, $`\theta (x)=0`$ if $`x<0`$ and 1 otherwise. Here $`V_{\text{ext}}(r)m\omega ^2r^2/2`$ is the trap potential, $`g`$ the strength of the repulsive interaction between the atoms, $`m`$ the mass of a single bosonic atom, $`\omega `$ the angular frequency, $`\mu `$ the chemical potential, and $`n_1(r)`$ represents the distribution function of the noncondensed bosons. As we are interested in studying the limits of validity of the semiclassical mean-field model and not in simulating a particular experimental situation, we consider a spherically symmetric trap both in two and three dimensions. The noncondensed particles are treated as non-interacting bosons in an effective potential
$$V_{\text{eff}}(r)=V_{\text{ext}}(r)+2gn_1(r)+2g|\mathrm{\Psi }(r)|^2.$$
(2)
Thermal averages are calculated with a standard Bose distribution of the noncondensed particles in chemical equilibrium with the condensate governed by the same chemical potential $`\mu `$. In particular the density $`n_1(r)`$ is given by
$$n_1(r)=\frac{1}{(2\pi \mathrm{})^𝒟}\frac{d^𝒟p}{\mathrm{exp}[\{p^2/2m+V_{\text{eff}}(r)\mu \}/k_BT]1},$$
(3)
where $`k_B`$ is the Boltzmann constant, and $`𝒟(2,3)`$ is the dimension of space. Equations (1) $``$ (3) above are the principal equations of the present model. The total number of particles $`N`$ of the system is given by the number equation
$`N=N_0+{\displaystyle \frac{\rho (E)dE}{\mathrm{exp}[(E\mu )/k_BT]1}},`$ (4)
where $`N_0|\mathrm{\Psi }(r)|^2d^𝒟r`$ is the total number of particles in the condensate. The critical temperature $`T_0`$ is obtained as the solution of Eq. (4) with $`N_0`$ and $`\mu `$ set equal to 0. The semiclassical density of states $`\rho (E)`$ of noncondensed particles is given by
$$\rho (E)=\frac{2\pi m^{𝒟/2}}{(2\pi \mathrm{})^𝒟}_{V_{\text{eff}}(r)<E}[8(EV_{\text{eff}}(r))]^{(𝒟2)/2}d^𝒟r.$$
(5)
The average single-particle energy of the noncondensed particles is given by
$$E_{\text{nc}}=\frac{E\rho (E)dE}{\mathrm{exp}[(E\mu )/k_BT]1}.$$
(6)
The kinetic energy of the condensate is assumed to be negligible and its interaction energy per particle is given by $`E_\text{c}=(g/2)\mathrm{\Psi }^4(r)d^𝒟r`$. The quantity of experimental interest is the average energy $`E=[E_{\text{nc}}(NN_0)/2+E_\text{c}]/N,`$ which we calculate in the following. The specific heat is defined by $`C=dE/dT`$ .
Equations (1) $``$ (5) are to be solved iteratively. The iteration is started at a fixed temperature with a trial chemical potential $`\mu `$ using $`n_1(r)=0`$. Then $`\mathrm{\Psi }(r)`$ and $`V_{\text{eff}}(r)`$ are calculated using Eqs. (1) and (2). With these results $`n_1(r),\mathrm{\Psi }(r),`$ and $`V_{\text{eff}}(r)`$ are recalculated. This procedure is repeated until desired precision is obtained. The results for the lowest order of iteration with $`n_1(r)=0`$ will be denoted by $`I=1`$, and successive orders by $`I=2,3,\mathrm{}`$ etc. With the solutions $`\mathrm{\Psi }(r)`$ and $`V_{\text{eff}}(r)`$ so obtained, the density of states $`\rho (E)`$ of Eq. (5) is calculated. Then it is seen if they satisfy the number equation (4). If Eq. (4) is satisfied the desired solution is obtained. If not, the initial trial $`\mu `$ is varied until the number equation is satisfied. In each order of iteration we calculate the condensate fraction $`N_0/N`$ and energy $`E`$.
## III Numerical Study
First we consider the three-dimensional case. In this case the coupling $`g`$ is given by $`g=4\pi \mathrm{}^2a/m`$, where $`a`$ is the scattering length. Usually, a dimensionless coupling is introduced via $`\eta (mg/\pi \mathrm{}^2)/a_{\text{ho}}=4a/a_{\text{ho}}`$, where $`a_{\text{ho}}=\sqrt{\mathrm{}/m\omega }`$. The semiclassical model under consideration should break down as either $`\eta `$ or $`N`$ is increased. This will violate the condition of weak interaction and diluteness. In the case of experiment on BEC of <sup>87</sup>Rb, $`\eta =0.025`$ and $`N=40000`$ . To test our calculational scheme, first we solve the present model for $`\eta =0.025`$ and $`N=40000`$. Our results are very similar to those of Ref. . The small difference between these two calculations is due to the use of an isotropic harmonic oscillator potential in this work and an anisotropic potential in Ref. . In this case specific heat is positive at all temperatures. Next we increase $`\eta `$ and $`N`$. We find that as $`\eta `$ and $`N`$ are increased, the energy develops a maximum at a temperature below $`T_0`$. Consequently, the specific heat becomes negative for an interval of temperature above this maximum. To show this violation in a pronounced way here we show the results for the following three cases: (a) $`\eta =0.1`$, $`N=10^6`$, (b) $`\eta =0.1`$, $`N=10^7`$, and (c) $`\eta =0.5,N=10^6`$. In the case of <sup>23</sup>Na, the experimental $`N`$ was as high as 10<sup>7</sup> . As the interaction is repulsive in both <sup>87</sup>Rb and <sup>23</sup>Na, it should be possible to have 10<sup>7</sup> atoms in a BEC of <sup>87</sup>Rb under favorable experimental conditions. Hence these values of the parameters are within the present experimental scenario.
Fig. 1. Condensate fraction $`N_0/N`$ and energy $`E/Nk_BT_0`$ in three dimensions as a function of $`T/T_0`$ for (a) $`\eta =0.1`$ and $`N=10^6`$, (b) $`\eta =0.1`$ and $`N=10^7`$, and (c) $`\eta =0.5`$ and $`N=10^6`$ for iterations $`I=1`$ (full line), 2 (dashed-dotted line), and 4 (dotted line). The straight line represents the classical Maxwell-Boltzman result of energy.
In Figs. 1 (a), (b), and (c) we plot $`E/Nk_BT_0`$ and $`N_0/N`$ versus $`T/T_0`$ for different iterations for the above three cases. We find that the energies are acceptable under conditions of diluteness and very weak interactions, but as $`\eta `$ and $`N`$ increases the average energy of the system may have a maximum leading to a negative specific heat above the maximum. From Figs. 1(a), (b), and (c) we find that this violation happens in the case (c) which has the largest $`\eta `$ and $`N`$. We verified that for sufficiently large values of $`\eta `$ and $`N`$ the lowest-order energy also leads to negative specific heat.
Next we consider the two-dimensional case. In this case, in analogy with the three-dimensional case, a dimensionless coupling is introduced by $`\eta (mg/\pi \mathrm{}^2)`$. This coupling is already dimensionless, whereas we needed to divide it by $`a_{\text{ho}}`$ in three dimensions to make it dimensionless. First we repeat the calculations reported in Ref. and our results are in agreement with that study. In addition, in agreement with our finding in three dimensions, we find that for small values of $`\eta `$ and $`N`$ the equations of the model converge well and lead to acceptable values for condensate fraction $`N_0/N`$ and energy $`E/Nk_BT_0`$. For larger values of $`\eta `$ and/or $`N`$, the condensate fraction $`N_0/N`$ is quite acceptable with a temperature dependence similar to that in three dimensions. However, the energy produces a maximum as $`N`$ and $`\eta `$ increase. Hence we shall be limited to a consideration of energy only.
As there is no experimental guideline for probable values of $`N`$ and $`\eta `$ in two dimensions, as in Ref. , we consider $`N=10^5`$ and $`\eta =0.1,`$ 1, and 10. In Fig. 2 we plot the temperature dependence of average energy $`E/Nk_BT_0`$ for different iterations. The lowest-order results for $`\eta =0.1`$ and 1 are in agreement with those of Ref. . The classical Maxwell-Boltzmann result for a noninteracting gas is also shown in Fig. 2. For a weakly interacting gas with $`\eta =0.1`$, the energies for all orders of iteration are acceptable. For a stronger interaction with $`\eta =1`$, the lowest-order energy is acceptable. However, in this case the energies for all orders of iteration produce maxima leading to negative specific heat. For $`\eta =10`$ the lowest-order result already leads to a negative specific heat.
Fig. 2. Energy $`E/Nk_BT_0`$ in two dimensions as a function of $`T/T_0`$ for $`N=10^5`$ and $`\eta =0.1`$, 1, and 10 for iterations $`I=1,2`$ and 4. Notations are the same as in Fig. 1. The curves are labeled by the values of $`\eta `$.
In addition to the trouble discussed above, the mean-field model may exhibit another unacceptable behavior. We see in Figs. 1 and 2, that the energy the system could be larger than the corresponding classical Maxwell-Boltzmann result for a noninteracting ideal gas. We recall that for BEC to materialize, the energy of the condensate should be smaller than the corresponding energy of the noncondensed system. However, one should consider the correction to the classical result above due to the presence of the trap and the interatomic interaction. In all our calculations we have noted that $`k_BT_0>>\mathrm{}\omega `$, so that for temperatures close to $`T_0`$ considered above the correction to the classical energy due to the presence of the trap can be neglected. The same is also true for the interatomic interaction at temperatures close to $`T_0`$, for the values of $`\eta `$ considered in this study. Consequently, the energy of the condensate of the mean-field model could be larger than the classical result signaling a breakdown of the model.
It is well known that the semiclassical model requires $`k_BT_0>>\mathrm{}\omega `$ and the Thomas-Fermi approximation does not hold for very small number of atoms in the condensate. We have verified that for the cases studied here $`k_BT_0`$ is typically 50 to 100 times larger than $`\mathrm{}\omega `$ and Thomas-Fermi approximation apparently seems to be valid for the condensates of size $`N10^6`$ to $`10^7`$ considered here. The diluteness condition for the validity of the GP equation $`na^3<<1`$ is also valid in three dimensions, where $`n`$ is the density. For the case considered in Fig. 1(a), $`N=10^6`$, $`a=50`$ nm, for a condensate of typical dimension 100 $`\mu `$m, $`na^310^3<<1`$. The values of $`na^3`$ are larger for Figs. 1(b) and 1(c). Nevertheless, we find that these values of diluteness and coupling set a limit to the applicability of the semiclassical mean-field two-fluid models for studying BEC thermodynamics.
## IV Conclusion
In conclusion, we reexamined the problem of BEC under the action of a trap potential using a two-fluid mean-field model in both two and three dimensions. We employed an iterative solution scheme of the system of equations. Although the system leads to rapid convergence for a weakly interacting dilute system, with the increase of coupling and particle number the iterative scheme leads to physically unacceptable results for thermodynamic observables. Specifically, this may lead to a maximum in energy responsible for a negative specific heat. In three dimensions the breakdown of the mean-field model happens for values of coupling and particle number, which are not so remote from present experimental scenario for repulsive interatomic interaction. In addition, the energy of the condensate could be larger than the corresponding classical energy of the system, which is another independent unacceptable result of the model. The larger values of the coupling $`\eta `$ and $`N`$ considered in this work possibly sets a limit to the applicability of the mean-field equations for BEC. Summarizing, the most important finding of this study is that the iterative solution of the mean-field model of Refs. for a condensate may lead to unphysical thermodynamical properties for medium to large coupling and number of particles.
Despite the above deficiency of the mean-field two-fluid thermodynamical models, they continue to be very useful in many cases. The virtue of these models is the simplicity and ability to yield results in agreement with experiment for weak interactions and dilute systems. As a theoretically sound description of the BEC thermodynamics seems to be unmanagably complicated, these simple mean-field two-fluid models remain as attractive simple alternatives to study thermodynamical properties provided that proper attention is paid to remain inside the domain of their validity.
AG thanks Prof. T. Frederico and Prof. L. Tomio for an introduction to the subject of BEC. The work is supported in part by the Conselho Nacional de Desenvolvimento Científico e Tecnológico and Fundação de Amparo à Pesquisa do Estado de São Paulo of Brazil.
|
warning/0005/astro-ph0005580.html
|
ar5iv
|
text
|
# On the velocity of the Vela pulsar
## 1 Introduction
The Vela pulsar is one of the best studied radio pulsars and was the first one found to be associated with the supernova remnant (SNR) (Large et al. (1968)). In spite of its relative proximity to the Earth, there is still no consensus on the value of its (transverse) velocity, which is connected with the yet unsolved problem of the distance to the Vela pulsar/Vela SNR. The first attempt to estimate the pulsar velocity was made by Bignami & Caraveo (1988), whose optical measurments gave an upper limit on the pulsar proper motion ($`<60\mathrm{mas}\mathrm{yr}^1`$). However, even the maximum admissible value of the pulsar proper motion was found to be too low to explain the pulsar offset from the apparent geometrical centre of the Vela SNR, which questions the pulsar/SNR association. Later, it was recognized (e.g. Seward (1990), Aschenbach et al. (1995)) that the real extent of the Vela SNR is much larger than was accepted in early studies, so now there can be no doubt that the Vela pulsar and the Vela SNR are the remnants of the same supernova explosion. However, this association has caused some problems in estimating the pulsar velocity.
Wallerstein & Silk (1971) questioned for the first time the ’canonical’ value of the distance to the Vela SNR of 500 pc given by Milne (m1 (1968); see also Taylor et al. (1993)) and suggested that this distance should be reduced to some smaller value ($`250`$ pc). Since that time many additional arguments in support of this suggestion have been put forward (Ögelman et al. ok (1989), Oberlack et al. o (1994), Jenkins & Wallerstein (1995), Aschenbach et al. (1995), Bocchino et al. (1999), Cha et al. ch2 (1999), Cha & Sembach (2000))<sup>1</sup><sup>1</sup>1We critically analysed these arguments (Gvaramadze 2000a,b ) and came to the conclusion that there are no weighty reasons to revise the ’canonical’ distance of 500 pc. . One of the arguments (proposed by Ögelman et al. (ok (1989)) and repeated by Oberlack et al. (o (1994)) and Cha et al. (ch2 (1999))) was based on the comparison of the Vela pulsar transverse velocity (inferred from the new estimate of the pulsar proper motion by Ögelman et al. (ok (1989))) with the scintillation velocity (reported by Cordes (c1 (1986))). It is known that proper motion velocities of pulsars show significant correlation with pulsar velocities inferred from interstellar scintillation measurements (e.g. Lyne & Smith (1982), Gupta gu1 (1995)). Therefore, assuming that the scintillation velocity $`V_{\mathrm{iss}}`$ is a ’true’ value of the pulsar transverse velocity, Ögelman et al. (ok (1989)) suggested that $`V_{\mathrm{iss}}=53\pm 5\mathrm{km}\mathrm{s}^1`$ (Cordes c1 (1986)) could be reconciled with the proper motion $`\mu 38\pm 8\mathrm{mas}\mathrm{yr}^1`$ if the distance to the Vela pulsar (and the Vela SNR) is about $`290\pm 80`$ pc. The distance reduction might be even more dramatic if one takes the recent high-precision estimate of the Vela pulsar proper motion ($`52\pm 3\mathrm{mas}\mathrm{yr}^1`$) obtained by De Luca et al. (d (2000); see also Nasuti et al. (1997)). The situation with the distance to the Vela SNR was ’improved’ after Gupta et al. (gu2 (1994)) showed that the scintillation velocity calculation formula used by Cordes (c1 (1986)) underestimates $`V_{\mathrm{iss}}`$ by factor of 3. The revised value of the scintillation velocity of the Vela pulsar of $`152\mathrm{km}\mathrm{s}^1`$ better corresponds to the proper motion velocity of $`123\mathrm{km}\mathrm{s}^1`$ (for the distance to the pulsar of 500 pc and $`\mu =52\mathrm{mas}\mathrm{yr}^1`$). However, the real situation is a bit more complicated.
It should be noted that the calculations of Gupta (gu1 (1995)) were based on the assumptions that the scattering material is concentrated in a thin screen and that the screen is placed midway between the observer and the pulsar. Although the first assumption is realistic, the second one is not suitable in the case of the Vela pulsar. Indeed, it is believed that the scattering irregularities responsible for the enhanced scattering of the Vela pulsar (Backer (1974)) are localized in a thin screen rather than uniformly distributed along our line of sight to the pulsar (Backer (1974), Lee & Jokipii (1976); see also Williamson (1974)), and that the scattering screen resides close to the pulsar and could be associated with the shell of the Vela SNR (Desai et al. de (1992), Taylor & Cordes (1993), Gwinn et al. (1993); see however Gwinn et al. gw2 (1997),gw3 (2000) and cf. Cordes & Rickett (1998)). The asymmetrical location of the screen implies (see Sect. 2) that the actual value of the scintillation velocity should be considerably larger than that given by Gupta (gu1 (1995)) and that the scintillation velocity is not equal to the proper motion velocity.
In this paper, we show that if the scattering of the Vela pulsar indeed occurs in the shell of the Vela SNR then the scintillation velocity could be reconciled with the pulsar proper motion velocity only if the scatterer has a nonzero transverse velocity. A possible origin of large-scale transverse motions in the Vela SNR’s shell is discussed.
## 2 Scintillation and proper motion velocities
The scintillation velocity for an asymmetrically placed thin scattering screen is (Gupta et al. gu2 (1994), Gupta gu1 (1995)):
$$V_{\mathrm{iss}}=3.85\times 10^4\frac{(\nu _{\mathrm{d},\mathrm{MHz}}D_{\mathrm{kpc}}x)^{1/2}}{f_{\mathrm{GHz}}t_\mathrm{d}}\mathrm{km}\mathrm{s}^1,$$
(1)
where $`\nu _{\mathrm{d},\mathrm{MHz}}`$ and $`t_\mathrm{d}`$ are the scintillation bandwidth and the time-scale measured respectively in MHz and seconds, $`D_{\mathrm{kpc}}`$ is the distance from observer to pulsar in kpc, $`x=D_\mathrm{o}/D_\mathrm{p}`$, $`D_\mathrm{o}`$ and $`D_\mathrm{p}`$ are the distances from observer to screen and from screen to pulsar, $`f_{\mathrm{GHz}}`$ is the frequency of observation in units of GHz. For $`\nu _{\mathrm{d},\mathrm{MHz}}=0.001`$, $`t_\mathrm{d}=5.6`$, $`f_{\mathrm{GHz}}=1`$ (Cordes c1 (1986)), $`D_{\mathrm{kpc}}=0.5`$, and assuming that $`x=1`$, one finds for the Vela pulsar that $`V_{\mathrm{iss}}=152\mathrm{km}\mathrm{s}^1`$ (Gupta gu1 (1995)). As we mentioned in Sect. 1, Desai et al. (de (1992)) showed that the scattering screen is close to the pulsar. Assuming that $`D=500`$ pc, they found that $`D_\mathrm{o}/D0.81`$, and that this value could be increased up to 0.96 if $`5\%`$ of the scattering of the Vela pulsar is due to the effect of the Gum Nebula. The latter value of $`D_\mathrm{o}/D`$ is expected if the scattering material is mainly concentrated in the shell of the Vela SNR of angular diameter of $`5^\mathrm{°}`$ (the figure accepted in early studies of the Vela SNR). Assuming that the Vela SNR’s shell is indeed the main scatterer of the Vela pulsar and using the currently adopted angular size of the Vela SNR of $`7^\mathrm{°}`$, one has $`D_\mathrm{o}/D0.94`$ or $`x=15.7`$, and correspondingly $`V_{\mathrm{iss}}600\mathrm{km}\mathrm{s}^1`$. In the observer’s reference frame, the scintillation velocity is connected with the pulsar proper motion velocity
$$V_{\mathrm{pm}}=4.74\mu D_{\mathrm{kpc}}\mathrm{km}\mathrm{s}^1,$$
(2)
where $`\mu `$ is measured in mas $`\mathrm{yr}^1`$, by the following relationship (cf. Gupta et al. gu2 (1994), Cordes & Rickett (1998)):
$`V_{\mathrm{iss}}`$ $`=`$ $`[x^2V_{\mathrm{pm}}^22x(1+x)V_{\mathrm{pm}}V_{\mathrm{scr},}`$ (3)
$`+`$ $`(1+x)^2V_{\mathrm{scr},}+(1+x)^2V_{\mathrm{scr},}^2]^{1/2},`$
where $`V_{\mathrm{scr},}`$ and $`V_{\mathrm{scr},}`$ are the components of the transverse velocity of the screen, correspondingly, parallel and perpendicular to the vector of the pulsar proper motion velocity. In (3) we neglected small contributions from the differential Galactic rotation and the Earth’s orbital motion around the Sun. If $`V_{\mathrm{scr}}=0`$, one has $`V_{\mathrm{pm}}=V_{\mathrm{iss}}/x38\mathrm{km}\mathrm{s}^1(D_{\mathrm{kpc}}=0.5)`$, i.e. about 3 times smaller than that from eq. (2). These velocity estimates could be reconciled only if the distance to the Vela pulsar is $`D_{\mathrm{kpc}}=0.05(x/15.7)^1`$, which is too small to be likely. The pulsar, however, could be placed at its ’canonical’ distance if $`V_{\mathrm{scr}}0`$.
Fig. 1 shows the 843 MHz image (Bock et al. (1998)) of the central part of the Vela SNR, known as the radio source Vela X (Milne m1 (1968)). A considerable fraction of the radio emission from Vela X originates in filamentary structures, one of which crosses the Vela pulsar position.
This filament (or rather its part to the south of the pulsar) is known as a radio counterpart of the Vela X-ray ’jet’ discovered by Markwardt & Ögelman (1995). In Gvaramadze (g1 (1998), g2 (1999)), we found that some of radio filaments of the Vela X and the X-ray ’jet’ show a fairly good correlation with optical filaments, and concluded that the ’jet’ is a dense filament in the Vela SNR’s shell, which is projected by chance near the line of sight to the Vela pulsar. We also suggested that filamentary structures visible throughout the Vela SNR in radio, optical and X-ray ranges have a common nature<sup>2</sup><sup>2</sup>2This suggestion implies that the radio source Vela X is a part of the Vela SNR’s shell (see also Milne & Manchester (1986), Milne (1995); but see e.g. Frail et al. (1997), Bock et al. (1998) and Chevalier (1998) for a different point of view). and that their origin is connected with projection effects in the Rayleigh-Taylor unstable shell of the remnant. The Rayleigh-Taylor instability results from the impact of the supernova ejecta/shock with the pre-existing wind-driven shell created by the supernova progenitor star (see Gvaramadze g2 (1999)), and induces in the shell large-scale transverse motions (laterally expanding domelike deformations of the shell). The existence of laterally expanding deformations naturally explains (Gvaramadze g2 (1999)) the ’unusual’ velocity field inferred by Jenkins et al. (1984) from the study of absorption lines in spectra of background stars (see also Jenkins et al. (1976), Danks & Sembach (1995)). The high-velocity absorption features were found not only in the central part of the Vela SNR but also near the edges of the remnant, which suggests that the expansion velocity of the shell deformations has comparable radial and transverse components. The same conclusion follows from the interpretation of UV spectra of face-on and edge-on shock waves in the Vela SNR (Raymond et al. (1997)). A characteristic expansion velocity of shell deformations inferred from the absorption data and UV spectra is about $`100\mathrm{km}\mathrm{s}^1`$, while some portions of the shell expand with much higher velocities.
Proceeding from the above, we suggest that the radio filament projected on the Vela pulsar is a large-scale deformation of the Vela SNR’s shell viewed edge-on, and that this deformation has a significant transverse velocity. We assume that the deformation lies on the approaching side of the Vela SNR’s shell and suggest that the turbulent material associated with the shell deformation is responsible for the scattering of the Vela pulsar (cf. Desai et al. de (1992)). The line of sight extent of the scattering material could be estimated to be $`1.21.7`$ pc (for $`D=500`$ pc) given that the width of the filament is $`2^{^{}}3^{^{}}`$ (Milne (1995), Bock et al. (1998)) and assuming that the characteristic size of the shell deformation is $`40^{^{}}50^{^{}}`$. The geometry (the curvature) of the filament projected on the Vela pulsar suggests that this part of the shell expands in the northwest direction, i.e. just parallel to the vector of the pulsar proper motion velocity (Ögelman et al. ok (1989), Bailes et al. (1989)). For $`V_{\mathrm{scr},}=0`$, one has from (3) that
$$V_{\mathrm{scr}}=V_{\mathrm{scr},}\frac{xV_{\mathrm{pm}}\pm V_{\mathrm{iss}}}{x+1}.$$
(4)
For the transverse velocity of the pulsar of $`120\mathrm{km}\mathrm{s}^1`$ (i.e. for $`D=500`$ pc) and for $`x=15.7`$ and $`V_{\mathrm{iss}}600\mathrm{km}\mathrm{s}^1`$, one has that $`V_{\mathrm{scr}}`$ is 80 or $`150\mathrm{km}\mathrm{s}^1`$. The first estimate is quite reasonable, while the second, though not impossible, looks less likely. It should be noted, however, that eq. (4) gives reasonable values for $`V_{\mathrm{scr}}`$ not only for $`D=500`$ pc. E.g. for $`D=250`$ pc, one finds that $`V_{\mathrm{scr}}`$ is equal to 30 or $`80\mathrm{km}\mathrm{s}^1`$. From this, we conclude that the scintillation data taken alone do not allow us to put meaningful limits on the distance to the Vela pulsar and therefore to get a reliable estimate of the transverse velocity of the pulsar. The forthcoming direct measurement of the distance to the Vela pulsar through its parallactic displacement (see De Luca et al. d (2000)) will solve the problem.
## 3 Discussion
The main assumption made in this paper is that the enhanced scattering of the Vela pulsar is due to the effect of the Vela SNR. This assumption is based on the result of Desai et al. (de (1992)) that the scattering material could be associated with the shell of the Vela SNR (see Sect. 2). In this case, the parameter $`x`$ has a fixed value, which depends only on the angular size of the Vela SNR’s shell, and therefore is independent of the distance to the pulsar.
It should be noted, however, that the result of Desai et al. (de (1992)) was derived from use of scintillation observables (see Blanford & Narayan (1985), Gwinn et al. (1993)). The referee (J.Cordes) pointed out that the quite large uncertainties in observables result in uncertainty in the value of the parameter $`x`$. He attracted our attention to the papers by Gwinn et al. (gw2 (1997),gw3 (2000)), from which follows that the value of $`x`$ could be much smaller than that adopted in our paper. The values of $`x`$ given in these papers (respectively, $`x=2.7(D/0.5\mathrm{kpc})^1`$ and $`x=1.5(D/0.5\mathrm{kpc})^1)`$ imply that the scattering material cannot be connected with the Vela SNR. We now discuss this possibility.
For $`x=2.7(D/0.5\mathrm{kpc})^1`$ (Gwinn et al. gw2 (1997)), one obtains<sup>3</sup><sup>3</sup>3For the sake of simplicity, we use here the same values of $`\nu _\mathrm{d},t_\mathrm{d}`$ and $`f`$ as in Sect. 2. $`V_{\mathrm{iss}}=250\mathrm{km}\mathrm{s}^1`$. Note that now $`V_{\mathrm{iss}}`$ is independent of $`D`$ since $`xD^1`$ (see e.g. Gwinn et al. (1993) and eq. (1)). One can see that $`V_{\mathrm{iss}}`$ could be reconciled with $`V_{\mathrm{pm}}`$ only if $`V_{\mathrm{scr}}0`$: $`V_{\mathrm{scr}}=157`$ or $`22\mathrm{km}\mathrm{s}^1`$ for $`D=0.5`$ kpc; $`V_{\mathrm{scr}}=91`$ or $`13\mathrm{km}\mathrm{s}^1`$ for $`D=0.25`$ kpc . These transverse velocities are too high to be attributed to the expansion of the shell of the Gum Nebula<sup>4</sup><sup>4</sup>4The enhanced scattering of the Vela pulsar was originally ascribed to the Gum Nebula (Backer (1974); see also Cordes et al. (1985)). (cf. Reynolds (1976) with Wallerstein et al. (1980)), though it is not impossible that they could characterize the expansion of a foreground small-scale H II region projected by chance on the Vela pulsar.
For $`x=1.5(D/0.5\mathrm{kpc})^1`$ (Gwinn et al. gw3 (2000)), one obtains $`V_{\mathrm{iss}}=186\mathrm{km}\mathrm{s}^1`$. One can show that just for this value of $`x`$ (or more exactly for $`x=(152/123)^2=1.53`$) the transverse velocity of the screen could be equal to zero (see eq. (4)), and therefore $`V_{\mathrm{pm}}=V_{\mathrm{iss}}/x`$.
Although these estimates show that the situation with the scattering screen is indeed quite uncertain, we believe that the enhanced scattering of the Vela pulsar is most likely connected with the shell of the Vela SNR. An argument in support of this belief is the fact that the elongated scattering disk of the Vela pulsar (Gwinn et al. gw2 (1997)) is nearly perpendicular to the magnetic field of the radio filament projected on the Vela pulsar.
###### Acknowledgements.
I am grateful to D.C.-J.Bock for providing the electronic version of the image of the central part of the Vela SNR, and to J.Cordes (the referee) for useful suggestions and comments.
|
warning/0005/astro-ph0005460.html
|
ar5iv
|
text
|
# Periodic modulation of the optical counterpart of the X-ray pulsator 1WGA J1958.2+3232. A new intermediate polar.Based on observations collected at Catania and Asiago Astrophysical Observatories.
## 1 Introduction
1WGA J1958.2+3232 is one of the objects found by Israel et al. (Isr98 (1998)) in a systematic search for pulsators in the catalogue of ROSAT X-ray sources compiled by White, Giommi & Angelini (WGA94 (1994)). The source appeared with a mean flux (in 0.1-2.4 keV) of $``$ 10<sup>-12</sup> erg cm<sup>-2</sup> s<sup>-1</sup>, modulated with a period 721 $`\pm `$ 14 s and pulsed fraction of about 80%. A subsequent observation performed by ASCA confirmed both the flux level and the strong periodic modulation at 734 $`\pm `$ 1 s (Israel et al. Isr99 (1999)). The energy spectrum, as measured by ROSAT, is consistent with a power law model with a photon index $`\mathrm{\Gamma }`$=0.8$`{}_{0.6}{}^{}{}_{}{}^{+1.2}`$ and a column density N<sub>H</sub>=(6$`{}_{5}{}^{}{}_{}{}^{+24}`$)$`\times `$10<sup>20</sup>cm<sup>-2</sup>.
On the basis of optical photometry and spectroscopy, Israel et al. (Isr99 (1999)) proposed a m<sub>V</sub>=15.7 star as the counterpart of 1WGA J1958.2+3232. The spectrum of this star, characterized by strong emission lines, was classified as B0Ve type, suggesting that 1WGA J1958.2+3232 could be a Be/X-Ray binary with an accreting neutron star. The assumed distance of 800 pc (resulting from interstellar absorption spectral features) yielded an X-ray luminosity of 8$`\times `$10<sup>31</sup> erg s<sup>-1</sup> in 0.1-2.4 keV.
Recently Negueruela et al. (Neg99 (2000)) reported new spectroscopic measurements, with a higher signal to noise ratio, of the Israel et al. (Isr99 (1999)) candidate. They found that optical properties of the source are incompatible with a Be/X-Ray binary, because photospheric features are absent, whereas some characteristics of the spectrum, such as the strong emission in all Balmer lines, are typical of a cataclysmic variable. The shape of the lines, strongly asymmetric and double peaked, suggests the presence of an accretion disc.
We included 1WGA J1958.2+3232 in our target list during a campaign for scientific performance evaluation of a photon counting detector that we carried out in June 1999. We report here on the results of this campaign and of subsequent observations.
## 2 Observations
The observations were carried out with the 91 cm telescope of the Astrophysical Observatory of Catania (OACT) - Serra la Nave and with the 182 cm telescope of the Astronomical Observatory of Padua (OAPD) - Asiago Cima Ekar (see the observations log in Table 1). Two kinds of detectors were used: a photon counting camera based on an Intensified CCD<sup>1</sup><sup>1</sup>1The ICCD detector has been developed at the Istituto di Fisica Cosmica ”G.Occhialini”, in collaboration with Dipartimento di Ingegneria Elettronica of Padua University. (Bergamini et al. Ber97 (1997, 1998, 2000)) and a standard photomultiplier photometer.
The Photon Counting ICCD (PC-ICCD) employs a high-gain Micro-Channel Plate (MCP) image intensifier as input device. The intensifier converts virtually each photon, interacting photoelectrically with the photocathode, into a luminous spot on a phosphor screen that preserves the (x,y) location of the event. The photon event is reimaged onto a fast framing CCD camera by means of a fiber optic taper with a 3.6:1 demagnification ratio. The camera detects the location of each event and a dedicated digital electronic system establishes the coordinates of the centroid of the events with subpixel accuracy. The detector can operate in image integration mode, where the collected photons are accumulated in a memory array, or in photon list mode where the photon events are recorded as a list of time tagged event coordinates with temporal resolution as low as 4.5 ms.
Thus, in photon list mode, the PC-ICCD is able to perform high speed space resolved photometry, allowing to investigate time variability down to few ms, simultaneously for all the stars in the field of view.
The PC-ICCD was developed for space UV astronomy and uses a RbTe photocathode: even though the photocathode shows its peak sensitivity in the vacuum UV range, it yields a residual quantum efficiency ($``$ 0.1 %) for ground based observations (Bergamini et al. Ber00 (2000)). In order to preserve this residual efficiency, no photometric filter has been used with the ICCD; the boundaries of instrumental band (3200-5500 Å) were set respectively by the atmosphere UV cut-off and by the detector photocathode sensitivity.
For measurements in photometric bands, we used a photon-counting cooled photometer of OACT, equipped with an EMI 9893QA/350 photomultiplier.
## 3 Data analysis and results
### 3.1 Source identification and short term variability
On several nights in June 1999, 1WGA J1958.2+3232 was observed in photon list mode with a time resolution of 16.8 ms for a total of 6.3 h distributed in five integrations (Table 1). The 27.6 $`\times `$ 27.6 mm<sup>2</sup> sensitive area of the PC-ICCD covered a 6.4’ $`\times `$ 6.4’ field of view with an effective resolution of 0.35 arcsec (FWHM).
A set of IDL routines were used to process the data stream of each observation and to perform data reduction adopting the following procedure. The observations were segmented into 60 s long data subsets. From each subset a 512 $`\times `$ 512 image array (pixel size 0.75$`\mathrm{}`$) was accumulated and flat–field corrected. In order to add consistently the field images acquired in the five separate observation runs and to compensate for small tracking inaccuracies within the single run, we selected a bright isolated source and used its centroid to coalign the 60 s images. The latter procedure, which improved the overall Point Spread Function (FWHM) from 4$`\mathrm{}`$ to 3$`\mathrm{}`$, was also applied to the light curve extraction process described below. A 2.8$`\mathrm{}\times `$2.8$`\mathrm{}`$ section of the field accumulated from the 6.3 h observation is shown in Fig. 1: the circles represent the position uncertainty (90% confidence) of ROSAT (30$`\mathrm{}`$) and ASCA (40$`\mathrm{}`$) X-ray detections.
From the sources found inside, or close to, the two position error circles we selected eight objects detected well above the sky background to be searched for variability. Note that the contribution of the detector noise to the background is virtually nil, being as low as 10<sup>-5</sup> counts s<sup>-1</sup> pix<sup>-1</sup> (sky background was $``$ 10<sup>-2</sup> counts s<sup>-1</sup> pix<sup>-1</sup>). From each object labeled in Fig. 1 (the candidates of Israel et al. Isr99 (1999) are reported with the same name, except star D which is too faint in our data) we extracted the source and background light curve binned at 1 s resolution. Great care was paid in the selection of the aperture radii used in the process, the field of 1WGA J1958.2+3232 being located in a crowded region near the galactic equator. The optimal source aperture radius was found at 2.25$`\mathrm{}`$ whereas for the background a concentric ring of inner and outer radii respectively measuring 9$`\mathrm{}`$ and 11.25$`\mathrm{}`$ was used. With the same procedure the light curves of three bright stars found well outside both ROSAT and ASCA error box were generated and, since a preliminary time series analysis excluded any intrinsic periodicity, they were used as comparison to remove any atmospheric effect.
To investigate the time variability, for each object we computed a Lomb-Scargle periodogram (Scargle Sca82 (1982)) on the combined data set. Some typical periodograms are shown in Fig. 2. Only star B shows a significant periodic signal. The periodogram shows two peaks, the most apparent corresponding to about 721 s period, consistent with the X-ray period. Star B is the source previously proposed as optical counterpart of 1WGA J1958.2+3232 by Israel et al. (Isr99 (1999)).
In order to look for longer periodicity, we planned new observations with the PC-ICCD at Asiago Observatory at the beginning of October but, due to poor weather conditions, we were able to collect only about 2<sup>h</sup>40<sup>m</sup> of data. The detector was operated in the highest time resolution mode (4.512 ms), corresponding to the widest dynamic range, but with a reduced field of view (a quarter of the sensitive area of the detector, resulting, with the Asiago plate scale, in 5.8’ $`\times `$ 1.45’). This choice was determined by the necessity to preserve, under higher count rates resulting from the better efficiency of the telescope, the linearity of the detector also for the brightest nearby stars, employed as comparison in data reduction of June observations. Thus, we were able to use the same comparison stars also for the October 1 run.
Observations in standard optical photometric bands have been carried out with the photoelectric photometer, again at the 91 cm telescope of the Catania Astrophysical Observatory. The photometer was operated in two different mode: i) in the standard mode to determine the U B V magnitudes and colours in the Johnson system, ii) in continuous acquisition at fixed filter, the U filter in our case. A diaphragm of $`\varphi `$ = 15 arcsec was adopted.
The V magnitude and colour indices of 1WGA J1958.2+3232 were determined calibrating on nearby field stars of known magnitudes and colours (e.g. HD 188992 V=8.29, B-V=-0.07, U-B=-0.38, and HD 189596 V= 7.54, B-V=-0.11, U-B=-0.44). Average seasonal atmospheric extinction and instrumental coefficients have been adopted to transform instrumental magnitudes to the Johnson system. The magnitudes and colours we obtain for 1WGA J1958.2+3232 are V= 15.713 $`\pm `$ 0.073, B-V = 0.231 $`\pm `$ 0.067, U-B = -0.784 $`\pm `$ 0.096.
Continuous monitoring observations on September 7 and 14, with the U filter and integration time of 5 sec for each measurement, have been done with the aim of improving the measured accuracy of the 12 m period. Uninterrupted sequences of 7000 second and of 9000 sec were obtained in September 7 and 14 respectively. The sky background was subtracted by linear interpolation of the values measured on ten samplings obtained at the beginning and at the end of each run. Correction for the atmospheric extinction have been made by adopting the average seasonal coefficient for the U filter.
No measurements of the sky background or of comparison stars have been made during the observing runs, to avoid interruptions of the data string and introduction of spurious frequencies in the power spectrum. This prevents us from using these data to study the long term variability.
The light curves of the source in the single nights, reduced as previously explained, are shown in Fig. 3.
A Lomb-Scargle periodogram was applied on the time series from separate nights, after removing low frequency trends in U band observations (carried out without comparison stars) by a polynomial fit on each data set. The periodograms are shown in Fig. 4: the best period and its first harmonic are marked by vertical dotted lines. In all but one observations a peak was clearly visible near the period of the X-ray modulation. The amplitude of the peak varies from night to night, but in the longest runs it is statistically significant. On the other hand, in the periodogram of June 13, there is only a hint of a peak. The different run length doesn’t justify the difference in the observability of the modulation: the peak of the longest run (October 1) is well below the 90% confidence level. In four observations a weaker peak is also present near the first harmonic of the X period, suggesting that pulse shape is time dependent and not simply sinusoidal. We notice also the presence of a peak corresponding to a period about twice the main period that can be seen in the periodograms of June 15 and September 7. In September 7, the confidence level for this frequency is greater than 99.999 %.
In order to improve the resolution in the main period measure, a Discrete Fourier Transform (DFT) was computed on the overall observation set (Deeming Dee75 (1975)), excluding the observation of October 1, because the modulation was very low. The spectrum is shown in Fig. 5, together with the spectral window. The clean spectrum, produced by the deconvolution of the raw spectrum with the spectral window by means of the CLEAN algorithm (Roberts et al. Rob87 (1987)), is also included in the lower panel of Fig. 5. The resulting main period is 727.06 $`\pm `$ 0.02 s, but, due to the complex spectral window, dominated by the 1/d aliasing, there is an ambiguity as to which of the alias peak is the true period. A MonteCarlo simulation showed that 1/d alias 721 and 733 s are acceptable (the resulting probability for the 727 s peak to be the true period is about 70 %, whereas there is about 30 % probability for either 721 s or 733 s peak). In particular, 733.24 $`\pm `$ 0.02 is compatible with the ASCA measure (734 $`\pm `$ 1 s), thus we consider this period as the most likely. In the CLEANed spectrum it is also apparent the presence of some weaker peaks,near 833 s and at some others lower frequency.
Instead, there is no significant peak near the half frequency of the main modulation. The FFT computed on the longest data set in which this modulation is apparent yields a best period determination of 1448$`\pm `$158 s. The DFT computed on both the data set showing this modulation doesn’t improve the accuracy. In fact, the June 15 observation is far shorter, with lower statistics and the large gap between the two observations results in a quite complex spectral window. In order to better estimate the period, the epoch folding technique (Leahy et al. Lea83 (1983)) was applied to the September 7 data, after removing the 733.24 modulation by subtracting a Fourier fit (in the epoch folding diagrams peaks are found at all the sub-harmonics of the signal frequencies). The resulting period was 1467 $`\pm `$ 25 s (the uncertainty has been derived from simulations), consistent with 2 $`\times `$ 733 s. After subtracting the long term variation, the light curves of September 7 and 14 were folded at twice the 733.24 s period (Fig. 6). A visual inspection of the folded light curves confirms the presence of an even-odd effect, more evident in the September 7 pulse shape.
### 3.2 The long term variability
Low frequency, high amplitude variations are clearly apparent in the longest differential photometry runs and from night to night. For example, the mean flux on October 1 is 50% lower than the mean flux averaged on the five nights of June. Moreover, on June 17, the source intensity increases by about 50% during $``$ 1<sup>h</sup>20<sup>m</sup>. In both observations carried out with the photometer, without comparison stars, the data show long term trends not easily attributable to instrumental or transparency changes. The data of September 7 show a slightly increasing trend from the beginning to the end of the observations, even before the correction for the atmospheric extinction is applied. After the atmospheric extinction correction, the average increase is about 15 %.
The data of September 14 show a steady increase in the first half of the observations, up to about the 40 %, and then a decrease to about the initial level. Although the changes in September 7 night could result from an improvement of the sky transparency, the trend observed on September 14 is more difficult to explain in the same way. In both cases the observations were made after the object meridian passage, therefore, the extinction effect should increase with time, contrary to the observed trends. The sky was very clear on both nights, so that an increase of the transparency of about the 40 %, as shown by the data of the second night, is very unlikely, also because the subsequent observations made during the rest of the night do not support this possibility. Changes on the photometer gain are excluded by the stability of the photon counting system over several years. Also changes in the sky background can be excluded because we were in dark period, and the observing direction is free of any close city light. The telescope was manually guided all the time through an intensified TV camera using an off-set nearby star. Normally the correction of the telescope wiggle produces random noise, or loss of signal, but no long term trends. We are, therefore, inclined to attribute the observed long term variation to real variations of intensity of 1WGA J1958.2+3232.
A peak is apparent in the DFT at low frequency (the second highest peak in the CLEANed spectrum showed in Fig. 5), corresponding to a period of 2.45<sup>h</sup> (with alias at 2.73<sup>h</sup> and 2.23<sup>h</sup>) with modulation of about 7.5%. However, due to the poor sampling, we are not able to reliably ascribe it to an orbital period. In fact the period corresponding to the peak is only slightly shorter than the two longest runs.
## 4 Discussion
Our observed B-V and U-B color indices are not compatible with colors of any unreddened normal stellar object. If it was a reddened object, adopting the Cygnus reddening law $`E(UB)=0.9E(BV)`$, 1WGA J1958.2+3232 would come out to be a hot star of spectral type about O5, with E(B-V) $``$ 0.6. Assuming a standard reddening law, the visual interstellar extinction $`A_V=3.1E(BV)=1.86`$ mag would lead to a distance of 1.5-2 Kpc, not compatible with the absence of interstellar absorption feature in Na I D2 lines and the weak interstellar feature observed in the Ca II K line (Negueruela et al. Neg99 (2000)), which are in agreement with a far lower distance, not above 500 pc.
In fact 1WGA J1958.2+3232 in the color-color diagram lies on the Black-Body and white dwarf region. According to our measurements (B-V = 0.231 $`\pm `$ 0.067 and U-B = - 0.784 $`\pm `$ 0.096), it falls a little above the B-B line and the main distribution of DA white dwarfs: it should be an object of about T=10,000 K whose optical emission comes from the disc accretion region.
Large modulation at the 733 s period (the same as measured in X-ray) was found in almost every night, even if with a variable amplitude. Weaker power peaks are present only in two data sets, but further measures are needed to confirm their significance. The detection of periodical optical modulation at the X-ray period rules out any possibility that 1WGA J1958.2+3232 could be a binary system hosting a Be, because the optical emission should be dominated by the early type star, whereas the modulated component comes from the compact object. A stable optical period, usually in the range from some tens of seconds to some tens of minutes, and X-ray modulation at or close to the same period, is a signature of the Intermediate Polar class of Cataclysmic Variables.
Generally, in the Intermediate Polar power spectra, the main power is found at the spin period in X-ray (with the exception of RX1712-24, Buckley et al. Buc97 (1997)), whereas the optical light curves can be dominated by the spin (e.g. V709 Cas, Norton et al. Nor99 (1999), BG CMi, De Martino et al. DM95 (1995), UU Col, Burwitz et al. Bur96 (1996)), or by a sideband (e.g. AO Psc, Hellier et al. Hel91 (1991), V1223 Sgr, Jablonski & Steiner Jab87 (1987), RX J1712-24, Buckley et al. Buc95 (1995)), or by the orbital modulation, depending on the strength of the magnetic field and the accretion geometry and rate. In particular, differences in the accretion rate can produce variation in the dominant frequency (e.g. FO Aqr, which in the past was dominated by the spin period, De Martino et al. DM94 (1994), recently was found with the orbital modulation dominating over all the other periods, De Martino et al. DM99 (1999)).
In the X-Ray light curve measured by ROSAT, Israel et al. (Isr98 (1998)) found modulation at the same period that we see in optical and at its first harmonic, whereas they didn’t find any evidence of a period shorter then the main one. Thus, the dominant period in X-rays and the optical should be the spin of the white dwarf, rather than a sideband of a shorter period. The dominance of the spin period and the non-detection of sidebands neither in X-ray nor in optical light curves suggest that 1WGA J1958.2+3232 is a disc accretor.
The amplitude of the modulation varies greatly from night to night, but this is a rather common behavior for the Intermediate Polars class (see, for example, Welsh & Martell Wel96 (1996)). Especially in the optical domain, IPs often show night to night variations in the relative amplitudes of the power spectrum at the characteristic frequencies. The presence of power in the DFT at the first harmonic of the 12 min period, in some nights only, suggests that the pulse shape could be variable.
The presence of another peak at half frequency suggests that the true spin period could be 1466 s: then, the 12 min period would be the first harmonic and the peak sometimes seen at 368$`\pm `$7 s would correspond to the third harmonic. Among the Intermediate Polars there are three systems (DQ Her, YY Dra and V405 Aur, Allan et al. All96 (1996)) which have optical and X-ray light curves dominated by the modulation at the first harmonic of the spin period, whereas at the fundamental frequency the modulation is very weak, or absent. However, all these objects are characterized by a fast spinning white dwarf (142 s, 529 s, 545 s respectively, Ritter & Kolb Rit98 (1998)). Commonly, the double peak pulse shape is interpreted as a signature of two poles accretion, but Norton et al. (Nor99 (1999)) point out that two poles accretors generally produce single peaked profiles (e.g. in EX Hya there is strong evidence of two poles accretion but the pulse profile is single peaked, Hellier Hel95 (1995)). They outline a model according to which, among the two poles accretors, only short period IPs (with a period below 700 s) could exhibit a double peaked pulse profile. For another system (BG CMi) there has been some uncertainty whether the dominant period (913 s) is the spin period or a sideband of its first harmonic (Patterson & Thomas Pat93 (1993)), but successive observations did not find any evidence confirming the longer 1693 s period (Garlick et al. Gar94 (1994), De Martino et al. DM95 (1995), Hellier Hel97 (1997)).
If 24 min were the true period, 1WGA J1958.2+3232 would be the slowest rotator which exhibits double peaked profile. In the Norton model it should have a single peaked pulse shape even if the accretion takes place on two poles. However, we point out the very high significance level of the 24 min peak in September 7 night and the evident even-odd asymmetry in the folded light curve in Fig. 6.
In conclusion, in spite that the evidence for this longer period is present only in one observation run and in spite that the run cover only less than 5 periods, we tend to identify the 24 min period as the true spin period, but further observations (in particular, time resolved spectroscopy) are needed in order to confirm this outcome.
In the time series from October 1 observation, the modulation at 12 min is very low and the corresponding peak in the periodogram is well below the 90% confidence level. In the same night, the mean flux is 20% lower than the flux averaged on the June observations and in the middle of the run a dip is present: a possible explanation of this behavior could be a partial eclipse of the accreting disc.
Regarding the long term variability, we are not able to ascribe it to an orbital modulation, because the low frequency peaks in the DFT correspond to long periods, poorly sampled by the observations. The presence of power at low frequency could be produced by instabilities in the accretion disc or curtain.
## 5 Conclusions
We have presented time resolved optical photometry of the stars included in the field of the X-ray pulsating source 1WGA J1958.2+3232. A stable, high amplitude, periodic modulation was found in the light curves of the star previously proposed by Israel et al. (Isr99 (1999)) as the optical counterpart of 1WGA J1958.2+3232. The 12 min period, compatible with the X-ray one, is likely associated with the first harmonic of the spin of a white dwarf in an Intermediate Polar Cataclysmic Variable. If this outcome will be confirmed, 1WGA J1958.2+3232 would be, among the Intermediate Polars, the slowest rotator which exhibit double peaked spin pulse shape.
###### Acknowledgements.
The authors are grateful to Domitilla de Martino, Gian Luca Israel, Sandro Mereghetti, Pablo Reig and Luigi Stella for helpful conversations and to Lucio Chiappetti and Constantinos Paizis for a critical reading of the manuscript. Financial support from the Catania Astrophysical Observatory and the Regione Sicilia, is gratefully acknowledged. This work was partially supported by the European Commission under contract ERB FMRXCT98-0195 and by Agenzia Spaziale Italiana (ASI).
|
warning/0005/astro-ph0005099.html
|
ar5iv
|
text
|
# The low surface brightness extent of the Fornax Cluster
## 1 Introduction
Over the last few decades new populations of galaxies have revealed themselves through the use of high sensitivity detectors at dark observing sites. From dwarfs to giants, galaxies of Low Surface Brightness (LSB) are a strikingly new addition to the galactic inventory of the Universe (Davies et al. 1999, Impey & Bothun 1997). Having identified this new class of object we have now moved on into the era of defining their cosmological importance in relation to the brighter, more well known galaxies. To do this it has been necessary to survey significant areas of sky at LSB levels (e.g. Morshidi-Esslinger et al. 1999a, Morshidi-Esslinger et al. 1999b, Bernstein et al. 1995, Schwartzenberg et al. 1995, Impey et al. 1996, Dalcanton et al. 1997, Kashikawa et al. 1998). In this paper we extend this survey work to a study of the nearby Fornax cluster and its immediate environment (see also Caldwell 1983, Davies et al. 1988, Ferguson 1990, Irwin et al. 1990, Bothun et al. 1991, Held & Mould 1994, Morshidi-Esslinger et al. 1999a, Morshidi-Esslinger et al. 1999b).
Two fundamental cosmological issues are the spatial scale over which mass is distributed in the Universe and the mass spectrum of the collapsed objects within the Universe. Given that dark matter dominates the mass the challenge is to measure these quantities for the dark, not visible, matter. One obvious solution is to measure the spatial distribution and mass spectrum of the most dark matter dominated galaxies we know of: these are the LSB and dwarf galaxies (McGaugh & de Blok 1998, Irwin & Hatzidimitriou 1995, Carignan & Freeman 1988, Dekel & Silk 1986).
Historically the intrinsically bright and high surface brightness galaxies have been used to define the spatial scale and distribution of structures in the Universe (see for example Maddox et al. 1990, da Costa et al. 1998). Their clustering length scale, as measured by such techniques as the two point spatial correlation function, defines the bright galaxy mass scale. The observed clustering scale is smaller than that inferred from simulations of structure growth using currently popular CDM models of dark haloes (see for example Frenk et al. 1996, Jenkins et al. 1998) and so it is often assumed that the bright galaxies are biased tracers of the underlying mass (Dekel & Rees 1987, Frenk et al. 1996). In these models the bright galaxies are more concentrated than the dark matter because they form at the peaks of the initial density fluctuations. They also undergo more mergers than the objects in the lower density outer regions of clusters and so they can become more ’visible’ due to enhanced star formation. In this paper we intend to show that, for one nearby cluster at least, there is a population of very LSB low luminosity galaxies that contribute about the same luminosity to the cluster as the brighter galaxies, but which defines a much more extended spatial scale than the bright galaxies. Our expectation is that these galaxies are dark matter dominated (cf. the ”dark galaxy” of Carignan & Freeman 1988), having mass-to-light ratios of 100 or more similar to the mass-to-light ratios determined for galaxy clusters (David et al. 1995).
It is also difficult to relate the mass spectrum inferred from numerical simulations of structure growth (for example CDM) to what we observe. The simulations model the mass spectrum of the dominant dark matter haloes yet what we observe is the luminosity function of galaxies. These two are related through the total mass-to-light ratio of each object, but the mass-to-light ratio is almost certainly a function of galaxy luminosity, type etc. and so relating luminosity to total mass is not easy. In addition, various forms of evolution (mergers, tidal stripping, star formation, galactic winds etc.) may have altered the mass spectrum over time. To overcome these problems numerous N-body simulations incorporating ever more complicated physics have been carried out in an attempt to relate the initial conditions (Press & Schector 1974) to what we observe (see for example Frenk et al. 1996, Kauffman et al. 1997). Even with these complications the models of hierarchical structure formation with galaxy evolution still predict a steep faint end slope to the local mass function, similar to the presumed initial conditions. This appears to be a rather robust conclusion. The local cluster galaxy mass function predicted by the models has a low mass slope of $`2`$ , (Frenk et al. 1996). This is much steeper than that generally found for the luminosity function of field galaxies (Loveday et al. 1992, Marke et al. 1994, but see also Loveday 1997). However there are a number of recent determinations of a steep luminosity function for clusters (Phillipps et al. 1998, Valotto et al. 1997, Smith et al. 1996, Bernstein et al. 1995). We will find observationally for the Fornax cluster a steep cluster luminosity function consistent with the expected mass spectrum inferred from N-body simulations and show that the steep slope extends to intrinsically faint magnitudes ($`M_B12`$).
Our prime objective in this paper is to compare the spatial distribution and total luminosity of three samples of Fornax cluster galaxies selected by their surface brightness. We define a bright galaxy sample typical of that used to measure the two point spatial correlation function and the galaxy luminosity function, a LSB sample typical of recent work to determine the numbers of LSB dwarf galaxies in clusters and a very LSB (VLSB) sample typical of what can now be detected using a large area CCD. The bright and LSB samples have been discussed previously by us (Disney et al. 1990, Morshidi-Esslinger et al. 1999ab, henceforth MDS). The VLSB sample is new and so we will start by describing these data.
## 2 Observation and image processing of the VLSB sample
The data were obtained using the Curtis Schmidt 96cm Telescope at the CTIO during the nights of 21-24th of October 1997, using the 2048$`\times `$2048 CCD with 2.3 $`arc`$ $`sec`$ pixels. The total area observed was 13.8 $`sq`$ $`deg`$, forming a strip $``$9 $`deg`$ long outwards from the cluster center (NGC1399) towards the lenticular barred galaxy NGC1291 (figure 1). We chose this particular strip of sky because MDS (the LSB sample, see below) had identified numerous quite large LSB galaxies in this area of sky around NGC1291.
There are 13 individual fields in the strip, each of $`1.3^o\times 1.3^o`$. Each field consists of four exposures of 1050sec, median stacked to give good signal to noise, while keeping saturation to a minimum. The different exposures were ”jittered” around the nominal field center to improve flat fielding.
### 2.1 Data reduction
The CCD uses two read out amplifiers, so all steps of data reduction were performed twice, once for each amplifier. The original images (data, bias frames, flats and standards) were cut into two parts (according to the reading amplifier) and then each side was processed separately. The data remained separated for the whole data reduction and detection process.
Prior to median stacking, the image data were bias subtracted and flat fielded using the standard techniques. After the stacking and trimming, the size of each field was decreased to $``$1 sq deg. It is very important for the automated detection of LSB and faint galaxies that the sky background is as flat as possible across each frame. Local background variations could lead to the detection of spurious objects in those areas where the sky appears brighter than its global value. In the opposite sense, real LSB objects would not be detected in regions where the local sky is fainter than the overall value (Davies et al. 1994). To overcome this problem we have modelled the sky background on each frame and then subtracted this from the data. The technique consists simply of creating a map of the background sky in which each pixel takes the median value of the pixels in the original frame in an n$`\times `$n box surrounding the pixel in question. This map is then subtracted from the original data. The choice of box size is critical, as it will constrain the maximum size of objects that can be detected. Any object with a size approximately half or greater than the box size will be lost or at least severely degraded. The large scale size of the background variations over the frames allowed for a large smoothing box size. A 340$`\times `$340 pixel box was used. Considering the 2.3 $`arc`$ $`sec`$ pixels of the CCD, this translates to objects with angular sizes of about 220 $`arc`$ $`sec`$, very much larger than the objects we might expect to detect and our size selection criteria (section 2.2, below). Previous work (Davies & Phillipps 1989) indicates that Fornax cluster LSB galaxies should have maximum diameters of about 50 $`arc`$ $`sec`$. We will show below that all objects detected are much smaller than this upper size limit. This technique allowed us to flatten the sky to $``$ 0.5% which equates to a 1$`\sigma `$ fluctuation of $``$26 $`R\mu `$.
Calibration was carried out using Landoldt standard stars (Landolt 1992) and the standard Harris R filter used for the observations. The observing conditions were not ideal (some high cloud) and we estimate a magnitude error of 0.2, which is sufficiently small for the comparison we intend to make. The sky brightness measured from the frames was $``$20.5 R$`\mu `$, a value typical for the location, season and lunar age. Finally, astrometry was carried out using the ESA Guide Star Catalogue and is accurate to about one pixel, 2.3 arc sec. We used this astrometric calibration to measure distances from NGC1399 (the central cluster galaxy) and to obtain surface densities of galaxies in annuli around NGC1399.
### 2.2 Image detection and object selection
The image detection was carried out in two steps. First the frames were scanned for objects using the SExtractor package (Source Extractor, Bertin & Arnouts 1996). All objects with 5 connected pixels brighter than 26 $`R\mu `$ (the brightest 1 sigma isophote from all of the frames) were included in the catalogue. SExtractor also includes a neural network trained program that classifies each object with a ”stellarity index” between 0 (galaxy) and 1 (star). A total of 322,505 objects were detected in this way, giving an average of 23,200 objects per sq deg. The objects were detected by their size with the faintest having an isophotal magnitude of 22.8 $`R\mu `$. The SExtractor catalogue was then used as the input into our selection program this eliminated any flagged objects, such as ones close to the edges of the frame, near bright stars and composite objects. Also, all objects indexed 1 (definite star) were eliminated at this stage. The catalogue was thus restricted to non-flagged objects with stellarity index up to 0.99 that met the selection criteria. With such large pixels it is difficult to define a star, but the second stage selection described below shows how we finally selected only very extended objects.
The final selection criteria are crucial because we wish to define a sample of galaxies that are predominately nearby cluster members and not distant background galaxies. Our first criteria was signal-to-noise ratio. To be sure we had real detections we set a minimum signal-to-noise ratio of 10. This equates to a minimum area of $`28`$ pixels (148 sq arc sec). Using numerical modelling MDS showed that by selecting objects of LSB (in this case $`\mu _o^B>22.5`$) and with relatively large exponential scale sizes ($`\alpha >3\mathrm{"}`$) a cluster sample was predominately selected. This conclusion was also reached independently by Schwartzenberg 1996 and Phillipps et al. 1998 (we will return to the question of background contamination in section 3). We also set an upper central surface brightness limit of 23 $`R\mu `$ so that we have a disjoint surface brightness distribution from the LSB sample (see section 3). So our final selection criteria for the VLSB sample were:
1. $`\mu _o23`$ $`R\mu `$
2. $`\alpha 3.0`$ arc sec
3. Isophotal area $`148`$ sq arc sec
Plotting the isophotal magnitude of each object against the logarithm of its isophotal area we can draw lines of constant surface brightness and exponential scale size and select only those objects satisfying the above selection criteria (fig. 2).
Using these criteria, 3462 objects were selected, an average of 251 objects per sq deg, less than 1$`\%`$ of the original number of detections.
### 2.3 Photometry
We have used the image parameters produced by SExtractor to determine the photometric properties of each object in the final catalogue. We will assume that the objects detected have exponential light profiles (Davies, 1989).
The mean intensity of a detected object is
$`<I>=\frac{L_{iso}}{A_{iso}}`$
where $`L_{iso}`$ and $`A_{iso}`$ are the isophotal luminosity and area respectively,
$`L_{iso}=2\pi I_o\alpha ^2f`$
and
$`A_{iso}=\pi r_L^2`$
where the fraction of the total light contained within the limiting isophote is
$`f=1e^x(1+x)`$
and
$`x=\frac{r_L}{\alpha }=\frac{\mu _L\mu _o}{1.086}`$.
$`\mu _L`$ is the limiting surface brightness in magnitudes per sq arc sec and $`r_L`$ is the limiting size at $`\mu _L`$. This leads to the following relationship between the measured mean surface brightness and the exponential central surface brightness.
$`<\mu >=\mu _o2.5\mathrm{log}f+5\mathrm{log}x0.75`$
Thus from the isophotal magnitude and area calculated by SExtractor we can calculate the central surface brightness $`\mu _o`$ (from above), the total magnitude using
$`m=m_{iso}+2.5\mathrm{log}f`$
and the scalelength using
$`\alpha =10^{\frac{\mu _om2}{5}}`$
It is worth commenting about the influence the large pixels may have had on our determination of the photometric parameters. We do not fit surface brightness profiles. Instead we derive the exponential parameters that correspond to the isophotal luminosity and area measured by SExtractor. Errors would arise if background pixels were included or excluded in the measured area because of the intrinsic fluctuations in the background. The background fluctuations ($`\sigma `$) are at about the same level as the limiting isophote so, the probability of including or excluding an extra pixel is at most 0.2 for Gaussian fluctuations ($`0.2^n`$ for n pixels). Including or excluding one extra pixel to the area, and ignoring the small contribution to the luminosity leads to an error in the surface brightness of only about 0.05 magnitudes even for the smallest galaxies detected. This will have little effect on our results.
## 3 Results
### 3.1 The three samples defined by their surface brightness and magnitude
In what follows we will compare the clustering scale and the contribution to the total cluster light of three samples selected by their surface brightness. These are three progressively fainter central surface brightness samples. A bright galaxy sample from Jones & Jones 1980, a LSB sample from MDS and our VLSB sample describe in this paper. We will also convert the VLSB sample surface brightnesses and magnitudes into the B band using (B-R)=1.5, a value typical of Fornax cluster dE galaxies (Evans et al., 1990), so that we can compare this sample with the bright and LSB samples.
Jones & Jones list 64 Fornax cluster galaxies with m$`{}_{B}{}^{}16`$. We have previously measured the surface brightness distribution of these galaxies (Disney et al., 1990) and we reproduce this in fig. 3. This sample illustrates the familiar Freeman result (Freeman, 1970). The distribution of surface brightness is sharply peaked at a value of $``$21.5 B$`\mu `$ with an rms width of $``$0.4 B$`\mu `$. These galaxies will be used to define the luminosity distribution as delineated by the bright galaxies.
The Fornax cluster LSB sample, from the catalogue of MDS, is a small part of a survey covering some 2400 $`sq`$ $`deg`$ of sky. The survey was carried out using UK Schmidt telescope photographic plates scanned by the Cambridge (U.K.) Automated Plate Measuring machine. All objectes were objectively selected to satisfy the selection criteria:
1. $`\mu _o`$$``$22.5 $`B\mu `$
2. $`\alpha `$$``$3.0 $`arc`$ $`sec`$
This sample is extensively discussed in MDS. The surface brightness distribution is disjoint from that of the bright galaxy sample with a peak at about 23.5 B$`\mu `$ (fig. 3).
The VLSB sample has been discussed in section 2, and its surface brightness distribution is again shown in fig. 3. Note that fig. 3 should not be seen as a true distribution of surface brightness for galaxies over some 7 magnitudes. Each sample was selected in a quite different way, at different isophotal levels against different sky backgrounds. Each will be incomplete for galaxies with central surface brightnesses close to the isophotal limit and the higher surface brightness galaxies in each sample will be selected against because at a given magnitude they will be smaller and so they will not satisfy the size criteria. We have made no attempt to correct the sample for surface brightness incompletness because the true distribution of surface brightness is not known. Comprehensive discussions of how such corrections may be made are given in Disney & Phillipps 1987, Davies 1990 and McGaugh et al. 1995. We will comment further on incompleteness in section 3.3.
The three samples also show a clear trend in apparent magnitude with the lowest surface brightness galaxies having the faintest magnitudes. The mean apparent magnitudes of the three samples are approximately 14.2, 18.1 and 19.8 respectively ($`M_B`$=-17.1, -13.2 and -11.5 for a distance modulus of 31.3, Ferguson 1990).
### 3.2 The radial distribution of the three samples
In this and the following section we will assume that all the galaxies in each sample are cluster members. We will justify this assumption at the end of this section. Using the three data sets described above we can measure the decrease in surface number density from the cluster centre, for galaxies of different surface brightnesses (fig. 4). The number density data for the bright and LSB samples are taken from MDS. The data for the VLSB are binned over annulii of 0.25 deg, with all radial distances measured from the central Fornax galaxy NGC1399. Fitting an exponential to the data we find scale sizes of $`0.5\pm 0.1`$, $`1.3\pm 0.4`$ and $`1.9\pm 0.3`$ $`deg`$ for the bright, LSB and VLSB samples respectively (The central surface density of the VLSB sample is flat within the inner 1 deg and so we do not fit over this region. Including these points would give an even longer scale length). These correspond to approximately 0.16, 0.42 and 0.64 $`Mpc`$ for a Fornax distance of 18.2 $`Mpc`$ (Ferguson 1990). The LSB and VLSB samples are almost 3 and 4 times respectively more spatially extended than the bright galaxy sample. The result for the LSB sample (clustering scales some three times larger than the bright galaxies) has also been shown to be true using a much larger sample of galaxies, over a larger area of sky that includes other nearby galaxy groups (see MDS). It is also consistent with previous work on the spatial extent of more luminous LSB galaxies - they are less strongly clustered than the brighter galaxies (Mo et al. 1994). The exponential fit central galaxy number densities are approximately 40 (HSB), 10 (LSB) and 800 (VLSB) per sq deg. This implies a sharp rise in the luminosity function at the faint end (slope of $`\alpha 2`$, see below).
### 3.3 The contribution to the total luminosity of the cluster
Having derived the parameters of the exponential fit to the number density distributions we can integrate to determine the total number of cluster galaxies expected in each sample. This corresponds approximately to the total numbers we actually have in the bright and LSB samples. For the VLSB sample the data only extend to large enough distances on one side of the cluster and so the integral takes account of the area not included. We can then use the mean luminosity of each sample to make a crude estimate of the ratio of total luminosities produced at each surface brightness. These are 1.0:0.1:0.5 for the bright:LSB:VLSB samples respectively. Adjusting the integral to account for the flattening of the VLSB number density within the central 1 deg makes little difference to the total luminosity (0.5 of the total luminosity becomes 0.45). Integrating to $`4^o`$ rather than infinity reduces the value from 0.5 to 0.3 of the total. Subtracting a background of 100 galaxies per sq deg (see below) alters the 0.5 to 0.3. If the cluster population ends at a radial distance of $`4^o`$ and there is a background contamination of about 100 galaxies per sq deg then the VLSB sample contributes about 0.1 to the total cluster luminosity. Over the inner $`2^o`$ radius of the cluster the total luminosity ratio is 1.0:0.05:0.2 ignoring any background contribution. We will argue below that effectively all of the VLSB galaxies belong to the cluster and so we conclude that the VLSB galaxy population contributes a significant fraction of the total cluster light.
It was our initial intention to use the range of luminosities in the VLSB sample to define the faint end of the luminosity function, but it is not possible to do this because of the way the sample has been selected, which itself was a consequence of trying to keep the background contamination to a minimum. The combination of the surface brightness and scale size limit forces a surface brightness size relation upon the data (see fig. 2). At a given size we preferentially select faint galaxies because these are the ones that have low enough surface brightness. This would have had the effect of making the luminosity function appear steeper than it really is (this will affect other samples selected in the same way, Schwartzenberg 1996). We cannot correct for this because we do not know the number of cluster galaxies that satisfy the individual surface brightness limits, but fail to be selected because of their size. For example Drinkwater et al. 1999 have identified (using redshifts) Fornax cluster galaxies with approximately the same surface brightness as the galaxies in the bright sample, but with much smaller sizes and fainter magnitudes. These were not included in the Jones & Jones 1980 magnitude limited sample.
What we have done is to estimate a lower limit to the contribution made by the LSB and VLSB samples to the luminosity function. We have done this by simply using the mean magnitude (standard deviation is about 0.8 mag) and total numbers (integrated to infinity) of each sample. We have then normalised this to the total numbers of bright galaxies. In fig. 5 we have scaled our numbers so that the bright galaxy numbers correspond to the normalised luminosity functions given in Smith et al. 1996 (converted to the B band). Smith et al. 1996 show that the luminosity function of three more distant clusters steepens beyond $`M_B19.5`$ with a faint end slope of $`1.8`$. Our data (lower limits) is consistent with this steep slope and extends this result by some 3.5 magnitudes to $`M_B12`$. It is consistent with a similar result obtained by Phillipps et al. 1998 for the Virgo cluster.
### 3.4 Contamination by non-cluster galaxies
The above result depends on there being little contamination of our sample by non-cluster galaxies. The radial surface density plots of course indicate a substantial number of cluster galaxies, but there is still the possibility of contamination of the sample by non-cluster galaxies.
The bright sample is one where cluster membership has been assigned because of a measured redshift and so there is no ambiguity. Possible background contamination of the LSB sample has been extensively discussed in MDS. The conclusion was that at most 20% of the galaxies could be in the background.
To try and assess the background contamination of the VLSB sample we have analysed the surface density of galaxies over the full $`9^o`$ extent of the data. The data terminate with one field centred on the peculiar lenticular galaxy NGC1291 (fig.1). Our original motivation for choosing this strip was that MDS had identified a number of relatively large LSB galaxies in the vicinity of NGC1291. A visual inspection of the field around NGC1291 on the CCD also clearly indicates a large number of LSB galaxies.
In fig. 6 we show the normalised surface density of galaxies as a function of distance from the centre of Fornax extending to $`9`$ $`deg`$. The most striking feature is the initial decline and then large increase in galaxy numbers as NGC1291 is approached. NGC1291 appears to be at the centre of a cluster of LSB low luminosity galaxies ! One can interpret fig. 6 in two ways. Either the Fornax cluster population comes to an end at a radius of about 4 deg and there is a background surface density of order 100 galaxies per sq deg or the two extended LSB populations of Fornax and NGC1291 overlap at about 4 deg.
Numerical simulations of the background contamination of the LSB sample by MDS predict about one background galaxy per sq deg with scale sizes greater than 3 arc sec. The same simulation but with the lower surface brightness cut-off of the VLSB sample predicts zero background galaxies per sq deg (the simulation assumes a flat faint-end to the luminosity function (Marke et al. 1994) and a Gaussian surface brightness distribution with a peak at Freeman’s value (Freeman 1970)). It is easy to show why this is so. A typical L\* spiral galaxy would have to be at a redshift of $`z0.3`$ to have sufficient cosmological dimming to enter the VLSB sample. At z=0.3 its scale size would be less than an arc second. To have an observed scale length of 3 arc sec it would need to be at $`z0.07`$ at which point it’s surface brightness is cosmologically dimmed by only about 0.5 magnitudes. The other option is that the field galaxy luminosity function is similar to that of the cluster ie. that beyond about 4 deg the galaxies in the VLSB sample are no longer associated with the cluster. We believe that there is observational evidence against this idea.
Observations of the general field have been made by Davies et al. 1994 and Dalcanton et al. 1997. Both of these surveys were optimised to detect LSB galaxies in the field with similar surface brightnesses and scale sizes to the galaxies in the VLSB sample. The result from Davies et al. 1994 was an upper limit of 10 per sq deg and from Dalcanton et al. 1997 7 per sq deg. This is far fewer than the 100 per sq deg suggested by our data. So, our conclusion is that these are two overlapping extended haloes of dwarf LSB galaxies about Fornax and NGC1291. We believe that the VLSB sample is essentially a Fornax cluster sample and that our three samples are not adversely affected by background contamination.
### 3.5 NGC1291
NGC1291 is not classified as a Fornax member. It is a peculiar lenticular barred galaxy in the field between us and Fornax at a velocity of 839km $`s^1`$ (De Vaucouleurs, 1975). The mean velocity of the Fornax cluster is 1366km $`s^1`$ (Ferguson, 1990). In the same way as we fitted an exponential to the surface density of galaxies around Fornax we have derived the exponential scale length of the surface density around NGC1291. The value is $``$ 1.4 +/- 0.1 $`deg`$ which equates to 0.22 $`Mpc`$ at the distance of NGC1291 (The Fornax scale length is 0.64 $`Mpc`$). This implies an extended, but not ridiculously large size for a halo of dwarf companions about an isolated galaxy (Zaritsky et al. 1993). We estimate the total luminosity of the VLSB population to be about 0.7 that of NGC1291 (apparent B magnitide of 9.4). This is a marginally larger number than Fornax, but the observations extend some 1.5 magnitudes further down the luminosity function and so we might expect about 4 times more galaxies if the luminosity function carried on with a slope of -2.0, each with about 4 times less luminosity. This suggests that if you could identify an ultra-LSB population you would conclude that there is more light in the extreme LSB constituents of the cluster than in the bright galaxies. Of course, at some point the luminosity function must turn over or terminate.
It is interesting to calculate the approximate mass contained within the VLSB galaxy population surrounding NGC1291. Irwin & Hatzidimitriou 1995 find that typical (M/L) ratios for dwarf galaxies lie within the range 10-100, whilst that for the central, optical, part of a ’normal’ galaxy is in the range 1-10 (e.g. Kent 1987). Given our above estimate that the total amount of light emitted from the VLSB galaxies is comparable to that from the central galaxy, over ten times as much mass may reside in the VLSB companions. This is comparable to the halo mass of $`2\times 10^{12}`$ $`M_{}`$ determined by Zaritsky et al. 1993 for isolated spiral galaxies. Thus the agreement between the scale-length of the NGC1291 dwarf population and the halo size estimated by Zaritsky et al. 1993 together with that between the halo mass and total mass of dwarfs, indicates that the ’dark halo’ of some spiral galaxies could be attributed to previously undetected VLSB galaxies.
The VLSB galaxies around NGC1291 may have a primordial origin, but another explanation may be found in the morphology of the bright galaxy; it is a peculiar galaxy. Its morphological classification ranges from ”the remarkable barred lenticular galaxy” (De Vaucouleurs, 1975) to a ringed SBa (NED) it has a dual personality because it also appears to be known as NGC1269. It is quite possible that NGC1291 has relatively recently undergone a merger event and that some or all the VLSB galaxies are the debris from this collision.
## 4 Discussion
The existence of such large numbers of previously undetected galaxies is an important result and ones first thoughts are to other observations that might contradict it. One concern is that we do not seem to find very large numbers of low luminosity VLSB galaxies in the Local Group. The observed luminosity function of the Local Group is relatively flat over the range of magnitudes sampled by our data ($`\alpha 1`$, van der Bergh 1992). Recently Blitz et al. 1999 have suggested that the Local Group mass function is really quite steep. They have proposed that the high velocity clouds detected at 21cm are in fact the remains of the smaller dark haloes predicted by the numerical simulations. This remains to be confirmed. It may be that all groups have flat luminosity functions. Muriel et al. 1998 have compared the luminosity function of galaxies in groups with those in clusters. They found that the luminosity function faint-end slope of the groups was much flatter ($`\alpha 1.0`$) than that of the clusters ($`\alpha 1.4`$). So the apparently flat luminosity function of the Local Group compared to a cluster like Fornax may not be unusual or a concern.
With an observed central galaxy number density of about 450 per sq deg we predict a diffuse intra-cluster light of about 31 $`B\mu `$ at the centre of the cluster. Such low surface brightness levels would be very difficult to detect against the relatively bright foreground sky (at best 23 $`B\mu `$). It is unlikely that a large population of low luminosity VLSB galaxies like this could be detected by their integrated light. Inter-cluster stars have previously been identified in Fornax by Theuns & Warren 1997. These stars may be associated with the cluster low luminosity VLSB galaxies. Considering the selection effects and an assumed stellar population Theuns & Warren 1997 estimate that these stars could account for 40% of the total cluster light a similar value to that derived by us for the VLSB galaxy population.
There has been an on-going debate about the nature of the objects that give rise to the absorption features seen in the spectra of quasars (Lanzetta et al. 1999). The issue is whether the Lyman limit and the damped Ly$`\alpha `$ absorption features arise in normal galaxies (in which case they have very large haloes) or in a previously unseen population of VLSB galaxies that are companions to the brighter galaxies (Phillipps et al. 1993, Linder 1998). The Fornax VLSB population has a total cross-sectional area on the sky about three times greater than that of the brighter galaxies (equating isophotal sizes). If the VLSB galaxies have sufficiently high gas masses then one would expect the majority of quasar absorption features, seen in sight lines through the cluster, to occur in these small galaxies and not in extended haloes of the brighter galaxies. Recently Impey et al. 1999 have compared absorption line features seen in the spectra of quasars behind the Virgo cluster with known cluster galaxies. They point out the ambiguity in trying to assign a given absorption feature with a particular galaxy. They also show that as you select Virgo galaxies of fainter magnitudes you are much more likely to conclude that a low luminosity galaxy is the absorber rather than a bright galaxy with a very large halo. Impey et al. 1999 also find that the correlation amplitude of the Ly<sub>α</sub> absorbers is 4-5 times smaller than that of the brighter galaxies, consistent with the absorption lines occuring in an extended VLSB galaxy population similar to that described here. Observations of Local Group dSph galaxies indicate low gas masses, $`10^7`$ $`M_{}`$ (Young & Lo 1997), but because of the small size of the systems the column densities are not extremely low $`10^{19}10^{20}`$ $`atoms`$ $`cm^2`$. This is typical of what one might expect through the outer regions of a galactic disc. We conclude Ly<sub>α</sub> absorption lines could thus arise from an extended distribution of VLSB low luminosity galaxies in clusters or around isolated bright galaxies like NGC1291 rather than in extended gaseous halos.
Others have measured steep cluster luminosity functions before (e.g. Phillipps et al. 1998, Trentham 1997, Wilson et al. 1997). One concern about these determinations has been the possible contamination by background galaxies. The measured background contamination was statistically removed from the data by assessing the numbers of background galaxies at each magnitude determined from fields away from the cluster. If this had been done incorrectly then a fit to the faint background number counts (slope $`0.6`$) would lead erroneously to a steep ($`\alpha 2.5`$) faint-end slope to the luminosity function. Our approach is different. We have specifically selected large LSB galaxies because our models (MDS) indicated very low background contamination if we selected in this way. In addition because of the large extent of our data we have been able to show a decrease in galaxy number density with cluster radius. This strongly supports our contention that we have predominately selected a cluster sample and confirms the previous observations and interpretations of Phillipps et al. 1998, Trentham 1997 and Wilson et al. 1997.
The clustering scale of the VLSB sample compared to the bright sample is similar to predictions of N-body hierarchical simulations of structure formation with ”bias” (White & Frenk 1991, Frenk et al. 1996, Kauffman et al. 1997). For example many features of the numerical simulations of a galaxy cluster by Frenk et al. 1996 are seen in the observational data. The galaxies show mass segregation with the most massive galaxies being more centrally clustered in relation to the smaller galaxies and to the dark matter (by about a factor of 3). Also the low mass slope of the simulated mass function is $`1.8`$. To be the dominant mass component of the cluster the VLSB galaxies require large values of (M/L). Previous work (McGaugh & de Blok 1998, Carignan & Freeman 1988) has suggested that LSB and dwarf galaxies are the most dark matter dominated galaxies of all. For example the mass to light ratios of LSB dwarf galaxies in the Local Group can be large ((M/L)=10-200, Irwin & Hatzidimitriou 1995). So, the VLSB galaxy population could account for all of the dark matter in clusters if the galaxy mass to light ratio is similar to the cluster mass to light ratio (of order 100-150, David et al. 1995) or there are significantly more galaxies below our present detection limits. The VLSB galaxies may well be the most reliable tracers of the dark matter content of clusters and so we might expect the dominant dark matter of a cluster to reside in its difficult-to-detect very LSB, low luminosity galaxy population.
## 5 Conclusions
We have identified a faint VLSB galaxy population associated with the nearby Fornax cluster. The VLSB galaxies appear to be associated with the brighter cluster galaxies, but their distribution is more spatially extended. Integrating the total numbers we estimate that they contribute almost as much to the total luminosity of the cluster as the brighter galaxies and they have a larger cross-sectional area on the sky. If they have mass-to-light ratios as high as some more nearby dwarf galaxies they will dominate the mass of the cluster. Faint galaxies like these have previously been predicted by extrapolations of recently determined cluster luminosity functions and by recent numerical models of cluster formation. In the future it will be important to investigate the nature of this VLSB population further and to try and identify similar galaxies in other environments.
|
warning/0005/hep-th0005216.html
|
ar5iv
|
text
|
# Untitled Document
p-ADIC AND ADELIC FREE RELATIVISTIC PARTICLE
G.S.DJORDJEVIĆ<sup>1</sup>, B. DRAGOVICH<sup>2∗</sup><sup>*</sup><sup>*</sup>E-mail:dragovic@phy.bg.ac.yu and LJ. NEŠIĆ<sup>1</sup>
<sup>1</sup>Department of Physics, University of Niš,P.O.Box 91,
18000 Niš, Yugoslavia
<sup>2</sup>Institute of Physics, P.O.Box 57, 11001 Belgrade, Yugoslavia
Abstract
We consider spectral problem for a free relativistic particle in p-adic and adelic quantum mechanics. In particular, we found p-adic and adelic eigenfunctions. Within adelic approach there exist quantum states that exhibit discrete structure of spacetime at the Planck scale.
1. Introduction
Since 1987, there have been many interesting applications of p-adic numbers in various parts of theoretical and mathematical physics (for a review, see, e.g. Refs. 1-3). It is likely that the most attractive investigations have been in the Planck scale physics and non-archimedean structure of spacetime. One of the greatest achievements in that direction is a formulation of p-adic quantum mechanics<sup>4,5</sup> with its adelic generalization<sup>6</sup>. Adelic quantum mechanics unifies ordinary and p-adic ones, for all primes $`p`$.
There is an interplay of physical and mathematical reasons to investigate possible role of p-adic numbers and adeles in physics. Namely, all numerical results of experiments belong to the field of rational numbers Q, which is dense in the field of real numbers R and p-adic ones $`\text{Q}_{\mathrm{}}`$ ($`p`$ is a prime number). R and $`\text{Q}_{\mathrm{}}`$ $`(p=2,3,5,\mathrm{})`$ exhaust all possible numbers which can be obtained by completion of Q. The set of adeles A enables to regard real and p-adic numbers simultaneosly. What is a limit in application of real numbers in description of spacetime? Do p-adic numbers play some role in physics? To answer these, and similar questions, one has to construct and examine p-adic and adelic models of quantum and relativistic physical systems.
Models at the Planck scale are of particular interest. In fact, if the Planck length $`l_0=(\mathrm{}G/c^3)^{1/2}10^{33}`$ cm is the elementary one then any other length $`x`$ should be an integer multiple of $`l_0`$. So, if one takes $`l_0=1`$ then one has $`x=n`$. Real distance is $`d_{\mathrm{}}(x,0)=n_{\mathrm{}}=n`$, while the p-adic one is $`d_p(x,0)=n_p1`$. Thus, p-adic geometry has to emerge approaching the Planck scale physics.
So far, different p-adic and adelic quantum models have been studied (see, e.g. Refs. 7-9).
In this letter we examine p-adic and adelic quantum properties of a free relativistic particle. As a result of adelic approach we find some discreteness of space and time at the Planck scale.
2. Some p-Adic and Adelic Mathematics
Since majority of physicists are still unfamiliar with p-adic numbers and adeles we give here some basic facts about these attractive parts of modern mathematics (for a profound knowledge, see, e.g. Refs. 2, 10-12).
Any p-adic number $`x\text{Q}_{\mathrm{}}`$ can be presented in the unique way as an infinite expansion
$$x=x_kp^k+x_{k+1}p^{k+1}+\mathrm{}+x_0+x_1p+x_2p^2+\mathrm{},k\text{N}\mathrm{}$$
$`\mathrm{}\mathrm{}\mathrm{}\mathrm{}\mathrm{}`$
where $`x_i=0,1,\mathrm{},p1`$ are digits. p-Adic norm of a term in (2.1) is $`x_ip^i_p=p^i`$. Since p-adic norm is the non-archimedean one, i.e. $`u+v_p\mathrm{max}\{u_p,v_p\}`$, it follows that $`x_p=p^k`$ in the representation (2.1).
There are mainly two types of analysis on $`\text{Q}_{\mathrm{}}`$. The first one (i) is based on the mapping $`f:\text{Q}_{\mathrm{}}\text{Q}_{\mathrm{}}`$, and the second one (ii) is related to map $`\phi :\text{Q}_{\mathrm{}}\text{C}`$, where C is the set of complex numbers. We use both of these analysis in p-adic generalization of physical models: (i) in classical and (ii) in quantum mechanics. Derivatives of $`f(x)`$ are defined as in the real case, but using p-adic norm instead of the usual absolute value function. For mapping $`\phi (x)`$ there is well-defined integration with the Haar measure. In particular, we use the Gauss integral<sup>2</sup>
$$_{x_pp^\nu }\chi _p(\alpha x^2+\beta x)𝑑x=\{\begin{array}{cc}p^\nu \mathrm{\Omega }(p^\nu \beta _p),\hfill & \alpha _pp^{2\nu },\hfill \\ \lambda _p(\alpha )2\alpha _p^{1/2}\chi _p\left(\frac{\beta ^2}{4\alpha }\right)\mathrm{\Omega }\left(p^\nu \frac{\beta }{2\alpha }_p\right),\hfill & 4\alpha _p>p^{2\nu }.\hfill \end{array}$$
$`(2.2)`$
$`\chi _p(u)=\mathrm{exp}(2\pi i\{u\}_p)`$ is a p-adic additive character, where $`\{u\}_p`$ denotes the fractional part of $`u\text{Q}_{\mathrm{}}`$. $`\lambda _p(\alpha )`$ is an arithmetic complex-valued function<sup>2</sup> with the following basic properties:
$$\lambda _p(0)=1,\lambda _p(a^2\alpha )=\lambda _p(\alpha ),\lambda _p(\alpha )\lambda _p(\beta )=\lambda _p(\alpha +\beta )\lambda _p(\alpha ^1+\beta ^1),\lambda _p(\alpha )_{\mathrm{}}=1.$$
$`(2.3)`$
$`\mathrm{\Omega }(u_p)`$ is the characteristic function on $`\text{Z}_{\mathrm{}}`$, i.e.
$$\mathrm{\Omega }(u_p)=\{\begin{array}{cc}1,\hfill & u_p1,\hfill \\ 0,\hfill & u_p>1,\hfill \end{array}$$
$`(2.4)`$
where $`\text{Z}_{\mathrm{}}\mathrm{}\{\mathrm{}\text{Q}_{\mathrm{}}\mathrm{}\mathrm{}_{\mathrm{}}\mathrm{}\}`$ is the ring of p-adic integers.
