id
stringlengths 24
24
| idx
int64 0
402
| paragraph
stringlengths 106
17.2k
|
|---|---|---|
61b1d32a0e35eb99589a8774
| 11 |
To test the ability of RBFE to predict antibiotic resistance we selected a small number of mutations in the RNAP and DNAG that confer resistance; to act as negative controls we added several more mutations known to have no clinical effect. We chose to test the most common resistanceconferring mutations for each drug. For RNAP this is S450L in the RRDR of rpoB and for the DNA gyrase these are A90V and D94G in the QRDR of gyrA (Fig. ). D94G is a robust test of RBFE as the mutation involves a significant change in amino acid properties and electrical charge.
|
61b1d32a0e35eb99589a8774
| 12 |
For rifampicin we also selected V170F and I491F in rpoB which both confer resistance, are proximal to both S450L and the antibiotic binding site, but are not in the RRDR (Fig. ). I491F is one of the so-called "disputed" mutations which either have variable or borderline rifampicin minimum inhibitory concentrations . For moxifloxacin we also tested E501D in gyrB which is close to the antibiotic binding site but not in the gyrA QRDR (Fig. ).
|
61b1d32a0e35eb99589a8774
| 13 |
When choosing negative controls, we prioritised mutations that were observed multiple times in clinical samples, are close to the drug binding site and do not involve a charge change or a proline residue. For the RNAP, L443F was selected since it lies within the RRDR and is close to the rifampicin binding site yet does not confer resistance and therefore is a good negative control (Fig. ). We also selected S388L and T585A which are further from the binding site and are seen in clinical samples. Finally, we choose an amino acid (Ser428) at which non-synonymous mutations are expected to confer resistance, since it lies in the RRDR, but for which no firm statistical association has been made, and choose a mutation (S428C) which minimally chemically perturbs the sidechain. We expect this to not confer resistance, since it has not been observed clinically and the sidechain points away from the drug and is therefore a good, if somewhat artificial, negative control.
|
61b1d32a0e35eb99589a8774
| 14 |
Testing different mutations at the same position which have different effects is a particularly stringent test of the ability of RBFE methods to predict antibiotic resistance. We therefore also tested the gyrA A90S mutation (Fig. ) -this is not seen clinically but a serine is present at the equivalent position in the DNA gyrase of other bacterial species and is suggested to help stabilise the gyrase-fluoroquinolone complex via participation in water-ion bridging interactions with the drug coordinated Mg 2+ . M. tuberculosis has some innate immunity to fluoroquinolones which has been suggested is due to the alanine at this position . The gyrA A90S mutation is therefore expected to strengthen the binding of moxifloxacin, thereby conferring hyper-susceptibility.
|
61b1d32a0e35eb99589a8774
| 15 |
The simplest approach is to assume that a positive value of the change in binding free energy of the antibiotic (ΔΔG < 0) indicates that the antibiotic binds less well to the target following the mutation and therefore would be predicted to confer resistance to that drug. Clinically, however, a sample is categorized as 'resistant' if its minimum inhibitory concentration (MIC) is greater than a critical concentration, often the epidemiological cutoff value (ECOFF/ECV), which is defined as the MIC of the 99 th percentile of a collection of phenotypically-wildtype samples. Such thresholds for both drugs were derived using published ECOFF/ECV values 28 as described previously . [1] N is the total number of free energy calculations used to calculate the DDG, excluding the rpoB restraints as their contributions were negligible (see Supplementary Information) N min and N max list the minimum and maximum number of repeat calculations used for apo or drug-bound de-charging (DG qoff ), van der Waals (DG vdW ) or re-charging (DG qon ) transitions, respectively (Figure ). * indicates a gyrB mutation.
|
61b1d32a0e35eb99589a8774
| 16 |
Three independent values of ΔΔG were first calculated. Each value of ΔΔG required the calculation of 6-8 alchemical free energies (Fig. , Methods). Repeats of the alchemical free energy components exhibiting the greatest variation were then run to efficiently reduce the confidence limits of the prediction as described in the Methods. First let us consider the overall values of ΔΔG and whether successful predictions can be made.
|
61b1d32a0e35eb99589a8774
| 17 |
Clinically the method as implemented would therefore return an 'Unknown' phenotype for these mutations. All three rifampicin-resistance conferring mutations, including the disputed mutation I491F, not only have positive values of ΔΔG but also lie above the clinical threshold derived from the ECOFF/ECV. These mutations are therefore correctly predicted to confer resistance to rifampicin.
|
61b1d32a0e35eb99589a8774
| 18 |
Both moxifloxacin negative controls (gyrA S95T & A90S) were correctly predicted to not affect the binding of moxifloxacin to the DNA gyrase. Although hyper-susceptibility is expected for A90S, the magnitude of the confidence limits prevents us drawing any conclusions. No definite prediction could be made for any of the three mutations associated with moxifloxacin resistance since the confidence limits of all three mutations straddled the clinical threshold. Unlike the RNA polymerase, two of the mutations to the DNA gyrase involved charged residues (gyrB E501D & gyrA D94G) and not surprisingly these had the largest estimated errors.
|
61b1d32a0e35eb99589a8774
| 19 |
To see how our ΔΔG values compared with clinical resistance measurements, we calculated an estimated 'expected ΔΔG' corresponding to the geometric mean of MICs associated with each of the resistance conferring mutations, using previously described methods . However, the errors in both the 'expected ΔΔG' and the ΔΔG values calculated by RBFE were too large to enable us to draw any conclusions about how well the values compare with one another (Fig. ).
|
61b1d32a0e35eb99589a8774
| 20 |
The magnitudes of the estimated errors prevented us from making a definite classification in six of the 12 mutations studied. One hypothesis is that the larger the alchemical perturbation, the larger the magnitude of error. We therefore examined whether there was a correlation between the number of atoms where the atom type was perturbed during the alchemical transition and the magnitude of error in calculated DDG values (Fig. ). There was a weak positive correlation for both RNAP and DNAG mutations, suggesting that whilst this does play a role, it is not the main driver behind the large errors observed here.
|
61b1d32a0e35eb99589a8774
| 21 |
To further examine what is driving the and confidence limits of the individual DDG values in Fig. , we analysed the alchemical free energy components from the de-charging (DG qoff ), van der Waals (DG vdW ) and re-charging (DG qon ) transitions (Fig. ) for both apo and drug-bound legs of the free energy calculations (Fig. ). As expected, for both the RNA polymerase and the DNA gyrase, there were no significant differences for the negative control mutations between the mean apo and drug-bound values of DG qoff , DG vdW and DG qon and the estimated error is generally low.
|
61b1d32a0e35eb99589a8774
| 22 |
For all three resistance-conferring mutations in rpoB the value of DG vdW when rifampicin is bound is significantly greater than the same transition for the apo protein and it is this that is mainly driving the positive value of DDG. The difference between the apo-and rifampicin-bound vdW transitions for V170F, S450L and I491F are 4.6, 5.6 and 5.3 kcal/mol, respectively. Since all three of these mutations involve the introduction of a larger sidechain that is oriented towards the bound drug, this is consistent with resistance arising primarily through steric hindrance of the rifampicin binding site. For comparison, despite a similar number of atoms being perturbed, there was no difference in the apo-and drug-bound values of DG vdW for the susceptible mutation rpoB L443F, which is also in the RRDR (Fig. ) and, whilst this also involves the introduction of a larger sidechain, in the crystal structure this is directed away from rifampicin. Differences in DG vdW between the apo-and complexed DNA gyrase also appear mainly responsible for the positive value of DDG for gyrA A90V, however the net effect is reduced. transitions. All results are normalized to the mean of the calculations for the apo leg for each transition for each mutation. Mean values are denoted by a cross and the error bars describe the 95% confidence limits, calculated from the SEM using the appropriate t-statistic. The free energy cost of removing the restraints for rifampicin is not shown since for all mutations it is negligible, indicating that restraints were likely not required to keep the drug in the binding site. An asterisk (*) indicates a gyrB mutation.