An adele $`a\text{A}`$ is an infinite sequence
$$a=(a_{\mathrm{}},a_2,\mathrm{},a_p,\mathrm{}),$$
$`(2.5)`$
where $`a_{\mathrm{}}\text{R}`$ and $`a_p\text{Q}_{\mathrm{}}`$ with the restriction that $`a_p\text{Z}_{\mathrm{}}`$ for all but a finite set $`S`$ of primes $`p`$. The set of all adeles A can be written in the form
$$\text{A}\mathrm{}\underset{\mathrm{𝕊}}{𝕌}\mathrm{𝔸}\mathrm{}\mathrm{𝕊}\mathrm{}\mathrm{}\mathrm{𝔸}\mathrm{}\mathrm{𝕊}\mathrm{}\mathrm{}\text{R}\times \underset{\mathrm{}\mathrm{𝕊}}{}\text{Q}_{\mathrm{}}\times \underset{\mathrm{}\mathrm{𝕊}}{}\text{Z}_{\mathrm{}}\mathrm{}$$
$`\mathrm{}\mathrm{}\mathrm{}\mathrm{}\mathrm{}`$
A is a topological space. It is a ring with respect to componentwise addition and multiplication. There is a natural generalization of analysis on R and $`\text{Q}_{\mathrm{}}`$ to analysis on A.
3. Free Relativistic Particle in p-Adic Quantum Mechanics
In the Vladimirov-Volovich formulation<sup>4</sup> (see also Ref. 5) one-dimensional p-adic quantum mechanics is a triple
$$(L_2(\text{Q}_{\mathrm{}}\mathrm{}\mathrm{}\mathrm{𝕎}_{\mathrm{}}\mathrm{}ϝ\mathrm{}\mathrm{}\mathrm{𝕌}_{\mathrm{}}\mathrm{}\mathrm{}\mathrm{}\mathrm{}\mathrm{}$$
$`\mathrm{}\mathrm{}\mathrm{}\mathrm{}\mathrm{}`$
where $`L_2(\text{Q}_{\mathrm{}}\mathrm{}`$ is the Hilbert space of complex-valued functions of p-adic variables, $`W_p(z)`$ is a unitary representation of the Heisenberg-Weyl group on $`L_2(\text{Q}_{\mathrm{}}\mathrm{}`$, and $`U_p(t)`$ is an evolution operator on $`L_2(\text{Q}_{\mathrm{}}\mathrm{}`$.
$`U_p(t)`$ is an integral operator
$$U_p(t)\psi _p(x)=_\text{Q}_{\mathrm{}}K_p(x,t;y,0)\psi _p(y)𝑑y$$
$`(3.2)`$
whose kernel is given by the Feynman path integral
$$K_p(x,t;y,0)=\chi _p\left(\frac{1}{h}S[q]\right)𝒟q=\chi _p\left(\frac{1}{h}_0^tL(q,\dot{q})𝑑t\right)\underset{t}{}dq(t),$$
$`(3.3)`$
where $`h`$ is the Planck constant. p-Adic Feynman path integral is investigated in Ref. 13, where it is shown that for quadratic classical actions $`\overline{S}(x,t;y,0)`$ the solution (3.3) becomes
$$K_p(x,t;y,0)=\lambda _p\left(\frac{1}{2h}\frac{^2\overline{S}}{xy}\right)\left|\frac{1}{h}\frac{^2\overline{S}}{xy}\right|_p^{1/2}\chi _p\left(\frac{1}{h}\overline{S}(x,t;y,0)\right).$$
$`(3.4)`$
Expression (3.4) has the same form as that one in ordinary quantum mechanics (i.e. with $`L_2(\text{R}\mathrm{}`$). For a particular physical system, p-adic eigenfunctions are subject of the spectral problem
$$U_p(t)\psi _p^{(\alpha )}(x)=\chi _p(\alpha t)\psi _p^{(\alpha )}(x).$$
$`(3.5)`$
The usual action for a free relativistic particle<sup>14</sup>
$$S=mc^2_{\tau _1}^{\tau _2}𝑑\tau \sqrt{\eta _{\mu \nu }\dot{x}^\mu \dot{x}^\nu }$$
$`(3.6)`$
is nonlinear and so unsuitable for quantum-mechanical investigations. However, a free relativistic particle can be treated as a system with the constraint<sup>15</sup> $`\eta _{\mu \nu }k^\mu k^\nu +m^2c^2=k^2+m^2c^2=0`$, which leads to the canonical Hamiltonian (with the Lagrange multiplier $`N`$)
$$H_c=N(k^2+m^2c^2),$$
$`(3.7)`$
and to the Lagrangian
$$L=\dot{x}_\mu k^\mu H_c=\frac{\dot{x}^2}{4N}m^2c^2N,$$
$`(3.8)`$
where $`\dot{x}_\mu =H_c/k^\mu =2k_\mu `$ and $`\dot{x}^2=\dot{x}^\mu \dot{x}_\mu `$. Instead of (3.6), the corresponding action for quantum treatment of a free relativistic particle is
$$S=_{\tau _1}^{\tau _2}𝑑\tau \left(\frac{\dot{x}^2}{4N}m^2c^2N\right).$$
$`(3.9)`$
From (3.9) it follows the classical trajectory
$$\overline{x}^\mu =\frac{x_2^\mu x_1^\mu }{\tau _2\tau _1}\tau +\frac{x_1\tau _2x_2\tau _1}{\tau _2\tau _1}$$
$`(3.10)`$
and the classical action
$$\overline{S}(x_2,T;x_1,0)=\frac{(x_2x_1)^2}{4T}m^2c^2T,$$
$`(3.11)`$
where $`T=N(\tau _2\tau _1)`$.
All the above expressions from (3.7) to (3.11) are valid in the real case and according to p-adic analysis they have place in the p-adic one. Note that the classical action (3.11) can be presented in the form
$$\begin{array}{cc}\hfill \overline{S}& =\left[\frac{(x_2^0x_1^0)^2}{4T}\frac{m^2c^2T}{4}\right]+\left[\frac{(x_2^1x_1^1)^2}{4T}\frac{m^2c^2T}{4}\right]\hfill \\ & +[\frac{(x_2^2x_1^2)^2}{4T}\frac{m^2c^2T}{4}]+\frac{(x_2^3x_1^3)^2}{4T}\frac{m^2c^2T}{4}]=\overline{S}^0+\overline{S}^1+\overline{S}^2+\overline{S}^3\hfill \end{array}$$
$`(3.12)`$
which is quadratic in $`x_2^\mu `$ and $`x_1^\mu `$ $`(\mu =0,1,2,3)`$.
Due to (3.4) and (3.12), the corresponding quantum-mechanical propagator may be written as product
$$K_p(x_2,T;x_1,0)=\underset{\mu =0}{\overset{3}{}}K_p^{(\mu )}(x_2^\mu ,T;x_1^\mu ,0),$$
$`(3.13)`$
$$\begin{array}{cc}\hfill K_p^{(\mu )}(x_2^\mu ,T;x_1^\mu ,0)& =\lambda _p((1)^{\delta _0^\mu }4hT)2hT_p^{\frac{1}{2}}\hfill \\ & \times \chi _p\left[\frac{1}{h}(1)^{\delta _0^\mu }\frac{(x_2^\mu x_1^\mu )^2}{4T}+\frac{1}{h}\frac{m^2c^2T}{4}\right]\hfill \end{array},$$
$`(3.14)`$
where $`\delta _0^\mu =1`$ if $`\mu =0`$ and $`0`$ otherwise.
Among all possible eigenstates which satisfy eq. (3.5), function $`\mathrm{\Omega }(x_p)`$, defined by (2.4), plays a central role in p-adic and adelic quantum mechanics. Therefore, let us first show existence of $`\mathrm{\Omega }`$-eigenfunction for the above relativistic particle. In fact, we have now 1+3 dimensional problem and the corresponding integral equation is
$$_{Q_p^4}K_p(x,T;y,0)\mathrm{\Omega }(y_p)d^4y=\mathrm{\Omega }(x_p),(\alpha =0),$$
$`(3.15)`$
where $`u_p=\mathrm{max}_{0\mu 3}\{u^\mu _p\}`$ is p-adic norm of $`u\text{Q}_{\mathrm{}}^{\mathrm{}}`$, and
$$K_p(x,T;y,0)=\frac{\lambda _p^2(4hT)}{2hT_p^2}\chi _p\left(\frac{(xy)^2}{4hT}+\frac{m^2c^2T}{h}\right).$$
$`(3.16)`$
Eq. (3.15), rewritten in a more explicite form, reads
$$\begin{array}{cc}& \frac{\lambda _p^2(4hT)}{2hT_p^2}\chi _p\left(\frac{m^2c^2T}{h}\frac{x^2}{4hT}\right)_\text{Z}_{\mathrm{}}\chi _p\left(\frac{(y^0)^2}{4hT}\frac{x^0y^0}{2hT}\right)𝑑y^0\hfill \\ & \times \underset{i=1}{\overset{3}{}}_\text{Z}_{\mathrm{}}\chi _p(\frac{(y^i)^2}{4hT}+\frac{x^iy^i}{2hT})dy^i=\mathrm{\Omega }(x_p).\hfill \end{array}$$
$`(3.17)`$
Using lower part of the Gauss integral (2.2) to calculate integrals in (3.17) for each coordinate $`y^\mu (\mu =0,\mathrm{},3)`$, we obtain
$$\chi _p\left(\frac{m^2c^2T}{h}\right)\underset{\mu =0}{\overset{3}{}}\mathrm{\Omega }(x^\mu _p)=\mathrm{\Omega }(x_p),hT_p<1.$$
$`(3.18)`$
Since $`_{\mu =0}^3\mathrm{\Omega }(x^\mu _p)=\mathrm{\Omega }(x_p)`$ is an identity, an equivalent assertion to (3.18) is
$$\left|\frac{m^2c^2T}{h}\right|_p1,hT_p<1.$$
$`(3.19)`$
Applying also the upper part of (2.2) to (3.17), we have
$$\frac{\lambda _p^2(4hT)}{2hT_p^2}\chi _p\left(\frac{m^2c^2T}{h}\frac{x^2}{4hT}\right)\underset{\mu =0}{\overset{3}{}}\mathrm{\Omega }\left(\left|\frac{x^\mu }{2hT}\right|_p\right)=\mathrm{\Omega }(x_p),4hT_p1,$$
$`(3.20)`$
what is satisfied only for $`p2`$. Namely, (3.20) becomes an equality if conditions
$$\left|\frac{m^2c^2T}{h}\right|_p1,hT_p=1,p2,$$
$`(3.21)`$
take place.
Thus, we obtained eigenstates
$$\psi _p(x,T)=\{\begin{array}{cc}\mathrm{\Omega }(x_p),|\frac{m^2c^2T}{h}|_p1,\hfill & hT_p1,p2,(3.22)\hfill \\ \mathrm{\Omega }(x_2),|\frac{m^2c^2T}{h}|_21,\hfill & hT_2<1,(3.23)\hfill \end{array}$$
which are invariant under $`U_p(t)`$ transformation.
We have also $`\mathrm{\Omega }`$-function in eigenstates
$$\psi _p(x,T)=\chi _p\left(\frac{m^2c^2}{h}T\right)\mathrm{\Omega }(p^\nu x_p),\nu \text{Z}\mathrm{}\mathrm{}\mathrm{𝕋}_{\mathrm{}}\mathrm{}\mathrm{}^\mathrm{}\mathrm{}\nu \mathrm{}$$
$`\mathrm{}\mathrm{}\mathrm{}\mathrm{}\mathrm{}\mathrm{}`$
This can be shown in the way similar to the previous case with $`\psi _p(x,T)=\mathrm{\Omega }(x_p)`$.
The eigenstates without $`\mathrm{\Omega }`$-functions are as follows:
$$\begin{array}{cc}\hfill \psi _p(x,T)=\chi _p\left(\frac{m^2c^2+k^2}{h}T\right)\chi _p\left(\frac{kx}{h}\right)& \end{array},$$
$`(3.25)`$
where $`k^2=k^0k^0+k^ik^i`$, and $`kx=k^0x^0+k^ix^i`$. Note that $`(m^2c^2+k^2)T=H_c\tau `$ (see (3.7) and (3.11)).
4. Free Relativistic Particle in Adelic Quantum Mechanics
According to Ref. 6, the main ingredients of adelic quantum mechanics are: (i) the Hilbert space $`L_2(\text{A}\mathrm{}`$ of complex-valued functions on the space of adeles A, (ii) a unitary representation $`W(z)`$ of the Heisenberg-Weyl group on $`L_2(\text{A}\mathrm{}`$, and (iii) a unitary representation of the evolution operator $`U(t)`$ on $`L_2(\text{A}\mathrm{}`$. In a sense, ingredients of quantum mechanics on adeles are some products of the corresponding objects from ordinary and p-adic quantum mechanics.
So, the evolution operator is defined by
$$U(t)\psi (x)=_\text{A}K(x,t;y,0)\psi (y)𝑑y,$$
$`(4.1)`$
where $`t\text{A}`$, $`x,y\text{A}`$, $`\psi L_2(\text{A}\mathrm{}`$, and
$$U(t)\psi (x)=U_{\mathrm{}}(t_{\mathrm{}})\psi _{\mathrm{}}(x_{\mathrm{}})\underset{p}{}U_p(t_p)\psi _p(x_p).$$
$`(4.2)`$
The spectral problem is given by
$$U(t)\psi ^{(\alpha )}(x)=\chi (\alpha t)\psi ^{(\alpha )}(x),\alpha \text{A}\mathrm{}$$
$`\mathrm{}\mathrm{}\mathrm{}\mathrm{}\mathrm{}`$
where $`\chi (\alpha t)=\chi _{\mathrm{}}(\alpha _{\mathrm{}}t_{\mathrm{}})_p\chi _p(\alpha _pt_p)`$ is the additive character on A.
Convergence of products and adelic consistency imply some conditions on p-adic constituents of an adelic object. For instance, any eigenfunction in (4.3) has the form
$$\psi (x)=\psi _{\mathrm{}}(x_{\mathrm{}})\underset{pS}{}\psi _p(x_p)\underset{pS}{}\mathrm{\Omega }(x_p_p),$$
$`(4.4)`$
where $`S`$ is a finite set of primes $`p`$. $`\psi _{\mathrm{}}(x_{\mathrm{}})L_2(\text{R}\mathrm{}`$ and $`\psi _p(x_p)`$, $`\mathrm{\Omega }(x_p_p)L_2(\text{Q}_{\mathrm{}}\mathrm{}`$ are eigenfunctions of ordinary and p-adic counterparts, respectively.
Adelic kernel of $`U(t)`$ for relativistic free particle is
$$K(x,T;y,0)=K_{\mathrm{}}(x_{\mathrm{}},T_{\mathrm{}};y_{\mathrm{}},0)\underset{p}{}K_p(x_p,T_p;y_p,0),$$
$`(4.5)`$
where $`K_p(x_p,T_p;y_p,0)`$ is given by (3.16), and $`K_{\mathrm{}}(x_{\mathrm{}},T_{\mathrm{}};y_{\mathrm{}},0)`$ is the real counterpart, which has the same form as $`K_p(x_p,T_p;y_p,0)`$. The corresponding arithmetic function $`\lambda _{\mathrm{}}(\alpha )`$ is defined by $`\lambda _{\mathrm{}}(\alpha )=(1i\text{sgn}\alpha )/\sqrt{2}`$ and satisfies the same basic properties as $`\lambda _p(\alpha )`$ (see (2.3)).
In order to complete the adelic spectral theory for a free relativistic particle let us now turn to the corresponding problem in ordinary quantum mechanics in the form
$$_\text{R}K_{\mathrm{}}(x,T_2;y,T_1)\psi _{\mathrm{}}(y,T_1)𝑑y=\chi _{\mathrm{}}(\alpha T_2)\psi _{\mathrm{}}(x),$$
$`(4.6)`$
where $`\chi _{\mathrm{}}(a)=\mathrm{exp}(2\pi ia)`$, $`\psi _{\mathrm{}}(x,T)=\chi _{\mathrm{}}(\alpha T)\psi _{\mathrm{}}(x)`$ and
$$K_{\mathrm{}}(x,T_2;y,T_1)=\frac{\lambda _{\mathrm{}}^2\left(4h(T_2T_1)\right)}{2h(T_2T_1)_{\mathrm{}}^2}\chi _{\mathrm{}}\left(\frac{(xy)^2}{4h(T_2T_1)}+\frac{m^2c^2(T_2T_1)}{h}\right).$$
$`(4.7)`$
(For simplicity, we omitted index $`\mathrm{}`$ for arguments in (4.7)). Using the Gauss integral
$$_\text{R}\chi _{\mathrm{}}(\alpha x^2+\beta x)𝑑x=\lambda _{\mathrm{}}(\alpha )2\alpha _{\mathrm{}}^{1/2}\chi _{\mathrm{}}\left(\frac{\beta ^2}{4\alpha }\right),\alpha 0,$$
$`(4.8)`$
we find the solution of (4.6) in the form
$$\psi _{\mathrm{}}(x,T)=\chi _{\mathrm{}}\left(\frac{m^2c^2+k^2}{h}T\right)\chi _{\mathrm{}}\left(\frac{kx}{h}\right),$$
$`(4.9)`$
where $`k^2=k_\mu k^\mu `$, $`kx=k_\mu x^\mu `$.
From the above investigation it follows adelic eigenfunction for a free relativistic particle:
$$\psi (x,T)=\chi _{\mathrm{}}\left(\frac{m^2c^2+k_{\mathrm{}}^2}{h}T_{\mathrm{}}\frac{k_{\mathrm{}}x_{\mathrm{}}}{h}\right)\underset{pS}{}\psi _p(x_p,T_p)\underset{pS}{}\mathrm{\Omega }(x_p_p),$$
$`(4.10)`$
where $`\psi _p(x_p,T)`$ are given by(3.24) and (3.25). Note that $`x\text{A}^{\mathrm{}}`$ in (4.10), i.e.
$$x=\left(\begin{array}{c}x^0\\ \mathrm{}\\ x^3\end{array}\right)=\left(\begin{array}{ccccc}x_{\mathrm{}}^0,& x_2^0,& \mathrm{},& x_p^0,& \mathrm{}\\ \mathrm{}& \mathrm{}& & \mathrm{}& & \\ x_{\mathrm{}}^3,& x_2^3,& \mathrm{},& x_p^3,& \mathrm{}\end{array}\right)=(x_{\mathrm{}},x_2,\mathrm{},x_p,\mathrm{}),$$
$`(4.11)`$
and $`k\text{A}^{\mathrm{}}`$, $`T\text{A}`$. Any adelic wave function may be obtained as superposition of eigenfunctions (4.10) by summation over $`S`$ and integration over four-momentum $`k`$.
5. Concluding Remarks
We shown that a free relativistic particle may be regarded as a subject not only of ordinary but also of p-adic and adelic quantum mechanics. It is an exactly soluble theoretical model. In particular, we found p-adic and adelic eigenfunctions.
In order to interpret adelic wave function for a free particle let us consider its norm, i.e.
$$|\psi (x,T)|_{\mathrm{}}^2=\underset{pS}{}|\psi _p(x_p,T_p)|_{\mathrm{}}^2\underset{pS}{}\mathrm{\Omega }(x_p_p),$$
$`(5.1)`$
where we used $`|\psi _{\mathrm{}}(x_{\mathrm{}},T_{\mathrm{}})|_{\mathrm{}}^2=1`$ and $`\mathrm{\Omega }^2(|x_p|_p)\mathrm{\Omega }(|x_p|_p)`$. As follows, (5.1) does not depend on real counterpart.
Comparison between theoretical predictions and experimental numerical data may be done only on rational numbers. Hence, consider (5.1) in points $`x_{\mathrm{}}=x_2=\mathrm{}=x_p=\mathrm{}=x\text{Q}`$. Since
$$\underset{pS}{}\mathrm{\Omega }(p^\nu x_p)=\{\begin{array}{cc}1,\hfill & xp^\nu \text{Z}\hfill \\ 0,\hfill & x\text{Q}\mathrm{}^\nu \text{Z}\mathrm{}\hfill \end{array}\underset{pS}{}\mathrm{\Omega }(x_p)=\{\begin{array}{cc}1,\hfill & x\text{Z}\hfill \\ 0,\hfill & x\text{Q}\text{Z}\mathrm{}\hfill \end{array}$$
$`(5.2)`$
it means that $`|\psi (x,T)|_{\mathrm{}}^2`$ may be different from zero only in a finite number of rational points which are not integers. Extending standard interpretation of ordinary wave function to the adelic one, it follows that the probability of finding the particle in integer points is dominant.
There is a special (vacuum) state $`(S=\mathrm{})`$ when all p-adic states are $`\mathrm{\Omega }(|x_p|_p)`$ and then particle can exist only in integer points of space and time. In ordinary quantum mechanics we label coordinates of space and time by real numbers which make continuum. However, using adeles to label space and time we obtain their discreteness in a natural way. Conditions $`|m^2c^2T/h|_p1`$ and $`|hT|_p|2|_p`$, which follow from (3.22) and (3.23), can be realized choosing $`h=c=m=1`$ as a system of units. If $`m`$ is the Planck mass then spacetime contains the Planck length as the elementary one. Since $`|T|_p=|N\tau |_p|2|_p`$ and $`N`$ is an arbitrary parameter, one can take $`N=2`$ and obtain $`|\tau |_p1`$ for every $`p`$. Thus, the invariant intervals $`\tau `$ as well as spacetime coordinates $`x^\mu `$ are discrete. Let us notice that there exist discrete subgroups of the Poincaré group that transform discrete spacetime (lattice) into itself (see, e.g. Ref. 16). It is reasonable to expect that spacetime discreteness is more manifest in quantum gravity models and we have such situation in adelic approach to quantum cosmology<sup>7,17</sup>.
Note that the above obtained results can be easily generalized to any number of spacetime dimensions.
References
1. L. Brekke and P. G. O. Freund, Phys. Rep. 233, 1 (1993).
2. V. S. Vladimirov, I. V. Volovich and E. I. Zelenov, p-Adic Analysis and Mathematical Physics (World Scientific, 1994).
3. A. Khrennikov, p-Adic Valued Distributions in Mathematical Physics (Kluwer, 1994).
4. V. S. Vladimirov and I. V. Volovich, Commun. Math. Phys. 123, 659 (1989).
5. Ph. Ruelle, E. Thiran, D. Verstegen and J. Weyers, J. Math.Phys 30, 2854 (1989).
6. B. Dragovich, Teor. Mat. Fiz. 101, 349 (1994); Int. J. Mod. Phys. A10, 2349 (1995).
7. B. Dragovich, ”Adelic Wave Function of the Universe”, in Proc. of the Third A. Friedmann Int. Seminar on Gravitation and Cosmology, St. Petersburg, 1996.
8. G. S. Djordjević and B. Dragovich, Facta Univ. 1, 204 (1996).
9. B. Dragovich and Lj. Nešić, Facta Univ. 1, 223 (1996).
10. W. H. Schikhof, Ultrametric Calculus (Cambridge Univ. Press, 1984).
11. I. M. Gel’fand, M. I. Graev and I. I. Piatetskii-Shapiro, Representation Theory and Automorphic Functions (Nauka, Moscow, 1966).
12. A. Weil, Adeles and Algebraic Groups ( Birkh$`\ddot{\text{a}}`$user, 1982).
13. G. S. Djordjević and B. Dragovich, Mod. Phys. Lett. A12, 1455 (1997).
14. B. Dragovich, P. H. Frampton and B. V. Urosevic, Mod. Phys. Lett. A5, 1521 (1990).
15. J. J. Halliwell and M. E. Ortiz, Phys. Rev. D48, 748 (1993).
16. P. A. M. Dirac, Discrete Subgroups of the Poincaré Group, in Problems of Theoretical Physics - A Memorial Volume to Igor E. Tamm (Nauka, Moscow, 1972).
17. B. Dragovich, Adelic Quantum Cosmology, to be published in Proc. of the Fourth A. Friedmann Int. Seminar on Gravitation and Cosmology, St. Petersburg, 1998.
|
warning/0005/hep-ph0005319.html
|
ar5iv
|
text
|
# 1 Why laser acceleration ?
## 1 Why laser acceleration ?
The problem of acceleration of charged particles by the laser field is, at present time, one of the most prestigeous problem in the accelerator physics. It is supposed that, in the future, the laser accelerator will play the same role in particle physics as the linear or circle accelerators working in today particle laboratories.
The acceleration effectiveness of the linear or circle accelerators is limitied not only by geometrical size of them but also by the energy loss of accelerated particles which is caused by bremsstrahlung during the acceleration. The amount of radiation, following from the Larmor formula, emitted by accelerated charged particle is given generally as follows (Maier, 1991):
$$P=\frac{2}{3}\frac{r_0m}{c}(\gamma ^6a_{||}^2+\gamma ^4a_{}^2);\gamma =\frac{1}{\sqrt{1\beta ^2}};\beta =v/c$$
(1)
where $`v`$ is the velocity of a particle, $`c`$ is the velocity of light in vacuum, $`a_{||}`$ is parallel acceleration of a particle and $`a_{}`$ is the perpendicular acceleration of a particle in the accelerator, $`m`$ is the rest mass of an electron. The quantity
$$r_0=\frac{1}{4\pi \epsilon _0}\frac{e^2}{mc^2}$$
(2)
is the electron classical radius in SI units.
In terms of momenta
$$\dot{p}_{||}=\frac{\dot{E}}{\beta c};\dot{p}_{}=m\gamma \dot{v}_{},$$
(3)
the radiated power can be written as
$$P=\frac{2}{3}\frac{r_0c}{E_0}(\dot{p}_{||}^2+\gamma ^2p_{}^2).$$
(4)
where $`E_0=mc^2`$. Equation (4) shows that the same acceleration force produces a $`\gamma ^2`$ times higher radiation power, if it is applied in perpendicular direction, compared to the parallel direction.
For a particle moving with a constant velocity in a circular machine with bending radius $`\varrho `$ the power radiated due to curvature of the orbit is
$$P=\frac{2}{3}r_0E_0c\frac{\beta ^4\gamma ^4}{\varrho ^2}.$$
(5)
So in the linear accelerator the energy loss caused by radiation is smaller than in circle accelerator and it means that to obtain high energy particles in linear accelerator is more easy than in the circle accelerator. In case of laser acceleration the situation radically changes . The classical idea of laser acceleration is to consider the laser light as the periodic electromagnetic field. The motion of electron in such a wave was firstly described by Volkov (Berestetzkii et al. 1989). However, it is possible to show that periodic electromagnetic wave does not accelerate electrons in classical and quantum theory, because the electric and magnetic components of the light field are mutually perpendicular and it means the motion caused by the classical periodic electromagnetic field is not linear but periodic (Landau et al., 1962).
The situation changes if we consider laser beam as a system of photons and the interaction of electron with laser light is via the Compton process
$$\gamma +e\gamma +e.$$
(6)
We can see that the right side of equation (6) involves no bremsstrahlung photons and it means that there are no energy loss caused by emission of photons. It means also that laser acceleration is more effective than the acceleration in the standard linear and circle accelerators. It is evident that acceleration by laser can be adequately described only by quantum field theory. Such viewpoint gives us the motivation to investigate theoretically the effectiveness of acceleration of charged particles by laser beam.
## 2 Historical view on laser acceleration
The acceleration of charged particles by laser beam has been studied by many authors (Tajima and Dawson, 1979; Katsouleas and Dawson, 1983; Scully and Zubary, 1991; Baranova and Zel’dovich, 1994).
Many designs for such devices has been proposed. Some of these are not sufficiently developed to be readily intelligible, others seem to be fallacious and others are unlikely to be relevant to ultra high energies. Some designs were developed only to observe pressure of laser light on microparticles in liquids and gas (Ashkin, 1970; Ashkin, 1972).
The idea of laser acceleration follows historically the idea that light exerts pressure. The former idea was for the first time postulated by Johanness Kepler, the King astronomer in Prague, in 1619. He wrote that the pressure of Sun light is what causes the tails of comets to point away from the Sun. The easy explanation of that effect was given by Newton in his corpuscular theory of light where this effect is evidently of the mechanical origin. Nevertheless, the numerical value of the light pressure was not known from time of Kepler to the formulation of the Maxwell theory of electromagnetism where Maxwell predicted in 1873 the magnitude of the light pressure.
The experimental terrestrial verification of the light pressure was given by the Russian physicist Lebedev and Nichols and Hull from USA (Nichols and Hull, 1903).The measurement consisted in determination of force acting on the torsion pendulum. At these experiments it was observed that the magnitude of the pressure of light confirmed the Maxwell prediction. It was confirmed that the pressure is very small and practically has no meaning if the weak terrestrial sources of light are used. Only after invention of lasers the situation changed because of the very strong intensity of the laser light which can cause the great pressure of the laser ray on the surface of the condensed matter. So, the problem of the determination of the light pressure is now physically meaningful because of the existence of high intensity lasers.
Here, we consider the acceleration of an electron by laser beam. We calculate the force due to Compton scattering of laser beam photons on electron using the methods of quantum field theory and quantum electrodynamics.
## 3 Quantum field theory of a laser beam acceleration
The dynamical equation of the relativistic particle with rest mass $`m`$ and the kinematical mass $`m(v)`$,
$$m(v)=\frac{m}{\sqrt{1\frac{v^2}{c^2}}},$$
(7)
is as follows (M$`ø`$ller, 1972):
$$𝐅=\frac{d(m(v)𝐯)}{dt}=m(v)\frac{d𝐯}{dt}+\frac{dm(v)}{dt}𝐯=m(v)\frac{d𝐯}{dt}+\frac{1}{c^2}\frac{dW}{dt}𝐯=$$
$$m(v)\frac{d𝐯}{dt}+\left(\frac{𝐅𝐯}{c^2}\right)𝐯,$$
(8)
or,
$$m(v)\frac{d𝐯}{dt}=𝐅\frac{𝐯}{c^2}(𝐅𝐯).$$
(9)
If we consider a particle that is acted upon by a force $`𝐅`$ and which has an initial velocity in the direction of the force, then, according to Eq. (9) the particle will continue to move in the direction of force. Therefore the path of the particle will be a straight line, and we can choose this line as the $`x`$-axis. From Eq. (9) then follows for $`𝐯𝐅`$, and $`F=|𝐅|`$:
$$\frac{d}{dt}\left\{\frac{v}{(1\frac{v^2}{c^2})^{1/2}}\right\}=\frac{F}{m},$$
(10)
where force $`F`$, in case it is generated by the laser beam, contains also the velocity $`v`$ of particle as an integral part of the Doppler frequency.
Now, let us consider the laser acceleration of an electron by the monochromatic laser beam. The force of photons acting on an electron in a laser beam depends evidently on the density of photons in this beam. Using the definition of the cross section of the electron-photon interaction and with the energy loss $`\omega \omega ^{}`$, it may be easy to define the force acting on electron by the laser beam, in the following way:
$$F=n_{\omega _1}^{\omega _2}(\omega \omega ^{})\frac{d\sigma (\omega \omega ^{})}{d(\omega \omega ^{})}d(\omega \omega ^{}),$$
(11)
where in the rest system of electron we have the following integral $`\omega ^{}`$-limits (Sokolov et al., 1983):
$$\omega _2=\frac{\omega }{1+\frac{2\omega }{m}}\omega ^{}\omega =\omega _1.$$
(12)
where $`\omega `$ can be identified with the frequency of the impinging photon on the rest electron.
Let us remark that Eq. (11) has the dimensionality of force if we correctly suppose that the dimensionality of $`d\sigma `$ is $`\mathrm{m}^2`$ and density of photons in laser beam is $`\mathrm{m}^3`$. At the same time the combination $`\omega \omega ^{}`$ in the cross section and differential can be replaced by $`\omega ^{}`$.
Since the dimensionality of the expression on the right side of the last equation is force, we at this moment connect $`m`$ with $`c^2`$ in order to get $`mc^2=E`$. Quantity $`\omega `$ will be later denoted as the frequency $`\omega _0`$ of the impinging photon. Using the expression for the differential Compton cross section (Berestetzkii et al., 1989), we get for the force:
$$F=n\pi r_e^2\frac{E}{\omega ^2}_{\omega _1}^{\omega _2}𝑑\omega ^{}(\omega ^{}\omega )\left[\frac{\omega }{\omega ^{}}+\frac{\omega ^{}}{\omega }+\left(\frac{m}{\omega ^{}}\frac{m}{\omega }\right)^22m\left(\frac{1}{\omega ^{}}\frac{1}{\omega }\right)\right],$$
(13)
where $`r_e=e^2/mc^2`$ is the classical radius of an electron.
After $`\omega ^{}`$-integration we obtain
$`F=n\pi r_e^2{\displaystyle \frac{E}{\omega ^2}}\{(3m^2\omega ^2+2m\omega )\mathrm{ln}{\displaystyle \frac{\omega _2}{\omega _1}}+m^2\omega ({\displaystyle \frac{1}{\omega _2}}{\displaystyle \frac{1}{\omega _1}})+`$
$`(\omega 4m{\displaystyle \frac{3m^2}{_\omega }})(\omega _2\omega _1)+{\displaystyle \frac{1}{3\omega }}(\omega _2^3\omega _1^3)+({\displaystyle \frac{m^2}{2\omega ^2}}+{\displaystyle \frac{m}{\omega }}{\displaystyle \frac{1}{2}})(\omega _2^2\omega _1^2)\},`$ (14)
where the corresponding the $`\omega _i`$-combination are as follows:
$$\frac{\omega _2}{\omega _1}=\frac{m}{m+2\omega };\omega _2\omega _1=\frac{2\omega ^2}{m+2\omega };\frac{1}{\omega _2}\frac{1}{\omega _1}=\frac{2}{m}$$
(15)
and
$$\omega _2^2\omega _1^2=\frac{4\omega ^3(m+\omega )}{(m+2\omega )^2};\omega _2^3\omega _1^3=\frac{2\omega ^4}{(m+2\omega )^3}(3m^2+6m\omega +4\omega ^2).$$
(16)
Then, after insertion of the $`\omega _i`$-combinations and putting $`\omega \omega _0`$, we obtain for the accelerating force instead of equation Eq. (14) the following equation:
$`F=n\pi r_e^2{\displaystyle \frac{E}{\omega _0^2}}\times `$
$`\{(3m^2\omega _0^2+2m\omega _0)\mathrm{ln}{\displaystyle \frac{m}{m+2\omega _0}}+2m\omega _0+{\displaystyle \frac{2\omega _0(4m\omega _0+3m^2\omega _0^2)}{m+2\omega _0}}`$
$`{\displaystyle \frac{2\omega _0^3(3m^2+6m\omega _0+4\omega _0^2)}{3(m+2\omega _0)^3}}+2\omega _0(\omega _0^2m^22m\omega _0){\displaystyle \frac{(m+\omega _0)}{(m+2\omega _0)^2}}\}.`$ (17)
Now, if we want to express the force $`F`$ in the MKS system where its dimensionality is kg.m<sup>2</sup> s<sup>-2</sup>, we are forced to introduce the physical constants; velocity of light c, Planck constant $`\mathrm{}`$ in the last formula. It is easy to see that the last formula expressed in the MKS system is as follows:
$$F=n\pi r_e^2\frac{E}{(\mathrm{}\omega _0)^2}\times $$
$$\{(3m^2c^4\mathrm{}^2\omega _0^2+2mc^2\mathrm{}\omega _0)\mathrm{ln}\frac{mc^2}{mc^2+2\mathrm{}\omega _0}+2mc^2\mathrm{}\omega _0+$$
$$\frac{2\mathrm{}\omega _0(4mc^2\mathrm{}\omega _0+3m^2c^4\mathrm{}^2\omega _0^2)}{mc^2+2\mathrm{}\omega _0}$$
$$\frac{2\mathrm{}^3\omega _0^3(3m^2c^4+6mc^2\mathrm{}\omega _0+4\mathrm{}^2\omega _0^2)}{3(mc^2+2\mathrm{}\omega _0)^3}+$$
$$2\mathrm{}\omega _0(\mathrm{}^2\omega _0^2m^2c^42mc^2\mathrm{}\omega _0)\frac{(mc^2+\mathrm{}\omega _0)}{(mc^2+2\mathrm{}\omega _0)^2}\}.$$
(18)
The last formula is valid approximately only for nonrelativistic velocities because for laser photons the Doppler effect plays substantional role for moving electron in the laser field. The formula of the relativistic Doppler effect is as follows:
$$\omega _0\omega _0\frac{1\frac{v}{c}}{\left(1\frac{v^2}{c^2}\right)^{1/2}},$$
(19)
which means that for ultrarelativistic electron the frequency of photons accelerating the electron will be very small and that the acceleration will be also very small with regard to the electron moving with the relativistic velocities in the laser field. In order to obtain the exact description of the electron motion in the laser field, it is necessary to insert the Doppler frequency equation Eq. (19) in the formula Eq. (10).
Of course, it is not easy to obtain the general solution of Eq. (10) because it is strongly nonlinear. For $`vc`$ we obtain the solution $`(E=mc^2,\mathrm{}\omega _0=\epsilon )`$:
$$v=\frac{1}{m}\pi ntr_e^2\times $$
$$\{(3E^2\epsilon ^2+2E\epsilon )\mathrm{ln}\frac{E}{E+2\epsilon }+2E\epsilon +$$
$$\frac{2\epsilon (4E\epsilon +3E^2\epsilon ^2)}{E+2\epsilon }\frac{2\epsilon ^3(3E^2+6E\epsilon +4\epsilon ^2)}{3(E+2\epsilon )^3}+$$
$$2\epsilon (\epsilon ^2E^22E\epsilon )\frac{(E+\epsilon )}{(E+2\epsilon )^2}\}.$$
(20)
We simplify the last result using the approximation $`\epsilon E`$. In this approximation we use with $`x=2\epsilon /E,\mathrm{ln}(1+x)^1(x+x^2/2x^3/3+x^4/4),(1+x)^11x+x^2x^3,(1+x)^212x+3x^24x^3,(1+x)^313x+6x^2`$, and we obtain:
$$v\frac{1}{m}nt\left(\frac{8\pi }{3}r_e^2\right)\frac{\epsilon ^2}{E}=\frac{1}{m}nt\sigma _{\gamma e}\frac{\epsilon ^2}{E}.$$
(21)
where $`\sigma _{\gamma e}`$ is the Thompson cross section. It is evident that the general formula of the velocity can be obtained in the form
$$v=\frac{1}{m}nt\sigma _{\gamma e}\epsilon f\left(\frac{\epsilon }{E}\right)$$
(22)
where Taylor coefficients of $`f`$ can be obtained by expansion of Eq. (20).
Now, let us calculate the laser acceleration of an electron by the monochromatic $`2MW`$ laser i. e. with photon energy $`\epsilon =\mathrm{}\omega _0=2\times 10^{20}`$ J and density of photons $`n=N/V=4\times 10^{24}\mathrm{m}^3`$. Let us determine the velocity of accelerated electron at $`t=1\mathrm{s}`$. We use the following physical constants: $`m=9.1\times 10^{31}\mathrm{kg}`$, $`c=3\times 10^8\mathrm{m}.\mathrm{s}^1`$, $`\sigma _{\gamma e}=6.65\times 10^{29}\mathrm{m}^2`$ (Muirhead, 1968).
After some numerical calculation we obtain the following numerical value of velocity of electron at time $`t=1\mathrm{s}`$:
$$v(t=1\mathrm{s})\frac{1}{9.1\times 10^{31}}\times 4\times 10^{24}\times (2\times 10^{20})^2\times 6.65\times 10^{29}\times $$
$$\frac{1}{9.1\times 10^{31}}\times \frac{1}{(3\times 10^8)}\times 11.4\mathrm{m}.\mathrm{s}^1.$$
(23)
So, we see that for very short time intervals the motion of an electron is nonrelativistic. For large time intervals it will be necessary to use the relativistic equation involving the relativistic mass and the Doppler effect. The derived velocity, as we can see, is very small. However it can be increased by increasing the the density of photons in a laser beam. The density of photons can be, for instance, increased by an appropriate focusation of a laser beam. So, by this way, we can achieve the sufficient velocity of electrons for the practical application.
## 4 Discussion
We have seen, in this article, that the force accelerating an electron by a laser beam can be determined by means of the quantum field theory. The derived formula Eq. (10) with Eq. (19) describes the force due to the Compton scattering of photons with electron moving in the laser monochromatic photon sea. We have used only the simple Compton process Eq. (6) and not the more complicated multiple Compton process defined by equation
$$n\gamma +e\gamma +e,$$
(24)
which follows for instance as a quantization of the Volkov equation (Berestetzkii et al., 1989).
The present article is the modification of the Pardy discussion on laser acceleration (Pardy, 1998), where the thermal statistical model of laser acceleration was proposed. The basic ansatz of that model was the energy loss formula
$$\frac{dW}{dx}=\frac{1}{v}𝑑\mathrm{\Gamma }(\omega \omega ^{}),$$
(25)
where $`\mathrm{\Gamma }`$ is the differential reaction rate defined in different manner in quark qluon plasma physics and in electrodynamical medium.
In that article, it was used the approach by Brown et al. (Brown and Steinke, 1997; Brown, 1992; Sokolov et al., 1983, Braaten and Thoma, 1991). At the same time we used the ideas of Blumenthal et al. (Blumenthal and Gould, 1970). Brown et al. (Brown and Steinke, 1997), applied the total electron scattering rate for determination of behaviour of electron in the Planckian photon sea inside of the pipe of the storage rings. While in the preceding article the thermal distribution function $`f(k)`$ of photons was considered, here, we used the nonthermal density of photons n.
The experimental perspective of the laser beam acceleration of elementary particles concerns not only the charged particles, however, also the neutral particles such as neutron, neutral $`\pi `$-meson, and so on. Also the system of particles with the opposite charges was considered to be simultaneously accelerated by the laser beam (Pardy, 1997). New experiments can be realized and new measurements performed by means of the laser accelerator, giving new results and discoveries. So, it is obvious that the acceleration of particles by the laser beam can form, in the near future, the integral part of the particle physics. In such laboratories as ESRF, CERN, DESY, SLAC and so on, there is no problem to install lasers with the sufficient power of the photon beam, giving opportunity to construct the laser accelerator.
To say the final words, we hope, the ideas of the present article open the way to laser accelerators and will be considered as the integral part of the today particle physics.
REFERENCES
Ashkin, A., (1970) Acceleration and trapping of particles by radiation pressure, Phys. Rev. Lett. 24, 156-159.
Ashkin, A., (1972) The pressure of laser light, Scientific American 226 (2), 63-71.
Baranova, N. B. and Zel’dovich, B. Ya., (1994) Acceleration of charged particles by laser beams, JETP 78, (3), 249-258.
Berestetzkii, B. B., Lifshitz, E. M. and Pitaevskii L. P., (1989) Quantum electrodynamics, (Moscow, Nauka).
Blumenthal, G. R., and Gould, R. J., (1970) Bremsstrahlung, synchrotron radiation, and Compton Scattering of High-Energy Electrons Traversing Dilute Gases, Rev. Mod. Phys. 42, 237-269.
Braaten, E. and Thoma, M. H., (1991) Energy loss of a heavy fermion in a hot QED plasma, Phys. Rev. D 44, No. 4, 1298-1310.
Brown, L. S., (1992) Quantum Field Theory, (Cambridge Univ. Press).
Brown, L. S. and Steinke, R. S., (1997), Compton scattering on black body photons, Am. J. Phys. 65 (4), 304-309.
Katsouleas, T. and Dawson, J. M., (1983) Unlimited electron acceleration in laser-driven plasma waves, Phys. Rev. Lett. 51, 392-395.
Landau, L. D., and Lifshitz, E. M., (1962) The Classical Theory of Fields, 2nd ed. (Pergamon Press, Oxford).
Maier, M., (1991) Synchrotron radiation, in: CAS-PROCEEDINGS, ed. S. Turner, CERN 91-04, 8 May, 97-115.
M$`ø`$ller, C., (1972) The Theory of Relativity, (2nd edition), (Clarendon Press, Oxford).
Muirhead, H., (1968) The Physics of Elementary Particles, 2nd ed. (Pergamon Press, Oxford, London).
Nichols, E. F. and Hull, G. F., (1903) Phys. Rev. 17, 26-50; ibid. 91-104.
Pardy, M., (1997) Čerenkov effect and the Lorentz Contraction, Phys. Rev. A 55, No. 3 , 1647-1652.
Pardy, M., (1998) The quantum field theory of laser acceleration, Phys. Lett. A 243, 223-228.
Scully, M. O. and Zubary, M. S., (1991) Simple laser accelerator: Optics and particle dynamics, Phys. Rev. A 44 , 2656-2663.
Sokolov, A. A., Ternov, I. M. , Žukovskii, V. Č. and Borisov, A. B., (1983) Quantum Electrodynamics, (Moscow Univ. Press) (in Russian).
Tajima, T. and Dawson, J. M., (1979) Laser Electron Accelerator, Phys. Rev. Lett. 43, 267-270.
|
warning/0005/math-ph0005010.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
Let $`YX`$ be a smooth fibre bundle of a field model. We study cohomology of the variational bicomplex of exterior forms on the infinite order jet space $`J^{\mathrm{}}Y`$ of $`YX`$. The exterior differential on $`J^{\mathrm{}}Y`$ splits into the sum of the vertical differential $`d_V`$ and the horizontal differential $`d_H`$. These differentials, together with the variational operator $`\delta `$, constitute the variational bicomplex of exterior forms on $`J^{\mathrm{}}Y`$.
Note that the two differential algebras of exterior forms $`𝒪_{\mathrm{}}^{}`$ and $`𝒬_{\mathrm{}}^{}`$ are usually considered on $`J^{\mathrm{}}Y`$. The $`𝒪_{\mathrm{}}^{}`$ consists of all exterior forms on finite order jet manifolds modulo the pull-back identification. Lagrangian field theory is phrased in terms of $`𝒪_{\mathrm{}}^{}`$. Its cohomology, except de Rham cohomology and a particular result of on $`\delta `$-cohomology, remains unknown. The $`𝒬_{\mathrm{}}^{}`$ is the structure algebra of the sheaf of germs of exterior forms on finite order jet manifolds. For short, one can say that it consists of exterior forms of locally finite jet order. The $`d_H`$\- and $`\delta `$-cohomology of $`𝒬_{\mathrm{}}^{}`$ has been investigated in . Due to Lemma 3 below, we simplify this investigation and complete it by the study of $`d_V`$-cohomology of $`𝒬_{\mathrm{}}^{}`$. We prove that the differential algebra $`𝒪_{\mathrm{}}^{}`$ has the same $`d_H`$\- and $`\delta `$-cohomology as $`𝒬_{\mathrm{}}^{}`$ (see Theorem 8 below). This provides a solution of the global inverse problem of the calculus of variations in the class of finite order Lagrangians. The main point for applications is that the obstruction to the exactness of the calculus of variations is given by closed forms on the fibre bundle $`Y`$, and is of first order.
## 2 The differential calculus on $`J^{\mathrm{}}Y`$
Smooth manifolds throughout are assumed to be real, finite-dimensional, Hausdorff, paracompact, and connected. Put further dim$`X=n1`$.
Jet spaces provide the standard framework in theory of non-linear differential equations and the calculus of variations . Recall that the $`r`$-order jet space $`J^rY`$ consists of sections of $`YX`$ identified by $`r+1`$ terms of their Taylor series. The key point is that $`J^rY`$ is a smooth manifold. It is coordinated by $`(x^\lambda ,y^i,y_\mathrm{\Lambda }^i)`$, where $`(x^\lambda ,y^i)`$ are bundle coordinates on $`YX`$ and $`\mathrm{\Lambda }=(\lambda _k\mathrm{}\lambda _1)`$, $`|\mathrm{\Lambda }|=kr`$, denotes a symmetric multi-index. The infinite order jet space $`J^{\mathrm{}}Y`$ is defined as a projective limit $`(J^{\mathrm{}}Y,\pi _r^{\mathrm{}})`$ of the inverse system
$$X\stackrel{\pi }{}Y\stackrel{\pi _0^1}{}\mathrm{}J^{r1}Y\stackrel{\pi _{r1}^r}{}J^rY\mathrm{}$$
(1)
of finite order jet manifolds $`J^rY`$ of $`YX`$, where $`\pi _{r1}^r`$ are affine bundles. The surjections
$$\pi _r^{\mathrm{}}:J^{\mathrm{}}YJ^rY$$
(2)
obey the composition condition $`\pi _i^{\mathrm{}}=\pi _i^j\pi _j^{\mathrm{}}`$, $`j>i`$. The set $`J^{\mathrm{}}Y`$ is provided with the coarsest topology such that all surjections (2) are continuous. The base of open sets of this topology consists of the inverse images of open subsets of finite order jet manifolds under the surjections (2), which thus are open maps. With this topology, $`J^{\mathrm{}}Y`$ is a paracompact Fréchet (but not Banach) manifold modelled on a locally convex vector space of formal series $`\{x^\lambda ,y^i,y_\lambda ^i,\mathrm{}\}`$ . Bearing in mind the well-known Borel theorem, one can say that $`J^{\mathrm{}}Y`$ consists of equivalence classes of sections of $`YX`$ identified by their Taylor series at points $`xX`$. A bundle coordinate atlas $`\{U_Y,(x^\lambda ,y^i)\}`$ of $`YX`$ yields the manifold coordinate atlas
$`\{(\pi _0^{\mathrm{}})^1(U_Y),(x^\lambda ,y_\mathrm{\Lambda }^i)\},0|\mathrm{\Lambda }|,`$
of $`J^{\mathrm{}}Y`$, together with the transition functions
$$y_{}^{}{}_{\lambda +\mathrm{\Lambda }}{}^{i}=\frac{x^\mu }{x^\lambda }d_\mu y_\mathrm{\Lambda }^i,$$
(3)
where $`\lambda +\mathrm{\Lambda }`$ is the multi-index $`(\lambda \lambda _k\mathrm{}\lambda _1)`$ and $`d_\lambda `$ are the total derivatives
$`d_\lambda =_\lambda +{\displaystyle \underset{|\mathrm{\Lambda }|0}{}}y_{\lambda +\mathrm{\Lambda }}^i_i^\mathrm{\Lambda }.`$
Moreover, $`Y`$ is a strong deformation retract of $`J^{\mathrm{}}Y`$ (see Appendix A for an explicit form of a homotopy map)
Since $`J^{\mathrm{}}Y`$ is not a Banach manifold, the familiar geometric definition of differential objects on $`J^{\mathrm{}}Y`$ is not appropriate (see, e.g., ). One uses the fact that $`J^{\mathrm{}}Y`$ is a projective limit of the inverse system of manifolds (1). Given this inverse system, we have the direct system
$$𝒪^{}(X)\stackrel{\pi ^{}}{}𝒪_0^{}\stackrel{\pi _0^1^{}}{}𝒪_1^{}\stackrel{\pi _1^2^{}}{}\mathrm{}\stackrel{\pi _{r1}^r^{}}{}𝒪_r^{}\mathrm{}$$
(4)
of differential algebras $`𝒪_r^{}`$ of exterior forms on finite order jet manifolds, where $`\pi _{r1}^r^{}`$ are pull-back monomorphisms. This direct system admits a direct limit $`(𝒪_{\mathrm{}}^{},\pi _r^{\mathrm{}})`$ in the category of $`R`$-modules. It consists of exterior forms on finite order jet manifolds modulo the pull-back identification, together with the $`R`$-module monomorphisms
$`\pi _k^{\mathrm{}}:𝒪_k^{}𝒪_{\mathrm{}}^{}`$
which obey the composition condition $`\pi _i^{\mathrm{}}=\pi _j^{\mathrm{}}\pi _i^j`$, $`j>i`$. Operations of the exterior product $``$ and the exterior differentiation $`d`$ of exterior forms on finite order jet manifolds commute with the pull-back maps $`\pi _{r1}^r^{}`$ and, thus, constitute the direct systems of the order-preserving endomorphisms of the direct system (4). These direct systems have the direct limits which make $`𝒪_{\mathrm{}}^{}`$ a graded differential algebra. The $`𝒪_{\mathrm{}}^{}`$ is a differential calculus over the $`R`$-ring $`𝒪_{\mathrm{}}^0`$ of continuous real functions on $`J^{\mathrm{}}Y`$ which are the pull-back of smooth real functions on finite order jet manifolds by surjections (2). Passing to the direct limit of de Rham complexes on finite order jet manifolds, de Rham cohomology of the differential algebra $`𝒪_{\mathrm{}}^{}`$ has only been found . This coincides with de Rham cohomology of the fibre bundle $`Y`$ (see Section 4). However, this is not a way of studying other cohomology of the graded differential algebra $`𝒪_{\mathrm{}}^{}`$.
To solve this problem, let us enlarge $`𝒪_{\mathrm{}}^0`$ to the $`R`$-ring $`𝒬_{\mathrm{}}^0`$ of continuous real functions on $`J^{\mathrm{}}Y`$ such that, given $`f𝒬_{\mathrm{}}^0`$ and any point $`qJ^{\mathrm{}}Y`$, there exists a neighborhood of $`q`$ where $`f`$ coincides with the pull-back of a smooth function on some finite order jet manifold. The reason lies in the fact that the paracompact space $`J^{\mathrm{}}Y`$ admits a partition of unity by elements of the ring $`𝒬_{\mathrm{}}^0`$ . Therefore, sheaves of $`𝒬_{\mathrm{}}^0`$-modules on $`J^{\mathrm{}}Y`$ are fine and, consequently, acyclic. Then, the abstract de Rham theorem on cohomology of a sheaf resolution can be called into play.
Remark 1. Throughout, we follow the terminology of where by a sheaf $`S`$ over a topological space $`Z`$ is meant a sheaf bundle $`SZ`$. Accordingly, $`\mathrm{\Gamma }(S)`$ denotes the canonical presheaf of sections of the sheaf $`S`$, and $`\mathrm{\Gamma }(Z,S)`$ is the group of global sections of $`S`$. All sheaves below are ringed spaces, but we omit this terminology if there is no danger of confusion.
Let us define a differential calculus over the ring $`𝒬_{\mathrm{}}^0`$. Let $`𝔒_r^{}`$ be a sheaf of germs of exterior forms on the $`r`$-order jet manifold $`J^rY`$ and $`\mathrm{\Gamma }(𝔒_r^{})`$ its canonical presheaf. There is the direct system of canonical presheaves
$`\mathrm{\Gamma }(𝔒_X^{})\stackrel{\pi ^{}}{}\mathrm{\Gamma }(𝔒_0^{})\stackrel{\pi _0^1^{}}{}\mathrm{\Gamma }(𝔒_1^{})\stackrel{\pi _1^2^{}}{}\mathrm{}\stackrel{\pi _{r1}^r^{}}{}\mathrm{\Gamma }(𝔒_r^{})\mathrm{},`$
where $`\pi _{r1}^r^{}`$ are pull-back monomorphisms with respect to open surjections $`\pi _{r1}^r`$. Its direct limit $`𝔒_{\mathrm{}}^{}`$ is a presheaf of graded differential $`R`$-algebras on $`J^{\mathrm{}}Y`$. The germs of elements of the presheaf $`𝔒_{\mathrm{}}^{}`$ constitute a sheaf $`𝔔_{\mathrm{}}^{}`$ on $`J^{\mathrm{}}Y`$. It means that, given a section $`\varphi \mathrm{\Gamma }(U,𝔔_{\mathrm{}}^{})`$ of $`𝔔_{\mathrm{}}^{}`$ over an open subset $`UJ^{\mathrm{}}Y`$ and any point $`qU`$, there exists a neighbourhood $`U_qU`$ of $`q`$ such that $`\varphi |_{U_q}`$ is the pull-back of a local exterior form on some finite order jet manifold. However, $`𝔒_{\mathrm{}}^{}`$ does not coincide with the canonical presheaf $`\mathrm{\Gamma }(𝔔_{\mathrm{}}^{})`$ the sheaf $`𝔔_{\mathrm{}}^{}`$. The structure algebra $`𝒬_{\mathrm{}}^{}=\mathrm{\Gamma }(J^{\mathrm{}}Y,𝔔_{\mathrm{}}^{})`$ of the sheaf $`𝔔_{\mathrm{}}^{}`$ is a desired differential calculus over the $`R`$-ring $`𝒬_{\mathrm{}}^{}`$. There are obvious $`R`$-algebra monomorphisms
$`𝒪_{\mathrm{}}^{}𝒬_{\mathrm{}}^{},𝔒_{\mathrm{}}^{}\mathrm{\Gamma }(𝔔_{\mathrm{}}^{}).`$
For short, we agree to call elements of $`𝒬_{\mathrm{}}^{}`$ the exterior forms on $`J^{\mathrm{}}Y`$. Restricted to a coordinate chart $`(\pi _0^{\mathrm{}})^1(U_Y)`$ of $`J^{\mathrm{}}Y`$, they can be written in a coordinate form, where horizontal forms $`\{dx^\lambda \}`$ and contact 1-forms $`\{\theta _\mathrm{\Lambda }^i=dy_\mathrm{\Lambda }^iy_{\lambda +\mathrm{\Lambda }}^idx^\lambda \}`$ constitute the set of generators of the differential calculus $`𝒬_{\mathrm{}}^{}`$. There is the canonical splitting
$`𝒬_{\mathrm{}}^{}=\underset{k,s}{}𝒬_{\mathrm{}}^{k,s},0k,0sn,`$
of $`𝒬_{\mathrm{}}^{}`$ into $`𝒬_{\mathrm{}}^0`$-modules $`𝒬_{\mathrm{}}^{k,s}`$ of $`k`$-contact and $`s`$-horizontal forms, together with the corresponding projections
$`h_k:𝒬_{\mathrm{}}^{}𝒬_{\mathrm{}}^{k,},0k,h^s:𝒬_{\mathrm{}}^{}𝒬_{\mathrm{}}^{,s},0sn.`$
Accordingly, the exterior differential on $`𝒬_{\mathrm{}}^{}`$ is decomposed into the sum $`d=d_H+d_V`$ of horizontal and vertical differentials such that
$`d_Hh_k=h_kdh_k,d_H(\varphi )=dx^\lambda d_\lambda (\varphi ),\varphi 𝒬_{\mathrm{}}^{},`$
$`d_Vh^s=h^sdh^s,d_V(\varphi )=\theta _\mathrm{\Lambda }^i_\mathrm{\Lambda }^i\varphi .`$
They are nilpotent, i.e.,
$`d_Hd_H=0,d_Vd_V=0,d_Vd_H+d_Hd_V=0.`$
Remark 2. It should be emphasized that, in the class of exterior forms of locally finite order, all local operators are well-defined because these forms depends locally on a finite number of variables and all sums over these variables converge.
Remark 3. Traditionally, one attempts to introduce the differential algebra $`𝒬_{\mathrm{}}^{}`$ of locally pull-back forms on $`J^{\mathrm{}}Y`$ in a standard geometric way . The difficulty lies in the geometric interpretation of derivations of the $`R`$-ring $`𝒬_{\mathrm{}}^0`$ as vector fields on the Fréchet manifold $`J^{\mathrm{}}Y`$.
## 3 The variational bicomplex
Being nilpotent, the differentials $`d_V`$ and $`d_H`$ provide the natural bicomplex $`\{𝔔_{\mathrm{}}^{k,m}\}`$ of the sheaf $`𝔔_{\mathrm{}}^{}`$ on $`J^{\mathrm{}}Y`$. To complete it to the variational bicomplex, one considers the projection $`R`$-module endomorphism
$`\tau ={\displaystyle \underset{k>0}{}}{\displaystyle \frac{1}{k}}\overline{\tau }h_kh^n,`$
$`\overline{\tau }(\varphi )=(1)^\mathrm{\Lambda }\theta ^i[d_\mathrm{\Lambda }(_i^\mathrm{\Lambda }\varphi )],0\mathrm{\Lambda },\varphi \mathrm{\Gamma }(𝔔_{\mathrm{}}^{>0,n}),`$
of $`𝔔_{\mathrm{}}^{}`$ such that
$`\tau d_H=0,\tau d\tau \tau d=0.`$
Introduced on elements of the presheaf $`𝔒_{\mathrm{}}^{}`$ (see, e.g., ), this endomorphism is induced on the sheaf $`𝔔_{\mathrm{}}^{}`$ and its structure algebra $`𝒬_{\mathrm{}}^{}`$. Put
$`𝔈_k=\tau (𝔔_{\mathrm{}}^{k,n}),E_k=\tau (𝒬_{\mathrm{}}^{k,n}),k>0.`$
Since $`\tau `$ is a projection operator, we have isomorphisms
$`\mathrm{\Gamma }(𝔈_k)=\tau (\mathrm{\Gamma }(𝔔_{\mathrm{}}^{k,n})),E_k=\mathrm{\Gamma }(J^{\mathrm{}}Y,𝔈_k).`$
The variational operator on $`𝔔_{\mathrm{}}^{,n}`$ is defined as the morphism $`\delta =\tau d`$. It is nilpotent, and obeys the relation
$$\delta \tau \tau d=0.$$
(5)
Let $`R`$ and $`𝔒_X^{}`$ denote the constant sheaf on $`J^{\mathrm{}}Y`$ and the sheaf of exterior forms on $`X`$, respectively. The operators $`d_V`$, $`d_H`$, $`\tau `$ and $`\delta `$ give the following variational bicomplex of sheaves of exterior forms on $`J^{\mathrm{}}Y`$:
$$\begin{array}{cccccccccccccccccccc}& & & & \hfill _{d_V}& \text{}\hfill & & \hfill _{d_V}& \text{}\hfill & & & \hfill _{d_V}& \text{}\hfill & & & \hfill _{d_V}& \text{}\hfill & & \hfill _\delta & \text{}\hfill \\ & & 0& & & 𝔔_{\mathrm{}}^{k,0}\hfill & \stackrel{d_H}{}& & 𝔔_{\mathrm{}}^{k,1}\hfill & \stackrel{d_H}{}& \mathrm{}& & 𝔔_{\mathrm{}}^{k,m}\hfill & \stackrel{d_H}{}& \mathrm{}& & 𝔔_{\mathrm{}}^{k,n}\hfill & \stackrel{\tau }{}& & 𝔈_k0\hfill \\ & & & & & \mathrm{}\hfill & & & \mathrm{}\hfill & & & & \mathrm{}\hfill & & & & \mathrm{}\hfill & & & \mathrm{}\hfill \\ & & & & \hfill _{d_V}& \text{}\hfill & & \hfill _{d_V}& \text{}\hfill & & & \hfill _{d_V}& \text{}\hfill & & & \hfill _{d_V}& \text{}\hfill & & \hfill _\delta & \text{}\hfill \\ & & 0& & & 𝔔_{\mathrm{}}^{1,0}\hfill & \stackrel{d_H}{}& & 𝔔_{\mathrm{}}^{1,1}\hfill & \stackrel{d_H}{}& \mathrm{}& & 𝔔_{\mathrm{}}^{1,m}\hfill & \stackrel{d_H}{}& \mathrm{}& & 𝔔_{\mathrm{}}^{1,n}\hfill & \stackrel{\tau }{}& & 𝔈_10\hfill \\ & & & & \hfill _{d_V}& \text{}\hfill & & \hfill _{d_V}& \text{}\hfill & & & \hfill _{d_V}& \text{}\hfill & & & \hfill _{d_V}& \text{}\hfill & & \hfill _\delta & \text{}\hfill \\ 0& & R& & & 𝔔_{\mathrm{}}^0\hfill & \stackrel{d_H}{}& & 𝔔_{\mathrm{}}^{0,1}\hfill & \stackrel{d_H}{}& \mathrm{}& & 𝔔_{\mathrm{}}^{0,m}\hfill & \stackrel{d_H}{}& \mathrm{}& & 𝔔_{\mathrm{}}^{0,n}\hfill & & & 𝔔_{\mathrm{}}^{0,n}\hfill \\ & & & & \hfill _\pi ^{\mathrm{}}& \text{}\hfill & & \hfill _\pi ^{\mathrm{}}& \text{}\hfill & & & \hfill _\pi ^{\mathrm{}}& \text{}\hfill & & & \hfill _\pi ^{\mathrm{}}& \text{}\hfill & & & \\ 0& & R& & & 𝔒_X^0\hfill & \stackrel{d}{}& & 𝔒_X^1\hfill & \stackrel{d}{}& \mathrm{}& & 𝔒_X^m\hfill & \stackrel{d}{}& \mathrm{}& & 𝔒_X^n\hfill & \stackrel{d}{}& \hfill 0& \\ & & & & & \text{}\hfill & & & \text{}\hfill & & & & \text{}\hfill & & & & \text{}\hfill & & & \\ & & & & & 0\hfill & & & 0\hfill & & & & 0\hfill & & & & 0\hfill & & & \end{array}$$
(6)
The second row and the last column of this bicomplex form the variational complex
$$0R𝔔_{\mathrm{}}^0\stackrel{d_H}{}𝔔_{\mathrm{}}^{0,1}\stackrel{d_H}{}\mathrm{}\stackrel{d_H}{}𝔔_{\mathrm{}}^{0,n}\stackrel{\delta }{}𝔈_1\stackrel{\delta }{}𝔈_2\mathrm{}.$$
(7)
The corresponding variational bicomplexes $`\{𝒬_{\mathrm{}}^{},E_k\}`$ and $`\{𝒪_{\mathrm{}}^{},\overline{E}_k\}`$ of the differential calculus $`𝒬_{\mathrm{}}^{}`$ and $`𝒪_{\mathrm{}}^{}`$ take place.
There are the well-known statements summarized usually as the algebraic Poincaré lemma (see, e.g., ).
Lemma 1. If $`Y`$ is a contractible fibre bundle $`R^{n+p}R^n`$, the variational bicomplex $`\{𝒪_{\mathrm{}}^{},\overline{E}_k\}`$ of the graded differential algebra $`𝒪_{\mathrm{}}^{}`$ is exact.
It follows that the variational bicomplex of sheaves (6) is exact for any smooth fibre bundle $`YX`$. Moreover, all sheaves $`𝔔^{k,m}`$ in this bicomplex are fine, and so are the sheaves $`𝔈_k`$ in accordance with the following lemma.
Lemma 2. Sheaves $`𝔈_k`$, $`k>0`$, are fine.
Proof. Though $`R`$-modules $`E_{k>1}`$ fail to be $`𝒬_{\mathrm{}}^0`$-modules , one can use the fact that the sheaves $`𝔈_{k>0}`$ are projections $`\tau (𝔔_{\mathrm{}}^{k,n})`$ of sheaves of $`𝒬_{\mathrm{}}^0`$-modules. Let $`𝔘=\{U_i\}_{iI}`$ be a locally finite open covering of $`J^{\mathrm{}}Y`$ and $`\{f_i𝒬_{\mathrm{}}^0\}`$ the associated partition of unity. For any open subset $`UJ^{\mathrm{}}Y`$ and any section $`\phi `$ of the sheaf $`𝔔_{\mathrm{}}^{k,n}`$ over $`U`$, let us put $`h_i(\phi )=f_i\phi `$. Then, $`\{h_i\}`$ provide a family of endomorphisms of the sheaf $`𝔔_{\mathrm{}}^{k,n}`$, required for $`𝔔_{\mathrm{}}^{k,n}`$ to be fine. Endomorphisms $`h_i`$ of $`𝔔_{\mathrm{}}^{k,n}`$ also yield the $`R`$-module endomorphisms
$`\overline{h}_i=\tau h_i:𝔈_k\stackrel{\mathrm{in}}{}𝔔_{\mathrm{}}^{k,n}\stackrel{h_i}{}𝔔_{\mathrm{}}^{k,n}\stackrel{\tau }{}𝔈_k`$
of the sheaves $`𝔈_k`$. They possess the properties required for $`𝔈_k`$ to be a fine sheaf. Indeed, for each $`iI`$, there is a closed set $`\mathrm{supp}f_iU_i`$ such that $`\overline{h}_i`$ is zero outside this set, while the sum $`\underset{iI}{}\overline{h}_i`$ is the identity morphism. $`\mathrm{}`$
This Lemma simplify essentially our cohomology computation of the variational bicomplex in comparison with that in . Since all sheaves except $`R`$ and $`\pi ^{\mathrm{}}𝔒_X^{}`$ in the bicomplex (6) are fine, the abstract de Rham theorem (, Theorem 2.12.1) can be applied to columns and rows of this bicomplex in a straightforward way. We will quote the following variant of this theorem (see Appendix B for its proof).
Theorem 3. Let
$$0S\stackrel{h}{}S_0\stackrel{h^0}{}S_1\stackrel{h^1}{}\mathrm{}\stackrel{h^{p1}}{}S_p\stackrel{h^p}{}S_{p+1},p>1,$$
(8)
be an exact sequence of sheaves on a paracompact topological space $`Z`$, where the sheaves $`S_p`$ and $`S_{p+1}`$ are not necessarily acyclic, and let
$$0\mathrm{\Gamma }(Z,S)\stackrel{h_{}}{}\mathrm{\Gamma }(Z,S_0)\stackrel{h_{}^0}{}\mathrm{\Gamma }(Z,S_1)\stackrel{h_{}^1}{}\mathrm{}\stackrel{h_{}^{p1}}{}\mathrm{\Gamma }(Z,S_p)\stackrel{h_{}^p}{}\mathrm{\Gamma }(Z,S_{p+1})$$
(9)
be the corresponding cochain complex of structure groups of these sheaves. The $`q`$-cohomology groups of the cochain complex (9) for $`0qp`$ are isomorphic to the cohomology groups $`H^q(Z,S)`$ of $`Z`$ with coefficients in the sheaf $`S`$.
## 4 De Rham cohomology of $`J^{\mathrm{}}Y`$
Let us start from de Rham cohomology of the graded differential algebra $`𝒪_{\mathrm{}}^{}`$.
Proposition 4. There is an isomorphism
$`H^{}(𝒪_{\mathrm{}}^{})=H^{}(Y)`$
between de Rham cohomology $`H^{}(𝒪_{\mathrm{}}^{})`$ of $`𝒪_{\mathrm{}}^{}`$ and de Rham cohomology $`H^{}(Y)`$ of the fibre bundle $`Y`$.
Proof. The proof is based on the fact that the de Rham complex
$$0R𝒪_{\mathrm{}}^0\stackrel{d}{}𝒪_{\mathrm{}}^1\stackrel{d}{}\mathrm{}$$
(10)
of $`𝒪_{\mathrm{}}^{}`$ is the direct limit of de Rham complexes of exterior forms on finite order jet manifolds. Since the exterior differential $`d`$ commutes with the pull-back maps $`\pi _{r1}^r^{}`$, these complexes form a direct system. Then, in accordance with the well-known theorem , the cohomology groups $`H^{}(𝒪_{\mathrm{}}^{})`$ of the de Rham complex (10) are isomorphic to the direct limit of the direct system
$`0H^{}(X)\stackrel{\pi ^{}}{}H^{}(Y)\stackrel{\pi _0^1}{}H^{}(J^1Y)\mathrm{}`$
of de Rham cohomology groups $`H^{}(J^rY)`$ of finite order jet manifolds $`J^rY`$. The forthcoming Lemma 4 completes the proof. $`\mathrm{}`$
Lemma 5. De Rham cohomology of any finite-order jet manifold $`J^rY`$ is equal to that of $`Y`$.
Proof. Since every fibre bundle $`J^rYJ^{r1}Y`$ is affine, $`J^{r1}Y`$ is a strong deformation retract of $`J^rY`$, and so is $`Y`$ (see Appendix A). Then, in accordance with the Vietoris–Begle theorem , cohomology $`H(J^rY,R)`$ of $`J^rY`$ with coefficients in the constant sheaf $`R`$ coincides with that $`H(Y,R)`$ of $`Y`$. The well-known de Rham theorem completes the proof. $`\mathrm{}`$
Turn now to de Rham cohomology of the graded differential algebra $`𝒬_{\mathrm{}}^{}`$. Let us consider the complex of sheaves
$$0R𝔔_{\mathrm{}}^0\stackrel{d}{}𝔔_{\mathrm{}}^1\stackrel{d}{}\mathrm{}$$
(11)
on $`J^{\mathrm{}}Y`$ and the de Rham complex of their structure algebras
$$0R𝒬_{\mathrm{}}^0\stackrel{d}{}𝒬_{\mathrm{}}^1\stackrel{d}{}\mathrm{}.$$
(12)
Proposition 6. There is an isomorphism
$`H^{}(𝒬_{\mathrm{}}^{})=H^{}(Y)`$
of de Rham cohomology $`H^{}(𝒬_{\mathrm{}}^{})`$ of the graded differential algebra $`𝒬_{\mathrm{}}^{}`$ to that $`H^{}(Y)`$ of the fiber bundle $`Y`$.
Proof. The complex (11) is exact due to the Poincaré lemma, and is a resolution of the constant sheaf $`R`$ on $`J^{\mathrm{}}Y`$ since $`𝔔_{\mathrm{}}^r`$ are sheaves of $`𝒬_{\mathrm{}}^0`$-modules. Then, by virtue of Theorem 3, we have the cohomology isomorphism
$$H^{}(𝒬_{\mathrm{}}^{})=H^{}(J^{\mathrm{}}Y,R).$$
(13)
Lemma 4 below completes the proof. $`\mathrm{}`$
Lemma 7. There is an isomorphism
$$H^{}(J^{\mathrm{}}Y,R)=H^{}(Y,R)=H^{}(Y)$$
(14)
between cohomology $`H^{}(J^{\mathrm{}}Y,R)`$ of $`J^{\mathrm{}}Y`$ with coefficients in the constant sheaf $`R`$, that $`H^{}(Y,R)`$ of $`Y`$, and de Rham cohomology $`H^{}(Y)`$ of $`Y`$.
Proof. Since $`Y`$ is a strong deformation retract of $`J^{\mathrm{}}Y`$, the first isomorphism in (14) follows from the above-mentioned Vietoris–Begle theorem , while the second one is a consequence of the de Rham theorem. $`\mathrm{}`$
Since the graded differential algebras $`𝒪_{\mathrm{}}^{}`$ and $`𝒬_{\mathrm{}}^{}`$ have the same de Rham cohomology, we agree to call
$`H^{}(J^{\mathrm{}}Y)=H^{}(𝒬_{\mathrm{}}^{})=H^{}(𝒪_{\mathrm{}}^{})`$
the de Rham cohomology of $`J^{\mathrm{}}Y`$.
Proposition 4 shows that every closed form $`\varphi 𝒬_{\mathrm{}}^{}`$ splits into the sum
$$\varphi =\phi +d\xi ,\xi 𝒬_{\mathrm{}}^{},$$
(15)
where $`\phi `$ is a closed form on the fiber bundle $`Y`$. Accordingly, Proposition 4 states that, if $`\varphi `$ in this splitting belongs to $`𝒪_{\mathrm{}}^{}`$, so is $`\xi `$. The decomposition (15) will play an important role in the sequel.
## 5 Cohomology of $`d_V`$
Let us consider the vertical exact sequence of sheaves
$$0𝔒_X^m\stackrel{\pi ^{\mathrm{}}}{}𝔔_{\mathrm{}}^{0,m}\stackrel{d_V}{}\mathrm{}\stackrel{d_V}{}𝔔_{\mathrm{}}^{k,m}\stackrel{d_V}{}\mathrm{},0mn,$$
(16)
in the variational bicomplex (6) and the corresponding complex of their structure algebras
$$0𝒪^m(X)\stackrel{\pi ^{\mathrm{}}}{}𝒬_{\mathrm{}}^{0,m}\stackrel{d_V}{}\mathrm{}\stackrel{d_V}{}𝒬_{\mathrm{}}^{k,m}\stackrel{d_V}{}\mathrm{}.$$
(17)
Proposition 8. There is an isomorphism
$$H^{}(m,d_V)=H^{}(Y,\pi ^{}𝔒_X^m)$$
(18)
of cohomology groups $`H^{}(m,d_V)`$ of the complex (17) to cohomology groups $`H^{}(Y,\pi ^{}𝔒_X^m)`$ of $`Y`$ with coefficients in the pull-back sheaf $`\pi ^{}𝔒_X^m`$ on $`Y`$.
Proof. The exact sequence (16) is a resolution of the pull-back sheaf $`\pi ^{\mathrm{}}𝔒_X^m`$ on $`J^{\mathrm{}}Y`$. Then, by virtue of Theorem 3, we have a cohomology isomorphism
$`H^{}(m,d_V)=H^{}(J^{\mathrm{}}Y,\pi ^{\mathrm{}}𝔒_X^m).`$
The isomorphism (18) follows from the facts that $`Y`$ is a strong deformation retract of $`J^{\mathrm{}}Y`$ and that $`\pi ^{\mathrm{}}𝔒_X^m`$ is the pull-back onto $`J^{\mathrm{}}Y`$ of the sheaf $`\pi ^{}𝔒_X^m`$ on $`Y`$ . $`\mathrm{}`$
Corollary 9. Cohomology groups $`H^{>\mathrm{dim}Y}(m,d_V)`$ vanish.
The cohomology groups $`H^{}(m,d_V)`$ have a $`C^{\mathrm{}}(X)`$-module structure. For instance, let
$`YX\times VX`$
be a trivial fibre bundle with a typical fibre $`V`$. There is an obvious isomorphism of $`R`$-modules
$$H^{}(m,d_V)=𝔒_X^mH^{}(V).$$
(19)
## 6 Cohomology of $`d_H`$
Turn now to the rows of the variational bicomplex (6). We have the exact sequence of sheaves
$`0𝔔_{\mathrm{}}^{k,0}\stackrel{d_H}{}𝔔_{\mathrm{}}^{k,1}\stackrel{d_H}{}\mathrm{}\stackrel{d_H}{}𝔔_{\mathrm{}}^{k,n}\stackrel{\tau }{}𝔈_k0,k>0.`$
Since the sheaves $`𝔔_{\mathrm{}}^{k,0}`$ and $`𝔈_k`$ are fine, this is a resolution of the fine sheaf $`𝔔_{\mathrm{}}^{k,0}`$. It states immediately the following.
Proposition 10. The cohomology groups $`H^{}(k,d_H)`$ of the complex
$$0𝒬_{\mathrm{}}^{k,0}\stackrel{d_H}{}𝒬_{\mathrm{}}^{k,1}\stackrel{d_H}{}\mathrm{}\stackrel{d_H}{}𝒬_{\mathrm{}}^{k,n}\stackrel{\tau }{}E_k0,k>0,$$
(20)
are trivial.
This result at terms $`𝒬_{\mathrm{}}^{k,<n}`$ recovers that of . The exactness of the complex (20) at the term $`𝒬_{\mathrm{}}^{k,n}`$ means that, if
$`\tau (\varphi )=0,\varphi 𝒬_{\mathrm{}}^{k,n},`$
then
$`\varphi =d_H\xi ,\xi 𝒬_{\mathrm{}}^{k,n1}.`$
Since $`\tau `$ is a projection operator, there is the $`R`$-module decomposition
$$𝒬_{\mathrm{}}^{k,n}=E_kd_H(𝒬_{\mathrm{}}^{k,n1}).$$
(21)
Remark 4. One can derive Proposition 6 from Theorem 3, without appealing to that sheaves $`𝔈_k`$ are acyclic.
Let us consider the exact sequence of sheaves
$`0R𝔔_{\mathrm{}}^0\stackrel{d_H}{}𝔔_{\mathrm{}}^{0,1}\stackrel{d_H}{}\mathrm{}\stackrel{d_H}{}𝔔_{\mathrm{}}^{0,n}`$
where all sheaves except $`R`$ are fine. Then, from Theorem 3 and Lemma 4, we state the following.
Proposition 11. Cohomology groups $`H^r(d_H)`$, $`r<n`$, of the complex
$$0R𝒬_{\mathrm{}}^0\stackrel{d_H}{}𝒬_{\mathrm{}}^{0,1}\stackrel{d_H}{}\mathrm{}\stackrel{d_H}{}𝒬_{\mathrm{}}^{0,n}$$
(22)
are isomorphic to de Rham cohomology groups $`H^r(Y)`$ of $`Y`$.
This result recovers that of , but let us say something more.
Proposition 12. Any $`d_H`$-closed form $`\sigma 𝒬_{\mathrm{}}^{,<n}`$ is represented by the sum
$$\sigma =h_0\phi +d_H\xi ,\xi 𝒬_{\mathrm{}}^{},$$
(23)
where $`\phi `$ is a closed form on the fibre bundle $`Y`$.
Proof. Due to the relation
$$h_0d=d_Hh_0,$$
(24)
the horizontal projection $`h_0`$ provides a homomorphism of the de Rham complex (12) to the complex
$$0R𝒬_{\mathrm{}}^0\stackrel{d_H}{}𝒬_{\mathrm{}}^{0,1}\stackrel{d_H}{}\mathrm{}\stackrel{d_H}{}𝒬_{\mathrm{}}^{0,n}\stackrel{d_H}{}0.$$
(25)
Accordingly, there is a homomorphism
$$h_0^{}:H^r(J^{\mathrm{}}Y)H^r(d_H),0rn,$$
(26)
of cohomology groups of these complexes. Proposition 4 and Proposition 6 show that, for $`r<n`$, the homomorphism (26) is an isomorphism (see the relation (34) below for the case $`r=n`$). It follows that a horizontal form $`\psi 𝒬^{0,<n}`$ is $`d_H`$-closed (resp. $`d_H`$-exact) if and only if $`\psi =h_0\varphi `$ where $`\varphi `$ is a closed (resp. exact) form. The decomposition (15) and Proposition 6 complete the proof. $`\mathrm{}`$
Proposition 13. If $`\varphi 𝒬^{0,<n}`$ is a $`d_H`$-closed form, then $`d_V\varphi =d\varphi `$ is necessarily $`d_H`$-exact.
Proof. Being nilpotent, the vertical differential $`d_V`$ defines a homomorphism of the complex (25) to the complex
$`0𝒬_{\mathrm{}}^{1,0}\stackrel{d_H}{}𝒬_{\mathrm{}}^{1,1}\stackrel{d_H}{}\mathrm{}\stackrel{d_H}{}𝒬_{\mathrm{}}^{1,n}\stackrel{d_H}{}0`$
and, accordingly, a homomorphism of cohomology groups $`H^{}(d_H)H^{}(1,d_H)`$ of these complexes. Since $`H^{<n}(1,d_H)=0`$, the result follows. $`\mathrm{}`$
## 7 Cohomology of the variational complex
Let us prolong the complex (22) to the variational complex
$$0R𝒬_{\mathrm{}}^0\stackrel{d_H}{}𝒬_{\mathrm{}}^{0,1}\stackrel{d_H}{}\mathrm{}\stackrel{d_H}{}𝒬_{\mathrm{}}^{0,n}\stackrel{\delta }{}E_1\stackrel{\delta }{}E_2\mathrm{}$$
(27)
of the graded differential algebra $`𝒬_{\mathrm{}}^{}`$. In accordance with Lemma 3, the variational complex (7) is a resolution of the constant sheaf $`R`$ on $`J^{\mathrm{}}Y`$. Then, Theorem 3 and Lemma 4 give immediately the following.
Proposition 14. There is an isomorphism
$$H_{\mathrm{var}}^{}=H^{}(Y)$$
(28)
between cohomology $`H_{\mathrm{var}}^{}`$ of the variational complex (27) and de Rham cohomology of the fibre bundle $`Y`$.
The isomorphism (28) recovers the result of and that of at terms $`𝒬_{\mathrm{}}^{0,n}`$, $`E_1`$, but let us say something more. The relation (5) for $`\tau `$ and the relation (24) for $`h_0`$ define a homomorphisms of the de Rham complex (12) of the algebra $`𝒬_{\mathrm{}}^{}`$ to the variational complex (27). The corresponding homomorphism of their cohomology groups is an isomorphism. Then, in accordance with the splitting (15), we come to the following assertion which completes Proposition 6.
Proposition 15. Any $`\delta `$-closed form $`\sigma 𝒬_{\mathrm{}}^{k,n}`$, $`k0`$, is represented by the sum
$`\sigma =h_0\phi +d_H\xi ,k=0,\xi 𝒬^{0,n1},`$ (29a)
$`\sigma =\tau (\phi )+\delta (\xi ),k=1,\xi 𝒬^{0,n},`$ (29b)
$`\sigma =\tau (\phi )+\delta (\xi ),k>1,\xi E_{k1},`$ (29c)
where $`\phi `$ is a closed $`(n+k)`$-form on $`Y`$.
## 8 Cohomology of $`𝒪_{\mathrm{}}^{}`$
Thus, we have the whole cohomology of the graded differential algebra $`𝒬_{\mathrm{}}^{}`$. The following theorem provides us with $`d_H`$\- and $`\delta `$-cohomology of the graded differential algebra $`𝒪_{\mathrm{}}^{}`$.
Theorem 16. Graded differential algebra $`𝒪_{\mathrm{}}^{}`$ has the same $`d_H`$\- and $`\delta `$-cohomology as $`𝒬_{\mathrm{}}^{}`$.
Proof. Let the common symbol $`D`$ stand for the coboundary operators $`d_H`$ and $`\delta `$ of the variational bicomplex. Bearing in mind the decompositions (23), (29a) – (29c), it suffices to show that, if an element $`\varphi 𝒪_{\mathrm{}}^{}`$ is $`D`$-exact with respect to the algebra $`𝒬_{\mathrm{}}^{}`$ (i.e., $`\varphi =D\phi `$, $`\phi 𝒬_{\mathrm{}}^{}`$), then it is $`D`$-exact in the algebra $`𝒪_{\mathrm{}}^{}`$ (i.e., $`\varphi =D\phi ^{}`$, $`\phi ^{}𝒪_{\mathrm{}}^{}`$). Lemma 3 states that, if $`Y`$ is a contractible fibre bundle and a $`D`$-exact form $`\varphi `$ on $`J^{\mathrm{}}Y`$ is of finite jet order $`[\varphi ]`$ (i.e., $`\varphi 𝒪_{\mathrm{}}^{}`$), there exists an exterior form $`\phi 𝒪_{\mathrm{}}^{}`$ on $`J^{\mathrm{}}Y`$ such that $`\varphi =D\phi `$. Moreover, a glance at the homotopy operators for $`d_H`$ and $`\delta `$ shows that the jet order $`[\phi ]`$ of $`\phi `$ is bounded for all exterior forms $`\varphi `$ of fixed jet order. Let us call this fact the finite exactness of the operator $`D`$. Given an arbitrary fibre bundle $`Y`$, the finite exactness takes place on $`J^{\mathrm{}}Y|_U`$ over any open subset $`U`$ of $`Y`$ which is homeomorphic to a convex open subset of $`R^{\mathrm{dim}Y}`$. Now, we show the following.
(i) Suppose that the finite exactness of the operator $`D`$ takes place on $`J^{\mathrm{}}Y`$ over open subsets $`U`$, $`V`$ of $`Y`$ and their non-empty overlap $`UV`$. Then, it is also true on $`J^{\mathrm{}}Y|_{UV}`$.