|
61b1d32a0e35eb99589a8774
| 23 |
Hence the variation in DDG arises mainly from the apo and complexed values of DG vdW -the notable exceptions being gyrB E501D and gyrA D94G. This is despite our efforts to minimize the overall error by running up to 4x the number of repeats for those transitions (Table ) to reduce their individual estimated errors. For gyrB E501D and gyrA D94G all three transitions contribute significant error, which since they add in quadrature, leads to a large overall error in DDG. This is not surprising since both mutations involve turning off (and on) electrical charge and D94G involves a net charge change that must be compensated for elsewhere in the system. To investigate how far we might reduce the errors, let us now consider the individual values of DG qoff , DG vdW and DG qon (Figure ). By starting each simulation from a different structural seed and discarding the first half of the alchemical free energy simulations and then applying statistics to the resulting values of DG we are assuming that they are independent. If true, then one would also expect the values to be normally distributed which would appear to be the case for most sets of DG values (Figure ).
|
61b1d32a0e35eb99589a8774
| 24 |
Applying the Shapiro-Wilks test of normality to the rpoB data confirms that, despite the small numbers of samples in some cases, the majority of DG values are indeed normally distributed with the exceptions of DG qon for the apo leg of S428C and DG qon for the drug bound leg of S450L. For two DNA gyrase mutations there was also evidence of non-normality in the DG vdW for the apo leg of gyrA S95T and DG on for both the apo and drug bound leg of gyrA D94G.
|
61b1d32a0e35eb99589a8774
| 25 |
To test how far our simulations are from normality, we extended four apo and five drug-bound simulations underlying the most variable component (qon, Fig. ) of the most complex mutation, gyrA D94G, by an order of magnitude (from 0.5 ns to 5 ns). As assessed by the Shapiro-Wilks, the resulting distributions of apo-and drug-bound free energies were indeed normal after 5 ns of simulation (p = 0.92 and p = 0.16) but the distribution of results for the repeated calculations, and therefore the error, remain large (Fig. ).
|
61b1d32a0e35eb99589a8774
| 26 |
We have shown how relative binding free energy (RBFE) techniques can be applied to large protein complexes to predict, with some success, the effect of individual protein mutations on the binding of an antibiotic, and thence whether resistance is conferred. When the size of the signal is large and the mutations do not involve significant changes in the electrical charge, as is the case for the rpoB mutations, one can successfully predict whether a mutation confers resistance to the antibiotic (in this case rifampicin). If the fold increase in minimum inhibitory concentration is small and/or there are significant charge changes, as is the case for most of the resistance conferring mutations in the DNA gyrase, then the estimated error of DDG will likely be so large that no definite prediction can be made. In addition, the observed non-normality of the DG qon free energy for gyrA D94G indicates that these values are also not independent: to solve this either the λ simulations would have to be extended or the equilibration simulations would need to be more numerous as well as longer.
|
61b1d32a0e35eb99589a8774
| 27 |
Despite the focus on resistance, it is more useful to be able to accurately and reproducibly predict susceptibility since clinically that will lead to immediate action i.e. starting the patient on the appropriate treatment regimen. A prediction of resistance will likely result in the sample being sent for further testing, at which point any incorrect predictions (false positives) will be detected.
|
61b1d32a0e35eb99589a8774
| 28 |
Unfortunately, it may be more difficult to predict susceptibility than resistance using RBFE, and for rpoB only one susceptible mutation could be confidently predicted. If we assume that most susceptible mutations will not affect the binding affinity for the antibiotic, then they would have a DDG of zero. The predicted DDG of such mutations would hence require the estimated error to be at least less than the value of the ECOFF (for rpoB and DNA gyrase 1.2 and 0.9 kcal/mol, respectively) to make a confident susceptible prediction. The magnitude of error for the mutations in this study was greater than the relevant ECOFF in all but one case (rpoB S388L, Table ). It is likely that the error could be reduced by running a greater number of repeats, however some mutations can result in a small increase in MIC but not enough to confer resistance (as susceptibility can be defined as any MIC up to the ECOFF), and in such cases even a lower level of error (± 0.5 kcal/mol) may still prove insufficient for prediction.
|
61b1d32a0e35eb99589a8774
| 29 |
Whilst alchemical binding free energy calculations therefore can play a role in predicting antibiotic resistance, the method currently works best when the target protein is small and the magnitude of the change in the binding free energy large, as is the case for S. aureus DHFR and trimethoprim . For this system it has also been shown that it is possible to reduce the length of the simulations yet further but still maintain an accurate qualitative prediction . Taking all this together, we appear to have probed the limits (for now at least) of using RBFE methods to predict antibiotic resistance de novo. Interestingly, unlike the majority of other applications of RBFE, one can tolerate large, estimated errors since we are ultimately only interested in the final binary classification of resistant or susceptible. A second and related application for RBFE is reducing the likelihood of a lead compound developing resistance by providing information during the development process of the likely mutations that could confer resistance and we hope to explore this in future work.
|
61b1d32a0e35eb99589a8774
| 30 |
Given we chose to use very short λ simulations the latter is almost certainly not true and whilst the majority of our calculated DG values appear to be normally distributed, some are not which is concerning. One would expect to have to run 4x the number of simulations to reduce the estimated error to half its original value if the simulations are independent. This makes simulations of these size prohibitively computationally intensive at the time of writing.
|
61b1d32a0e35eb99589a8774
| 31 |
It is not in doubt that how the structure and dynamics of a protein change upon mutation contains valuable information that can, in theory, be used to predict whether individual mutations confer antibiotic resistance. An alternative route, which is much less computationally intensive, is to train machine learning models using structural and chemical input features. This is the focus of future
|
628bf40843d1f07e2e36c6e1
| 0 |
Molecular dynamics (MD) simulations are a cornerstone of modern-day computational chemistry and biology, and in the last 40 years a range of general-purpose codes have been developed to make simulations more reproducible and accessible to the community . Although the preparation of typical biomacromolecular systems, e.g. proteins, lipid membranes or nucleic acids containing standard residues, has become extremely streamlined and automated , the corresponding topology files -containing a full description of the system's connectivity and energetics -can be unreadable and not amenable to simple manual manipulation. As a result, any non-automated departure from these standards, such as removing atoms or adding bonds, can become a painstakingly complex process, in particular for novice users not aware of nuances and caveats of these fixed file formats.
|
628bf40843d1f07e2e36c6e1
| 1 |
To deal with these routine but non-standard tasks, high-level utility libraries are needed that can accommodate the increasing complexity of inputs and dynamically respond to popular requests from the community. One notable example of such a library, Parmed, has already been incorporated into Amber and contains many OpenMM-specific functions , but its primary goal is to provide easy access to general properties and allow for smooth interconversion between standard file formats. Another one, pmx, serves to prepare hybrid topologies for "alchemical" simulations and the analysis of thus obtained results . Many routine tasks, though, are not covered by any popular library, leaving plenty of room for new developments.
|
628bf40843d1f07e2e36c6e1
| 2 |
In an attempt to fill that gap, we present Gromologist, a Pythonic utility library oriented at custom manipulation of Gromacs topologies. For many years now, Gromacs has been not only the most broadly used but also one of the fastest-growing MD engines in the field of molecular simulations and has a well-established user base, whichcombined with the ease of format interconversion allowed by other tools -makes Gromologist a broadly applicable tool for all types of non-standard operations including, but not limited to, adding and removing atoms and bonds, introducing amino acid mutations in both the structure and topology, checking and listing force field parameter compatibilities, looking for mismatches between structures and topologies, merging files, manipulating alchemical states, modifying force field types or combination rules, parameter optimization or automated editing of structures. In addition, we maintain a strict zero-dependency policy (except for a few very specific features), so that the library is conveniently lightweight and can be easily installed with any modern Python distributions without causing dependency conflicts.