(ii) Given a family $`\{U_\alpha \}`$ of disjoint open subsets of $`Y`$, let us suppose that the finite exactness takes place on $`J^{\mathrm{}}Y|_{U_\alpha }`$ over every subset $`U_\alpha `$ from this family. Then, it is true on $`J^{\mathrm{}}Y`$ over the union $`\underset{\alpha }{}U_\alpha `$ of these subsets.
If the assertions (i) and (ii) hold, the finite exactness of $`D`$ on $`J^{\mathrm{}}Y`$ takes place since one can construct the corresponding covering of the manifold $`Y`$ (, Lemma 9.5).
Proof of (i). Let $`\varphi =D\phi 𝒪_{\mathrm{}}^{}`$ be a $`D`$-exact form on $`J^{\mathrm{}}Y`$. By assumption, it can be brought into the form $`D\phi _U`$ on $`(\pi _0^{\mathrm{}})^1(U)`$ and $`D\phi _V`$ on $`(\pi _0^{\mathrm{}})^1(V)`$, where $`\phi _U`$ and $`\phi _V`$ are exterior forms of finite jet order. Due to the decompositions (23), (29a) – (29c), one can choose the forms $`\phi _U`$, $`\phi _V`$ such that $`\phi \phi _U`$ on $`(\pi _0^{\mathrm{}})^1(U)`$ and $`\phi \phi _V`$ on $`(\pi _0^{\mathrm{}})^1(V)`$ are $`D`$-exact forms. Let us consider their difference $`\phi _U\phi _V`$ on $`(\pi _0^{\mathrm{}})^1(UV)`$. It is a $`D`$-exact form of finite jet order which, by assumption, can be written as $`\phi _U\phi _V=D\sigma `$ where an exterior form $`\sigma `$ is also of finite jet order. Lemma 8 below shows that $`\sigma =\sigma _U+\sigma _V`$ where $`\sigma _U`$ and $`\sigma _V`$ are exterior forms of finite jet order on $`(\pi _0^{\mathrm{}})^1(U)`$ and $`(\pi _0^{\mathrm{}})^1(V)`$, respectively. Then, putting
$`\phi _U^{}=\phi _UD\sigma _U,\phi _V^{}=\phi _V+D\sigma _V,`$
we have the form $`\varphi `$ equal to $`D\phi _U^{}`$ on $`(\pi _0^{\mathrm{}})^1(U)`$ and $`D\phi _V^{}`$ on $`(\pi _0^{\mathrm{}})^1(V)`$, respectively. Since the difference $`\phi _U^{}\phi _V^{}`$ on $`(\pi _0^{\mathrm{}})^1(UV)`$ vanishes, we obtain $`\varphi =D\phi ^{}`$ on $`(\pi _0^{\mathrm{}})^1(UV)`$ where
$`\phi ^{}\stackrel{\mathrm{def}}{=}\{\begin{array}{cc}\phi ^{}|_U=\phi _U^{},\hfill & \\ \phi ^{}|_V=\phi _V^{}\hfill & \end{array}`$
is of finite jet order.
Proof of (ii). Let $`\varphi 𝒪_{\mathrm{}}^{}`$ be a $`D`$-exact form on $`J^{\mathrm{}}Y`$. The finite exactness on $`(\pi _0^{\mathrm{}})^1(U_\alpha )`$ holds since $`\varphi =D\phi _\alpha `$ on every $`(\pi _0^{\mathrm{}})^1(U_\alpha )`$ and, as was mentioned above, the jet order $`[\phi _\alpha ]`$ is bounded on the set of exterior forms $`D\phi _\alpha `$ of fixed jet order $`[\varphi ]`$. $`\mathrm{}`$
Lemma 17. Let $`U`$ and $`V`$ be open subsets of a fibre bundle $`Y`$ and $`\sigma 𝔒_{\mathrm{}}^{}`$ an exterior form of finite jet order on the non-empty overlap $`(\pi _0^{\mathrm{}})^1(UV)J^{\mathrm{}}Y`$. Then, $`\sigma `$ splits into a sum $`\sigma _U+\sigma _V`$ of exterior forms $`\sigma _U`$ and $`\sigma _V`$ of finite jet order on $`(\pi _0^{\mathrm{}})^1(U)`$ and $`(\pi _0^{\mathrm{}})^1(V)`$, respectively.
Proof. By taking a smooth partition of unity on $`UV`$ subordinate to the cover $`\{U,V\}`$ and passing to the function with support in $`V`$, one gets a smooth real function $`f`$ on $`UV`$ which is 0 on a neighborhood of $`UV`$ and 1 on a neighborhood of $`VU`$ in $`UV`$. Let $`(\pi _0^{\mathrm{}})^{}f`$ be the pull-back of $`f`$ onto $`(\pi _0^{\mathrm{}})^1(UV)`$. The exterior form $`((\pi _0^{\mathrm{}})^{}f)\sigma `$ is zero on a neighborhood of $`(\pi _0^{\mathrm{}})^1(U)`$ and, therefore, can be extended by 0 to $`(\pi _0^{\mathrm{}})^1(U)`$. Let us denote it $`\sigma _U`$. Accordingly, the exterior form $`(1(\pi _0^{\mathrm{}})^{}f)\sigma `$ has an extension $`\sigma _V`$ by 0 to $`(\pi _0^{\mathrm{}})^1(V)`$. Then, $`\sigma =\sigma _U+\sigma _V`$ is a desired decomposition because $`\sigma _U`$ and $`\sigma _V`$ are of finite jet order which does not exceed that of $`\sigma `$. $`\mathrm{}`$
It is readily observed that Theorem 8 is applied to de Rham cohomology of $`𝒪_{\mathrm{}}^{}`$ whose isomorphism to that of $`𝒬_{\mathrm{}}^{}`$ has been stated by Proposition 4 and Proposition 4
## 9 The global inverse problem in the calculus of variations
The variational complex (27) provides the algebraic approach to the calculus of variations on fiber bundles in the class of exterior forms of locally finite jet order . For instance, the variational operator $`\delta `$ acting on $`𝒬_{\mathrm{}}^{0,n}`$ is the Euler–Lagrange map, while $`\delta `$ acting on $`E_1`$ is the Helmholtz–Sonin map. Let
$`L=\omega 𝒬_{\mathrm{}}^{0,n},\omega =dx^1\mathrm{}dx^n,`$
be a horizontal density on $`J^{\mathrm{}}Y`$. One can think of $`L`$ as being a Lagrangian of locally finite order. Then, the canonical decomposition (21) leads to the first variational formula
$$dL=\tau (dL)+(\mathrm{Id}\tau )(dL)=\delta _1(L)+d_H(\varphi ),\varphi 𝒬_{\mathrm{}}^{1,n1},$$
(31)
where the exterior form
$`\delta _1(L)=(1)^\mathrm{\Lambda }d_\mathrm{\Lambda }(_i^\mathrm{\Lambda })\theta ^i\omega `$
is the Euler–Lagrange form associated with the Lagrangian $`L`$.
Let us relate the cohomology isomorphism (28) to the global inverse problem of the calculus in variations. As a particular repetition of Proposition 7, we come to its following solution in the class of Lagrangians of locally finite order.
Theorem 18. A Lagrangian $`L𝒬_{\mathrm{}}^{0,n}`$ is variationally trivial, i.e., $`\delta (L)=0`$ if and only if
$$L=h_0\phi +d_H\xi ,\xi 𝒬_{\mathrm{}}^{0,n1},$$
(32)
where $`\phi `$ is a closed $`n`$-form on $`Y`$ (see the expression (29a)).
Theorem 19. An Euler–Lagrange-type operator $`E_1`$ satisfies the Helmholtz condition $`\delta ()=0`$ if and only if
$$=\delta (L)+\tau (\varphi ),L𝒬_{\mathrm{}}^{0,n},$$
(33)
where $`\varphi `$ is a closed $`(n+1)`$-form on $`Y`$ (see the expression (29b)).
Theorem 9 recovers the result of .
Remark 5. As a consequence of Theorem 9, one obtains that the cohomology group $`H^n(d_H)`$ of the complex (25) obeys the relation
$$H^n(d_H)/H^n(Y)=\delta (𝒬_{\mathrm{}}^{0,n}),$$
(34)
where $`\delta (𝒬_{\mathrm{}}^{0,n})`$ is the $`R`$-module of Euler–Lagrange forms on $`J^{\mathrm{}}Y`$.
Theorem 8 leads us to the similar solution of the global inverse problem in the class of finite order Lagrangians. This is the case of higher order Lagrangian field theory. Namely, the theses of Theorem 9 and Theorem 9 remain true if all exterior forms in expressions (32) and (33) belong to $`𝒪_{\mathrm{}}^{}`$. Thus, the obstruction to the exactness of the finite order calculus of variations is the same as for exterior forms of locally finite order, without minimizing the order of Lagrangians. In particular, we recover the result of .
Note that the local exactness of the calculus of variations has been proved in the class of exterior forms of finite order by use of homotopy operators which do not minimize the order of Lagrangians (see, e.g., ). The infinite variational complex of such exterior forms on $`J^{\mathrm{}}Y`$ has been studied by many authors (see, e.g., ). However, these forms on $`J^{\mathrm{}}Y`$ fail to constitute a sheaf. Therefore, the cohomology obstruction to the exactness of the calculus of variations has been obtained in the class of exterior forms of locally finite jet order which make up the differential algebra $`𝒬_{\mathrm{}}^{}`$ Several statements without proof were announced in . A solution of the global inverse problem in the calculus of variations in the class of exterior forms of a fixed jet order has been suggested in by a computation of cohomology of the fixed order variational sequence (see for another variant of such a variational sequence). The key point of this computation lies in the local exactness of the finite order variational sequence which however requires rather sophisticated ad hoc technique in order to be reproduced (see also ). Therefore, the results of were not called into play. The first thesis of agrees with Theorem 9 for finite order Lagrangians, but says that the jet order of the form $`\xi `$ in the expression (32) is $`k1`$ if $`L`$ is a $`k`$-order variationally trivial Lagrangian. The second one states that a $`2k`$-order Euler–Lagrange operator can be always associated with a $`k`$-order Lagrangian.
Theorem 9 and Theorem 9 for elements of $`𝒪_{\mathrm{}}^{}`$ provide a solution of the global inverse problem in time-dependent mechanics treated as a particular field theory on smooth fiber bundles over $`X=R`$ . Note that, in time-dependent mechanics, the inverse problem is more intricate than in field theory. Given a second order dynamic equation, one studies the existence of an associated Newtonian system and its equivalence to a Lagrangian one . Since a fiber bundle $`YR`$ is trivial, de Rham cohomology of $`Y`$ is equal to that of its typical fiber $`M`$, and so is de Rham cohomology $`H^{}(J^{\mathrm{}}Y)`$ of $`J^{\mathrm{}}Y`$. The $`d_V`$-cohomology groups of the differential algebra $`𝒪_{\mathrm{}}^{}`$ are given by the isomorphism (19) such that
$`H^{}(0,d_V)=H^{}(1,d_V)=C^{\mathrm{}}(R)H^{}(M).`$
The variational complex (27) in time-dependent mechanics takes the form
$`0R𝒬_{\mathrm{}}^0\stackrel{d_t}{}𝒬_{\mathrm{}}^{0,1}\stackrel{\delta }{}E_1\stackrel{\delta }{}E_2\mathrm{}.`$
Its cohomology coincides with de Rham cohomology of $`M`$. In particular, Theorem 9 states that a Lagrangian $`L`$ of time-dependent mechanics is variationally trivial if and only if it takes the form
$`L=(\phi _t+\phi _iy_t^i)dt+d_t\xi ,`$
where $`\phi =\phi _tdt+\phi _idy^i`$ is a closed 1-form on $`Y`$ (see also ).
## 10 Cohomology of conservation laws
Let us concern briefly cohomology of conservation laws in Lagrangian formalism on $`J^{\mathrm{}}Y`$, but everything below is also true for a finite order Lagrangian formalism. Let $`u`$ be a vertical vector field on a fibre bundle $`YX`$, treated as a generator of a local 1-parameter group of gauge transformations of $`Y`$. Its infinite order jet prolongation
$`J^{\mathrm{}}u=d_\mathrm{\Lambda }u_i^\mathrm{\Lambda },0\mathrm{\Lambda },`$
is a derivation of the ring $`𝒬_{\mathrm{}}^0`$, and also defines the contraction $`u\varphi `$ and the Lie derivative
$`𝐋_{J^{\mathrm{}}u}\varphi \stackrel{\mathrm{def}}{=}J^{\mathrm{}}ud\varphi +d(J^{\mathrm{}}u\varphi )`$
of elements of the differential algebra $`𝒬_{\mathrm{}}^{}`$. It is easily justified that
$`J^{\mathrm{}}ud_H\varphi =d_H(J^{\mathrm{}}u\varphi ),\varphi 𝒬^{}_{\mathrm{}}.`$
Let $`L`$ be a Lagrangian on $`J^{\mathrm{}}Y`$. By virtue of the first variational formula (31), the Lie derivative of the Lagrangian $`L`$ along $`J^{\mathrm{}}u`$ reads
$$𝐋_{J^{\mathrm{}}u}L=J^{\mathrm{}}udL=u\delta Ld_H(J^{\mathrm{}}u\varphi ),$$
(35)
where
$`J_u=J^{\mathrm{}}u\varphi 𝒬^{0,n1}_{\mathrm{}}`$
is called the symmetry current along the vector field $`u`$. If $`L`$ is an $`r`$-order Lagrangian, we come to the familiar expression for a symmetry current
$`J_u=J^{\mathrm{}}u\varphi =h_0(J^{2r1}u\rho _L)+\phi `$
where $`\rho _L`$ is a $`(2r1)`$-order Lepagean equivalent of the Lagrangian $`L`$ , and $`\phi `$ is a $`d_H`$-closed form. Of course, a symmetry current $`J_u`$ in the expression (35) is not defined uniquely, but up to a $`d_H`$-closed form. In finite order Lagrangian formalism, one usually sets
$`J_u=h_0(J^{2r1}u\rho _L),`$
but the problem of a choice of a Lepagean equivalent $`\rho _L`$ remains .
If the Lie derivative (35) vanishes, we obtain the weak conservation law
$`d_HJ_u=d_\lambda J_u^\lambda \omega 0`$
on the shell Ker$`\delta (L)`$, i.e., the global section $`d_HJ_u`$ of the sheaf $`𝔔_{\mathrm{}}^{0,n}`$ on $`J^{\mathrm{}}Y`$ takes zero values at points of the subspace Ker$`\delta (L)J^{\mathrm{}}Y`$ given by the condition $`\delta (L)=0`$. Then, one can say that the divergence $`d_HJ_u`$ is a relative $`d_H`$-cocycle on the pair of topological spaces $`(J^{\mathrm{}}Y,\mathrm{Ker}\delta (L))`$. Of course, it is a $`d_H`$-coboundary, but not necessarily a relative $`d_H`$-coboundary since $`J_u0`$. Therefore, the divergence $`d_HJ_u`$ of a conserved current $`J_u`$ can be characterized by elements of the relative $`d_H`$-cohomology group $`H_{\mathrm{rel}}^n(J^{\mathrm{}}Y,\mathrm{Ker}\delta (L))`$ of the pair $`(J^{\mathrm{}}Y,\mathrm{Ker}\delta (L))`$.
For instance, any conserved Noether current in the Yang–Mills gauge theory on a principal bundle $`P`$ with a structure group $`G`$ is well known to reduce to a superpotential, i.e., $`J_u=W+d_HU`$ where $`W0`$ . Its divergence $`d_HJ_u`$ belongs to the trivial element of the relative cohomology group $`H_{\mathrm{rel}}^n(J^2Y,\mathrm{Ker}\delta (L_{\mathrm{YM}}))`$, where $`Y=J^1P/G`$.
Let now $`N^nX`$ be an $`n`$-dimensional submanifold of $`X`$ with a compact boundary $`N^n`$. Let $`s`$ be a section of the fibre bundle $`YX`$ and $`\overline{s}=J^{\mathrm{}}s`$ its infinite order jet prolongation, i.e., $`y_\mathrm{\Lambda }^i\overline{s}=d_\mathrm{\Lambda }s^i`$, $`0<|\mathrm{\Lambda }|`$. Let us assume that $`\overline{s}(N^n)\mathrm{Ker}\delta (L)`$. Then, the quantity
$$\underset{N^n}{}\overline{s}^{}d_HJ_u=\underset{N^n}{}\overline{s}^{}J_u$$
(36)
depends only on the relative cohomology class of the divergence $`d_HJ_u`$. For instance, in the above mentioned case of gauge theory, the quantity (36) vanishes.
Let $`N^{n1}`$ be a compact $`(n1)`$-dimensional submanifold of $`X`$ without boundary, and $`s`$ a section of $`YX`$ such that $`\overline{s}(N^{n1})\mathrm{Ker}\delta (L)`$. Let $`J_u`$ and $`J_u^{}`$ be two currents in the first variational formula (35). They differ from each other in a $`d_H`$-closed form $`\phi `$. Then, the difference
$$\underset{N^{n1}}{}\overline{s}^{}(J_uJ_u^{})$$
(37)
depends only on the homology class of $`N^{n1}`$ and the de Rham cohomology class of $`\overline{s}^{}\phi `$. The latter is an image of the $`d_H`$-cohomology class of $`\phi `$ under the morphisms
$`H^{n1}(d_H)\stackrel{h_0}{}H^{n1}(Y)\stackrel{s^{}}{}H^{n1}(X).`$
In particular, if $`N^{n1}=N^n`$ is a boundary, the quantity (37) always vanishes.
## 11 Appendix A
If $`QZ`$ is an affine bundle coordinated by $`(z^\lambda ,q^i)`$, the map
$`[0,1]\times Q(t,z^\lambda ,q^i)(z^\lambda ,tq^i+(1t)s^i(z)),`$
where $`s`$ is a global section of $`QZ`$, provides a homotopy from $`Q`$ to $`Z`$ identified with $`s(Z)Q`$. Similarly, a desired homotopy from $`J^{\mathrm{}}Y`$ to $`Y`$ is constructed
Let $`\gamma _{(k)}`$, $`k1`$, be global sections of the affine jet bundles $`J^kYJ^{k1}Y`$. Then, we have a global section
$$\gamma :Y(x^\lambda ,y^i)(x^\lambda ,y^i,y_\mathrm{\Lambda }^i=\gamma _{(|\mathrm{\Lambda }|)}{}_{\mathrm{\Lambda }}{}^{i}\gamma _{(|\mathrm{\Lambda }|1)}\mathrm{}\gamma _{(1)})J^{\mathrm{}}Y.$$
(38)
of the open surjection $`\pi _0^{\mathrm{}}:J^{\mathrm{}}YY`$. Let us consider the map
$`[0,1]\times J^{\mathrm{}}Y(t;x^\lambda ,y^i,y_\mathrm{\Lambda }^i)(x^\lambda ,y^i,y_\mathrm{\Lambda }^i)J^{\mathrm{}}Y,0<|\mathrm{\Lambda }|,`$ (39)
$`y_\mathrm{\Lambda }^i=f_k(t)y_\mathrm{\Lambda }^i+(1f_k(t))\gamma _{(k)}{}_{\mathrm{\Lambda }}{}^{i}(x^\lambda ,y^i,y_\mathrm{\Sigma }^i),|\mathrm{\Sigma }|<k=|\mathrm{\Lambda }|,`$
where $`f_k(t)`$ is a continuous monotone real function on $`[0,1]`$ such that
$$f_k(t)=\{\begin{array}{cc}0,\hfill & t12^k,\hfill \\ 1,\hfill & t12^{(k+1)}.\hfill \end{array}$$
(40)
A glance at the transition functions (3) shows that, although written in a coordinate form, this map is globally defined. It is continuous because, given an open subset $`U_kJ^kY`$, the inverse image of the open set $`(\pi _k^{\mathrm{}})^1(U_k)J^{\mathrm{}}Y`$, is the open subset
$`(t_k,1]\times (\pi _k^{\mathrm{}})^1(U_k)(t_{k1},1]\times (\pi _{k1}^{\mathrm{}})^1(\pi _{k1}^k[U_k\gamma _{(k)}(J^{k1}Y)])\mathrm{}`$
$`[0,1]\times (\pi _0^{\mathrm{}})^1(\pi _0^k[U_k\gamma _{(k)}\mathrm{}\gamma _{(1)}(Y)])`$
of $`[0,1]\times J^{\mathrm{}}Y`$, where $`[t_r,1]=\mathrm{supp}f_r`$. Then, the map (39) is a desired homotopy from $`J^{\mathrm{}}Y`$ to $`Y`$ which is identified with its image under the global section (38).
## 12 Appendix B
Proof. For $`q=0`$, the manifested isomorphism follows from the fact that $`H^0(Z,S)=\mathrm{\Gamma }(Z,S)`$ for any sheaf $`S`$ on $`Z`$. To prove other ones, let us replace the exact sequence (8) with
$`0S\stackrel{h}{}S_0\stackrel{h^0}{}S_1\stackrel{h^1}{}\mathrm{}\stackrel{h^{p2}}{}S_{p1}\stackrel{h^{p1}}{}\mathrm{Ker}h^p0`$
and consider the short exact sequences
$`0S\stackrel{h}{}S_0\stackrel{h^0}{}\mathrm{Ker}h^10,`$
$`0\mathrm{Ker}h^{r1}\stackrel{\mathrm{in}}{}S_{r1}\stackrel{h^{r1}}{}\mathrm{Ker}h^r0,1<rp.`$
They give the corresponding exact cohomology sequences
$`0H^0(Z,S)H^0(Z,S_0)H^0(Z,\mathrm{Ker}h^1)H^1(Z,S)`$
$`H^1(Z,S_0)\mathrm{},`$ (41)
$`0H^0(Z,\mathrm{Ker}h^{r1})H^0(Z,S_{r1})H^0(Z,\mathrm{Ker}h^r)`$
$`H^1(Z,\mathrm{Ker}h^{r1})H^1(Z,S_{r1})\mathrm{}.`$ (42)
Since sheaves $`S_r`$, $`0r<p`$, are acyclic, the exact sequence (41) falls into
$`0H^0(Z,S)H^0(Z,S_0)H^0(Z,\mathrm{Ker}h^1)H^1(Z,S)0,`$
$`H^k(Z,\mathrm{Ker}h^1)=H^{k+1}(Z,\mathrm{Ker}h^0),1k,`$ (43)
and, similarly, the exact sequence (42) does
$`0H^0(Z,\mathrm{Ker}h^{r1})H^0(Z,S_{r1})H^0(Z,\mathrm{Ker}h^r)`$
$`H^1(Z,\mathrm{Ker}h^{r1})0,`$ (44)
$`H^k(Z,\mathrm{Ker}h^r)=H^{k+1}(Z,\mathrm{Ker}h^{r1}),1k.`$ (45)
The equalities (45) for the couples of numbers $`(k=m,r=qm)`$, $`1mq2`$, and the equality (43) for $`k=q1`$ lead to the chain of isomorphisms
$$H^1(Z,\mathrm{Ker}h^{q1})=H^2(Z,\mathrm{Ker}h^{q2})=\mathrm{}=H^q(Z,\mathrm{Ker}h^0)=H^q(Z,S).$$
(46)
The exact sequence (44) for $`r=q`$ contains the exact sequence
$$H^0(Z,S_{q1})\stackrel{h_{}^{q1}}{}H^0(Z,\mathrm{Ker}h^q)H^1(Z,\mathrm{Ker}h^{q1})0.$$
(47)
Since $`H^0(Z,S_{q1})=\mathrm{\Gamma }(Z,S_{q1})`$ and $`H^0(Z,\mathrm{Ker}h^q)=\mathrm{Ker}h_{}^q`$, the result follows from (46) and (47) for $`0<qp`$. $`\mathrm{}`$
|
warning/0005/quant-ph0005062.html
|
ar5iv
|
text
|
# The Entanglement of Formation for Isotropic States
\[
## Abstract
We give an explicit expression for the entanglement of formation for isotropic density matrices in arbitrary dimensions in terms of the convex hull of a simple function. For two qutrit isotropic states we determine the convex hull and we have strong evidence for its exact form for arbitrary dimension. Unlike for two qubits, the entanglement of formation for two qutrits or more is found to be a nonanalytic function of the maximally entangled fraction in the regime where the density matrix is entangled.
\]
One of the main goals in quantum information theory is to develop a theory of entanglement. A cornerstone of this theory will be a good measure of bipartite entanglement. Such a measure must obey the essential property that the entanglement of a bipartite density matrix $`\rho `$ which is shared by Alice and Bob cannot increase, on average, under local quantum operations and classical communication ($`LO+CC`$) between Alice and Bob. In this way, the entanglement captures the truly quantum correlations in a bipartite density matrix. For pure bipartite states a good measure of entanglement has been found, it is the following quantity:
$$E(|\psi \psi |)=S(\mathrm{Tr}_B(|\psi \psi |)),$$
(1)
where $`S(\rho )`$ is the von Neumann entropy of $`\rho `$, i.e. $`S(\rho )=\mathrm{Tr}\rho \mathrm{log}\rho `$ and $`\mathrm{Tr}_B(|\psi \psi |)`$ is the reduced density matrix that we obtain by tracing out over Bob’s quantum system. This measure $`E`$ is unique if one requires the entanglement to obey a set of natural properties, such as convexity, non-increase under local measurements, asymptotic continuity, partial additivity and normalization. Moreover, $`E`$ is a measure of the asymptotic entanglement costs of making the state $`|\psi `$ out of a canonical set of states, which we can choose to be EPR singlets $`\frac{1}{\sqrt{2}}(|01|10)`$, which have $`E=1`$. This process is reversible, in the sense that one can concentrate a set of $`n`$ states $`|\psi `$ with entanglement $`E`$ to a smaller set $`m=En`$ EPR singlets.
The situation for mixed states is much more complex. In Ref. a first measure of mixed state entanglement, called the entanglement of formation, was introduced. This measure is a candidate for measuring the asymptotic costs of making the density matrix out of a supply of EPR singlets. There are no mixed density matrices for which this statement has been proved, but neither have counterexamples been found so far. The search for a possible discrepancy between the entanglement of formation and the asymptotic entanglement costs is hampered by the fact that we know the entanglement of formation only for two qubit systems; Wootters found an analytic expression for the entanglement of formation for all two qubit density matrices.
In this Letter we present the first calculation of the entanglement of formation of a class of density matrices in dimensions higher than $`𝐂^2𝐂^2`$. We explicitly determine the entanglement of formation for two qutrit density matrices in this class and we find an expression in arbitrary dimension in terms of the convex hull of a simple function. We conjecture the explicit form of this convex hull, which can be easily verified in a given dimension. Surprisingly, the entanglement of formation is found to be a nonanalytic function of the parameter characterizing the class of states that we consider.
Let us start by recalling the definition of the entanglement of formation. Let $`_\rho =\{p_i,|\psi _i\}`$ be an ensemble of pure states which form a decomposition of $`\rho =_ip_i|\psi _i\psi _i|`$. The entanglement of formation for mixed states $`\rho `$ is defined as
$$E(\rho )=\underset{=\{p_i,|\psi _i\}}{\mathrm{min}}\underset{i}{}p_iE(|\psi _i\psi _i|).$$
(2)
In this Letter we will consider the class of density matrices, sometimes called isotropic density matrices, which are convex mixtures of a maximally entangled state and the maximally mixed state:
$$\rho _F=\frac{1F}{d^21}\left(\mathrm{𝟏}|\mathrm{\Psi }^+\mathrm{\Psi }^+|\right)+F|\mathrm{\Psi }^+\mathrm{\Psi }^+|,$$
(3)
for $`0F1`$ and $`|\mathrm{\Psi }^+=\frac{1}{\sqrt{d}}_{i=1}^d|ii`$. For $`F1/d`$ these density matrices are separable . The entanglement of formation for states with $`d=2`$ is equal to
$$H_2(\mu ),\mu =\frac{1}{2}+\sqrt{F(1F)},$$
(4)
where $`H_2(.)`$ is the binary entropy function. The states $`\rho _F`$ have the important property that they are invariant under the operation $`UU^{}`$ for any unitary transformation $`U`$. The $`LO+CC`$ “twirling” superoperator $`𝒮^{UU^{}}`$ is defined as
$$𝒮^{UU^{}}(\rho )=\frac{1}{Vol(U)}𝑑UUU^{}\rho U^{}U_{}^{}{}_{}{}^{}.$$
(5)
In Ref. the Schmidt number of the isotropic states was determined. Instead of making an isotropic state out of a set of maximally entangled states, we ask how to construct an isotropic state with a given $`F`$ out of some state characterized by a Schmidt vector $`\stackrel{}{\mu }`$. So, let us take an arbitrary initial pure state $`|\psi =_{i=1}^d\sqrt{\mu _i}|a_i,b_i`$ and consider the effect of twirling. We can write $`|\psi =U_AU_B_i\sqrt{\mu _i}|i,i`$ and thus
$`𝒮^{UU^{}}\left({\displaystyle \underset{i,j}{}}\sqrt{\mu _i\mu _j}|a_i,b_ia_j,b_j|\right)=`$ (6)
$`𝒮^{UU^{}}((\mathrm{𝟏}V){\displaystyle \underset{i,j}{}}\sqrt{\mu _i\mu _j}|i,ij,j|(\mathrm{𝟏}V^{})),`$ (7)
where $`V=U_A^TU_B`$. We define $`v_{ij}=i|V|j`$. The twirled state becomes
$$𝒮^{UU^{}}(|\psi \psi |)=\frac{|_i\nu _i|^2}{d}P_++\frac{1|_i\nu _i|^2/d}{d^21}(\mathrm{𝟏}P_+),$$
(8)
where $`\nu _i=\sqrt{\mu _i}v_{ii}`$ and $`P_+=|\mathrm{\Psi }^+\mathrm{\Psi }^+|`$. When we choose $`V=\mathrm{𝟏}`$ we find the density matrix $`\rho _F`$ at $`F=[_i\sqrt{\mu _i}]^2/d`$. For general $`V`$ one can bound
$$|\underset{i}{}\nu _i|^2[\underset{i}{}|\nu _i|]^2[\underset{i}{}\sqrt{\mu _i}]^2,$$
(9)
since $`|v_{ii}|1`$ for all $`i`$. Thus the largest value for $`F`$ is obtained by choosing the initial state $`_{i=1}^d\sqrt{\mu _i}|ii`$.
The use of symmetry makes it possible to give a simplified expression for the entanglement of formation for isotropic states:
Lemma 1 The entanglement of formation for isotropic states in $`𝐂^d𝐂^d`$ ($`d2`$) for $`F(1/d,1]`$ is given by
$$E(\rho _F)=co(R(F)),$$
(10)
where $`co(g)`$ denotes the convex hull of the function $`g`$ and $`R(F)`$ is defined as
$$R(F)=\underset{\stackrel{}{\mu }}{\mathrm{min}}\left\{H(\stackrel{}{\mu })|F=[\underset{i=1}{\overset{d}{}}\sqrt{\mu _i}]^2/d\right\},$$
(11)
and $`\stackrel{}{\mu }`$ is a Schmidt vector.
Proof Assume that there exists an optimal decomposition of $`\rho _F`$ formed by the ensemble $`\{p_i,|\psi _i(\stackrel{}{\mu }^i)\}`$, where $`\stackrel{}{\mu }^i`$ denotes the Schmidt vector of the state $`|\psi _i`$. By twirling the l.h.s. and r.h.s. of the equation $`\rho _F=p_i|\psi _i\psi _i|`$ we obtain that $`\rho _F=_ip_i\rho _{F_i}=\rho _{_ip_iF_i}`$ where
$$F_i(V_i,\stackrel{}{\mu }^i)=\frac{|_kv_{kk}\sqrt{\mu _k^i}|^2}{d},$$
(12)
as in Eq. (8). Since the decomposition is optimal, each Schmidt vector $`\stackrel{}{\mu }^i`$ has minimal entropy under this constraint. Consider the function
$$R_V(F)=\underset{\stackrel{}{\mu }}{\mathrm{min}}\{H(\stackrel{}{\mu })|F=|\underset{i=1}{\overset{d}{}}v_{ii}\sqrt{\mu _i}|^2/d\}.$$
(13)
An optimal decomposition of $`\rho _F`$ is a convex combination of pure states each of which corresponds to a certain $`F`$ under twirling. Thus the entanglement of formation $`E(\rho _F)`$ can be obtained by taking the convex hull of the functions $`co(R_V(F))`$. We can make an additional simplification. Eq. (9) implies that $`R_V(F)=R_\mathrm{𝟏}(F^{})R(F^{})`$ where $`F^{}F`$ for every $`V`$. Thus instead of taking the convex hull of all functions $`co(R_V(F))`$, we can take the convex hull of function $`R_+(F)=\mathrm{min}_x\{R(x)|xF\}`$. In Lemma 2 we will determine $`R(F)`$ and it is not hard to show that $`R(F)`$ is a monotonically increasing function of $`F`$. It follows then that $`R_+(F)=R(F)`$ and $`co(R_+(F)=co(R(F))`$. $`\mathrm{}`$
We now determine the function $`R(F)`$ defined in Eq. (11). Since all the equations are symmetric in $`\mu _i`$, we can restrict ourselves to solutions which satisfy $`\mu _1\mu _2\mathrm{}\mu _d`$. With the method of Lagrange multipliers we get a necessary condition for the minimum
$$1\mathrm{log}\mu _i+\mathrm{\Lambda }_1+\frac{\mathrm{\Lambda }_2}{2}\mu _i^{\frac{1}{2}}=0,$$
(14)
where $`\mathrm{\Lambda }_1,\mathrm{\Lambda }_2`$ denote the Lagrange multipliers. For fixed $`\mathrm{\Lambda }_1,\mathrm{\Lambda }_2`$ this determines the whole set $`\{\mu _i\}`$. Setting $`\mu _i=\frac{1}{q_i^2}`$ we obtain an expression of the form $`\mathrm{log}q_i=Aq_i+B`$ where $`A,B`$ only depend on $`\mathrm{\Lambda }_1,\mathrm{\Lambda }_2`$. Since a convex and a concave function cross each other in at most two points, this equation has maximally two possible nonzero solutions for $`q_i`$. Therefore all Schmidt vectors $`\stackrel{}{\mu }`$ that are possible candidates for the minimum have to satisfy the condition $`\mu _i\{\gamma ,\delta ,0\}`$. Let $`n`$ be the number of entries where $`\mu _i=\gamma `$ and $`m`$ the number of entries where $`\mu _i=\delta `$. The minimization problem has been reduced considerably: For fixed $`n,m`$, $`n+md`$, we minimize the function
$$nh(\gamma )+mh(\delta ),$$
(15)
where $`h(x)=x\mathrm{log}x`$, under the constraints
$$\{n\gamma +m\delta =1,n\sqrt{\gamma }+m\sqrt{\delta }=\sqrt{dF}\},$$
(16)
The constraints give rise to a quadratic equation in $`\sqrt{\gamma }`$ which provides two possible solutions for $`\gamma `$ for every choice of $`n,m`$:
$$\gamma _{nm}^\pm (F)=\left(\frac{\sqrt{dF}n\pm \sqrt{mn(m+ndF)}}{n(n+m)}\right)^2.$$
(17)
With the first constraint we get the corresponding $`\delta _{nm}^\pm (F)=(1n\gamma _{nm}^\pm (F))/m`$. Since $`\gamma _{mn}^{}=\delta _{nm}^+`$, the function in Eq. (15) takes the same value for $`\gamma _{nm}^+`$ and $`\gamma _{mn}^{}`$. Therefore we can restrict ourselves to the solutions $`\gamma _{nm}:=\gamma _{nm}^+`$. The pointwise minimum over all possible choices for $`n,m`$ of
$$R_{nm}(F)=H_2(n\gamma _{nm})+n\gamma _{nm}\mathrm{log}\frac{n}{m}+\mathrm{log}m,$$
(18)
defined on the domain $`\frac{n}{d}F\frac{n+m}{d}`$, is the required function $`R(F)`$. The restriction on the domain comes from requiring that $`\gamma _{nm}`$ is a proper solution of Eq. (17) which implies that $`F\frac{n+m}{d}`$. On the other hand we demand that $`\delta _{nm}0`$ which implies that $`F\frac{n}{d}`$. In this regime one can verify that $`\gamma _{nm}\delta _{nm}`$.
When $`m=0`$, $`\gamma `$ and $`F`$ are uniquely determined by the constraints, i.e. $`F=\frac{n}{d}`$. Since $`R_{n0}(\frac{n}{d})=R_{n^{}m^{}}(\frac{n}{d})`$ for all $`n^{}+m^{}=n`$, we can neglect these cases.
When $`d=3`$, what remains is a minimization over the three functions $`R_{12}(F)`$, $`R_{21}(F)`$ and $`R_{11}(F)`$, which are plotted in Fig. 1. For $`d=3`$ we get $`R(F)=R_{12}(F)`$ . Thus the optimal vector $`\stackrel{}{\mu }`$ is always of the form
$$\stackrel{}{\mu }=\{\gamma ,\delta ,\delta \},$$
(19)
satisfying $`\gamma \delta `$.
The case $`d=3`$ is the important one, since it turns out that we can relate all the higher dimensional minima to $`d=3`$ and prove that
Lemma 2 For $`d3`$ the function $`R(F)=R_{1,d1}(F)`$.
Proof The case $`d=3`$ is discussed above. Note that $`R_{1,d1}(1/d)=0`$, which is clearly minimal, so we provide a proof for $`F>1/d`$. Let the minimum be attained in $`d>3`$ dimensions by a vector $`\stackrel{}{\mu }=\{\mu _i\}`$. Let us select some subset of the entries of $`\stackrel{}{\mu }`$, the set $`\{\mu _{i_j}\}_{j=1}^d^{}`$, where $`_{j=1}^d^{}\mu _{i_j}=k1`$. Since $`\stackrel{}{\mu }`$ is the minimum, it follows that the set $`\{\mu _{i_j}\}_{j=1}^d^{}`$ is the minimum when we keep the other entries of the vector $`\stackrel{}{\mu }`$ fixed. Let $`\mu _j^{}:=\frac{\mu _{i_j}}{k}`$. The vector $`\stackrel{}{\mu ^{}}`$ is the solution for the minimization of
$$\underset{i=1}{\overset{d^{}}{}}h(\mu _i^{}k)=k\left[\underset{i=1}{\overset{d^{}}{}}h(\mu _i^{})\right]+h(k),$$
(20)
under the constraints $`_{j=1}^d^{}\mu _j^{}=1`$ and $`_{j=1}^d^{}\sqrt{\mu _j^{}}=𝒞`$ where
$$𝒞=\sqrt{\frac{dF}{k}}\frac{1}{\sqrt{k}}\underset{i|j,ii_j}{}\sqrt{\mu _i}.$$
(21)
This last equation can always be written as $`𝒞=\sqrt{d^{}F^{}}`$ for some $`F^{}`$. Thus the restricted minimization problem is equivalent to a $`d^{}`$-dimensional version of the original problem, up to the scaling factor $`k`$ and the additive term $`h(k)`$. When $`F^{}\frac{1}{d^{}}`$, we know that solution of this minimization problem is given by a Schmidt vector $`\stackrel{}{\mu }^{}`$ which corresponds to an unentangled state, i.e. it is of the form $`\stackrel{}{\mu }^{}=\{1,0,\mathrm{},0\}`$. Let us choose three arbitrary $`\mu _i`$ out of the optimal vector $`\stackrel{}{\mu }`$. When the resulting $`F^{}\frac{1}{3}`$, it follows that $`\stackrel{}{\mu }^{}=\{1,0,0\}`$. When $`F^{}>\frac{1}{3}`$, the three entries of $`\stackrel{}{\mu }^{}`$ have to satisfy Eq. (19). So in fact, in both cases they satisfy Eq. (19). Suppose now that one entry of $`\stackrel{}{\mu }`$ is equal to zero. Then it follows that $`\stackrel{}{\mu }`$ cannot have two nonzero entries since this would violate condition (19), in other words it must be that $`\stackrel{}{\mu }=\{1,0,\mathrm{},0\}`$. But this is a solution for $`F=1/d`$. Therefore we get $`n+m=d`$ for $`F>1/d`$. Suppose that $`n2`$. We can choose the vector $`\{\gamma ,\gamma ,\delta \}`$ satisfying $`\gamma \delta `$. Then condition (19) implies that $`\gamma =\delta `$. This implies that all entries of $`\stackrel{}{\mu }`$ are identical, or $`\stackrel{}{\mu }=\{\frac{1}{d},\mathrm{},\frac{1}{d}\}`$. This corresponds to a maximally entangled state, which is the unique solution for $`F=1`$. Therefore $`n=1`$ and $`m=d1`$. $`\mathrm{}`$
Lemma 1 and Lemma 2 together result in
Theorem 1 The entanglement of formation $`E(\rho _F)`$ for isotropic states in $`𝐂^d𝐂^d`$ ($`d2`$) for $`F(1/d,1)`$ is given by
$$E(\rho _F)=co(R_{1,d1}(F)),$$
(22)
where
$$R_{1,d1}(F)=H_2(\gamma (F))+(1\gamma (F))\mathrm{log}(d1),$$
(23)
with
$$\gamma (F)=\frac{1}{d}\left(\sqrt{F}+\sqrt{(d1)(1F)}\right)^2.$$
(24)
For $`d=3`$ the first and second derivative of the function $`R_{12}(F)`$ are plotted in Fig. 2. The figure shows that the function $`R_{12}(F)`$ is not convex near $`F=1`$; its second derivative is not positive. In order to determine $`co(R_{12}(F))`$ for $`d=3`$ we solve the following equations. Let $`E_{line}(F)=aF+\mathrm{log}3a`$ be the line crossing through the point $`(1,\mathrm{log}3)`$. We solve (1) $`E_{line}(F)=R_{12}(F)`$ and (2) $`\frac{dE_{line}}{dF}=a=\frac{dR_{12}}{dF}`$ for $`a`$ and $`F`$. Figure 2 indicates that $`R_{12}(F)`$ is monotonically increasing and that there is only one region where $`R_{12}(F)`$ is not convex, namely near $`F=1`$. Therefore the solution to the equations will be unique: we find that $`F=8/9`$ and $`a=3`$. For higher dimensions, we conjecture, based on examining these two equations, that the entanglement of formation in $`𝐂^d𝐂^d`$ is given by
$$E(\rho _F)=\{\begin{array}{cc}0,\hfill & F\frac{1}{d},\hfill \\ R_{1,d1}(F),\hfill & F(\frac{1}{d},\frac{4(d1)}{d^2}),\hfill \\ \frac{d\mathrm{log}(d1)}{d2}(F1)+\mathrm{log}d,\hfill & F[\frac{4(d1)}{d^2},1].\hfill \end{array}$$
(25)
The correctness of this solution can easily be verified for a given $`d`$ by plotting the function $`R_{1,d1}(F)`$ and its second derivative and noting the convex hull of $`R_{1,d1}(F)`$ is obtained by calculating where $`R_{1,d1}(F)`$ meets the line going through the point $`(F=1,E=\mathrm{log}d)`$ and the tangent of $`R_{1,d1}^{}(F)`$ equals the slope of this line.
It is surprising to find that $`E(\rho _F)`$ is nonanalytic in the region where $`\rho _F`$ is an entangled density matrix. Another feature of our solution is that for, say, $`d=3`$ and $`F>8/9`$ an optimal decomposition of $`\rho _F`$ is not one in which every pure state has an equal amount of entanglement. Indeed, the optimal decomposition that we find, is a mixture of the maximally entangled state and the ensemble of states $`|\psi `$ obtained by twirling, each of which has entanglement $`E=1/3+\mathrm{log}3`$. Since every state in the optimal decomposition of $`\rho _F`$ has, under twirling, a value of entanglement on $`R_{1,d1}(F)`$, every optimal decomposition of $`\rho _F`$ for $`d=3`$ in the range $`F>8/9`$ will be a mixture of the maximally entangled state and some less entangled states. This is in contrast with optimal decomposition for $`E`$ for two qubits. For $`F>8/9`$ more than $`d^2=9`$ pure states must be used in the optimal decomposition of $`\rho _F`$. We make $`\rho _F`$ from a maximally entangled state and the state $`\rho _{F=8/9}`$ which has rank 9, and thus needs at least 9 states in its optimal decomposition. In total, this gives 10 states. For $`F>8/9`$ there is no optimal decomposition with fewer states: one always has to mix in the maximally entangled state with some probability. The remaining state $`\rho _F^{}`$ either has rank 9 (like $`\rho _{F=8/9}`$) or a lower rank. If it has a lower rank, it must be separable, which would imply that the optimal decomposition is made from mixing a separable state with a maximally entangled state which we know to be false. This is the first example of an entangled state for which it is proved that the number of pure states in the optimal decomposition exceeds the rank of the state (see Ref. for separable states with this property).
Crucial in our method is the invariance of the isotropic states under a symmetry group of local operations. A result similar to Lemma 2 will hold for example for the class of Werner states which are invariant under the transformation $`UU`$ for all $`UU(d)`$. Let $`𝒮^{UU}`$ be defined as in Eq. (5), but with omission of the complex conjugation. The Werner states $`\rho _p^W`$ are characterized by a single parameter $`p`$. One can prove that $`E(\rho _p^W)=co(R(p))`$ where
$$R(p)=\mathrm{min}\{E(|\psi \psi |)|𝒮^{UU}(|\psi \psi |)=\rho _p^W\}.$$
(26)
It may thus be possible to carry out a similar analysis as was done here for the Werner states. In a further generalization one could consider the entanglement of formation for $`gg`$ or $`gg^{}`$ invariant states where $`gG`$ and $`G`$ is a subgroup of $`U(d)`$.
Acknowledgments: BMT would like to thank David DiVincenzo, Julia Kempe, John Smolin and Armin Uhlmann for interesting discussions. BMT acknowledges support of the ARO under contract number DAAG-55-98-C-0041. KGHV would like to thank R.F. Werner for discussions. KGHV is supported by Deutsche Forschungs Gemeinschaft (DFG).
|
warning/0005/cond-mat0005172.html
|
ar5iv
|
text
|
# The dynamics of the spherical p-spin model: from microscopic to asymptotic
## Abstract
We have numerically investigated the mean-field dynamics of the the $`p`$-spin interaction spin glass model with p=3 using an efficient method of integrating the dynamic equations. We find a new time scale associated with the onset of the breakdown of the fluctuation-dissipation theorem in the intermediate time regime. We also find that the off-equilibrium relaxation exhibits a sub-aging behavior in the intermediate times and crosses over to a simple aging in the asymptotic regime.
Since the recognition of possible deep connection between structural glass and some class of spin glass systems such as the $`p`$-spin model , the spherical $`p`$-spin model , due to its analytic accessibility, has been the subject of intense research in both statics and dynamics. Here we examine some open questions (such as scaling behavior) of the off-equilibrium dynamics of the mean field spherical $`p`$-spin model by developing a new method of integrating the dynamic equations of the model.
We consider the system of $`N`$ spins $`S_1,S_2,\mathrm{}S_N`$ which interact via Hamiltonian
$$H=\underset{1i_1<i_2\mathrm{}<i_pN}{}J_{i_1i_2\mathrm{}i_p}S_{i_1}S_{i_2}\mathrm{}S_{i_p}$$
(1)
where the spins are continuous variables subject to the spherical constraint $`_{i=1}^NS_i^2=N`$. The coupling $`J_{i_1i_2\mathrm{}i_p}`$ is a Gaussian random variable with zero mean and variance $`[J_{i_1i_2\mathrm{}i_p}^2]_J=p!/2N^{p1}`$. It is well known that this model (with $`p>2`$) exhibits an equilibrium phase transition at a finite temperature $`T=T_S`$ from paramagnetic phase to spin glass phase characterized by one-step replica-symmetry breaking. This temperature is lower than the than the dynamical freezing temperature $`T_D`$, below which the dynamics is non ergodic.
For the time evolution of the system, we consider the following dissipative Langevin dynamics
$$\mathrm{\Gamma }^1_tS_i(t)=z(t)S_i(t)\frac{H}{S_i(t)}+\eta _i(t)$$
(2)
where $`z(t)`$ is a Lagrange multiplier to enforce the spherical constraint for all times. In order to satisfy the detailed balance, the thermal noise $`\eta _i(t)`$ is chosen to be Gaussian with zero mean and variance $`\eta _i(t)\eta _j(t^{})=2\mathrm{\Gamma }^1T\delta _{ij}\delta (tt^{})`$. The inverse of the kinetic coefficient $`\mathrm{\Gamma }`$ sets the microscopic time scale, which henceforth is set to unity.
This Langevin dynamics governs the time evolution of the system starting from, for example, the random disordered spin configuration toward thermal equilibrium. Two physical quantities of interest, which quantitatively characterize this time evolution, are the two-time correlation function $`C(t,t_w)=_{i=1}^N[<S_i(t)S_i(t_w)>]_J/N`$ where $`\mathrm{}`$ and $`[\mathrm{}]_J`$ represent averages over thermal noise and random coupling, respectively, and the response function $`R(t,t_w)=_{i=1}^N[\frac{<S_i(t)>}{h_i(t_w)}]_J/N`$ where $`h_i(t_w)`$ is the external field turned on at time $`t_w`$. In equilibrium, these quantities become time-translation invariant, i.e., $`C(t,t_w)=C(\tau )`$ and $`R(t,t_w)=R(\tau )`$, $`\tau tt_w`$. Moreover, there exists a fundamental relationship between the correlation and the response, known as the fluctuation-dissipation theorem (FDT), which takes the form for the present irreversible dynamics
$$R(\tau )=\frac{\theta (\tau )}{T}\frac{C(\tau )}{\tau }$$
(3)
where $`\theta (\tau )`$ is the unit step function which reflects the causality for the response function. We will see below that the non-equilibrium dynamics of the present system manifests an explicit strong dependence of the two times $`t`$ and $`t_w`$ (aging) and an interesting modification of FDT .
At this stage, we find it very useful to introduce an integrated response function $`F(t,t_w)`$ defined as
$$F(t,t_w)_{t_w}^t𝑑sR(t,s).$$
(4)
From (4) $`F(t,t^{})=0`$ and $`F(t,t_w)/t_w=R(t,t_w)`$. FDT (3) then takes the form $`F(\tau )=\theta (\tau )[C(\tau )1]/T`$.
In the limit of $`N`$ going to infinity, one can treat the dynamics exactly using the standard functional method such as Martin-Siggia-Rose formalism . In particular, the dynamics is governed by the closed set of coupled equations of $`C`$ and $`F`$:
$`_tC(t,t_w)`$ $`=`$ $`z(t)C(t,t_w)+{\displaystyle \frac{p(p1)}{2}}{\displaystyle _0^t}𝑑sC^{p2}(t,s)(_sF(t,s))C(t_w,s)`$ (5)
$`+`$ $`{\displaystyle \frac{p}{2}}{\displaystyle _0^{t_w}}𝑑sC^{p1}(t,s)_sF(t_w,s),`$ (6)
$$_tF(t,t_w)=1z(t)F(t,t_w)+\frac{p(p1)}{2}_{t_w}^t𝑑sC^{p2}(t,s)(_sF(t,s))F(s,t_w).$$
(7)
where $`z(t)TpE(t)`$, $`E(t)`$ being the average energy per spin. The energy density $`E(t)`$ is related to $`C`$ and $`F`$ as $`E(t)=(p/2)_0^t𝑑sC^{p1}(t,s)_sF(t,s)`$. The same type of the equations have been derived in various physical contexts such as the dynamics of a long-range superconducting wire network ($`p=4`$), the dynamics of Amit-Roginsky model($`p=3`$) , and the dynamics of a particle in the random potential in large dimensional space .
As Cugliandolo and Kurchan (CK) have shown, the dynamics exhibits the two distinct regimes depending on the relative magnitude of the two times $`\tau (t\tau +t_w)`$ and $`t_w`$. The first regime (while $`t,t_w\mathrm{}`$, the time difference $`\tau tt_w`$ is finite, i.e., $`\tau /t_w0`$) is the regime where the time-translation invariance and FDT hold. That is, $`C(\tau +t_w,t_w)=C(\tau )`$, $`F(\tau +t_w,t_w)=F(\tau )`$ and $`TF(\tau )=(C(\tau )1)`$ independent of a given $`t_w`$. In this regime, the dynamic equation for $`C(\tau )`$ can be easily derived from the equation for the integrated response and is given by
$$(_\tau +T)C(\tau )+p\left(E_{\mathrm{}}+\frac{1}{2T}\right)(1C(\tau ))+\frac{p}{2T}_0^\tau 𝑑sC^{p1}(\tau s)\frac{dC(s)}{ds}=0.$$
(8)
where $`E_{\mathrm{}}`$ is the long time limit of the energy density $`E(t)`$, $`E_{\mathrm{}}lim_t\mathrm{}E(t)`$. One can recognize that this equation is quite similar to a schematic model developed in the context of the mode-coupling theory (MCT) for the glass transition in structural glasses . In particular, apart from the term involving $`(1C(\tau ))`$, (8) with $`p=3`$ is identical to the Leutheusser model (without the inertial term) . Hence one can sense that this equation, as in the Leutheusser model, may lead to a dynamic transition to a nonergodic phase where the long-time limit of $`C(\tau )`$ is non-vanishing. It is easy to show from (8) that the non-ergodicity parameter $`qlim_\tau \mathrm{}C(\tau )`$ is then related to $`E_{\mathrm{}}`$ via
$$(TpE_{\mathrm{}})+\frac{p}{2T}(1q^{p1})=\frac{T}{1q}$$
(9)
But the full understanding of the FDT dynamics requires the asymptotic value of the energy density, $`E_{\mathrm{}}`$, for which one has to consider the dynamics in different regime (aging regime); the two regimes are closely coupled to each other.
When the two times $`t`$ and $`t_w`$ are large and well separated, the relaxation is very slow, and hence the time derivatives in (6) and (7) can be ignored. In this situation, it is found that the scaling Ansatz for $`C`$ and $`F`$
$$C(t,t_w)=𝒞\left[\frac{h(t_w)}{h(t)}\right],F(t,t_w)=\left[\frac{h(t_w)}{h(t)}\right]$$
(10)
leads to
$$\frac{p(p1)}{2T^2}(1q)^2q^{p2}=1,$$
(11)
$$T(\lambda )=x𝒞(\lambda )[1(1x)q],$$
(12)
and
$$E_{\mathrm{}}=\frac{1}{2T}\left[1(1x)q^p\right]$$
(13)
with $`x=(p2)(1q)/q`$ for non vanishing $`q`$. The equation (12) with $`x<1`$ is a modification of FDT in aging regime and is one of the most important results obtained from the asymptotic analysis. Note that although the actual dynamics will select the form of $`h(t)`$ uniquely, the above asymptotic analysis holds for an arbitrary monotonically increasing function $`h(t)`$ (known as the time-reparametrization invariance): the function $`h(t)`$ remains undetermined within the asymptotic analysis.
We now go back to the equation (8) and discuss the dynamics of FDT regime. First note from (9) that $`q=0`$ is always a solution for all temperatures. We see from (11) that the non-vanishing $`q`$ starts to appear at the temperature $`T^{}=[2p^{1p}(p1)(p2)^{p2}]^{1/2}`$ at which $`q=q^{}=(p2)/p`$ ($`q^{}=1/3`$ and $`T^{}=2/3`$ for $`p=3`$). Thus below $`T^{}`$, starting from the initial state with zero energy density (e.g., the state with all spins up), the system always chooses the solution with higher energy (which is the highest TAP state at a given $`T`$ ) among these two solutions. Thus for $`T_D<TT^{}`$, $`q=0`$ is the genuine solution, and $`E_{\mathrm{}}=1/(2T)`$, $`x=1`$. The system is ergodic. The dynamic transition temperature $`T_D`$ is determined by $`x(T=T_D)=1`$ for $`q0`$. This condition with (11) lead to $`q_D=(p2)/(p1)`$ and $`T_D=[p(p2)^{p2}/(2(p1)^{p1})]^{1/2}`$ (for $`p=3`$, $`T_D=\sqrt{3/8}0.612\mathrm{}`$ and $`q_D=1/2`$). Therefore the term involving $`(1C(\tau ))`$ drops out in (8) and the resulting equation is the same as a schematic mode-coupling equation for supercooled liquids. Therefore the relaxation dynamics above the dynamic transition exhibits the well-known scaling laws derived with MCT . For $`TT_D`$ the system chooses $`q0`$, $`x<1`$, and $`E_{\mathrm{}}+1/(2T)>0`$. Hence the system is non-ergodic. Thus the term $`p(E_{\mathrm{}}+1/2T)(1C(\tau ))`$ in (8) is turned on. This additional term makes the dynamics in the FDT regime in the present model differ from that of the MCT for supercooled liquids. In particular, near the transition $`q(T)`$ shows a linear behavior $`q(T)=q_D+\text{const.}(T_DT)`$ instead of a square-root singularity observed in MCT. Also whereas the critical relaxation is seen very near and at the transition in MCT, in the present situation the correlation exhibits a critical relaxation $`C(\tau )=q+\text{const.}\tau ^a`$ for all temperatures below the transition with the exponent $`a`$ related to the FDT violating factor $`x`$ as $`\mathrm{\Gamma }^2[1a]/\mathrm{\Gamma }[12a]=x/2`$ , $`\mathrm{\Gamma }`$ being the gamma function.
Though the asymptotic analysis put forward by CK provided quantitative informations on the both FDT and the asymptotic aging dynamics, it leaves unexplored the non-equilibrium dynamics in the intermediate time regime. Moreover, the time reparametrization invariance leaves undetermined the form of the function $`h(t)`$ which the original set of causal dynamic equations would select in the course of its evolution. Therefore, we do not know what type of scaling feature the long time aging dynamics might exhibit.
In this work, we have developed an efficient way of integrating the mean-field dynamic equations (6) and (7), which can extend the solution to the time regime long enough so that one can clearly investigate the aforementioned important open questions. Basic idea is that since the relaxation becomes very slow at long times one can employ an adaptive integration time step . There are two crucial ingredients in the present method. First, one has to work with the integrated response instead of the response function since the integrated response relax much more slowly than the response (which is basically a time derivative of the correlation). Another ingredient is to increase the integration time step along the both two time directions in such a way that the fast relaxing time regime should have many integration points. The details of the present method will be given elsewhere.
Now we turn to the discussion of the numerical results. All presented results are for $`p=3`$. Here we focus on the dynamics below the transition. Shown in Fig. 1(a) is $`C`$ and $`F`$ for $`T=0.5`$. Both functions exhibit a strong $`t`$-dependence (aging effect) which persists up to longest time $`t`$. In contrast to the situation above the transition, here the aging will not be interrupted, and the system remains perennially out-of-equilibrium. The parametric plot $`TF(t,t\tau )`$ versus $`C(t,t\tau )`$ for $`T=0.5`$ shown in Fig. 2(b) exhibits several interesting features. After some time $`t10^2`$, independent of time $`t`$, the FDT is established in the range $`qC1`$. Then, after still longer time $`t10^5`$, all the curves with different times $`t`$ merge into a single straight line with slope close to the FDT violation parameter $`x0.565\mathrm{}`$. Note the huge time interval (more than 9 decades!) it takes to reach the asymptotic regime. Therefore in the asymptotic regime, the FDT and its violation can be characterized by the two straight lines with slopes $`1`$ and $`x`$ meeting at $`C=q`$: $`TF=1C`$ for $`qC1`$ and $`TF=x(1C)+(1x)(1q)`$ for $`Cq`$. This is in accordance with the result of the asymptotic analysis (12). Similar breakdown of FDT has been observed in simulations of supercooled liquids and spin glass .
One interesting new feature we find in the off-equilibrium dynamics in the intermediate time region is the existence of a new time scale associated with the onset of FDT violation. Note that the breakdown of FDT occurs sharply at $`C=q`$ (see Fig. 2(b)). Hence we are able to measure this time scale by reading off the times $`\tau _p`$ defined as $`C(t,t\tau _p)=q`$ for each fixed $`t`$. The double-log plot $`\tau _p`$ versus $`t`$ in the inset of Fig. 2(a) indeed demonstrates that $`\tau _p`$ shows a power law dependence on $`t`$ as $`\tau _pt^\varphi `$ with $`\varphi 0.68`$. Furthermore, the big difference in $`\tau _p`$ and $`t`$ for large $`t`$ implies that $`\tau _pt_w^\varphi `$ with the same exponent. Since $`\varphi `$ is smaller than $`1`$, we find a new time scale characterizing the aging dynamics in intermediate time region. It is thus highly desirable to have a theoretical development which can provide a detailed information on the dynamics in the intermediate times and on the crossover to the asymptotic time regime. The presence of this time scale in the spherical SK model ($`p=2`$) was demonstrated by Zippold et al .
We now discuss the scaling behavior in the aging regime. Recall that the asymptotic analysis loses information on the unique selection of $`h(t)`$. The scaling property in our numerical solution is important since it will be able to determine the specific form of the function $`h(t)`$ that the system actually selects. Figure 2(a) shows the relaxation of the correlation function $`C(\tau +t_w,t_w)`$ for different waiting times $`t_w`$ at $`T=0.5`$. Due to the adaptive integration steps along the both $`t`$ and $`t_w`$ axes, it is inevitable that the short time data for each fixed $`t_w`$ is lost. It is natural to first examine the possibility of $`h(t)`$ being a power of $`t`$, $`h(t)t^\gamma `$. This means that the correlation in the aging regime shows a scaling behavior $`C_A(\tau +t_w,t_w)=\widehat{C}(\tau /t_w)`$ (known as the simple aging). To check this scenario, we plot in Fig. 2(b) the correlation function $`C(\tau +t_w,t_w)`$ against the rescaled time $`\tau /t_w`$. Indeed, we observe that the relaxation data with longest waiting times collapse onto a single scaling curve. Thus we find that the relaxation obeys the simple aging in the asymptotic time regime. It is also very interesting to observe that the relaxation in the intermediate waiting times exhibit a sub-aging behavior (the characteristic relaxation time grows slower than the waiting time), which is often observed in the experimental data of thermoremanent magnetization (TRM) of real spin glass systems. Note that the crossover from the sub-aging to the simple aging occurs over a rather broad time region. We find that close to the dynamic transition, as shown in Fig. 3(a), the simple scaling is not yet fulfilled for waiting times which are long enough to reach simple aging at $`T=0.5`$. This suggests that the crossover time from sub-aging to simple aging becomes broader as the transition is approached.
Though the sub-aging eventually makes a crossover to the simple aging behavior, due to the broadness of the crossover regime (particularly near the transition), we wanted to know whether there is a scaling function $`h(t)`$, which can collapse the data in the intermediate times. One available empirical form is $`h(t)=\mathrm{exp}[\frac{1}{1\mu }(t/t_0)^{1\mu }]`$, which has been successfully used to collapse the TRM data in spin glasses . The time $`t_0`$ is the microscopic time scale ($`t_0=1`$ for our case). The case of $`\mu =0`$ corresponds to the absence of aging (no waiting time dependence), and the case of $`\mu =1`$ to the simple aging. Thus, the sub-aging behavior will give $`0<\mu <1`$. As shown in Fig. 3(b), with this empirical form of $`h(t)`$ with $`\mu 0.81`$ for $`T=0.61`$, we can collapse both the correlation and the integrated response with largest waiting times. We find that the exponent $`\mu `$ becomes larger in order to collapse the sub-aging data (with the same range of waiting times) at lower temperatures. We should emphasize that $`\mu `$ is an effective exponent, i.e., it tends to increase in order to incorporate the data with longer waiting times into scaling collapse due to the crossover to simple aging.
In summary, we have developed an efficient method of integrating the dynamic equations of a mean field spin glass model with $`p`$-spin interaction. This method allows us to see numerically for the first time the entire dynamic regime covering from microscopic to asymptotic regime. Tio cover such a broad dynamic range will also be of major importance in the sudy of aging in structural glasses . We observe several new dynamic features in the intermediate time regime including the existence of a new time scale at which the breakdown of FDT sets in, the broad sub-aging regime and its crossover dynamics to the simple aging. These findings offer a new theoretical challenge for further understanding of the off-equilibrium dynamics of the $`p`$-spin model.
###### Acknowledgements.
We thank K. Binder, L. Cugliandolo, H. Horner, K. Kawasaki, W. Kob, J. Kurchan, M. Mézard, R. Schilling and T. Theenhaus for valuable discussions and suggestions. B.K. thanks R. Schilling and K. Binder for hospitality during his sabbatical at Mainz. This work was supported by the SFB 262 and the MWFZ of the Johannes Gutenberg-University of Mainz.
|
warning/0005/math0005282.html
|
ar5iv
|
text
|
# On the Moduli Space of Classical Dynamical r-matrices
### Introduction.
A classical dynamical r-matrix is an $`𝔩`$-equivariant function $`r:𝔩^{}𝔤𝔤`$ (where $`𝔩`$, $`𝔤`$ are Lie algebras), such that $`r^{21}+r=\mathrm{\Omega }`$ is $`𝔤`$-invariant, which satisfies the classical dynamical Yang-Baxter equation (CDYBE). CDYBE is a differential equation, which generalizes the usual classical Yang-Baxter equation. It was introduced in 1994 by G.Felder \[Fe\], in the context of conformal field theory. Solutions of CDYBE and their quantizations appear naturally in several mathematical theories: the theory of integrable systems, special functions, representation theory (see \[ES\] for a review).
Since classical dynamical r-matrices were introduced, several authors tried to study and classify them (\[EV\],\[S\],\[Xu\]). The goal of this paper is to describe the moduli space of classical dynamical r-matrices modulo gauge transformations. In particular, we improve and generalize the results of \[EV\], \[S\], as well as correct some errors that occurred in these papers (See remarks 3 and 5).
The main achievement of this paper, compared to the previous ones, is that its results are valid for dynamical r-matrices for a nonabelian Lie algebra $`𝔩`$. It turns out that this generalization not only brings in new interesting examples (see \[EV\],\[AM\]) but also makes the general theory much more clear and natural.
The composition of the paper is as follows.
In Section 1, we recall the definition of a dynamical r-matrix.
In Section 2, we extend to the nonabelian case the notion of a gauge transformation of dynamical r-matrices, introduced in \[EV\].
In Section 3, we decribe the space of dynamical r-matrices modulo gauge transformations (the moduli space). Here we formulate our main theorem, stating that under some technical conditions, the moduli space can be identified with a certain explicitly given affine variety. For instance, if $`𝔩=𝔤`$, this variety consists of one point, which is the Alekseev-Meinrenken solution \[AM\] (for semisimple Lie algebras, it was also constructed in \[EV\]).
In Section 4 we prove the main theorem.
In the appendix, we construct a generalization of the Alekseev-Meinrenken classical dynamical r-matrix, associated to any finite-dimensional Lie algebra $`𝔤`$ with a nondegenerate invariant form and an automorphism $`B`$ of $`𝔤`$ of finite order which preserves this form.
## 1 The Dynamical Yang-Baxter Equation
###
Let $`𝔤`$ be a Lie algebra over $``$, and $`𝔩𝔤`$ a finite dimensional Lie subalgebra. Let $`x_1,\mathrm{},x_r`$ be a basis of $`𝔩`$.
Let $`D𝔩^{}`$ be the formal neighborhood of $`0`$. Let $`V`$ be a complex vector space. By functions from $`D`$ to $`V`$ we will mean elements of the space $`V[[x_1,\mathrm{},x_r]]`$, where we regard $`x_i`$ as coordinates on $`D`$. Finally, if $`\omega \mathrm{\Omega }^k(D,V)`$ is a $`k`$-form with values in any vector space $`V`$, we denote by $`\overline{\omega }:D\mathrm{\Lambda }^k𝔩V`$ the associated function. For an element $`r𝔤𝔤`$ we define the classical Yang-Baxter operator
$$CYB(r)=[r^{12},r^{13}]+[r^{12},r^{23}]+[r^{13},r^{23}].$$
The classical dynamical Yang-Baxter equation (CDYBE) is the following differential equation for an $`𝔩`$-equivariant function $`r:D𝔤𝔤`$ :
$$\mathrm{Alt}(\overline{dr})+CYB(r)=0,$$
(1.1)
where for $`x𝔤^3`$, we let $`\mathrm{Alt}(x)=x^{123}x^{213}+x^{312}`$.
It is useful to consider solutions of CDYBE which satisfy an additional quasi-unitarity condition:
$$r+r^{21}=\mathrm{\Omega }(S^2𝔤)^𝔤.$$
(1.2)
It is easy to show that if $`r`$ satisfies CDYBE and the quasi-unitarity condition then $`\mathrm{\Omega }`$ is a constant function of $`\lambda `$.
An $`𝔩`$-equivariant solution of CDYBE which satisfies the quasi-unitarity condition is called a dynamical r-matrix. The set of all dynamical r-matrices satisfying (1.2) will be denoted by $`Dynr(𝔤,𝔩,\mathrm{\Omega })`$.
## 2 Gauge transformations
###
Here we will reproduce some results from \[EV\], but unlike \[EV\], we will not assume that $`𝔩`$ is abelian. We will assume, however, that $`𝔤`$ is finite dimensional.
Let $`G`$ be the simply connected complex Lie group such that $`\mathrm{Lie}(G)=𝔤`$. Let $`g:DG`$ be any regular, $`𝔩`$-equivariant map. Consider the $`1`$-form $`\eta _g=g^1dg`$ and set $`\zeta _g=[\overline{\eta _g}^{12},\overline{\eta _g}^{13}]`$. Define an $`𝔩`$-equivariant function $`\tau _g:D\mathrm{\Lambda }^2𝔤`$ by the formula $`\tau _g(\lambda )=(\lambda 11)\zeta _g(\lambda )`$. For any $`𝔩`$-equivariant function $`r:D𝔤𝔤`$ we set
$$r^g=(gg)(r\overline{\eta _g}+\overline{\eta _g}^{21}+\tau _g)(g^1g^1).$$
(2.1)
The following theorem is a nonabelian generalization of Proposition 1.2 of \[EV\].
###### Proposition 2.1.
The function $`r`$ is a dynamical r-matrix if and only if the function $`r^g`$ is.
Proof. Let us show that if $`r`$ is a dynamical r-matrix then so is $`r^g`$. The other direction is analogous. Let $`X=(D\times G\times D,\{,\})`$ be the dynamical Poisson groupoid associated to $`r`$ in \[EV\]. Consider the automorphism $`\sigma `$ of $`X`$ given by $`\sigma (u_1,x,u_2)=(u_1,g(u_1)xg(u_2)^1,u_2)`$. Then $`\sigma `$ transforms $`\{,\}`$ into the Poisson bracket $`\{f,g\}_\sigma =\sigma ^1\{\sigma f,\sigma g\}`$. It is straightforward to calculate that the corresponding transformation at the level of dynamical r-matrices is exactly (2.1).$`\mathrm{}`$
###
The transformation $`rr^g`$ is called a gauge transformation. Note that (2.1) defines an action of the group $`\mathrm{Map}(D,G)^𝔩`$ on $`Dynr(𝔤,𝔩,\mathrm{\Omega })`$, i.e we have $`(r^{g_1})^{g_2}=r^{g_2g_1}`$ for any $`g_1,g_2\mathrm{Map}(D,G)^𝔩`$ and $`rDynr(𝔤,𝔩,\mathrm{\Omega })`$. Let us denote by $`\mathrm{Map}_0(D,G)^𝔩`$ the subgroup consisting of maps $`g`$ satisfying $`g(0)=1`$. We would like to understand the moduli space
$$(𝔤,𝔩,\mathrm{\Omega })=Dynr(𝔤,𝔩,\mathrm{\Omega })/\mathrm{Map}_0(D,G)^𝔩.$$
In the triangular case (i.e when $`\mathrm{\Omega }=0`$) this space was considered by P. Xu in \[Xu\].
### Remark 1.
It is clear that $`\mathrm{Map}(D,G)^𝔩/\mathrm{Map}_0(D,G)^𝔩G^𝔩`$. Hence the complete moduli space $`\overline{}(𝔤,𝔩,\mathrm{\Omega })=Dynr(𝔤,𝔩,\mathrm{\Omega })/\mathrm{Map}(D,G)^𝔩`$ is equal to $`(𝔤,𝔩,\mathrm{\Omega })/G^𝔩`$ where $`gG^𝔩`$ acts by $`r^g=\mathrm{Ad}(gg)(r)`$.
## 3 The structure of $`(𝔤,𝔩,\mathrm{\Omega })`$
From now on we will assume that
1. $`𝔩𝔤\mathrm{has}\mathrm{an}𝔩\mathrm{invariant}\mathrm{complement}𝔪.`$
The following theorem is a generalization of Theorem 1.4 in \[EV\]. It shows that the space of dynamical r-matrices is, up to gauge equivalence, finite dimensional.
###### Theorem 1.
Let $`\rho ,r:D𝔤^2`$ be two dynamical r-matrices such that $`r(0)=\rho (0)`$. Then there exists $`g\mathrm{Map}(D,G)^𝔩`$ such that $`\rho =r^g`$.
The proof is a generalization of the proof in \[EV\]. Before giving it we state the following auxiliary result.
###### Lemma 3.1 (equivariant Poincaré lemma).
Let $`𝔩`$ be a finite-dimensional Lie algebra, $`V`$ a finite-dimensional $`𝔩`$-module, $`k1`$ and $`\omega \mathrm{\Omega }^k(D,V)`$ an $`𝔩`$-equivariant closed $`k`$-form with values in $`V`$. Then there exists an $`𝔩`$-equivariant $`k1`$-form $`\zeta \mathrm{\Omega }^{k1}(D,V)`$ such that $`d\zeta =\omega `$.
Proof. The proof is the same as that for the usual Poincaré lemma. It is enough to assume that $`\omega `$ is homogeneous, of degree $`l`$. Let $`E=_ix_i\frac{}{x_i}`$ be the Euler vector field on $`D`$. Then by Cartan’s homotopy formula,
$$l\omega =L_E\omega =i_Ed\omega +di_E\omega =d(i_E\omega )$$
and we can set $`\zeta =i_E\omega /l`$. Note that $`E`$ is $`𝔩`$-equivariant, hence so is $`\zeta `$. $`\mathrm{}`$
Proof of Theorem 1. The dynamical r-matrices $`r,\rho `$ are by definition formal power series in the variables $`x_i`$. Let us assume that the statement of the theorem holds modulo terms of degree $`K`$. Let $`g_k:UG`$ be a gauge transformation such that $`E:=r^{g_k}\rho `$ has degree $`K`$ and let $`E_K`$ be the homogeneous component of $`E`$ of degree $`K`$. Then $`\mathrm{Alt}(\overline{dE_K})=\left[CYB(r^{g_K})CYB(\rho )\right]_{K1}`$ where $`[]_{K1}`$ denotes the homogeneous component of degree $`K1`$. But $`\left[r^{g_K}\rho \right]_l=0`$ for all $`l<K`$ by assumption, hence
$$\mathrm{Alt}(\overline{dE_K})=0.$$
(3.1)
###### Lemma 3.2.
The exists an $`𝔩`$-equivariant closed 1-form $`\zeta \mathrm{\Omega }^1(D,𝔤)`$ such that $`E_K=\overline{\zeta }^{21}\overline{\zeta }`$.
Proof. Let us write $`E_K=E_{𝔩𝔩}+E_{𝔩𝔪}E_{𝔩𝔪}^{21}+E_{𝔪𝔪}`$ where $`E_{𝔩𝔩}\mathrm{\Lambda }^2𝔩`$, $`E_{𝔩𝔪}𝔩𝔪`$ and $`E_{𝔪𝔪}\mathrm{\Lambda }^2𝔪`$. From (3.1) it follows that $`dE_{𝔪𝔪}=0`$ hence $`E_{𝔪𝔪}=0`$. Now let $`\xi \mathrm{\Omega }^1(D,𝔪)`$ be such that $`\overline{\xi }=E_{𝔩𝔪}`$. Then (3.1) implies that $`\xi `$ is closed. Note that the assumption i) guarantees that $`\xi `$ is equivariant. Finally, let $`\omega \mathrm{\Omega }^2(D,)`$ be such that $`\overline{\omega }=E_{𝔩𝔩}`$. Then (3.1) says that $`\omega `$ is closed. By the equivariant Poincaré lemma, there exists an equivariant 1-form $`\eta `$ such that $`d\eta =\omega `$. Set $`\theta =d\overline{\eta }`$, so that $`\overline{\theta }\overline{\theta }^{21}=\overline{\omega }`$. Then $`\zeta =\xi +\theta `$ satisfies the conditions of the lemma.$`\mathrm{}`$
###
We now conclude the proof of Theorem 1. Let $`\chi :D𝔤`$ be any $`𝔩`$-equivariant function of order $`K+1`$ such that $`d\chi =\zeta `$. Set $`g=e^\chi `$. Then $`\eta _g=g^1dg`$ is of order $`K`$ and $`\overline{\zeta }\overline{\eta _g}`$ is of order $`K+1`$. But then $`\overline{\zeta }^{21}\overline{\zeta }(\overline{\eta _g}^{21}\overline{\eta _g}+\tau _g)`$ is also of order $`K+1`$. Set $`g_{K+1}=gg_K`$. Then, by the above $`r^{g_{K+1}}\rho `$ is of degree $`K+1`$. The proof follows by induction.$`\mathrm{}`$
###### Proposition 3.1.
Any dynamical r-matrix $`r`$ is gauge-equivalent to a dynamical r-matrix $`\rho `$ such that $`\rho (0)\frac{\mathrm{\Omega }}{2}+\mathrm{\Lambda }^2𝔪`$.
Proof. Let $`\overline{\eta }_0𝔩𝔤`$ such that $`r(0)\overline{\eta }_0+\overline{\eta }_0^{21}\frac{\mathrm{\Omega }}{2}+\mathrm{\Lambda }^2𝔪`$. Since $`𝔪`$ is $`𝔩`$-invariant, we have $`\overline{\eta }_0(𝔩𝔤)^𝔩`$. By the equivariant Poincaré lemma, there exists an equivariant function $`\chi :D𝔤`$ satisfying $`\chi (0)=0`$, $`d\chi =\eta _0`$. Set $`g=e^\chi `$. Then $`\rho :=r^g`$ satisfies the CDYBE and $`\rho (0)\frac{\mathrm{\Omega }}{2}+\mathrm{\Lambda }^2𝔪`$. $`\mathrm{}`$
###
Consider the following algebraic variety
$$_\mathrm{\Omega }=\{x\frac{\mathrm{\Omega }}{2}+(\mathrm{\Lambda }^2𝔪)^𝔩|CYB(x)=0in\mathrm{\Lambda }^3(𝔤/𝔩)\}.$$
It is immediate from (2.1) that if $`\rho `$ and $`r`$ are gauge-equivalent and if $`\rho (0)\frac{\mathrm{\Omega }}{2}+\mathrm{\Lambda }^2𝔪`$ and $`r(0)\frac{\mathrm{\Omega }}{2}+\mathrm{\Lambda }^2𝔪`$ then $`r(0)=\rho (0)`$. Moreover, it follows from the CDYBE (1.1) that for every dynamical r-matrix $`rDynr(𝔤,𝔩,\mathrm{\Omega })`$ such that $`r(0)\frac{\mathrm{\Omega }}{2}+\mathrm{\Lambda }^2𝔪`$ we have $`r(0)_\mathrm{\Omega }`$.
Hence Theorem 1 and Proposition 3.1 give the following corollary.
###### Corollary 3.1.
The map $`(𝔤,𝔩,\mathrm{\Omega })_\mathrm{\Omega }`$ which sends a class $`𝒞`$ to $`r(0)`$ where $`r𝒞`$ is any representative such that $`r(0)\frac{\mathrm{\Omega }}{2}+\mathrm{\Lambda }^2𝔪`$, is an embedding.
### Remark 2.
If condition i) fails then the space $`(𝔤,𝔩,\mathrm{\Omega })`$ may be infinite-dimensional. This is demonstrated by the following example due to P. Xu \[Xu\]. Let $`𝔤=xy`$ be the two-dimensional Lie algebra with $`[x,y]=y`$, and set $`𝔩=y`$. Then $`\mathrm{\Lambda }^3𝔤=0`$ and $`\mathrm{\Lambda }^2𝔤`$ is a trivial $`𝔩`$-module. Thus any function $`r:D\mathrm{\Lambda }^2𝔤`$ is a dynamical r-matrix. On the other hand, $`𝔤^𝔩=𝔩`$ and all gauge transformations act trivially.
### Remark 3.
We would like to use this opportunity to correct the statement of Theorem 1.4 of \[EV\]. This theorem is incorrect as stated (as shown by Xu’s counterexample, see Remark 1). The mistake is in the proof of Lemma 1.5, which uses the incorrect statement that
$$(𝔤𝔩𝔩𝔤)^𝔩=(𝔤^𝔩𝔩𝔩𝔤^𝔩)^𝔩$$
(3.2)
for commutative $`𝔩`$. This statement, however, is correct with the additional assumption i); in this case Theorem 1.4 of \[EV\] and its proof are correct, and Theorem 1.4 of \[EV\] is a special case of Theorem 1 above.
###
Now suppose that $`𝔩=𝔤`$. Note that i) automatically holds in this case. Then by Proposition 3.1 and Theorem 1 there is at most one gauge-equivalence class of dynamical r-matrices $`r:D\frac{\mathrm{\Omega }}{2}+\mathrm{\Lambda }^2𝔤`$. Such a class in fact always exists, as was discovered by Alekseev and Meinrenken \[AM\]. A representative of this class is constructed as follows.
###
Let $`𝔤_\mathrm{\Omega }`$ be the ideal of $`𝔤`$ spanned by the components of $`\mathrm{\Omega }`$, and let $`D_\mathrm{\Omega }`$ be the formal neighborhood of $`0`$ in $`𝔤_\mathrm{\Omega }^{}`$. Let us identify $`𝔤_\mathrm{\Omega }`$ with $`𝔤_\mathrm{\Omega }^{}`$ via $`\mathrm{\Omega }`$. Set $`f(s)=\frac{1}{s}\frac{1}{2}\mathrm{cotanh}(\frac{s}{2})`$. Then $`f`$ is smooth at the origin. Consider the following map
$`T:D_\mathrm{\Omega }`$ $`\mathrm{End}(𝔤)𝔤^{}𝔤𝔤𝔤`$
$`u`$ $`f(\mathrm{ad}\mu )`$
Let $`\pi ^{}:𝔤^{}𝔤_\mathrm{\Omega }^{}`$ be the projection and set
$$r_{AM}^𝔤=\frac{\mathrm{\Omega }}{2}+T\pi ^{}:D𝔤𝔤.$$
###### Theorem 2 (\[AM\]).
The map $`r_{AM}^𝔤`$ is a dynamical r-matrix.
This theorem is proved in \[AM\] in the case of compact Lie algebras, but the proof can be adapted to the general case. Another proof is given in the appendix.
###### Corollary 3.2.
The moduli space $`(𝔤,𝔤,\mathrm{\Omega })`$ consists of the single class $`r_{AM}^𝔤`$.
### Remark 4.
When $`𝔤`$ is a simple Lie algebra and $`𝔩=𝔤`$ these results easily follow from \[EV\], Section 3.8.
###
We will now show that, under some technical conditions on $`\mathrm{\Omega }`$, the embedding defined in Corollary 3.1 is actually an isomorphism. From now on we assume that
1. We have $`\mathrm{\Omega }(𝔩𝔩)(𝔪𝔪)`$.
We will write $`\mathrm{\Omega }_𝔩`$ (resp. $`\mathrm{\Omega }_𝔪`$) for the component of $`\mathrm{\Omega }`$.
Condition ii) is satisfied in particular in the triangular case ($`\mathrm{\Omega }=0`$). It is also satisfied when $`𝔩`$ acts semisimply on $`𝔤`$ and when the restriction of the inverse form $`(,)=\mathrm{\Omega }^1`$ to $`𝔩_\mathrm{\Omega }=𝔩𝔤_\mathrm{\Omega }`$ is nondegenerate. Indeed, let $`𝔤^{}`$ be an $`𝔩`$-invariant complement of $`𝔩+𝔤_\mathrm{\Omega }`$ in $`𝔤`$ and let $`𝔪_\mathrm{\Omega }`$ be the orthogonal complement of $`𝔩_\mathrm{\Omega }`$ i $`𝔤_\mathrm{\Omega }`$. Then $`𝔪=𝔤^{}𝔪_\mathrm{\Omega }`$ satisfies conditions i) and ii).
###### Proposition 3.2.
Any dynamical r-matrix $`r`$ is gauge-equivalent to a dynamical r-matrix of the form $`\rho =r_{AM}^𝔩+\frac{\mathrm{\Omega }_𝔪}{2}+t`$ with $`t:D\mathrm{\Lambda }^2𝔪`$.
Proof. By Proposition 3.1 there exists a dynamical r-matrix $`\rho _0`$ gauge-equivalent to $`r`$ such that $`\rho _0(0)\frac{\mathrm{\Omega }}{2}+\mathrm{\Lambda }^2𝔪`$. We will first construct a sequence of gauge transformations $`g_i,i=1,\mathrm{}`$ such that $`\rho _0^{g_i}\frac{\mathrm{\Omega }}{2}+\left(\mathrm{\Lambda }^2𝔩\mathrm{\Lambda }^2𝔪\right)`$ modulo terms of degree $`i`$. We set $`g_1=1`$. Suppose that we have constructed $`g_i`$ and let $`E_i`$ be the term of degree exactly $`i`$ of $`\rho _0^{g_i}`$. From the CDYBE we have
$$\mathrm{Alt}(\overline{dE_i})=\left[CYB(\rho _0^{g_i})\right]_{i1}$$
(3.3)
where $`[]_{i1}`$ denotes the component of degree $`i1`$. But by our assumption we have $`\rho _0^{g_i}\frac{\mathrm{\Omega }}{2}+\left(\mathrm{\Lambda }^2𝔩\mathrm{\Lambda }^2𝔪\right)`$ in degrees $`i1`$. Using the $`𝔩`$-invariance of $`𝔪`$ it is easy to see that this implies that
$$\left[CYB(\rho _0^{g_i})\right]_{i1}\mathrm{Alt}\left((𝔩𝔪𝔪)(𝔩𝔩𝔩)(𝔪𝔪𝔪)\right).$$
(3.4)
Let $`\xi \mathrm{\Omega }^1(D,𝔪)`$ such that $`E_i+\overline{\xi }^{21}\overline{\xi }\mathrm{\Lambda }^2𝔩\mathrm{\Lambda }^2𝔪`$. Then from (3.3) and (3.4) it follows that $`d\xi =0`$. By the equivariant Poincaré lemma there exists an equivariant map $`\chi :D𝔪`$ such that $`\xi =d\chi `$. Moreover, $`\xi `$ is of degree $`i`$, hence $`\chi `$ is of degree $`i+1`$. Now set $`g=e^\chi `$. Then $`\eta _g\xi `$ is of order $`i+1`$. Thus
$$(gg)\left(\rho _0^{g_i}+\overline{\eta _g}^{21}\overline{\eta _g}+\tau _g\right)(g^1g^1)$$
is in $`\frac{\mathrm{\Omega }}{2}+\left(\mathrm{\Lambda }^2𝔩\mathrm{\Lambda }^2𝔪\right)`$ modulo terms of degree $`i+1`$, and we put $`g_{i+1}=g_ig`$. This allows to define the sequence $`g_i`$ inductively.
It is clear that the sequence $`\rho _0^{g_i}`$ converges, in the sense of formal power series, to a dynamical r-matrix $`\rho _1`$ which is gauge-equivalent to $`\rho _0`$. Moreover $`\rho _1`$ takes values in $`\frac{\mathrm{\Omega }}{2}+\left(\mathrm{\Lambda }^2𝔩\mathrm{\Lambda }^2𝔪\right)`$ by construction. Let us write $`\rho _1=\rho _1^𝔩+\rho _1^𝔪`$ where $`\rho _1^𝔩`$ and $`\rho _1^𝔪`$ take values in $`𝔩𝔩`$ and $`𝔪𝔪`$ respectively. Observe that $`\rho _1^𝔩:D\frac{\mathrm{\Omega }_𝔩}{2}+\mathrm{\Lambda }^2𝔩`$ is itself a dynamical r-matrix. Hence by Corollary 3.2 we can perform a gauge-transformation for $`𝔩`$ to reduce it to $`r_{AM}^𝔩`$.$`\mathrm{}`$
###
The following theorem is a generalization to the nonabelian case of \[S\], Theorem 3, and will be proved in the next section.
###### Theorem 3.
Let $`r_0_\mathrm{\Omega }`$. Then there exists a unique dynamical r-matrix $`r=r_{AM}^𝔩+\frac{\mathrm{\Omega }_𝔪}{2}+t`$ with $`t:D\mathrm{\Lambda }^2𝔪`$, such that $`r(0)=r_0`$.
###### Corollary 3.3.
Under conditions i) and ii) the moduli space $`(𝔤,𝔩,\mathrm{\Omega })`$ of gauge-equivalence classes of dynamical r-matrices is isomorphic to $`_\mathrm{\Omega }`$.
### Remark 5.
We use this opportunity to correct the statement of Theorem 3 in \[S\] which is false as stated. The mistake is in the proof of Lemma 1, which uses the incorrect statement (3.2). However, the theorem and its proof are correct if one makes in addition the assumption i). In this case it is a special case of Theorem 3 above. Moreover the genericity assumption made in \[S\] Theorem 3 is not necessary, as the flow constructed in \[S\] Lemma 2 is well-defined on the whole $`(\mathrm{\Lambda }^2𝔪)^𝔩`$.
### Remark 6.
Let us identify $`𝔪`$ with $`𝔤/𝔩`$ via the decomposition $`𝔤=𝔩𝔪`$. This allows to define an action of $`G^𝔩`$ on $`𝔪`$, hence also an action of $`G^𝔩`$ on $`_\mathrm{\Omega }`$. It is clear from (2.1) that the isomorphism $`(𝔤,𝔩,\mathrm{\Omega })_\mathrm{\Omega }`$ is $`G^𝔩`$-equivariant. In particular, $`\overline{}(𝔤,𝔩,\mathrm{\Omega })_\mathrm{\Omega }/G^𝔩`$.
## 4 Proof of Theorem 3
Proof of Theorem 3. We will construct by induction a formal power series $`t=_kt_k`$ with $`t_k:D\mathrm{\Lambda }^2𝔪`$ of degree $`k`$, such that $`r=r_{AM}^𝔩+\frac{\mathrm{\Omega }_𝔪}{2}+t`$ is a dynamical r-matrix satisfying $`r(0)=r_0`$. Set $`t_0=r_0\frac{\mathrm{\Omega }}{2}\mathrm{\Lambda }^2𝔪`$ and let us suppose that we have defined an $`𝔩`$-equivariant polynomial $`t_{<k}=_{l<k}t_l`$. Set $`s=r_{AM}^𝔩\frac{\mathrm{\Omega }_𝔩}{2}`$, $`Z_\mathrm{\Omega }=CYB(\mathrm{\Omega })`$ and $`Z_{\mathrm{\Omega }_𝔩}=CYB(\mathrm{\Omega }_𝔩)`$. Then the CDYBE for $`r_{AM}^𝔩`$ is equivalent to the following equation for $`s`$ :
$$\mathrm{Alt}(\overline{ds})+CYB(s)+\frac{1}{4}Z_{\mathrm{\Omega }_𝔩}=0.$$
(4.1)
Let $`\pi :𝔤𝔩`$ be the projection along $`𝔪`$. Consider, for $`lk`$ the following system of differential equations for $`i=1,\mathrm{}r`$.
$$\frac{t_l}{x_i^{}}=(x_i^{}11)\left[[t_{<l}^{12},t_{<l}^{13}]+[s^{12}+s^{13},t_{<l}]+\frac{1}{4}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})\right]_{l1}$$
($`E_l`$)
where by definition $`x^{}(y)=x^{}(\pi (y))`$ for all $`x^{}𝔩^{}`$, $`y𝔤`$.
###### Lemma 4.1.
Suppose that ($`E_l`$) is satisfied for all $`l<k`$. Then ($`E_k`$) admits a unique solution $`t_k`$ of degree $`k`$, which is $`𝔩`$-equivariant.
Proof. By the equivariant Poincaré lemma, it is enough to show that
$$\begin{array}{cc}\hfill \frac{}{x_j^{}}& (x_i^{}11)\left\{[t_{<k}^{12},t_{<k}^{13}]+[s^{12}+s^{13},t_{<k}]+\frac{1}{4}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})\right\}\hfill \\ & =\frac{}{x_i^{}}(x_j^{}11)\left\{[t_{<k}^{12},t_{<k}^{13}]+[s^{12}+s^{13},t_{<k}]+\frac{1}{4}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})\right\}\hfill \end{array}$$
(4.2)
Let us write $`_i`$ for $`\frac{}{x_i^{}}`$ and $`t`$ for $`t_{<k}`$. All equations below will be understood modulo terms of degree $`k`$. Let $`X_i`$ and $`X_j`$ denote the r.h.s and l.h.s of (4.2). Using the assumption that $`t`$ is a solution of the system ($`E_l`$) for all $`l<k`$, we have
$$\begin{array}{cc}\hfill X_i& X_j=\hfill \\ & =(x_i^{}x_j^{}11)\{[[t^{12},t^{13}]+[s^{12}+s^{13},t^{23}]+\frac{1}{4}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})^{123},t^{24}]\hfill \\ & +[t^{23},[t^{12},t^{14}]+[s^{12}+s^{14},t^{24}]+\frac{1}{4}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})^{124}]\hfill \\ & +[s^{23}+s^{24},[t^{13},t^{14}]+[s^{13}+s^{14},t^{34}]+\frac{1}{4}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})^{134}]\hfill \\ & [_i(s^{23}+s^{24}),t^{34}]+[_j(s^{13}+s^{14}),t^{34}]\hfill \\ & [[t^{12},t^{23}]+[s^{12}+s^{23},t^{13}]\frac{1}{4}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})^{123},t^{14}]\hfill \\ & [t^{13},[t^{12},t^{24}]+[s^{12}+s^{24},t^{14}]\frac{1}{4}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})^{124}]\hfill \\ & [s^{13}+s^{14},[t^{23},t^{24}]+[s^{23}+s^{24},t^{34}]+\frac{1}{4}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})^{234}]\}.\hfill \end{array}$$
(4.3)
By the Jacobi identity we have
$$[[t^{12},t^{13}],t^{24}]+[t^{13},[t^{12},t^{24}]]=[[t^{12},t^{23}],t^{14}]+[t^{23},[t^{12},t^{14}]]=0.$$
(4.4)
Moreover,
$$\begin{array}{cc}\hfill (x_i^{}x_j^{}11)\{& [\frac{1}{4}(Z_{\mathrm{\Omega }_𝔩})^{123},t^{24}]+[t^{23},\frac{1}{4}(Z_{\mathrm{\Omega }_𝔩})^{124}]\hfill \\ & +[\frac{1}{4}(Z_{\mathrm{\Omega }_𝔩})^{123},t^{14}]+[t^{13},\frac{1}{4}(Z_{\mathrm{\Omega }_𝔩})^{124}]\}=0\hfill \end{array}$$
(4.5)
since $`Z_{\mathrm{\Omega }_𝔩}\mathrm{\Lambda }^3𝔩`$, $`t\mathrm{\Lambda }^2𝔪`$ and $`𝔪`$ is $`𝔩`$-invariant. Furthermore, $`Z_\mathrm{\Omega }`$ is
$`𝔤`$-invariant, hence
$`(x_i^{}x_j^{}11)\{`$ $`[{\displaystyle \frac{1}{4}}(Z_\mathrm{\Omega })^{123},t^{24}]+[{\displaystyle \frac{1}{4}}(Z_\mathrm{\Omega })^{123},t^{14}]`$
$`+[t^{23},{\displaystyle \frac{1}{4}}(Z_\mathrm{\Omega })^{124}]+[t^{13},{\displaystyle \frac{1}{4}}(Z_\mathrm{\Omega })^{124}]\}`$
$`=(x_i^{}x_j^{}1`$ $`1)\{[{\displaystyle \frac{1}{4}}(Z_\mathrm{\Omega })^{123},t^{34}]+[t^{34},{\displaystyle \frac{1}{4}}(Z_\mathrm{\Omega })^{124}]\}`$
$`=(x_i^{}x_j^{}1`$ $`1)\{[{\displaystyle \frac{1}{4}}((Z_{\mathrm{\Omega }_𝔩})^{123}+(Z_{\mathrm{\Omega }_𝔩})^{124}),t^{34}]\}.`$ (4.6)
In a similar way, $`Z_\mathrm{\Omega }`$ and $`Z_{\mathrm{\Omega }_𝔩}`$ are $`𝔩`$-invariant, hence
$`[s^{23}+s^{24},{\displaystyle \frac{1}{4}}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})^{134}]`$ $`=[s^{12},{\displaystyle \frac{1}{4}}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})^{134}],`$
$`[s^{13}+s^{14},{\displaystyle \frac{1}{4}}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})^{234}]`$ $`=[s^{12},{\displaystyle \frac{1}{4}}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})^{234}].`$ (4.7)
From the Jacobi identity again we deduce
$`[[s^{12},t^{23}],t^{24}]+[t^{23},[s^{12},t^{24}]]`$ $`=[s^{12},[t^{23},t^{24}]],`$
$`[[s^{12},t^{13}],t^{14}]+[t^{13},[s^{12},t^{14}]]`$ $`=[s^{12},[t^{13},t^{14}]],`$
$`[[s^{13},t^{23}],t^{24}][s^{13},[t^{23},t^{24}]]`$ $`=0,`$
$`[s^{24},[t^{13},t^{14}]][t^{13},[s^{24},t^{14}]]`$ $`=0,`$
$`[t^{23},[s^{14},t^{24}]][s^{14},[t^{23},t^{24}]]`$ $`=0,`$
$`[s^{23},[t^{13},t^{14}]][[s^{23},t^{13}],t^{14}]`$ $`=0.`$ (4.8)
and
$$\begin{array}{cc}\hfill [s^{23}+s^{24},[s^{13}+s^{14},t^{34}]][s^{13}& +s^{14},[s^{23}+s^{24},t^{34}]]\hfill \\ & =[[s^{23},s^{13}],t^{34}]+[[s^{24},s^{14}],t^{34}]\hfill \\ & =[[s^{13},s^{23}],t^{34}][[s^{14},s^{24}],t^{34}].\hfill \end{array}$$
(4.9)
Collecting terms from (4.4),(4.5),(4),(4), (4),(4.9) and replacing in (4.3), we obtain
$$\begin{array}{cc}\hfill X_i& X_j=\hfill \\ & =(x_i^{}x_j^{}11)\{[\frac{1}{4}((Z_{\mathrm{\Omega }_𝔩})^{123}+(Z_{\mathrm{\Omega }_𝔩})^{124}),t^{34}]\hfill \\ & [[s^{13},s^{23}],t^{34}][[s^{14},s^{24}],t^{34}][_i(s^{23}+s^{24}),t^{34}]\hfill \\ & +[_j(s^{13}+s^{14}),t^{34}]+[s^{12},[t^{23},t^{24}]+\frac{1}{4}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})^{234}]\hfill \\ & +[s^{12},[t^{13},t^{14}]+\frac{1}{4}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})^{134}]\}.\hfill \end{array}$$
(4.10)
Using the fact that $`t`$ is a solution to the system ($`E_l`$) again we have
$$[s^{12},[t^{23},t^{24}]+\frac{1}{4}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})^{234}]+[s^{12},[t^{13},t^{14}]+\frac{1}{4}(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩})^{134}]$$
$$\begin{array}{cc}& =[s^{12},\underset{k}{}(x_k1+1x_k)(_kt[s_k1+1s_k,t])]\hfill \\ & =[s^{12},\underset{k}{}(x_k1+1x_k)_kt][s^{12},[s^{23}+s^{24},t^{34}]]\hfill \\ & [s^{12},[s^{13}+s^{14},t^{34}]]\hfill \end{array}$$
(4.11)
where we set $`s_k=(x_k^{}1)s`$. But $`s`$ is $`𝔩`$-equivariant, i.e for $`y𝔩`$ we have
$$[s,y1+1y]=\underset{l}{}[x_l,y]_ls$$
where $`[x_l,y]`$ is considered as a function $`D`$. Thus,
$$\begin{array}{cc}\hfill [s^{12},\underset{k}{}(x_k1+1x_k)_kt]& =\underset{l,k}{}_ls^{12}_kt^{34}[x_l,x_k]\hfill \\ & =\underset{l}{}_ls^{12}[x_l^3+x_l^4,t^{34}].\hfill \end{array}$$
(4.12)
Using Jacobi identity, we can write
$`[s^{12},[s^{23}+s^{24},t^{34}]]`$ $`=[[s^{12},s^{23}],t^{34}]+[[s^{12},s^{24}],t^{34}],`$
$`[s^{12},[s^{23}+s^{24},t^{34}]]`$ $`=[[s^{12},s^{13}],t^{34}]+[[s^{12},s^{14}],t^{34}].`$ (4.13)
Using (4.10), (4.11), (4.12) and (4) we finally get, by (4.1)
$$\begin{array}{cc}\hfill X_i& X_j=(x_i^{}x_j^{}11)\hfill \\ & \{[\mathrm{Alt}(\overline{ds})^{123}+[s^{13},s^{23}]+[s^{12},s^{13}]+[s^{12},s^{23}]+\frac{1}{4}(Z_{\mathrm{\Omega }_𝔩})^{123},t^{34}]\hfill \\ & +[\mathrm{Alt}(\overline{ds})^{124}+[s^{14},s^{24}]+[s^{12},s^{14}]+[s^{12},s^{24}]+\frac{1}{4}(Z_{\mathrm{\Omega }_𝔩})^{124},t^{34}]\}\hfill \\ & =0\hfill \end{array}$$
$`\mathrm{}`$
###
Let $`t=t_i:D\mathrm{\Lambda }^2𝔪`$ be the $`𝔩`$-equivariant series constructed by applying Lemma 4.2 succesively, starting from $`t_0`$.
###
Consider the algebraic variety
$$𝒯_\mathrm{\Omega }=\{t\mathrm{\Lambda }^2𝔪|CYB(t+\frac{\mathrm{\Omega }}{2})=0in\mathrm{\Lambda }^3(𝔤/𝔩)\}.$$
###
Let $`x^{}𝔩^{}`$ and consider the flow on $`\mathrm{\Lambda }^2𝔪`$ defined by the equation
$$\frac{u}{ϵ}=(x^{}11)\left([u^{12},u^{13}]+[s^{12}+s^{13},u^{23}]+\frac{1}{4}\left(CYB(\mathrm{\Omega })CYB(\mathrm{\Omega }_𝔩)\right)\right).$$
(4.14)
###### Lemma 4.2.
The flow (4.14) preserves $`𝒯_\mathrm{\Omega }`$.
Proof. Let $`u𝒯_\mathrm{\Omega }`$. Set $`h_1=(x^{}11)\left([s^{12}+s^{13},u^{23}]\right)`$,
$$h_2=(x^{}11)\left([u^{12},u^{13}]+\frac{1}{4}\left(Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩}\right)\right).$$
Note that $`h_1\mathrm{\Lambda }^2𝔪`$ by condition i) and that $`h_2\mathrm{\Lambda }^2𝔪`$ since $`u\mathrm{\Lambda }^2𝔪`$ and since by ii),
$$Z_\mathrm{\Omega }Z_{\mathrm{\Omega }_𝔩}(𝔪𝔤𝔤)(𝔩𝔪𝔪).$$
It thus remains to check that the vector field defined by (4.14) is tangent to $`𝒯_\mathrm{\Omega }`$, i.e that $`CYB(u,h_1+h_2)\mathrm{Alt}(𝔩𝔤𝔤)`$, where we use the notation
$$CYB(a,b)=[a^{12},b^{13}]+[a^{13},b^{23}]+[a^{12},b^{23}]+[b^{12},a^{13}]+[b^{13},a^{23}]+[b^{12},a^{23}].$$
But
$$CYB(u,h_1)=\mathrm{ad}\left((x^{}1)s\right)CYB(u)\mathrm{Alt}(𝔩𝔤𝔤),$$
and $`CYB(u,h_2)\mathrm{Alt}(𝔩𝔤𝔤)`$ by \[S\], Lemma 3 (note that the commutativity of $`𝔩`$, assumed in \[S\], is not used in the proof of Lemma 3).$`\mathrm{}`$
###### Corollary 4.1.
The map $`t:D\mathrm{\Lambda }^2𝔪`$ takes values in $`𝒯_\mathrm{\Omega }`$.
Proof. Note that $`t(0)𝒯_\mathrm{\Omega }`$ by assumption, and that for any $`x^{}D`$ the function $`u(ϵ)=t(ϵx^{})`$ on the formal disc satisfies (4.14) by construction. Hence $`t`$ takes values in $`𝒯_\mathrm{\Omega }`$. $`\mathrm{}`$
###
We now conclude the proof of Theorem 3 by showing that $`r=r_{AM}^𝔩+\frac{\mathrm{\Omega }_𝔪}{2}+t`$ is a dynamical r-matrix. Setting $`s_{AM}^𝔩=r_{AM}^𝔩\frac{\mathrm{\Omega }_𝔩}{2}`$ we have
$$\begin{array}{cc}\hfill \mathrm{Alt}& (\overline{dr})+CYB(r)\hfill \\ & =\mathrm{Alt}(\overline{ds_{AM}^𝔩})+\mathrm{Alt}(\overline{dt})+CYB(s_{AM}^𝔩+t)+\frac{1}{4}CYB(\mathrm{\Omega })\hfill \\ & =\mathrm{Alt}(\overline{ds_{AM}^𝔩})+\mathrm{Alt}(\overline{dt})+CYB(s_{AM}^𝔩)+CYB(t)+CYB(s_{AM}^𝔩,t)\hfill \\ & +\frac{1}{4}CYB(\mathrm{\Omega }).\hfill \end{array}$$
Using the CDYBE for $`r_{AM}^𝔩`$ we see that $`r`$ is a dynamical r-matrix if and only if
$$\mathrm{Alt}(\overline{dt})+CYB(t)+CYB(s_{AM}^𝔩,t)+\frac{1}{4}\left(CYB(\mathrm{\Omega })CYB(\mathrm{\Omega }_𝔩)\right)=0.$$
(4.15)
Since $`t`$ takes values in $`𝒯_\mathrm{\Omega }`$, (4.15) is equivalent to the system
$$\frac{t}{x_i^{}}=(x_i^{}11)\left(CYB(t)+CYB(s_{AM}^𝔩,t)+\frac{1}{4}\left(CYB(\mathrm{\Omega })CYB(\mathrm{\Omega }_𝔩)\right)\right)$$
for $`i=1,\mathrm{}r`$. It is easy to see from conditions i) and ii) that this last system is itself equivalent to the collection of systems ($`E_l`$) for all $`l`$.$`\mathrm{}`$
## 5 Appendix. Generalized Alekseev-Meinrenken <br>dynamical r-matrices
###
In this appendix we give a generalization of the dynamical r-matrix $`r_{AM}^𝔩`$.
###
Let $`𝔤`$ be a finite-dimensional complex Lie algebra and $`B:𝔤𝔤`$ an automorphism of order $`n`$. Then $`𝔤=_{j/n}𝔤_j`$ where $`𝔤_j=\mathrm{Ker}(Be^{\frac{2i\pi j}{n}})`$. Set $`𝔩=𝔤_0`$. Then $`𝔤_0`$ acts on $`𝔤_j`$ for all $`j`$.
Assume that $`𝔤`$ carries a nondegenerate invariant form $`(,)`$, which is stable under $`B`$. Set $`\mathrm{\Omega }=(,)^1(S^2𝔤)^𝔤`$. We will identify $`𝔤`$ with $`𝔤^{}`$ and $`𝔩`$ with $`𝔩^{}`$ using $`(,)`$.
###
Let $`D`$ be the formal neighborhood of zero in $`𝔩^{}𝔩`$. Consider the function $`\widehat{\rho }:D\mathrm{End}(𝔤)`$ such that $`\widehat{\rho }(A)_{|𝔤_i}=f_i(\mathrm{ad}A),`$ with
$`f_0(s)`$ $`={\displaystyle \frac{1}{s}}{\displaystyle \frac{1}{2}}\mathrm{cotanh}({\displaystyle \frac{1}{2}}s),`$
$`f_j(s)`$ $`={\displaystyle \frac{1}{2}}\mathrm{cotanh}({\displaystyle \frac{1}{2}}(s+{\displaystyle \frac{2i\pi j}{n}})),j0.`$
The element $`\widehat{\rho }`$ defines a map $`\rho :D\mathrm{\Lambda }^2𝔤`$. Let us set $`r_B=\frac{\mathrm{\Omega }}{2}+\rho `$.
###### Theorem A 1.
The map $`r_B`$ is a dynamical r-matrix.
### Remark 7.
If $`B=1`$ then $`r_B`$ is equal to the Alekseev-Meinrenken dynamical r-matrix $`r_{AM}^𝔩`$.
###
The rest of this appendix is devoted to the proof of Theorem A.1. We start by recalling the following result from \[EV\]. Let $`𝔩`$ be a reductive Lie algebra with Cartan subalgebra $`𝔥`$, $`𝔤`$ any finite-dimensional Lie algebra containing $`𝔩`$ and let $`\mathrm{\Omega }(S^2𝔤)^𝔤`$. The projection $`𝔩𝔥`$ defines an embedding $`𝔥^{}𝔩^{}`$. Let $`\mathrm{\Delta }`$ be the root system of $`𝔩`$ and $`𝔩_\alpha `$ the weight subspace corresponding to $`\alpha \mathrm{\Delta }`$. Choose an nondegenerate invariant inner product on $`𝔩`$. Let us fix $`e_\alpha 𝔩_\alpha `$ for all $`\alpha \mathrm{\Delta }`$ such that $`(e_\alpha ,e_\alpha )=1`$. Define a function $`\rho _0:D\mathrm{\Lambda }^2𝔩\mathrm{\Lambda }^2𝔤`$ by
$$\rho _0(\lambda )=\underset{\alpha >0}{}\frac{e_\alpha e_\alpha e_\alpha e_\alpha }{(\alpha ,\lambda )}.$$
It is clear that $`\rho _0`$ does not depend on the choice of the inner product.
Let $`r:𝔩^{}𝔤𝔤`$ be an $`𝔩`$-equivariant meromorphic function satisfying the quasi-unitarity condition $`r+r^{21}=\mathrm{\Omega }`$.
###### Theorem A 2 (\[EV\], Theorem 3.14).
The map $`r`$ is a classical dynamical r-matrix if and only if $`r_{|𝔥^{}}+\rho _0`$ is a classical dynamical r-matrix for $`𝔥`$.
Proof. This is proved in \[EV\] under the assumption that $`𝔤`$ is simple and $`𝔥𝔤`$ is a Cartan subalgebra. However, this assumption is not used in the proof and the result is valid in general.$`\mathrm{}`$
###### Proposition A 1.
Theorem A1 is valid if $`𝔩`$ is reductive, $`𝔤=𝔩_1\mathrm{}𝔩_n`$ with $`𝔩_i=𝔩`$, and $`B`$ is the cyclic permutation automorphism $`B:𝔩_i\stackrel{}{}𝔩_{i+1\mathrm{mod}n}`$.
Proof. Let $`(x_i)_{iI}`$ be an orthonormal basis of $`𝔥𝔩`$. For $`i=1,\mathrm{},n`$ we will write $`e_\alpha ^{(i)}`$ for the element of $`𝔩_i𝔤`$ corresponding to $`e_\alpha `$. With this notation, we have $`B(e_\alpha ^{(i)})=e_\alpha ^{(i+1)}`$ and
$$𝔤_j=\{g_1^{(1)}\mathrm{}g_n^{(n)};|g_{k+1}=e^{\frac{2i\pi j}{n}}g_k\}.$$
Finally, let $`𝔨_i𝔥^{(i)}`$ be the orthogonal complement to $`𝔥`$. Note that $`1B`$ restricts to an invertible operator on $`𝔨`$. A direct computation shows that
$$\begin{array}{cc}\hfill r_{B|𝔥^{}}& +\rho _0\hfill \\ & =\frac{\mathrm{\Omega }}{2}+\underset{\alpha >0,i}{}e_\alpha ^{(i)}\left(\frac{1}{2}\left(\frac{1+Be^{(\alpha ,\lambda )}}{1Be^{(\alpha ,\lambda )}}\right)e_\alpha ^{(i)}\right)+\underset{i}{}y_i\left(\frac{1+B}{2(1B)}y_i\right)\hfill \\ & =\underset{i}{}x_ix_i+\underset{\alpha >0,i}{}e_\alpha ^{(i)}e_\alpha ^{(i)}\underset{\alpha >0,i}{}\underset{l1}{}e^{l(\alpha ,\lambda )}e_\alpha ^{(i)}e_\alpha ^{(i+l)}\hfill \\ & +\frac{1}{2}\underset{i}{}\frac{B+1}{B1}y_iy_i\hfill \end{array}$$
where $`(y_i)_{iJ}`$ is an orthonormal basis of $`𝔨`$. By \[S\], Theorem 4 this expression is a dymamical r-matrix. Hence, by Theorem A.2, $`r_B`$ is a dynamical r-matrix.$`\mathrm{}`$
###
Define a map $`W:D\mathrm{\Lambda }^3𝔤`$ by
$$W(A)=\mathrm{Alt}(\overline{d\rho (A)})+CYB(\rho (A))+\frac{1}{4}Z$$
where $`Z=CYB(\mathrm{\Omega })`$. For any $`i,j/n`$, $`X𝔤_i`$, $`Y𝔤_j`$ and $`A𝔩`$, consider the expression
$$K_{ij}(A,X,Y)=(1XY,W(A))𝔤_{i+j}.$$
###### Lemma A 1.
The expression $`K_{ij}(A,X,Y)`$ is given by a universal Lie series in $`A,X`$ and $`Y`$.
Proof. Straightforward.$`\mathrm{}`$
###
Moreover, from Proposition A.1 we deduce the following result.
###### Proposition A 2.
We have $`K_{ij}(A,X,Y)=0`$ for all $`A,X,Y`$ if $`𝔩=𝔤𝔩_n()`$, $`𝔤=𝔩_1\mathrm{}𝔩_n`$ with $`𝔩_k=𝔩`$, and $`B`$ is the cyclic permutation automorphism.
###
Finally, we recall the following standard fact.
###### Lemma A 2.
Let $`P(X_1,\mathrm{},X_n)`$ be a Lie polynomial which vanishes identically for all $`X_1,\mathrm{}X_n𝔤𝔩_k()`$ for all $`k`$. Then $`P(X_1,\mathrm{},X_n)=0`$.
Proof. Let $`F_n`$ be the free Lie algebra in $`n`$ generators and let $`U_n`$ be its enveloping algebra (the free associative algebra). Let $`d`$ be the degree of $`P`$ and let $`I`$ be the ideal in $`U_n`$ generated by elements of degree at least $`d+1`$. Then $`U_n/I`$ is a finite-dimensional algebra. Let $`\sigma :U_n/I𝔤𝔩(U_n/I)`$ be the left regular representation. Then $`\sigma \left(P(X_1,\mathrm{},X_n)\right)=0`$. Hence $`P(X_1,\mathrm{},X_n)=0`$.$`\mathrm{}`$
###
Now, let us write $`K_{ij}=_kK_{ij}^{(k)}`$ where $`K_{ij}^{(k)}`$ is the homogeneous component of degree $`k`$. By Proposition A.2, $`K_{ij}^{(k)}(A,X,Y)=0`$ whenever $`A,X,Y𝔤𝔩_m()`$ for some $`m`$. Hence $`K_{ij}^{(k)}=0`$ by Lemma A.2. Thus $`W(A)=0`$ for all $`A𝔩`$. Theorem A.1 is proved.
### Examples.
Let $`𝔤`$ be a simple complex Lie algebra and $`𝔩𝔤`$ a semisimple subalgebra with same rank as $`𝔤`$. Such pairs are classified in \[BdS\]. Let $`Q_𝔩`$ and $`Q_𝔤`$ be the root lattices of $`𝔩`$ and $`𝔤`$ respectively and set $`\mathrm{\Gamma }=Q_𝔤/Q_𝔩`$. It follows from \[BdS\] that $`\mathrm{\Gamma }`$ is one of the groups $`/2`$, $`/3`$ or $`/5`$ (the case $`\mathrm{\Gamma }=/2`$ corresponds to symmetric spaces). Let $`\chi `$ be a nontrivial character $`\chi `$ of $`\mathrm{\Gamma }`$. Then $`\chi `$ gives rise to an automorphism $`B_\chi `$ of $`𝔤`$ whose set of fixed points is $`𝔩`$, defined by
$$B_{\chi |𝔤_\alpha }=\chi (\alpha )Id,$$
where $`𝔤_\alpha `$ is the root space of weight $`\alpha `$. Let $`\mathrm{\Omega }(S^2𝔤)^𝔤`$ be a Casimir element ($`\mathrm{\Omega }0`$). Let $`r:D𝔤𝔤`$ be a dynamical r-matrix such that $`r+r^{21}=\mathrm{\Omega }`$. It follows from Theorem A.2 and \[EV\] Theorem 3.1, that, up to gauge-equivalence,
$$r_{|𝔥^{}}+\rho _0=\frac{\mathrm{\Omega }}{2}+\underset{\alpha \mathrm{\Delta }_𝔤}{}\frac{1}{2}\mathrm{cotanh}(\frac{1}{2}(\alpha ,\lambda \nu ))e_\alpha e_\alpha ,$$
for some $`\nu 𝔥^{}`$, where $`\mathrm{\Delta }_𝔤`$ is the root system of $`𝔤`$. But then $`r`$ is regular at $`\lambda =0`$ if and only if $`(\alpha ,\nu )=0`$ modulo $`2\pi i`$ for all $`\alpha \mathrm{\Delta }_𝔩`$ and $`(\alpha ,\nu )0`$ modulo $`2\pi i`$ for all $`\alpha \mathrm{\Delta }_𝔤\backslash \mathrm{\Delta }_𝔩`$, where $`\mathrm{\Delta }_𝔩\mathrm{\Delta }_𝔤`$ is the root system of $`𝔩`$. Such $`\nu `$ defines a nontrivial character $`\chi `$ of $`\mathrm{\Gamma }`$, and it follows from \[EV\], Section 3.8 that $`r`$ is gauge-equivalent to the generalized Alekseev-Meinrenken dynamical r-matrix $`r_{B_\chi }`$. Hence the moduli space $`(𝔤,𝔩,\mathrm{\Omega })`$ consists of $`|\mathrm{\Gamma }|1`$ points.
### Acknowledgments.
The authors are grateful to P. Xu for useful discussions. The first author was supported by the NSF grant DMS-9700477. P.E performed this research as a CMI prize fellow. O.S conducted this research partially for the Clay Mathematics Institute, and thanks the MIT mathematics department for hospitality.
|
warning/0005/hep-ph0005173.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
The Sudbury Neutrino Observatory (SNO) is capable of detecting solar neutrinos through charged current and neutral current neutrino deuteron reactions as well as through $`\nu e`$ scattering. It is expected to perform precision measurements of a number of the characteristics of the solar neutrino flux. The forthcoming data from SNO will complement in a very important way the already available data of Homestake, SAGE, GALLEX, Kamiokande and Super-Kamiokande experiments , allowing crucial tests of the proposed solutions of the solar neutrino problem.
Recently, a comprehensive new study of the possible implications of the forthcoming SNO data for neutrino oscillation solutions of the solar neutrino problem has been performed by Bahcall, Krastev and Smirnov (hereafter BKS) . In the present paper we analyse possible implications of the SNO measurements for another particle physics solution of the solar neutrino problem which is currently consistent with all the data – the neutrino magnetic moment scenario. We also compare our predictions with those of BKS and discuss the possibilities of discriminating between the two scenarios.
Neutrinos with transition magnetic moments experience a simultaneous rotation of their spin and flavour in external magnetic fields (spin-flavour precession) . This precession can be resonantly enhanced in matter . The resonance spin-flavour precession of neutrinos (RSFP) in the matter and magnetic field of the sun can efficiently transform solar $`\nu _{eL}`$ into, e.g., $`\nu _{\mu R}`$ (or $`\overline{\nu }_{\mu R}`$ in the case of the transition magnetic moment of Majorana neutrinos) provided that the neutrino magnetic moment $`\mu _\nu \stackrel{>}{_{}}10^{11}\mu _B`$ and the average strength of the solar magnetic field in the region where the transition takes place $`B\stackrel{>}{_{}}40`$ kG. As $`\nu _{\mu R}`$ or $`\overline{\nu }_{\mu R}`$ either do not interact with the detectors or interact with them more weakly than $`\nu _{eL}`$, RSFP can account for the deficiency of observed solar neutrino flux. Analyses performed in the framework of this scenario show that a very good fit of the currently available data can be achieved. In particular, the fits of the total detection rates are much better than those in the case of the Mikheyev-Smirnov-Wolfenstein (MSW ) effect whereas the fits of the recoil electron spectrum in Super-Kamiokande in the frameworks of the RSFP and MSW effect are of nearly the same quality .
In the present paper we calculate, within the RSFP mechanism, the expected values of a number of observables to be measured by SNO. For three model solar magnetic field profiles that produce the best fits of the previous data we calculate the charged current event rate, the (neutral current)/(charged current) event ratio and the charged current electron spectrum as well as its first and second moments. We study the dependence of the calculated observables on the choice of the magnetic field profile and on the value of the solar $`hep`$ neutrino flux. We then compare our results with those obtained by BKS assuming that the solar neutrino problem is due to neutrino oscillations in vacuum or in matter.
The implications of the RSFP of solar neutrinos for the SNO experiment have been analysed in the past . Our analysis is more detailed, especially as far as the charged current electron spectrum is concerned; in addition, we use the magnetic field profiles which provide a good fit of all existing solar neutrino data, including the recent Super-Kamiokande ones which were not available when the analyses of were performed.
The paper is organized as follows: in sec. 2 we present the three magnetic field profiles to be used, in sec. 3 we define and evaluate the physical quantities which are relevant for the SNO experiment, for both the charged current (CC) and neutral current (NC) $`\nu d`$ reactions. Finally in section 4 we discuss the obtained results and draw our main conclusions.
## 2 Solar magnetic field profiles
Unfortunately, very little is known about the inner magnetic field of the sun, and one is forced to use various model magnetic filed profiles in order to account for the solar neutrino data in the framework of the RSFP mechanism. In a previous paper we investigated seven different magnetic field profiles and found that only some of them produce acceptable fits of the data. Here we present the three profiles which produced the best fits and which will be used in the present study. They are shown in figs. 1 and 2 of ref. (notice that our present profiles 1, 2 and 3 are respectively profiles 2, 5 and 6 of ref. ).
All three profiles show a sudden rise around the bottom of the convective zone, at 0.65-0.71 of the solar radius followed by a smoother decrease up to the surface. Here the field intensity is at most of the order of a few hundred Gauss. The best fits quoted here were taken from ref. and correspond to a neutrino magnetic moment $`\mu _\nu =10^{11}\mu _B`$ <sup>1</sup><sup>1</sup>1We recall that only the product of the magnetic moment and the magnetic field enters in the neutrino evolution equation and so our results apply to any other value of $`\mu _\nu `$ provided that the magnetic field is rescaled accordingly.. The first profile (profile 1) is
$$B=0,x<x_R.$$
(1)
$$B=B_0\frac{xx_R}{x_Cx_R},x_Rxx_C$$
(2)
$$B=B_0\left[1\frac{xx_C}{1x_C}\right],x_C<x1.$$
(3)
Here $`x=r/R_{}`$, $`R_{}`$ being the solar radius, and we take $`x_R=0.65`$, $`x_C=0.80`$. The best fit for the Homestake, SAGE, GALLEX, Kamiokande and Super-Kamiokande rates in this case corresponds to the values of the mass squared difference and peak field strength, respectively, $`\mathrm{\Delta }m_{21}^2=1.20\times 10^8`$ eV<sup>2</sup> and $`B_0=1.23\times 10^5`$ G ($`\chi ^2/d.o.f.=0.10/1`$).
The next profile (profile 2) is
$$B=0,x<x_R$$
(4)
$$B=\frac{B_0}{\mathrm{cosh}30(xx_R)},xx_R$$
(5)
with the best fit at $`\mathrm{\Delta }m_{21}^2=2.1\times 10^8`$ eV<sup>2</sup>, $`B_0=1.45\times 10^5`$ G ($`\chi ^2/d.o.f.=0.055/1`$).
Profile 3 is a modification of the previous one:
$$B=2.16\times 10^3\mathrm{G},x0.7105$$
(6)
$$B=B_1\left[1\left(\frac{x0.75}{0.04}\right)^2\right],0.7105<x<0.7483$$
(7)
$$B=\frac{B_0}{\mathrm{cosh}30(x0.7483)},0.7483x1$$
(8)
with $`B_0=0.998B_1`$. The best fit is in this case is $`\mathrm{\Delta }m_{21}^2=1.6\times 10^8`$ eV<sup>2</sup>, $`B_0=9.6\times 10^4`$ G with $`\chi ^2/d.o.f.=0.048/1`$ . The best fits of the Super-Kamiokande recoil electron spectrum found for these three profiles, at values of $`\mathrm{\Delta }m_{21}^2`$ and magnetic field strength rather close to those indicated above, were respectively $`\chi ^2=23.9,23.5,23.6`$ for 16 d.o.f. . In what follows we will be using the rate best fits.
We have determined the areas on the plane $`\mathrm{\Delta }m_{21}^2`$, $`B_0`$ corresponding to the 90% c.l. fits of the data ($`\chi ^2=\chi _{min}^2+2.7`$ for 1 d.o.f.) for each of the three magnetic field profiles described above. The resulting areas are shown in fig. 1 together with the best fit values.
## 3 Charged and neutral currents at SNO
SNO can detect solar neutrinos through the CC reaction
$$\nu _e+dp+p+e^{}$$
(9)
(energy threshold $`Q=1.44`$ MeV), NC reaction
$$\nu _x+dn+p+\nu _x$$
(10)
(energy threshold equal to the deuteron binding energy $`E_B=2.225`$ MeV, $`x=e,\mu `$ or $`\tau `$), and through the neutrino electron scattering $`\nu _x+e\nu _x+e`$. The rate and the recoil electron energy spectrum in the latter process at SNO are expected to be very similar to those at Super-Kamiokande studied in ; for this reason we shall not discuss them here.
We first consider the relative electron spectrum of reaction (9) which is defined as the ratio of the electron energy distribution and the corresponding distribution for standard model neutrinos. By standard model neutrinos we mean neutrinos described by the standard electroweak model (i.e. with no magnetic moment or mass) with their fluxes given by the standard solar model. We subdivide the electron energy spectrum into 0.5 MeV bins, with the relative spectrum for the i$`th`$ bin given by
$$S_i=\frac{_{T_i}^{T_{i+1}}𝑑T_Q^{\mathrm{}}𝑑Ef(E)P(E)_0^{EQ}𝑑T^{^{}}\frac{d\sigma _{CC}}{dT^{^{}}}(E,T^{^{}})R(T,T^{^{}})}{_{T_i}^{T_{i+1}}𝑑T_Q^{\mathrm{}}𝑑Ef(E)_0^{EQ}𝑑T^{^{}}\frac{d\sigma _{CC}}{dT^{^{}}}(E,T^{^{}})R(T,T^{^{}})}\frac{_{T_i}^{T_{i+1}}𝑑T𝑑R_{CC}/𝑑T}{_{T_i}^{T_{i+1}}𝑑T𝑑R_{CC}^{st}/𝑑T}$$
(11)
Here $`P(E)`$ denotes the electron neutrino survival probability, $`T`$ is the measured recoil electron kinetic energy, in contrast to the physical one, $`T^{^{}}`$. Following , we approximate the energy resolution function of the detector $`R(T^{},T)`$ by a Gaussian
$$R(T,T{}_{}{}^{})=\frac{1}{\mathrm{\Delta }_T^{^{}}\sqrt{2\pi }}\mathrm{exp}[\frac{(T^{^{}}T)^2}{2\mathrm{\Delta }_T^{^{}}^2}],\mathrm{\Delta }_T^{^{}}=(1.1\mathrm{MeV})\sqrt{\frac{T^{^{}}}{10\mathrm{MeV}}}.$$
(12)
The cross sections of the CC and NC $`\nu d`$ reactions used throughout this paper are those of Kubodera et al. . The SNO experiment is only sensitive to the high-energy <sup>8</sup>B and $`hep`$ components of the solar neutrino flux. We use the BP98 solar neutrino spectrum $`f(E)`$ except that the $`hep`$ neutrino flux is allowed to deviate from its nominal BP98 value. The $`hep`$ flux is about 3 orders of magnitude smaller than that of <sup>8</sup>B neutrinos and so gives a negligible contribution to the CC and NC rates. However, its contribution to the highest energy part of the CC electron spectrum may be important as its endpoint energy is 18.8 MeV whereas that of <sup>8</sup>B neutrinos is only 15 MeV.
Unless otherwise specified, we consider the case of Majorana neutrinos in the present paper. The survival probability $`P(E)`$ is calculated by solving the RSFP evolution equation numerically . The relative electron spectra for the three magnetic field profiles discussed in sec. 2 are shown in fig 2. The highest energy bin ($`E_e14`$ MeV) represents the average value of $`S`$ for measured electron energies between 14 and 20 MeV. The solid lines correspond to the best fits whereas the dashed and dotted lines show the maximum and minimum values obtained within the 90% c.l. allowed regions of values of $`\mathrm{\Delta }m_{21}^2`$ and $`B_0`$ obtained by fitting the previous data and shown in fig. 1. In order to check the sensitivity of the spectrum to the poorly known $`hep`$ flux we perform all the calculations for three values of the flux scaling factor, $`f_{hep}=0`$, 1 and 20, with $`f_{hep}=1`$ corresponding to the nominal $`hep`$ flux value in the BP98 standard solar model (SSM).
We also calculate the first and second moments $`T`$ and $`\sigma `$ of the electron spectrum, defined through
$$T^n=\frac{1}{R_{CC}}_{T_m}^{\mathrm{}}T^n\frac{dR_{CC}}{dT}𝑑T,R_{CC}=_{T_m}^{\mathrm{}}\frac{dR_{CC}}{dT}𝑑T,\sigma =\sqrt{T^2T^2},$$
(13)
with $`dR_{CC}/dT`$ defined in (11). The deviations of these moments from their values predicted by the SSM are convenient quantitative characteristics of the electron spectrum distortion . The results of the calculations for two different values of the threshold electron kinetic energies $`T_m`$ which correspond to the total energies of 5 MeV and 8 MeV and for $`f_{hep}=1`$ and 20 are presented in tables I and II. We have also calculated $`T`$ and $`\sigma `$ for $`f_{hep}=0`$ and found that their values are practically indistinguishable from those for $`f_{hep}=1`$. This also applies to the CC event ratio $`r_{CC}`$ and NC/CC double ratio $`\overline{r}_{NC}`$ discussed below. The errors presented in tables I and II include the 90% c.l. errors coming from the fits of the parameters $`\mathrm{\Delta }m_{21}^2`$ and $`B_0`$ (fig. 1) as well as from the uncertainties in the energy resolution and scale, <sup>8</sup>B neutrino spectrum, reaction cross section and statistics (assuming 5000 CC events). These uncertainties were taken from table II of ref. .
Next we examine the ratio of the CC event rate for neutrinos undergoing RSFP and that for standard model neutrinos,
$$r_{CC}=\frac{_{T_m}^{\mathrm{}}(dR_{CC}/dT)𝑑T}{_{T_m}^{\mathrm{}}(dR_{CC}^{st}/dT)𝑑T}=\frac{R_{CC}}{R_{CC}^{st}}.$$
(14)
This ratio depends on the chosen threshold energy $`T_m`$ and we again use the same values of $`T_m`$ that we used in calculating the moments of the electron spectrum. The results are presented in table III. The absolute values of the CC event rates for standard model neutrinos for the total electron threshold energies $`E_e=5`$ MeV are 5.132 SNU and 5.253 SNU for $`f_{hep}=1`$ and 20 respectively, and those for the electron threshold energy $`E_e=8`$ MeV are 2.355 SNU and 2.451 SNU.
We have also calculated the total event rate of the NC reaction (10)
$$R_{NC}=_{E_B}^{\mathrm{}}f(E)\{P(E)\sigma _{NC}^{\nu d}(E)+[1P(E)]\sigma _{NC}^{\overline{\nu }d}(E)\}ϵ(E)𝑑E$$
(15)
Here $`ϵ(E)`$ is the NC detection efficiency; following refs. we have taken $`ϵ(E)=0.50`$. For neutrino energies $`E15`$ MeV the cross sections of the NC $`\nu d`$ and $`\overline{\nu }d`$ reactions differ by less than 7.3%, the difference at $`E=9`$ MeV being about 4% . These differences are within the uncertainty of the value of the NC cross sections itself, and to a good approximation one can put $`\sigma _{NC}^{\overline{\nu }d}(E)=\sigma _{NC}^{\nu d}(E)`$. The NC rate (15) then does not depend on the $`\nu _e`$ survival neutrino probability $`P(E)`$, and one therefore expects $`R_{NC}`$ in the case of neutrinos undergoing RSFP to coincide with that for standard neutrinos $`R_{NC}^{st}`$ . Notice that for $`\mu _\nu \stackrel{<}{_{}}10^{11}\mu _B`$ the electromagnetic contribution to the NC neutrino-deuteron disintegration reaction (10) due to the neutrino magnetic moment is more than eight orders of magnitude smaller than the standard electroweak one and so can be safely neglected. For $`f_{hep}=1`$ and 20 we obtain $`R_{NC}^{st}=1.216`$ SNU and 1.242 SNU respectively.
An important characteristics of the $`\nu d`$ reactions at SNO is the ratio $`\overline{r}_{NC}`$ of the ratios of the NC and CC event rates to their respective values for standard neutrinos. This double ratio is free of many uncertainties which are present in the single ratios $`r_{CC}`$ and $`r_{NC}R_{NC}/R_{NC}^{st}`$. In particular, the errors in $`r_{CC}`$ and $`r_{NC}`$ due to the uncertainties in the flux of <sup>8</sup>B neutrinos practically cancel out in the double ratio, and those due to the uncertainties in the CC and NC cross sections cancel out to a large extent. As was mentioned above, for Majorana neutrinos one has $`r_{NC}=1`$ to a good accuracy, and so $`\overline{r}_{NC}=1/r_{CC}`$. The values of $`\overline{r}_{NC}`$ for the three magnetic field profiles that we consider are given in table IV. The indicated errors are the combined 90% c.l. ones coming from the uncertainties in the fitted values of $`\mathrm{\Delta }m_{21}^2`$ and $`B_0`$ (see fig. 1) as well as from the uncertainties in the electron energy resolution and scale in the CC reaction, <sup>8</sup>B neutrino spectrum, CC and NC cross sections and statistics (assuming 5000 CC events). The latter uncertainties were taken from table II of ref. .
Finally, we remark on the case of Dirac neutrino transition moments. In a sense this case is similar to neutrino oscillations into sterile neutrinos because the RSFP due to Dirac neutrino magnetic moments transforms $`\nu _{eL}`$ into sterile $`\nu _{\mu R}`$ or $`\nu _{\tau R}`$. The CC rate ratio $`r_{CC}`$ in this case is similar to that in the Majorana neutrino case since the survival probabilities in the two cases are very close to each other. However, the NC rate $`R_{NC}`$ and the rate ratio $`r_{NC}`$ (and so the double ratios $`\overline{r}_{NC}`$) in the Majorana and Dirac cases are drastically different: while for Majorana neutrinos $`R_{NC}R_{NC}^{st}`$ ($`r_{NC}1`$), in the case of Dirac neutrinos $`R_{NC}`$ exhibits a suppression similar to that of $`R_{CC}`$ and so the double ratio $`\overline{r}_{NC}`$ is close to unity. Thus, the double ratio $`\overline{r}_{NC}`$ can be used to discriminate between Dirac and Majorana neutrinos. We would like to stress that this is one of a very few known quantities that hold such a potential, the best known other example being the neutrinoless double beta decay. Notice that in general the RSFP of Dirac neutrinos leads to a fit of the solar neutrino data that is worse than that due to the RSFP of Majorana neutrinos and therefore we do not pursue here this possibility in detail.
## 4 Summary and discussion
We have investigated the expectations from the RSFP solution to the solar neutrino problem for the SNO experiment using the solar magnetic field profiles that provide the best fits for the rates of the Homestake, SAGE, GALLEX, Kamiokande and Super-Kamiokande experiments. We considered Majorana neutrinos and discussed the quantities that are relevant for the CC and NC processes (9) and (10) in SNO. To this end we have examined the expected relative CC electron spectrum $`S(E_e)`$ as well as the first and second moments of the spectrum $`T`$ and $`\sigma `$, the ratio $`r_{CC}`$ of the CC event rate to that of standard model neutrinos, and the ratio $`\overline{r}_{NC}`$ of the ratios of the NC and CC event rates to their respective values for standard neutrinos.
The values of the CC event rate ratio $`r_{CC}`$ calculated for the electron threshold energies 5 MeV and 8 MeV are very close to each other, the difference being only 2 – 4% (table III). This is related to the the fact that, for the magnetic field profiles that we consider, the high-energy part of the survival probability $`P(E)`$ for neutrinos undergoing the RSFP is rather flat (see figs. 3 and 4 in ). Therefore both $`R_{CC}`$ and $`R_{CC}^{st}`$ decrease with increasing electron energy threshold to nearly the same extent so that their ratio $`r_{CC}`$ changes very little. For the same reason the double ratio $`\overline{r}_{NC}`$, which depends on the CC electron threshold energy only through $`r_{CC}`$, is rather insensitive to the value of this threshold energy. In contrast to this, the first and second moments of the CC electron spectrum $`T`$ and $`\sigma `$ depend sensitively on the chosen electron threshold energy (see tables I and II).
To assess the sensitivity of the observables to be measured by SNO to the poorly known value of the $`hep`$ neutrino flux we have performed all the calculations for three values of this flux: zero flux ($`f_{hep}=0`$), the nominal flux of the BP98 standard solar model ($`f_{hep}=1`$) and a factor of 20 larger one ($`f_{hep}=20`$). Changing the value of $`f_{hep}`$ from 0 to 1 does not lead to any noticeable difference in the calculated observables. The CC event rate ratio $`r_{CC}`$ and the NC/CC double ratio $`\overline{r}_{CC}`$ only weakly depend on the value of the $`hep`$ flux. This is mainly because the expected contributions of the $`hep`$ neutrinos to the CC and NC event rates $`R_{CC}`$ and $`R_{NC}`$ are very small and even increasing these contributions by a factor of 20 would not change $`R_{CC}`$ and $`R_{NC}`$ much. The sensitivity to $`f_{hep}`$ of the ratio $`r_{CC}`$ and double ratio $`\overline{r}_{NC}`$ is further reduced due to a partial cancellation of the $`f_{hep}`$ dependences of the numerators and denominators. The first moment of the CC electron spectra $`T`$ is also rather insensitive to the $`hep`$ neutrino flux: changing $`f_{hep}`$ from 1 to 20 modifies $`T`$ by less than 1%. The second moment $`\sigma `$ is more sensitive to the $`hep`$ neutrino flux: changing $`f_{hep}`$ from 1 to 20 increases $`\sigma `$ by about 4% for the electron energy threshold $`E_{min}=5`$ MeV and by about 9% for $`E_{min}=8`$ MeV. These features are independent of whether or not neutrinos undergo RSFP.
As can be seen from fig. 2, the high energy part of the relative CC electron spectrum ($`E\stackrel{>}{_{}}12`$ MeV) depends sensitively on the magnitude of the $`hep`$ neutrino flux. For $`f_{hep}=20`$ the excess of the number of the high energy events ($`E14`$ MeV) over the SSM prediction with nominal $`hep`$ flux can be quite significant. It should be noted, however, that, although it is quite likely that the actual $`hep`$ neutrino flux exceeds the nominal one of the BP98 model, the illustrative value $`f_{hep}=20`$ that we used may in fact be too high. The most recent calculation gives $`f_{hep}5`$.
All quantities that we have calculated have very similar values for the three magnetic field profiles that we chose. In particular, the values $`T`$ and $`\sigma `$ differ by less than 1% for different profiles, and those of $`r_{CC}`$ and $`\overline{r}_{NC}`$ differ by 3 - 4%. This is the consequence of the fact that the magnetic field profiles used in our calculations, although rather different, lead to very similar $`\nu _e`$ survival probabilities $`P(E)`$ .
We shall now compare our predictions with those for standard neutrinos (with no magnetic moment or mass) and for neutrinos undergoing oscillations in vacuum or in matter. As can be seen from table III, the CC detection rates for neutrinos undergoing RSFP constitute about 40% of those for the standard neutrinos and SSM fluxes. The errors in $`r_{CC}`$ are rather large (90% c.l. errors $`3540\%`$), with about a half of their values coming from the uncertainty of the <sup>8</sup>B neutrino flux. The predicted values of $`r_{CC}`$ differ from the SSM one $`r_{CC}=1`$ by about $`5\sigma `$. However, for neutrinos of vanishing mass and magnetic moments, the <sup>8</sup>B neutrino flux measured by Super-Kamiokande, $`\varphi _{{}_{}{}^{8}\mathrm{B}}=(0.475\pm 0.015)\varphi _{{}_{}{}^{8}\mathrm{B}}^{\mathrm{SSM}}`$, implies that the actual <sup>8</sup>B neutrino flux is about 0.48 of the one given by the SSM and so the expected value of the CC event rate ratio for standard neutrinos is $`r_{CC}0.48`$ rather than $`r_{CC}=1`$. This value is less than $`1\sigma `$ away from our predictions given in table III, and so $`r_{CC}`$ is not a suitable parameter for discriminating between standard and non-standard neutrinos.
The NC/CC double ratio $`\overline{r}_{NC}`$ has smaller errors (except for profile 1); it is practically independent of the uncertainty of the <sup>8</sup>B neutrino flux and in addition a number of other uncertainties of $`R_{CC}`$ and $`R_{NC}`$ drop out from this ratio. The calculated values of $`\overline{r}_{NC}`$ are of the order 2.5 and exceed the prediction for standard neutrinos $`\overline{r}_{NC}=1`$ by more than $`7\sigma `$ for profiles 2 and 3 and by about $`4\sigma `$ for profile 1. Thus $`\overline{r}_{NC}`$ is an ideal indicator of non-standard neutrinos.
The predicted relative CC electron spectrum $`S(E_e)`$ for $`f_{hep}=1`$ is rather flat (see fig. 2). For $`f_{hep}=20`$, one expects an excess of the high energy events. In both cases there is also a small decrease at low energies, $`E\stackrel{<}{_{}}8`$ MeV. The shapes of the relative spectrum that we obtained for all three studied magnetic field profiles are very similar to each other and to that predicted for the SMA solution of the solar neutrino problem in the case of neutrino oscillations (see fig. 3 in ). They are also similar to the $`S(E_e)`$ shape in the case of VAC<sub>S</sub> oscillation solution, but the latter has a steeper increase at $`E_e\stackrel{>}{_{}}10`$ MeV. The shapes of the relative CC electron spectrum for LOW and LMA and VAC<sub>L</sub> solutions differ from ours in that the former two are almost horizontal at low energies, whereas the latter one has a distinct dip in the energy range $`E_e8`$ – 10 MeV.
Our predicted values of $`r_{CC}`$ are typically slightly larger than those obtained in for the LMA and LOW solution of the solar neutrino problem in the neutrino oscillation scenario, and in most part of the allowed range, also larger than those of the VAC<sub>S</sub> solution studied by BKS (compare our table III with table V of ). Our values of $`r_{CC}`$ are rather close to those for the SMA and VAC<sub>L</sub> neutrino oscillation solutions. However, even in the case of the LMA and LOW solutions, there is a partial overlap with the RSFP predictions because the allowed regions of $`r_{CC}`$ are rather large. The values of the NC/CC double ratio $`\overline{r}_{NC}`$ that we obtained are typically larger than those for the VAC<sub>L</sub> solution, lower than those for the LMA and LOW solutions and similar to those for the SMA and VAC<sub>S</sub> solutions studied in .
For an electron energy threshold of 5 MeV the best fit value of the first moment of the CC electron spectrum $`T`$ in the case of the RSFP is systematically up-shifted compared to the SSM prediction by about (55 – 75) keV. The values of the shift $`\mathrm{\Delta }T`$ allowed at 90% c.l. range between $`110`$ and +230 keV, which is about three times the expected combined calculational and measurement uncertainty, $`\pm 96`$ keV . For the electron threshold energy of 8 MeV the predicted range of $`\mathrm{\Delta }T`$ is somewhat larger, between $`180`$ and +215 keV. Therefore in principle SNO holds a potential of measuring the shift in $`T`$ due to non-standard neutrino properties. The shifts in $`T`$ in the case of the RSFP have partial overlap with those predicted in the case of neutrino oscillations in vacuum or in matter . Our predicted values of the shift in the second moment $`\sigma `$ are smaller than those in $`T`$. Their values allowed at 90% c.l. range between $`70`$ and $`82`$ keV, which has to be compared with the $`1\sigma `$ uncertainty of $`\pm 44`$ keV. Again, there is a partial overlap between our predictions for $`\mathrm{\Delta }\sigma `$ and those in the case of neutrino oscillations , but there are also rather large regions of no overlap, especially in the case of VAC<sub>S</sub> and VAC<sub>L</sub> oscillation solutions.
We therefore conclude that the possibility of experimentally disentangling the two types of solutions to the solar neutrino problem (neutrino oscillations and magnetic moments) depends to a large extent on where in the allowed region the neutrino parameters lie. Given the uncertainties in the calculations and the expected uncertainties in the experimental results, the unambiguous discrimination on the basis of the average rates and electron spectrum distortions appears to be difficult.
On the other hand, the RSFP mechanism can lead to time dependence of the solar neutrino signal due to the variability of the solar magnetic field strength. Such time dependence should be manifest in the CC reaction rate and electron spectrum, but should not be observable in the NC rate provided that neutrinos are Majorana particles . In particular, the CC signal can have an 11-year periodicity related to the solar activity cycle . There may also be seasonal variations of the CC signal due to the surface equatorial gap in the toroidal magnetic field of the sun and a non-zero angle between the solar equatorial plane and the earth’s ecliptic . These seasonal variations are expected to be different from those expected in the case of VAC<sub>S</sub> and VAC<sub>L</sub> oscillation solutions. However, the gap in the magnetic field seen at the surface of the sun may not be present in the inner regions where the RSFP effectively takes the place. It is therefore difficult to make an unambiguous prediction on whether or not the seasonal variations of the CC signal should take place in the case of the RSFP scenario.
If both transition magnetic moments and mixing of massive neutrinos are present, the combined action of the RSFP and neutrino oscillations may lead to an observable flux of solar $`\overline{\nu }_e`$’s provided that the mixing angle is not too small, $`\mathrm{sin}2\theta _0\stackrel{>}{_{}}0.1`$. SNO can detect $`\overline{\nu }_e`$ through the CC reaction $`\overline{\nu }_e+dn+n+e^+`$ (for a recent discussion of the $`\overline{\nu }_e`$ signal in SNO see the second reference in ).
Thus, the possible time dependence of the CC signal and spectrum distortion together with the time independence of the NC signal in the case of the magnetic moment solution remains the best hope for a discrimination between the neutrino magnetic moment and oscillation scenarios. For a hybrid solution (neutrino magnetic moment plus flavour mixing) the smoking gun signature would be an observation of $`\overline{\nu }_e`$’s from the sun.
We are grateful to K. Kubodera for useful correspondence. The work of E. A. was supported by Fundação para a Ciência e a Tecnologia through the grant PRAXIS XXI/BCC/16414/98 and also in part by the TMR network grant ERBFMRX-CT960090 of the European Union.
|
warning/0005/hep-th0005139.html
|
ar5iv
|
text
|
# Exact Cosmological Solution and Modulus Stabilization in the Randall-Sundrum Model with Bulk Matter
## Abstract
We provide the exact time-dependent cosmological solutions in the Randall-Sundrum (RS) setup with bulk matter, which may be smoothly connected to the static RS metric. In the static limit of the extra dimension, the solutions are reduced to the standard Friedmann equations. In view of our solutions, we also propose an explanation for how the extra dimension is stabilized in spite of a flat modulus potential at the classical level.
preprint: SNUTP 00-010
As a possible solution of the gauge hierarchy problem, Randall and Sundrum (RS) proposed an $`S^1/Z_2`$ orbifold model with non-factorizable geometry of space-time , which has immediately attracted a great deal of attention. The model employs two branes, Brane 1 (B1) with a positive cosmological constant(or brane tension) $`\mathrm{\Lambda }_16k_1M^3`$ and Brane 2 (B2) with a negative cosmological constant $`\mathrm{\Lambda }_26k_2M^3`$, and introduces a negative bulk cosmological constant $`\mathrm{\Lambda }_b6k^2M^3`$. B1 is interpreted as the hidden brane and B2 is identified with the visible brane. Then the metric has an exponential warp factor which could be used to understand the huge gap between the Planck and eletroweak scales. Although the RS setup introduces cosmological constants $`k`$ in the bulk and $`k_1`$ and $`k_2`$ on the branes, it still describes a static universe because of the fine-tuning between the bulk and brane cosmological constants $`k=k_1=k_2`$, which is a consistency condition in the model. Hence, if the fine-tuning is not exact, the solution has the time dependence and the universe expands exponentially but its form is not suitable for the standard Big Bang universe after the inflation. To circumvent this cosmological problem, some approximation schemes regarding the brane matter in the static limit has been taken into account and(or) some conditions (such as the positive brane tension) for the brane and bulk cosmological constants are required . Thus B1 was thought of as the visible brane even though it is inconsistent with the original motivation for the gauge hierarchy solution . This problem, however, can be resolved with the Gauss-Bonnet interaction , bulk matter effects and extra dimension stabilization process , etc.
In this paper, we will present exact cosmological solutions in the RS setup with bulk matter. Although there have been many cosmological solutions in the RS setup , the graceful exit problem from the inflation phase to the standard Big Bang cosmology has not been seriously considered yet. In addition, the role of the extra dimension in the presence of the bulk matter is not well understood. Our exact solutions converge to the RS metric if the space time is made to be static, and leads to the standard Friedmann equations if the fifth dimension is stabilized. In view of our exact solutions, we can find a clue for a stabilization mechanism of the fifth dimension and obtain a small compactified fifth dimension naturally.
Throughout this paper we consider a (4+1) dimensional universe with coordinate indexed by (0, 1, 2, 3, 5). The action describing the bulk matter as well as bulk gravity and brane matter is
$$S=d^5x\sqrt{g}\left(\frac{R}{2}\mathrm{\Lambda }_b+^{(M)}\right)+\underset{j=1,2\mathrm{branes}}{}d^4x\sqrt{g^{(j)}}\left(_j^{(M)}\mathrm{\Lambda }_j\right),$$
(1)
where we set the fundamental scale $`M=1`$. $`^{(M)}`$ and $`_j^{(M)}`$ represent matter contributions in the bulk and on the branes. For compatibility with the cosmological principle that our three dimensional space is homogeneous and isotropic, we assume that the metric of the universe has the following form,
$$ds^2=e^{2N(\tau ,y)}d\tau ^2+e^{2A(\tau ,y)}\delta _{ij}dx^idx^j+e^{2B(\tau ,y)}dy^2,$$
(2)
where $`\tau `$ denotes time and $`y`$ denotes the fifth component. From the metric ansatz, the Einstein tesor $`G_{MN}`$ is derived through the standard calculation,
$`G_{00}^{(1)}=3e^{2(NB)}\left[A^{\prime \prime }+2A^2A^{}B^{}\right]`$ (3)
$`G_{00}^{(2)}=3\left[\dot{A}^2+\dot{A}\dot{B}\right]`$ (4)
$`G_{ii}^{(1)}=e^{2(AB)}\left[2A^{\prime \prime }+3A^2+N^{\prime \prime }+N^2+2A^{}N^{}2A^{}B^{}B^{}N^{}\right]`$ (5)
$`G_{ii}^{(2)}=e^{2(AN)}\left[2\ddot{A}+3\dot{A}^2+\ddot{B}+\dot{B}^2+2\dot{A}\dot{B}2\dot{A}\dot{N}\dot{B}\dot{N}\right]`$ (6)
$`G_{55}^{(1)}=3\left[A^2+A^{}N^{}\right]`$ (7)
$`G_{55}^{(2)}=3e^{2(BN)}\left[\ddot{A}+2\dot{A}^2\dot{A}\dot{N}\right]`$ (8)
$`G_{05}=3\left[A^{}\dot{B}+\dot{A}N^{}\dot{A}^{}\dot{A}A^{}\right],`$ (9)
where dot and prime denote the derivatives with respect to $`\tau `$ and $`y`$, respectively, and the $`i`$ runs through $`1`$, $`2`$, and $`3`$. Here diagonal Einstein tensors are split into two parts, $`G_{AA}^{(1)}`$ and $`G_{AA}^{(2)}`$, depending on the nontrivial $`y`$ and $`\tau `$ derivatives, respectively. Thus the original Einstein tensor is, of course, expressed as the sum, $`G_{AA}G_{AA}^{(1)}+G_{AA}^{(2)}`$, where the $`A`$ is $`(0,i,5)`$. The source part of the Einstein equation is composed of the cosmological constant and the energy-momentum tensor of matter. In this paper, we will regard the matter as perfect fluid. For the future convenience, we divide also the source tensor into two parts,
$`T^{(1)}_B^A`$ $`=`$ $`(1\eta )\mathrm{diag}[\mathrm{\Lambda }_b,\mathrm{\Lambda }_b,\mathrm{\Lambda }_b,\mathrm{\Lambda }_b,\mathrm{\Lambda }_b]`$ (12)
$`{\displaystyle \underset{j=1,2\mathrm{branes}}{}}\delta (yy_j)e^B\mathrm{diag}[\mathrm{\Lambda }_j,\mathrm{\Lambda }_j,\mathrm{\Lambda }_j,\mathrm{\Lambda }_j,0]`$
$`+{\displaystyle \underset{j=1,2\mathrm{branes}}{}}\delta (yy_j)e^B\mathrm{diag}[\widehat{\rho }_j,\widehat{p}_j,\widehat{p}_j,\widehat{p}_j,0]`$
$`T^{(2)}_B^A`$ $`=`$ $`\mathrm{diag}[(\rho +\eta \mathrm{\Lambda }_b),P\eta \mathrm{\Lambda }_b,P\eta \mathrm{\Lambda }_b,P\eta \mathrm{\Lambda }_b,P_5\eta \mathrm{\Lambda }_b],`$ (13)
where the $`\widehat{\rho }_j`$ and $`\widehat{p}_j`$ are nontrivial components of the energy-momentum tensor of the matter living only on the $`j`$-th brane, and $`\eta `$ is a number representing how $`\mathrm{\Lambda }_b`$ is split into $`T^{(1)}_B^A`$ and $`T^{(2)}_B^A`$. The total source tensor is described as $`T_B^A=T^{(1)}{}_{B}{}^{A}+T^{(2)}_B^A`$. Here we set $`T_{05}=0`$ because it is believed that there is no flow of matter along the fifth direction. The continuity equation of the energy-momentum tensor $`T_{B;A}^A=0`$ must be satisfied, whose $`B=0`$ and $`B=5`$ components are
$`\dot{\rho }+3\dot{A}\left(\rho +P\right)+\dot{B}\left(\rho +P_5\right)=0`$ (14)
$`P_5^{}+3A^{}\left(P_5P\right)+N^{}\left(\rho +P_5\right)=0.`$ (15)
The $`B=i`$ component is identically zero.
Now let us take some ansatze,
$`G_{AA}^{(1)}`$ $`=`$ $`T_{AA}^{(1)}(\mathrm{or}G_{AA}^{(2)}=T_{AA}^{(2)})`$ (16)
$`A^{}(\tau ,y)`$ $`=`$ $`N^{}(\tau ,y).`$ (17)
The above ansatze have been chosen to fulfill our purpose of restoring the Randall-Sundrum metric in the static limit. The ansatz Eq. (16) and $`G_{05}`$ read
$`3e^{2B}`$ $`\left[A^{\prime \prime }+2A^2A^{}B^{}\right]`$ (19)
$`=(1\eta )\mathrm{\Lambda }_be^B\left[\delta (y)\left(\mathrm{\Lambda }_1+\rho _1\right)+\delta (y1/2)\left(\mathrm{\Lambda }_2+\rho _2\right)\right]`$
$`e^{2B}`$ $`\left[2A^{\prime \prime }+3A^2+N^{\prime \prime }+N^2+2A^{}N^{}2A^{}B^{}B^{}N^{}\right]`$ (21)
$`=(1\eta )\mathrm{\Lambda }_be^B\left[\delta (y)\left(\mathrm{\Lambda }_1p_1\right)+\delta (y1/2)\left(\mathrm{\Lambda }_2p_2\right)\right]`$
$`3e^{2B}`$ $`\left[A^2+A^{}N^{}\right]=(1\eta )\mathrm{\Lambda }_b`$ (22)
$`A^{}\dot{B}+`$ $`\dot{A}N^{}\dot{A}^{}\dot{A}A^{}=0.`$ (23)
Under the ansatz Eq. (17), Eq. (22) becomes
$$A^2=(1\eta )\frac{\mathrm{\Lambda }_b}{6}\times e^{2B}k^2e^{2B}.$$
(24)
Hence, de Sitter($`\mathrm{\Lambda }_b>0,\eta >1`$), anti-de Sitter ($`\mathrm{\Lambda }_b<0,\eta <1`$), and flat Minkowski space ($`\mathrm{\Lambda }_b=0`$) are possible in the bulk. The solution consistent with the $`S^1/Z_2`$ orbifold symmetry is
$$A(\tau ,|y|)kF(\tau ,|y|)+J(\tau )\mathrm{and}F(\tau ,|y|)^{}=e^{B(\tau ,|y|)}sgn(y),$$
(25)
where the $`sgn(y)`$ is defined as $`sgn(y)|y|^{}=2[\theta (y)\theta (y1/2)]1`$. Then, because of the ansatz Eq. (17), the exponential factor $`N`$ of the $`g_{00}`$ component in our metric tensor is written as
$$N(\tau ,|y|)=kF(\tau ,|y|)+K(\tau )kF(\tau ,|y|),$$
(26)
where $`K(\tau )`$ is removed by the redefinition of time $`\tau `$ in the second part of the above equation. Therefore, we ignore $`K(\tau )`$ below.
The above result Eq. (25) leads to some useful relations,
$`A^{\prime \prime }`$ $`=`$ $`ke^BB^{}sgn(y)2\left[\delta (y)\delta (y1/2)\right]ke^B`$ (27)
$`=`$ $`A^{}B^{}2\left[\delta (y)\delta (y1/2)\right]ke^B`$ (28)
$`\dot{A}^{}`$ $`=`$ $`ke^Bsgn(y)\dot{B}=A^{}\dot{B}.`$ (29)
Eq. (29) implies that the $`N(\tau ,|y|)`$ should be stabilized if the $`B(\tau ,|y|)`$ can be stabilized somehow, since $`A^{}=N^{}`$. With Eqs. (17) and (29), we can show directly that our ansatz is consistent with Eq. (23). Because of Eqs. (17) and (28), Eqs. (19) and (21) just require matching the boundary conditions,
$$k=\frac{1}{6}\left(\mathrm{\Lambda }_1+\widehat{\rho }_1\right)=\frac{1}{6}\left(\mathrm{\Lambda }_2+\widehat{\rho }_2\right)\mathrm{and}\widehat{\rho }_j=\widehat{p}_j.$$
(30)
Hence, considering the fluid continuity equation on the branes, $`\dot{\widehat{\rho }}_j+3\dot{A}(\widehat{\rho }_j+\widehat{p}_j)=0`$, we can arrive at a result $`\widehat{\rho }_j=constant`$ and so the $`\widehat{p}_j`$ is a constant also, which are expected results from our assumption $`T_{05}=0`$. The non-dynamical properties of the $`\widehat{\rho }_j`$ and $`\widehat{p}_j`$ obstruct our intention to explain the dynamics of the space-time exactly with brane matter, which is a reason for us to consider bulk matter. \[The solution with the brane matter condition $`\widehat{\rho }_j=\widehat{p}_j`$ is equivalent to that we do not have a solution with the radiation or matter dominated phases by brane matter. Anyway, the radiation or matter dominated universe must results with a vanishing effective cosmological constant, in which case we have not obtained a most general solution.\]
Now that we have fulfilled the ansatz Eq. (16) already, the remaining equations, $`G_{AA}^{(2)}=T_{AA}^{(2)}`$ are
$`\rho +\eta \mathrm{\Lambda }_b`$ $`=`$ $`3e^{2N}\left[\dot{A}^2+\dot{A}\dot{B}\right]`$ (31)
$`P\eta \mathrm{\Lambda }_b`$ $`=`$ $`e^{2N}\left[2\ddot{A}+3\dot{A}^2+\ddot{B}+\dot{B}^2+2\dot{A}\dot{B}2\dot{A}\dot{N}\dot{B}\dot{N}\right]`$ (32)
$`P_5\eta \mathrm{\Lambda }_b`$ $`=`$ $`3e^{2N}\left[\ddot{A}+2\dot{A}^2\dot{A}\dot{N}\right],`$ (33)
which describe the relation between matter and geometry dynamics. They are nothing but the extended Friedmann equations. With Eqs. (17) and (29), we can check that the above equations Eqs. (31), (32) and (33) satisfy both fluid continuity equations, Eqs. (14) and (15) identically, that is, the constraints, Eqs. (14) and (15) are just redundant equations, which are interesting results. Compare with the standard cosmology, where two Freedmann equations and one continuity equation lead to one dependent equation. Therefore, the remaining required conditions for the solution are only Eqs. (25) and (26). The relations among the $`\rho `$, $`P`$ and $`P_5`$ may be, of course, governed by particle physics.
Toward a simple solution, let us consider the case that the size of the extra dimension is stabilized, i.e. $`\dot{B}=0`$, which leads also to $`B^{}=0`$ generically by redefinition of $`y`$. Because of Eqs. (17), (26) and (29), then, $`\dot{N}`$ is also generically set to zero. After all we have
$$\dot{B}=B^{}=\dot{N}=0.$$
(34)
Then, from Eqs. (25) and (26), the function $`F(\tau ,|y|)`$ is determined to $`F(\tau ,|y|)=ke^B|y|`$ and so
$`N(\tau ,|y|)`$ $`=`$ $`ke^B|y|kb_0|y|`$ (35)
$`A(\tau ,|y|)`$ $`=`$ $`kb_0|y|+J(\tau )kb_0|y|+{\displaystyle ^\tau }H(t)𝑑t,`$ (36)
where the interval scale $`b_0`$ is a small constant and $`H(\tau )`$ is a time dependent arbitrary function but may be determined by the equation of state. Thus the metric is read off as
$$ds^2=e^{2kb_0|y|}\left(d\tau ^2+e^{2^\tau H(t)𝑑t}d\stackrel{}{x}^2\right)+b_0^2dy^2.$$
(37)
Note that the metric is the same as that of the Randall-Sundrum except for the factor $`e^{2^\tau H(t)𝑑t}`$. Then Eqs. (31), (32) and (33) become
$`\rho (\tau ,|y|)+\eta \mathrm{\Lambda }_b`$ $`=`$ $`3e^{2kb_0|y|}H^2(\tau )`$ (38)
$`P(\tau ,|y|)\eta \mathrm{\Lambda }_b`$ $`=`$ $`e^{2kb_0|y|}\left[2\dot{H}(\tau )+3H^2(\tau )\right]`$ (39)
$`P_5(\tau ,|y|)\eta \mathrm{\Lambda }_b`$ $`=`$ $`3e^{2kb_0|y|}\left[\dot{H}(\tau )+2H^2(\tau )\right]`$ (40)
$`=`$ $`{\displaystyle \frac{1}{2}}\left[\rho (\tau ,|y|)3P(\tau ,|y|)\right]2\eta \mathrm{\Lambda }_b`$ (41)
$`=`$ $`{\displaystyle \frac{1}{2}}T_\mu ^{(2)\mu },`$ (42)
where $`\mu `$ runs through 0, 1, 2, 3. \[$`B(\tau ,|y|)`$ is associated with the vacuum expectation value of a massless four-dimensional scalar field.\] The above equations, Eqs. (38), (39) and (41), show that due to the exponential factor matter in the bulk is accumulated mainly near the B2 brane (negative tension brane), which is a similar result to those of the Refs., which are the exact solutions of bulk scalar, gauge boson and fermion fields to their field equations under the fixed RS static background geometry.
Of course, any $`H(\tau )`$ with $`H(\tau )0`$ and $`\dot{H}(\tau )0`$ as $`\tau \mathrm{}`$ can lead to an exit from an inflationary phase to a static Randall-Sundrum ($`k0`$) or Minkowski ($`k=0`$) universe. In this paper, however, we will not specify a model because we are more interested in the real expanding universe.
In Eqs.(38), (39) and (41), we should remember that $`\rho +\eta \mathrm{\Lambda }_b`$, $`P\eta \mathrm{\Lambda }_b`$ and $`P_5\eta \mathrm{\Lambda }_b`$ are non-trivial components of the 5 dimensional energy-momentum tensor. To derive effective 4 dimensional energy-momentum tensor $`\stackrel{~}{T}_B^A`$, it is necessary to consider the definition of the original 5 dimensional energy-momentum tensor,
$$\delta S_{\mathrm{matter}}=d^5x\sqrt{g}\delta (g_A^B)T_B^{(2)A}=d^4x𝑑yb_0\sqrt{g_4}\delta (\delta _A^B)T_B^{(2)A},$$
(43)
where $`g_4\mathrm{det}[g_{\mu \nu }]`$ ($`\mu ,\nu =0,1,2,3`$). As the 4 dimensional metric at a 4 dimensional slice in the bulk is $`\stackrel{~}{g}_{\mu \nu }=e^{2kb_0|y|}g_{\mu \nu }`$, which was introduced by Randall and Sundrum to solve the gauge hierarchy problem , the effective 4 dimensional energy-momentum tensor is given as
$`\stackrel{~}{T}_\nu ^\mu `$ $`=`$ $`b_0{\displaystyle 𝑑ye^{4kb_0|y|}T_\nu ^{(2)\mu }}`$ (44)
$`=`$ $`b_0{\displaystyle 𝑑ye^{2kb_0|y|}}`$ (46)
$`\mathrm{diag}[3H^2(\tau ),2\dot{H}(\tau )+3H^2(\tau ),2\dot{H}(\tau )+3H^2(\tau ),2\dot{H}(\tau )+3H^2(\tau )]`$
$``$ $`\mathrm{diag}[(\stackrel{~}{\rho }(\tau )+\eta \stackrel{~}{\mathrm{\Lambda }}),\stackrel{~}{P}(\tau )\eta \stackrel{~}{\mathrm{\Lambda }},\stackrel{~}{P}(\tau )\eta \stackrel{~}{\mathrm{\Lambda }},\stackrel{~}{P}(\tau )\eta \stackrel{~}{\mathrm{\Lambda }}].`$ (47)
According to RS, $`b_0𝑑ye^{2kb_0|y|}`$ is nothing but the induced 4 dimensional Planck scale $`M_{Pl}^2`$ . Thus, from Eqs.(38), (39), (46) and (47), we can get the Friedmann equations,
$`\left[{\displaystyle \frac{\dot{a}(\tau )}{a(\tau )}}\right]^2`$ $`=`$ $`{\displaystyle \frac{e^{2kb_0|y|}}{3}}\left[\rho (\tau ,|y|)+\eta \mathrm{\Lambda }_b\right]={\displaystyle \frac{1}{3M_{Pl}^2}}\left[\stackrel{~}{\rho }(\tau )+\eta \stackrel{~}{\mathrm{\Lambda }}\right]\mathrm{and}`$ (48)
$`{\displaystyle \frac{\ddot{a}(\tau )}{a(\tau )}}`$ $`=`$ $`{\displaystyle \frac{e^{2kb_0|y|}}{6}}\left[\rho (\tau ,|y|)+3P(\tau ,|y|)2\eta \mathrm{\Lambda }_b\right]={\displaystyle \frac{1}{6M_{Pl}^2}}\left[\stackrel{~}{\rho }(\tau )+3\stackrel{~}{P}(\tau )2\eta \stackrel{~}{\mathrm{\Lambda }}\right],`$ (49)
where $`a(\tau )`$ is a scale factor of our three dimensional space, $`a(\tau )e^{^\tau H(t)𝑑t}`$. Therefore, we could have saved the whole standard cosmological scenario by only requiring stabilization of $`B`$ in our framework (i.e. $`\dot{B}=0`$). Thus, we do not have to require additional conditions except for $`\dot{B}=0`$ such as the bulk to have non-zero cosmological constant or the visible brane to have a positive tension, which were essential in other models .
As an example, let us consider the vacuum dominated era. The equation of state is $`P=\rho `$ and then $`P_5`$ is given by $`P_5=2\rho \eta \mathrm{\Lambda }_b`$. Then Eqs. (38) and (39) lead to
$$H=constantH_0,$$
(50)
which gives the inflationary universe. $`H_0`$ can be considered as a parameter representing the degree of fine-tuning as given in Eq. (38). If $`H_0`$ vanishes, the fine-tuning is successful and the universe is static. On the other hand, a non-zero $`H_0`$ does not satisfy the fine-tuning condition and gives rise to an inflationary universe. Note that our solution is for $`\dot{B}=0`$ and any modulus potential is not generated at the classical level since $`b_0`$ is arbitrary.
We proceed to discuss the possibility of stabilizing the size of the fifth dimension in our framework. Since we obtained already the solutions for the $`\dot{B}=0`$ case, we conclude that if $`\dot{B}`$ goes to zero asymptotically, there exist solutions converging asymptotically to our $`\dot{B}=0`$ solutions.
We suppose that the stabilization era occurs before or during the conventional inflation era. Thus, we suppose that the system is governed by the same equations of state as those of the inflation era discussed above. Then, from Eqs. (31), (32) and (33), we obtain
$`2\ddot{A}+\ddot{B}+\dot{B}^2\dot{A}\dot{B}`$ $`2\dot{A}\dot{N}\dot{B}\dot{N}=0`$ (51)
$`\ddot{A}\dot{A}\dot{N}`$ $`2\dot{A}\dot{B}=0,`$ (52)
which are the equations of state in the inflationary era. The above equations and Eqs. (25) and (26) convince us that any potential for the modulus field is not generated also since the $`B`$ is not fixed yet.
Eq. (52) is easily solved,
$$\dot{A}=e^{s(|y|)+N}b^2=e^{s(|y|)+kF}b^2\left(=k\dot{F}+\dot{J}\right),$$
(53)
where $`s(|y|)`$ is an integration constant and $`b`$ is defined as $`b(\tau ,|y|)e^{B(\tau ,|y|)}`$. Note that $`A`$ is an increasing function of time since $`\dot{A}>0`$. Removing $`\ddot{A}`$ from Eqs. (51) and (52), we obtain
$$\ddot{B}+\dot{B}^2+3\dot{A}\dot{B}\dot{B}\dot{N}=0,$$
(54)
which can be solved to give
$$\dot{b}=\frac{1}{e^{t(|y|)N+3A}}=\frac{1}{e^{t(|y|)+2kF+3J}},$$
(55)
where $`t(|y|)`$ is an integration constant. Note that $`b`$ is an increasing function of time but $`\dot{b}`$ could be zero asymptotically. The extra dimension scale $`b`$ can be stabilized if the combination $`N+3A(=2kF+3J)`$ is an increasing function of time without limit. Especially, if $`NA`$ and $`A`$ increases as $`\tau \mathrm{}`$ without a limit at least with a power law, which inflates the three dimensional space, then $`\dot{b}`$ would decrease exponentially and so $`b`$ could be stabilized. Below we show that it is possible.
From Eqs. (53) and (55), we obtain
$$\frac{\dot{A}}{\dot{b}}=e^{s(|y|)+t(|y|)+3A}b^2,$$
(56)
which is integrable. The solution is
$$b(\tau ,|y|)^3=u^3(|y|)e^{s(|y|)t(|y|)3A},$$
(57)
where $`u(|y|)`$ is a $`|y|`$ dependent arbitrary function. The remaining constraints to satisfy are only Eqs. (25) and (26), i.e. $`A^{}=N^{}=kb(\tau ,|y|)sgn(y)`$, from which Eq. (57) can be written as
$$3\times \frac{u^2u^{}b^2b^{}}{u^3(|y|)b^3}+s^{}(|y|)+t^{}(|y|)=3A^{}=3kbsgn(y).$$
(58)
We intend to make our solution become the inflationary solution asymptotically that was obtained before. Since $`Nkb_0|y|`$, $`\dot{A}H_0`$, and $`bb_0`$ as $`\tau \mathrm{}`$, let us take the integration constants in Eqs. (53) and (57) as
$`s(|y|)`$ $`=`$ $`kb_0|y|+\mathrm{ln}H_02\mathrm{ln}b_0`$ (59)
$`u(|y|)`$ $`=`$ $`b_0,`$ (60)
where $`b_0`$ and $`H_0`$ were defined above. So far the solutions were exact. Note that any value for $`H_0`$ is possible, which does not influence Eq. (58). Although Eq. (58) is difficult to solve exactly, we can argue that $`|N|=k|F|\{J(\tau )\mathrm{and}H_0\tau \}`$ is sufficient to draw a meaningful conclusion.
$`J(\tau )`$ and $`H_0`$ can be chosen always such that $`|N|=k|F|\{J(\tau )\mathrm{and}H_0\tau \}`$. It is compatible with the phenomenological requirement of a large $`H_0`$ so that our solutions are derived during the inflationary epoch. (Actually during the inflationary era, $`H_0^1`$ is (a very large value)<sup>-1</sup> $`10^{34}`$ sec.) In this early inflationary era, $`N`$ and $`b`$ are regarded as being much smaller than $`H_0\tau `$, since the scale of the universe was small right after the Big-Bang. Then $`N+3A`$ increases with time and the extra dimension scale $`b`$ is stabilized rapidly to $`b_0`$. Once $`b`$ is stabilized, the conditions Eq. (34) become valid and $`N`$ and $`\dot{A}`$ are forced to $`kb_0|y|`$ and $`H_0`$, respectively.
To see our argument explicitly, let us take somewhat large $`J(\tau )`$ assumption and solve Eqs. (53), (55) and (58) approximately under the condition $`k|F|J`$. Our aim is to show $`\dot{A}H_0`$, $`\dot{b}0`$, and $`bb_0`$. Here, we set $`kF`$ and $`b`$ as $`O(1)`$ initially, and then $`J1`$. Then Eq. (57) gives
$$b_0^3b^3(\tau ,|y|)=\frac{H_0}{b_0^2}e^{kb_0|y|3kF}e^{3J}1.$$
(61)
With Eqs. (58) and (61) we are led to the following results,
$$b^{}1\mathrm{or}bb(\tau ),$$
(62)
and we obtain
$$F(\tau ,|y|)b(\tau )|y|.$$
(63)
Here we must set $`t(|y|)0`$ in view of Eq. (55). Then, Eqs. (53) and (55) become
$$\dot{J}H_0\frac{b^2}{b_0^2}\mathrm{and}\dot{b}e^{3J}1,$$
(64)
which shows that during the period of the three space inflation it is quite difficult for $`b`$ to be dynamical. It is interpreted as the stabilization of the extra dimension in spite of the flat potential for $`b`$. From Eq. (55) we can derive an expression for $`b(\tau )`$,
$$\frac{1}{6b_0^2}\mathrm{ln}\left[\frac{(b_0b)^2}{b_0^2+b_0b+b^2}\right]+\frac{1}{b_0^2\sqrt{3}}\mathrm{tan}^1\left[\frac{b_0+2b}{b_0\sqrt{3}}\right]=\frac{H_0}{b_0^2}\tau ,$$
(65)
or
$$b(\tau )b_0e^{3H_0\tau }.$$
(66)
Here we can see that as $`\tau \mathrm{}`$, $`b(\tau )`$ grows to $`b_0`$ and $`\dot{J}`$ tends to $`H_0`$ asymptotically. In other words, to obtain an inflationary universe $`b`$ should be stabilized to $`b_0`$ exponentially. Note that Eq. (65) becomes an exact result provided the warp factor vanishes, which corresponds the cases of $`\mathrm{\Lambda }_b=0`$ or $`\eta =1`$.