|
628bf40843d1f07e2e36c6e1
| 3 |
To handle Gromacs topologies, Gromologist implements a hierarchical representation of sections, subsections and entries. Sections represent high-level abstractions, such as individual molecule definitions, force field parameter sets or headers/footers. Subsections correspond to individual fragments of the topology file, delimited with the [ subsection name ] syntax, and directly handle most operations. Entries correspond to individual lines and provide a low-level access to all key features of the contents of a topology.
|
628bf40843d1f07e2e36c6e1
| 4 |
Splitting and merging. By default, topologies are saved as a single standalone file, including all .itp files referenced therein, but splitting to individual .itp is also supported. If needed, files can be made lightweight by keeping (a) only molecules listed in the [ system ] subsection and/or (b) only parameters used by the system.
|
628bf40843d1f07e2e36c6e1
| 5 |
Bonds and atoms. Primitive operations on topologies include adding, removing or swapping atoms in molecules while preserving a correct numbering scheme across all sections. Adding or removing bonds is also supported, including new bonds between two molecules, in which case the two become merged into a single molecule entry, and all new bonded terms (pairs, angles, dihedrals) are automatically identified and added, while existing ones are preserved and renumbered correctly. This allows for easy creation of chemical adducts or cyclic molecules, as well as editing of protonation states or simple chemical modifications without the need of extending the basic residue library (.rtp/.hdb files). Another heavily requested feature was the automated introduction of ,,special bonds" -disulfides and transition metal coordination bonds -that are not always correctly guessed by default Gromacs tools. Finally, Gromologist also provides a low-level interface for the mutation library, and can be easily incorporated into custom modification workflows.
|
628bf40843d1f07e2e36c6e1
| 6 |
Mutations. In this vein, a mutations module allows to introduce standard amino acid mutations to the topology without the need to go through the full process of topology generation. This has proven particularly useful in the study of heavily glycosylated systems, such as the SARS-CoV-2 Spike , where ready-made topologies and systems were often shared publicly but introducing even minor changes (such as a point mutation) would entail going through the same complicated server-based pipeline, risking introducing additional errors at this stage. This idea is also extended to protonation states, so that individual titratable residues can be protonated or deprotonated easily and in a consistent manner. As long as standard atom name conventions are followed, the mutation module is force-field agnostic, and the user can point to a specific .rtp file containing the respective residue definitions.
|
628bf40843d1f07e2e36c6e1
| 7 |
Force fields. Finally, some features are meant to facilitate force field development and the implementation of selected methods such as solute tempering (REST2) . These include the possibility of automatically cloning types along with all their bonded interaction parameters and renaming all selected atoms to the new type, or by automatically adding NBFIX-type modifications just by specifying the deviation from the standard Lorentz-Berthelot rules . An additional module implements a generic strategy for multiple dihedral fitting to QM data, akin to that available in the FFTK module of VMD for NAMD topologies .
|
628bf40843d1f07e2e36c6e1
| 8 |
In addition, many often-required features of the PDB files can be filled in automatically. Chain assignment, an information typically lost in PDB to GRO conversion can be inferred based on a simple distance criterion, similarly to the CONECT entries required by some analysis programs. Elements can be inferred from atom names, and beta-factors can be set to reflect external data sets, including an option to smooth the values out spatially.
|
628bf40843d1f07e2e36c6e1
| 9 |
To assist with these operations, Gromologist implements a robust selection language akin to that of VMD , allowing for logical operations, selections "within X of Y", and a number of predefined molecule classes such as "protein", "nucleic" or "solvent". Periodic boundary condition (PBC) treatment is also implemented for any generalized triclinic box.
|
628bf40843d1f07e2e36c6e1
| 10 |
In Gromologist, topologies can be matched with a compatible structure, so that where relevant, topology operations can be simultaneously performed on the structure attribute (and vice versa), facilitating the work with simulation-ready systems, and fully leveraging the information available in both files. Examples of such procedures include atom addition/removal, or introduction of mutations. When a topology is available, chain assignment can be made using molecule definitions therein. Moreover, when Gromacs often only indicates an unspecified mismatch between e.g. the number of atoms in structure and topology, Gromologist provides a specific list of all unmatched atom names, facilitating the identification of missing atoms or residues in either one.