If $`b`$ is $`O(1)`$ but small right after the Big Bang, $`b_0`$ should be $`O(1)`$ but small. As the extra dimension gets stabilized ($`\dot{B}=0`$) soon after the beginning of the inflationary era, while the three dimensional space inflates (to $`e^{H_0\tau }`$), $`b`$ remains small ($`1/M`$) and the universe is reduced effectively to 4-dimension.
In a similar method, we can show that $`\dot{b}`$ is made asymptotically to zero also in the radiation ($`P=\rho /3`$, $`\eta =0`$) and matter dominated era ($`P=0`$, $`\eta =0`$). However, as the initial condition for $`\dot{b}`$ is zero (through the above solution in the inflationary epoch), $`b`$ should have been stabilized already and so Eq. (34) should have been valid since the beginning of the radiation dominated era. Therefore, the Friedmann equations Eqs. (48) and (40) hold good in the radiation and matter dominated eras.
In conclusion, we have provided exact cosmological solutions in the RS setup with bulk matter. In the static limit of all components of the metric, the solutions become the RS metric and in static limit of the extra dimension, they are reduced to the standard Friedmann equations, which implies that bulk matter is accumulated mainly near the negative tension brane (visible brane B2). In this case the modulus potential is not generated effectively at the classical level. With our solution, however, we have shown that the extra dimension could be stabilized (the $`\dot{B}=0`$ solution) even if the modulus potensial is flat ($`b_0`$ is arbitrary) and it should be small since the three dimensional space inflates during the inflationary era.
###### Acknowledgements.
This work is supported in part by the BK21 program of Ministry of Education.
|
warning/0005/math0005102.html
|
ar5iv
|
text
|
# Good representations and solvable groups
## 1. Introduction
The purpose of this paper is to provide a characterization of solvable linear algebraic groups in terms of a geometric property of representations. Representations with a related property played an important role in the proof of the equivariant Riemann-Roch theorem \[EG2\]. In that paper, we constructed representations with that property (which we call freely good) for the group of upper triangular matrices in $`GL_n`$. We noted that it seemed unlikely that such representations exist for arbitrary groups; the main result of this paper implies that they do not.
To state our results, we need some definitions. A representation $`V`$ of a linear algebraic group $`G`$ is said to be good (resp. freely good) if there exists a non-empty $`G`$-invariant open subset $`UV`$ such that
(i) $`G`$ acts properly (resp. freely) on $`U`$.
(ii) $`VU`$ is the union of a finite number of $`G`$-invariant linear subspaces.
Note that freely good representations were called good in \[EG2\].
The main result of the paper is the following theorem.
###### Theorem 1.1.
Let $`G`$ be a connected algebraic group over a field $`k`$ of characteristic not equal to $`2`$. Then $`G`$ is solvable if and only if $`G`$ has a good representation. Moreover, if $`G`$ is solvable and $`k`$ is perfect then $`G`$ has a freely good representation.
In characteristic $`2`$, a solvable group still has good representations, and a partial converse holds (Corollary 4.1). A key step in the proof of the main result is Theorem 4.1, which is inspired by an example of Mumford \[GIT, Example 0.4\].
In characteristic 0 solvable groups are characterized by a weaker property which does not require the action to be proper. (In general, if $`G`$ acts properly on $`X`$, then $`G`$ acts with finite stabilizers on $`X`$, but the converse need not hold.)
###### Theorem 1.2.
Let $`G`$ be a connected algebraic group over a field of characteristic $`0`$. Suppose that $`G`$ has a representation $`V`$ that contains a nonempty open set $`U`$ such that
$`(1)`$ The complement of $`U`$ is a finite union of invariant linear subspaces, and
$`(2)`$ $`G`$ acts with finite stabilizers on $`U`$.
Then $`G`$ is solvable.
Examples (see Section 6) show that this weaker property does not characterize solvability in positive characteristic.
## 2. Preliminaries
Groups and representations We let $`k`$ denote a field, with algebraic closure $`\overline{k}`$ and separable closure $`k_s`$. If $`Z`$ is a $`k`$-variety and $`k^{}k`$ is any extension of $`k`$ then $`Z(k^{})`$ denotes the $`k^{}`$-valued points of $`Z`$, while $`Z_k^{}`$ denotes the $`k^{}`$-variety $`Z\times _kk^{}`$.
All groups in this paper are assumed to be linear algebraic groups over a field $`k`$. We assume that such a group $`G`$ is geometrically reduced, that is, that $`G_{\overline{k}}`$ is reduced. The identity component of a group $`G`$ is denoted $`G^0`$.
Unless otherwise stated, a representation $`V`$ of a group $`G`$ is assumed to be $`k`$-rational; i.e. $`V`$ is a $`k`$-vector space and the action map $`G\times VV`$ is a morphism of $`k`$-varieties.
If $`k^{}k`$ is a field extension we call a $`k^{}`$-rational representation $`V`$ of $`G_k^{}`$ a $`k^{}`$-representation of $`G`$. We say that $`V`$ is defined over $`k`$ if it is obtained by base change from a $`k`$-rational representation.
If $`k^{}k`$ be a Galois field extension, then $`\mathrm{Gal}(k^{}/k)`$ acts on $`k^{}`$-representations of $`G`$. Indeed, let $`V`$ be a $`k^{}`$-representation of $`G`$ corresponding to a $`k^{}`$ morphism $`\rho :G_k^{}GL(V)`$. For $`gG(k_s)`$ $`{}_{}{}^{\sigma }\rho (g)`$ is defined as follows (cf. \[Borel, AG14.3, 24.5\]). Because $`\rho `$ is defined over $`k^{}`$, for any $`\tau \mathrm{Gal}(k_s/k^{})`$ and any $`gG(k_s)`$, $`\tau (\rho (\tau ^1(g)))=\rho (g)`$. Thus, if $`\sigma \mathrm{Gal}(k^{}/k)=\mathrm{Gal}(k_s/k)/\mathrm{Gal}(k^{}/k)`$, then
$$\sigma ^{}(\rho ((\sigma ^{})^1g))GL(V)(k_s)$$
is independent of the lift of $`\sigma `$ to an element $`\sigma ^{}\mathrm{Gal}(k_s/k)`$. We will call this point $`\sigma \rho (\sigma ^1(g))`$ and set $`\rho (g)=\sigma (\rho (\sigma ^1g))`$.
The $`k^{}`$-representation $`V`$ is obtained by base change from a representation defined over $`k`$ if and only if $`{}_{}{}^{\sigma }\rho =\rho `$ for all $`\sigma \mathrm{Gal}(k^{}/k)`$.
Free and proper actions The action of a group $`G`$ on a scheme $`X`$ is said to be free if the action map $`G\times XX\times X`$ is a closed embedding. The action is said to be proper if the map $`G\times XX\times X`$ is proper. If the action is proper then the stabilizer of every point is finite. If the stabilizer of every geometric point is a trivial group-scheme then we say that the action is set theoretically free. An action which is set theoretically free and proper is free \[EG1\].
Let $`HG`$ be a finite morphism of algebraic groups. If $`G`$ acts properly on a scheme $`X`$ then $`H`$ also acts properly on $`X`$. Thus, if $`V`$ is a good representation of $`G`$ then $`V`$ is also a good representation of $`H`$ via the action induced by the map $`HG`$. Moreover, if $`H`$ is a closed subgroup and $`V`$ is a freely good representation of $`G`$, then $`V`$ is a freely good representation of $`H`$.
###### Example 2.1.
Let $`B`$ be the group of upper triangular matrices in $`\mathrm{GL}(n)`$. The group $`B`$ acts by left multiplication on the vector space $`V`$ of upper triangular matrices; it acts with trivial stabilizers on the open subset $`U`$ of invertible upper triangular matrices. Since the matrices are upper triangular, $`VU`$ is the union of the invariant subspaces $`L_i=\{AV|A_{ii}=0\}`$. This representation is freely good because the action of $`B`$ on $`U`$ is identified with $`B`$ acting itself by left multiplication. The map $`B\times BB\times B`$ given by $`(A,A^{})(A,AA^{})`$ is an isomorphism, so the action of $`B`$ on $`U`$ is free.
By contrast, the action of $`\mathrm{GL}(n)`$ by left multiplication on the vector space $`M_n`$ of $`n\times n`$ matrices is not good.
## 3. Existence of good representations
In this section we show that every connected solvable group has good representations, and if $`k`$ is perfect, freely good representations.
By the Lie-Kolchin theorem $`G_{\overline{k}}`$ is trigonalizable; i.e., it can be embedded in the group $`B_{\overline{k}}\mathrm{GL}_n`$ of upper triangular matrices.
Let $`V_{\overline{k}}`$ be the vector space of upper triangular $`n\times n`$ matrices. The group $`B_{\overline{k}}`$ acts on $`V_{\overline{k}}`$ by left multiplication and we have seen that this representation is freely good. By restriction $`V_{\overline{k}}`$ is a good representation of $`G_{\overline{k}}`$. Consider the morphism $`\rho :G_{\overline{k}}\mathrm{GL}(V_{\overline{k}})`$ corresponding to the action of $`G_{\overline{k}}`$ on $`V_{\overline{k}}`$.
Since $`\rho `$ is a morphism of schemes of finite type, it is defined over a field extension $`k^{}k`$ of finite degree. Write $`V=V_k^{}`$ for the corresponding $`k^{}`$-representation; then we have $`\rho :G_k^{}\mathrm{GL}(V)`$.
Case I. $`k^{}`$ is separable over $`k`$. (This will occur when $`k`$ is perfect.) In this case we will use Galois descent to construct a freely good representation of $`G`$.
Replacing $`k^{}`$ by a possibly bigger field extension we may assume that $`k^{}k`$ is Galois. Enumerate the elements of $`\mathrm{Gal}(k^{}/k)`$ as $`\{1=\sigma _1,\sigma _2,\mathrm{}\sigma _d\}`$ and consider the representation $`\mathrm{\Phi }:G_k^{}\mathrm{GL}(V^d)`$ where $`G_k^{}`$ acts on the $`j`$-th factor by the representation $`{}_{}{}^{\sigma _j}\rho :G_k^{}\mathrm{GL}(V)`$.
We define $`U_dV^d`$ to be the open set whose $`k_s`$-rational points are the $`d`$-tuples $`(A_1,\mathrm{},A_d)`$ where some $`A_i`$ is invertible. We realize $`U_d`$ as a complement of $`G_k^{}`$-invariant linear subspaces as follows. Let $`L_j=\{AV|A_{jj}=0\}`$, a $`G_k^{}`$-invariant subspace of $`V`$. Given a $`d`$-tuple $`(j_1,\mathrm{},j_d)`$, define
$$L_{(j_1,\mathrm{},j_d)}=L_{j_1}\mathrm{}L_{j_d}.$$
This is a $`G_k^{}`$-invariant subspace of $`V^d`$ and $`U_d=V_dL_{(j_1,\mathrm{},j_d)}`$.
###### Lemma 3.1.
(cf. \[EG2, Theorem 2.2\]) $`G_k^{}`$ acts freely on $`U_d`$.
###### Proof.
Since $`G_k^{}`$ is a closed subgroup of $`B_k^{}`$ and the open set $`U_d`$ is $`B_k^{}`$ invariant, it suffices to show that $`B_k^{}`$ acts freely on $`U_d`$.
To do this we must show that the map $`B_k^{}\times U_dU_d\times U_d`$ given on $`k_s`$ points by
$$(A,A_1,\mathrm{},A_d)(AA_1,\sigma _2(A\sigma _2^1(A_2)),\mathrm{},\sigma _d(A\sigma _d^1(A_d))).$$
is a closed embedding.
First we show that the image $`Z`$ of $`B_k^{}\times U_d`$ is closed in $`U_d\times U_d`$. Let $`(A_1,A_2,\mathrm{}A_d,C_1,\mathrm{},C_d)`$ be matrix coordinates on $`U_d\times U_d`$. Expanding the inverse out in terms of the adjoint we see that the image is contained in the subvariety defined by the matrix equations
$$\sigma _j\sigma _i^1(detA_i)C_j=\sigma _j\sigma _i^1(C_i\text{ }\text{Adj }A_i)A_j.$$
Suppose that a 2d-tuple of matrices $`(A_1,A_2,\mathrm{},A_d,C_1,C_2,\mathrm{}C_d)U_d\times U_d`$ satisfies the matrix equations above. At least one of the $`A_i`$ and one of the $`C_j`$ is invertible because we are in $`U_d\times U_d`$. Let $`A=\sigma _i^1(C_iA_i^1)`$. Substituting into our equations we see that $`C_l=\sigma _l(A\sigma _l^1(A_l))`$ for all $`l`$. Moreover $`A`$ is invertible since $`C_j`$ is invertible and $`C_j=\sigma _j(A\sigma _j^1(A_j))`$. Hence every point satisfying the matrix equations is in the image $`Z`$ of $`B_k^{}\times U_d`$, so $`Z`$ is closed.
The variety $`Z`$ is covered by open sets of the form
$$\{(A_1,A_2,\mathrm{}A_j,\mathrm{}A_d,AA_1,\mathrm{},\mathrm{}AA_j,\mathrm{}A_d)|detA_j0\}.$$
These open sets are isomorphic to $`V^{d1}\times B_k^{}`$, where $`V^{d1}`$ is the $`d1`$-fold cartesian product of $`V`$. Hence the image is smooth, in particular, normal. The action of $`G_k^{}`$ on $`U_d`$ is set-theoretically free so $`G_k^{}\times U_dZ`$ is a birational bijection. By Zariski’s main theorem (cf. \[Borel, AG18\]) a birational bijection of a normal varieties is an isomorphism, so $`G_k^{}\times U_dZ`$ is an isomorphism. Therefore, $`G_k^{}\times U_dU_d\times U_d`$ is a closed embedding. $`\mathrm{}`$
Remark. The proof of \[EG2, Theorem 2.2\] is incomplete; the last paragraph of the above argument is needed.
For any basis of $`V`$, there is a natural choice of basis so that with respect to this basis, if $`gG(k_s)`$, $`\mathrm{\Phi }(g)`$ is represented by the block diagonal matrix
$$\left[\begin{array}{ccccc}\rho (g)& & & & \\ & {}_{}{}^{\sigma _2}\rho (g)& & & \\ & & ..& & \\ & & & ..& \\ & & & & {}_{}{}^{\sigma _d}\rho (g)\end{array}\right]$$
This representation is not defined over $`k`$ because the Galois group acts by permuting the blocks. More precisely, we have the following. Given a $`d\times d`$ matrix $`M`$, let $`M[n]`$ denote the $`nd\times nd`$ matrix whose $`ij`$ block is $`M_{ij}I_n`$, where $`I_n`$ is the $`n\times n`$ identity matrix. If $`\sigma \mathrm{Gal}(k^{}/k)`$, let $`J_\sigma `$ denote the permutation matrix corresponding to the permutation $`\sigma _i\sigma \sigma _i`$. In matrix form, for $`gG(k_s)`$,
$${}_{}{}^{\sigma }\mathrm{\Phi }(g)=J_\sigma [n]^1\mathrm{\Phi }(g)J_\sigma [n].$$
We will show that $`\mathrm{\Phi }`$ is $`k^{}`$-isomorphic to a freely good representation defined over $`k`$. Choose a primitive element $`\alpha `$ for the extension $`k^{}k`$, and let $`A`$ be the $`d\times d`$ matrix with $`A_{ij}=\sigma _j(\alpha ^i)`$. $`A`$ is invertible since $`\alpha ,\sigma _2(\alpha ),\mathrm{}\sigma _d(\alpha )`$ are exactly the roots of the irreducible polynomial $`fk[x]`$ of $`\alpha `$ over $`k`$, so $`detA=_{i<j}(1)^d(\sigma _i(\alpha )\sigma _j(\alpha ))0`$. The Galois group acts by $`\sigma (A)=AJ_\sigma `$. Consider the morphism $`\mathrm{\Psi }:G_k^{}\mathrm{GL}(V^d)`$ defined by by $`\mathrm{\Psi }(g)=A[n]\mathrm{\Phi }(g)A[n]^1`$ for $`gk_s`$ . Then $`{}_{}{}^{\sigma }\mathrm{\Psi }(g)=\mathrm{\Psi }(g)`$ for any $`\sigma \mathrm{Gal}(k^{}/k)`$; hence $`\mathrm{\Psi }`$ is defined over $`k`$. Each of the subspaces $`L_{(j_1,\mathrm{},j_d)}`$ is $`G_k^{}`$-invariant under the action $`\mathrm{\Psi }`$. Moreover, because each $`L_{(j_1,\mathrm{},j_d)}`$ is a vector subspace of $`V^d`$ defined over $`k`$, the corresponding sub-representations $`G_k^{}\mathrm{GL}(L_{(j_1,\mathrm{},j_d)})`$ are also defined over $`k`$. Therefore $`\mathrm{\Psi }`$ is obtained by base change from a freely good representation of $`G`$.
Case II. The general case In this case we may assume that there is a freely good $`k^{}`$-rational representation $`\rho :G_k^{}\mathrm{GL}(V)`$ defined over a finite normal extension $`k^{}`$ of $`k`$. Then $`k^{}k`$ factors as $`k^{}k^{\prime \prime }k`$, with $`k^{}/k^{\prime \prime }`$ purely inseparable of degree $`p^n`$ and $`k^{\prime \prime }/k`$ Galois. The Frobenius endomorphism on $`V`$ induces a group homomorphism of $`\mathrm{GL}(V)`$. Composing $`\rho `$ with the $`n`$-th power of Frobenius on $`\mathrm{GL}(V)`$ we obtain a representation defined over $`k^{\prime \prime }`$. Because the Frobenius has finite kernel, this representation will no longer be faithful. However, the action of Frobenius is trivial on geometric points, so $`G`$ will act properly on an open set whose complement is a union of linear subspaces. We can now use the Galois descent argument of Case I to obtain a good $`k`$-rational representation of $`G`$.
## 4. Characterization of solvable groups by good representations
In this section we show that if $`\text{char }k2`$, every reductive group with a good representation is a torus. However, many of the results of this section are valid in arbitrary characteristic, and we only need that $`\text{char }k2`$ in part of the proof of Theorem 1.1. We will explicitly say when we start assuming this; until then $`\text{char }k`$ is arbitrary.
Let $`T`$ be the diagonal torus in $`SL_2`$ and $`N(T)`$ the normalizer of $`T`$. We will first show that $`N(T)`$ has no good representations. We begin by recalling some facts about $`N(T)`$. Let
$$J=\left[\begin{array}{cc}0& 1\\ 1& 0\end{array}\right].$$
We will also write
$$H(t)=\left[\begin{array}{cc}t& 0\\ 0& t^1\end{array}\right].$$
The group $`N(T)`$ is generated by $`T`$ and $`J`$; it has two components, $`T`$ and $`J(T)`$. The action of $`SL_2`$ on its two dimensional standard representation $`V`$ induces an action on $`S(V^{})k[x,y]`$, given by
$$\left[\begin{array}{cc}a& b\\ c& d\end{array}\right]:yaycx,xby+dx.$$
Let $`W_i`$ denote the subspace of $`S(V^{})`$ spanned by $`x^i`$ and $`y^i`$; this is an irreducible representation of $`N(T)`$, of dimension $`2`$ (if $`i>0`$). Let $`W_0^{}`$ denote the $`1`$-dimensional irreducible representation of $`N(T)`$ on which $`T`$ acts trivially and $`J`$ acts by multiplication by $`1`$.
If $`\text{char }k2`$, the group $`N(T)`$ is linearly reductive; that is, its action on any representation is completely reducible \[GIT, p.191\]. The next lemma shows that much of this survives in arbitrary characteristic.
###### Lemma 4.1.
Let $`V`$ be a representation of $`N(T)`$.
$`(1)`$ As a representation of $`N(T)`$, $`V`$ splits as a direct sum of $`N(T)`$-submodules:
$$V=V_0\underset{i>0}{}V_{\pm i}.$$
Here $`V_{\pm i}`$ is the sum of the $`i`$ and $`i`$ weight spaces of $`T`$ on $`V`$, and $`V_j`$ is the $`j`$-weight space.
$`(2)`$ The action of $`N(T)`$ on $`V_{\pm i}`$ $`(i>0)`$ is completely reducible, and $`V_{\pm i}`$ is isomorphic as $`N(T)`$-module to a direct sum of copies of $`W_i`$.
$`(3)`$ If $`\text{char }k2`$, then $`V_0`$ is isomorphic to a direct sum of copies of $`W_0`$ and $`W_0^{}`$.
###### Proof.
(1) Because the action of $`T`$ on $`V`$ is completely reducible, we can decompose $`V=V_i`$ as $`T`$-module. As $`JH(t)J^1=H(t^1)`$, we have $`JV_i=V_i`$. Hence $`V_{\pm i}`$ is an $`N(T)`$-submodule and we get the desired direct sum decomposition of $`V`$.
(2) Let $`v_1,\mathrm{},v_d`$ be a basis for $`V_i`$ $`(i>0)`$. The map $`v_ry,Jv_rx`$ defines an isomorphism of the span of $`v_r,Jv_r`$ (denoted $`v_r,Jv_r`$) with $`W_i`$, and the map $`W_i^dV_{\pm i}`$, taking the $`r`$-th component to $`v_r,Jv_r`$, is an $`N(T)`$-module isomorphism.
(3) Decompose the $`0`$-weight space of $`V`$ into the $`+1`$ and $`1`$ eigenspaces of $`J`$; these are isomorphic to sums of copies of $`W_0`$ and $`W_0^{}`$, respectively. $`\mathrm{}`$
The proof of the following result was motivated by \[GIT, Example 0.4\].
###### Theorem 4.1.
The group $`N(T)`$ has no good representations.
###### Proof.
If a group $`G`$ has good representations, then so does $`G_{\overline{k}}`$, so we may assume that $`k`$ is algebraically closed. Suppose that $`V`$ is a representation of $`N(T)`$, and let $`UV`$ be the complement of a finite set of invariant linear subspaces $`𝒮`$. We will show that $`N(T)`$ does not act properly on $`U=V_{L𝒮}L`$. The strategy of the proof is as follows. Consider the action map $`\mathrm{\Phi }:N(T)\times UU\times U`$. We will find a closed subvariety $`Z`$ of $`N(T)\times U`$ whose closed points are of the form
$$(\left[\begin{array}{cc}0& \lambda ^1\\ \lambda & 0\end{array}\right],v_\lambda )$$
whose image is not closed in $`U\times U`$. Hence $`\mathrm{\Phi }`$ is not proper, so the representation is not good.
We now carry out the proof. Decompose $`V=V_i`$, where $`V_i`$ is the $`i`$-weight space of $`V`$ for $`T`$. Pick $`uU`$, and write $`u=u_i`$ where $`u_iV`$. Some of the $`u_i`$ may be $`0`$; let $`d`$ be the dimension of the space spanned by the nonzero $`u_i`$.
Step 1. If $`a_i0`$ for all $`i`$ with $`u_i0`$, then $`w=a_iu_iU`$. Indeed, suppose not; then $`wL`$ for some $`L𝒮`$. For almost all choices $`t_1,\mathrm{},t_d`$ of $`d`$ elements of $`k^{}`$, the vectors
$$H(t_q)w=\underset{p}{}t_q^{i_p}a_{i_p}u_{i_p}L$$
are linearly independent. (Here $`i_1,\mathrm{},i_d`$ are the indices $`i_p`$ with $`u_{i_p}0`$.) This follows because the $`d\times d`$ matrix $`A`$ with entries
$$A_{pq}=t_q^{i_p}$$
is nonsingular for almost all $`t_1,\mathrm{},t_d`$. (This is because $`detA`$ is a sum of monomials, where each monomial is a product of one term from each row and each column. Each monomial has different multi-degree, so $`detA`$ is not the zero polynomial.) Therefore, the vectors $`H(t_q)w`$ span the same space as the $`u_i`$, so $`uL`$, contradicting our assumption that $`uU`$. We conclude that $`wU`$, as claimed.
A similar argument shows that $`Ju_0+_{i0}u_iU`$.
Step 2. There exists an element $`u^{}=u_i^{}U`$ with $`Ju_i^{}=u_i^{}`$ for all $`i>0`$. To see this, suppose $`u_jJu_j`$ for some $`j>0`$. Let $`W_jV_jV_j`$ be the subspace of vectors of the form $`v_j+Jv_j`$ $`(v_jV_j)`$. Note that $`W_j`$ generates $`V_jV_j`$ as $`N(T)`$-module. Consider the affine linear subspace
$$B=\underset{i\pm j}{}u_i+W_j.$$
We claim that $`BU`$ is nonempty. If it is empty, then because $`B`$ is affine linear and is contained in a finite union of the subspaces in $`𝒮`$, we see that $`BL`$ for some $`L𝒮`$. But then the span of $`B`$ is contained in $`L`$, so $`_{i\pm j}u_iL`$ and $`W_jL`$. Because $`L`$ is $`N(T)`$-stable, $`V_jV_jL`$ as well; so
$$u\underset{i\pm j}{}u_i+(V_jV_j)L,$$
contradicting $`uU`$. We conclude that $`BU`$ is nonempty. Replacing $`u`$ by an element of $`BU`$, which we again call $`u`$, we do not change $`u_i`$ for $`i\pm j`$, but we obtain $`u_j=Ju_j`$. Iterating this process, we obtain $`u^{}`$ of the desired form.
Replacing $`u`$ by $`u^{}`$, we will assume that $`Ju_i=u_i`$ for all $`i>0`$. From the $`N(T)`$-module isomorphism of $`u_i,Ju_i`$ with $`x^i,y^i`$, we see that for $`i>0`$,
$$\left[\begin{array}{cc}0& \lambda ^1\\ \lambda & 0\end{array}\right]:u_i\lambda ^iu_i,u_i(1)^i\lambda ^iu_i.$$
Note also that
$$\left[\begin{array}{cc}0& \lambda ^1\\ \lambda & 0\end{array}\right]u_0=Ju_0.$$
Step 3. Define
$$v_\lambda =u_0+\underset{i>0}{}(u_i+\lambda ^iu_i)$$
$$v_\lambda ^{}=Ju_0+\underset{i>0}{}(u_i+(1)^i\lambda ^iu_i).$$
For all $`\lambda 0`$, both $`v_\lambda `$ and $`v_\lambda ^{}`$ are in $`U`$ (by Step 1). Define $`Z`$ to be the closed subvariety of $`N(T)\times U`$ whose points are the pairs
$$(\left[\begin{array}{cc}0& \lambda ^1\\ \lambda & 0\end{array}\right],v_\lambda ).$$
Then
$$\mathrm{\Phi }(Z)=\{(v_\lambda ,v_\lambda ^{})\}_{\lambda 0}U\times U.$$
Consider the point
$$(v,v^{})=(u_0+\underset{i>0}{}u_i,Ju_0+\underset{i>0}{}u_i).$$
Reasoning as in Step 1 shows that $`u`$ is in the $`N(T)`$-module generated by $`v`$ or $`v^{}`$, so if either $`v`$ or $`v^{}`$ were in $`L`$ then $`u`$ would be, but this is impossible as $`uU`$. Hence $`v`$ and $`v^{}`$ are in $`U`$, so $`(v,v^{})U\times U`$. Also, $`(v,v^{})`$ is not in $`\mathrm{\Phi }(Z)`$, but is in the closure of $`\mathrm{\Phi }(Z)`$ in $`U\times U`$. We conclude that $`\mathrm{\Phi }`$ is not proper, so the representation is not good. $`\mathrm{}`$
### 4.1. Proof of Theorem 1.1
Let $`G`$ be a connected nonsolvable linear algebraic group. Consider the surjective map $`\pi :GG_1=G/_uG`$, where $`_uG`$ is the unipotent radical of $`G`$ and $`G_1`$ is reductive. Because $`G`$ is not solvable, $`G_1`$ is not trivial or a torus. Let $`T`$ be a maximal torus of $`G`$. Then $`T_1=\pi (T)`$ is a maximal torus of $`G_1`$, and $`\pi `$ induces an isomorphism of Weyl groups $`W(T,G)W(T_1,G_1)`$ \[Borel, 11.20\]. (Here $`W(T,G)=N_G(T)/Z_G(T)`$ where $`N_G`$ and $`Z_G`$ denote normalizer and centralizer of $`T`$ in $`G`$, and similarly for $`G_1`$.) Because $`\mathrm{ker}\pi `$ is a unipotent group, $`\pi |_T:TT_1`$ is an isomorphism. Note that $`Z_G(T)=T(_uG)^T`$ \[Borel, 13.17\]. Because $`G_1`$ is reductive, this fact (applied to $`G_1`$) implies that $`Z_{G_1}(T_1)=T_1`$. Moreover, any $`g_1N_{G_1}(T_1)`$ can be lifted to $`gN_G(T)`$. This follows because the isomorphism of Weyl groups above, and the structure of the centralizers, imply that each component of $`N_{G_1}(T_1)`$ is the image of a surjective map of a component of $`N_G(T)`$.
As $`G_1`$ is not a torus, there is a root $`\alpha `$ and a homomorphism $`\varphi _\alpha :\mathrm{SL}_2G_1`$ with kernel either trivial, or the set of matrices $`\left[\begin{array}{cc}a& 0\\ 0& a\end{array}\right]`$ with $`a^2=1`$. Moreover (using the subscript $`\mathrm{SL}_2`$ to denote terms for $`\mathrm{SL}_2`$ defined in the previous subsection), $`\varphi _\alpha (T_{\mathrm{SL}_2})T`$ and $`J_1:=\varphi _\alpha (J_{\mathrm{SL}_2})N_{G_1}(T_1)`$. (See \[J, p.176\] for these facts.) Let $`H_1=\varphi _\alpha (N(T_{\mathrm{SL}_2}))`$; its identity component $`H_1^0=\varphi _\alpha (T_{\mathrm{SL}_2})T_1`$. Because $`H_1`$ is a finite image of $`N(T_{\mathrm{SL}_2})`$, it has no good representations (and hence neither does $`G_1`$).
Up to this point, $`\text{char }k`$ has been arbitrary; now we assume that $`\text{char }k2`$.
Because $`\pi |_T`$ is an isomorphism there is a unique subgroup $`H^0T`$ projecting isomorphically to $`H_1^0`$. As noted above, we can choose a lift $`JN_G(T)`$ of $`J_1N_{G_1}(T_1)`$. Write $`J=J_sJ_u`$ for the Jordan decomposition of $`J`$. Because $`\text{char }k2`$, $`J_1`$ is semisimple, so $`\pi (J_u)=1`$. Therefore we can replace $`J`$ by $`J_s`$ and assume $`J`$ is semisimple. Now, $`J^2`$ corresponds to the identity element in the Weyl group (as $`J_1^2`$ does), so $`J^2Z_G(T)=T(_uG)^T`$. Since $`J`$ is semisimple, we conclude that $`J^2T`$. As $`J_1^2`$ is in the subgroup $`H_1^0`$ of $`T_1`$ and $`T`$ maps isomorphically to $`T_1`$, we conclude that $`J^2H^0`$. Therefore the group $`H`$ generated by $`H^0`$ and $`J`$ maps isomorphically to $`H_1`$, and thus has no good representations. Therefore $`G`$ has no good representations. This proves Theorem 1.1.
The proof of Theorem 1.1 yields the following weaker statement in characteristic $`2`$. Note that Levi decompositions need not exist in positive characteristic \[Borel, 11.22\].
###### Corollary 4.1.
Suppose $`\text{char }k=2`$. If the connected algebraic group $`G`$ has a Levi decomposition and if $`G`$ has a good representation, then $`G`$ is solvable. In particular, any connected reductive group with a good representation is diagonalizable.
###### Proof.
Suppose $`G=LN`$ where $`L`$ is reductive and $`N`$ unipotent. If $`G`$ has a good representation then so does $`L`$. As proved above, this implies that $`L`$ is a torus, so $`G`$ is solvable. $`\mathrm{}`$
## 5. Proof of Theorem 1.2
If $`G`$ has a representation $`V`$ that contains an open set $`U`$ whose complement is a finite union of invariant subspaces such that $`G`$ acts with finite stabilizers on $`U`$, then $`G_{\overline{k}}`$ also has such a representation. Thus we can assume that $`k`$ is algebraically closed.
Assume that $`G`$ is not solvable and let $`V`$ be a representation of $`G`$. Since the characteristic is 0, $`G`$ has a Levi subgroup $`L`$. Since $`G`$ is assumed to be non-solvable, $`L`$ contains a Borel subgroup which is not a torus. Hence $`L`$ contains a non-trivial unipotent subgroup $`N`$.
Since the characteristic is 0 and $`L`$ is reductive, $`V`$ decomposes as a direct sum $`V=V_1V_2\mathrm{}V_p`$ of irreducible $`L`$-modules. Every vector in the subspace $`V^N=V_1^NV_2^N\mathrm{}V_p^N`$ has positive dimensional stabilizer. Since $`N`$ is unipotent, $`V_i^N0`$ for each $`i`$, so $`L(V_i^N)`$ spans all of $`V_i`$. Hence the subset $`LV^N=L(V_1^N\mathrm{}V_p^N)`$ which consists of vectors with positive dimensional stabilizers cannot be contained in any proper $`L`$-invariant subspace. Since $`L`$ is a subgroup of $`G`$ this means that $`LV^N`$ is not contained in any proper $`G`$-invariant subspace. Hence $`V`$ does not have properties (1) and (2). $`\mathrm{}`$
## 6. Examples and complements
In this section we discuss “set-theoretic” versions of the conditions freely good and good. We will say a representation $`V`$ is set-theoretically freely good (resp. set-theoretically good) if it contains a nonempty open subset $`U`$ whose complement is a union of invariant subspaces, such that $`G`$ acts with trivial stabilizers (resp. finite stabilizers) on $`U`$ (cf. Theorem 1.2). Surprisingly, these conditions are not enough to characterize solvability in arbitrary characteristic.
###### Example 6.1.
Let $`V`$ be the standard representation of $`\mathrm{SL}_2`$ and let $`V_d=S(V^{})`$ be the vector space of homogeneous forms of degree $`d`$. As in Section 4, $`\mathrm{SL}_2`$ acts on $`V_d`$. If $`p=\text{char }k`$ is an odd prime then $`W_p=V_{2p2}V_1`$ is a set-theoretically freely good representation of $`\mathrm{SL}_2`$. The reason is as follows. The stabilizer of any pair of forms $`(f(x,y),l(x,y))`$ is trivial as long as $`l(x,y)0`$ and the coefficient of $`x^{p1}y^{p1}`$ in $`f`$ is non-zero. Since the characteristic is $`p`$, the subspace $`L_{2p2}V_{2p2}`$ of forms with no $`x^{p1}y^{p1}`$ term is an $`\mathrm{SL}_2`$ invariant subspace (cf. \[J, II2.16\]). Thus, $`\mathrm{SL}_2`$ acts with trivial stabilizers on the open set $`U_p=W_p(W_{2p2}V_1V_{2p2}0)`$.
In characteristic 2, the representation $`W_2=V_2V_1`$ is not set-theoretically freely good because the matrix $`\left[\begin{array}{cc}0& 1\\ 1& 0\end{array}\right]`$ stabilizes the pair $`(x^2y^2,x+y)`$. However, $`W_2`$ is set-theoretically good.
In positive characteristic, we do not know if the group $`\mathrm{SL}_n`$ admits set-theoretically good representations for $`n3`$.
###### Example 6.2.
Assume $`k`$ is algebraically closed and $`\text{char }k2`$. Then $`G=PGL_2`$ has no representation which is set-theoretically freely good. Indeed, let
$$g=\left[\begin{array}{cc}0& 1\\ 1& 0\end{array}\right]$$
and let $`H=\{1,g\}`$. If $`V`$ is any representation of $`G`$, $`V^H`$ generates $`V`$ as a representation of $`G`$. Indeed, this holds if $`V`$ is irreducible, since for any vector $`v`$, the vector $`v+gv`$ is a nonzero $`H`$-invariant. Because $`\text{char }k2`$, the action of $`H`$ is completely reducible, so if
$$0V_1V_2V_30$$
is an exact sequence of $`G`$-modules, then the corresponding sequence of $`H`$-invariants is also exact. By induction, we may assume that $`V_1^H`$ and $`V_3^H`$ generate $`V_1`$ and $`V_3`$ as $`G`$-modules, and a diagram-chase then shows that $`V_2^H`$ generates $`V_2`$ as a $`G`$-module.
Hence if $`V`$ is any representation, there is no proper invariant linear subspace of $`V`$ containing $`V^H`$. Therefore $`V`$ is not set-theoretically freely good.
A similar argument shows that $`\mathrm{PGL}_n`$ and $`\mathrm{GL}_n`$ do not have set-theoretically freely good representations.
We conclude with a proposition about the inductive construction of good representations.
###### Proposition 6.1.
Let $`G`$ be a connected linear algebraic group and $`H`$ a normal subgroup. Assume $`k`$ is algebraically closed. If $`H`$ and $`G/H`$ have set-theoretically freely good representations, then so does $`G`$.
###### Proof.
For this proof only, we will use “good” to mean “set-theoretically freely good”. Let $`W`$ be a good representation of $`H`$, with $`M_i`$ a finite set of proper invariant subspaces containing the vectors with nontrivial stabilizers. Because $`G/H`$ is affine \[Borel, Theorem 6.8\], the vector bundle $`G\times ^HW`$ is generated by a finite dimensional space of global sections $`\mathrm{\Gamma }`$. We will view sections of the vector bundle as regular functions $`\gamma :GW`$ satisfying $`\gamma (gh)=h^1\gamma (g)`$, where on the right side we are using the action of $`H`$ on $`W`$. The action of $`G`$ on the space of sections of the vector bundle corresponds to the left action of $`G`$ on regular functions: $`(g\gamma )(g_0)=\gamma (g^1g_0)`$. Because the action of $`G`$ on regular functions is locally finite, by enlarging the space $`\mathrm{\Gamma }`$, we may assume $`\mathrm{\Gamma }`$ is stable under the $`G`$-action.
Define $`L_i`$ to be the subspace of $`\mathrm{\Gamma }`$ consisting of those elements of $`\mathrm{\Gamma }`$ which are sections of $`G\times ^HM_i`$. Each $`L_i`$ is a $`G`$-stable subspace of $`\mathrm{\Gamma }`$. Let $`\mathrm{\Gamma }^0`$ denote the complement of the $`L_i`$ in $`\mathrm{\Gamma }`$.
Let $`V`$ be a good representation of $`G/H`$, viewed as a representation of $`G`$ via the map $`GG/H`$. We claim that $`V\mathrm{\Gamma }`$ is a set-theoretically good representation of $`G`$. Indeed, let $`V_j`$ be a finite set of invariant subspaces of $`V`$ containing the vectors with nontrivial stabilizer. It suffices to show that the vectors with nontrivial stabilizer in $`V\mathrm{\Gamma }`$ are contained in the union of the subspaces $`V_j\mathrm{\Gamma }`$ and $`VL_i`$. To see this, let $`(v,\gamma )`$ be in the complement of these subspaces. so $`vV_j`$ and $`\gamma L_i`$ for any $`i,j`$. We must show that $`\mathrm{stab}_G(v,\gamma )`$ is trivial. First, $`\mathrm{stab}_G(v,\gamma )\mathrm{stab}_G(v)=H`$. Let $`h\mathrm{stab}_G(v,\gamma )`$. As above, we will view $`\gamma `$ as a function $`GW`$. Because $`\gamma `$ is not in any $`L_i`$, we have $`\gamma `$ is not a section of $`G\times ^HM_i`$ for any $`i`$. In other words, the open subsets $`\gamma ^1(WM_i)`$ of $`G`$ are nonempty. Choose $`g_0`$ in the intersection of these sets, so $`s(g_0)M_i`$ for any $`i`$. Our hypothesis implies that $`h\gamma =\gamma `$. By definition, we have
$$(h\gamma )(g_0)=\gamma (h^1g_0)=\gamma (g_0(g_0^1h^1g_0))=(g_0^1hg_0)\gamma (g_0).$$
But $`\mathrm{stab}_H\gamma (g_0)=\{1\}`$, so we conclude $`h=1`$ as desired. $`\mathrm{}`$
|
warning/0005/hep-ph0005181.html
|
ar5iv
|
text
|
# Untitled Document
hep-ph/0005181 WIS/8/00-May-DPP SLAC-PUB-8457 FERMILAB-Pub-00/102-T IASSNS–HEP–00–42 JHU-TIPAC-200003
Lessons from CLEO and FOCUS Measurements of $`D^0\overline{D^0}`$ Mixing Parameters
S. Bergmann<sup>a</sup>, Y. Grossman<sup>b</sup>, Z. Ligeti<sup>c</sup>, Y. Nir<sup>d,a</sup> and A.A. Petrov<sup>e</sup>
<sup>a</sup>Department of Particle Physics, Weizmann Institute of Science, Rehovot 76100, Israel
<sup>b</sup>Stanford Linear Accelerator Center, Stanford University, Stanford, CA 94309, USA<sup>1</sup> Research supported by the Department of Energy under contract DE-AC03-76SF00515.
<sup>c</sup>Theory Group, Fermilab, P.O. Box 500, Batavia, IL 60510, USA
<sup>d</sup>School of Natural Sciences, Institute for Advanced Study, Princeton, NJ 08540, USA<sup>2</sup> Address for academic year 1999-2000.
<sup>e</sup> Department of Physics and Astronomy, The Johns Hopkins University
3400 North Charles Street, Baltimore, MD 21218, USA
If the true values of the $`D^0\overline{D^0}`$ mixing parameters lie within the one sigma ranges of recent measurements, then there is strong evidence for a large width difference, $`y>0.01`$, and large $`SU\left(3\right)`$ breaking effects in strong phases, $`\delta >\pi /4`$. These constraints are model independent, and would become stronger if $`\left|M_{12}/\mathrm{\Gamma }_{12}\right|1`$ in the $`D^0\overline{D^0}`$ system. The interesting fact that the FOCUS result cannot be explained by a large mass difference is not trivial and depends on the small $`D^0/\overline{D^0}`$ production asymmetry in FOCUS and the bounds on CP violating effects from CLEO. The large value of $`\delta `$ might help explain why $`y\mathrm{sin}^2\theta _c`$.
5/00
1. Introduction
Recent studies of time-dependent decay rates of $`D^0K^+\pi ^{}`$ by the CLEO collaboration and measurements of the combination of $`D^0K^+K^{}`$ and $`D^0K^{}\pi ^+`$ rates by the FOCUS collaboration have provided highly interesting results concerning $`D^0\overline{D^0}`$ mixing. (For previous, related results, see \[3--8\].) Each of the two experiments finds a signal for mixing at a level that is close to $`2\sigma `$. It is not unlikely that these signals are just the results of statistical fluctuations and the true mixing parameters lie well below the experimental sensitivity. In this work, however, we interpret the experimental results assuming that their central values are not far from the true values and that $`D^0\overline{D^0}`$ mixing has indeed been observed.
The interpretation of the results and, in particular, testing the consistency of the two recent measurements with each other, require a careful treatment of signs and phase conventions. We present the relevant model-independent formalism in section 2. In section 3 we carefully explain what parameters have the FOCUS and CLEO experiments actually measured. We emphasize that, in principle, both CLEO and FOCUS results can be accounted for even if the width difference is negligibly small. This fact was known for the CLEO result , but it is much more subtle for the FOCUS result.
In section 4, we analyze the theoretical implications of the FOCUS and CLEO results in a model independent framework. We do however make some reasonable assumptions. With new physics, it is possible that there are large, CP violating new contributions to the mass difference. On the other hand, it is very unlikely that the width difference and relevant decay amplitudes are significantly affected by new physics. In such a framework, the measured observables depend on the mass difference $`x`$, the width difference $`y`$, two independent CP violating parameters, $`\varphi `$ and $`A_m`$, and a strong phase $`\delta `$. We find that the experimental results have strong implications for the width difference $`y`$ and for the strong phase $`\delta `$. The qualitative features are independent of the other parameters, though the detailed quantitative results are not.
It could be that the $`D^0\overline{D^0}`$ system is a unique example of a case where the dispersive part of the $`D^0\overline{D^0}`$ transition amplitude is much smaller than the absorptive part, $`|M_{12}||\mathrm{\Gamma }_{12}|`$. (For the $`K^0\overline{K^0}`$ the two are comparable, while for the $`B^0\overline{B^0}`$ and $`B_s\overline{B_s}`$ systems the situation is opposite, $`|M_{12}||\mathrm{\Gamma }_{12}|`$.) This situation, which is rarely discussed in the literature, is analyzed in section 5. We point out that, if this approximation is valid, the dependence on $`x`$ and on the CP violating parameters can be neglected. Consequently, the FOCUS and CLEO results depend on $`y`$ and $`\delta `$ only, and the implications become much clearer, both qualitatively and quantitatively.
Within the Standard Model, $`D^0\overline{D^0}`$ mixing vanishes in the limit of exact $`SU(3)`$ flavor symmetry of the strong interactions. For example, the sum of the contributions to the width difference from intermediate $`K^+K^{}`$, $`\pi ^+\pi ^{}`$, $`K^+\pi ^{}`$ and $`K^{}\pi ^+`$ states vanishes in the $`SU(3)`$ limit. The fact that the one sigma ranges of the FOCUS and CLEO results constrain $`\mathrm{cos}\delta `$ allows, for the first time, a calculation of this contribution based entirely on experimental data. We carry out such a calculation in section 6 and find a surprisingly large contribution to $`y`$, of order one percent.
A summary of our results is given in section 7.
2. Notations and Formalism
We investigate neutral $`D`$ decays. The two mass eigenstates, $`|D_1`$ of mass $`m_1`$ and width $`\mathrm{\Gamma }_1`$ and $`|D_2`$ of mass $`m_2`$ and width $`\mathrm{\Gamma }_2`$, are linear combinations of the interaction eigenstates:
$$\begin{array}{cc}\hfill |D_1=& p|D^0+q|\overline{D^0},\hfill \\ \hfill |D_2=& p|D^0q|\overline{D^0}.\hfill \end{array}$$
The average mass and width are given by
$$m\frac{m_1+m_2}{2},\mathrm{\Gamma }\frac{\mathrm{\Gamma }_1+\mathrm{\Gamma }_2}{2}.$$
The mass and width difference are parametrized by
$$x\frac{m_2m_1}{\mathrm{\Gamma }},y\frac{\mathrm{\Gamma }_2\mathrm{\Gamma }_1}{2\mathrm{\Gamma }}.$$
Decay amplitudes into a final state $`f`$ are defined by
$$A_ff|_d|D^0,\overline{A}_ff|_d|\overline{D^0}.$$
It is useful to define the complex parameter $`\lambda _f`$:
$$\lambda _f\frac{q}{p}\frac{\overline{A}_f}{A_f}.$$
The processes that are relevant to the CLEO and FOCUS experiments are the doubly-Cabibbo-suppressed $`D^0K^+\pi ^{}`$ decay, the singly-Cabibbo-suppressed $`D^0K^+K^{}`$ decay, the Cabibbo-favored $`D^0K^{}\pi ^+`$ decay, and the three CP-conjugate decay processes. We now write down approximate expressions for the time-dependent decay rates that are valid for times $`t<1/\mathrm{\Gamma }`$. We take into account the experimental information that $`x`$, $`y`$ and $`\mathrm{tan}\theta _c`$ are small, and expand each of the rates only to the order that is relevant to the CLEO and FOCUS measurements:
$$\begin{array}{cc}\hfill \mathrm{\Gamma }[D^0(t)& K^+\pi ^{}]=e^{\mathrm{\Gamma }t}|\overline{A}_{K^+\pi ^{}}|^2|q/p|^2\hfill \\ \hfill \times & \left\{|\lambda _{K^+\pi ^{}}^1|^2+[e(\lambda _{K^+\pi ^{}}^1)y+m(\lambda _{K^+\pi ^{}}^1)x]\mathrm{\Gamma }t+\frac{1}{4}(y^2+x^2)(\mathrm{\Gamma }t)^2\right\},\hfill \\ \hfill \mathrm{\Gamma }[\overline{D^0}(t)& K^{}\pi ^+]=e^{\mathrm{\Gamma }t}|A_{K^{}\pi ^+}|^2|p/q|^2\hfill \\ \hfill \times & \left\{|\lambda _{K^{}\pi ^+}|^2+[e(\lambda _{K^{}\pi ^+})y+m(\lambda _{K^{}\pi ^+})x]\mathrm{\Gamma }t+\frac{1}{4}(y^2+x^2)(\mathrm{\Gamma }t)^2\right\},\hfill \end{array}$$
$$\begin{array}{cc}\hfill \mathrm{\Gamma }[D^0(t)K^+K^{}]=& e^{\mathrm{\Gamma }t}|A_{K^+K^{}}|^2\left\{1+[e(\lambda _{K^+K^{}})ym(\lambda _{K^+K^{}})x]\mathrm{\Gamma }t\right\},\hfill \\ \hfill \mathrm{\Gamma }[\overline{D^0}(t)K^+K^{}]=& e^{\mathrm{\Gamma }t}|\overline{A}_{K^+K^{}}|^2\left\{1+[e(\lambda _{K^+K^{}}^1)ym(\lambda _{K^+K^{}}^1)x]\mathrm{\Gamma }t\right\},\hfill \end{array}$$
$$\begin{array}{cc}\hfill \mathrm{\Gamma }[D^0(t)& K^{}\pi ^+]=e^{\mathrm{\Gamma }t}|A_{K^{}\pi ^+}|^2,\hfill \\ \hfill \mathrm{\Gamma }[\overline{D^0}(t)& K^+\pi ^{}]=e^{\mathrm{\Gamma }t}|\overline{A}_{K^+\pi ^{}}|^2.\hfill \end{array}$$
Within the Standard Model, the physics of $`D^0\overline{D^0}`$ mixing and of the tree level decays is dominated by the first two generations and, consequently, CP violation can be safely neglected. In all ‘reasonable’ extensions of the Standard Model, the six decay modes of eqs. (2.1), (2.1) and (2.1) are still dominated by the Standard Model CP conserving contributions . On the other hand, there could be new short distance, possibly CP violating contributions to the mixing amplitude $`M_{12}`$. Allowing for only such effects of new physics, the picture of CP violation is simplified since there is no direct CP violation. The effects of indirect CP violation can be parametrized in the following way :
$$\begin{array}{cc}\hfill |q/p|=& R_m,\hfill \\ \hfill \lambda _{K^+\pi ^{}}^1=& \sqrt{R}R_m^1e^{i(\delta +\varphi )},\hfill \\ \hfill \lambda _{K^{}\pi ^+}=& \sqrt{R}R_me^{i(\delta \varphi )},\hfill \\ \hfill \lambda _{K^+K^{}}=& R_me^{i\varphi }.\hfill \end{array}$$
Here $`R`$ and $`R_m`$ are real and positive dimensionless numbers. CP violation in mixing is related to $`R_m1`$ while CP violation in the interference of decays with and without mixing is related to $`\mathrm{sin}\varphi 0`$. The choice of phases and signs in (2.1) is consistent with having $`\varphi =0`$ in the Standard Model and $`\delta =0`$ in the $`SU(3)`$ limit (see below). We further define
$$\begin{array}{cc}\hfill x^{}& x\mathrm{cos}\delta +y\mathrm{sin}\delta ,\hfill \\ \hfill y^{}& y\mathrm{cos}\delta x\mathrm{sin}\delta .\hfill \end{array}$$
With our assumption that there is no direct CP violation in the processes that we study, and using the parametrizations (2.1) and (2.1), we can rewrite eqs. (2.1)$``$(2.1) as follows:
$$\begin{array}{cc}\hfill \mathrm{\Gamma }[D^0(t)& K^+\pi ^{}]=e^{\mathrm{\Gamma }t}|A_{K^{}\pi ^+}|^2\hfill \\ \hfill \times & \left[R+\sqrt{R}R_m(y^{}\mathrm{cos}\varphi x^{}\mathrm{sin}\varphi )\mathrm{\Gamma }t+\frac{R_m^2}{4}(y^2+x^2)(\mathrm{\Gamma }t)^2\right],\hfill \\ \hfill \mathrm{\Gamma }[\overline{D^0}(t)& K^{}\pi ^+]=e^{\mathrm{\Gamma }t}|A_{K^{}\pi ^+}|^2\hfill \\ \hfill \times & \left[R+\sqrt{R}R_m^1(y^{}\mathrm{cos}\varphi +x^{}\mathrm{sin}\varphi )\mathrm{\Gamma }t+\frac{R_m^2}{4}(y^2+x^2)(\mathrm{\Gamma }t)^2\right]\hfill \end{array}$$
$$\begin{array}{cc}\hfill \mathrm{\Gamma }[D^0(t)& K^+K^{}]=e^{\mathrm{\Gamma }t}|A_{K^+K^{}}|^2[1R_m(y\mathrm{cos}\varphi x\mathrm{sin}\varphi )\mathrm{\Gamma }t],\hfill \\ \hfill \mathrm{\Gamma }[\overline{D^0}(t)& K^+K^{}]=e^{\mathrm{\Gamma }t}|A_{K^+K^{}}|^2[1R_m^1(y\mathrm{cos}\varphi +x\mathrm{sin}\varphi )\mathrm{\Gamma }t],\hfill \end{array}$$
$$\mathrm{\Gamma }[D^0(t)K^{}\pi ^+]=\mathrm{\Gamma }[\overline{D^0}(t)K^+\pi ^{}]=e^{\mathrm{\Gamma }t}|A_{K^{}\pi ^+}|^2.$$
3. CLEO and FOCUS Measurements
The FOCUS experiment fits the time dependent decay rates of the singly-Cabibbo suppressed (2.1) and the Cabibbo-favored (2.1) modes to pure exponentials. We define $`\widehat{\mathrm{\Gamma }}`$ to be the parameter that is extracted in this way. More explicitly, for a time dependent decay rate with $`\mathrm{\Gamma }[D(t)f]e^{\mathrm{\Gamma }t}(1z\mathrm{\Gamma }t+\mathrm{})`$, where $`|z|1`$, we have $`\widehat{\mathrm{\Gamma }}(Df)=\mathrm{\Gamma }(1+z)`$. The above equations imply the following relations:
$$\begin{array}{cc}\hfill \widehat{\mathrm{\Gamma }}(D^0K^+K^{})=& \mathrm{\Gamma }[1+R_m(y\mathrm{cos}\varphi x\mathrm{sin}\varphi )],\hfill \\ \hfill \widehat{\mathrm{\Gamma }}(\overline{D^0}K^+K^{})=& \mathrm{\Gamma }[1+R_m^1(y\mathrm{cos}\varphi +x\mathrm{sin}\varphi )],\hfill \\ \hfill \widehat{\mathrm{\Gamma }}(D^0K^{}\pi ^+)=& \widehat{\mathrm{\Gamma }}(\overline{D^0}K^+\pi ^{})=\mathrm{\Gamma }.\hfill \end{array}$$
Note that deviations of $`\widehat{\mathrm{\Gamma }}(DK^+K^{})`$ from $`\mathrm{\Gamma }`$ do not require that $`y0`$. They can be accounted for by $`x0`$ and $`\mathrm{sin}\varphi 0`$, but then they have a different sign in the $`D^0`$ and $`\overline{D^0}`$ decays. FOCUS combines the two $`DK^+K^{}`$ modes. To understand the consequences of such an analysis, one has to consider the relative weight of $`D^0`$ and $`\overline{D^0}`$ in the sample. Let us define $`A_{\mathrm{prod}}`$ as the production asymmetry of $`D^0`$ and $`\overline{D^0}`$:
$$A_{\mathrm{prod}}\frac{N(D^0)N(\overline{D^0})}{N(D^0)+N(\overline{D^0})}.$$
Then
$$\begin{array}{cc}\hfill y_{\mathrm{CP}}& \frac{\widehat{\mathrm{\Gamma }}(DK^+K^{})}{\widehat{\mathrm{\Gamma }}(D^0K^{}\pi ^+)}1\hfill \\ \hfill =& y\mathrm{cos}\varphi \left[\frac{1}{2}(R_m+R_m^1)+\frac{A_{\mathrm{prod}}}{2}(R_mR_m^1)\right]\hfill \\ & x\mathrm{sin}\varphi \left[\frac{1}{2}(R_mR_m^1)+\frac{A_{\mathrm{prod}}}{2}(R_m+R_m^1)\right].\hfill \end{array}$$
The one sigma range measured by FOCUS is
$$y_{\mathrm{CP}}=(3.42\pm 1.57)\times 10^2.$$
The interpretation of this measurement simplifies when the following two facts are taken into account:
(i) The E687 data suggest that $`A_{\mathrm{prod}}`$ is small for FOCUS, of order 0.03.
(ii) The CLEO data suggest that $`R_m`$ is not very different from one (see below). Actually, CLEO implicitly assume that this is the case in their analysis by using
$$R_m^{\pm 2}=1\pm A_m.$$
Evaluating (3.1) to linear order in the small quantities $`A_{\mathrm{prod}}`$ and $`A_m`$ yields
$$y_{\mathrm{CP}}=y\mathrm{cos}\varphi x\mathrm{sin}\varphi \left(\frac{A_m}{2}+A_{\mathrm{prod}}\right).$$
The CLEO measurement gives the coefficient of each of the three terms ($`1`$, $`\mathrm{\Gamma }t`$ and $`(\mathrm{\Gamma }t)^2`$) in the doubly-Cabibbo suppressed decays (2.1). Such measurements allow a fit to the parameters $`R`$, $`R_m`$, $`x^{}\mathrm{sin}\varphi `$, $`y^{}\mathrm{cos}\varphi `$, and $`x^2+y^2`$. Fit A of ref. quotes the following one sigma ranges:<sup>3</sup> CLEO quote a range for $`y^{}`$. It is obvious however that, with our conventions, their range applies to $`y^{}\mathrm{sign}(\mathrm{cos}\varphi )`$ or perhaps to $`y^{}\mathrm{cos}\varphi `$. Since the one sigma range is $`|\mathrm{cos}\varphi |>0.8`$, the difference between these two possibilities is unimportant for our purposes.
$$\begin{array}{cc}\hfill R=& (0.48\pm 0.13)\times 10^2,\hfill \\ \hfill y^{}\mathrm{cos}\varphi =& (2.5_{1.6}^{+1.4})\times 10^2,\hfill \\ \hfill x^{}=& (0.0\pm 1.5)\times 10^2,\hfill \\ \hfill A_m=& 0.23_{0.80}^{+0.63}.\hfill \end{array}$$
We would like to point out that the interpretation of the FOCUS and CLEO results in terms of $`y`$, $`x`$, $`\varphi `$, $`\delta `$ and $`A_m`$ is almost independent of our assumption that there is no CP violation in decay. To understand this point, let us parametrize CP violation in decay in the following way:
$$\begin{array}{cc}\hfill A_{\mathrm{CP}}(f)& \frac{\mathrm{\Gamma }(D^0f)\mathrm{\Gamma }(\overline{D^0}\overline{f})}{\mathrm{\Gamma }(D^0f)+\mathrm{\Gamma }(\overline{D^0}\overline{f})}\hfill \\ \hfill =& \frac{1|\overline{A}_{\overline{f}}/A_f|^2}{1+|\overline{A}_{\overline{f}}/A_f|^2}.\hfill \end{array}$$
Experimentally, we have the following constraints on the asymmetries in the Cabibbo-favored , singly-Cabibbo-suppressed \[15--17\] and doubly-Cabibbo-suppressed decays:
$$\begin{array}{cc}\hfill A_{\mathrm{CP}}(K^{}\pi ^+)=& 0.001\pm 0.011,\hfill \\ \hfill A_{\mathrm{CP}}(K^{}K^+)=& 0.0004\pm 0.0234,\hfill \\ \hfill A_{\mathrm{CP}}(K^+\pi ^{})=& 0.01\pm 0.17.\hfill \end{array}$$
For FOCUS, eq. (3.1) would be corrected by terms of order $`A_{\mathrm{CP}}(K^{}K^+)A_{\mathrm{prod}}`$ and $`A_{\mathrm{CP}}(K^{}K^+)A_m`$, which are negligible. For CLEO, the results in eq. (3.1) have been obtained allowing for CP violation in decay. There is however another subtle aspect of direct CP violation where our theoretical assumption does play a role. In the presence of new CP violating contributions to the decay amplitudes, the CP violating phases in $`\lambda _f`$ are not necessarily universal. Therefore, the use of a single phase $`\varphi `$ in eq. (2.1) and consequently in eqs. (3.1) and (3.1) is valid only in the absence of direct CP violation.
4. Theoretical Interpretation
We now assume that the true values of the various mixing parameters are within the one sigma ranges measured by FOCUS and CLEO. That means in particular that we hypothesize that $`D^0\overline{D^0}`$ mixing is being observed in the FOCUS measurement of $`y_{\mathrm{CP}}`$ and in the CLEO measurement of $`y^{}\mathrm{cos}\varphi `$. The combination of these two results is particularly powerful in its theoretical implications.
Let us first focus on the FOCUS result (3.1). We argue that it is very unlikely that this result is accounted for by the second term in (3.1). Even if we take all the relevant parameters to be close to their one sigma upper bounds, say, $`|x|0.04`$ (we use Fig. 3 of ref. to extract this upper bound), $`|\mathrm{sin}\varphi |0.6`$, $`|A_m/2|0.4`$ and $`A_{\mathrm{prod}}0.03`$, we get $`y_{\mathrm{CP}}0.01`$, about a factor of two too small. We can make then the following model independent statement: if the true values of the mixing parameters are within the one sigma ranges of CLEO and FOCUS measurements, then $`y`$ is of order of a (few) percent. Note that this is true even in the presence of CP violation, which does allow a mass difference, $`x0`$, to mimic a deviation from the average lifetime. Practically, we can take the FOCUS result to be given to a good approximation by
$$y\mathrm{cos}\varphi 0.034\pm 0.016.$$
This is a rather surprising result. Most theoretical estimates are well below the one percent level (for a review, see ). These estimates have however been recently criticized \[19,,20\]. We will have more to say about this issue in section 6.
Second, we examine the consistency of the FOCUS and CLEO results. The two most significant measurements, that of $`y\mathrm{cos}\varphi `$ in eq. (4.1) and that of $`y^{}\mathrm{cos}\varphi `$ in eq. (3.1) are consistent if
$$\mathrm{cos}\delta (x/y)\mathrm{sin}\delta =0.73\pm 0.55.$$
This requirement allows us to make a second model independent statement: if the true values of the mixing parameters are within the one sigma ranges of CLEO and FOCUS measurements, then the difference in strong phases between the $`D^0K^+\pi ^{}`$ and $`D^0K^{}\pi ^+`$ decays is very large. For $`\delta =0`$ we get $`y^{}/y=1`$ instead of the range given in eq. (4.1). To satisfy (4.1), we need, for example,
$$\mathrm{cos}\delta <\{\begin{array}{cc}+0.65\hfill & |x||y|\text{,}\hfill \\ 0.18\hfill & |x||y|\text{.}\hfill \end{array}$$
The result in eq. (4.1) is also rather surprising. The strong phase $`\delta `$ vanishes in the $`SU(3)`$ flavor symmetry limit . None of the models in the literature \[9,,22,,23\] finds such a large $`\delta `$. Eq. (4.1) implies a very large $`SU(3)`$ breaking effect in the strong phase. For comparison, the experimental value of $`\sqrt{R}0.07`$ in eq. (3.1) is enhanced compared to its $`SU(3)`$ value of $`\mathrm{tan}^2\theta _c0.051`$ by a factor $`1.4`$. On the other hand, there are other known examples of $`SU(3)`$ breaking effects of order one in $`D`$ decays,<sup>4</sup> For example, $`\mathrm{\Gamma }(D^0K^+K^{})/\mathrm{\Gamma }(D^0\pi ^+\pi ^{})=2.75\pm 0.15\pm 0.16`$ experimentally , while the ratio is predicted to be one in the $`SU(3)`$ limit. so perhaps we should not be prejudiced against a very large $`\delta `$.
Before concluding this section, we would like to explain the consequences of the CLEO and FOCUS measurements in the context of the Standard Model. Within the Standard Model, $`D^0\overline{D^0}`$ mixing and $`D^0`$ decays into $`K^+K^{}`$, $`\pi ^+\pi ^{}`$ and $`\pi ^\pm K^{}`$ are described to an excellent approximation by physics of the first two generations. Consequently, the Standard Model makes a clean prediction that any CP violating effects in these processes are negligibly small. We can thus safely set $`\varphi =0`$ and $`R_m=1`$. The statements below hold in any model where CP is a good symmetry in the relevant processes.
It is important to realize that the choice of $`\varphi =0`$ is equivalent to choosing $`|D_1`$ ($`|D_2`$) to be the CP-odd (CP-even) state, $`|D_{}`$ ($`|D_+`$). This can be seen from eq. (2.1). It gives $`\lambda _{K^+K^{}}=1`$. We define the CP-odd state as the mass eigenstate that does not decay into $`K^+K^{}`$. Indeed, we now have
$$K^+K^{}||D_1=pA_{K^+K^{}}(1+\lambda _{K^+K^{}})=0.$$
In the CP limit, a non-zero value of $`y_{\mathrm{CP}}`$ (see eq. (3.1)) requires unambiguously that the width difference is large:
$$y=\frac{\mathrm{\Gamma }_+\mathrm{\Gamma }_{}}{2\mathrm{\Gamma }}=(3.42\pm 1.39\pm 0.74)\times 10^2.$$
The fact that $`y>0`$ is preferred suggests that the CP-even state has a shorter lifetime, that is $`|D_{+,}=|D_{S,L}`$ where $`S`$ and $`L`$ stands for ‘short’ and ‘long’ lifetimes, respectively. This important result holds in the CP limit model independently.
5. The Case of $`|M_{12}/\mathrm{\Gamma }_{12}|1`$
It could be the case that $`SU(3)`$ breaking effects are stronger for the absorptive part of the $`D^0\overline{D^0}`$ transition amplitude, $`\mathrm{\Gamma }_{12}`$, than for the dispersive part, $`M_{12}`$. In this section we investigate the implications of the FOCUS and CLEO results in case that indeed
$$|M_{12}/\mathrm{\Gamma }_{12}|1.$$
When we neglect small effects of $`𝒪(|M_{12}/\mathrm{\Gamma }_{12}|)`$, several simplifications occur. Define
$$\varphi _{12}\mathrm{arg}(M_{12}/\mathrm{\Gamma }_{12}).$$
Then, to leading order in $`|M_{12}/\mathrm{\Gamma }_{12}|`$, we have:
$$\begin{array}{cc}\hfill x/y=& 2\left|M_{12}/\mathrm{\Gamma }_{12}\right|\mathrm{cos}\varphi _{12},\hfill \\ \hfill A_m=& 4\left|M_{12}/\mathrm{\Gamma }_{12}\right|\mathrm{sin}\varphi _{12},\hfill \\ \hfill \varphi =& 2\left|M_{12}/\mathrm{\Gamma }_{12}\right|^2\mathrm{sin}2\varphi _{12}.\hfill \end{array}$$
We learn that in the limit (5.1), $`x`$ can be neglected and all CP violating effects can be neglected. This should be contrasted with the case of $`|\mathrm{\Gamma }_{12}/M_{12}|1`$, which holds for the $`B`$ and $`B_s`$ mesons, where the effects of $`A_m`$ can be neglected but those of $`\varphi `$ are not suppressed. There are two interesting consequnces of this difference. First, in the $`B_s`$ system, a lifetime difference between CP eigenstates and flavor specific final states (analoguous to $`y_{\mathrm{CP}}`$ of eq. (3.1)) measures $`\mathrm{\Delta }\mathrm{\Gamma }(B_s)`$ only if there is no new CP violation in the mixing . In the $`D`$ system, if (5.1) holds, $`y_{\mathrm{CP}}y`$ model independently. Second, even in the case that new physics dominates $`M_{12}(D)`$, the sensitivity of any physical observable to it is suppressed by $`|M_{12}/\mathrm{\Gamma }_{12}|`$.
Neglecting $`x`$, $`A_m`$ and $`\varphi `$, the FOCUS and CLEO results can be written as follows:
$$\begin{array}{cc}\hfill y=& (3.42\pm 1.57)\times 10^2,\hfill \\ \hfill y\mathrm{cos}\delta =& (2.5_{1.6}^{+1.4})\times 10^2,\hfill \\ \hfill y\mathrm{sin}\delta =& (0.0\pm 1.5)\times 10^2.\hfill \end{array}$$
The FOCUS measurement determines directly $`y`$. The first two equations give
$$\mathrm{cos}\delta =0.73_{0.27}^{+0.55}.$$
The third equation requires that $`|\mathrm{sin}\delta |`$ is not large and consequently narrows the range for $`\delta `$ even further,
$$\mathrm{cos}\delta <0.5.$$
The conclusion of our discussion here is that if the $`D^0\overline{D^0}`$ system provides a (unique!) example of $`|M_{12}||\mathrm{\Gamma }_{12}|`$, then the FOCUS and CLEO measurements determine $`y`$ to be at the few percent level and the strong phase $`\delta `$ is well above $`\pi /2`$.
6. Implications for the Width Difference
The value of the phase $`\delta `$ has important implications for another aspect of our study, that is the width difference. The contributions of the four charged two-body states,
$$n_{2c}=K^+K^{},\pi ^+\pi ^{},K^+\pi ^{},K^{}\pi ^+,$$
to $`\mathrm{\Gamma }_{12}`$, the absorptive part of the transition amplitude $`D^0||\overline{D^0}`$, can be written as
$$(\mathrm{\Gamma }_{12})_{2c}=\underset{n_{2c}}{}A_{n_{2c}}^{}\overline{A}_{n_{2c}},$$
which leads to the following contribution to $`y`$:
$$y_{2c}=\mathrm{BR}(D^0K^{}K^+)+\mathrm{BR}(D^0\pi ^{}\pi ^+)2\mathrm{cos}\delta \sqrt{R}\mathrm{BR}(D^0K^+\pi ^{}).$$
There are two points that we would like to extract from eq. (6.1). First, in the $`SU(3)`$ limit, $`\mathrm{BR}(D^0K^{}K^+)=\mathrm{BR}(D^0\pi ^{}\pi ^+)=\sqrt{R}\mathrm{BR}(D^0K^+\pi ^{})`$. The phase $`\delta `$ defined in (2.1) vanishes in the $`SU(3)`$ limit which is consitent with the fact that $`y_{2c}=0`$ in this limit. Second, we can use the measured branching ratios for the four decay modes and the value of the phase $`\delta `$ as fitted to the CLEO and FOCUS results to estimate $`y_{2c}`$. We use
$$\begin{array}{cc}\hfill \mathrm{BR}(D^0K^{}\pi ^+)=& (3.83\pm 0.09)\times 10^2,\hfill \\ \hfill \mathrm{BR}(D^0\pi ^{}\pi ^+)=& (1.52\pm 0.09)\times 10^3,\hfill \\ \hfill \mathrm{BR}(D^0K^{}K^+)=& (4.24\pm 0.16)\times 10^3,\hfill \end{array}$$
and (see eq. (3.1))
$$\sqrt{R}=0.069\pm 0.009.$$
Using central values for the branching ratios, we get:
$$y_{2c}(5.765.29\mathrm{cos}\delta )\times 10^3.$$
Taking $`1<\mathrm{cos}\delta <0`$ from (4.1), we find
$$0.6\times 10^2<y_{2c}<1.1\times 10^2,$$
to be compared with the range (4.1) for $`y`$. Note that the sign of this contribution is consistent with the overall sign of $`y`$ as measured by FOCUS. There are of course other intermediate states that contribute to $`y`$. Eq. (6.1) suggests that, if the strong phases strongly violate $`SU(3)`$ as required for consistency of the CLEO and FOCUS results, such contributions could easily be at the percent level as required by the same experiments.
7. Conclusions
The FOCUS and CLEO collaborations have provided new measurements of the $`D^0\overline{D^0}`$ mixing parameters that are sensitive to effects of order a few percent. FOCUS obtains a 2.2$`\sigma `$ signal and CLEO obtains a 1.8$`\sigma `$ signal of such effects. It could well be that these signals are just statistical fluctuations and that the mixing parameters are much smaller than the percent level. This is the theoretical wisdom, based on the Standard Model and on approximate flavor $`SU(3)`$. If, however, the central values of the two measurements are close to the true values, then at least the assumption of approximate $`SU(3)`$ for the strong interactions has to be modified. In particular, there are two independent pieces of evidence that the strong phase in $`DK^\pm \pi ^{}`$ decays is very large, $`\delta >\pi /4`$ and perhaps $`\delta 3\pi /4`$ (while $`\delta =0`$ in the $`SU(3)`$ limit):
(i) Either a negative sign for $`\mathrm{cos}\delta `$ or large $`x`$ and large $`\mathrm{sin}\delta `$ are necessary to make the signs of the mixing parameters measured by FOCUS and by CLEO consistent with each other.
(ii) $`\mathrm{cos}\delta `$ far from its $`SU(3)`$ limit value of one implies that some contributions to the width difference are at the percent level.
We also discussed the possibility that in the $`D^0\overline{D^0}`$ system $`|M_{12}/\mathrm{\Gamma }_{12}|1`$, in contrast to the neutral $`B`$ meson systems. In such a case, the $`D^0\overline{D^0}`$ system is not sensitive to new physics, even if new physics dominates $`M_{12}`$. In particular, CP is expected to be a good symmetry regardless of whether there are large CP violating contributions to $`M_{12}`$. The above statements about large $`SU(3)`$ breaking effects become even sharper in this case.
A much clearer picture would emerge if the accuracy of the measurements improves and, in particular, if the mixing parameters are measured separately in the $`D^0`$ and $`\overline{D^0}`$ decays. For example, the FOCUS collaboration has summed over the $`D^0K^+K^{}`$ and $`\overline{D^0}K^+K^{}`$ modes, but there is much to learn from comparing them to each other. Explicitly, we obtain from eq. (3.1):
$$\frac{\widehat{\mathrm{\Gamma }}(D^0K^+K^{})\widehat{\mathrm{\Gamma }}(\overline{D^0}K^+K^{})}{\widehat{\mathrm{\Gamma }}(D^0K^+K^{})+\widehat{\mathrm{\Gamma }}(\overline{D^0}K^+K^{})}=\frac{A_m}{2}y\mathrm{cos}\varphi x\mathrm{sin}\varphi .$$
A difference between the fitted decay width of the two CP conjugate modes will provide important information on the CP violating parameters.
Acknowledgments
We thank David Asner, Adam Falk, Harry Nelson and Jim Wiss for useful discussions. S.B. thanks the School of Natural Sciences at the Institute for Advanced Study for the hospitality. Y.G. is supported by the U.S. Department of Energy under contract DE-AC03-76SF00515. Fermilab is operated by Universities Research Association, Inc., under DOE contract DE-AC02-76CH03000. Y.N. is supported by the Department of Energy under contract No. DE–FG02–90ER40542, by the Ambrose Monell Foundation, by AMIAS (Association of Members of the Institute for Advanced Study), by the Israel Science Foundation founded by the Israel Academy of Sciences and Humanities, and by the Minerva Foundation (Munich).
References
relax R. Godang et al. \[CLEO Collaboration\], hep-ex/0001060. relax J.M. Link et al. \[FOCUS Collaboration\], hep-ex/0004034. relax J.C. Anjos et al. \[E691 Collaboration\], Phys. Rev. Lett. 60 (1988) 1239. relax D. Cinabro et al. \[CLEO Collaboration\], Phys. Rev. Lett. 72 (1994) 1406. relax E.M. Aitala et al. \[E791 Collaboration\], Phys. Rev. Lett. 77 (1996) 2384, hep-ex/9606016. relax E.M. Aitala et al. \[E791 Collaboration\], Phys. Rev. D57 (1998) 13, hep-ex/9608018. relax R. Barate et al. \[ALEPH Collaboration\], Phys. Lett. B436 (1998) 211, hep-ex/9811021. relax E.M. Aitala et al. \[E791 Collaboration\], Phys. Rev. Lett. 83 (1999) 32, hep-ex/9903012. relax A.F. Falk, Y. Nir and A.A. Petrov, JHEP 12 (1999) 019, hep-ph/9911369. relax G. Blaylock, A. Seiden and Y. Nir, Phys. Lett. B355 (1995) 555, hep-ph/9504306. relax S. Bergmann and Y. Nir, JHEP 09 (1999) 031, hep-ph/9909391. relax Y. Nir, hep-ph/9911321. relax P.L. Frabetti et al. \[E687 Collaboration\], Phys. Lett. B370 (1996) 222. relax J. Bartelt et al. \[CLEO Collaboration\], Phys. Rev. D52 (1995) 4860. relax E.M. Aitala et al. \[E791 Collaboration\], Phys. Lett. B421 (1998) 405, hep-ex/9711003. relax J. Wiss, Fermilab Joint Experimental Theoretical Seminar (April 2000), available from http://www-hep.colorado.edu/FocusPublic/ConferencesPublic. relax D. Asner, T. Hill, S. McGee and C. Prescott \[CLEO Collaboration\], Presented at APS April Meeting 2000, Long Beach, CA. relax H.N. Nelson, hep-ex/9908021. relax I.I. Bigi and N.G. Uraltsev, hep-ph/0005089. relax A.F. Falk, Y. Grossman, Z. Ligeti and A.A. Petrov, to appear. relax L. Wolfenstein, Phys. Rev. Lett. 75 (1995) 2460, hep-ph/9505285. relax L. Chau and H. Cheng, Phys. Lett. B333 (1994) 514, hep-ph/9404207. relax T.E. Browder and S. Pakvasa, Phys. Lett. B383 (1996) 475, hep-ph/9508362. relax Y. Grossman, Phys. Lett. B380 (1996) 99, hep-ph/9603244. relax C. Caso et al., The Eur. Phys. J. C3 (1998) 1, and 1999 off-year partial update for the 2000 edition available on the PDG WWW pages (URL: http://pdg.lbl.gov/).
|
warning/0005/hep-th0005027.html
|
ar5iv
|
text
|
# Untitled Document
RUNHETC-2000-17 hepth/nnnmmyy
Finite Temperature Expectation Values of Local Fields in the sinh-Gordon model
Sergei Lukyanov
Department of Physics and Astronomy, Rutgers University
Piscataway, NJ 08855-0849, USA
and
L.D. Landau Institute for Theoretical Physics
Kosygina 2, Moscow, Russia
Abstract
Sklyanin’s method of separation of variables is employed in a calculation of finite temperature expectation values. An essential element of the approach is Baxter’s $`Q`$-function. We propose its explicit form corresponding to the ground state of the sinh-Gordon theory. With the method of separation of variables we calculate the finite temperature expectation values of the exponential fields to one-loop order of the semi-classical expansion.
April, 00
1. Introduction
One-point functions have numerous applications in Statistical Mechanics and Condensed Matter Physics \[1,,2\]. They determine “generalized susceptibilities” i.e. the linear response of a system to external fields. In a path integral formulation the one-point function of a local field $`𝒪`$ is represented by a Euclidean path integral of the form
$$𝒪=Z^1𝒟[\phi ]𝒪e^𝒜.$$
Recently some progress has been achieved in the calculation of one-point functions in integrable Quantum Field Theory (QFT) defined on a two dimensional Euclidean plane \[3,,4\]. In this case, the integral (1.1) can also be viewed as a Vacuum Expectation Value (VEV) in $`1+1`$-dimensional QFT associated with the action $`𝒜`$. For many applications, especially in Condensed Matter Physics, it is important to generalize the results of Refs.\[3,,4\] to the case of Euclidean path integrals defined on an infinite cylinder. In the Matsubara imaginary time formalism such path integrals are interpreted as thermal averages
$$𝒪_R=\frac{\mathrm{Tr}\left[e^{R𝐇}𝒪\right]}{\mathrm{Tr}\left[e^{R𝐇}\right]},$$
where $`𝐇`$ is the Hamiltonian of the corresponding QFT and the temperature coincides with the inverse circumference of the cylinder.
The path integral defined on a cylinder also allows another interpretation. It is an expectation value for the ground state $`|vac_R`$ of the $`1+1`$-dimensional theory in the finite geometry where the spatial coordinate is compactified on a circle. Hence, the VEVs contain important information about Renormalization Group flow controlled by the parameter $`R`$.
An exact calculation of the finite volume (finite $`R`$) VEVs is a challenge even in integrable QFT. Recent progress made in papers \[5,,6\] should be mentioned here. In A. Leclair and G. Mussardo proposed an integral representation which makes it possible to generate a low-temperature ($`R\mathrm{}`$) expansion for the VEVs in terms of infinite volume form-factors of local fields and some thermodynamical data. Their conjecture works for theories with trivial $`S`$-matrices such as Ising and Free Dirac Fermion models , but its validity remains questionable for models with non-trivial scattering amplitudes (see e.g. ). Another line of research was proposed in the work . F. Smirnov applied the method of separation of variables \[9,,10\] to the semi-classical study of finite volume matrix elements in the quantum KdV equation. The model does not constitute a relativistic field theory. Nevertheless, it is of prime importance for the sinh-Gordon QFT since both of the equations are in the same integrable hierarchy.
In this paper we will implement the method of separation of variables in the case of the quantum sinh-Gordon theory. The problem is defined by a Euclidean action,
$$𝒜_{\mathrm{shG}}=_{\mathrm{}}^{\mathrm{}}𝑑x_1_0^R𝑑x_2\left\{\frac{1}{16\pi }\left(_\sigma \phi \right)^2+2\mu \mathrm{cosh}(b\phi )\right\},$$
where $`\phi `$ is a scalar field with periodic boundary condition along the $`x_2`$-coordinate. We are focusing on the VEVs of the exponential fields,
$$𝒪=e^{a\phi }.$$
For our purposes it will be useful to rewrite (1.1) in the form,
$$e^{a\phi }_R=Z^1𝒟[\chi ]\mathrm{\Psi }_0^2[\chi ]e^{a\chi }.$$
Here $`\mathrm{\Psi }_0[\chi ]`$ is an integral taken over field configurations on the half-cylinder, $`x_1<0`$, satisfying the boundary condition,
$$\phi (x_1,x_2)|_{x_1=0}=\chi (x_2),$$
i.e. it is a wave functional corresponding to the ground state $`|vac_R`$. The method of separation of variables \[9,,10,,5\] allows one to introduce a change of integration variables in (1.1),
$$\chi (x_2)\left\{\gamma _k\right\}_{k=\mathrm{}}^{\mathrm{}},$$
from the function $`\chi (x_2)`$ to the infinite discrete set of $`\gamma _k`$. A notable advantage of the new variables is that the wave functional in the “$`\gamma `$-representation” has a factorizable form,
$$\mathrm{\Psi }_0\left[\{\gamma _k\}\right]\underset{k}{}𝒬[\gamma _k].$$
Notice that the integration measure $`𝒟\left[\{\gamma _k\}\right]`$ does not factorize in the variables $`\gamma _k`$. At this moment, we are not able to elaborate on all steps of the changing of variables on a rigorous basis. Therefore, we suggest the deduced integral representation for $`e^{a\phi }_R`$ as a conjecture rather than a well established result. To test the validity of this integral representation, we carry out a semi-classical expansion of the VEV.
More explicitly, the parameter $`b^2`$ in the action (1.1) can readily be identified with the Planck constant. Then, for finite $`\alpha =ab`$ and $`b^20`$, the functional integral (1.1) is dominated by a non-trivial saddle-point configuration and admits the semi-classical expansion,
$$e^{a\phi }_R=e^{\frac{S}{b^2}}D\left(\mathrm{\hspace{0.17em}1}+O(b^2)\right).$$
Here $`S`$ is a Euclidean action on the cylinder evaluated in the saddle-point configuration and the pre-exponential factor $`D`$ is the result of evaluating the functional integral (1.1) in the Gaussian approximation around the classical solution. With the proposed integral representation we calculate the functions $`S`$ and $`D`$ and find complete agreement with the expected high- and low-temperature behavior of the VEVs. In particular, our result matches well with the Leclair-Mussardo conjecture.
2. Integral representation for VEVs
2.1. Flaschka-McLaughlin variables
In the paper H. Flaschka and D. McLaughlin found remarkable canonically conjugate variables in the phase spaces of the classical Toda chain and KdV equations. Their approach can be straightforwardly adapted to the classical sinh-Gordon equation. Here we give a brief review of the Flaschka-McLaughlin variables for this model. For more information and proofs, the reader is referred to Refs.\[11,,12\].
The sinh-Gordon equation admits a zero curvature formulation: There exists a $`sl(2,R)`$-valued connection 1-form, depending on an auxiliary parameter $`\lambda `$, such that the condition of vanishing curvature is equivalent to the equation of motion. Dealing with the theory on cylinder, one can integrate this 1-form along some cycle, say,
$$x_1=0,0x_2<R,$$
and obtain the so-called monodromy matrix,
$$𝐌(\lambda )=\left(\begin{array}{cc}𝙰(\lambda )& \lambda 𝙱(\lambda )\\ \lambda ^1𝙲(\lambda )& 𝙳(\lambda )\end{array}\right)SL(2,R)(\mathrm{}m\lambda =0).$$
This matrix satisfies important analytical conditions which are readily obtained from an explicit form of the connection. In particular, the matrix elements in (2.1) are real analytical functions of the variable $`\lambda ^2`$ with two essential singularities at the points $`\lambda ^2=0,\mathrm{}`$. Zeroes of $`𝙱(\lambda )`$,
$$\lambda _k^2:𝙱(\lambda _k)=0,$$
are of prime importance in the construction. It is possible to show that all zeroes (2.1) are simple, real, positive and accumulate towards the essential singularities. Thus we can order them,
$$0\mathrm{}\lambda _N^2<\lambda _{N+1}^2\mathrm{}<\lambda _0^2<\mathrm{}\lambda _{N1}^2<\lambda _N^2\mathrm{}+\mathrm{}$$
and define two infinite sets:
$$\begin{array}{cc}& \left\{\gamma _k\right\}_{k=\mathrm{}}^{\mathrm{}}:\gamma _k=\mathrm{log}\lambda _k^2,\hfill \\ & \left\{\pi _k\right\}_{k=\mathrm{}}^{\mathrm{}}:\pi _k=4\mathrm{log}|𝙰(\lambda _k)|.\hfill \end{array}$$
The mapping of the canonical Poisson data, $`\phi `$ and $`_{x_1}\phi `$, to the variables (2.1) is found to be a canonical transformation, i.e.
$$\{\pi _k,\gamma _m\}=\delta _{km},\{\pi _k,\pi _m\}=\{\gamma _k,\gamma _m\}=0.$$
Hence (2.1) are canonically conjugate variables in the phase space of the sinh-Gordon model. At the same time, we can treat $`\left\{\gamma _k\right\}_{k=\mathrm{}}^{\mathrm{}}`$ as coordinates in the corresponding configuration space.
2.2. $`\gamma `$-representation
The Flaschka-McLaughlin variables proved to be useful in quantum theory as was demonstrated in the seminal work on the example of the Toda chain equation (see also \[13,,14\]). We refer to a quantization in these variables as a quantization in $`\gamma `$-representation. Recently the $`\gamma `$-representation was employed to quantize “real” KdV theory <sup>1</sup> The monodromy matrix in the “real” KdV model has the form (2.1), whereas the monodromy matrix of the “imaginary” equation belongs to the group $`SU(2)(\mathrm{}m\lambda =0)`$. The imaginary KdV model is related to the sine-Gordon theory and (perturbed) CFT with the central charge $`c<1`$ (see e.g. ). A sensible $`\gamma `$-representation for the imaginary equation has, to our knowledge, not been found.. In fact, Smirnov presented heuristic, but convincing, model-independent arguments which can also be applied to the sinh-Gordon equation. Following these arguments we introduce the integral,
$$_N(R,a)=\underset{k=N}{\overset{N}{}}_{\mathrm{}}^+\mathrm{}\frac{d\gamma _k}{b}\underset{Nk>mN}{}\mathrm{sinh}(\gamma _k\gamma _m)\underset{k=N}{\overset{N}{}}𝒬^2[\gamma _k]e^{\frac{2\gamma _k}{b^2}(ab+k)}.$$
We shall also use slightly different form of $`_N`$: With the identity,
$$\underset{Nk>mN}{}2\mathrm{sinh}\left(\frac{\gamma _k\gamma _m}{b^2}\right)=\mathrm{Det}\left|\mathrm{exp}\left(\frac{2k\gamma _j}{b^2}\right)\right|_{Nj,kN},$$
the integral (2.1) can be rewritten in the form
$$\begin{array}{cc}\hfill _N(R,a)& =\frac{1}{(2N+1)!}\underset{k=N}{\overset{N}{}}_{\mathrm{}}^+\mathrm{}\frac{d\gamma _k}{b}\times \hfill \\ & \underset{Nk>mN}{}2\mathrm{sinh}(\gamma _k\gamma _m)\mathrm{sinh}\left(\frac{\gamma _k\gamma _m}{b^2}\right)\underset{k=N}{\overset{N}{}}𝒬^2[\gamma _k]e^{\frac{2\gamma _ka}{b}}.\hfill \end{array}$$
The function $`𝒬[\gamma ]`$ appearing in Eqs.(2.1), (2.1) is the so-called Baxter’s $`Q`$-function. It is a non-singular function for real $`\gamma `$ with leading asymptotic behavior
$$𝒬[\gamma ]e^{2C_0\mathrm{cosh}(bq\gamma )}\mathrm{as}\gamma \mathrm{}.$$
Here and below we use the notation,
$$q=b+b^1.$$
The positive constant $`C_0`$ in (2.1) reads explicitly,
$$C_0=\frac{mR}{4\mathrm{sin}\left(\frac{\pi b}{q}\right)},$$
and $`m`$ is a mass of the sinh-Gordon particle. Therefore, $`_N`$ (2.1) is a convergent integral for any finite $`N`$. Notice that the configuration space of the model under consideration is an infinite-dimensional space. In writing the $`2N+1`$-fold integral, we truncate it to $`2N+1`$-dimensional space with coordinates $`\left\{\gamma _k\right\}_{k=N}^N`$. As well as in the KdV theory , we can treat
$$\mathrm{\Psi }_N\left[\{\gamma _k\}\right]=\underset{k=N}{\overset{N}{}}𝒬[\gamma _k]$$
as a wave functional in the $`\gamma `$-representation. For $`N\mathrm{}`$, it corresponds to the ground state $`|vac_R`$ of the sinh-Gordon theory with periodic boundary conditions. Furthermore, the double product in (2.1) is an integration measure and we shall denote it $`𝒟_N\left[\{\gamma \}\right]`$. It was calculated in the semi-classical approximation in . The semi-classical analysis suggests also that the product
$$𝒪_N=\underset{k=N}{\overset{N}{}}e^{\frac{2\gamma _ka}{b}}$$
represents (in the limit $`N\mathrm{}`$) the exponential field $`e^{a\phi }`$ located at the point $`(R/2,R/2)`$ on the cylinder<sup>2</sup> The position of the insertion is determined by choosing of the integration contour for the monodromy matrix. Here we assume that the contour is given by (2.1).. Therefore, the integral (2.1) has the form of a quantum mechanical diagonal matrix element,
$$_N=𝒟_N\left[\{\gamma \}\right]\mathrm{\Psi }_N^{\mathrm{}}𝒪_N\mathrm{\Psi }_N.$$
We shall consider the Vacuum Expectation Values only. In this case, $`𝒬`$ is an eigenvalue of the Baxter $`Q`$-operator corresponding to the ground state.
2.3. Baxter’s $`Q`$-function in sinh-Gordon theory
We now turn to the most delicate point of our construction: an explicit form of the function $`𝒬`$. To the best of our knowledge a rigorous derivation of $`𝒬`$ is not currently available. Here we formulate a conjecture for the sinh-Gordon $`Q`$-function based on the following heuristic arguments. First, we note that the substitution $`bi\beta `$ transforms $`𝒜_{\mathrm{shG}}`$ (1.1) to the action of the sine-Gordon model. Naively we could try to obtain $`𝒬`$ by means of analytical continuation from $`1<b^2<0`$. In this coupling constant domain, the Baxter $`Q`$-operator is relatively well studied . Unfortunately, the sine-Gordon $`Q`$-function has an essential singularity at $`b^2=0`$ the analytical structure of which is unknown. This makes the analytical continuation to the domain of the sinh-Gordon model a highly questionable procedure. One can guess an explicit form for $`𝒬`$ by examining its asymptotic behavior. The $`Q`$-function in the sine-Gordon model admits the following asymptotic expansion for $`\gamma +\mathrm{}`$ <sup>3</sup> In this work we use a convention for the $`Q`$-function which differs from the one of by an overall shift of the argument.,
$$\mathrm{log}Q_{sinG}C_0e^\theta +\underset{n=1}{\overset{\mathrm{}}{}}C_nI_{2n1}e^{(2n1)\theta }+\underset{n=1}{\overset{\mathrm{}}{}}\stackrel{~}{G}_ne^{2n\theta (b^2+1)}(1<b^2<0).$$
Here we introduce a new notation,
$$\theta \gamma bq.$$
The leading term of this expansion has already appeared in our consideration (see (2.1)) while $`I_{2n1}`$ and $`\stackrel{~}{G}_n`$ are vacuum eigenvalues of the so-called local and dual unlocal Integrals of Motion (IM) respectively. The constants $`C_n`$ depend on the normalization of the local IM. In Appendix A (see Eq.(A.2)) we present their form for the normalization adopted in . A similar asymptotic form holds for $`\gamma \mathrm{}`$. The eigenvalues $`I_{2n1}`$ are regular functions of $`b^2`$, and can be continued to the domain of the sinh-Gordon model without problems. Contrary to $`I_{2n1}`$, the eigenvalues of the dual unlocal IM, $`\stackrel{~}{G}_n`$, are highly singular functions at $`b^2=0`$. One can expect that the appearance of these IM is a consequence of the existence of the soliton sector of the sine-Gordon QFT. This sector is absent for $`b^2>0`$. All these observations suggest the following large $`\gamma `$ asymptotic behavior in the sinh-Gordon model,
$$\mathrm{log}𝒬C_0e^\theta \underset{n=1}{\overset{\mathrm{}}{}}C_nI_{2n1}e^{(2n1)\theta }(b^2>0).$$
In Appendix A we give numerical evidence that the values of local IM can be expressed in terms of a solution of the Thermodynamic Bethe Ansatz (TBA) equation:
$$C_nI_{2n1}=C_0\delta _{n1}+(1)^n_{\mathrm{}}^{\mathrm{}}\frac{d\theta }{\pi }e^{(2n1)\theta }\mathrm{log}\left(\mathrm{\hspace{0.17em}1}+e^{ϵ(\theta )}\right).$$
Here the function $`ϵ(\theta )`$ solves the TBA equation \[16,,17,,18\]
$$ϵ(\theta )mR\mathrm{cosh}(\theta )+_{\mathrm{}}^{\mathrm{}}\frac{d\theta ^{}}{2\pi }\mathrm{\Phi }(\theta \theta ^{})\mathrm{log}\left(1+e^{ϵ(\theta ^{})}\right)=0,$$
with the kernel
$$\mathrm{\Phi }(\theta )=\frac{4\mathrm{sin}\left(\frac{\pi b}{q}\right)\mathrm{cosh}(\theta )}{\mathrm{cosh}(2\theta )\mathrm{cos}\left(\frac{2\pi b}{q}\right)}.$$
The series (2.1) is an asymptotic expansion. In fact, it is a divergent geometrical series which can easily be summed up and one can guess an explicit form of $`𝒬`$:
$$\mathrm{log}𝒬(\theta )=\frac{mR}{2\mathrm{sin}\left(\frac{\pi b}{q}\right)}\mathrm{cosh}(\theta )+_{\mathrm{}}^{\mathrm{}}\frac{d\theta ^{}}{2\pi }\frac{\mathrm{log}\left(1+e^{ϵ(\theta ^{})}\right)}{\mathrm{cosh}(\theta \theta ^{})}.$$
We leave an examination of the properties of this function for future publications.
One more aspect of the $`\gamma `$-representation deserves a comment. The sinh-Gordon model manifests an important non-perturbative symmetry. The couplings $`b`$ and $`b^1`$ correspond to physically indistinguishable theories. $`𝒬`$ in (2.1) is a self-dual function in a sense that it is invariant under the substitution $`bb^1`$. Furthermore, it is easy to see that
$$_N|_b=_N|_{b^1}.$$
This supports our choice of the measure in (2.1). Strictly speaking, the measure was obtained in in the semi-classical approximation. The exact invariance of the semi-classical measure suggests its applicability for an arbitrary value of the coupling constant $`b^2`$.
2.4. Large $`N`$ limit
As was noted above, in writing the $`2N+1`$-fold integral (2.1), we truncate the configuration space of the theory to $`2N+1`$-dimensional one. In fact, the truncation amounts to an ultraviolet regularization with a momentum cutoff given by
$$\mathrm{\Lambda }_N=\frac{2\pi N}{R}.$$
Now we let $`N\mathrm{}`$. From (2.1), it is clear that the ratio
$$\overline{}_N(R,a)=\frac{_N(R,a)}{_N(R,0)}$$
represents the VEV $`e^{a\phi }_R`$ in the large $`N`$ limit. More explicitly, dimensional analysis suggests that
$$\overline{}_N(R,a)\left(\frac{\mathrm{\Lambda }_N}{m}\right)^{2a^2},$$
thus the correct relation has the form
$$e^{a\phi }_R=\kappa _a\mathrm{lim}_N\mathrm{}\left(\frac{4\pi N}{mR}\right)^{2a^2}\overline{}_N(R,a).$$
Here $`\kappa _a`$ is an arbitrary $`R`$-independent constant. To eliminate this ambiguity one has to impose some normalization condition on the fields. For example, the so-called conformal normalization stipulates that the exponential fields with sufficiently small $`|a|`$ are normalized in accordance with the short distance behavior of the two-point function
$$e^{a\phi }(x)e^{a\phi }(y)|xy|^{4a^2}\mathrm{as}|xy|0.$$
The key result of Refs.\[3,,4\] is a calculation of the limit,
$$\underset{R\mathrm{}}{lim}e^{a\phi }_R=𝒢_a$$
with this normalization. Explicitly,
$$\begin{array}{cc}\hfill 𝒢_a=& \left[\frac{m\mathrm{\Gamma }\left(\frac{1}{2bq}\right)\mathrm{\Gamma }\left(1+\frac{b}{2q}\right)}{4\sqrt{\pi }}\right]^{2a^2}\times \hfill \\ & \mathrm{exp}\{_0^{\mathrm{}}\frac{dt}{t}[\frac{\mathrm{sinh}^2(2abt)}{2\mathrm{sinh}(b^2t)\mathrm{sinh}(t)\mathrm{cosh}(qbt)}+2a^2e^{2t}]\}.\hfill \end{array}$$
Once we adopt the conformal normalization for the exponential fields, the constant $`\kappa _a`$ is uniquely determined by the condition (2.1). In particular, the semi-classical consideration (see below) leads to the relation,
$$\kappa _a=𝒢_a\left(\mathrm{\hspace{0.17em}1}+O(b^2)\right).$$
3. Semi-classical expansion
In this section we will study the VEVs in the semi-classical approximation. The VEV (2.1) can be represented by the Euclidean path integral (1.1) on a cylinder with $`𝒜`$ and $`𝒪`$ given by (1.1), (1.1). For fixed $`\alpha `$,
$$\alpha =ab$$
and $`b^20`$, the path integral is dominated by a saddle-point configuration $`\varphi =b\phi `$ and the VEV has the form (1.1), where $`S`$ coincides with the regularized Euclidean classical action on the cylinder:
$$S=\underset{\epsilon 0}{lim}\left[_{|xy|>\epsilon }\frac{d^2x}{8\pi }\left\{\frac{(_\sigma \varphi )^2}{2}+m^2\mathrm{sinh}^2\left(\frac{\varphi }{2}\right)\right\}+\alpha _{|xy|=\epsilon }\frac{ds}{2\pi }\varphi 2\alpha ^2\mathrm{log}\epsilon \right].$$
Here the function $`\varphi `$ is a solution of the classical equation of motion,
$$_\sigma ^2\varphi =m^2\mathrm{sinh}(\varphi ),$$
such that,
$$\varphi \mathrm{\hspace{0.17em}4}\alpha \mathrm{log}|xy|+O(1)\mathrm{as}|xy|0,$$
and
$$\begin{array}{cc}& \varphi 0\mathrm{as}|xy|\mathrm{},\hfill \\ & \varphi (x_1,x_2+R)=\varphi (x_1,x_2).\hfill \end{array}$$
The field configuration $`\varphi `$ develops a singularity at the point $`y`$ where the exponential field is inserted. Therefore, in the definition (3.1) we cut the small disc of radius $`\epsilon `$ around this point and add the boundary term to the action to ensure (3.1). We also add a field independent term such that the action is finite at $`\epsilon 0`$. The pre-exponential factor $`D`$ (1.1) is a result of evaluating the path integral (1.1) in the Gaussian approximation around the classical solution defined above,
$$D=\left(\frac{m\epsilon }{2}e^{\gamma _E}\right)^{2\alpha ^2}\left[\mathrm{Det}^{}\left(\frac{_\sigma ^2+m^2\mathrm{cosh}(\varphi )}{_\sigma ^2+m^2}\right)\right]^{\frac{1}{2}},$$
where $`\gamma _E=0.577216\mathrm{}`$ is the Euler constant. The first factor in (3.1) appears as a result of the mass renormalization.
3.1. Main semi-classical order
We now proceed to the semi-classical calculation of the VEV $`e^{a\phi }`$ using the representation (2.1). In order to apply the saddle-point machinery, it is convenient to begin with the form (2.1) for the integral $`_N`$. To the lowest semi-classical order,
$$𝒬[\gamma ]e^{\frac{r\mathrm{cosh}(\gamma )}{2\pi b^2}}.$$
Here and below the notation
$$r=mR$$
is used. The corresponding saddle-point equations have the form,
$$r\mathrm{sinh}(\rho _k)=2\pi (k+\alpha ),k=0,\pm 1,\mathrm{},\pm N.$$
In writing (3.1) we have assumed that $`b^20`$ and $`\alpha =ab`$ is fixed. Thus we find,
$$\mathrm{log}_N=\frac{1}{b^2}\underset{k=N}{\overset{N}{}}\left\{\frac{r}{\pi }\mathrm{cosh}(\rho _k)2\rho _k(k+\alpha )\right\}+O(1).$$
The sum can be evaluated with the result,
$$\begin{array}{cc}\hfill & \underset{k=N}{\overset{N}{}}\left\{\frac{r}{\pi }\mathrm{cosh}(\rho _k)2\rho _k(k+\alpha )\right\}=T\left(\frac{r}{2\pi }\right)^2\mathrm{log}\left(\frac{4\pi Ne}{r}\right)+2\alpha ^2\mathrm{log}N+\hfill \\ & \left(\mathrm{\hspace{0.17em}2}N(N+1)+\frac{1}{3}+2\alpha ^2\right)\mathrm{log}\left(\frac{4\pi }{r}\right)+4\underset{k=1}{\overset{N}{}}k\mathrm{log}(k/e)4\mathrm{log}A_G+O\left(\frac{1}{N}\right),\hfill \end{array}$$
where $`A_G=1.282427\mathrm{}`$ is the Glaisher constant and the function $`T=T(r,\alpha )`$ reads explicitly,
$$T=r_{\mathrm{}}^{\mathrm{}}\frac{d\tau }{\pi ^2}\tau \mathrm{sinh}(\tau )\mathrm{}e\left[\mathrm{log}\left(\mathrm{\hspace{0.17em}1}e^{r\mathrm{cosh}(\tau )+2\pi i\alpha }\right)\right].$$
Now, combining Eqs.$`(2.1),(3.1),(3.1)`$ one obtains
$$e^{a\phi }_Re^{\frac{S}{b^2}},$$
with
$$S=S_0(\alpha )T(r,\alpha )+T(r,0),$$
and
$$S_0=2\alpha ^2\mathrm{log}\left(\frac{m}{4}\right)+_0^{\mathrm{}}\frac{dt}{t}\left\{\frac{\mathrm{sinh}^2(2\alpha t)}{t\mathrm{sinh}(2t)}2\alpha ^2e^{2t}\right\}.$$
Thus the function (3.1) coincides with the regularized Euclidean action (3.1)<sup>4</sup> This result was obtained by a different method in ..
3.2. Semi-classical expansion of Q-function
To compute the VEV to one-loop order, we have to find the next term in the semi-classical expansion of $`𝒬`$. It can be obtained by means of the TBA equation (2.1). The kernel $`\mathrm{\Phi }`$ (2.1) allows an expansion in $`b^2`$,
$$\mathrm{\Phi }(\theta )=2\pi \delta (\theta )+2\pi b^2P.V.\frac{\mathrm{cosh}(\theta )}{\mathrm{sinh}^2(\theta )}+O(b^4),$$
thus to lowest order the solution of the TBA equation has the form,
$$e^ϵ=e^{r\mathrm{cosh}(\theta )}1+O(b^2).$$
With this equation and the definition (2.1) we calculate
$$\mathrm{log}𝒬[\gamma ]=\frac{r\mathrm{cosh}(\gamma )}{2\pi b^2}\frac{r\mathrm{cosh}(\gamma )}{2\pi }+\frac{r\gamma \mathrm{sinh}(\gamma )}{2\pi }_{\mathrm{}}^{\mathrm{}}\frac{d\tau }{2\pi }\frac{\mathrm{log}\left(1e^{r\mathrm{cosh}(\tau )}\right)}{\mathrm{cosh}(\gamma \tau )}+O(b^2).$$
In order to see an analytical structure of this function it is instructive to represent $`𝒬`$ in the form of an infinite product,
$$\begin{array}{cc}\hfill \frac{1}{𝒬[\gamma ]}=& e^{\frac{r\mathrm{cosh}(\gamma )}{2\pi b^2}}\left(\frac{re^{\gamma _E}}{4\pi }\right)^{\frac{r\mathrm{cosh}(\gamma )}{2\pi }}\sqrt{2r}\times \hfill \\ & \mathrm{cosh}\left(\frac{\gamma }{2}\right)\underset{n=1}{\overset{\mathrm{}}{}}\left\{\sqrt{1+\left(\frac{r}{2\pi n}\right)^2}+\frac{r\mathrm{cosh}(\gamma )}{2\pi n}\right\}e^{\frac{r\mathrm{cosh}(\gamma )}{2\pi n}}\left(\mathrm{\hspace{0.17em}1}+O(b^2)\right).\hfill \end{array}$$
Here $`\gamma _E`$ is the Euler constant.
3.3. One-loop order
The saddle-point approximation allows one to find the one-loop order in the semi-classical expansion. To this order,
$$_N=W\underset{Nk>mN}{}\mathrm{sinh}(\rho _k\rho _m)\underset{k=N}{\overset{N}{}}𝒬^2[\rho _k]e^{2\rho _k(\alpha +k)}\left(\mathrm{\hspace{0.17em}1}+O(b^2)\right).$$
Here the function $`W`$ is a result of the Gaussian integrations in (2.1) around the saddle points $`\gamma _k=\rho _k`$ (3.1),
$$W=(2\pi ^2)^{N+\frac{1}{2}}\underset{k=N}{\overset{N}{}}\frac{1}{\sqrt{r\mathrm{cosh}(\rho _k)}}.$$
The main steps in the calculation of (3.1) are given in Appendix B. Our final result has the form (1.1) with the function $`S`$ given by (3.1) and
$$\begin{array}{cc}\hfill \mathrm{log}D=& \mathrm{log}D_0+T(r,\alpha )T(r,0)\alpha _\alpha T(r,\alpha )\hfill \\ & _r^{\mathrm{}}\frac{dr}{8}r\left\{\left(_r_\alpha T\right)^2\frac{1}{2\pi ^2}\left(_\alpha ^2T|_{\alpha =0}_\alpha ^2T\left(_\alpha ^2T\right)^2|_{\alpha =0}\right)\right\}.\hfill \end{array}$$
Here
$$\mathrm{log}D_0=2\alpha ^2\mathrm{log}2+\frac{1}{2}_0^{\mathrm{}}𝑑t\frac{\mathrm{sinh}^2(2\alpha t)}{\mathrm{cosh}^2(t)}.$$
Recall that (3.1) should coincide with the functional determinant (3.1).
4. High-temperature behavior
Now that we have computed (1.1), let us check the result for some limiting cases. Here we argue for $`R0`$ behavior of the VEVs.
Due to the scaling properties of the interaction operator in (1.1) one can rescale the problem to a circle of circumference $`2\pi `$. Thus, the Hamiltonian of the model under consideration takes the form
$$𝐇_{\mathrm{shG}}=\frac{2\pi }{R}_0^{2\pi }𝑑\tau \left\{\mathrm{\hspace{0.17em}4}\pi \mathrm{\Pi }^2+\frac{1}{16\pi }(_\tau \chi )^2+\mu \left(\frac{R}{2\pi }\right)^{2bq}\left(e^{b\chi }+e^{b\chi }\right)\right\},$$
where $`\mathrm{\Pi }=\frac{1}{i}\frac{\delta }{\delta \chi }`$ is the momentum conjugate to $`\chi =\phi |_{x_1=0}`$. The mass of the sinh-Gordon particle is related to the parameter $`\mu `$ by ,
$$\mu =\frac{\mathrm{\Gamma }(b^2)}{\pi \mathrm{\Gamma }(1+b^2)}\left[\frac{m\mathrm{\Gamma }(\frac{1}{2bq})\mathrm{\Gamma }(1+\frac{b}{2q})}{4\sqrt{\pi }}\right]^{2bq}.$$
For $`r0`$ and $`a>0`$ the main contribution in the path integral (1.1) comes from a region of the configuration space corresponding to
$$\chi 2q\mathrm{log}(r)1.$$
In this region we can neglect the term $`e^{b\chi }`$ in the Hamiltonian (4.1), and approximate the ground state wave functional $`\mathrm{\Psi }_0[\chi ]`$ by a proper wave functional from the Liouville Conformal Field Theory (CFT).
More explicitly, the Hilbert space of the Liouville CFT contains a continuous set of primary states $`|p`$ parameterized by $`p>0`$ with the conformal dimension ,
$$\mathrm{\Delta }_p=p^2+\frac{q^2}{4}.$$
We will assume that these states are canonically normalized,
$$p^{}|p=2\pi \delta (pp^{}).$$
Let $`\mathrm{\Psi }_p[\chi ]`$ be a normalized wave functional corresponding to the state $`|p.`$ As was discussed in \[17,,18\], the following relation holds
$$\mathrm{\Psi }_0[\chi ]L_p\mathrm{\Psi }_p[\chi ]$$
in the region (4.1). Here $`p=p(r)`$ solves the equation,
$$p(r):\mathrm{\hspace{0.17em}2}pq\mathrm{log}\left[\frac{r\mathrm{\Gamma }(\frac{1}{2bq})\mathrm{\Gamma }(1+\frac{1}{2bq})}{8\pi ^{\frac{3}{2}}(b^2)^{\frac{1}{bq}}}\right]=\frac{\pi }{2}+\mathrm{}m\left[\mathrm{log}\left\{\mathrm{\Gamma }(1+2ip/b)\mathrm{\Gamma }(1+2ipb)\right\}\right].$$
We emphasize that $`L_p`$ is a unique coefficient provided a normalization of $`\mathrm{\Psi }_0`$ is chosen. Therefore, one can expect the following relation for $`r1`$:
$$e^{a\phi }_RL_pL_p\frac{p|e^{a\phi }|p_{\mathrm{Liouv}}}{{}_{R}{}^{}vac|vac_{R}^{}}.$$
The matrix element $`p|e^{a\phi }|p_{\mathrm{Liouv}}`$ was found in \[21,,18\]. It reads explicitly,
$$p|e^{a\phi }|p_{\mathrm{Liouv}}=\left(\frac{R}{2\pi }\right)^{2a(qa)}\left[\frac{\pi \mu \mathrm{\Gamma }(b^2)b^{22b^2}}{\mathrm{\Gamma }(1b^2)}\right]^{a/b}\frac{\mathrm{{\rm Y}}_0\mathrm{{\rm Y}}(2a)\mathrm{{\rm Y}}(2ip)\mathrm{{\rm Y}}(2ip)}{\mathrm{{\rm Y}}^2(a)\mathrm{{\rm Y}}(a+ip)\mathrm{{\rm Y}}(aip)}.$$
Here we use the notations
$$\mathrm{log}\mathrm{{\rm Y}}(a)=\left(\frac{q}{2}a\right)^2\mathrm{log}(2b)+_0^{\mathrm{}}\frac{dt}{t}\left[\left(\frac{q}{2}a\right)^2e^{2t}\frac{\mathrm{sinh}^2\left((qb2ab)t\right)}{\mathrm{sinh}(2t)\mathrm{sinh}(2tb^2)}\right],$$
and
$$\mathrm{{\rm Y}}_0=_a\mathrm{{\rm Y}}(a)|_{a=0}.$$
It is easy to see that the function $`p=p(r)`$ (4.1) satisfies the condition,
$$\underset{r0}{lim}p(r)=0.$$
Using Eqs.(4.1)-(4.1), we can derive
$$e^{a\phi }_R𝒩^2\left[\frac{m\mathrm{\Gamma }(\frac{1}{2bq})\mathrm{\Gamma }(1+\frac{b}{2q})}{4\sqrt{\pi }}\right]^{2aq}b^{2aq}\left(\frac{R}{2\pi }\right)^{2a(aq)}\frac{\mathrm{{\rm Y}}(2a)\mathrm{{\rm Y}}_0^3}{\mathrm{{\rm Y}}^4(a)},$$
where
$$𝒩^2=\frac{\underset{p0}{lim}\left\{\mathrm{\hspace{0.17em}4}p^2L_pL_p\right\}}{{}_{R}{}^{}vac|vac_{R}^{}}$$
does not depend on $`a`$. In writing (4.1) we also used the relation (4.1). Unfortunately, the function $`𝒩=𝒩(r,b)`$ is not known in closed form for an arbitrary value of the coupling constant. One can obtain its limiting value as $`b^20`$. For $`b^2r1`$, it is sufficient to consider the dynamics of the zero-mode $`X`$ \[17,,18\]:
$$X=_0^{2\pi }\frac{d\tau }{2\pi }\chi (\tau ).$$
In this approximation, known as the mini-superspace approach , the Hamiltonian (4.1) is substituted by,
$$𝐇_{ms}=\frac{2\pi }{R}\left\{2_X^2+\left(\frac{r}{4\pi b}\right)^2\mathrm{cosh}(bX)\right\}.$$
The Schr$`\ddot{\mathrm{o}}`$dinger equation
$$𝐇_{ms}\mathrm{\Psi }_0(X)=E_{ms}\mathrm{\Psi }_0(X),$$
coincides with the modified Mathieu equation and the wave functional $`\mathrm{\Psi }_0`$ is represented by its lowest eigenfunction. We will use the common normalization condition
$$_{\mathrm{}}^{\mathrm{}}𝑑X\mathrm{\Psi }_0^2(X)=1.$$
The Liouville wave functionals $`\mathrm{\Psi }_p`$ in the mini-superspace approximation have the form,
$$\mathrm{\Psi }_p(X)=\frac{2}{\mathrm{\Gamma }(2ip/b)}\left(\frac{r}{8\pi b^2}\right)^{2ip/b}K_{\frac{2ip}{b}}\left(\frac{r}{4\pi b^2}e^{\frac{bX}{2}}\right).$$
Here $`K_\nu (z)`$ is the MacDonald function. Now it is clear that the mini-superspace approximation for $`𝒩`$ (4.1) can be obtained from the large $`X`$ behavior of the normalized Mathieu function $`\mathrm{\Psi }_0(X)`$ (4.1),
$$\mathrm{\Psi }_0(X)\sqrt{\frac{2}{r}}\frac{\pi 𝒩_{ms}}{\mathrm{cosh}(bX/4)}\mathrm{exp}\left\{\frac{r}{2\pi b^2}\mathrm{cosh}(bX/2)\right\}\mathrm{as}X\pm \mathrm{}.$$
Of concern to us is the behavior of $`𝒩_{ms}`$ in the domain $`b^2r1`$. In this case, we replace $`\mathrm{\Psi }_0(X)`$ by its WKB asymptotic (4.1) and readily obtain,
$$\underset{b^20}{lim}𝒩^2=\frac{\sqrt{2}r^{\frac{3}{2}}}{(2\pi )^3},r1.$$
Having arrived at Eq. (4.1), we can straightforwardly expand (4.1) in a power series of $`b^2`$,
$$e^{a\phi }_RF\left(\mathrm{\hspace{0.17em}1}+O(b^2)\right)(b^2r1),$$
with
$$\begin{array}{cc}\hfill F=& \left(\frac{R}{2\pi }\right)^{\frac{2\alpha (\alpha 1)}{b^2}}\mathrm{\hspace{0.17em}2}^{\frac{2\alpha }{b^2}}\mathrm{exp}\left\{\frac{1}{2b^2}(4S_0(1/2\alpha )S_0(1/22\alpha ))\right\}\times \hfill \\ & 2^{2+\frac{1}{2b^2}}m^{\frac{3}{4b^2}}A_G^{\frac{9}{b^2}}\left(\frac{r}{4\pi }\right)^{\frac{3}{2}2\alpha }\sqrt{\frac{\mathrm{\Gamma }(12\alpha )}{\mathrm{\Gamma }(2\alpha )}}\frac{\mathrm{\Gamma }^2(\alpha )}{\mathrm{\Gamma }^2(1\alpha )},\hfill \end{array}$$
where $`S_0`$ is given by Eq.(3.1) and $`A_G`$ is the Glaisher constant. It is possible to show that the high-temperature expansion of (1.1) exactly matches (4.1) (see Appendix B for some details).
5. Low-temperature expansion
We have mentioned in the Introduction that the finite volume VEVs can be understood as thermal averages (1.1). Hence, $`e^{a\phi }_R`$ admits the low-temperature ($`R\mathrm{}`$) expansion in the form,
$$\mathrm{log}\left(e^{a\phi }_R/𝒢_a\right)=1+\underset{k=1}{\overset{\mathrm{}}{}}G_k(r).$$
Here $`G_k`$ represents $`k`$-particle contributions in the infinite-volume channel and
$$G_k(r)e^{kr}.$$
Recently A. Leclair and G. Mussardo proposed an integral representation which is sufficient to generate $`G_k(r)`$ systematically in terms of form-factors of the field $`e^{a\phi }`$ at $`R=\mathrm{}`$ and the solution of the TBA equation (2.1). Taking into account contributions of one- and two-particle states to the trace (1.1), they obtained
$$\begin{array}{cc}\hfill \mathrm{log}\left(e^{a\phi }_R/𝒢_a\right)=& 4[a]_{\mathrm{}}^{\mathrm{}}\frac{d\theta }{2\pi }f_{}(\theta )+\hfill \\ & [2a]_{\mathrm{}}^{\mathrm{}}\frac{d\theta _1}{2\pi }\frac{d\theta _2}{2\pi }\frac{\mathrm{\Phi }(\theta _1\theta _2)}{\mathrm{cosh}(\theta _1\theta _2)}f_{}(\theta _1)f_{}(\theta _2)+\mathrm{},\hfill \end{array}$$
where the notations
$$f_{}(\theta )=\frac{1}{1+e^{ϵ(\theta )}},$$
and
$$[a]=\frac{\mathrm{sin}^2(\frac{\pi a}{q})}{\mathrm{sin}(\frac{\pi b}{q})}$$
are used. The function $`\mathrm{\Phi }`$ is the kernel in the TBA equation (2.1). With (5.1) and the TBA equation one can calculate the first two terms in the low-temperature expansion (5.1):
$$\begin{array}{cc}& G_1=\frac{4[a]}{\pi }K_0(r),\hfill \\ & G_2=4[a]_{\mathrm{}}^{\mathrm{}}\frac{d\theta _1}{2\pi }\frac{d\theta _2}{2\pi }\left(\mathrm{\Phi }(\theta _1\theta _2)2\pi \delta (\theta _1\theta _2)\right)e^{r\mathrm{cosh}\theta _1+r\mathrm{cosh}\theta _2}+\hfill \\ & [2a]_{\mathrm{}}^{\mathrm{}}\frac{d\theta _1}{2\pi }\frac{d\theta _2}{2\pi }\frac{\mathrm{\Phi }(\theta _1\theta _2)}{\mathrm{cosh}(\theta _1\theta _2)}e^{r\mathrm{cosh}\theta _1+r\mathrm{cosh}\theta _2},\hfill \end{array}$$
where $`K_n(r)`$ is the MacDonald function. We now expand (5.1) as a power series in $`b^2`$,
$$G_1=4\left\{\frac{s^2(\alpha )}{b^2}+s^2(\alpha )\alpha s(2\alpha )+O(b^2)\right\}K_0(r),$$
where
$$s(\alpha )=\frac{\mathrm{sin}(\pi \alpha )}{\pi }.$$
To expand $`G_2`$ one needs to use Eq.(3.1),
$$\begin{array}{cc}\hfill G_2& =\left\{\frac{s^2(2\alpha )}{b^2}+s^2(2\alpha )2\alpha s(4\alpha )\right\}K_0(2r)\hfill \\ & 4s^2(\alpha )r^2\left(K_1^2(r)K_0^2(r)\right)s^2(2\alpha )r^2\left(K_2(r)K_0(r)K_1^2(r)\right)+O(b^2).\hfill \end{array}$$
It is quite straightforward to verify that the low-temperature expansion of (1.1) exactly reproduces (5.1) and (5.1).
6. Conclusion
The proposed $`\gamma `$-representation (2.1) is the main result of this paper. Its rigorous derivation has not yet been achieved. Although (2.1) are conjectures, the evidence presented in this paper appears to make it reasonable to take them as the starting point for further investigation.
One can expect that similar representations exist for non-minimal CFT, say, the Liouville theory and $`SL(2,R)/U(1)`$ non-compact $`\sigma `$-model. It may cast new light on many unsolved problems of $`2D`$ Quantum Gravity. In this connection an intriguing similarity between the integrals appeared in the $`\gamma `$-representation and Matrix Models of $`2D`$ Quantum Gravity \[23,,24,,25\] can be mentioned.
Acknowledgments
I am grateful to A.B. Zamolodchikov for interesting discussions. The research is supported in part by the DOE grant #DE-FG05-90ER40559.
Note added
After finishing this paper it was drawn to my attention that Al.B. Zamolodchikov independently introduced and studied the function (2.1) in Ref.. I am grateful to him for the communication of that paper, and sharing insights.
Appendix A.
The QFT defined by (1.1) possesses infinitely many local IM $`\widehat{I}_{2n1}`$ whose vacuum eigenvalues have appeared in the equation (2.1). They can be represented in the form,
$$\widehat{I}_{2n1}=_0^R\frac{dx_2}{2\pi }\left(T_{2n}(x_2+ix_1,x_2ix_1)+\mathrm{\Theta }_{2n2}(x_2+ix_1,x_2ix_1)\right),$$
where the local fields $`T_{2n}`$ and $`\mathrm{\Theta }_{2n2}`$ satisfy the continuity equations,
$$_{\overline{z}}T_{2n}(z,\overline{z})=_z\mathrm{\Theta }_{2n2}(z,\overline{z}).$$
Although a general expression for the densities $`T_{2n}`$, $`\mathrm{\Theta }_{2n2}`$ is not known, they are determined up to normalization by the commutativity conditions,
$$[\widehat{I}_{2n1},\widehat{I}_{2m1}]=0.$$
In Refs.\[15,,27\] the following normalization was adopted,
$$T_{2n}=2^{2n}(_z\phi )^{2n}+\mathrm{},$$
where omitted terms contain higher derivatives of $`\phi `$ and exponential fields. Notice that the condition (A.1) does not depend on the regularization scheme defining the composite field $`(_z\phi )^{2n}`$. With normalization (A.1) the constants $`C_n`$ (2.1) were found in Refs.\[28,,15\],
$$C_n=\frac{\mathrm{\Gamma }\left(\frac{(2n1)b}{2q}\right)\mathrm{\Gamma }\left(\frac{2n1}{2bq}\right)}{2\sqrt{\pi }n!q}\left[\frac{m\mathrm{\Gamma }\left(\frac{b}{2q}\right)\mathrm{\Gamma }\left(\frac{1}{2qb}\right)}{8q\sqrt{\pi }}\right]^{12n}.$$
The local IM $`\widehat{I}_{2n1}`$ are certain deformations of the local IM of the Liouville CFT. Let $`I_{2n1}(p)`$ be an eigenvalue of the Liouville local IM corresponding to the state $`|p`$ (4.1), while $`I_{2n1}`$ is the sinh-Gordon ground state eigenvalues of $`\widehat{I}_{2n1}`$. Eq.(4.1) suggests the following relation for $`r1`$:
$$I_{2n1}=I_{2n1}\left(p(r)\right)+O(r^{4qb},r^{4q/b}),$$
where $`p(r)`$ solves (4.1)<sup>5</sup> For $`n=1`$ this relation was discussed in Ref... The power corrections in $`r`$ (A.1) can be obtained by means of Conformal Perturbation Theory. Explicit forms of the functions $`I_{2n1}(p)`$ (for $`n=1,\mathrm{},8`$) are given in Appendix B of Ref.. Here we present only the first two of them,
$$\begin{array}{cc}& I_1(p)=\frac{2\pi }{R}\left(p^2\frac{1}{24}\right),\hfill \\ & I_3(p)=\left(\frac{2\pi }{R}\right)^3\left(p^4\frac{p^2}{4}+\frac{4b^4+17b^2+4}{960b^2}\right).\hfill \end{array}$$
In Tables 1-4 we list numerical values of the local IM $`I_{2n1}`$ for some $`0.01r1`$ and $`b^2=0.81`$ which were obtained by means of numerical solution of the TBA equation (2.1) with use of Eqs.(2.1) and (A.1). These data are compared against values of the Liouville local IM $`I_{2n1}\left(p(r)\right)`$. We consider the content of Tables 1-4 to be an impressive evidence in support of the relation (2.1).
$`r`$ $`I_1`$ $`I_1((p(r))`$ 1.0 -0.0059897196942029 -0.0059891933248581 0.8 -0.0123637695005731 -0.0123636397906571 0.6 -0.0183955019094376 -0.0183954816200409 0.4 -0.0242191017573388 -0.0242191003738075 0.2 -0.0301582132435655 -0.0301582132312211 0.1 -0.0335250692109914 -0.0335250692108919 0.01 -0.0381898149656469 -0.0381898149656469
$`r`$ $`I_3`$ $`I_3((p(r))`$ 1.0 0.01857760504756 0.01858088002375 0.8 0.01975951264163 0.01976027692024 0.6 0.02095100471765 0.02095111804696 0.4 0.02216988492643 0.02216989225147 0.2 0.02348269733500 0.02348269739676 0.1 0.02425825249985 0.02425825250033
$`r`$ $`I_5`$ $`I_5((p(r))`$ $`I_7`$ $`I_7((p(r))`$ 1.0 -0.02830178173 -0.02830149724 0.0731035360 0.0731032652 0.8 -0.02936366562 -0.02936359281 0.0750097817 0.0750097140 0.6 -0.03040878256 -0.03040877074 0.0768644292 0.0768644184 0.4 -0.03145618592 -0.03145618508 0.0787032289 0.0787032281 0.2 -0.03256438579 -0.03256438578 0.0806285556 0.0806285556
$`r`$ $`I_9`$ $`I_9((p(r))`$ $`I_{11}`$ $`I_{11}((p(r))`$ 1.0 -0.29532264 -0.29532199 1.73235 1.73235 0.8 -0.30116225 -0.30116209 1.75983 1.75983 0.6 -0.30680503 -0.30680500 1.78627 1.78627 0.4 -0.31236318 -0.31236317 1.81220 1.81220 0.2 -0.31814537 -0.31814537 1.83907 1.83907
Tables 1-4. Comparison of the LHS and RHS of equation (A.1) ($`b^2=0.81`$).
Appendix B.
Here we proceed with calculation of the products in (3.1) and give some technical hints on the study of their high-temperature behavior. First, let us consider the product,
$$e^{M_1}=W\underset{Nk>jN}{}\mathrm{sinh}(\rho _k\rho _m).$$
Using the relation
$$_r\rho _k=\frac{1}{r}\mathrm{tanh}(\rho _k),$$
which follows from the saddle-point equations (3.1), one obtains
$$_rM_1=\frac{1}{2r}\underset{k,m=N}{\overset{N}{}}\frac{\mathrm{cosh}(\rho _k\rho _m)}{\mathrm{cosh}(\rho _k)\mathrm{cosh}(\rho _m)}.$$
This sum can be rewritten in the form,
$$_rM_1=\frac{(2N+1)^2}{2r}+\frac{1}{2r}\left[\underset{k=N}{\overset{N}{}}\mathrm{tanh}(\rho _k)\right]^2.$$
Notice that the sum in (B.1) converges for $`N\mathrm{}`$,
$$\underset{N\mathrm{}}{lim}\underset{k=N}{\overset{N}{}}\mathrm{tanh}(\rho _k)=\frac{r}{2}_r_\alpha T+2\alpha .$$
To derive this formula we used (3.1) and the saddle-point equations (3.1). Thus we obtain,
$$\begin{array}{cc}\hfill M_1=& _N\frac{(2N+1)^2}{2}\mathrm{log}\left(\frac{r}{4\pi }\right)\hfill \\ & 2\alpha ^2\mathrm{log}\left(\frac{4\pi N}{r}\right)\alpha _\alpha T(r,\alpha )_r^{\mathrm{}}\frac{dr}{8}r\left(_r_\alpha T\right)^2+O\left(N^1\right).\hfill \end{array}$$
The constant $`_N`$ here does not depend on $`r`$. To find how it depends on $`\alpha `$, let us consider $`_\alpha M_1`$. A similar calculation leads to the equation,
$$\begin{array}{cc}& _\alpha M_1=\frac{1}{4}\left(r_r_\alpha T4\alpha \right)\left(_\alpha ^2T+4\mathrm{log}(4\pi N/r)\right)+\frac{2\pi }{r}\underset{m=N}{\overset{N}{}}\frac{\mathrm{tanh}(\rho _0)}{\mathrm{cosh}(\rho _0)+\mathrm{cosh}(\rho _m)}+\hfill \\ & \frac{2\pi }{r}\underset{m=N}{\overset{N}{}}\underset{k=1}{\overset{N}{}}\left[\frac{\mathrm{tanh}\left(\rho _k(\alpha )\right)}{\mathrm{cosh}(\rho _k(\alpha ))+\mathrm{cosh}(\rho _m)}\frac{\mathrm{tanh}\left(\rho _k(\alpha )\right)}{\mathrm{cosh}(\rho _k(\alpha ))+\mathrm{cosh}(\rho _m)}\right].\hfill \end{array}$$
It follows immediately from the last equation that
$$_\alpha M_1|_r\mathrm{}=4\alpha \mathrm{log}\left(\frac{4\pi N}{r}\right).$$
Therefore, we conclude that the constant $`_N`$ in (B.1) does not depend on $`\alpha `$. Notice that Eq.(B.1) is very convenient for studying the high-temperature limit $`r0`$. It is straightforward to show that for $`\alpha >0`$,
$$\begin{array}{cc}& e^{M_1}|_{r0}\left(\frac{4\pi }{r}\right)^{2N(N+1)}N^{2\alpha ^2}e^_N\times \hfill \\ & \frac{2^{2\alpha }}{\mathrm{\Gamma }(\frac{1}{2}+\alpha )}\sqrt{\frac{2\pi \mathrm{\Gamma }(\alpha )}{\mathrm{\Gamma }(1\alpha )}}e^{2\alpha ^2\gamma }\underset{k=1}{\overset{\mathrm{}}{}}\frac{\mathrm{\Gamma }^2\left(\frac{k+1}{2}\right)e^{\frac{2\alpha ^2}{k}}}{\mathrm{\Gamma }\left(\frac{k+1}{2}\alpha \right)\mathrm{\Gamma }\left(\frac{k+1}{2}+\alpha \right)}.\hfill \end{array}$$
To finish the calculation of (3.1) one needs to evaluate the product,
$$e^{M_2}=\underset{k=N}{\overset{N}{}}𝒬^2[\rho _k]e^{2\rho _k(\alpha +k)}.$$
The first two terms of the semi-classical expansion for $`𝒬`$ are given by (3.1). With the saddle-point equation (3.1) $`M_2`$ in (B.1) can be written as,
$$M_2=M_2^{}+M_2^{\prime \prime },$$
where
$$M_2^{}=\left(b^2+1\right)\underset{k=N}{\overset{N}{}}\left\{\frac{r}{\pi }\mathrm{cosh}(\rho _k)2\rho _k(k+\alpha )\right\}$$
and
$$M_2^{\prime \prime }=2_{\mathrm{}}^{\mathrm{}}\frac{d\tau }{2\pi }\mathrm{log}\left(1e^{r\mathrm{cosh}(\tau )}\right)\underset{k=N}{\overset{N}{}}\frac{1}{\mathrm{cosh}(\rho _k\tau )}.$$
The sum $`M_2^{}`$ is evaluated by means of Eq. (3.1). To calculate $`M_2^{\prime \prime }`$ we note that
$$_rM_2^{\prime \prime }=2\underset{k=N}{\overset{N}{}}\frac{1}{\mathrm{cosh}(\theta _k)}_{\mathrm{}}^{\mathrm{}}\frac{d\tau }{2\pi }\frac{m}{e^{r\mathrm{cosh}(\tau )}1}.$$
With the relations
$$\underset{k=N}{\overset{N}{}}\frac{1}{\mathrm{cosh}(\theta _k)}=\frac{r}{2\pi }\underset{k=N}{\overset{N}{}}_\alpha \theta _k=\frac{r}{4\pi }\left\{_\alpha ^2T+4\mathrm{log}\left(\frac{4\pi N}{r}\right)\right\}+O\left(\frac{1}{N}\right)$$
and
$$_{\mathrm{}}^{\mathrm{}}\frac{d\tau }{e^{r\mathrm{cosh}(\tau )}1}=\frac{1}{4}_\alpha ^2T|_{\alpha =0},$$
one obtains,
$$M_2^{\prime \prime }=_r^{\mathrm{}}\frac{dr}{16\pi ^2}r_\alpha ^2T|_{\alpha =0}_\alpha ^2T+_r^{\mathrm{}}\frac{dr}{4\pi ^2}r_\alpha ^2T|_{\alpha =0}\mathrm{log}\left(\frac{4\pi N}{r}\right).$$
We specify the integration constant here using the condition,
$$M_2^{\prime \prime }|_r\mathrm{}0,$$
which follows from the definition (B.1). The function $`T`$ (3.1) satisfies Laplace’s equation,
$$r^1_r\left(r_rT\right)+\frac{1}{4\pi ^2}_\alpha ^2T=0.$$
This allows one to calculate the second integral in (B.1). Thus we find,
$$M_2^{\prime \prime }=_r^{\mathrm{}}\frac{dr}{16\pi ^2}r_\alpha ^2T|_{\alpha =0}_\alpha ^2T+T|_{\alpha =0}+r_rT|_{\alpha =0}\mathrm{log}\left(\frac{4\pi N}{r}\right).$$
Finally we note that the most efficient way to study the $`r0`$ limit of $`M_2`$ (B.1) is based on the representation (3.1). It shows that for $`r0`$ with $`r\mathrm{cosh}(\gamma )`$ fixed,
$$𝒬[\gamma ]|_{r0}\frac{\mathrm{\Gamma }\left(1+\frac{r\mathrm{cosh}(\gamma )}{2\pi }\right)}{\sqrt{2r}\mathrm{cosh}(\gamma /2)}\left(\frac{r}{4\pi }\right)^{\frac{r\mathrm{cosh}(\gamma )}{2\pi }}\left(\mathrm{\hspace{0.17em}1}+O(b^2)\right).$$
Hence, examination of the product (B.1) at this limit creates no difficulties at all.
References
relax Patashinskii, A.Z. and Pokrovskii, V.N.: Fluctuation theory of phase transition, Oxford: Pergamon Press 1979 relax Tsvelik, A.M.: Quantum Field Theory and Condensed Matter Physics, Cambridge Univ. Press 1995 relax Lukyanov, S. and Zamolodchikov, A.: Exact expectation values of local fields in quantum sine-Gordon model. Nucl. Phys. B493, 571-587 (1997) relax Fateev, V., Lukyanov, S., Zamolodchikov, A. and Zamolodchikov, Al.: Expectation values of local fields in Bullough-Dodd model and integrable perturbed conformal field theories. Nucl. Phys. B516, 652-674 (1998) relax Smirnov, F.A.: Quasi-classical study of form-factors in finite volume. (Paris U., VI-VII). LPTHE-98-10, Feb 1998, pp.21 (hep-th/9802132) relax Leclair, A. and Mussardo, G.: Finite temperature correlation functions in integrable QFT. Nucl. Phys. B552, 624-642 (1999) relax Lukyanov, S. and Zamolodchikov, A.: unpublished relax Saleur, H.: A comment on finite temperature correlations in integrable QFT. Nucl. Phys. B567, 602-610 (2000) relax Sklyanin, E.K.: The quantum Toda chain. Lect. Notes in Physics 226, 196-233 (1985); J. Sov. Math. 31, 3417 (1985) relax Sklyanin, E.K.: Separation of variables. New Trends. Preprint UTMS 95-9 (solv-int/9504001) relax Flaschka, H. and McLaughlin, D.: Canonically conjugate variables for the Korteweg-de Vries equation and the Toda Lattice with periodic boundary conditions. Progress Theor. Phys. 55, 438-456 (1976) relax Faddeev, L.D. and Takhtajan, L.A.: Hamiltonian Method in the Theory of Solitons. New York: Springer 1987 relax Smirnov, F.A.: Structure of matrix elements in quantum Toda chain. J. Phys. A.: Math. Gen. 31 8953-8971 (1998) relax Lebedev, D. and Krachev, S.: Integral representation for the eigenfunctions of quantum periodic Toda chain. Preprint ITEP-TH-54/99 (hep-th/9911016) relax Bazhanov, V.V., Lukyanov, S.L. and Zamolodchikov, A.B.: Integrable structure of conformal field theory, I. Quantum KdV theory and thermodynamic Bethe Ansatz. Commun. Math. Phys. 177, 381-398 (1996); II. Q-operator and DDV equation. Commun. Math. Phys. 190, 247-278 (1997) ; III. The Yang-Baxter Relation. Commun. Math. Phys. 200, 297-324 (1999) relax Zamolodchikov, Al.B.: Thermodynamic Bethe ansatz in relativistic models: Scaling 3-state Potts and Lee-Yang models. Nucl. Phys. B342, 695-720 (1990) relax Zamolodchikov, Al.B.: Resonance factorized scattering and roaming trajectories. ENS-LPS-335 pp.27 (1991) relax Zamolodchikov, A.B. and Zamolodchikov, Al.B.: Structure Constants and Conformal Bootstrap in Liouville Field Theory. Nucl. Phys. B477, 577-605 (1996) relax Zamolodchikov, Al.B.: Mass scale in the Sine-Gordon model and its reductions. Int. J. Mod. Phys. A10, 1125-1150 (1995) relax Curtright, T. and Thorn, C.: Conformally invariant quantization of the Liouville Theory. Phys. Rev. Lett. 48, 1309-1313 (1982); Braaten, E., Curtright, T. and Thorn, C.: An exact operator solution of the Quantum Liouville Field Theory. Ann. Phys. 147, 365-416 (1983) relax Dorn, H. and Otto, H.-J.: On correlation functions for noncritical strings with $`c1,D1`$. Phys. Lett. B291, 39-43 (1992); Two and three point functions in Liouville Theory. Nucl. Phys. B429, 375-388 (1994) relax Seiberg, N.: Notes on Quantum Liouville Theory and Quantum Gravity. Prog. Theor. Phys. Supp. 102, 319-349 (1990) relax Brezin, E. and Kazakov, V.: Exactly Solvable Field Theories of Closed strings. Phys. Lett. B236, 144-150 (1990) relax Douglas, M. and Shenker S.: Strings in less than one-dimension. Nucl. Phys. B335, 635-654 (1990) relax Gross, D. and Migdal, A.: Nonperturbative two-dimensional Quantum Gravity. Phys. Rev. Lett. 64, 127-130 (1990) relax Zamolodchikov, Al.B.: On the Thermodynamic Bethe Ansatz equation in sinh-Gordon model. To appear relax Bazhanov, V.V., Lukyanov, S.L. and Zamolodchikov, A.B.: Quantum field theories in finite volume: Excited state energies. Nucl. Phys. B489 \[FS\], 487-531 (1997) relax Zamolodchikov, Al.B.: unpublished
|
warning/0005/math0005018.html
|
ar5iv
|
text
|
# Electron Wavefunctions and Densities for Atoms
## 1. Introduction and Results
Let $`V`$ be the Coulomb potential for an atom consisting of a nucleus of charge $`Z`$ (fixed at the origin) and $`N`$ electrons:
$$V(𝐱)=V(x_1,\mathrm{},x_N)=\underset{j=1}{\overset{N}{}}\frac{Z}{|x_j|}+\underset{1j<kN}{}\frac{1}{|x_jx_k|},$$
$$𝐱=(x_1,\mathrm{},x_N)^{3N},x_j=(x_{j,1},x_{j,2},x_{j,3})^3,j=1,\mathrm{},N,$$
(1.1)
and let $`H`$ be the corresponding $`N`$ \- electron Hamilton operator:
$`HH^N=\mathrm{\Delta }+V`$ (1.2)
with
$`\mathrm{\Delta }={\displaystyle \underset{j=1}{\overset{N}{}}}\mathrm{\Delta }_j,\mathrm{\Delta }_j={\displaystyle \underset{i=1}{\overset{3}{}}}{\displaystyle \frac{^2}{x_{j,i}^2}}`$
being the kinetic energy operator of the $`N`$ electrons. The quadratic form domain of $`H`$ is $`W^{1,2}(^{3N})`$, see Reed and Simon . Assume $`\psi L^2(^{3N})`$ is a real-valued normalised eigenfunction of the operator $`H`$:
$`(HE)\psi =0,\psi \psi _{L^2(^{3N})}=1.`$ (1.3)
It is known that then $`\psi `$ is continuous with bounded derivatives, and $`\psi W^{2,2}(^{3N})`$ (Kato ) and that $`\psi `$ is in fact analytic away from the singularities (in $`^{3N}`$) of $`V`$, since $`V`$ is here real analytic (see Hopf ). In this paper we derive various qualitative and quantitative properties of the wave function $`\psi `$ and of the corresponding one-electron density
$`\rho (x)={\displaystyle _{^{3(N1)}}}|\psi (x,x_2,\mathrm{},x_N)|^2dx_2\mathrm{}dx_N,x^3,`$ (1.4)
as well as of its spherical average ($`x=r\omega ,r=|x|,\omega =x/|x|𝕊^2`$)
$`\stackrel{~}{\rho }(r)`$ $`={\displaystyle _{𝕊^2}}\rho (r\omega )𝑑\omega `$
$`={\displaystyle _{𝕊^2}}{\displaystyle _{^{3(N1)}}}|\psi (r\omega ,x_2,\mathrm{},x_N)|^2dx_2\mathrm{}dx_Nd\omega ,r[0,\mathrm{}).`$ (1.5)
###### Remark 1.1.
Aside from Kato’s classical results (see Kato ), the local behaviour of electron wavefunctions has been investigated more recently by Hoffmann-Ostenhof et al. , . The electron density itself has been studied extensively in the large-$`Z`$-limit, see Lieb . Except for the spatial asymptotics, see Ahlrichs et al. , there are virtually no recent rigorous results on $`\rho `$ despite the fact that the density is the central object in various popular numerical approximation schemes, as Density Functional Theory (DFT) and all the various descendants of Hartree-Fock theory.
We now present our results.
###### Theorem 1.2.
Let $`\psi `$ be as in (1.3). For all $`R(0,\mathrm{})`$, there exists a constant $`C=C(R)`$ such that
$`\underset{𝐲B(𝐱,R)}{sup}|\psi (𝐲)|C\underset{𝐲B(𝐱,2R)}{sup}|\psi (𝐲)|\text{ for all }𝐱^{3N}.`$
###### Remark 1.3.
This result complements the result by Simon \[14, Thm. C.2.5 (C14)\] for the case of operators of the form (1.2), but with $`V`$ in the Kato-class $`K^{n,1}(^n)`$: for $`\delta [0,2)`$ $`(\delta =0:n3)`$,
$`VK^{n,\delta }(^n)\underset{ϵ0}{lim}\underset{x^n}{sup}{\displaystyle _{|xy|<ϵ}}{\displaystyle \frac{|V(y)|}{|xy|^{n2+\delta }}}𝑑y=0.`$
The Coulomb potential (1.1) is in $`K^{3N,\delta }(^{3N})`$ for all $`\delta [0,1)`$, but is not in $`K^{3N,1}(^{3N})`$.
We recall the definition of Hölder continuity:
###### Definition 1.4.
For $`\mathrm{\Omega }^n`$ an open set, $`k`$, and $`\alpha (0,1]`$, we say that the function $`u`$ belongs to $`C_{\text{loc}}^{k,\alpha }(\mathrm{\Omega })`$ whenever $`uC^k(\mathrm{\Omega })`$, and for all $`\beta ^n`$ with $`|\beta |=k`$, and all open balls $`B(x_0,r)\mathrm{\Omega }`$, we have
$`\underset{x,yB(x_0,r),xy}{sup}{\displaystyle \frac{|D^\beta u(x)D^\beta u(y)|}{|xy|^\alpha }}C(x_0,r).`$
As a consequence of the proof of Theorem 1.2 we get:
###### Proposition 1.5.
Let
$`F(𝐱)=F(x_1,\mathrm{},x_N)={\displaystyle \underset{j=1}{\overset{N}{}}}{\displaystyle \frac{Z}{2}}|x_j|+{\displaystyle \underset{1j<kN}{}}{\displaystyle \frac{1}{4}}|x_jx_k|.`$ (1.6)
Then the eigenfunction $`\psi `$ given in (1.3) can be represented as
$`\psi =e^F\varphi `$
with
$`\varphi C_{\text{loc}}^{1,\alpha }(^{3N})\text{ for all }\alpha (0,1).`$
###### Remark 1.6.
This result classifies the singularities of $`\psi `$ as those coming from $`F`$: $`\psi =\psi F+e^F\varphi `$. Kato proved that $`\psi `$ is bounded, but as the ground state of Hydrogen-like systems ($`N=1,E=Z^2/4`$, $`\psi (x)=c_0e^{Z|x|/2},x^3`$) shows, it is not in general continuous.
###### Remark 1.7.
The results of Theorem 1.2 and Proposition 1.5 easily generalise to the case of molecules: $`L`$ nuclei, of charge $`Z_l`$, fixed at $`R_l^3`$, $`l=1,\mathrm{},L`$, with corresponding $`N`$ \- electron Hamilton operator
$`H^{N,L}={\displaystyle \underset{j=1}{\overset{N}{}}}\left(\mathrm{\Delta }_j{\displaystyle \underset{l=1}{\overset{L}{}}}{\displaystyle \frac{Z_l}{|x_jR_l|}}\right)+{\displaystyle \underset{1j<kN}{}}{\displaystyle \frac{1}{|x_jx_k|}}.`$
We assume throughout when studying $`\rho `$ and $`\stackrel{~}{\rho }`$ that $`E`$ and $`\psi `$ in (1.3) are such that there exist constants $`C_0,\gamma >0`$ such that
$`|\psi (𝐱)|C_0e^{\gamma |𝐱|}\text{ for all }𝐱^{3N}.`$ (1.7)
For references on the exponential decay of eigenfunctions, see e. g. Simon .
###### Remark 1.8.
Inequality (1.7) holds when $`E<inf\sigma _{\text{ess}}(H^N)`$. In this case, we let $`\epsilon E_0^{N1}E`$ with $`E_0^{N1}`$ the ground state energy of the $`(N1)`$ \- electron operator:
$`H^{N1}={\displaystyle \underset{j=2}{\overset{N}{}}}\left(\mathrm{\Delta }_j{\displaystyle \frac{Z}{|x_j|}}\right)+{\displaystyle \underset{2j<kN}{}}{\displaystyle \frac{1}{|x_jx_k|}}.`$ (1.8)
By the HVZ-theorem (see Cycon et al. \[2, Theorem 3.7\], $`inf\sigma _{\text{ess}}(H^N)=E_0^{N1}`$, and so $`\epsilon >0`$ if $`E<inf\sigma _{\text{ess}}(H^N)`$. When we study $`H`$ in a symmetry sector, so that $`\psi `$ transforms according to this symmetry, then $`E_0^{N1}`$ stands for the groundstate energy of the ionized particle system described by the Hamiltonian $`H^{N1}`$ in the appropriate symmetry subspace as determined by the symmetry behaviour of $`\psi `$. For this case a modified version of the HVZ-theorem holds (see e. g. Reed and Simon \[13, Thm. XIII. 17’\] and Zhislin and Sigalov ). This includes in particular the physically important case of real atoms (Pauli principle). So if $`E`$ lies below the beginning of the essential spectrum of $`H`$ considered in a symmetry sector, then analogously to the above the ionisation energy $`\epsilon >0`$ and $`\psi `$ satisfies (1.7).
###### Remark 1.9.
When assuming (1.7), Theorem 1.2 implies that $`|\psi (𝐱)|`$ also decays exponentially for $`|𝐱|\mathrm{}`$.
###### Remark 1.10.
From (1.7) and Lebesgue’s Dominated Convergence Theorem follows that the density $`\rho `$ is continuous in $`^3`$.
###### Theorem 1.11.
Let $`\psi `$ be given according to (1.3) and assume that (1.7) holds. Then:
1. The function $`\rho `$ defined in (1.4) satisfies, in the distributional sense, the equation
$`{\displaystyle \frac{1}{2}}\mathrm{\Delta }\rho {\displaystyle \frac{Z}{r}}\rho +h=0\text{ in }^3,`$ (1.9)
where
$`h`$ $`C^\alpha (^3\{0\})L^{\mathrm{}}(^3)\text{ for all }\alpha (0,1)`$
and
$`\rho `$ $`C^{2,\alpha }(^3\{0\})C^{0,1}(^3)\text{ for all }\alpha (0,1).`$
2. The function $`\stackrel{~}{\rho }`$ defined in (1) satisfies
$`{\displaystyle \frac{1}{2}}\mathrm{\Delta }\stackrel{~}{\rho }{\displaystyle \frac{Z}{r}}\stackrel{~}{\rho }+\stackrel{~}{h}=0\text{ for }r>0`$ (1.10)
where $`\stackrel{~}{h}(r)=_{𝕊^2}h(r\omega )𝑑\omega `$. Thereby,
$`\stackrel{~}{h}C^\alpha ((0,\mathrm{}))C^0([0,\mathrm{}))\text{ for all }\alpha (0,1)`$
and
$`\stackrel{~}{\rho }C^{2,\alpha }((0,\mathrm{}))C^2([0,\mathrm{}))\text{ for all }\alpha (0,1).`$
3. $`h(x)`$ $`C(R)\left({\displaystyle _{B(x,R)}}\rho (y)𝑑y+\rho (x)\right)\text{ for all }x^3,`$ (1.11)
$`h(x)`$ $`\epsilon \rho (x)\text{ for all }x^3,\text{ if }\epsilon =E_0^{N1}E>0.`$ (1.12)
4. $`\left({\displaystyle \frac{d^2}{dr^2}}\stackrel{~}{\rho }\right)(0)={\displaystyle \frac{2}{3}}\left(\stackrel{~}{h}(0)+Z^2\stackrel{~}{\rho }(0)\right).`$ (1.13)
###### Remark 1.12.
The results in (i) generalize to the case of molecules, where the continuity results for $`\rho `$ and $`h`$ hold in the complement of the set $`\{R_1,\mathrm{},R_L\}^3`$ (see Remark 1.7).
###### Remark 1.13.
It is known that eigenfunctions obey (Kato’s) Cusp Condition (see Kato ), and similar properties hold for particle densities. For more recent results see Hoffmann-Ostenhof et al. , Hoffmann-Ostenhof and Seiler . In the proof of Theorem 1.11, (iv) we make use of the Cusp Condition for $`\stackrel{~}{\rho }`$, namely:
$`\stackrel{~}{\rho }^{}(0)=\underset{r0}{lim}{\displaystyle \frac{\stackrel{~}{\rho }(r)\stackrel{~}{\rho }(0)}{r}}=Z\stackrel{~}{\rho }(0)\text{ and }\underset{r0}{lim}\stackrel{~}{\rho }^{}(r)=\stackrel{~}{\rho }^{}(0)`$ (1.14)
and also present a proof for it.
###### Remark 1.14.
Of course our results are only first steps in a thorough investigation of qualitative properties of the one-electron density. Here are some obvious open questions:
1. Is $`\rho (x)>0`$ for all $`x^3`$? We remark that this cannot be true in general, since it is false for some exited states of Hydrogen.
2. Is $`\rho C^{\mathrm{}}(^3\{0\})`$ or even $`C^\omega (^3\{0\})`$ ?
3. Is $`\stackrel{~}{\rho }`$ smooth in $`[0,\mathrm{})`$, in the sense that $`\left(\frac{d^k}{dr^k}\stackrel{~}{\rho }\right)(r)`$ exists for $`r0`$ for all $`k`$?
4. Is $`\frac{d}{dr}\stackrel{~}{\rho }(r)0`$ for $`r0`$ ? This is expected to be true for groundstate densities, but not known even for the bosonic case like Helium. Our results imply that $`\frac{d}{dr}\stackrel{~}{\rho }(r)0`$ for $`rR_0`$ for the bosonic case, where $`R_0`$ depends on the constant $`C`$ in Theorem 1.2. Note that because of (1.9) and (1.12) we have $`\mathrm{\Delta }\rho 0`$ for $`|x|Z/\epsilon `$, and so the Maximum Principle gives that $`\frac{d}{dr}\stackrel{~}{\rho }(r)<0`$ for $`r>Z/\epsilon `$.
###### Remark 1.15.
In the proof of Theorem 1.11 we obtain (see Proposition 3.1): With $`_1=(\frac{}{x_{1,1}},\frac{}{x_{1,2}},\frac{}{x_{1,3}})`$, the function
$`t_1(r)={\displaystyle _{𝕊^2}}{\displaystyle _{^{3(N1)}}}|_1\psi (r\omega ,x_2,\mathrm{},x_N)|^2𝑑x_2\mathrm{}𝑑x_N𝑑\omega `$
is continuous on $`[0,\mathrm{})`$.
## 2. Proofs
Throughout the proofs, we will denote by $`C`$ generic constants.
Crucial for our investigations is Corollary 8.36 in Gilbarg and Trudinger . We shall make use of this result several times and for convenience we state it already here, adapted for our special case:
###### Proposition 2.1.
Let $`\mathrm{\Omega }`$ be a bounded domain in $`^n`$ and suppose $`uW^{1,2}(\mathrm{\Omega })`$ is a weak solution of $`\mathrm{\Delta }u+_{j=1}^nb_jD_ju+Wu=g`$ in $`\mathrm{\Omega }`$, where $`b_j,W,gL^{\mathrm{}}(\mathrm{\Omega })`$. Then $`uC^{1,\alpha }(\mathrm{\Omega })`$ for all $`\alpha (0,1)`$ and for any domain $`\mathrm{\Omega }^{}`$, $`\overline{\mathrm{\Omega }^{}}\mathrm{\Omega }`$ we have
$`|u|_{C^{1,\alpha }(\mathrm{\Omega }^{})}C\left(\underset{\mathrm{\Omega }}{sup}|u|+\underset{\mathrm{\Omega }}{sup}|g|\right)`$
for $`C=C(n,M,\mathrm{dist}(\mathrm{\Omega }^{},\mathrm{\Omega }))`$, with
$`\underset{j=1,\mathrm{},n}{\mathrm{max}}\{1,b_j_{L^{\mathrm{}}(\mathrm{\Omega })},W_{L^{\mathrm{}}(\mathrm{\Omega })},g_{L^{\mathrm{}}(\mathrm{\Omega })}\}M.`$
Thereby
$`|u|_{C^{1,\alpha }(\mathrm{\Omega }^{})}=u_{L^{\mathrm{}}(\mathrm{\Omega }^{})}+u_{L^{\mathrm{}}(\mathrm{\Omega }^{})}+\underset{x,y\mathrm{\Omega }^{},xy}{sup}{\displaystyle \frac{|u(x)u(y)|}{|xy|^\alpha }}.`$
Proof of Theorem 1.2 and Proposition 1.5.
Let the function $`F`$ be as in (1.6) and define the function $`F_1`$ by
$`F_1(x_1,\mathrm{},x_N)={\displaystyle \underset{j=1}{\overset{N}{}}}{\displaystyle \frac{Z}{2}}\sqrt{|x_j|^2+1}+{\displaystyle \underset{1j<kN}{}}{\displaystyle \frac{1}{4}}\sqrt{|x_jx_k|^2+1}.`$ (2.1)
A computation shows that
$`\mathrm{\Delta }F=V,`$ (2.2)
$`FF_1_{L^{\mathrm{}}(^{3N})},(FF_1)_{L^{\mathrm{}}(^{3N})},D^\beta F_1_{L^{\mathrm{}}(^{3N})}`$
$`C(\beta ,N,Z),|\beta |1.`$ (2.3)
Make the ‘Ansatz’ $`\psi =e^{FF_1}\psi _1`$. Using $`(HE)\psi =0`$ and (2.2) we get that $`\psi _1`$ satisfies the equation
$`\mathrm{\Delta }\psi _1+2(FF_1)\psi _1+(|(FF_1)|^2\mathrm{\Delta }F_1+E)\psi _1=0.`$ (2.4)
Due to (2) the coefficients in (2.4) are bounded in $`^{3N}`$. Then Proposition 2.1 implies that $`\psi _1`$ is $`C^{1,\alpha }`$ for all $`\alpha (0,1)`$, in any ball $`B(𝐱,R)^{3N}`$, and
$`|\psi _1|_{C^{1,\alpha }(B(𝐱,R))}C\underset{yB(𝐱,2R)}{sup}|\psi _1(y)|`$ (2.5)
with $`C`$ depending on $`R`$ but not on $`𝐱`$. Since
$`|\psi (𝐲)||(FF_1)||\psi (𝐲)|+|e^{FF_1}\psi _1(𝐲)|`$
we obtain, via (2) and (2.5),
$`\underset{𝐲B(𝐱,R)}{sup}|\psi (𝐲)|C(\underset{𝐲B(𝐱,R)}{sup}|\psi (𝐲)|+\underset{𝐲B(𝐱,R)}{sup}|\psi _1(𝐲)|)`$
$`C(\underset{𝐲B(𝐱,2R)}{sup}|\psi (𝐲)|+\underset{𝐲B(𝐱,2R)}{sup}|\psi _1(𝐲)|)C\underset{𝐲B(𝐱,2R)}{sup}|\psi (𝐲)|,`$
with $`C=C(R)`$. This proves Theorem 1.2.
Proposition 1.5 follows from $`\psi =e^{FF_1}\psi _1`$, $`\psi _1C_{\text{loc}}^{1,\alpha }(B(𝐱,R))`$, since $`e^{F_1}`$ is smooth. ∎
Proof of Theorem 1.11.
Multiplying the equation $`(HE)\psi =0`$ with $`\psi `$ and integrating over $`x_2,\mathrm{},x_N`$ gives the equation
$`{\displaystyle _{^{3(N1)}}}`$ $`\psi \mathrm{\Delta }_1\psi dx_2\mathrm{}dx_N+{\displaystyle \frac{Z}{|x_1|}}\rho (x_1)=`$
$`={\displaystyle \underset{j=2}{\overset{N}{}}}{\displaystyle _{^{3(N1)}}}\psi \left(\mathrm{\Delta }_j{\displaystyle \frac{Z}{|x_j|}}\right)\psi 𝑑x_2\mathrm{}𝑑x_N`$
$`+{\displaystyle \underset{1j<kN}{}}{\displaystyle _{^{3(N1)}}}{\displaystyle \frac{1}{|x_jx_k|}}\psi ^2𝑑x_2\mathrm{}𝑑x_N`$
$`={\displaystyle _{^{3(N1)}}}\psi \left(H^{N1}E\right)\psi 𝑑x_2\mathrm{}𝑑x_N`$
$`+{\displaystyle \underset{j=2}{\overset{N}{}}}{\displaystyle _{^{3(N1)}}}{\displaystyle \frac{1}{|x_1x_j|}}\psi ^2𝑑x_2\mathrm{}𝑑x_N`$ (2.6)
where $`H^{N1}`$ is the $`(N1)`$ \- electron operator defined in (1.8). Since $`\mathrm{\Delta }_1\left(\psi ^2\right)=2|_1\psi |^2+2\psi \mathrm{\Delta }_1\psi `$ and $`\mathrm{\Delta }_1(\psi ^2)(x_1,x^{})𝑑x^{}=\mathrm{\Delta }_1\rho `$ in the distributional sense, we get that
$`{\displaystyle \frac{1}{2}}\mathrm{\Delta }_1\rho (x_1)`$ $`={\displaystyle _{^{3(N1)}}}|_1\psi |^2𝑑x_2\mathrm{}𝑑x_N`$
$`+{\displaystyle _{^{3(N1)}}}\psi \mathrm{\Delta }_1\psi 𝑑x_2\mathrm{}𝑑x_N`$ (2.7)
which, together with (2), gives the equation ($`r=|x|`$)
$`{\displaystyle \frac{1}{2}}\mathrm{\Delta }\rho (x)+{\displaystyle \frac{Z}{r}}\rho (x)`$ $`={\displaystyle _{^{3(N1)}}}\psi \left(H^{N1}E\right)\psi 𝑑x_2\mathrm{}𝑑x_N`$
$`+{\displaystyle \underset{j=2}{\overset{N}{}}}{\displaystyle _{^{3(N1)}}}{\displaystyle \frac{1}{|x_1x_j|}}\psi ^2𝑑x_2\mathrm{}𝑑x_N`$
$`+{\displaystyle _{^{3(N1)}}}|_1\psi |^2𝑑x_2\mathrm{}𝑑x_Nh(x),`$ (2.8)
hence we obtain (1.9). Integration of (1.9) over $`𝕊^2`$ yields (1.10).
The proof of the regularity properties of the functions $`h`$ and $`\stackrel{~}{h}`$ are rather technical, and therefore postponed to the next section.
We now verify the regularity properties of the functions $`\rho `$ and $`\stackrel{~}{\rho }`$ under the assumption that the regularity properties of $`h`$ and $`\stackrel{~}{h}`$ stated in Theorem 1.11 have been shown. Define the function $`\mu `$ by the equation ($`r=|x|`$)
$`\rho (x)=e^{Zr}\left(\rho (0)+\mu (x)\right).`$ (2.9)
Then $`\mu =e^{Zr}\rho \rho (0)`$, $`\mu (0)=0`$, and (2) implies that
$`\mathrm{\Delta }\mu 2Z{\displaystyle \frac{x}{r}}\mu +Z^2\mu =2he^{Zr}Z^2\rho (0).`$ (2.10)
Since $`hL_{\text{loc}}^{\mathrm{}}`$, all coefficients of (2.10) are $`L_{\text{loc}}^{\mathrm{}}`$, and since $`\rho W_{\text{loc}}^{1,2}`$, also $`\mu W_{\text{loc}}^{1,2}`$. Therefore Proposition 2.1 leads to $`\mu C_{\text{loc}}^{1,\alpha }`$, for all $`\alpha (0,1)`$. Due to (2.9), $`\rho C^{0,1}(^3)`$ follows.
Now consider
$`\mathrm{\Delta }\rho ={\displaystyle \frac{2Z}{r}}\rho +2hg\text{ in }^3\{0\}.`$ (2.11)
Since $`hC^\alpha (^3\{0\})`$, for all $`\alpha (0,1)`$ and due to the above, $`\rho /rC^\alpha (^3\{0\})`$, for all $`\alpha (0,1)`$, we have
$`gC^\alpha (^3\{0\})\text{ for all }\alpha (0,1).`$ (2.12)
From (2.11) and (2.12) we obtain from regularity theory for the Poisson equation that $`\rho C^{2,\alpha }(^3\{0\})`$ (see e. g. Gilbarg and Trudinger \[3, Thm. 4.3 and 4.6\] or Lieb and Loss \[10, Thm. 10.3\]).
We proceed analogously for $`\stackrel{~}{\rho }`$: integrating (2.11) over $`𝕊^2`$, we get the equation
$`\mathrm{\Delta }\stackrel{~}{\rho }={\displaystyle \frac{2Z}{r}}\stackrel{~}{\rho }+2\stackrel{~}{h}\stackrel{~}{g}\text{ in }^3\{0\}`$ (2.13)
with
$`\stackrel{~}{h}(r)={\displaystyle _{𝕊^2}}h(r\omega )𝑑\omega .`$ (2.14)
Since the R.H.S. of (2.13) is in $`C^\alpha (^3\{0\})`$, we obtain that $`\stackrel{~}{\rho }`$ as a (radially symmetric) function in $`^3`$ is $`C^{2,\alpha }`$ away from the origin, and therefore $`\stackrel{~}{\rho }:_+`$ satisfies $`\stackrel{~}{\rho }C^{2,\alpha }((0,\mathrm{}))`$.
That $`\stackrel{~}{\rho }C^2([0,\mathrm{}))`$ is shown in the proof of (iv).
Next we prove (iii). To prove the bound (1.12), let $`E_0^{N1}`$ be the groundstate energy for the operator $`H^{N1}`$. From Remark 1.8 and the Variational Principle we get that for almost all $`x_1^3`$,
$`{\displaystyle _{^{3(N1)}}}\psi \left(H^{N1}E\right)\psi 𝑑x_2\mathrm{}𝑑x_N(E_0^{N1}E)\rho (x_1),`$
and so $`h(x)\epsilon \rho (x)`$ with $`\epsilon =E_0^{N1}E>0`$.
As for the bound (1.11), note that due to the operator inequality $`\mathrm{\Delta }\beta /r\beta ^2/4`$ (true in dimension $`3`$) and the translation invariance of $`\mathrm{\Delta }`$ we have, for almost all $`x_k^3`$ (fixed), $`k\{1,\mathrm{},N\}`$, $`kj`$,
$`{\displaystyle _^3}{\displaystyle \frac{1}{|x_jx_k|}}|\psi |^2𝑑x_j{\displaystyle _^3}|_j\psi |^2𝑑x_j+{\displaystyle \frac{1}{4}}{\displaystyle _^3}|\psi |^2𝑑x_j.`$
In this way, using (2.7),
$`h(x)C\left({\displaystyle _{^{3(N1)}}}|\psi |^2𝑑x_2\mathrm{}𝑑x_N+{\displaystyle _{^{3(N1)}}}|\psi |^2𝑑x_2\mathrm{}𝑑x_N\right).`$ (2.15)
Due to Theorem 1.2 and a subsolution estimate (see Simon \[14, Theorem C.1.2.\]) we get, with $`𝐱=(x_1,\mathrm{},x_N)=(x_1,x^{})`$, $`x^{}^{3(N1)}`$ and $`\chi _\mathrm{\Omega }`$ the characteristic function of the set $`\mathrm{\Omega }`$:
$`|\psi (𝐱)|^2C\underset{𝐲B(𝐱,R)}{sup}|\psi (𝐲)|^2`$ $`C{\displaystyle _{𝐲B(𝐱,2R)}}|\psi (𝐲)|^2𝑑𝐲`$
$`=C{\displaystyle _{^{3N}}}\chi _{B(𝐱,2R)}(𝐲)|\psi (𝐲)|^2𝑑𝐲.`$
Using this, and that for fixed $`𝐱,𝐲^{3N}`$:
$`\chi _{B(𝐱,2R)}(𝐲)=\chi _{B(𝐲,2R)}(𝐱)=\{\begin{array}{cc}& 1\text{ if }|𝐱𝐲|<2R,\hfill \\ & 0\text{ otherwise}\hfill \end{array}`$
we get, by Fubini,
$`{\displaystyle _{^{3(N1)}}}|\psi (x_1,x^{})|^2𝑑x^{}`$
$`C{\displaystyle _{^{3(N1)}}}\left({\displaystyle _{^{3N}}}\chi _{B(𝐱,2R)}(𝐲)|\psi (𝐲)|^2𝑑𝐲\right)𝑑x^{}`$
$`=C{\displaystyle _{^{3N}}}|\psi (𝐲)|^2\left({\displaystyle _{^{3(N1)}}}\chi _{B(𝐲,2R)}(𝐱)𝑑x^{}\right)𝑑𝐲.`$ (2.16)
Note that with $`𝐳=𝐱𝐲`$ we have
$`\chi _{B(𝐲,2R)}(𝐱)=\chi _{B(𝐲,2R)}(𝐳+𝐲)=\chi _{B(\mathrm{𝟎},2R)}(𝐳)`$
and so (with $`r^{}=|z^{}|`$ and $`\omega ^{}=z^{}/r^{}`$)
$`{\displaystyle _{^{3(N1)}}}\chi _{B(𝐲,2R)}((x_1,x^{}))𝑑x^{}={\displaystyle _{^{3(N1)}}}\chi _{B(\mathrm{𝟎},2R)}((z_1,z^{}))𝑑z^{}`$
$`={\displaystyle _{|(z_1,z^{})|2R}}dz^{}=\chi _{B(0,2R)}(z_1){\displaystyle _{𝕊^{3(N1)1}}}{\displaystyle _0^{\sqrt{4R^2|z_1|^2}}}r_{}^{}{}_{}{}^{3(N1)1}dr^{}d\omega ^{}`$
$`=C(N)(4R^2|z_1|^2)^{3(N1)/2}\chi _{B(0,2R)}(z_1)`$
$`\stackrel{~}{C}(N)R^{3(N1)}\chi _{B(x_1,2R)}(y_1).`$ (2.17)
From (2) and (2) we get
$`{\displaystyle _{^{3(N1)}}}|\psi (x_1,x^{})|^2𝑑x^{}`$
$`C(R){\displaystyle _{|x_1y_1|<2R}}{\displaystyle _{^{3(N1)}}}|\psi (y_1,y^{})|^2𝑑y^{}𝑑y_1=C(R){\displaystyle _{B(x_1,2R)}}\rho (y_1)𝑑y_1.`$ (2.18)
Combining (2.15) and (2) proves (1.11).
We now prove (iv). We first prove Kato’s Cusp Condition (1.14) for the function $`\stackrel{~}{\rho }`$:
$`\stackrel{~}{\rho }^{}(0)=\underset{r0}{lim}{\displaystyle \frac{\stackrel{~}{\rho }(r)\stackrel{~}{\rho }(0)}{r}}=Z\stackrel{~}{\rho }(0)\text{ and }\underset{r0}{lim}\stackrel{~}{\rho }^{}(r)=\stackrel{~}{\rho }^{}(0).`$
First, define the function $`\stackrel{~}{\mu }`$ by the equation (see also (2.9))
$`\stackrel{~}{\rho }(r)=e^{Zr}\left(\stackrel{~}{\rho }(0)+\stackrel{~}{\mu }(r)\right).`$ (2.19)
Note that $`\stackrel{~}{\mu }(0)=0`$. Then, using (1.10), $`\stackrel{~}{\mu }`$ satisfies the equation
$`\mathrm{\Delta }\stackrel{~}{\mu }2Z{\displaystyle \frac{x}{r}}\stackrel{~}{\mu }+Z^2\stackrel{~}{\mu }=2\stackrel{~}{h}e^{Zr}Z^2\stackrel{~}{\rho }(0),`$
and, since $`\stackrel{~}{h}`$ is continuous, Proposition 2.1 gives that $`\stackrel{~}{\mu }`$, as a (radially symmetric) function in $`^3`$, is $`C^{1,\alpha }`$ in a neighbourhood of the origin. In particular, $`lim_{r0}\stackrel{~}{\mu }^{}(r)=\stackrel{~}{\mu }^{}(0)`$. Since (see (2.19))
$`\stackrel{~}{\rho }^{}(r)=Z\stackrel{~}{\rho }(0)+e^{Zr}\stackrel{~}{\mu }^{}(r)`$ (2.20)
this means that
$`\underset{r0}{lim}\stackrel{~}{\rho }^{}(r)=Z\stackrel{~}{\rho }(0)+\underset{r0}{lim}\stackrel{~}{\mu }^{}(r)=\stackrel{~}{\mu }^{}(0)Z\stackrel{~}{\rho }(0).`$ (2.21)
From (2.19) we also get that
$`{\displaystyle \frac{\stackrel{~}{\mu }(r)}{r}}=e^{Zr}{\displaystyle \frac{\stackrel{~}{\rho }(r)\stackrel{~}{\rho }(0)}{r}}+{\displaystyle \frac{e^{Zr}1}{r}}\stackrel{~}{\rho }(0).`$
This, together with (2.21) and $`\stackrel{~}{\mu }(0)=0`$, implies that
$`\underset{r0}{lim}\stackrel{~}{\rho }^{}(r)`$ $`=\stackrel{~}{\mu }^{}(0)Z\stackrel{~}{\rho }(0)=\underset{r0}{lim}{\displaystyle \frac{\stackrel{~}{\mu }(r)}{r}}\stackrel{~}{\rho }(0)\underset{r0}{lim}{\displaystyle \frac{e^{Zr}1}{r}}`$
$`=\underset{r0}{lim}e^{Zr}{\displaystyle \frac{\stackrel{~}{\rho }(r)\stackrel{~}{\rho }(0)}{r}}=\stackrel{~}{\rho }^{}(0),`$ (2.22)
and by (2.20) and (2), $`\stackrel{~}{\rho }^{}(0)=Z\stackrel{~}{\rho }(0)`$. This proves (1.14).
Next, define the function $`\stackrel{~}{\eta }`$ by the equations
$`\stackrel{~}{\rho }(r)`$ $`=e^{Zr}\left(\stackrel{~}{\rho }(0)+\beta r^2+\stackrel{~}{\eta }(r)\right),`$ (2.23)
$`\beta `$ $`={\displaystyle \frac{1}{3}}\left(\stackrel{~}{h}(0){\displaystyle \frac{Z^2}{2}}\stackrel{~}{\rho }(0)\right).`$
Then $`\stackrel{~}{\eta }(0)=0`$, and due to (1.14), $`\stackrel{~}{\eta }^{}(0)=0`$. Together with (1.9) this gives
$`\mathrm{\Delta }`$ $`\stackrel{~}{\eta }(r)=\left({\displaystyle \frac{1}{r^2}}{\displaystyle \frac{}{r}}r^2{\displaystyle \frac{}{r}}\right)\stackrel{~}{\eta }`$
$`=2\left[e^{Zr}\stackrel{~}{h}(r){\displaystyle \frac{Z^2}{2}}\stackrel{~}{\rho }(r)e^{Zr}3\beta +2Z\beta r+Z\stackrel{~}{\eta }^{}(r)\right]`$
$`2G(r).`$ (2.24)
From the foregoing regularity properties of $`\stackrel{~}{\rho }`$, in particular (1.14) and (2.23), we obtain that $`\stackrel{~}{\eta }^{}C^0([0,\mathrm{}))`$. From this, together with the regularity properties of $`\stackrel{~}{h}`$ shown in Section 3, we conclude that $`GC^0([0,\mathrm{}))`$ and
$`G(r)\stackrel{~}{h}(0){\displaystyle \frac{Z^2}{2}}\stackrel{~}{\rho }(0)3\beta =0\text{as}r0.`$ (2.25)
From (2.25) and (2) we get that
$`\stackrel{~}{\eta }(r)=2{\displaystyle _0^r}{\displaystyle \frac{1}{t^2}}{\displaystyle _0^t}G(s)s^2𝑑s𝑑t,`$
and
$`{\displaystyle \frac{\stackrel{~}{\eta }^{}(r)\stackrel{~}{\eta }^{}(0)}{r}}={\displaystyle \frac{2}{r^3}}{\displaystyle _0^r}G(s)s^2𝑑s={\displaystyle \frac{2}{3}}{\displaystyle \frac{1}{\mathrm{Vol}_^3(B(0,r))}}{\displaystyle _{B(0,r)}}G(|x|)d^3x,`$
so that
$`\eta ^{\prime \prime }(0)=\underset{r0}{lim}{\displaystyle \frac{\eta ^{}(r)\eta ^{}(0)}{r}}={\displaystyle \frac{2}{3}}G(0)=0.`$ (2.26)
Then by (2.23) $`\stackrel{~}{\rho }^{\prime \prime }(0)`$ exists, and
$`\stackrel{~}{\rho }^{\prime \prime }(0)=Z^2\stackrel{~}{\rho }(0)+2\beta ={\displaystyle \frac{2}{3}}\left(\stackrel{~}{h}(0)+Z^2\stackrel{~}{\rho }(0)\right).`$
This verifies (1.13).
Furthermore, by (2),
$`\mathrm{\Delta }\stackrel{~}{\eta }(r)=\stackrel{~}{\eta }^{\prime \prime }(r)+{\displaystyle \frac{2}{r}}\stackrel{~}{\eta }^{}(r)=2G(r)`$
and so
$`\stackrel{~}{\eta }^{\prime \prime }(r)=2G(r){\displaystyle \frac{2}{r}}\stackrel{~}{\eta }^{}(r)=2G(r)2\left({\displaystyle \frac{\stackrel{~}{\eta }^{}(r)\stackrel{~}{\eta }^{}(0)}{r}}\right)`$
since $`\stackrel{~}{\eta }^{}(0)=0`$. This implies, by (2.26),
$`\underset{r0}{lim}\stackrel{~}{\eta }^{\prime \prime }(r)=2\left(\underset{r0}{lim}G(r)\underset{r0}{lim}\left({\displaystyle \frac{\stackrel{~}{\eta }^{}(r)\stackrel{~}{\eta }^{}(0)}{r}}\right)\right)=\stackrel{~}{\eta }^{\prime \prime }(0)=0,`$
so that due to (2.23) $`\stackrel{~}{\rho }^{\prime \prime }(r)`$ is continuous at $`r=0`$. Hence formula (1.13) follows from (2.23) and $`\stackrel{~}{\eta }^{\prime \prime }(0)=0`$. This finishes the proof of Theorem 1.11. ∎
## 3. Regularity of $`h`$ and $`\stackrel{~}{h}`$
In this section we prove the statements in Theorem 1.11 on the regularity of the functions $`h`$ and $`\stackrel{~}{h}`$. More precisely, we prove the following:
###### Proposition 3.1.
Let $`\psi `$ satisfy (1.3) and let $`h`$ be as in (2):
$`h(x)`$ $`={\displaystyle _{^{3(N1)}}}|\psi |^2𝑑x_2\mathrm{}𝑑x_N`$
$`{\displaystyle \underset{j=2}{\overset{N}{}}}{\displaystyle _{^{3(N1)}}}{\displaystyle \frac{Z}{|x_j|}}\psi ^2𝑑x_2\mathrm{}𝑑x_N`$
$`+{\displaystyle \underset{1j<kN}{}}{\displaystyle _{^{3(N1)}}}{\displaystyle \frac{1}{|x_jx_k|}}\psi ^2𝑑x_2\mathrm{}𝑑x_N,`$ (3.1)
and $`\stackrel{~}{h}`$ as in (2.14):
$`\stackrel{~}{h}(r)`$ $`={\displaystyle _{𝕊^2}}h(r\omega )𝑑\omega .`$
Then $`hC_{\text{loc}}^\alpha (^3\{0\})`$ and $`\stackrel{~}{h}C^0([0,\mathrm{}))C_{\text{loc}}^\alpha ((0,\mathrm{}))`$ for all $`\alpha (0,1)`$.
###### Remark 3.2.
From the proof of Proposition 3.1 follows the continuity of the function $`t_1`$ in Remark 1.15.
###### Proof.
For convenience, we shall often write $`_{^{3(N1)}}`$. Let
$`J_1(x)`$ $`={\displaystyle |\psi |^2𝑑x_2\mathrm{}𝑑x_N},`$
$`J_2(x)`$ $`={\displaystyle \underset{j=2}{\overset{N}{}}}{\displaystyle \frac{Z}{|x_j|}\psi ^2𝑑x_2\mathrm{}𝑑x_N},`$
$`J_3(x)`$ $`={\displaystyle \underset{1j<kN}{}}{\displaystyle \frac{1}{|x_jx_k|}\psi ^2𝑑x_2\mathrm{}𝑑x_N}.`$ (3.2)
For the proof of the regularity of the functions $`J_1,J_2,J_3`$ we shall make use of the following lemmas. The proof of the first lemma is trivial, using $`|(x_1,x_2,\mathrm{},x_N)||(x_2,\mathrm{},x_N)|`$.
###### Lemma 3.3.
Let $`\alpha (0,1)`$ and $`x_0^3`$. Assume that the real function $`G=G(x_1,\mathrm{},x_N)`$ satisfies: For all $`R>0`$ there exists constants $`C,\gamma `$ such that
$`\underset{x,yB(x_0,R)}{sup}{\displaystyle \frac{|G(x,x_2,\mathrm{},x_N)G(y,x_2,\mathrm{},x_N)|}{|xy|^\alpha }}`$
$`C\mathrm{exp}\left(\gamma |(x_0,x_2,\mathrm{},x_N)|\right)\text{for all}(x_0,x_2,\mathrm{},x_N)^{3N}.`$ (3.3)
Then the function
$`\eta (x){\displaystyle _{^{3(N1)}}}G(x,x_2,\mathrm{},x_N)𝑑x_2\mathrm{}𝑑x_N`$
is in $`C_{\text{loc}}^\alpha (^3)`$.
We next prove the following lemma.
###### Lemma 3.4.
Let $`\alpha (0,1)`$. Assume that the real valued function $`K=K(x_1,\mathrm{},x_N)`$ satisfies (3.3) and that there exists constants $`C,\gamma `$ such that
$`|K(x_1,\mathrm{},x_N)|`$
$`C\mathrm{exp}\left(\gamma |(x_1,\mathrm{},x_N)|\right)\text{ for all }(x_1,\mathrm{},x_N)^{3N}.`$ (3.4)
Then:
(a)
For all $`j,k\{1,\mathrm{},N\},jk`$, the function
$`\zeta (x_1){\displaystyle _{^{3(N1)}}}{\displaystyle \frac{1}{|x_jx_k|}}K(x_1,\mathrm{},x_N)𝑑x_2\mathrm{}𝑑x_N`$
is in $`C_{\text{loc}}^\alpha (^3)`$.
(b)
For $`j2`$ the function
$`\mu (x_1){\displaystyle _{^{3(N1)}}}{\displaystyle \frac{1}{|x_j|}}K(x_1,\mathrm{},x_N)𝑑x_2\mathrm{}𝑑x_N`$
is in $`C_{\text{loc}}^\alpha (^3)`$.
###### Proof.
Assume first that $`j1k`$. Let $`x,yB(x_0,R)`$, then by (3.3),
$`{\displaystyle \frac{|\zeta (x)\zeta (y)|}{|xy|^\alpha }}`$
$`{\displaystyle \frac{1}{|x_jx_k|}\frac{|K(x,x_2,\mathrm{},x_N)K(y,x_2,\mathrm{},x_N)|}{|xy|^\alpha }𝑑x_2\mathrm{}𝑑x_N}`$
$`C{\displaystyle \frac{1}{|x_jx_k|}\mathrm{exp}\left(\gamma |(x_0,x_2,\mathrm{},x_N)|\right)𝑑x_2\mathrm{}𝑑x_N}.`$
By equivalence of norms in $`^{3N}`$ there is a constant $`c_0`$ such that
$`{\displaystyle \frac{|\zeta (x)\zeta (y)|}{|xy|^\alpha }}C\left({\displaystyle \underset{l=2,lj,k}{\overset{N}{}}}{\displaystyle _^3}\mathrm{exp}(\gamma c_0|x_l|)dx_l\right)\times `$
$`\times {\displaystyle _^6}{\displaystyle \frac{1}{|x_jx_k|}}\mathrm{exp}(\gamma c_0(|x_j|+|x_k|))dx_jdx_kC,x,yB(x_0,R).`$
The last inequality is an application of the following inequality (with $`n=3,\lambda =1,p=r=6/5`$): (see Lieb and Loss \[10, Theorem 4.3\])
Hardy-Littlewood-Sobolev Inequality: *Let $`p,r>1`$ and $`0<\lambda <n`$ with $`1/p+\lambda /n+1/r=2`$. Let $`fL^p(^n)`$ and $`hL^r(^n)`$. Then there exists a sharp constant $`C(n,\lambda ,p)`$, independent of $`f`$ and $`h`$, such that*
$`\left|{\displaystyle _^n}{\displaystyle _^n}f(x)|xy|^\lambda h(y)𝑑x𝑑y\right|C(n,\lambda ,p)f_ph_r.`$
This proves Lemma 3.4 (a) when $`j1k`$.
Assume now that $`j=1`$. We assume without loss that $`k=2`$. Then, with $`x_1,\overline{x}_1B(x_0,R)`$,
$`{\displaystyle \frac{|\zeta (x_1)\zeta (\overline{x}_1)|}{|x_1\overline{x}_1|^\alpha }}`$
$`{\displaystyle \frac{1}{|x_1x_2|}\frac{|K(x_1,x_2,\mathrm{},x_N)K(\overline{x}_1,x_2,\mathrm{},x_N)|}{|x_1\overline{x}_1|^\alpha }𝑑x_2\mathrm{}𝑑x_N}`$
$`+{\displaystyle \left|\frac{1}{|x_1x_2|}\frac{1}{|\overline{x}_1x_2|}\right||x_1\overline{x}_1|^\alpha |K(\overline{x}_1,x_2,\mathrm{},x_N)|𝑑x_2\mathrm{}𝑑x_N}`$
$`(\text{A})+(\text{B}).`$
For $`(\text{A})`$, using (3.3) and equivalence of norms in $`^{3N}`$, we get
$`(\text{A})`$ $`C{\displaystyle \frac{1}{|x_1x_2|}\mathrm{exp}\left(\gamma c_0(|x_0|+|x_2|+\mathrm{}+|x_N|)\right)𝑑x_2\mathrm{}𝑑x_N}`$
$`C{\displaystyle _^3}{\displaystyle \frac{1}{|x_1x_2|}}\mathrm{exp}(\gamma c_0|x_2|)𝑑x_2C(x_0,R).`$
As for $`(\text{B})`$, we apply the following inequality; for the convenience of the reader, we give the proof (borrowed from Lieb and Loss \[10, (3) p. 225\]).
For $`\alpha (0,1)`$:
$`\left|{\displaystyle \frac{1}{|xz|}}{\displaystyle \frac{1}{|yz|}}\right||xy|^\alpha `$
$`|xz|^{1\alpha }+|yz|^{1\alpha }\text{for all}x,y,z^3.`$ (3.5)
*Proof* of (3). By Hölder’s inequality we have, for $`b>1`$, $`\alpha (0,1)`$,
$`1b^1={\displaystyle _1^b}t^2𝑑t\left({\displaystyle _1^b}𝑑t\right)^\alpha \left({\displaystyle _1^{\mathrm{}}}t^{2/(1\alpha )}𝑑t\right)^{1\alpha }(b1)^\alpha .`$
Substituting $`b/a`$ for $`b`$, with $`a>0`$, this gives
$`\left|b^1a^1\right||ba|^\alpha \mathrm{max}\{a^{1\alpha },b^{1\alpha }\}.`$
So, for $`x,y,z^3`$, using $`\mathrm{max}\{s,t\}s+t`$ and the triangle inequality in $`^3`$, we have
$`\left||xz|^1|yz|^1\right||xy|^\alpha \left\{|xz|^{1\alpha }+|yz|^{1\alpha }\right\}.`$
In this way, by (3.4) and equivalence of norms in $`^{3N}`$,
$`(\text{B})`$ $`{\displaystyle \left(\frac{|K(\overline{x}_1,x_2,\mathrm{},x_N)|}{|x_1x_2|^{1+\alpha }}+\frac{|K(\overline{x}_1,x_2,\mathrm{},x_N)|}{|\overline{x}_1x_2|^{1+\alpha }}\right)𝑑x_2\mathrm{}𝑑x_N}`$
$`C{\displaystyle \underset{j=3}{\overset{N}{}}}{\displaystyle _^3}\mathrm{exp}(\gamma c_0|x_j|)𝑑x_j`$
$`\times {\displaystyle _^3}({\displaystyle \frac{1}{|\overline{x}_1x_2|^{1+\alpha }}}+{\displaystyle \frac{1}{|x_1x_2|^{1+\alpha }}})\mathrm{exp}(\gamma c_0|x_2|)dx_2`$
$`C(x_0,R),`$
since $`x_1,\overline{x}_1B(x_0,R)`$. This finishes the proof of Lemma 3.4 (a).
The proof of $`(b)`$ is similar to that of $`(a)`$ so we omit the details. ∎
The proof of the following fact is straightforward:
There exist constants $`C=C(\gamma ,R)`$ and $`\stackrel{~}{\gamma }=\stackrel{~}{\gamma }(\gamma )`$ such that
$`\mathrm{exp}(\gamma |(x,\mathrm{},x_N)|)C\mathrm{exp}(\stackrel{~}{\gamma }|(x_0,\mathrm{},x_N)|)`$ (3.6)
for all $`xB(x_0,R)`$.
Using this and Lemma 3.3, 3.4, we shall prove the following lemma on the regularity of the functions $`J_1,J_2`$ and $`J_3`$ from (3).
###### Lemma 3.5.
Let $`J_1,J_2`$ and $`J_3`$ be as in (3). Then
1. $`J_2,J_3C_{\text{loc}}^\alpha (^3)`$ for all $`\alpha (0,1)`$.
2. $`J_1C_{\text{loc}}^\alpha (^3\{0\})`$ for all $`\alpha (0,1)`$.
Herefrom follow the regularity properties of the function $`h`$ stated in Proposition 3.1.
*Proof* of Lemma 3.5 (i).
Firstly, by Theorem 1.2 and Remark 1.8,
$`|\psi (𝐱)|,|\psi (𝐱)|C\mathrm{exp}(\gamma |𝐱|),𝐱^{3N},`$ (3.7)
which gives (3.4) for $`K=\psi ^2`$.
Next we verify that for $`G=\psi ^2`$, (3.3) is fulfilled. Then Lemma 3.4 can be applied with $`K=\psi ^2`$, and the Hölder-continuity of $`J_2`$ and $`J_3`$ follows.
Given $`x_0^3,R>0,\alpha (0,1)`$, and $`x,yB(x_0,R)`$. Using that (see e. g. Malý and Ziemer \[11, Theorem 1.41\]) (here, $`(x,x_2,\mathrm{},x_N)=(x,x^{})`$, $`x^{}^{3(N1)}`$)
$`\psi ^2(x,x^{})`$ $`\psi ^2(y,x^{})={\displaystyle _0^1}{\displaystyle \frac{}{s}}\left[\psi ^2(sx+(1s)y,x^{})\right]𝑑s`$
$`={\displaystyle _0^1}\left[_1(\psi ^2)(sx+(1s)y,x^{})\right](xy)𝑑s`$ (3.8)
and that $`sx+(1s)yB(x_0,R)`$ for all $`s[0,1]`$ we get, with (3.7) and (3.6),
$`{\displaystyle \frac{|\psi ^2(x,x^{})\psi ^2(y,x^{})|}{|xy|^\alpha }}`$
$`2|xy|^{1\alpha }{\displaystyle _0^1}|_1\psi (sx+(1s)y,x^{})||\psi (sx+(1s)y,x^{})|𝑑s`$
$`2(2R)^{1\alpha }{\displaystyle _0^1}C\mathrm{exp}(2\gamma |(sx+(1s)y,x^{})|)𝑑s`$
$`C\mathrm{exp}(\stackrel{~}{\gamma }|(x_0,\mathrm{},x_N)|),`$ (3.9)
so (3.3) follows for $`\alpha (0,1)`$.
This proves (i) of Lemma 3.5.
To prove (ii), we write $`\psi `$ as in the proof of Theorem 1.2: $`\psi =e^{FF_1}\psi _1`$, with $`F`$ and $`F_1`$ as in (1.6) and (2.1). Then
$`J_1`$ $`(x)={\displaystyle |\psi |^2𝑑x^{}}={\displaystyle |F|^2\psi ^2𝑑x^{}}+{\displaystyle |F_1|^2\psi ^2𝑑x^{}}`$
$`2{\displaystyle \left(FF_1\right)\psi ^2𝑑x^{}}+{\displaystyle e^{2(FF_1)}|\psi _1|^2𝑑x^{}}`$
$`+2{\displaystyle \left(F\psi _1\right)e^{2(FF_1)}\psi _1𝑑x^{}}2{\displaystyle \left(F_1\psi _1\right)e^{2(FF_1)}\psi _1𝑑x^{}}`$
$`I_1(x)+I_2(x)+I_3(x)+I_4(x)+I_5(x)+I_6(x).`$ (3.10)
Using the idea from (3) twice (on $`|F_1|^2`$ and $`\psi ^2`$, respectively), the estimates (2), (3.7), (3.9), and (3.6), we have, with $`x_0^3,R>0,\alpha (0,1)`$, and $`x,yB(x_0,R)`$:
$`{\displaystyle \frac{\left||F_1|^2\psi ^2(x,x^{})|F_1|^2\psi ^2(y,x^{})\right|}{|xy|^\alpha }}`$
$`{\displaystyle \frac{\left||F_1(x,x^{})|^2|F_1(y,x^{})|^2\right|}{|xy|^\alpha }}|\psi (x,x^{})|^2`$
$`+|F_1(y,x^{})|^2{\displaystyle \frac{\left|\psi ^2(x,x^{})\psi ^2(y,x^{}x)\right|}{|xy|^\alpha }}`$
$`C|xy|^{1\alpha }\left(|F_1|^2\right)_{\mathrm{}}\mathrm{exp}(2\gamma |(x,x_2,\mathrm{},x_N)|)`$
$`+2\stackrel{~}{C}|xy|^{1\alpha }|F_1|^2_{\mathrm{}}\mathrm{exp}(\stackrel{~}{\gamma }|(x_0,\mathrm{},x_N)|)`$
$`\overline{C}\mathrm{exp}(\overline{\gamma }(x_0,\mathrm{},x_N)|).`$
By Lemma 3.3, with $`G=|F_1|^2\psi ^2(x_1,\mathrm{},x_N)`$, this implies that $`I_2C_{\text{loc}}^\alpha (^3)`$.
Using the same ingredients, writing $`\psi _1=(e^{F_1F}\psi )`$, gives (3.3) and (3.4) with
$`G`$ $`=K=e^{2(FF_1)}|\psi _1|^2`$
and
$`G`$ $`=K=\left(F_1\psi _1\right)e^{2(FF_1)}\psi _1,`$
and so by Lemma 3.3, $`I_4,I_6C_{\text{loc}}^\alpha (^3)`$.
The remaining terms are those involving the function $`F`$, namely $`I_1`$, $`I_3`$ and $`I_5`$.
Note that
$`F(x_1,\mathrm{},x_N)={\displaystyle \frac{Z}{2}}({\displaystyle \frac{x_1}{|x_1|}},\mathrm{},{\displaystyle \frac{x_N}{|x_N|}})`$
$`+{\displaystyle \frac{1}{4}}({\displaystyle \underset{k=2}{\overset{N}{}}}{\displaystyle \frac{x_1x_k}{|x_1x_k|}},\mathrm{},{\displaystyle \underset{k=1,kj}{\overset{N}{}}}{\displaystyle \frac{x_jx_k}{|x_jx_k|}},\mathrm{},{\displaystyle \underset{k=1}{\overset{N1}{}}}{\displaystyle \frac{x_Nx_k}{|x_Nx_k|}})`$ (3.11)
and so
$`|F(x_1,\mathrm{},x_N)|^2`$ $`={\displaystyle \frac{NZ^2}{4}}{\displaystyle \frac{Z}{8}}{\displaystyle \underset{j,k=1,kj}{\overset{N}{}}}{\displaystyle \frac{x_j}{|x_j|}}{\displaystyle \frac{x_jx_k}{|x_jx_k|}}`$
$`+{\displaystyle \frac{1}{16}}{\displaystyle \underset{j,k,l=1,kj,lj}{\overset{N}{}}}{\displaystyle \frac{x_jx_k}{|x_jx_k|}}{\displaystyle \frac{x_jx_l}{|x_jx_l|}}.`$
In this way,
$`I_1(x_1)={\displaystyle \frac{NZ^2}{4}}{\displaystyle \psi ^2𝑑x_2\mathrm{}𝑑x_N}`$
$`{\displaystyle \frac{Z}{8}}{\displaystyle \underset{j,k=1,kj}{\overset{N}{}}}{\displaystyle \frac{x_j}{|x_j|}\frac{x_jx_k}{|x_jx_k|}\psi ^2𝑑x_2\mathrm{}𝑑x_N}`$
$`+{\displaystyle \frac{1}{16}}{\displaystyle \underset{j,k,l=1,kj,lj}{\overset{N}{}}}{\displaystyle \frac{x_jx_k}{|x_jx_k|}}{\displaystyle \frac{x_jx_l}{|x_jx_l|}}\psi ^2dx_2\mathrm{}dx_N`$
$`={\displaystyle \frac{NZ^2}{4}}\rho (x_1){\displaystyle \frac{Z}{8}}{\displaystyle \underset{j,k=1,kj}{\overset{N}{}}}\kappa _{j,k}(x_1)+{\displaystyle \frac{1}{16}}{\displaystyle \underset{j,k,l=1,kj,lj}{\overset{N}{}}}\nu _{j,k,l}(x_1).`$ (3.12)
Note that $`\nu _{j,k,l}=\nu _{j,l,k}`$.
Using the ideas above, Lemma 3.3 implies that the following functions from (3.12) (with the mentioned choices of $`G`$ satisfying (3.3)) are all in $`C_{\text{loc}}^\alpha (^3)`$:
$`\rho `$ $`:G=\psi ^2,`$
$`\kappa _{j,k},j,k1,jk`$ $`:G={\displaystyle \frac{x_j}{|x_j|}}{\displaystyle \frac{x_jx_k}{|x_jx_k|}}\psi ^2,`$
$`\nu _{j,k,k},jk`$ $`:G={\displaystyle \frac{x_jx_k}{|x_jx_k|}}{\displaystyle \frac{x_jx_k}{|x_jx_k|}}\psi ^2=\psi ^2,`$
$`\nu _{j,k,l},j,k,l1,ljk`$ $`:G={\displaystyle \frac{x_jx_k}{|x_jx_k|}}{\displaystyle \frac{x_jx_l}{|x_jx_l|}}\psi ^2.`$
Likewise, Lemma 3.4 implies (with the mentioned choices of $`G=K`$ satisfying (3.3) and (3.4)) that the following functions from (3.12) are all in $`C_{\text{loc}}^\alpha (^3)`$:
$`\kappa _{j,1},j1`$ $`:G=K={\displaystyle \frac{x_j(x_jx_1)}{|x_j|}}\psi ^2,`$
$`\nu _{j,1,l},j,l1,jl`$ $`:G=K={\displaystyle \frac{(x_jx_1)(x_jx_l)}{|x_jx_l|}}\psi ^2,`$
From the decomposition of $`I_1`$ in (3.12) we are left with
$`\kappa _{1,k}(x_1)=`$ $`{\displaystyle }{\displaystyle \frac{x_1}{|x_1|}}{\displaystyle \frac{x_1x_k}{|x_1x_k|}}\psi ^2dx^{},k=2,\mathrm{},N,`$ (3.13)
and
$`\nu _{1,k,l}(x_1)=`$ $`{\displaystyle }{\displaystyle \frac{x_1x_k}{|x_1x_k|}}{\displaystyle \frac{x_1x_l}{|x_1x_l|}}\psi ^2dx^{},k,l\{2,\mathrm{},N\},kl.`$ (3.14)
Note that
$`{\displaystyle \frac{x_1}{|x_1|}\frac{x_1x_k}{|x_1x_k|}\psi ^2𝑑x_2\mathrm{}𝑑x_N}`$
$`={\displaystyle \frac{1}{|x_1|}}{\displaystyle \frac{1}{|x_1x_k|}\left(x_1(x_1x_k)\psi ^2\right)𝑑x_2\mathrm{}𝑑x_N}.`$
The function $`1/|x_1|`$ is smooth for $`x_10`$ and therefore in $`C_{\text{loc}}^\alpha (^3\{0\})`$. The function $`x_1(x_1x_k)\psi ^2`$ satisfies (3.3) and (3.4) (by the same ideas as above), so Lemma 3.4 (a) implies that the function
$`{\displaystyle \frac{1}{|x_1x_k|}\left(x_1(x_1x_k)\psi ^2\right)𝑑x_2\mathrm{}𝑑x_N}`$
is in $`C_{\text{loc}}^\alpha (^3)`$. The functions in (3.13) are therefore in $`C_{\text{loc}}^\alpha (^3\{0\})`$.
As for the functions in (3.14), these are all in $`C_{\text{loc}}^\alpha (^3)`$, which can be seen by applying the previous ideas, in particular Lemma 3, (3.6), (3.7) and (3.9).
This proves that $`I_1C_{\text{loc}}^\alpha (^3\{0\})`$.
As for $`I_3`$ (see (3) and (3)), with $`=(_1,\mathrm{},_N)`$,
$`I_3(x)`$ $`=Z{\displaystyle \underset{j=1}{\overset{N}{}}}{\displaystyle \left(\frac{x_j}{|x_j|}_jF_1\right)\psi ^2𝑑x_2\mathrm{}𝑑x_N}`$
$`{\displaystyle \frac{1}{2}}{\displaystyle \underset{j,k=1,jk}{\overset{N}{}}}{\displaystyle \left(\frac{x_jx_k}{|x_jx_k|}_jF_1\right)\psi ^2𝑑x_2\mathrm{}𝑑x_N}.`$ (3.15)
The terms in the first sum with $`j1`$ are in $`C_{\text{loc}}^\alpha (^3)`$, due to Lemma 3.4 (b), with $`G=K=(x_j_jF_1)\psi ^2`$ satisfying (3.3) and (3.4). (To see this, use the previous ideas; to apply the idea from (3) to $`_jF_1`$ we use that $`F_1`$ is smooth). The terms in the second sum in (3) are all in $`C_{\text{loc}}^\alpha (^3)`$, due to Lemma 3.4 (a), applied with $`G=K=\left((x_jx_k)_jF_1\right)\psi ^2`$. The term with $`j=1`$ is in $`C_{\text{loc}}^\alpha (^3\{0\})`$. This can be seen by following the ideas in the proof of the regularity properties of the function in (3.13), now using Lemma 3.3 with $`G=(x_1_1F_1)\psi ^2`$.
The statements and proofs are similar for
$`I_5(x)`$ $`=Z{\displaystyle \underset{j=1}{\overset{N}{}}}{\displaystyle \left(\frac{x_j}{|x_j|}_j\psi _1\right)e^{2(FF_1)}\psi _1𝑑x_2\mathrm{}𝑑x_N}`$
$`+{\displaystyle \frac{1}{2}}{\displaystyle \underset{j,k=1,jk}{\overset{N}{}}}{\displaystyle \left(\frac{x_jx_k}{|x_jx_k|}_j\psi _1\right)e^{2(FF_1)}\psi _1𝑑x_2\mathrm{}𝑑x_N}.`$ (3.16)
That is, the functions in the first sum in (3) with $`j2`$ and those in the second sum are all in $`C_{\text{loc}}^\alpha (^3)`$, whereas the function in the first sum with $`j=1`$ is only in $`C_{\text{loc}}^\alpha (^3\{0\})`$. To prove this we use the inequality (with $`𝐱=(x_1,\mathrm{},x_N)`$):
$`|\psi _1|_{C^{1,\alpha }(B(𝐱,R/2))}C\underset{y(B(𝐱,R))}{sup}|\psi _1(y)|C\mathrm{exp}(\gamma |(x_1,\mathrm{},x_N)|).`$
This inequality follows from (2), (2.5) and (3.7) (remember that $`\psi _1=e^{F_1F}\psi `$).
This proves that $`I_5C_{\text{loc}}^\alpha (^3\{0\})`$, and so finishes the proof that $`J_1C_{\text{loc}}^\alpha (^3\{0\})`$. (See (3)). This proves (ii) and therefore Lemma 3.5. ∎
That $`\stackrel{~}{h}C_{\text{loc}}^\alpha ((0,\mathrm{}))`$ is a consequence of the foregoing and of the following proposition:
###### Proposition 3.6.
Assume $`fC_{\text{loc}}^\alpha (^3\{0\})`$, $`\alpha (0,1)`$. Then $`\stackrel{~}{f}C_{\text{loc}}^\alpha ((0,\mathrm{}))`$, where
$`\stackrel{~}{f}(r)={\displaystyle _{𝕊^2}}f(r\omega )𝑑\omega .`$
###### Proof.
Let $`r(0,\mathrm{})`$. For all $`x_0A=\{x^3||x|=r\}`$, choose $`R=R(x_0)`$ and $`C=C(x_0)`$ such that
$`\underset{x,yB(x_0,R(x_0))}{sup}{\displaystyle \frac{|f(x)f(y)|}{|xy|^\alpha }}C(x_0).`$ (3.17)
This is possible, since $`fC_{\text{loc}}^\alpha (^3\{0\})`$. Then
$`A{\displaystyle \underset{x_0A}{}}B(x_0,R(x_0)).`$
Using compactness of $`A`$, choose $`x_1,\mathrm{},x_mA`$ such that
$`A{\displaystyle \underset{j=1}{\overset{m}{}}}B(x_j,R(x_j)).`$
Choose $`ϵ(0,r)`$ such that
$`\{y^3|rϵ<|y|<r+ϵ\}{\displaystyle \underset{j=1}{\overset{m}{}}}B(x_j,R(x_j)).`$
Then, for all $`s,t(rϵ,r+ϵ)`$ and all $`\omega 𝕊^2`$ there exists $`j\{1,\mathrm{},m\}`$ such that $`s\omega ,t\omega B(x_j,R(x_j))`$ and therefore by (3.17),
$`{\displaystyle \frac{|f(s\omega )f(t\omega )|}{|st|^\alpha }}={\displaystyle \frac{|f(s\omega )f(t\omega )|}{|s\omega t\omega |^\alpha }}C(x_j).`$
So with $`C=\mathrm{max}\{C(x_1),\mathrm{},C(x_m)\}`$,
$`{\displaystyle \frac{|f(s\omega )f(t\omega )|}{|st|^\alpha }}C,\text{ for all }s,t(rϵ,r+ϵ)\text{and all }\omega 𝕊^2.`$
This implies that
$`{\displaystyle \frac{|\stackrel{~}{f}(s)\stackrel{~}{f}(t)|}{|st|^\alpha }}={\displaystyle \frac{|_{𝕊^2}(f(s\omega )f(t\omega ))𝑑\omega |}{|st|^\alpha }}`$
$`{\displaystyle _{𝕊^2}}{\displaystyle \frac{|f(s\omega )f(t\omega )|}{|s\omega t\omega |^\alpha }}𝑑\omega C,\text{ for all }s,t(rϵ,r+ϵ).`$
This proves that $`\stackrel{~}{f}C_{\text{loc}}^\alpha ((0,\mathrm{}))`$. ∎
To prove that $`\stackrel{~}{h}C^0([0,\mathrm{}))`$, we apply the following:
###### Proposition 3.7.
Assume $`fC_{\text{loc}}^\alpha (^3)`$. Then $`\stackrel{~}{f}C^0([0,\mathrm{}))`$, where
$`\stackrel{~}{f}(r)={\displaystyle _{𝕊^2}}f(r\omega )𝑑\omega .`$
###### Proof.
The function $`f`$ is continuous in $`^3`$, since it is in $`C_{\text{loc}}^\alpha (^3)`$. Let $`r[0,\mathrm{})`$. Then
$`\underset{sr}{lim}f(s\omega )=f(r\omega )\text{ for all }\omega 𝕊^2.`$
Using the supremum of $`f`$ on a sufficiently large compact set in $`^3`$ as a dominant, Lebesque’s Dominated Convergence Theorem gives us that
$`\underset{sr}{lim}{\displaystyle _{𝕊^2}}f(s\omega )𝑑\omega ={\displaystyle _{𝕊^2}}f(r\omega )𝑑\omega .`$
Therefore $`fC^0([0,\mathrm{}))`$. ∎
Recall the proof of the fact that $`hC_{\text{loc}}^\alpha (^3\{0\})`$. In fact, the only terms in the decomposition of $`h`$ (see (3.1), (3), (3), and (3)) that are only in $`C_{\text{loc}}^\alpha (^3\{0\})`$ and not in $`C_{\text{loc}}^\alpha (^3)`$ are the functions
$`{\displaystyle }{\displaystyle \frac{x_1}{|x_1|}}{\displaystyle \frac{x_1x_k}{|x_1x_k|}}\psi ^2dx_2\mathrm{}dx_N,k=2,\mathrm{},N,`$
$`{\displaystyle \left(\frac{x_1}{|x_1|}_1F_1\right)\psi ^2𝑑x_2\mathrm{}𝑑x_N},`$
$`{\displaystyle \left(\frac{x_1}{|x_1|}_1\psi _1\right)e^{2(FF_1)}\psi _1𝑑x_2\mathrm{}𝑑x_N}.`$ (3.18)
Comparing (3), (3.13), (3) and (3), it can be seen that all the terms in (3) stem from the function $`J_1`$, namely from $`I_1`$, $`I_3`$, and $`I_5`$.
All other terms in the decomposition of $`h`$ are in $`C_{\mathrm{loc}}^\alpha (^3)`$. When integrating them over $`𝕊^2`$, we get something continuous in $`[0,\mathrm{})`$, according to Proposition 3.7 above.
For the terms in (3) we note that they are all of the form
$`{\displaystyle \frac{x_1}{|x_1|}𝐊(x_1,x^{})𝑑x^{}}={\displaystyle \frac{x_1}{|x_1|}}{\displaystyle 𝐊(x_1,x^{})𝑑x^{}}.`$ (3.19)
In each case, we have
$`𝐋(x_1)=(L_1(x_1),L_2(x_1),L_3(x_1))={\displaystyle 𝐊(x_1,x^{})𝑑x^{}},L_jC_{\text{loc}}^\alpha (^3).`$ (3.20)
To see this, apply Lemma 3.3 to each of the coordinate functions $`L_j`$, $`j=1,2,3`$. The integrands are easily seen to satisfy (3.3) in each case, by the previous ideas. To get continuity in $`[0,\mathrm{})`$ of the functions in (3) we use (3.19) and (3.20), and the following lemma:
###### Proposition 3.8.
Assume $`𝐟=(f_1,f_2,f_3),f_jC_{\text{loc}}^\alpha (^3)`$. Then $`\overline{f}C^0([0,\mathrm{}))`$, where
$`\overline{f}(r)={\displaystyle _{𝕊^2}}\left(\omega 𝐟(r\omega )\right)𝑑\omega .`$
###### Proof.
The same as for Proposition 3.7, noting that for all $`r[0,\mathrm{})`$ and fixed $`\omega 𝕊^2`$:
$`\underset{sr}{lim}\omega 𝐟(s\omega )=\omega 𝐟(r\omega ).`$
This holds even in the case $`r=0`$, for which
$`\underset{s0}{lim}{\displaystyle _{𝕊^2}}\omega 𝐟(s\omega )𝑑\omega ={\displaystyle _{𝕊^2}}\omega 𝐟(0)𝑑\omega =0.`$
This proves that the functions in (3) are in $`C^0([0,\mathrm{}))`$. Therefore $`\stackrel{~}{h}C^0([0,\mathrm{}))`$, which finishes the proof of Proposition 3.1. ∎
Acknowledgement: The authors wish to thank G. Friesecke and S. Fournais for stimulating discussions.
|
warning/0005/hep-ph0005140.html
|
ar5iv
|
text
|
# 1 The spin structure in inclusive DIS
## 1 The spin structure in inclusive DIS
In deep-inelastic scattering (DIS) the transition from hadrons to quarks and gluons is described in terms of distribution and fragmentation functions. In general, the distribution functions for a quark can be obtained from the lightcone correlation functions .
$$\mathrm{\Phi }_{ij}(x)=\frac{d\xi ^{}}{2\pi }e^{ip\xi }P,S|\overline{\psi }_j(0)\psi _i(\xi )|P,S|_{\xi ^+=\xi _T=0},$$
(1)
depending on the lightcone fraction $`x=p^+/P^+`$. The hadron momentum $`P`$ is chosen so that it has no transverse component, $`P_T=0`$. At leading order, the relevant part of the correlator is $`\mathrm{\Phi }\gamma ^+`$
$`(\mathrm{\Phi }\gamma ^+)_{ij}`$ $`=`$ $`{\displaystyle \frac{d\xi ^{}}{2\pi \sqrt{2}}e^{ip\xi }P,s^{}|\psi _{+j}^{}(0)\psi _{+i}(\xi )|P,s}|_{\xi ^+=\xi _T=0}`$ (2)
where $`\psi _+P_+\psi =\frac{1}{2}\gamma ^{}\gamma ^+\psi `$ is the good component of the quark field .
The correlator contains all the soft parts appearing in the scattering processes and, as shown in Fig. 1, is related to the forward amplitude for antiquark-hadron scattering. By considering the quantity $`M=(\mathrm{\Phi }\gamma ^+)^T`$, one finds that for any antiquark-hadron state $`|a`$ the expectation value $`a|M|a`$ must be larger than or equal to zero.
Thus our strategy is the following: express the forward scattering matrix M as a matrix in the \[parton chirality space $``$ hadron spin space\] and obtain our bounds by requiring it to be positive semi-definite.
At leading twist and when no partonic intrinsic transverse momentum is taken into account, $`\mathrm{\Phi }(x)\gamma ^+`$ is simply given by the contribution of three distribution functions
$$\mathrm{\Phi }(x)\gamma ^+=\left\{f_1(x)+\lambda g_1(x)\gamma _5+h_1(x)\gamma _5\text{/}S_T\right\}P_+.$$
(3)
The first step consists in writing this quantity as a matrix in the parton chirality space; this is easily done by using the explicit expression of $`P_+=\frac{1}{2}\gamma ^{}\gamma ^+`$ and $`\gamma _5`$ as $`4\times 4`$ Dirac matrices. In chiral representation we find
$$M_{ij}=\left(\begin{array}{cccc}f_1(x)+\lambda g_1(x)& 0& 0& (S_T^1+iS_T^2)h_1(x)\\ 0& 0& 0& 0\\ 0& 0& 0& 0\\ (S_T^1iS_T^2)h_1(x)& 0& 0& f_1(x)\lambda g_1(x)\end{array}\right)$$
(4)
As it is clear from the above expression, at leading twist there are only two relevant basis states, corresponding to (good components of) the right and the left handed partons. Thus, instead of using the full four dimensional Dirac space, we can effectively use only a two-dimensional chirality space.
Step two will be to write the above matrix explicitly in the hadron spin space. In order to study the correlation function in a spin $`1/2`$ target we introduce a spin vector $`S`$ that parameterizes the spin density matrix
$$\rho (P,S)=\frac{1}{2}(1+𝑺\mathbf{}𝝈).$$
(5)
The spin vector satisfies $`PS`$ = 0 and $`S^2=1`$ (spacelike) for a pure state, $`1<S^20`$ for a mixed state. Using $`\lambda MS^+/P^+`$ and the transverse spin vector $`S_T`$, the condition becomes $`\lambda ^2+𝑺_T^21`$, as can be seen from the rest-frame expression $`S`$ = $`(0,𝑺_T,\lambda )`$. The precise equivalence of a $`2\times 2`$ matrix $`\stackrel{~}{M}_{ss^{}}`$ in the target spin space and the $`S`$-dependent function $`M(S)`$ is $`M(S)=\text{Tr}\left[\rho (S)\stackrel{~}{M}\right]`$. Explicitly, the $`S`$-dependent function
$$M(S)=M_O+\lambda M_L+S_T^1M_T^1+S_T^2M_T^2,$$
(6)
corresponds to a matrix, which in the target rest-frame with as basis the spin 1/2 states with $`\lambda =+1`$ and $`\lambda =1`$ becomes
$$\stackrel{~}{M}_{ss^{}}=\left(\begin{array}{cc}M_O+M_L& M_T^1iM_T^2\\ & \\ M_T^1+iM_T^2& M_OM_L\end{array}\right)$$
(7)
Thus, each element of matrix (4) will transform in a $`2\times 2`$ matrix, according to Eqs.(6) and (7).
At leading twist and in absence of intrinsic transverse momentum, in the combined \[parton chirality $``$ hadron spin space\], the final result is
$$\stackrel{~}{M}_{is,js}=\left(\begin{array}{cccc}f_1+g_1& 0& 0& 2h_1\\ & & & \\ 0& f_1g_1& 0& 0\\ & & & \\ 0& 0& f_1g_1& 0\\ & & & \\ 2h_1& 0& 0& f_1+g_1\end{array}\right).$$
(8)
From the positivity of the diagonal elements one recovers the trivial bounds $`f_1(x)0`$ and $`|g_1(x)|f_1(x)`$, but requiring the eigenvalues of the matrix to be positive gives the stricter Soffer bound ,
$$|h_1(x)|\frac{1}{2}\left(f_1(x)+g_1(x)\right).$$
(9)
## 2 The full spin structure in SIDIS
We now turn to the more general case in which non-collinear configurations are taken into account. Transverse momenta of the partons inside the proton play an important role in hard processes with more than one hadron , like semi-inclusive deep inelastic scattering (SIDIS), $`e^{}He^{}hX`$ , or Drell-Yan scattering, $`H_1H_2\mu ^+\mu ^{}X`$ .
Analogous bounds can be obtained for transverse momentum dependent distribution and fragmentation functions. The soft parts involving the distribution functions are contained in the lightfront correlation function
$$\mathrm{\Phi }_{ij}(x,𝒑_T)=\frac{d\xi ^{}d^2𝝃_T}{(2\pi )^3}e^{ip\xi }P,S|\overline{\psi }_j(0)\psi _i(\xi )|P,S|_{\xi ^+=0},$$
(10)
depending on $`x=p^+/P^+`$ and the quark transverse momentum $`𝒑_T`$ in a target with $`P_T=0`$.
Separating the terms corresponding to unpolarized ($`O`$), longitudinally polarized ($`L`$) and transversely polarized targets ($`T`$), the most general parameterizations with $`p_T`$-dependence, relevant at leading order, are
$`\mathrm{\Phi }_O(x,𝒑_T)\gamma ^+`$ $`=`$ $`\left\{f_1(x,𝒑_T^2)+ih_1^{}(x,𝒑_T^2){\displaystyle \frac{\text{/}p_T}{M}}\right\}P_+,`$
$`\mathrm{\Phi }_L(x,𝒑_T)\gamma ^+`$ $`=`$ $`\left\{\lambda g_{1L}(x,𝒑_T^2)\gamma _5+\lambda h_{1L}^{}(x,𝒑_T^2)\gamma _5{\displaystyle \frac{\text{/}p_T}{M}}\right\}P_+,`$
$`\mathrm{\Phi }_T(x,𝒑_T)\gamma ^+`$ $`=`$ $`\{f_{1T}^{}(x,𝒑_T^2){\displaystyle \frac{ϵ_{T\rho \sigma }p_T^\rho S_T^\sigma }{M}}+g_{1T}(x,𝒑_T^2){\displaystyle \frac{𝒑_T𝑺_T}{M}}\gamma _5`$ (11)
$`+h_{1T}(x,𝒑_T^2)\gamma _5\text{/}S_T+h_{1T}^{}(x,𝒑_T^2){\displaystyle \frac{𝒑_T𝑺_T}{M}}{\displaystyle \frac{\gamma _5\text{/}p_T}{M}}\}P_+,`$
where indeed $`\mathrm{\Phi }(x,𝒑_T)=\mathrm{\Phi }_O(x,𝒑_T)+\mathrm{\Phi }_L(x,𝒑_T)+\mathrm{\Phi }_T(x,𝒑_T)`$. As before, $`f_{\mathrm{}}`$, $`g_{\mathrm{}}`$ and $`h_{\mathrm{}}`$ indicate unpolarized, chirality and transverse spin distributions. The subscripts $`L`$ and $`T`$ indicate the target polarization, and the superscript $``$ signals explicit presence of transverse momentum of partons. Using the notation $`f^{(1)}(x,𝒑_T^2)(|𝒑_T|^2/2M^2)f(x,𝒑_T^2)`$, one sees that $`f_1(x,𝒑_T^2)`$, $`g_1(x,𝒑_T^2)=g_{1L}(x,𝒑_T^2)`$ and $`h_1(x,𝒑_T^2)=h_{1T}(x,𝒑_T^2)+h_{1T}^{(1)}(x,𝒑_T^2)`$ are the functions surviving $`p_T`$-integration.
To put bounds on the transverse momentum dependent functions, we again make the matrix structure explicit, following the same procedure we used in the previous simpler case in which no $`p_T`$ was taken into account. We find for $`M=(\mathrm{\Phi }(x,𝒑_T)\gamma ^+)^T`$ the full spin matrix (for simplicity we do not explicitly indicate the $`x`$ and $`𝒑_T^2`$ dependence of the distribution functions)
$$\stackrel{~}{M}=\left(\begin{array}{cccc}f_1+g_{1L}& \frac{|p_T|}{M}e^{i\varphi }g_{1T}& \frac{|p_T|}{M}e^{i\varphi }h_{1L}^{}& 2(h_{1T}+h_{1T}^{(1)})\\ & & & \\ \frac{|p_T|}{M}e^{i\varphi }g_{1T}^{}& f_1g_{1L}& \frac{|p_T|^2}{M^2}e^{2i\varphi }h_{1T}^{}& \frac{|p_T|}{M}e^{i\varphi }h_{1L}^{}\\ & & & \\ \frac{|p_T|}{M}e^{i\varphi }h_{1L}^{}& \frac{|p_T|^2}{M^2}e^{2i\varphi }h_{1T}^{}& f_1g_{1L}& \frac{|p_T|}{M}e^{i\varphi }g_{1T}^{}\\ & & & \\ 2(h_{1T}+h_{1T}^{(1)})& \frac{|p_T|}{M}e^{i\varphi }h_{1L}^{}& \frac{|p_T|}{M}e^{i\varphi }g_{1T}& f_1+g_{1L}\end{array}\right)$$
where $`\varphi `$ is the azimuthal angle of the transverse momentum vector. Here, we have left out the T-odd functions. But time-reversal invariance was not imposed in the parameterization of $`(\mathrm{\Phi }(x,𝒑_T)\gamma ^+)`$ in Eqs. (11), allowing for non-vanishing T-odd functions $`f_{1T}^{}(x,𝒑_T^2)`$ and $`h_1^{}(x,𝒑_T^2)`$. They can be easily incorporated as the imaginary parts of the functions $`g_{1T}(x,𝒑_T^2)`$ and $`h_{1L}^{}(x,𝒑_T^2)`$, to be precise $`g_{1T}g_{1T}+if_{1T}^{}`$ and $`h_{1L}^{}h_{1L}^{}+ih_1^{}`$. Possible sources of T-odd effects in the initial state have been discussed in Refs .
This matrix is particularly relevant, as it illustrates the full quark spin structure accessible in a polarized nucleon , which is equivalent to the full spin structure of the forward antiquark-nucleon scattering amplitude. Bounds to insure positivity of any matrix element can be obtained by looking at the 1-dimensional and 2-dimensional subspaces and at the eigenvalues of the full matrix. The 1-dimensional subspaces give the trivial bounds
$$f_1(x,𝒑_T^2)0,$$
(12)
$$|g_{1L}(x,𝒑_T^2)|f_1(x,𝒑_T^2).$$
(13)
From the 2-dimensional subspaces we get
$`|h_1|{\displaystyle \frac{1}{2}}\left(f_1+g_{1L}\right)f_1,`$ (14)
$`|h_{1T}^{(1)}|{\displaystyle \frac{1}{2}}\left(f_1g_{1L}\right)f_1,`$ (15)
$`|g_{1T}^{(1)}|^2{\displaystyle \frac{𝒑_T^2}{4M^2}}\left(f_1+g_{1L}\right)\left(f_1g_{1L}\right){\displaystyle \frac{𝒑_T^2}{4M^2}}f_1^2,`$ (16)
$`|h_{1L}^{(1)}|^2{\displaystyle \frac{𝒑_T^2}{4M^2}}\left(f_1+g_{1L}\right)\left(f_1g_{1L}\right){\displaystyle \frac{𝒑_T^2}{4M^2}}f_1^2,`$ (17)
where, once again, we did not explicitly indicate the $`x`$ and $`𝒑_T^2`$ dependence to avoid too heavy a notation. Besides the Soffer bound, Eq. (14), new bounds for the distribution functions are found. In particular, one sees that functions like $`g_{1T}^{(1)}(x,𝒑_T^2)`$ and $`h_{1L}^{(1)}(x,𝒑_T^2)`$ appearing in azimuthal asymmetries in leptoproduction are proportional to $`|𝒑_T|`$ for small $`p_T`$.
Before sharpening these bounds via the eigenvalues, it is convenient to introduce two positive definite functions $`A(x,𝒑_T^2)`$ and $`B(x,𝒑_T^2)`$ such that $`f_1=A+B`$ and $`g_1=AB`$ and define
$`h_1(x,𝒑_T^2)=\alpha A,`$ (18)
$`h_{1T}^{(1)}(x,𝒑_T^2)=\beta B,`$ (19)
$`g_{1T}^{(1)}(x,𝒑_T^2)=\gamma {\displaystyle \frac{|p_T|}{M}}\sqrt{AB},`$ (20)
$`h_{1L}^{(1)}(x,𝒑_T^2)=\delta {\displaystyle \frac{|p_T|}{M}}\sqrt{AB},`$ (21)
where the functions $`\alpha (x,𝒑_T^2)`$, $`\beta (x,𝒑_T^2)`$, $`\gamma (x,𝒑_T^2)`$ and $`\delta (x,𝒑_T^2)`$ have absolute values in the interval $`[1,1]`$. Note that $`\alpha `$ and $`\beta `$ are real-valued but $`\gamma `$ and $`\delta `$ are complex-valued, the imaginary part determining the strength of the T-odd functions.
Next we sharpen these bounds using the eigenvalues of the matrix, which are given by
$`e_{1,2}=(1\alpha )A+(1+\beta )B\pm \sqrt{4AB|\gamma +\delta |^2+((1\alpha )A(1+\beta )B)^2},`$ (22)
$`e_{3,4}=(1+\alpha )A+(1\beta )B\pm \sqrt{4AB|\gamma \delta |^2+((1+\alpha )A(1\beta )B)^2}.`$ (23)
Requiring them to be positive can be converted into the conditions
$`A+B0.`$ (24)
$`|\alpha A\beta B|A+B,\text{i.e.}|h_{1T}(x,𝒑_T^2)|f_1(x,𝒑_T^2)`$ (25)
$`|\gamma +\delta |^2(1\alpha )(1+\beta ),`$ (26)
$`|\gamma \delta |^2(1+\alpha )(1\beta ).`$ (27)
It is interesting for the phenomenology of deep inelastic processes that a bound for the transverse spin distribution $`h_1`$ is provided not only by the inclusively measured functions $`f_1`$ and $`g_1`$, but also by the functions $`g_{1T}(x,𝒑_T^2)`$ and $`h_{1L}^{}(x,𝒑_T^2)`$, responsible for specific azimuthal asymmetries . This is illustrated in Fig. 2.
A perfectly analogous calculation con be performed for fragmentation functions, which describe the hadronization process of a parton into the final detected hadron. In this case the transverse momentum dependent correlator is
$$\mathrm{\Delta }_{ij}(z,𝒌_T)=\underset{X}{}\frac{d\xi ^{}d^2𝝃_T}{(2\pi )^3}e^{ik\xi }0|\psi _i(\xi )|P_h,XP_h,X|\overline{\psi }_j(0)|0|_{\xi ^+=0},$$
(28)
(see Fig. 3) depending on $`z=P_h^+/k^+`$ and the quark transverse momentum $`k_T`$ leading to a hadron with $`P_{hT}=0`$. A simple boost shows that this is equivalent to a quark producing a hadron with transverse momentum $`P_h=zk_T`$ with respect to the quark.
Like $`\mathrm{\Phi }`$, $`\mathrm{\Delta }`$ is parameterized in terms of unpolarized, chirality and transverse-spin fragmentation functions , denoted by capital letters $`D_{\mathrm{}}`$, $`G_{\mathrm{}}`$, and $`H_{\mathrm{}}`$, respectively. For the fragmentation process, time-reversal invariance cannot be imposed , and the T-odd fragmentation functions $`D_{1T}^{}`$ and $`H_1^{}`$ play a crucial role in some azimuthal spin asimmetries as we shall discuss later on.
All the bounds obtained for the distribution functions can be rephrased in terms of the corresponding fragmentation functions. For instance, the relevant bounds for the Collins function $`H_1^{(1)}`$, describing the fragmentation of a transversely polarized quark into a (spin zero) pion becomes
$$H_1^{(1)}(z_\pi ,𝑷_\pi ^2)\frac{|𝑷_\pi |}{2z_\pi M_\pi }D_1(z_\pi ,𝑷_\pi ^2),$$
(29)
while for the other T-odd function $`D_{1T}^{(1)}`$, describing fragmentation of an unpolarized quark into a polarized hadron such as a $`\mathrm{\Lambda }`$ one has
$$D_{1T}^{(1)}(z_\mathrm{\Lambda },𝑷_\mathrm{\Lambda }^2)\frac{|𝑷_\mathrm{\Lambda }|}{2z_\mathrm{\Lambda }M_\mathrm{\Lambda }}D_1(z,𝑷_\mathrm{\Lambda }^2).$$
(30)
Similarly to what happened for the distribution functions, a bound for the transverse spin fragmentation $`H_1^{}`$ is provided not only by the inclusive function $`D_1`$ but, when we sharpen the bounds by requiring positivity of the eigenvalues of the full matrix, the magnitude of $`H_1^{}`$ also constrains the magnitude of $`H_1`$ .
Recently SMC , HERMES and LEP have reported preliminary results for azimuthal asymmetries. More results are likely to come in the next few years from HERMES, RHIC and COMPASS experiments. Although much theoretical work is needed, for instance on factorization, scheme ambiguities and the stability of the bounds under evolution , these future experiments may provide us with the knowledge of the full helicity structure of quarks in a nucleon. The elementary bounds derived in this paper can serve as important guidance to estimate the magnitudes of asymmetries expected in the various processes.
## 3 An example, relevant for JLAB@12 GeV
The asymmetries for which evidence recently has been found are mostly single spin asymmetries involving T-odd fragmentation functions, such as the Collins function $`H_1^{}`$. We would like to discuss here a measurement that can be combined with the measurement of the inclusive structure function $`g_2(x)`$ in deep inelastic scattering off a transversely polarized target at large $`x`$.
The relevant kinematic variables are illustrated in Fig. 4, where also the scaling variables are introduced.
Most often one considers the cross section integrated over all transverse momenta. But we emphasize that in principle the cross section can depend on a transverse vector, for which we use $`𝒒_T`$ = $`P_h/z_h`$. This vector either represents the transverse momentum of the photon momentum $`q`$ (with respect to the two hadrons, target and produced hadron) or the transverse momentum of the produced hadron (with respect to the target and photon momenta). Introducing the weighted cross sections
$`W_{P_eP_HP_h}{\displaystyle 𝑑\varphi ^{\mathrm{}}d^2𝒒_TW(Q_T,\varphi _h^{\mathrm{}},\varphi _S^{\mathrm{}},\varphi _{S_h}^{\mathrm{}})\frac{d\sigma _{P_eP_HP_h}}{dx_Bdydz_hd\varphi ^{\mathrm{}}d^2𝒒_T}},`$ (31)
where $`W`$ is some weight depending on azimuthal angles and transverse momentum and the subscripts $`P_e`$, $`P_H`$ and $`P_h`$ are the polarizations of lepton, target and produced hadron respectively, we can construct several asymmetries.
To illustrate the weights, let’s consider an easy example: the standard $`𝒒_T`$-integrated 1-particle inclusive unpolarized cross section,
$$\frac{d\sigma _{OO}}{dx_Bdydz_h}=\frac{2\pi \alpha ^2s}{Q^4}\underset{a,\overline{a}}{}e_a^2\left(1+(1y)^2\right)x_Bf_1^a(x_B)D_1^a(z_h),$$
(32)
in this language becomes
$$1_{OO}=\frac{2\pi \alpha ^2s}{Q^4}\underset{a,\overline{a}}{}e_a^2\left(1+(1y)^2\right)x_Bf_1^a(x_B)D_1^a(z_h).$$
(33)
One of the leading asymmetries involving the function $`g_{1T}(x,𝒑_T^2)`$ discussed in the previous section is an asymmetry for longitudinally polarized leptons off a transversely polarized nucleon
$$\frac{Q_T}{M}\mathrm{cos}(\varphi _h^{\mathrm{}}\varphi _S^{\mathrm{}})_{LT}=\frac{2\pi \alpha ^2s}{Q^4}\lambda _e|𝑺_T|y(2y)\underset{a,\overline{a}}{}e_a^2x_Bg_{1T}^{(1)a}(x_B)D_1^a(z_h),$$
(34)
Since the fragmentation function involved is the standard leading one for unpolarized quarks into unpolarized or spin 0 hadrons, one can consider it for pion production, for which the fragmentation functions are reasonably well-known .
As we mentioned before, it is interesting to do this measurements together with the $`g_2`$-measurement which, expressed as an (inclusive) asymmetry is given by
$$\mathrm{cos}\varphi _S^{\mathrm{}}_{LT}=\lambda _e|𝑺_T|y\sqrt{1y}\underset{a,\overline{a}}{}e_a^2\frac{Mx_B^2}{Q}g_T^a(x_B)$$
(35)
where $`g_T^a(x)=g_1^a(x)+g_2^a(x)`$. The comparison of the inclusive measurement of $`g_2`$ and the semi-inclusive measurement of $`g_{1T}^{(1)}`$ would enable one to test the relation ,
$$g_2^a(x)=\frac{d}{dx}g_{1T}^{(1)a}.$$
(36)
relating a twist three function to a transverse momentum dependent function. At present this relation can be used to estimate the function $`g_{1T}(x)`$ from existing $`g_2`$ measurements, of course in the same flavor averaged way as an inclusive measurement allows. The result is shown in Fig. 5, taken from ref. . A measurement of the asymmetry in Eq. 34 allows an independent measurement of $`g_{1T}`$ and a test of the above relation.
## Acknowledgments
We would like to thank Elliot Leader for useful discussions. This work is part of the research program of the Foundation for Fundamental Research on Matter (FOM) and the Dutch Organization for Scientific Research (NWO). It is also part of the TMR program ERB FMRX-CT96-0008.
|
warning/0005/gr-qc0005128.html
|
ar5iv
|
text
|
# New Developments in the Search for the Topology of the Universe
## 1 BASICS
### 1.1 Historical elements
In the standard cosmological framework, the universe is decribed by a Friedmann-Lemaitre solution, the spatial sections of which being usually assumed to be simply–connected. Einstein equations being local, they allow us to determine the local geometry but they give no complete information about the global structure of the universe, i.e. about its topology, even if the geometry constrains to some extent the topology. This was pointed out by Friedmann himself just after he proposed its cosmological solution . The indeterminacy about the global topology of the cosmological solutions was raised out as soon as Einstein proposed the first cosmological solution of his equations ; the Einstein static universe assumed spatial sections with the topology of the hypersphere $`S^3`$ , although de Sitter stressed that the same geometry could admit as well the projective space $`P^3=S^3/Z^2`$ as spatial sections . These two solutions were locally identical (same metric) but differed by their topology, i.e. by the choice of boundary conditions.
Many arguments were then advanced in favour of a simply–connected universe, such as the simplicity or economy principle stating that one should introduce as few parameters as possible in physical modelling. Indeed, such an argument is very unclear. There has been some tendency to favor spaces with finite volume. The eternal and infinite space of Newtonian physics was for instance leading to logical difficulties such as the Olbers paradox , and the finite cosmological solution proposed by Einstein in 1917 was received as a smart way of solving such paradoxes. The Mach’s principle based on the idea that the local inertia is determined by the distribution of masses in the whole universe, also tends to favor universes with a finite volume. If space is simply–connected then it is finite if and only if it is locally elliptic, whereas if multi–connected it can also be locally hyperbolic or Euclidean. It has also been argued that infinite spaces were unaesthetic since “all phenomenon with a non vanishing probability must happen somewhere else” . This argument was used by Ellis to conclude that if space was infinite then one can avoid the former conclusion by dropping the assumption of spatial homogeneity and by arguing that we were living in a particular place of the universe . To finish, many arguments coming from some ideas in quantum cosmology are in favor of a finite volume space . Indeed, none of these arguments can help us to determine the shape and size of our universe. So, we are lead to the question, can we detect or constrain observationnally the topology of our universe ?
Astonishingly enough, our century has on one hand seen the birth of the geometrical description of the universe and of a non static curved spacetime. On the other, mathematicians have developped the classification of 3D–manifolds. The first motivation came from crystallography, which lead to the classification of Euclidean 3D–manifolds and was achieved in 1934 . The classification of locally elliptic 3-manifold was set by F. Klein and W. Killing and solved by J.A. Wolf in 1960 . The classification of locally hyperbolic manifold was started in the 70’s by W. Thurston and is not yet achieved . Thus both elements necessary to answer the question of cosmic topology were developped at the same time without interacting so much. For more details concerning this interesting modern quest, one can see .
### 1.2 Mathematical elements
In relativistic cosmology, our universe is described by a globally hyperbolic 4–manifold $`=\mathrm{\Sigma }\times R`$ , where the spatial sections $`\mathrm{\Sigma }`$ are homogeneous and isotropic Riemannian 3–manifolds. From a topological point of view, it is convenient to decribe such a manifold by its fundamental polyhedron (hereafter FP), which is convex. Its faces are associated by pairs through the elements of a holonomy group $`\mathrm{\Gamma }`$ which is acting freely and discontinuously on $`\mathrm{\Sigma }`$ (see for mathematical definitions and for an introduction to topology in the cosmological context). The holonomy group is isomorphic to the fundamental group $`\pi _1(\mathrm{\Sigma })`$. Using the property (see e.g. ):
$$\pi _1()=\pi _1(\mathrm{\Sigma }\times R)\pi _1(R)\times \pi _1(\mathrm{\Sigma })\pi _1(\mathrm{\Sigma }),$$
(1)
we deduce that the study of the topology of the universe reduces to the study of the topology of its spatial sections.
What are the allowed homogeneous 3–manifolds usable in cosmology ? According to the sign of the spatial curvature $`K`$, the universal covering space of the spatial sections (which will be referred to as $`\stackrel{~}{\mathrm{\Sigma }}`$) can be described by the Euclidean space $`E^3`$, the hypersphere $`S^3`$ or the 3-hyperboloid $`H^3`$ if $`K`$ vanishes, is positive or negative respectively. Thus, the homogeneous and isotropic 3-manifolds will be of the form
$$E^3/\mathrm{\Gamma },S^3/\mathrm{\Gamma }\text{and}H^3/\mathrm{\Gamma }.$$
(2)
Let us just summarize some of the properties of each family and also introduce $`G`$, the full isometry group of $`\stackrel{~}{\mathrm{\Sigma }}`$ keeping the metric invariant. Indeed $`\mathrm{\Gamma }`$ will be a discrete sub-group of $`G`$.
1. locally Euclidean 3–manifolds: The covering space in the Euclidean space $`\stackrel{~}{\mathrm{\Sigma }}=E^3`$ and the isometry group is $`G=R^3\times SO(3)`$. The metric can be written under the form
$$ds^2=a^2(\eta )\left\{d\eta ^2+d\chi ^2+\chi ^2d\mathrm{\Omega }\right\},$$
(3)
where $`\eta `$ is the conformal time, $`a(\eta )`$ the scale factor, $`\chi `$ the radial coordinate and $`d\mathrm{\Omega }`$ the unit solid angle. The generator of $`\mathrm{\Gamma }`$ are the identity, the translations, the reflexions and the helicoidal motions.
These transformations generate 18 different spaces among which 17 are multi–connected and correspond to the 17 crystallographic groups . 10 spaces are compact and among them 6 are orientable. The description of these spaces can be found in e.g. .
2. locally elliptic 3–manifolds : The covering space in the elliptic space $`\stackrel{~}{\mathrm{\Sigma }}=S^3`$ and the isometry group is $`G=SO(4)`$. The metric can be written as
$$ds^2=a^2(\eta )\left\{d\eta ^2+d\chi ^2+\mathrm{sin}^2\chi d\mathrm{\Omega }\right\}.$$
(4)
With the curvature radius as unit length, the volume of the hypersphere is
$$\text{Vol}(S^3)=_0^\pi 4\pi \mathrm{sin}^2\chi d\chi =2\pi ^2.$$
(5)
Wolf gives an exhaustive description of the allowed discrete groups $`\mathrm{\Gamma }`$. They are the cyclic groups of order $`p`$ ($`p2`$), the dihedral groups of order $`2m`$, and the symmetry groups of the tetrahedron, octaedron and icosahedron.
If we denote $`|\mathrm{\Gamma }|`$ the order of the holonomy group, it is straigthforward to show that
$$\text{Vol}(S^3/\mathrm{\Gamma })=2\pi ^2/|\mathrm{\Gamma }|,$$
(6)
which tells us that the volume of an elliptic manifold is a topological invariant (degenerate since two different groups with same order will have same volume).
3. locally hyperbolic 3–manifolds: The covering space in the hyperbolic space $`\stackrel{~}{\mathrm{\Sigma }}=H^3`$ and the isometry group is $`G=PSL(2,C)SL(2,C)/Z_2`$. The metric can be written as
$$ds^2=a^2(\eta )\left\{d\eta ^2+d\chi ^2+\mathrm{sinh}^2\chi d\mathrm{\Omega }\right\}.$$
(7)
The classification of compact hyperbolic manifolds is not achieved yet and we will just describe two of them and give some useful properties. This classification relies on the rigidity theorem stating that the geometry is fixed by the topology, a consequence of which being that the volume and other characteristic lengths are topological invariants, so that
$$\pi _1(X)\pi _1(Y)\text{Vol}(X)=\text{Vol}(Y).$$
(8)
It has also been shown that there is a minimal allowed volume
$$\text{Vol}(\mathrm{\Sigma })\text{Vol}_{\mathrm{min}}=0.166.$$
(9)
We also define the outside radius $`r_+`$, the radius of the smallest geodesic ball that contains the FP, the inside radius $`r_{}`$, the radius of the biggest geodesic ball contained in the FP, and the injectivity radius $`r_{\mathrm{inj}}`$, half the length of the shortest closed geodesic. $`r_+`$, $`r_{}`$ and $`r_{\mathrm{inj}}`$ are altogether topological invariants.
The classification and the description of the known compact hyperbolic manifolds can be obtained by using the software SnapPea which provides the FP, the generators of the holomy group and the characteristic lengths. Examples of such manifolds can be found in e.g. . Here we mention for a later use a manifold of special interest, the Weeks space , which is the smallest known compact hypermabolic manifold. Its FP has 18 faces and its geometrical characteristics are
$$\text{Vol}=0.94272,r_+=0.7525,r_{}=0.5192,r_{\mathrm{inj}}=0.2923$$
(10)
### 1.3 Physical and observational introduction
At the time being, two classes of methods to detect and/or constrain the topology of the spatial sections of our universe have been proposed and investigated. They use two classes of cosmological observations, namely the 2D cosmic microwave background data and the 3D catalogs of discrete sources.
1. 2D–methods: The cosmic microwave background is composed of all the photons emitted during the decoupling period between matter and radiation at a redshift of $`z1100`$. It is observed as a black body with a temperature of $`2.7`$ K with small fluctuations of order $`10^5`$. The temperature anisotropies reflect the small inhomogeneities that ultimately lead to the observed structures in the universe.
When studying these anisotropies in a multi–connected universe, one has to take into account the fact that we are in a compact manifold and thus that the wavelengths must be discretized. Stevens et al. showed that in a cubic hypertorus one should see a cut–off in the two–point correlation function if $`L/R_H0.8`$. This analysis was then generalised to non cubic hypertorus and to the six flat manifolds . However, in compact hyperbolic manifolds there are super-curvature modes , which implies that no such cut–off exists .
It has then been realised that the patterns on such an extended surface must be correlated if the universe is multi-connected. Levin et al. studied these correlations in the flat manifolds and Cornish et al. developped a topology independent method to look for a topological signature in the cosmic microwave background map as expected from the MAP and Planck satellites.
All these studies have been performed assuming that the small initial inhomogeneities where generated during an inflationary phase. There is however another process that could generate such inhomogeneities, namely topological defects produced during a symmetry breaking phase transition in the early universe. Uzan and Peter showed that the topology implies a constraint on the topological defects network from which it was deduced that there will be a cut–off in the angular power spectrum of the temperature anisotropies, the cut–off value depending only on the characteristic size of the universe and on the cosmological parameters. Note that such considerations also provide another method for constraining the topology, namely the observation of an extended defect at a redshift $`z`$ will give a lower bound on the size of the universe.
In conclusion, at the moment 2D–methods constrain flat manifolds to $`L/R_H0.8`$, and there is no convincing constraint on compact hyperbolic manifolds . Some powerful methods can be hoped to be used when high resolution cosmic microwave background maps are achieved.
2. 3D–methods: In the framework of pre–relativistic cosmology, it was already pointed out by Schwarzschild that if space is multi–connected, one could see multiple images of astronomical objects , whereas Friedmann did the same in the framework of general relativity. All 3D–methods for testing the space topology are based on this fact. We shall develop their description in the following sections.
In these Proceedings, we describe the developments of the 3D–methods, from the Schwarzschild initial idea of the existence of multiple images of the same object, until its statistical implementation under the form of cosmic cristallography. We then carefully study the method and explain why, contrary to what was previously thought, it won’t work in locally hyperbolic universes. We then present attempts to generalise it and finish by describing a new and promising “CCP” method. This will be examplified by numerical simulations with depiction of color pictures in order to make the subject as clear as possible.
## 2 FROM THE ORIGINAL IDEA TO THE CRYSTALLOGRAPHIC METHOD
### 2.1 Initial idea
As stressed above, if space is multi–connected, one should see multiple “topological” images of the same object. This is illustrated on figure 1 where we have used a 2–D example. In the universal covering space, the spacetime trajectories of the different images ($`A,B,\mathrm{}`$) of a given object will intersect the observer past light cone at different times ($`a,b,\mathrm{}`$). The same object will then been seen at different stages of its evolution. Indeed in a 4D universe, these images will also be seen in different directions of the sky.
The question is indeed to find methods to implement such a property in order to detect or to put lower bounds on the size of the universe.
### 2.2 Multiple images of individual objects
Some specific objects have been tried to be recognised individually. An extensive description of these attempts can be found in and we just stick to a brief description.
* Milky Way: Can we recognise our own Galaxy ? The maximum distance up to which we would be able to do that has been estimated as $`7.5h^1\text{Mpc}`$ by Sokolov and Shvartsmann which gave the observational bound $`r_+15h^1\text{Mpc}`$. Then, Fagundes and Wichoski proposed that our galaxy was a quasar long ago but indeed, for physical reasons due to realistic quasar models, our Galaxy cannot have been a quasar in its past. Fagundes also studied the occurrence of images of the Milky Way in the Best hyperbolic space model . Presently, no source has been identified as an image of our own Galaxy.
* Galaxy clusters : Gott used the Coma cluster has a candidate for the search of ghosts images and obtained the bound $`r_+60h^1\text{Mpc}`$. He also performed simulations in $`T^3`$ to provide a pattern of clustering and a correlation function in agreement with observations of nearby galaxies. Roukema and Edge proposed 3 candidates as ghost images for the Coma cluster. Further investigations are needed.
Sokoloff and Shvartsmann considered the Abell catalog of clusters and gave the bound $`r_+600h^1\text{Mpc}`$ in Euclidean space, whereas Demianski and Lapucha searched unsuccessfully for opposite pairs in a catalog of 1889 clusters. Lehoucq et al. developped a statistical method that is presented below to obtain $`r_+650h^1\text{Mpc}`$ in Euclidean space and Fetisova et al. reported a spike about $`125h^1\text{Mpc}`$ in the correlation function of rich clusters, but such an evidence does not necessarily supports multi–connectedness (see e.g. ).
* Quasars: Quasars occupy a large volume of the universe ($`z3`$). However, they are probably short-lived compared to the expected time necessary for a photon to go around the universe. If $`\tau _q10^9\text{years}`$, then one can hope to test topology on scale $`r_+c\tau _q`$. So, they are of interest even if $`\tau _q10^8\text{years}`$. Paál remarked that although individual quasars may have short lifetime, they may be part of larger associations which survive longer.
Roukema tried to identify quintuplets and quadruplets of quasars in a catalog of 4554 quasars. He found 27 identifications whereas he found $`26\pm 1`$ identifications in simulations of a simply–connected universe. Fagundes tried to discuss the redshift contreversy in the Best model.
Until now, the best bound obtained by 3D–methods was $`r_+650h^1\text{Mpc}`$ and hold only in Euclidean universe. There is still some observational room for the topologies of an hyperbolic universe, even on scales significantly shorter than the horizon radius.
All the methods trying to recognize a particular object suffer from the same major problems:
1. distance determination: catalogs give the angular position and the redshift. One has then to go from redshift space to real space, which depends on the knowledge of the cosmological constant and of the density parameter. This explained why most of the former studies were performed assuming $`\mathrm{\Omega }_0=1,\mathrm{\Omega }_\mathrm{\Lambda }=0`$. This point will be develop later.
2. peculiar velocities: objects are not comoving, so that the position of their topological images are shifted. If their velocity is of order $`v_p500\text{km.s}^1`$, then, assuming that the time for a photon to go around the universe can be estimated by $`tr_+/c`$, one obtains the maximal precision on the determination of position $`\delta \chi (v_p/c)\times r_+`$.
3. morphology: when one tries to associate two images to the same physical object, one has to be aware that, in general, the object will not be seen in the same angle and at the same stage of its evolution. It will then be very difficult to recognise it and the identification will strongly depend on galactic evolution models.
Conversely, the discovery of topology by any means could provide informations such as peculiar velocities with a better accuracy . We will also be able to see the same object at different stages of its evolution and on different angles, which will constrain galactic evolution models.
### 2.3 The crystallographic method
Since there is little chance to recognise different images of a given object, one can try to detect these images statistically, since for instance the number of topological images of a single object in a toroidal universe has been estimated in table 1 of .
The crystallographic method is based on a property of multi–connected universes according to which each topological image of a given object is linked to each other one by the holonomies of space. Indeed, we do not know these holomies as far as we have not determined the topology, but we know that they are isometries. For instance in locally Euclidean universes, to each holonomy is associated a distance $`\lambda `$, equal to the length of the translation by which the fundamental domain is moved to produce the tessellation in the covering space. Assuming the FP contains $`N`$ objects (e.g. galaxy clusters), if we calculate the mutual 3D–distances between every pair of topological images (inside the particle horizon), the distances $`\lambda `$ will occur $`N`$ times for each copy of the fundamental domain, and all other distances will be spread in a smooth way between zero and two times the horizon distance. In a histogram plotting the number of pairs versus their 3D separations, the distances $`\lambda `$ will thus produce spikes. Simulations indeed showed that the pairs between two topological images of the same object drastically emerge from ordinary pairs in the histogram. The applicability of the method in Euclidean spaces has also been discussed by Fagundes and Gaussmann when the size of the physical space is comparable to the horizon size.
Two kind of catalogs of astronomical objects can be thought of to apply this method: the galaxy cluster catalogs, which typically have a redshift depth $`z=1`$, and the quasars catalogs, which typically extend to $`z=3`$. Concerning quasars, even if their lifetime is probably too short to be good candidates for producing topological images, they are usually part of systems that have a much larger lifetime . Let us stress that a recent survey in the Hubble Deep Field south NICMOS field found 17 galaxies with a redshift between 5 and 10 and 5 galaxies with a redshift above 10, among a total of 323 galaxies. This can let us hope to apply this method to deeper catalogs in the future. The angular resolution needed is given by the fact that the objects have a peculiar velocity and that they will not be seen at exactly the same position . Note that the crystallographic method, contrary to the “direct” method which would try to recognize topological images of individual objects, is not plagued by the fact that topological images of the same object are seen at different stages of its evolution.
We present on figure 2 an example from for the flat torus with a cubic FP.
Now, we apply this method to a catalog containing more than 11,000 quasars up to a redshift of $`z3.25`$ (see figure 3 where we have depicted its projection on the celestial sphere and the redshift distribution of its objects). The pair separation histogram is plotted on figure 4 and it exhibits no topological signature. Since respectively 60 %, 80 % and 95 % of the sources are below the redshifts 2, 2.3 and 3, we only test scales smaller that $`z3`$.
In a locally Euclidean universe, the distance-redshift relation is given by
$$\chi (z)=\frac{2c}{a_0H_0}\left(1\frac{1}{\sqrt{1+z}}\right).$$
(11)
We simulated a catalog with the same number of objects and the same redshift distribution to reproduce the real catalog depicted in figure 3. Varying the size of cubic hypertorus, a spike appears when $`L_03000h^1\text{Mpc}`$ for the simulated catalog and thus we obtain the new bound on the size of a locally Euclidean manifold as
$$L_03000h^1\text{Mpc},$$
(12)
which is the best constraint presently available with 3D–methods for locally Euclidean manifolds. Note that it corresponds to
$$\frac{L_0}{R_H}0.5,$$
(13)
which is comparable to the bound obtained with 2D–methods. Note that this result has not been published elsewhere.
## 3 WHY THE CRYSTALLOGRAPHIC METHOD DOES NOT WORK IN HYPERBOLIC SPACES
When one looks carefully at the crystallographic method, one must wonder about the origin of the spikes. As it has recently been explained , two kinds of pairs can create a spike, namely (we keep the notation and vocabulary introduced in )
1. Type I pairs of the form $`\{g(x),g(y)\}`$, since $`\text{dist}[g(x),g(y)]=\text{dist}[x,y]`$ for all points and all elements $`g`$ of $`\mathrm{\Gamma }`$.
2. Type II pairs of the form $`\{x,g(x)\}`$ if $`\text{dist}[x,g(x)]=\text{dist}[y,g(y)]`$ for at least some points and elements $`g`$ of $`\mathrm{\Gamma }`$.
Both families of pairs are depicted on figure 5 for the example of the 2–torus.
### 3.1 Clifford translations
A Clifford translation is an isometry $`g`$ such that the displacement function $`\text{dist}(x,g(x))`$ is constant . This is precisely what is required to get type II pairs in the histogram.
In compact hyperbolic manifolds, $`\text{dist}[x,g(x)]`$ always depends on the position $`x`$ of the source (see for details).
In elliptic spaces, distances are position-independent whenever the holonomy is a Clifford translation. All finite groups of Clifford translations of spheres are the cyclic group, the binary dihedral, tetrahedral, octahedral and icosahedral groups , from which we deduce that the covering transformations which move a source to its nearest neighbours are Clifford translations (although the transformations to more distant neighbours might not be), and type II pairs can be produced.
In locally Euclidean universes, type I and type II pairs are both present. The reason is that the 3-torus has the very special property that the separation distance of gg-pairs (i.e. any pair of images comprising an original and one of its ghosts, or two ghosts of the same object) is independent of the location of the source. In other Euclidean spaces the spectrum of gg-pair distances varies with the location of the source. However all closed Euclidean 3-manifolds have the 3-torus as a covering space, so for each such manifold there will be some distances which are independent of the location of the source. As a consequence, the topological signal expected in the histogram from type I and type II pairs clearly stands out, as was shown in the simulations of .
In conclusion, type II pairs exist only in elliptic and Euclidean universes. The question we still have to answer to judge the efficiency of the crystallographic method in the general case is: what about the amplitude of the spikes related to type I pairs only ?
### 3.2 Cosmological parameters
If the injectivity radius of the space manifold is smaller than the calatog’s length scale, type I pairs will always be present, whatever the curvature. Their number roughly equals the number of copies of the fundamental domain within the catalog’s limits. Following , this number can easily been estimated by computing the ratio between the volume of the geodesic sphere of radius $`\chi (z)`$ and the volume of the manifold. The calculation (in the case of hyperbolic universes) leads to
$$N(\mathrm{\Omega }_{m0},\mathrm{\Omega }_{\mathrm{\Lambda }0};z<Z)=\frac{\pi \left(\mathrm{sinh}2\chi (Z)2\chi (Z)\right)}{\text{Vol}(\mathrm{\Sigma })},$$
(14)
where $`\chi (Z)`$ is the radial distance corresponding to the redshift $`Z`$, given
$$\chi (z)_{a_0}^a\frac{da}{a\dot{a}}=_{\frac{1}{1+z}}^1\frac{\sqrt{1\mathrm{\Omega }_{m0}\mathrm{\Omega }_{\mathrm{\Lambda }0}}dx}{x\sqrt{\mathrm{\Omega }_{\mathrm{\Lambda }0}x^2+(1\mathrm{\Omega }_{m0}\mathrm{\Omega }_{\mathrm{\Lambda }0})+\frac{\mathrm{\Omega }_{m0}}{x}}},$$
(15)
where $`\mathrm{\Omega }_{m0}`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$ are respectively the density parameter and the cosmological constant. We have plotted on figure 6 an estimation of the number of copies of an object taken respectively in a catalog of galaxy clusters and in a catalog of quasars, as a function of the cosmological parameters.
The small number of type I pairs in hyperbolic manifolds is due to the property that the volume of the manifold is fixed once the topology is determined (the rigidity theorem), contrarily to Euclidean spaces where the characteristic sizes and the volume of the FP can be chosen at will (since $`K=0`$, the geometry does notimpose any characteristic size).
In conclusion, type I pairs will indeed contribute to spikes in the pair separation histogram but the amplitude of these spikes will be of the same order than the statistical noise.
### 3.3 Numerical simulations
The former conclusion can be tested numerically. For that purpose, we concentrate on locally hyperbolic three–dimensional manifolds. It can be first embedded in a four–dimensional Minkowski space by introducing the set of coordinates $`(x^\mu )_{\mu =\mathrm{0..3}}`$ related to the intrinsic coordinates $`(\chi ,\theta ,\phi )`$ through (see e.g. )
$`x_0`$ $`=`$ $`\mathrm{cosh}\chi `$
$`x_1`$ $`=`$ $`\mathrm{sinh}\chi \mathrm{sin}\theta \mathrm{sin}\phi `$
$`x_2`$ $`=`$ $`\mathrm{sinh}\chi \mathrm{sin}\theta \mathrm{cos}\phi `$
$`x_3`$ $`=`$ $`\mathrm{sinh}\chi \mathrm{cos}\theta ,`$ (16)
so that the three–dimensional hyperboloid $`H^3`$ has the equation
$$x_0^2+x_1^2+x_2^2+x_3^2=1.$$
(17)
With these notations, the comoving spatial distance between two points of comoving coordinates $`x`$ and $`y`$ can be computed directly in the Minkowski space by
$$d[x,y]=\mathrm{arg}\mathrm{cosh}\left[x^\mu y_\mu \right],$$
(18)
where $`x_\mu =\eta _{\mu \nu }x^\nu `$, $`\eta _{\mu \nu }`$ being the Minkowskian metric coefficients.
To generate an idealised catalog, $`𝒞`$, we first distribute homogeneously $`N`$ objects in the FP, then we by unfold the catalog by applying the generators of the holonomy group. Now, to generate a catalog with the required depth $`z_{\mathrm{max}}`$ in redshift, the set of all points obtained as above must be truncated in order to keep only the points $`x`$ such that $`\text{dist}[0,x]\chi (\mathrm{\Omega }_{m0},\mathrm{\Omega }_{\mathrm{\Lambda }0};z_{\mathrm{max}})`$, given by equation (15). This amounts to select objects located within the geodesic ball of radius $`\chi (\mathrm{\Omega }_{m0},\mathrm{\Omega }_{\mathrm{\Lambda }0};z_{\mathrm{max}})`$ centered onto an observer right at the centre.
The unfolding of Weeks manifold is shown in figure 7, where we have depicted the central cell and its 18 nearest neighbours in Klein coordinates. In figure 8, we have plotted a simulated catalog showing all points, including those outside the horizon distance.
We then compute all the 3D separations between all the pairs of $`𝒞(z)`$ and we plot the histogram of the number of pairs with a given separation.
Indeed the former procedure applies when the observer stands at the center of the polyhedron $`(\chi =0)`$. In , we explain how to take into account an off-centered observer.
In Lehoucq et al. , we have generated some pair histograms in a universe whose spatial sections have the topology of the Weeks manifold, using $`N=1000`$ objects in the FP. In figure 9 we give an example wher $`\mathrm{\Omega }_\mathrm{\Lambda }=0`$, $`\mathrm{\Omega }_{m0}=0.2`$ and where the observer stands at the center of the polyhedron ($`\chi =0`$). The dependence on the cosmological parameters, on the position of the observer and on the catalog depth where studied in details in .
As expected, none of the plots exhibit spikes. This is due to the combined effects of the geometry and of the cosmological parameters values.
To summarize the failure of cosmic crystallography for hyperbolic manifolds:
1. hyperbolic manifolds are such that $`\text{dist}[x,g(x)]`$ depends on $`x`$, so that there is no amplification for the type II pairs $`\{x,g(x)\}`$, whereas $`\text{dist}[x,g(x)]=\text{dist}[y,g(y)]`$ in the Euclidean case. This suppresses the corresponding spikes.
2. The spikes associated to the isometries (i.e. such that $`g\mathrm{\Gamma },\text{dist}[g(x),g(y)]=\text{dist}[x,y]`$) must remain. But, given the cosmological parameters, we have shown that the number of topological images is too low to create such spikes associated to type I pairs.
## 4 GENERALISING THE CRYSTALLOGRAPHIC METHOD
The next question is: can we generalise the crystallographic method in order to make it work in any case ?
Two solutions were proposed almost immediately to improve the crystallographic method and to extract from 3D catalogs the signature of type I pairs:
1. Fagundes and Gaussman substracted the pair separation histogram for a simulated catalog (with the same number of objects and the same cosmological parameters) in a simply-connected universe to the observed pair separation histogram. The result is “a plot with much oscillation on small scale, modulated by a long wavelength quasi-sinusoidal pattern”. The statistical relevance of this signature still has to be investigated. To illustrate this point we show on figure 10 the result of a simulation where we have substracted to realisations on a simply-connected Friedmann-Lemaitre universe. This has to be compared with figure 1 from and let us think that is method is not efficient for determining a relevant signature of the topology of the universe.
2. Gomero et al. proposed to split the catalog in “smaller” catalogs and average the pair separation histograms built from each sub–catalog to reduce the statistical noise and extract a visible signal of the non-translational isometries. The feasability of this method has not yet been demonstrated.
## 5 A NEW EFFICIENT STATISTICAL METHOD
We now present a new method based on the property that whatever the topology, type I pairs will always exist. Indeed, as shown ahead, this does not lead to any observable spike in the pair separation histogram. We thus had to find a method to enhance the topological signal by “collecting” all the homologous pairs together. This can be achieved by building a collecting correlated pairs method (hereafter CCP). Such an approach (described in details in and below) is not able to determine the exact topology, but will provide a signature of the compactness of the spatial sections, which is indeed a first step toward the determination (or the rejection) of the cosmic topology.
In this section, we recall the construction of the CCP–index and the main resuts we have obtained.
### 5.1 Basic idea and general method
As stressed in the previous section, type I pairs will always exist in multi–connected spaces as soon as one of the characteristic scales of the fundamental domain is smaller than the Hubble scale. Defining $`g_i|_{1i2K}`$ as the $`2K`$ generators of $`\mathrm{\Gamma }`$ and referring to $`x`$ as the position of the image in the universal covering space $`X`$, we have:
1. $`x,yX`$, $`g\mathrm{\Gamma }`$,
$`\text{dist}[g(x),g(y)]`$ $`=`$ $`\text{dist}[x,y].`$ (19)
We will refer to these pairs as $`xy`$–pairs.
2. $`xX`$, $`g_1,g_2\mathrm{\Gamma }`$,
$`\text{dist}[g_1(x),g_1g_2(x)]`$ $`=`$ $`\text{dist}[x,g_2(x)].`$ (20)
We will refer to these pairs as $`xg(x)`$-pairs.
Both the $`xy`$–pairs and the $`xg(x)`$-pairs are type I pairs.
To collect all these pairs together and enhance the topological signal, we define the CCP–index of a catalog containing $`N`$ objects as follows.
1. We compute all the 3D–distances $`\text{dist}[x,y]`$ for all points within the catalog’s limit.
2. We order all these distances in a list $`d_i|_{1iP}`$, where $`PN(N1)/2`$ is the number of pairs, such that $`d_{i+1}d_i`$.
3. We create a new list of increments defined by
$$i[1\mathrm{}P1],\mathrm{\Delta }_id_{i+1}d_i$$
(21)
(keeping all the equal distances, if any, in the list).
4. We then define the CCP–index $``$ as
$$\frac{𝒩}{P1},$$
(22)
where $`𝒩\text{card}(\{i,\mathrm{\Delta }_i=0\})`$, so that $`01`$.
With such a procedure, all type I pairs will contribute to $`𝒩`$. For instance, if a given distance appears 4 times in the list $`d_i|_{1iP}`$, it will contribute to 3 counts in $`𝒩`$. Compared to the old crystallographic method, all the correlated pairs are gathered into a single spike, instead of being smoothed out into the noise of the histogram pair separation.
Indeed, in a more realistic situation, one has to take into account bins of finite width $`ϵ`$ and replace $`𝒩`$ by
$`𝒩_ϵ`$ $``$ $`\text{card}(\{i,\mathrm{\Delta }_i[0,ϵ[\})`$ (23)
in the computation of $``$. The effect of this “binning” is discussed below. We now focus on the “idealised” version of the procedure by studying the amplitude of the CCP–index in some multi-connected models.
For that purpose, let us assume that the catalog is obtained from an initial set of $`A`$ objects lying in the fundamental domain and that $`B`$ copies of the domain are within the catalog’s limits ($`B=0`$ if the whole observable universe up to the catalog’s limit is included inside a fundamental domain). The total number of images is $`N=A(B+1)`$. Indeed $`B`$ is usually not an integer but we assume it is, in order to estimate the amplitude of $``$ and compare it with the result in a simply-connected model. In , it has been shown that
$$𝒩_{\mathrm{min}}=A\left[(A1)\frac{B}{2}+A\nu _1(\mathrm{\Sigma },B)\right],$$
(24)
where $`\nu _1(\mathrm{\Sigma },B)`$ is a function characterising the manifold $`\mathrm{\Sigma }`$. It is an increasing function of $`B`$ which vanishes for $`B=0`$.
Indeed, if the holonomy group $`\mathrm{\Gamma }`$ contains Clifford translations allowing for type II pairs, or if there are “fake” pairs (i.e. such that $`\text{dist}[x,y]=\text{dist}[u,v]`$), $`𝒩_{\mathrm{min}}`$ computed from (24) will give a lower bound for the true $`𝒩`$.
The normalised CCP–index (22) follows straightforwardly. $``$ is a good index for extracting the topological signal since
1. when $`B=0`$ (i.e. when the fundamental domain is greater than the catalog’s spatial scale), $`=0`$,
2. when the number of sources in the fundamental domain becomes large, it behaves as
$$\frac{B+2\nu _1(\mathrm{\Sigma },B)}{(B+1)^2}\text{as}A\mathrm{}.$$
(25)
As shown in , one can compute analytically the CCP–index in various examples and show that it is a fair indicator of the existence of at least one compact spatial dimension.
### 5.2 Simulations and statistical relevance
Now that the CCP–index can be computed for any multi-connected space, we can compare it with its value in a simply-connected Friedmann-Lemaitre model containing the same number objects (i.e. $`A(B+1)`$). $``$ is given as a function of $`\nu _1`$, $`A`$ (the number of sources in the fundamental domain) and $`B`$ (the number of copies of the domain within the observable universe) by:
$$=\frac{A[(2\nu _1B)AB]}{(B+1)^2A^2(B+1)A2}.$$
(26)
Indeed, when working with a real catalog, we do not know the radial distance of cosmic objects but only their redshift. As seen before, the determination of the radial distance requires the knowledge of the cosmological parameters $`\mathrm{\Omega }_0`$ and $`\mathrm{\Lambda }`$ and can be obtained analytically only when $`\mathrm{\Lambda }=0`$.
Thus, if the universe is multi-connected on sub–horizon scale, the plot of $``$ in terms of $`\mathrm{\Omega }_0`$ and $`\mathrm{\Omega }_\mathrm{\Lambda }`$ will exhibit a spike only when the cosmological parameters have the right value (as shown on plot 7 of ). If the cosmological parameters are not exactly known, the distance determination (15) will be wrong and the topological signature will be destroyed (see figure 6 in ).
Two consequences follow:
1. One should span the parameters space $`(\mathrm{\Omega }_0,\mathrm{\Omega }_\mathrm{\Lambda })`$ in order to detect the topological signal, plotting $`(\mathrm{\Omega }_0,\mathrm{\Omega }_\mathrm{\Lambda })`$.
2. If there is any topological signal, the position of the spike gives the values of the cosmological parameters on the scale of the catalog’s limit (see figure 11).
We proceed as follows. We first generate a catalog by choosing the number of objects in the fundamental domain ($`A=30`$), the topology (Weeks manifold) and the cosmological parameters (e.g. $`\mathrm{\Omega }_0=0.2`$, $`\mathrm{\Omega }_\mathrm{\Lambda }=0.1`$), we then use a second code to apply the test and we draw $``$ in terms of the two cosmological parameters. As shown in , the method works pretty well in the sense that there is a strong spike which signals a non trivial topology and which determines the cosmological parameters. But we also see that a slight deviation in the evaluation of the cosmological parameters would make the spike to disappear. This effect is now discussed.
### 5.3 Real data
When one wants to apply the CCP–method to real data, one has to face a number of problems.
First, we cannot use a zero width bin, one of the reasons being that the sources are not comoving. In , we estimate the precision needed on the cosmological parameters when working with a bin width $`ϵ`$ such that
$$\left|\frac{\delta \mathrm{\Omega }_0}{\mathrm{\Omega }_0}\right|ϵ.$$
(27)
Indeed, using a catalog with a smaller depth $`z_{\mathrm{max}}=3`$ will allow us to use a smaller resolution for the cosmological parameters. One has thus to find a compromise between depth and resolution as discussed in . On figure 11, we give an example with a bin width of $`ϵ=10^6`$ which produces a background noise.
Now, we apply our test to the quasars catalog used ahead (see figure 3). No topological signature was found . Does it mean that there is no topological effect on scales smaller than $`z_{\mathrm{max}}3`$ ? Not necessarily, since we applied the test with presicions $`ϵ=10^7,10^6,10^5`$ and were unable to span the full cosmological parameters’ space with the required accuracy. The computational time is one of the main limitations of our technique (this point is discussed extensively in )
Now, another limitation comes from the peculiar velocities of the sources. In , we discussed in details the effect of peculiar velocities both on $`xy`$–pairs and on $`xg(x)`$–pairs, we and showed that even if the amplitude of the topological signal is reduced, it is not destroyed.
## 6 CONCLUSIONS
We have reviewed the main allowed topological structures for relativistic universe models and we discussed the various methods aimed to detect this topology. Focusing on 3D-methods, we described the crystallographic method and we applied it to a quasars catalog to obtain a new constraint for Euclidean manifolds, $`L_03000h^1\text{Mpc}`$. Then we discussed the failure of this method for detecting the topology of compact hyperbolic manifolds. We thus generalised the method by introducing a new technique based on the construction of a CCP–index. The main difference with the cosmic crystallography method is the collecting process of all correlated pairs of the catalog, which enhances the signal associated to the existence of a non trivial topology. We gave examples of computation of such an index in order to show its statistical relevance. We finally discussed the implementation difficulties of this new method when working with observational data.
## References
|
warning/0005/quant-ph0005033.html
|
ar5iv
|
text
|
# 1 Introduction
## 1 Introduction
In a recent paper the symplectic manifold
$$𝒮=\{(\phi mod2\pi ,p>\mathrm{\hspace{0.17em}0})\}$$
(1)
(associated with the symplectic form $`d\phi dp`$) was quantized group theoretically by means of the group $`SO^{}(1,2)`$ (identity component of the proper Lorentz group in 2+1 space-time dimensions). The purpose was the quantization of Schwarzschild black holes . In the meantime I realized that that quantization also sheds new light on the old unsolved problem how to represent phase and modulus as self-adjoint operators in a Hilbert space associated with the corresponding physical system (see the Reviews ).
The crucial point is that the manifold (1) has the nontrivial topology $`S^1\times ^+,^+`$: real numbers $`>0`$. Such a manifold cannot be quantized in the usual naive way used for a phase space with the trivial topology $`^2`$ by converting the classical canonical pair $`(q,p)`$ of phase space variables into operators and their Poisson bracket into a commutator. Here the group theoretical quantization scheme developed in the eighties of the last century as a generalization of the conventional one (see the reviews ) helps: The group $`SO^{}(1,2)`$ acts symplectically, transitively, effectively and (globally) Hamilton-like on the manifold (1) and, therefore, its irreducible representations (or those of its covering groups) can provide the basic self-adjoint quantum observables and their Hilbert space of states (see Ref. for more details): In the course of the group theoretical quantization one finds that the three basic classical observables
$$P_3=p>0,P_1=p\mathrm{cos}\phi ,P_2=p\mathrm{sin}\phi $$
(2)
correspond to the three self-adjoint Lie algebra generators $`K_3,K_1,`$ and $`K_2`$ of a positive discrete series irreducible unitary representation of the group $`SO^{}(1,2)`$ or one of its infinitely many covering groups, e.g. the double covering $`SU(1,1)`$ which is isomorphic to the groups $`SL(2,)`$ and $`Sp(1,)`$. The generators $`K_i`$ obey the commutation relations
$$[K_3,K_1]=iK_2,[K_3,K_2]=iK_1,[K_1,K_2]=iK_3.$$
(3)
Here $`K_3`$ is the generator of the compact sub-group $`SO(2)`$.
The corresponding Poisson brackets for the classical observables (2),
$$\{P_3,P_1\}=P_2,\{P_3,P_2\}=P_1,\{P_1,P_2\}=P_3,$$
(4)
form the real Lie algebra of $`SO^{}(1,2)(P_3iK_3,P_1iK_1,P_2iK_2`$), where $`\{f_1,f_2\}_\phi f_1_pf_2_pf_1_\phi f_2`$ for any two smooth functions $`f_i(\phi ,p),i=1,2`$. As any function $`f(\phi ,p)`$ periodic in $`\phi `$ with period $`2\pi `$ can – under certain conditions – be expanded in a Fourier series and as $`\mathrm{cos}(n\phi )`$ and $`\mathrm{sin}(n\phi )`$ can be expressed by polynomials of order $`n`$ in $`\mathrm{cos}\phi `$ and $`\mathrm{sin}\phi `$, the observables (2) are indeed the basic ones on the phase manifold (1).
Their relationship to corresponding observables in interferences (optical or otherwise) is the following: Consider the sum
$$A=a_1e^{i\phi _1}+a_2e^{i\phi _2}$$
(5)
of two complex numbers, where the phases $`\phi _i`$ can be chosen such that $`a_i>0,i=1,2`$. The quantities $`a_i`$ and $`\phi _i`$ may be functions of other parameters, e.g. space or/and time variables etc. The absolute square of $`A`$ has the form
$$|A|^2(a_1,a_2,\phi =\phi _2\phi _1)=(a_1)^2+(a_2)^2+2a_1a_2\mathrm{cos}\phi .$$
(6)
The corresponding ”quadrature” quantity is
$$|A|^2(a_1,a_2,\phi +\frac{\pi }{2})=(a_1)^2+(a_2)^22a_1a_2\mathrm{sin}\phi ,$$
(7)
obtained by an appropriate $`\pi /2`$-phase shift of either $`\phi _1`$ or $`\phi _2`$.
Knowing the quantities $`p=a_1a_2>0,a_1a_2\mathrm{cos}\phi `$ and $`a_1a_2\mathrm{sin}\phi `$ allows for a complete description of the classical interference pattern against the uniform intensity background $`(a_1)^2+(a_2)^2`$. Thus, the basic observables of an interference pattern generate the Lie algebra $`SO^{}(1,2)`$!
It is essential to realize that a group theoretical quantization does $`not`$ assume that the generators of the basic Lie algebra themselves may be expressed by some conventional canonical variables like in the case of angular momentum. This may be the case locally in special examples, but in general, like in the case of the manifold (1), it will not be possible globally. For more details see the discussion below and the Refs. .
In order to calculate expectation values and fluctuations we have to know the actions of the operators $`K_i,i=1,2,3`$ on the Hilbert spaces associated with the positive discrete series of the irreducible unitary representations of $`SO^{}(1,2)`$ (or its covering groups). In the following I rely heavily on Ref. where more (mathematical) details and Refs. to the corresponding literature can be found.
As the eigenfunctions of $`K_3`$ – the generator of the compact subgroup – form a complete basis of the associated Hilbert spaces, it is convenient to use them as a starting point. The operators $`K_+=K_1+iK_2,K_{}=K_1iK_2`$ act as ladder operators. The positive discrete series is characterized by the property that there exists a state $`|k,0`$ for which $`K_{}|k,0=0.`$ The number $`k>0`$ characterizes the representation: For a general normalized eigenstate $`|k,n`$ of $`K_3`$ we have
$`K_3|k,n`$ $`=`$ $`(k+n)|k,n,n=0,1,\mathrm{},`$ (8)
$`K_+|k,n`$ $`=`$ $`\omega _n[(2k+n)(n+1)]^{1/2}|k,n+1,|\omega _n|=1,`$ (9)
$`K_{}|k,n`$ $`=`$ $`{\displaystyle \frac{1}{\omega _{n1}}}[(2k+n1)n]^{1/2}|k,n1.`$ (10)
In irreducible unitary representations the operator $`K_{}`$ is the adjoint operator of $`K_+:(f_1,K_+f_2)=(K_{}f_1,f_2)`$. The phases $`\omega _n`$ serve to guarantee this property. Their choice depends on the concrete realization of the representations. In the examples discussed in Ref. they have the values 1 or $`i`$. In the following we assume $`\omega _n`$ to be independent of $`n:\omega _n=\omega `$.
The Casimir operator $`Q=K_1^2+K_2^2K_3^2`$ has the eigenvalues $`q=k(1k)`$. The allowed values of $`k`$ depend on the group: For $`SO^{}(1,2)`$ itself one has $`k=1,2,\mathrm{}`$ and for the double covering $`SU(1,1)`$ $`k=1/2,1,3/2,\mathrm{}`$. The appropriate choice will depend on the physics to be described.
The relation (9) implies
$$|k,n=\omega ^n\left[\frac{\mathrm{\Gamma }(2k)}{n!\mathrm{\Gamma }(2k+n)}\right]^{1/2}(K_+)^n|k,0.$$
(11)
The expectation values of the self-adjoint operators $`K_1=(K_++K_{})/2`$ and $`K_2=(K_+K_{})/2i`$ (which correspond to the classical observables $`p\mathrm{cos}\phi `$ and $`p\mathrm{sin}\phi `$) with respect to the eigenstates $`|k,n`$ and the associated fluctuations may be calculated with the help of the relations (8)-(10) :
$$k,n|K_i|k,n=0,i=1,2.$$
(12)
The corresponding fluctuations are
$$(\mathrm{\Delta }K_i)_{k,n}^2=k,n|K_i^2|k,n=\frac{n}{2}(2k+n)+\frac{k}{2},i=1,2.$$
(13)
For very large $`n`$ the correspondence principle, $`(p\mathrm{cos}\phi )^2+(p\mathrm{sin}\phi )^2=p^2`$, is fulfilled:
$$k,n|K_1^2|k,n+k,n|K_2^2|k,nn^2k,n|K_3^2|k,n$$
(14)
This follows already from the Casimir operator which for an irreducible representation can be rewritten as $`(K_1)^2+(K_2)^2=(K_3)^2+k(1k).`$ For $`k=1`$ we even have $`(K_1)^2+(K_2)^2=(K_3)^2`$.
Next I define the self-adjoint operators $`\widehat{\mathrm{cos}\phi }`$ and $`\widehat{\mathrm{sin}\phi }`$ as follows:
$$\widehat{\mathrm{cos}\phi }=\frac{1}{2}(K_3^1K_1+K_1K_3^1),\widehat{\mathrm{sin}\phi }=\frac{1}{2}(K_3^1K_2+K_2K_3^1).$$
(15)
Notice that $`K_3^1`$ is well-defined because $`K_3`$ is a positive definite operator for the positive discrete series. One has $`K_3^1|k,n=|k,n/(n+k)`$. Using again the relations (8)-(10) we get
$`\widehat{\mathrm{cos}\phi }|k,n`$ $`=`$ $`{\displaystyle \frac{\omega }{4}}f_{n+1}^{(k)}|k,n+1+{\displaystyle \frac{1}{4\omega }}f_n^{(k)}|k,n1,`$ (16)
$`\widehat{\mathrm{sin}\phi }|k,n`$ $`=`$ $`{\displaystyle \frac{\omega }{4i}}f_{n+1}^{(k)}|k,n+1+{\displaystyle \frac{1}{4i\omega }}f_n^{(k)}|k,n1,`$ (17)
$`f_n^{(k)}`$ $`=`$ $`[n(2k+n1)]^{1/2}\left({\displaystyle \frac{1}{k+n}}+{\displaystyle \frac{1}{k+n1}}\right),`$ (18)
and therefore
$$k,n|\widehat{\mathrm{cos}\phi }|k,n=0,k,n|\widehat{\mathrm{sin}\phi }|k,n=0$$
(19)
and
$$k,n|(\widehat{\mathrm{cos}\phi })^2|k,n=k,n|(\widehat{\mathrm{sin}\phi })^2|k,n=\frac{1}{16}[(f_{n+1}^{(k)})^2+(f_n^{(k)})^2]$$
(20)
$$k,n|[\widehat{\mathrm{sin}\phi },\widehat{\mathrm{cos}\phi }]|k,n=\frac{1}{8i}[(f_{n+1}^{(k)})^2(f_n^{(k)})^2].$$
(21)
For the ground state $`|k,n=0`$ one gets
$$k,0|(\widehat{\mathrm{cos}\phi })^2|k,0=k,0|(\widehat{\mathrm{sin}\phi })^2|k,0=\frac{(2k+1)^2}{8k(k+1)^2}.$$
(22)
It follows that an upper bound $`k,0|(\widehat{\mathrm{cos}\phi })^2|k,01`$ implies for $`k`$ the lower bound $`kk_1[(0.5+0.5\sqrt{23/27})^{1/3}+(0.50.5\sqrt{23/27})^{1/3}1]/2=0.162\mathrm{}`$. A slightly higher lower bound for allowed values of $`k`$ will be discussed below.
For very large $`n`$ we have the (correct) correspondence principle limits
$$k,n|(\widehat{\mathrm{cos}\phi })^2|k,n=k,n|(\widehat{\mathrm{sin}\phi })^2|k,n\frac{1}{2}(1+O(n^2))\text{for}n\mathrm{},$$
(23)
$$k,n|[\widehat{\mathrm{sin}\phi },\widehat{\mathrm{cos}\phi }]|k,nO(n^2)\text{for}n\mathrm{}.$$
(24)
I next turn to some properties of coherent states. Contrary to the conventional coherent states (i.e. the eigenstates of the Bose annihilation operator associated with the harmonic oscillator, see e.g. the reviews and the modern exposition ) there are several inequivalent ways to define coherent states related to the group $`SO^{}(1,2)`$ or $`SU(1,1)`$ (see also the Refs. ). For our purposes the definition
$$K_{}|z=z|z,z,$$
(25)
seems to be an interesting one: Using the property (11) we get
$$k,n|z=\frac{1}{\overline{\omega }^n}\left[\frac{\mathrm{\Gamma }(2k)}{n!\mathrm{\Gamma }(2k+n)}\right]^{1/2}z^nk,0|z,$$
(26)
($`\overline{\omega }:\text{compl. conj. of}\omega `$) so that
$`z|z`$ $`=`$ $`{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}=z|k,nk,n|z=\mathrm{\Gamma }(2k)|k,0|z|^2{\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{|z|^{2n}}{n!\mathrm{\Gamma }(2k+n)}}`$ (27)
$`=`$ $`\mathrm{\Gamma }(2k)|k,0|z|^2|z|^{12k}I_{2k1}(2|z|),,`$
where
$$I_\nu (x)=\left(\frac{x}{2}\right)^\nu \underset{n=0}{\overset{\mathrm{}}{}}\frac{1}{n!\mathrm{\Gamma }(\nu +n+1)}\left(\frac{x}{2}\right)^{2n}$$
(28)
is the usual modified Bessel function of the first kind which has the asymptotic expansion
$$I_\nu (x)\frac{e^x}{\sqrt{2\pi x}}\left(1\frac{4\nu ^21}{8x}+O(x^2)\right)\text{for}x+\mathrm{}.$$
(29)
If $`z|z=1`$ we have
$$|k,0|z|^2|C_z|^2=\frac{|z|^{2k1}}{\mathrm{\Gamma }(2k)I_{2k1}(2|z|)}.$$
(30)
Choosing the phase of $`C_z`$ appropriately and absorbing the phase $`\omega `$ into a redefinition of $`z`$ we finally get the expansion
$$|z=\frac{|z|^{k1/2}}{\sqrt{I_{2k1}(2|z|)}}\underset{n=0}{\overset{\mathrm{}}{}}\frac{z^n}{\sqrt{n!\mathrm{\Gamma }(2k+n)}}|k,n.$$
(31)
Notice that $`|z=0=|k,n=0`$.
Two different coherent states are not orthogonal. They are complete, however, because with $`z=\rho e^{i\alpha }`$ we have the completeness relation
$`{\displaystyle \frac{2}{\pi }}{\displaystyle _0^{\mathrm{}}}𝑑\rho \rho K_{2k1}(2\rho )I_{2k1}(2\rho ){\displaystyle _0^{2\pi }}𝑑\alpha k,n|z=\rho e^{i\alpha }z=\rho e^{i\alpha }|k,n`$
$`=k,n|k,n=1,`$ (32)
where $`K_\nu (x)`$ is the modified Bessel function of the third kind .
The following expectation values are associated with the states $`|z`$:
$`K_3_z`$ $``$ $`z|K_3|z=k+|z|{\displaystyle \frac{I_{2k}(2|z|)}{I_{2k1}(2|z|)}},`$ (33)
$`K_3^2_z`$ $`=`$ $`k^2+|z|^2+|z|{\displaystyle \frac{I_{2k}(2|z|)}{I_{2k1}(2|z|)}},`$ (34)
so that
$$(\mathrm{\Delta }K_3)_z^2=|z|^2\left(1\frac{I_{2k}^2(2|z|)}{I_{2k1}^2(2|z|)}\right)+(12k)|z|\frac{I_{2k}(2|z|)}{I_{2k1}(2|z|)}.$$
(35)
For very large $`|z|`$ we have the leading terms
$$K_3_z|z|,(\mathrm{\Delta }K_3)_z^2\frac{1}{2}|z|\text{for}|z|+\mathrm{}.$$
(36)
This, together with the probability
$$|k,n|z|^2=\frac{|z|^{2(n+k)1}}{n!\mathrm{\Gamma }(2k+n)I_{2k1}(2|z|)}2\sqrt{\pi }\frac{|z|^{2(n+k)1/2}}{n!\mathrm{\Gamma }(2k+n)}e^{2|z|}$$
(37)
shows that the corresponding distribution for large $`|z|`$ is not Poisson-like!
In addition we have the following expectation values
$`K_1_z={\displaystyle \frac{1}{2}}(\overline{z}+z)=\rho \mathrm{cos}\alpha `$ , $`K_2_z={\displaystyle \frac{1}{2i}}(\overline{z}z)=\rho \mathrm{sin}\alpha ,`$ (38)
$`(\mathrm{\Delta }K_1)_z^2=(\mathrm{\Delta }K_2)_z^2`$ $`=`$ $`{\displaystyle \frac{1}{2}}K_3_z.`$ (39)
Comparing Eqs. (2) and (38) we see the close relationship between the expectation values $`K_i_z,i=1,2`$ and their classical counterparts. This supports the above choice (25) as coherent states. Further support comes from their property to realize the minimal uncertainty relation: ¿From the third commutator in Eqs. (3) we get the general inequality
$$(\mathrm{\Delta }K_1)^2(\mathrm{\Delta }K_2)^2\frac{1}{4}|K_3|^2.$$
(40)
The relations (39) show that the coherent states (31) realize the minimum of the uncertainty relation (40). One can, of course, extend the discussion to associated squeezed states .
For the operators $`\widehat{\mathrm{cos}\phi }`$ and $`\widehat{\mathrm{sin}\phi }`$ we have (from Eqs. (16), (17) and (31)) the expectation values
$`\widehat{\mathrm{cos}\phi }_z`$ $`=`$ $`{\displaystyle \frac{\overline{z}+z}{4|z|}}{\displaystyle \frac{g^{(k)}(|z|)}{I_{2k1}(2|z|)}}={\displaystyle \frac{1}{2}}\mathrm{cos}\alpha {\displaystyle \frac{g^{(k)}(|z|)}{I_{2k1}(2|z|)}},`$ (41)
$`\widehat{\mathrm{sin}\phi }_z`$ $`=`$ $`{\displaystyle \frac{\overline{z}+z}{4i|z|}}{\displaystyle \frac{g^{(k)}(|z|)}{I_{2k1}(2|z|)}}={\displaystyle \frac{1}{2}}\mathrm{sin}\alpha {\displaystyle \frac{g^{(k)}(|z|)}{I_{2k1}(2|z|)}},`$ (43)
$`g^{(k)}(|z|)={\displaystyle \underset{n=0}{\overset{\mathrm{}}{}}}{\displaystyle \frac{|z|^{2(n+k)}}{n!\mathrm{\Gamma }(2k+n)}}\left({\displaystyle \frac{1}{n+k}}+{\displaystyle \frac{1}{n+k+1}}\right).`$
One has
$$g^{(k)}(|z|)=_0^{2|z|}𝑑uI_{2k1}(u)+\frac{1}{4|z|^2}_0^{2|z|}𝑑uu^2I_{2k1}(u).$$
(44)
The right hand side may be expressed by a combination of modified Bessel and Lommel functions . For large $`|z|`$ one obtains
$$\frac{g^{(k)}(|z|)}{I_{2k1}(2|z|)}2\left(1\frac{1}{4|z|}+O(|z|^2)\right)\text{for large}|z|$$
(45)
which again gives the expected correspondence principle limits for $`\widehat{\mathrm{cos}\phi }_z`$ and $`\widehat{\mathrm{sin}\phi }_z`$.
The expectation values $`\widehat{\mathrm{cos}\phi }^2_z`$ etc. may be calculated by observing that $`\widehat{\mathrm{cos}\phi }^2_z=_nz|\widehat{\mathrm{cos}\phi }|k,nk,n|\widehat{\mathrm{cos}\phi }|z=_n|k,n|\widehat{\mathrm{cos}\phi }|z|^2`$ etc.
The operators $`\widehat{\mathrm{cos}\phi }`$ and $`\widehat{\mathrm{sin}\phi }`$ are bounded self-adjoint operators (see Eqs. (20)) with a continuous spectrum within the interval $`[1,+1]`$ for $`k0.32`$. The last assertion follows from Eqs. (41) and (42) together with a numerical analysis of the ratio $`g^{(k)}(|z|)/I_{2k1}(2|z|)`$ which shows that ratio to be $`<2`$ for all finite $`|z|`$ if $`k0.32`$. That is not so e.g. for $`k=0.25`$. This and the relation (45) imply that at least for $`k0.32`$ we have $`sup_z|\widehat{\mathrm{cos}\phi }_z|=sup_z|\widehat{\mathrm{sin}\phi }_z|=1`$ from which the support of the spectrum follows . Thus, for the groups $`SO^{}(1,2)`$ and $`SU(1,1)`$ which have $`k=1`$ and $`k=1/2`$ respectively as their lowest $`k`$-values we are on the safe side.
The ansatz $`|\mu =_{n=0}^{\mathrm{}}a_n|k,n`$ for the improper “eigenfunctions” of $`\widehat{\mathrm{cos}\phi }`$ with “eigenvalues” $`\mu ,\widehat{\mathrm{cos}\phi }|\mu =\mu |\mu ,`$ leads to the recursion formula $`a_{n+1}=(4\mu a_nf_n^{(k)}a_{n1})/f_{n+1}^{(k)},f_0^{(k)}=0,`$ which allows to express the $`a_n`$ by $`\mu ,a_0`$ and the $`f_n^{(k)}`$.
Up to now I have not specified the concrete form of the Hilbert space, the operators $`K_i,i=1,2,3`$ and the eigenfunctions $`|k,n`$. Several interesting examples may be found in Ref. .
I finally come – very preliminary – to some subtle points of the physical interpretation of the results. Let us take the example represented by the Eqs. (6) and (7): Here the eigenvalues of the operator $`K_3`$ correspond to the square root $`\sqrt{I_1I_2}`$ of the product of the intensities $`I_1`$ and $`I_2`$ of the two interfering classical waves. In an interference experiment with photons we therefore expect the natural number $`n`$ characterizing the eigenvalues $`n+k`$ of $`K_3`$ to count the number of photons registered by an appropriate device. The number $`k`$ characterizes the ground state and it is not so obvious which value it will take. The group theoretical quantization requires its own interpretation.
On the other hand we are very much used to the conventional quantization in which the intensities $`I_1`$ and $`I_2`$ become proportional to associated number operators $`\widehat{N}_1`$ and $`\widehat{N}_2`$ built up in the simplest case from two pairs of bosonic creation and annihilation operators $`b_i^+`$ and $`b_i,i=1,2`$ which may also be used to construct the associated Hilbert space: $`\widehat{N}_i^{(b)}=b_i^+b_i`$ and $`b_i^+|n_i_i=\sqrt{n_i+1}|n_i+1_i,b_i|n_i_i=\sqrt{n_i}|n_i1_i`$. One can take the square root of $`\widehat{N}_i^{(b)}`$ naively or (in order to have those square roots linear in the $`b_i^+`$ and $`b_i`$) in a Dirac-type manner:
$`\sqrt{\widehat{N}_i^{(b)}}=\sigma _+b_i^++\sigma _{}b_i`$ $`=`$ $`\left(\begin{array}{cc}0& b_i^+\\ b_i& 0\end{array}\right),\sigma _\pm ={\displaystyle \frac{1}{2}}(\sigma _1\pm i\sigma _2),`$ (48)
$`\sqrt{\widehat{N}_i^{(b)}}|\sqrt{n_i}_i`$ $`=`$ $`\sqrt{n_i}|\sqrt{n_i}_i,|\sqrt{n_i}_i=\left(\begin{array}{c}|n_i_i\\ |n_i1_i\end{array}\right),`$ (51)
$`\left(\begin{array}{cc}0& b_i^+\\ b_i& 0\end{array}\right)^2`$ $`=`$ $`\left(\begin{array}{cc}\widehat{N}_i^{(b)}& 0\\ 0& \widehat{N}_i^{(b)}+1\end{array}\right),i=1,2,`$ (56)
where $`\sigma _i,i=1,2`$ are the usual Pauli matrices. If we now consider the operator $`\sqrt{\widehat{N}_1^{(b)}}\sqrt{\widehat{N}_2^{(b)}}`$ acting on $`|\sqrt{n_1}_1|\sqrt{n_2}_2`$ to be the quantized counterpart of $`\sqrt{I_1I_2}`$ then we obviously can have the eigenvalues $`n_0`$ for all $`n_1,n_2`$ only if $`n_1=n_2`$, that is for an ideal 50% beam splitter. Another possibility is that $`n_2n_1`$ but $`\sqrt{n_1n_2}=n=0,1,2,\mathrm{}`$.
Now there is a close relationship between another pair $`a_i^+,a_i,i=1,2,`$ of bosonic creation and annihilation operators and the irreducible unitary representations of $`SU(1,1)`$ (see Ref. and the literature quoted there): The operators
$$K_3^{(a)}=\frac{1}{2}(a_1^+a_1+a_2^+a_2+1),K_+^{(a)}=a_1^+a_2^+,K_{}^{(a)}=a_1a_2$$
(57)
obey the associated commutation relations (3) and the tensor product $`_1^{osc}_2^{osc}`$ of the two harmonic oscillator Hilbert spaces contains all the irreducible unitary representations of the group $`SU(1,1)`$ in the following way: Let $`|m_i_i,m_i_0,i=1,2,`$ be the eigenstates of the number operators $`\widehat{N}_i^{(a)}`$ generated by $`a_i^+`$ from the oscillator ground states. Then each of those two subspaces of $`_1^{osc}_2^{osc}=\{|m_1_1|m_2_2\}`$ for which $`|m_1m_2|0`$ is fixed contains an irreducible representation with $`k=1/2+|m_1m_2|/2=1,3/2,2,\mathrm{}`$ and for which the number $`n`$ in the eigenvalue $`n+k`$ is given by $`\mathrm{min}\{m_1,m_2\}`$.
For the “diagonal” case $`m_2=m_1`$ one gets the unitary representation with $`k=1/2`$. In general the two systems of oscillators $`a_i^+,a_i`$ and $`b_i^+,b_i`$ will represent different physical systems, but perhaps there can be a correspondence between the diagonal states $`|\sqrt{n_1}_1|\sqrt{n_2=n_1}_2`$ and $`|m_1_1|m_2=m_1_2`$.
This suggests that the representation with $`k=1/2`$ may play a special role for the applications. Such a suggestion is supported intuitively by the following argument: For $`k=1/2`$ the operator $`K_3`$ has the same spectrum $`\{n+1/2,n_0\}`$ as the harmonic oscillator. Since $`n`$ counts the number of quanta (e.g. photons) the additional term $`1/2`$ will represent the usual ground state contribution to the energy! Experiments will have to decide which value of $`k`$ one has to choose.
Finally I would like to stress again that a group theoretical quantization like the one above does $`not`$ suppose that there is a “deeper” conventional canonical structure in terms of the usual $`q`$ and $`p`$. It claims to provide an appropriate quantization for topologically nontrivial symplectic manifolds like (1) by itself. Quantum optics (or other quantum interference phenomena) may very well be able to test such claims experimentally. In addition it may test the identification (15) as an operator version of $`\mathrm{cos}\phi `$ and $`\mathrm{sin}\phi `$. That definition is a new ansatz within \- not a basic ingredient of \- the group theoretical quantization scheme.
In the beginning of this work I enjoyed the hospitality of the DESY Theory group. I thank W. Buchmüller and P. Zerwas for their kind invitation. I thank H. Walther for very kindly letting me use the library of the MPI for Quantum Optics at the very early stages of this work during a brief visit of mine to the University of Munich. I am indebted to N. Düchting for an numerical analysis of the ratio $`g^{(k)}(|z|)/I_{2k1}(2|z|)`$ and to him and to M. Bojowald for a critical reading of the manuscript. Most thanks go to my wife Dorothea.
|
warning/0005/astro-ph0005111.html
|
ar5iv
|
text
|
# A New Perspective on Cosmic Coincidence Problems[1]
## Abstract
Cosmological data suggest that we live in an interesting period in the history of the universe when $`\rho _\mathrm{\Lambda }\rho _M\rho _R`$. The occurence of any epoch with such a “triple coincidence” is puzzling, while the question of why we happen to live during this special epoch is the “Why now?” problem. We introduce a framework which makes the triple coincidence inevitable; furthermore, the “Why now?” problem is transformed and greatly ameliorated. The framework assumes that the only relevant mass scales are the electroweak scale, $`M_{EW}`$, and the Planck scale, $`M_{Pl}`$, and requires $`\rho _\mathrm{\Lambda }^{1/4}M_{EW}^2/M_{Pl}`$ parametrically. Assuming that the true vacuum energy vanishes, we present a simple model where a false vacuum energy yields a cosmological constant of this form.
preprint: LBNL-42967, UCB-PTH-99/07
1. Recent observations using high-redshift Type-IA supernovae indicate a small but non-zero cosmological constant ($`\mathrm{\Lambda }`$) is present in the universe today. Evidence for a finite $`\mathrm{\Lambda }`$ comes also from the cosmic age, large scale structure, and probably most convincingly from the combination of the cosmic microwave background anisotropy and cluster dynamics (see for recent reviews). The data suggest a flat Universe, $`\mathrm{\Omega }_{\mathrm{total}}=1`$, with about 70% of the energy density coming from $`\mathrm{\Lambda }`$ and 30% in the form of non-relativistic matter.
The tiny energy density of the cosmological constant $`\rho _\mathrm{\Lambda }(2`$meV$`)^4`$ poses the most severe naturalness problem in theoretical physics: why is $`\rho _\mathrm{\Lambda }10^{120}M_{Pl}^4`$? Moreover, the near equivalence between the two components in the energy density raises a serious question: Why should we observe them to be so nearly equivalent now? While a cosmological constant is by definition time-independent, the matter energy density is diluted as $`1/R^3`$ as the Universe expands. Thus, despite evolution of $`R`$ over many orders of magnitude, we appear to live in an era during which the two energy densities are roughly the same (see Fig. 1). This is the “Why now?” problem. If we believe the data, there appears to be no choice but to believe that we live in a special time in the history of the universe. This possibility is thought by many to be sufficiently distasteful to warrant disbelieving the data.
One possible partial remedy to this problem would be to assume that the “cosmological constant” is not a constant at all, but has always been comparable to the rest of the energy densities. Such a possibility is realized in some models of quintessence which possess “scaling” behavior (see for a recent attempt along this line). Though such an idea may make the closeness of the effective $`\mathrm{\Lambda }`$ to the total energy density natural, it does not explain the coincidence that the quintessence field becomes settled with a finite energy density comparable to the matter energy density just now.
Returning to Fig. 1, we notice two remarkable features. The first is that there is an era in the history of the universe where all three forms of energy, in matter, radiation and $`\mathrm{\Lambda }`$, become comparable within a few orders of magnitude. The second is that the universe happens to be in this interesting era now. These observations lead to two different questions. The first is the “cosmic triple coincidence” problem: why should three forms of energy density very nearly cross at any point in the evolution of the universe? The second question is a generalization of the “Why now?” problem: why is the universe in this era of triple coincidence now? The first problem is epoch independent: in principle, at any point in the history of the universe the three forms of energy could be measured and extrapolated in time, and the triple coincidence could be inferred. The existence of a triple coincidence is reminiscent of the unification of gauge coupling constants in Grand Unified Theories, and suggests an underlying theory in which the coincidence has a simple explanation.
In this letter, we propose a solution to the triple coinicidence problem. We begin with one of the central observations of particle physics, the special significance played by two mass scales: the reduced Planck scale, $`M_{Pl}10^{18}`$ GeV, and the electroweak scale, $`M_{EW}10^3`$ GeV. Given these two scales, the existence of the triple coincidence of matter, radiation and the cosmological constant can follow in a completely natural way. In particular, once one assumes that $`\mathrm{\Lambda }`$ is order $`M_{EW}^2/M_{Pl}`$ (and we will give explicit models in which this arises), this triple coincidence is inevitable, split only by $`𝒪(1)`$ coefficients such as $`\alpha `$ and $`\pi `$. As we will see, this solution to the cosmic coincidence problem also allows for a new understanding of the “Why now?” problem.
Another apparent “coincidence” in cosmology arises when one examines the components that comprise the matter density. Current data favors cold dark matter (CDM) as the dominant component of $`\mathrm{\Omega }_M`$, but baryons ($`\mathrm{\Omega }_B0.05`$) and neutrinos ($`\mathrm{\Omega }_\nu 0.003`$–0.15) are also appreciable. Having argued for the triple coincidence, we outline how the framework can be extended to give a five-fold coincidence: it is natural for $`\mathrm{\Omega }_B`$ and $`\mathrm{\Omega }_\nu `$ to also be $`𝒪(1)`$ at the present time.
2. We begin with the minimal standard model. The physics of electroweak symmetry breaking is well-known to suffer from a hierarchy problem: the energy scale at which the electroweak symmetry is broken, $`\mathrm{TeV}`$, is 15 orders of magnitude smaller than the Planck scale $`10^{18}`$ GeV, which we take to be the fundamental length scale of the universe. Proposed models in which this hierarchy is explained are substantially more complicated at the electroweak scale than the standard model itself, with a plethora of new particles. Among these, there are typically stable particles which can be produced as thermal relics, such as the lightest supersymmetric particle in models with hidden-sector supersymmetry breaking , the lightest messenger particle in the models of gauge-mediated supersymmetry breaking , and technibaryons in technicolor models , to name a few. It is oft-remarked upon that the relic density of a stable particle at the electroweak scale gives the correct order of magnitude for the cold dark matter component of the Universe, which is believed to dominate the matter energy density. This important result is apparently a numerical coincidence, but in our new perspective it will be guaranteed.
It is simple to estimate the relic density solely from the fundamental parameters of the theory. For a stable particle $`\chi `$ of mass $`m`$ with electroweak interactions, its annihilation cross section is given roughly by $`\sigma 1/m^2`$. When the temperature falls below the mass $`m`$, the expansion rate $`HT_f^2/M_{Pl}`$ wins over the annihilation rate $`\mathrm{\Gamma }\sigma n_\chi `$ and the abundance $`n_\chi /s`$ freezes out when $`H\mathrm{\Gamma }`$. After freeze-out, the abundance $`n_\chi /s`$ is constant, so that the matter energy density (dominated by $`\chi `$) is
$$\rho _M=m_\chi n_\chi \frac{1}{\sigma }\frac{m^3}{M_{Pl}}\left(\frac{T}{m}\right)^3\frac{M_{EW}^2T^3}{M_{Pl}},$$
(1)
where the factor $`(T/m)^3(R_f/R)^3`$ is the usual dilution due to the expansion of the universe after freeze-out. Given this result, it is easy to estimate the temperature at which matter-radiation equality occurs, simply by equating $`\rho _R=T^4`$ with $`\rho _M`$, and we find
$$T_{eq}\frac{M_{EW}^2}{M_{Pl}}.$$
(2)
Now, we would like to understand the presence of a triple coincidence, where $`\mathrm{\Lambda }`$ becomes equal to matter and radiation. If this is not to be a numerical accident, it must be that $`\mathrm{\Lambda }`$ itself is determined in terms of the fundamental mass scales of the theory as
$$\rho _\mathrm{\Lambda }\left(\frac{M_{EW}^2}{M_{Pl}}\right)^4.$$
(3)
In such a theory, the existence of a triple coincidence between radiation, matter and $`\mathrm{\Lambda }`$ is guaranteed when the universe is at the temperature $`T_{eq}M_{EW}^2/M_{Pl}`$, regardless of the numerical value of $`M_{EW}`$. For $`M_{EW}1`$ TeV, $`T_{eq}10^3`$ eV $``$ 10 K.
We will shortly present models where $`\mathrm{\Lambda }`$ is plausibly determined in terms of the fundamental scales in the correct combination. However, before discussing this, it is important to note that our triple coincidence is significantly split by $`O(1)`$ dimensionless factors. A correct version of Eq. (1) is :
$$\rho _M=\frac{0.756(n+1)x_f^{n+1}}{g^{1/2}\sigma _0M_{Pl}}s(T),$$
(4)
where $`g`$ is the effective number of degrees of freedom at the time of the freeze-out, $`s(T)=\frac{2\pi ^2}{45}g_0T^3`$ is the entropy of the Universe, and $`x_f=m/T_f20`$ is a parameter which describes the freeze-out temperature $`T_f`$. The annihilation cross section is thermally averaged as $`\sigma v=\sigma _0x_f^n`$ with $`n=0`$ ($`n=1`$) for $`S`$-wave ($`P`$-wave) annihilations. These dimensionless factors are not very important. However, the cross-section $`\sigma _0`$ is suppressed by weak coupling factors so that $`\sigma _0\pi \alpha ^2/M_{EW}^2`$, which makes $`\rho _M`$ bigger by a factor of $`1/\pi \alpha ^210^3`$. This enhancement of the matter energy density causes the temperature at the matter-radiation equality to become roughly a factor of $`10^3`$ bigger, $`T_{eq}1`$ eV instead of $`10^3`$ eV. The existence of the (approximate) triple coincidence is not a numerical accident, however, since it is parametrically guaranteed for any values of $`M_{EW}`$ and $`M_{Pl}`$.
The following picture emerges from the above considerations. Based on rough order of magnitude estimates, the triple coincidence of the matter, radiation energy densities and $`\mathrm{\Lambda }`$ can be natural consequences of electroweak physics. Indeed, the evolution of our Universe shows a near triple coincidence (Fig. 1). If the coincidence were perfect, the Universe would be rather boring: before $`T_{eq}M_{EW}^2/M_{Pl}`$ it is radiation dominated and structure cannot form; after $`T_{eq}`$ the Universe starts to inflate, leaving a completely empty Universe. However the $`O(10^3)`$ enhancement in $`\rho _M`$ causes matter-radiation equality to occur before matter-$`\mathrm{\Lambda }`$ equality. Thus, as can be seen in Fig. 1, there is a small triangle during which density fluctuations can grow, resulting in structure. If the cosmological constant is truly a constant, then this interesting period does not last long; the matter energy density soon dilutes to the level of $`\mathrm{\Lambda }`$ and this interesting period ends. On the other hand, the apparent cosmological constant could be due to an energy density in a quintessence field $`\varphi (M_{Pl})`$ of mass $`m_\varphi ^2M_{EW}^8/M_{Pl}^6`$, which has been prevented from falling by the Hubble friction until the triple coincidence era. In this case, $`\varphi `$ will fall to the minimum of its potential a few e-foldings into the coincidence era, and the universe will return to being matter-dominated.
3. We would now like to demonstrate that it is natural to expect $`\mathrm{\Lambda }`$ of order $`(M_{EW}^2/M_{Pl})^4`$. We will start with a particularly simple model which uses false vacua as the origin of the $`\mathrm{\Lambda }`$ . We do not insist that this model should be the true origin of $`\mathrm{\Lambda }`$, but we find it extremely encouraging that such a simple model serves the purpose. We assume supersymmetry, which is broken at the TeV scale by an order parameter chiral superfield $`S=\theta ^2M_{EW}^2`$, so that the electroweak symmetry breaking is indeed the direct consequence of the supersymmetry breaking. We also assume that there is a “hidden sector,” which couples to the supersymmetry breaking only by Planck-scale suppressed operators. This is reminiscent of the conventional hidden sector supersymmetry breaking models in supergravity except that we now reverse the roles of the hidden and the observable sectors. In the hidden sector, we have a supersymmetric QCD with $`SU(n_c)`$ gauge group (or any other non-abelian gauge theory) with $`n_f`$ flavors $`Q+\stackrel{~}{Q}`$, and assume that the beta function $`3n_cn_f`$ is somewhat small. Once supersymmetry is broken in the observable sector by the order parameter $`S`$, it can generate masses for the hidden sector quarks by
$$d^4\theta \frac{S^{}}{M_{Pl}}\stackrel{~}{Q}Q=d^2\theta \frac{M_{EW}^2}{M_{Pl}}\stackrel{~}{Q}Q.$$
(5)
Therefore, the masses of quarks and squarks are of the order of $`m_QM_{EW}^2/M_{Pl}`$. Similarly, the gluino $`\lambda `$ of the $`SU(n_c)`$ gauge group also acquires a mass via the operator
$$d^2\theta \frac{S}{M_{Pl}}W_\alpha W^\alpha =\frac{M_{EW}^2}{M_{Pl}}\lambda \lambda ,$$
(6)
and again the mass is of the order of $`m_\lambda /g^2M_{EW}^2/M_{Pl}`$. For the simplicity of the analysis, we take the gluino mass somewhat smaller than the others. In the absence of the gluino mass, the decoupling of the quarks generates the low-energy dynamical scale:
$$\mathrm{\Lambda }_{\mathrm{low}}^{3n_c}=\mathrm{\Lambda }_{\mathrm{high}}^{3n_cn_f}m_Q^{n_f}.$$
(7)
Note that the low-energy dynamical scale is determined essentially by $`m_Q`$ if $`3n_cn_fn_f`$. It is basically that the gauge coupling constant evolves slowly above $`m_Q`$, while it grows quickly below $`m_Q`$.
The low-energy theory, supersymmetric pure Yang–Mills, develops a gluino condensate $`\lambda \lambda `$ with superpotential
$$d^2\theta \mathrm{\Lambda }_{\mathrm{low}}^3e^{2\pi ik/n_c}.$$
(8)
Here, the integer $`k=1,\mathrm{},n_c`$ parameterizes the degenerate vacua consistent with the Witten index $`n_c`$. The discrete $`R`$-symmetry $`Z_{2n_c}`$ is broken spontaneously down to $`Z_2`$ by the gluino condensate. In the presence of the (small) gluino mass, we replace the gauge coupling constant as $`8\pi ^2/g^28\pi ^2(1+\theta ^2m_\lambda )/g^2`$ which breaks the $`R`$-symmetry explicitly to $`Z_2`$. The vacuum energy can be calculated exactly to lowest order in $`m_\lambda `$:
$$V_k=\frac{16\pi ^2}{g^2}|m_\lambda \mathrm{\Lambda }_{\mathrm{low}}^3|\left(c\mathrm{cos}\frac{2\pi k+\mathrm{\Theta }}{n_c}\right),$$
(9)
where $`\mathrm{\Theta }`$ is the $`\mathrm{\Theta }`$-parameter of $`SU(n_c)`$. We assume the ground state energy is tuned to zero by choosing the constant $`c`$ appropriately. However the system can drop into any of the states labeled by $`k`$ and in general has the vacuum energy of the order of $`m_\lambda \mathrm{\Lambda }_{\mathrm{low}}^3(M_{EW}^2/M_{Pl})^4`$. Although in this model we assumed low-energy supersymmetry breaking, a simple variation can be extended to the case of gravity-mediated supersymmetry breaking . If the $`\mathrm{\Theta }`$-parameter is dynamical, i.e., $`\mathrm{\Theta }=a/M_{Pl}`$ for an axion-like field $`a`$, then $`a`$ has the properties of a quintessence field, beginning a slow roll down its potential only recently and contributing $`\rho _a(M_{EW}^2/M_{Pl})^4`$ .
In the above model, we assumed that an unknown mechanism cancelled the vacuum energy at the global minimum of the potential. It may be that the true mechanism for cancelling the cosmological constant has the feature that only part of the vacuum energy is cancelled. Indeed, propose to eliminate the purely Standard Model loop contributions to the cosmological constant, by embedding the Standard Model fields on a 3-brane in large extra dimensions. In these models, the effective four dimensional cosmological constant is a power series expansion in powers of the Standard Model vacuum energy $`V_{SM}M_{EW}^4`$ of the form
$$\rho _\mathrm{\Lambda }=\underset{n=0}{\overset{\mathrm{}}{}}c_n\frac{V_{SM}^{n+1}}{M_{}^{4n}}$$
(10)
where $`M_{}`$ is the higher dimensional Planck scale related to the 4D Planck scale as $`M_{Pl}^2=M_{}^6/V_{SM}`$, and the coefficient $`c_0=0`$ naturally. Unfortunately, the terms with $`n=1,2`$ are still too large, but $`n=3`$ gives the perfect combination $`\rho _{\mathrm{\Lambda },4}(M_{EW}^2/M_{Pl})^4`$. These models demonstrate that an incomplete cancellation mechanism for truly solving the cosmological constant problem may yield a $`\mathrm{\Lambda }`$ appropriately determined by $`M_{EW}`$ and $`M_{Pl}`$.
4. What about neutrino and baryon energy densities? The fact that the neutrino energy density is comparable to the rest has a natural explanation if the neutrino mass arises from the seesaw mechanism. In terms of the two fundamental scales $`M_{EW}`$ and $`M_{Pl}`$, the seesaw mechanism gives $`m_\nu M_{EW}^2/M_{Pl}`$. Then we find
$$\rho _\nu m_\nu T_0^3\frac{M_{EW}^2}{M_{Pl}}T_0^3\frac{M_{EW}^8}{M_{Pl}^4},$$
(11)
again of the same order of magnitude as $`\rho _M`$, $`\rho _R`$ and $`\rho _\mathrm{\Lambda }`$. One can also find models of baryogenesis where the baryon asymmetry is generated naturally as $`n_B/s10^5M_{EW}/M_{Pl}`$ where $`10^5(2\pi /\alpha )^2`$ . Then the baryon energy density is
$$\rho _B10^5\frac{M_{EW}}{M_{Pl}}T_0^3m_p10^5\frac{m_p}{M_{EW}}\frac{M_{EW}^8}{M_{Pl}^4},$$
(12)
which is also parametrically the same combination of $`M_{EW}`$ and $`M_{Pl}`$ if one regards $`m_p/M_{EW}10^3`$ as an $`O(1)`$ coefficient.
5. Given our solution to the triple coincidence problem, we gain a significant insight into the “Why now?” problem. For this purpose, we need to define the problem more precisely. One aspect of the “Why now?” problem is that the time when structure starts to form $`t_{\mathrm{str}}`$ and the time of matter-$`\mathrm{\Lambda }`$ equality $`t_\mathrm{\Lambda }`$ are close to each other. Within our solution to the triple coincidence problem, this means
$$t_{\mathrm{str}}(\pi \alpha )^2\frac{M_{Pl}^3}{M_{EW}^4}\left(\frac{\delta \rho }{\rho }\right)_p^{3/2}t_\mathrm{\Lambda }\frac{M_{Pl}^3}{M_{EW}^4}.$$
(13)
Here, $`(\delta \rho /\rho )_p10^5`$ is the primordial density fluctuation generated during inflation. Note that both time scales have the same parametric dependences on $`M_{Pl}`$ and $`M_{EW}`$, which makes this near equality natural. This is essentially the same as the statement that the triple coincidence is natural in our framework. In fact, the equality is surprisingly good, given that our solution by itself does not fix unknown $`O(1)`$ coefficients.
Another aspect of the “Why now?” problem is why we do not live at a time far beyond $`t_{\mathrm{str}}`$. This is an old mystery present even if the cosmological constant were zero and not our main concern here. Nonetheless it is interesting to discuss it briefly. Once structure forms, life as we know it is limited to the time before which all the available hydrogen has been burnt up in stars. The time scale for this is given by
$$t_{\mathrm{burn}}10^2\frac{\pi \alpha ^28\pi M_{Pl}^2}{m_pm_e^2}\frac{M_{Pl}^2}{M_{EW}^3}\frac{M_{EW}}{m_ph_e^2}$$
(14)
where $`h_e`$ is the electron Yukawa coupling. It is an old puzzle why $`t_{\mathrm{burn}}t_{\mathrm{str}}`$. Having expressed $`t_{\mathrm{str}}`$ in terms of the fundamental energy scales, the dependence on $`M_{Pl}`$ and $`M_{EW}`$ suggests that $`t_{\mathrm{burn}}`$ is in general much shorter than $`t_{\mathrm{str}}`$. This would actually strengthen our solution to the “Why now?” problem since it would leave little room between structure formation at $`t_{\mathrm{str}}`$ and the time when all nuclear fuel is used up at $`t_{\mathrm{str}}+t_{\mathrm{burn}}`$; then $`t_{\mathrm{str}}t_\mathrm{\Lambda }`$ would be enough to solve the puzzle. For a wide range of dimensionless parameters $`m_p/M_{EW}`$ and $`h_e`$, this near equality remains true. In reality, $`h_e`$ is so small that $`t_{\mathrm{burn}}100t_{\mathrm{str}}`$. The fact that $`t_{\mathrm{str}}`$ and $`t_{\mathrm{burn}}`$ are nearly equal in orders of magnitude is a purely numerical coincidence beyond the scope of our current understanding. Note, however, that the only requirement for solving the “Why now?” problem is that the two dimensionless quantities $`m_p/M_{EW}`$ and $`h_e`$ are not too small.
If $`m_p/M_{EW}`$ and/or $`h_e`$ were smaller than observed, stars could still exist much later than $`t_{\mathrm{str}}t_\mathrm{\Lambda }`$. However, it is intriguing to note that cosmology cannot be done long after the matter-$`\mathrm{\Lambda }`$ equality. Structures at cosmological distances go beyond the horizon, the cosmic microwave background is exponentially redshifted and nuclei become highly processed. Even if some life-forms remained, they might not be able to perform cosmological observations to discover the Big Bang.
6. In summary, we have introduced a new perspective on the cosmic coincidence problems. We begin with the assumption that there are only two fundamental energy scales in the problem, the Planck scale $`M_{Pl}10^{18}`$ GeV and the electroweak scale $`M_{EW}1`$ TeV. Electroweak scale physics produces a cold dark matter relic, while the cosmological constant is also tied to the electroweak physics as $`\rho _\mathrm{\Lambda }^{1/4}M_{EW}^2/M_{Pl}`$. Then we find that a triple coincidence among the matter, radiation, and $`\mathrm{\Lambda }`$ energy densities is a necessary consequence. Putting $`O(1)`$ coefficients back into the discussion, the matter energy density is enhanced by a factor of $`1/\pi \alpha ^210^3`$, and a cosmological window opens in the evolution of the Universe between the matter-radiation equality and $`\mathrm{\Lambda }`$-dominance during which structure forms. We presented a simple model in which the cosmological constant is indeed generated at $`\rho _\mathrm{\Lambda }^{1/4}M_{EW}^2/M_{Pl}`$. Moreover, we pointed out that the coincidence of the neutrino as well as the baryon energy density can also be simply understood. This framework allows us to understand why the time scale for structure formation and matter-$`\mathrm{\Lambda }`$ equality are comparable in orders of magnitude, solving the aspect of the “Why now?” problem involving the cosmological constant.
|
warning/0005/hep-th0005214.html
|
ar5iv
|
text
|
# An Extended Abelian Chern-Simons Model and the Symplectic Projector Method
## 1 Introduction
Nearly ten years ago, a method was developped that treats the fundamental question of canonically quantising field theories based on gauge symmetries. In this method, a crucial point is to identify, among the original constrained coordinates, those quantities related to the true degrees of freedom, which we refer to as the physical variables. We shall call this procedure the Symplectic Projector Method (SPM).
Along this line of investigation, the SPM has been tested through a number of relevant situations, such as Classical Electrodynamics, the 2-Dimensional Bosonised Schwinger Model , the Christ-Lee Model and the Chern-Simons-Maxwell Theory .
In this work, we reassess the efficacy of the method in picking up the true (physical) field coordinates for a sort of extended Abelian gauge model with a Chern-Simons term coupling a pair of gauge potentials. Such a model is the 3D descent of a 4D gauge theory with a topological mass term involving the Kalb-Ramond 2-form gauge potential. Also, the model discussed here is the core of the (gauge) bosonic sector (we leave behind an additive bilinear for a massive scalar field) of an $`N=2`$-supersymmetric gauge theory that leads to (after suitable identifications of fields) an $`N=2`$ model endowed with a rich structure of topological magnetic vortices .
Since our model displays a vector potential with a peculiar gauge transformation and without any dynamics, if taken on-shell, we believe it could also provide an interesting working example to test the consistency and the efficacy of the SPM.
Our paper is presented according to the following outline: in Section 2, we show explicitly the 4D origin of our model and perform its dimensional reduction towards the 3D mixed Chern-Simons theory; in Section 3, we establish the set of constraints and apply the SPM to pick up the physical variables. Finally, we present our General Conclusions.
## 2 The 4-Dimensional Model and Its Reduction
The 4-dimensional- $`U(1)\times U(1)`$ model we start off is based on the presence of a vector potential, $`A_{\widehat{\mu }}`$, together with a rank-2 gauge potential, $`B_{\widehat{\mu }\widehat{\nu }}=B_{\widehat{\nu }\widehat{\mu }}`$, the latter playing the rôle of a Kalb-Ramond field . We use hatted indices to denote components with respect to 4-dimensional space, while the bare ones refer to D=3 ($`\widehat{\mu }=0,1,2,3and\mu =0,1,2`$). The corresponding field strengths are given as below:
$`F_{\widehat{\mu }\widehat{\nu }}`$ $``$ $`_{\widehat{\mu }}A_{\widehat{\nu }}_{\widehat{\nu }}A_{\widehat{\mu }},`$ (1)
$`G_{\widehat{\mu }\widehat{\nu }\widehat{\kappa }}`$ $``$ $`_{\widehat{\mu }}B_{\widehat{\nu }\widehat{\kappa }}+_{\widehat{\nu }}B_{\widehat{\kappa }\widehat{\mu }}+_{\widehat{\kappa }}B_{\widehat{\mu }\widehat{\nu }},`$ (2)
and the coupling with a general matter field is carried out by means of the extended gauge-covariant derivative as below:
$$D_{\widehat{\mu }}\mathrm{\Phi }(_{\widehat{\mu }}+ieA_{\widehat{\mu }}+igG_{\widehat{\mu }})\mathrm{\Phi },$$
(3)
where $`G_{\widehat{\mu }}`$ is the dual of the field strength 3-form:
$$G_{\widehat{\mu }}\frac{1}{3!}ϵ_{\widehat{\mu }\widehat{\kappa }\widehat{\lambda }\widehat{\rho }}G^{\widehat{\kappa }\widehat{\lambda }\widehat{\rho }}.$$
(4)
This means that charged matter couples minimally to $`A_{\widehat{\mu }}`$ (coupling constant $`e`$) and non-minimally to $`B_{\widehat{\mu }\widehat{\nu }}`$ ($`g`$ is the coupling parameter governing the non-minimal interaction).
We propose to begin with the 4D action as follows:
$$_{4D}=\frac{1}{4}F_{\widehat{\mu }\widehat{\nu }}^{}{}_{}{}^{2}+\frac{1}{12}G_{\widehat{\mu }\widehat{\nu }\widehat{\kappa }}^{}{}_{}{}^{2}+\frac{1}{2}mϵ^{\widehat{\mu }\widehat{\nu }\widehat{\kappa }\widehat{\lambda }}A_{\widehat{\mu }}_{\widehat{\nu }}B_{\widehat{\kappa }\widehat{\lambda }}+(\text{matter-gauge terms});$$
(5)
we do not specify the matter-gauge coupling terms as we wish to discuss the quantisation of the gauge potential sector exclusively. For a discussion of the complete model, we refer the reader to the works of refs. . $`m`$ is the mass parameter associated to the presence of a massive spin-1 gauge boson in the spectrum .
The idea is now to dimensionally reduce the model to (1+2) dimensions, adopting the Scherk ansatz , by simply assuming that all fields do not depend on $`x^3`$:
$$_3(\text{fields})=0.$$
(6)
$`A_{\widehat{\mu }}`$ yields two independent fields in 3D, namely, a gauge potential, $`A_\mu `$, and a scalar, $`\phi A_3`$; on the other hand, two vector potentials stem from $`B_{\widehat{\mu }\widehat{\nu }}`$:
$$B_\mu B_{\mu 3}$$
(7)
and
$$Z_\mu \frac{1}{2}ϵ_{\mu \kappa \lambda }B^{\kappa \lambda }.$$
(8)
The gauge transformations read now as given below:
$`A_\mu ^{}`$ $`=`$ $`A_\mu +_\mu \alpha ,`$ (9)
$`B_\mu ^{}`$ $`=`$ $`B_\mu +_\mu \beta ,`$ (10)
$`Z_\mu ^{}`$ $`=`$ $`Z_\mu +ϵ_{\mu \nu \lambda }^\nu \xi ^\lambda ,`$ (11)
where $`\alpha ,\beta `$ and $`\xi ^\lambda `$ are arbitrary functions: $`\alpha `$ is the gauge parameter associated to the gauge symmetry of $`A_{\widehat{\mu }}`$, whereas $`\beta `$ and $`\xi ^\lambda `$ are the 3D descents ($`\beta \xi _3`$) of the (vector) gauge parameter associated to $`B_{\widehat{\mu }\widehat{\nu }}`$. The reduced model exhibits a $`[U(1)]^3`$-symmetry; the extra Abelian factor comes about by virtue of the 4D gauge symmetry of the Kalb-Ramond field . Moreover, eq.(11) displays an unusual gauge transformation for the vector potential $`Z_\mu `$, as we have anticipated. Such an exchange of rôles between the longitudinal and transverse sectors of the vector potential $`Z_\mu `$ (the longitudinal part is now gauge-invariant) has as a counterpart odd expressions for some of the constraints, as we shall see in Section 3.
The relationships between the 4D and 3D field strengths are readily worked out and the following expressions can be shown to hold:
$`F_{\widehat{\mu }\widehat{\nu }}^2`$ $`=`$ $`F_{\mu \nu }^22(_\mu \phi )^2,`$ (12)
$`G_{\widehat{\mu }\widehat{\nu }\widehat{\kappa }}^2`$ $`=`$ $`3G_{\mu \nu }^2+6(_\mu Z^\mu )^2,`$ (13)
$`ϵ^{\widehat{\mu }\widehat{\nu }\widehat{\kappa }\widehat{\lambda }}A_{\widehat{\mu }}_{\widehat{\nu }}B_{\widehat{\kappa }\widehat{\lambda }}`$ $`=`$ $`2ϵ^{\mu \nu \kappa }A_\mu _\nu B_\kappa 2\phi (_\mu Z^\mu ),`$ (14)
so that the 3D Lagrangian for the gauge fields takes over the form:
$$_{3D}^{\text{gauge}}=\frac{1}{4}F_{\mu \nu }^{}{}_{}{}^{2}+\frac{1}{2}(_\mu \phi )^2\frac{1}{4}G_{\mu \nu }^{}{}_{}{}^{2}+\frac{1}{2}(_\mu Z^\mu )^2+mϵ^{\mu \nu \kappa }A_\mu _\nu B_\kappa m\phi (_\mu Z^\mu ),$$
(15)
where $`F_{\mu \nu }`$ and $`G_{\mu \nu }`$ are the field strengths corresponding to $`A_\mu `$ and $`B_\mu `$, respectively. One should also notice the presence of a non-diagonal Chern-Simons term along with a partner that mixes $`\phi `$ and $`Z^\mu `$. The 3D action of eq.(15) is invariant under the $`U(1)\times U(1)\times U(1)`$ -symmetry quoted in eqs. (9), (10) and (11).
In the work of ref. , the potentials $`A_\mu `$ and $`B_\mu `$ were suitably identified with each other in a consistent way in connection with an N=2-D=3 supersymmetric version of the Maxwell-Chern-Simons model with anomalous magnetic couplings of matter to the gauge fields.
Before going ahead to discuss the symplectic quantisation of the model under consideration, we should perhaps mention that the 3 massive degrees of freedom that propagate in 4D are now accommodated in $`A_\mu `$ (1 d.f.), $`B_\mu `$ (1 d.f.) and $`\phi `$ (1 d.f.); the vector potential $`Z_\mu `$, featured with the unusual gauge transformation given in eq. (11) and with the potentially dangerous longitudinal kinetic term $`(_\mu Z^\mu )^2`$, does not propagate any on-shell degree of freedom (in fact, as displayed in eq.(8), $`Z_\mu `$ is just the dual of the well-known non-propagating 3D Kalb-Ramond field). This means that one of the $`U(1)`$-factors has no dynamical significance. The application of the Symplectic Projector Method to reassess the quantisation of this peculiar 3D gauge model in the sequel shall illustrate this procedure in a more evident way.
## 3 Constraints and relevant degrees of freedom
In view of what we have set previously, we define our model by means of the Lagrangian density:
$$=\frac{1}{4}F^{\mu \nu }F_{\mu \nu }\frac{1}{4}G^{\mu \nu }G_{\mu \nu }+\frac{1}{2}^\mu \phi _\mu \phi +\frac{1}{2}^\mu Z_\mu ^\nu Z_\nu m\left(^\mu Z_\mu \right)\phi +m\epsilon ^{\mu \nu \rho }B_\mu _\nu A_\rho ,$$
(16)
with
$`F^{\mu \nu }`$ $`=`$ $`^\mu A^\nu ^\nu A^\mu ,`$ (17)
$`G^{\mu \nu }`$ $`=`$ $`^\mu B^\nu ^\nu B^\mu ,`$ (18)
and the metric signature $`\eta ^{\mu \nu }=(+,,)`$. Picking up the canonically conjugate momenta, we have:
$$\pi ^\mu \frac{\delta }{\delta \left(_0A_\mu \right)}=F^{0\mu }+m\epsilon ^{\nu 0\mu }B_\nu ,$$
(19)
which yields
$`\pi ^0`$ $`=`$ $`0`$ (20)
$`\pi ^i`$ $`=`$ $`F^{0i}+m\epsilon ^{0ik}B_k.`$ (21)
Also,
$$P^\mu \frac{\delta }{\delta \left(_0B_\mu \right)}=G^{0\mu },$$
(22)
or
$`P^0`$ $`=`$ $`0`$ (23)
$`P^i`$ $`=`$ $`G^{0i}.`$ (24)
For the scalar field,
$$\pi _\phi \frac{\delta }{\delta \left(_0\phi \right)}=_0\phi .$$
(25)
Finally,
$$\pi ^\mu \frac{\delta }{\delta \left(_0Z_\mu \right)}=\left(m\phi +^\beta Z_\beta \right)\eta ^{\mu 0}$$
(26)
or
$`\pi ^0`$ $`=`$ $`\left(+^\beta Z_\beta m\phi \right)`$ (27)
$`\pi ^i`$ $`=`$ $`0.`$ (28)
Now, we are ready to write down the canonical Hamiltonian of the theory:
$`_c`$ $`=`$ $`\pi ^\mu _0A_\mu +P^\mu _0B_\mu +\pi ^\mu _0Z_\mu +\pi _\phi _0\phi `$ (29)
$`=`$ $`{\displaystyle \frac{1}{2}}\pi _i^2+{\displaystyle \frac{1}{2}}P_i^2+{\displaystyle \frac{1}{2}}\pi _\phi ^2+{\displaystyle \frac{1}{2}}\pi _0^2+A_0\left(_i\pi _i\right)+B_0\left(_iP_im\epsilon _{0ij}_iA_j\right)+`$
$`+{\displaystyle \frac{1}{4}}F_{ij}F_{ij}+m\epsilon _{0ik}\pi _iB_k+{\displaystyle \frac{1}{2}}m^2B_kB_k+{\displaystyle \frac{1}{4}}G_{ij}G_{ij}+{\displaystyle \frac{1}{2}}\left(_i\phi \right)^2+{\displaystyle \frac{1}{2}}m^2\phi ^2`$
$`+\pi _0^{}\left(m\phi +_iZ_i\right).`$
The primary Hamiltonian is just
$$_p=_c+v_i\mathrm{\Omega }_i,$$
(30)
where the primary constraints are
$`\mathrm{\Omega }_1`$ $`=`$ $`\pi _00`$
$`\mathrm{\Omega }_2`$ $`=`$ $`P_00`$
$`\mathrm{\Omega }_3`$ $`=`$ $`\pi _1^{}0`$ (31)
$`\mathrm{\Omega }_4`$ $`=`$ $`\pi _2^{}0.`$
As one can immediately notice, $`\mathrm{\Omega }_3`$ and $`\mathrm{\Omega }_4`$ are the first counterparts of the $`Z_\mu `$’s unusual features to be brought about in the set of constraints. The consistency condition imposed on the latter yields:
$`\mathrm{\Omega }_5`$ $`=`$ $`_i\pi _i0`$
$`\mathrm{\Omega }_6`$ $`=`$ $`_iP_im\epsilon _{0ij}_iA_j0`$ (32)
$`\mathrm{\Omega }_7`$ $`=`$ $`\pi _0^{}f\left(t\right)0,`$
for some arbitrary $`f(t)`$. They are all first class constraints; the gauge-fixing conditions will be so chosen that
$`\mathrm{\Omega }_8`$ $`=`$ $`A_00`$
$`\mathrm{\Omega }_9`$ $`=`$ $`B_00`$
$`\mathrm{\Omega }_{10}`$ $`=`$ $`Z_10`$
$`\mathrm{\Omega }_{11}`$ $`=`$ $`Z_20`$ (33)
$`\mathrm{\Omega }_{12}`$ $`=`$ $`_iA_i0`$
$`\mathrm{\Omega }_{13}`$ $`=`$ $`_iB_i0`$
$`\mathrm{\Omega }_{14}`$ $`=`$ $`Z_00,`$
where we have imposed the ”Coulomb” gauge-fixing on $`Z_\mu `$ in straight analogy to the usual procedure as applied onto $`A_\mu `$ (and $`B_\mu `$). As a net result, the collection of constraints related to $`Z_\mu `$ already indicate that its phase space variables are to be excluded from the dynamical subset.
We have now to build the matrix $`g_{ij}(x,y)=\{\mathrm{\Omega }_i(x),\mathrm{\Omega }_j(y)\}`$. Adopting the notational convention $`\delta \delta ^2\left(xy\right)`$, we get:
$`g(x,y)=`$ (48)
$`(\begin{array}{cccccccccccccc}0& 0& 0& 0& 0& 0& 0& \delta & 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& \delta & 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& \delta & 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \delta & 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& _i^x_i^y\delta & 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& _i^x_i^y\delta & 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \delta \\ \delta & 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& \delta & 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& \delta & 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& \delta & 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& _i^x_i^y\delta & 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& _i^x_i^y\delta & 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& \delta & 0& 0& 0& 0& 0& 0& 0\end{array}),`$
whose inverse reads as below:
$`g^1(x,y)=`$ (63)
$`\left(\begin{array}{cccccccccccccc}0& 0& 0& 0& 0& 0& 0& \delta & 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& \delta & 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& \delta & 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \delta & 0& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& +^2& 0& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& +^2& 0\\ 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& \delta \\ \delta & 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& \delta & 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& \delta & 0& 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& \delta & 0& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& ^2& 0& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& ^2& 0& 0& 0& 0& 0& 0& 0& 0\\ 0& 0& 0& 0& 0& 0& \delta & 0& 0& 0& 0& 0& 0& 0\end{array}\right).`$
We shall label the fields and their corresponding momenta as follows:
$`(A^0,A^1,A^2,\phi ,B^0,B^1,B^2,Z^0,Z^1,Z^2,\pi _0,\pi _1,\pi _2,\pi _\phi ,P_0,P_1,P_2,\pi _0^{},\pi _1^{},\pi _2^{})`$ $``$
$`(\xi _1,\xi _2,\xi _3,\xi _4,\xi _5,\xi _6,\xi _7,\xi _8,\xi _9,\xi _{10},\xi _{11},\xi _{12},\xi _{13},\xi _{14},\xi _{15},\xi _{16},\xi _{17},\xi _{18},\xi _{19},\xi _{20}).`$
At this stage, we are able to calculate the symplectic projector defined by the expression :
$$\mathrm{\Lambda }_j^i(x,y)=\delta _j^i\delta ^2\left(xy\right)\epsilon ^{ik}d^2zd^2w\left[\delta _k(x)\mathrm{\Omega }^m\left(z\right)\right]g_{mn}^1(z,w)\left[\delta _j(y)\mathrm{\Omega }^n\left(w\right)\right],$$
(64)
where
$$\delta _k(x)\mathrm{\Omega }^m\left(z\right)\frac{\delta \mathrm{\Omega }^m\left(z\right)}{\delta \xi _k(x)},$$
and $`\epsilon \left\{\epsilon ^{ik}\right\}`$ is the symplectic matrix. The prescription to obtain the projected variables is
$$\xi ^i\left(x\right)=d^2y\mathrm{\Lambda }_j^i(x,y)\xi ^j\left(y\right),$$
(65)
yielding the following expressions for those which are non-trivial:
$`\xi ^2\left(x\right)`$ $`=`$ $`A^1\left(x\right)`$ (66)
$`\xi ^3\left(x\right)`$ $`=`$ $`A^2\left(x\right)`$ (67)
$`\xi ^4\left(x\right)`$ $`=`$ $`\phi \left(x\right)`$ (68)
$`\xi ^6\left(x\right)`$ $`=`$ $`B^1\left(x\right)`$ (69)
$`\xi ^7\left(x\right)`$ $`=`$ $`B^2\left(x\right)`$ (70)
$`\xi ^{12}\left(x\right)`$ $`=`$ $`\pi _1^{}\left(x\right)+m{\displaystyle d^2y_{x_2}^2(x,y)\left[_{y_1}B_1\left(y\right)+_{y_2}B_2\left(y\right)\right]}`$ (71)
$`=`$ $`\overline{\pi }_1^{}\left(x\right)`$
$`\xi ^{13}\left(x\right)`$ $`=`$ $`\pi _2^{}\left(x\right)m{\displaystyle d^2y_{x_1}^2(x,y)\left[_{y_1}B_1\left(y\right)+_{y_2}B_2\left(y\right)\right]}`$ (72)
$`=`$ $`\overline{\pi }_2^{}\left(x\right)`$
$`\xi ^{14}\left(x\right)`$ $`=`$ $`\pi _\phi \left(x\right)`$ (73)
$`\xi ^{16}\left(x\right)`$ $`=`$ $`P_1^{}\left(x\right)+m_{x_1}{\displaystyle d^2y^2(x,y)\left[_{y_1}A_2\left(y\right)_{y_2}A_1\left(y\right)\right]}`$ (74)
$`=`$ $`P_1^{}\left(x\right)+m_{x_1}{\displaystyle d^2y^2(x,y)\left[\times A^{}\right]_y}`$
$`\xi ^{17}\left(x\right)`$ $`=`$ $`P_2^{}\left(x\right)+m_{x2}{\displaystyle d^2y^2(x,y)\left[_{y_1}A_2\left(y\right)_{y_2}A_1\left(y\right)\right]}`$ (75)
$`=`$ $`P_2^{}\left(x\right)+m_{x_2}{\displaystyle d^2y^2(x,y)\left[\times A^{}\right]_y}.`$
The reduced canonical Hamiltonian is obtained from the previous primary Hamiltonian taking into account the constraints and the gauge conditions, now looked upon as strong equalities. We obtain:
$`_c^r`$ $`=`$ $`{\displaystyle \frac{1}{2}}\pi _i^2+{\displaystyle \frac{1}{2}}P_i^2+{\displaystyle \frac{1}{2}}\pi _\phi ^2+{\displaystyle \frac{1}{2}}\pi _0^2+{\displaystyle \frac{1}{4}}F_{ij}F_{ij}+m\epsilon _{0ik}\pi _iB_k+{\displaystyle \frac{1}{2}}m^2B_kB_k+`$ (76)
$`+{\displaystyle \frac{1}{4}}G_{ij}G_{ij}+{\displaystyle \frac{1}{2}}\left(_i\phi \right)^2+{\displaystyle \frac{1}{2}}m^2\phi ^2+\pi _0^{}\left(m\phi \right),`$
or, in symplectic notation,
$`_c^r`$ $`=`$ $`{\displaystyle \frac{1}{2}}\left(\xi _{12}^2+\xi _{13}^2\right)+{\displaystyle \frac{1}{2}}\left(\xi _{16}^2+\xi _{17}^2\right)+{\displaystyle \frac{1}{2}}\xi _{14}^2+{\displaystyle \frac{1}{2}}\xi _{18}^2+{\displaystyle \frac{1}{2}}\left(_1\xi _3\right)^2+{\displaystyle \frac{1}{2}}\left(_2\xi _2\right)^2+`$ (77)
$`\left(_1\xi _3\right)\left(_2\xi _2\right)m\left(\xi _{12}\xi _7\xi _{13}\xi _6\right)+{\displaystyle \frac{1}{2}}m^2\left(\xi _6^2+\xi _7^2\right)+{\displaystyle \frac{1}{2}}\left(_1\xi _7\right)^2+{\displaystyle \frac{1}{2}}\left(_2\xi _6\right)^2+`$
$`\left(_1\xi _7\right)\left(_2\xi _6\right)+{\displaystyle \frac{1}{2}}\left(_1\xi _4\right)^2+{\displaystyle \frac{1}{2}}\left(_2\xi _4\right)^2+{\displaystyle \frac{1}{2}}m^2\xi _4^2+\xi _{18}\left(m\xi _4\right).`$
The physical Hamiltonian density is obtained by rewriting the one given above in terms of the projected variables:
$`^{}`$ $`=`$ $`{\displaystyle \frac{1}{2}}\left(\xi _{12}^2+\xi _{13}^2\right)+{\displaystyle \frac{1}{2}}\left(\xi _{16}^2+\xi _{17}^2\right)+{\displaystyle \frac{1}{2}}\xi _{14}^2+{\displaystyle \frac{1}{2}}\left(_1\xi _3^{}\right)^2+{\displaystyle \frac{1}{2}}\left(_2\xi _2^{}\right)^2+`$ (78)
$`\left(_1\xi _3^{}\right)\left(_2\xi _2^{}\right)m\left(\xi _{12}^{}\xi _7^{}\xi _{13}^{}\xi _6^{}\right)+{\displaystyle \frac{1}{2}}m^2\left(\xi _6^2+\xi _7^2\right)+{\displaystyle \frac{1}{2}}\left(_1\xi _7^{}\right)^2+`$
$`+{\displaystyle \frac{1}{2}}\left(_2\xi _6^{}\right)^2\left(_1\xi _7^{}\right)\left(_2\xi _6^{}\right)+{\displaystyle \frac{1}{2}}\left(_1\xi _4^{}\right)^2+{\displaystyle \frac{1}{2}}\left(_2\xi _4^{}\right)^2+{\displaystyle \frac{1}{2}}m^2\xi _4^2.`$
Finally, the equations of motion are obtained directly from the Hamilton-Jacobi equations by means of Poisson parentheses of the projected variables with the Hamiltonian $`d^2y^{}(y)`$. In so doing, we arrive at:
$$\stackrel{.}{\xi }_4^{}(x)=d^2y\{\xi _4^{}(x),^{}(y)\}=\xi _{14}^{}(x);$$
(79)
this yields
$$\stackrel{..}{\xi }_4^{}=\stackrel{.}{\xi }_{14}^{}=d^2y\{\xi _{14}^{},^{}(y)\}=m^2\xi _4^{}+^2\xi _4^{},$$
or
$$(\mathrm{}+m^2)\xi _4^{}=0.$$
(80)
Analogously, we obtain
$`\stackrel{..}{\xi }_2^{}`$ $`=`$ $`_2_2\xi _2^{}_1_2\xi _3^{}m\xi _{17}^{}`$
$`\stackrel{..}{\xi }_3^{}`$ $`=`$ $`_1_1\xi _3^{}_1_2\xi _2^{}+m\xi _{16}^{}`$
$`\stackrel{..}{\xi }_6^{}`$ $`=`$ $`m^2\xi _6^{}+_2_2\xi _6^{}_1_2\xi _7^{}m\xi _{13}^{}`$ (81)
$`\stackrel{..}{\xi }_7^{}`$ $`=`$ $`m^2\xi _7^{}+_1_1\xi _7^{}_1_2\xi _6^{}+m\xi _{12}^{}.`$
Now, eqs.(81) can be rephrased in a much simpler form if one chooses, without loss of generality, the momentum to lay upon the x-axis ($`\stackrel{}{k}=(k,0)`$), selecting the components ($`A_2,B_2`$) to be the transverse ones. One can easily notice that such a choice would cancel the variables $`\xi _2^{}`$, $`\xi _6^{}`$ and $`\xi _{12}^{}`$, and render the variables $`\xi _{16}^{}`$ and $`\xi _{17}^{}`$ equal to:
$`\xi _{16}^{}`$ $`=`$ $`m\xi _3^{},`$
$`\xi _{17}^{}`$ $`=`$ $`P_2^{}.`$
The set of independent variables would correspondingly be specified by the pairs ($`\xi _3^{},\xi _{13}^{}`$), ($`\xi _4^{},\xi _{14}^{}`$) and ($`\xi _7^{},\xi _{17}^{}`$). The remaining equations of motion would then read:
$`\mathrm{}\xi _4^{}`$ $`=`$ $`m^2\xi _4^{},`$
$`\mathrm{}\xi _3^{}`$ $`=`$ $`m^2\xi _3^{},`$
$`\mathrm{}\xi _7^{}`$ $`=`$ $`m^2\xi _7^{}.`$ (82)
So, we have a massive scalar, $`\xi _4^{}`$, and two massive transverse vector fields, $`\xi _3^{}`$ and $`\xi _7^{}`$, according to what could be expected from our counting of degrees of freedom in the framework of the 3D Lagrangian given by eq.(16), if we were to keep the mapping $`B_{\mu \nu }Z_\mu `$ in mind. Such an outcome also matches the natural allocation of physical degrees of freedom that could be inferred from the original 4D model. Moreover, the fact that the two vectors provide the physical sector with equivalent transverse massive contributions indicates the room for a consistent mapping into a new model, a procedure that can be implemented through the identification of the vector fields (and partners, in a supersymmetric context), as performed in Ref..
## 4 General Conclusions
We have reassessed the efficacy of the SPM in selecting the true dynamical set of phase space variables for a gauge 3D model hosting a peculiar topological term, namely, a mixed Chern-Simons bilinear. As a consequence of the dimensional reduction procedure, which defines our model as a descent of the 4D Cremmer-Scherk-Kalb-Ramond model , a counterpart for the mixed Chern-Simons term shows up as a source of mass for the scalar field. The SPM has proven to be efficient in casting the physical variables and exhibiting their dynamics through the field equations (82), where the expected topological mass generation is explicitly displayed.
Concerning the results obtained through Dirac’s method for the canonical quantisation of gauge systems, as applied to our 3D model, the dynamics turns out to be the same. This convergence of outcomes can be seen as another check of consistency for the SPM. Moreover, in contrast to the possibility, in principle, to reduce the phase space through Dirac’s approach, by using the SPM we always place ourselves in the right context for totally identifying, with the very explicit corresponding expressions, the true physical variables, an outcome lying on the very heart of the geometrical nature of the method.
The analysis we have pursued in this work happens to be an achievement in the sense of fully and rigorously specifying the dynamics generated in the gauge sector of an interesting N=2 off-shell supersymmetric model, one that can be mapped (through suitable identifications of fields) into another N=2 system in which topological self-dual vortex solutions can be found on-shell . The presence of two massive transverse vector excitations along with a massive scalar mode in the (bosonic) gauge sector is thus the relevant information confirmed as part of our knowledge about the above mentioned N=2-D=3 model. Also, applying the Symplectic Projector Method has proven to be a consistent choice for the complete clarification of the symplectic structure underlying the phase space spanned by this gauge model.
## Acknowledgements
The authors are grateful to Prof. J.A. Helayël-Neto for useful suggestions, discussions and the always kindest support to our work. L.R.U. Manssur and M.A. Santos also express their gratitude to the GFT-UCP (Group of Theoretical Physics), at Universidade Católica de Petrópolis - UCP (Petrópolis, Brazil), where part of this work has been done, for the kind and warm hospitality.
E-mail contact:
$``$leon@cbpf.br, $``$ nogue@cbpf.br
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.