|
628bf40843d1f07e2e36c6e1
| 11 |
The last area of applications of Gromologist covers extracting information about the topologies and structures, or making it better visible to the user. On the most basic level, one can list molecules in the system, atoms in the molecule, as well as selected bonded terms (bonds, 1-4 pairs, angles, dihedrals) using atom names or atom types. Charges and masses of both the whole system and individual molecules are easy-to-access attributes. For a protein or a nucleic acid structure, the sequence can be printed chain by chain, and per-residue missing atoms identified, helping identify cases in which residues have to be rebuilt. Gradually diagnostic functions are being incorporated to quickly identify structural issues prior to simulation, like one currently allowing for validation of all chiral centres in a protein.
|
628bf40843d1f07e2e36c6e1
| 12 |
When required, atom names can be explicitly stated in the comments of bonded interaction entries to facilitate debugging, and numerical values of the parameters can be explicitly included in the topology. In the same vein, fields set using the define syntax can be set to their explicit values. Parameters whose numerical definitions are missing can also be identified with a single function call.
|
628bf40843d1f07e2e36c6e1
| 13 |
To be applicable as broadly as possible, Gromologist uses no dependencies other than base Python libraries. Although a few optional features, such as smoothing of beta-factors or dihedral optimization, do require numpy and scipy, these make up extremely rare use cases and therefore do not form part of the official dependency list.
|
628bf40843d1f07e2e36c6e1
| 14 |
Given the existing set of capabilities, Gromologist is becoming an easily extensible platform for the introduction of further convenience functions, editing tools and workflows. Further development of the platform could benefit the Gromacs community by rapidly implementing utilities tailored to new or popular protocols, and a closer collaboration with the Gromacs development team is expected to directly address these issues. Moreover, extensive testing and some code reorganization will be required to ensure that the library covers all features supported by Gromacs, including more cryptic ones. In the long run, however, we expect Gromologist to become a stable element of the Gromacs environment and a valuable contribution to the broad simulation community.
|
6787f5eb81d2151a024c0f07
| 0 |
On-road vehicle traffic is one of the largest sources of hazardous air pollutant emissions in the United States (U.S.), with a disproportionate impact on historically disadvantaged communities. Numerous studies have shown that these communities bear a higher burden of exposure to traffic-related air pollution (TRAP), which is typically highest near major roadways and declines to background levels within a few hundred meters of the road corridors . Proximity to high-traffic roads has been linked to increased risks of adverse health outcomes, including respiratory and cardiovascular diseases, as evidenced by numerous health effect studies . A recent systematic review and meta-analysis by the Health Effects Institute Panel (2022) underscored strong evidence linking TRAP exposure to all-cause mortality, circulatory mortality, heart disease in adults, and varying degrees of evidence for lung cancer mortality, asthma in adults, and respiratory infections in children . Racial, ethnic, and socioeconomic disparities in exposure to TRAP in the U.S. are well-documented and persist despite overall reductions in vehicle emissions levels , exacerbating public health challenges in underserved communities.
|
6787f5eb81d2151a024c0f07
| 1 |
Recent research has focused on the contributions of medium-and heavy-duty truck traffic, which emit significantly higher levels of particulate matter (PM) and nitrogen oxides (NOX) than lightduty vehicles . These diesel-powered vehicles are a major source of TRAP emissions, even though they make up a smaller proportion of the U.S. vehicle fleet by distance traveled. Despite their smaller numbers, medium-and heavy-duty trucks are responsible for a disproportionate amount of pollution, which contributes to exposure disparities among racial, ethnic, and socioeconomically disadvantaged groups . The elevated emissions from these vehicles, particularly in urban and freight corridor environments, heighten the urgency of addressing the differential exposure levels faced by underserved communities.
|
6787f5eb81d2151a024c0f07
| 2 |
In epidemiology studies, TRAP surrogates such as proximity to high traffic roads and traffic density have commonly been used as proxies for estimating traffic-related exposure, given the complexities and impracticalities of directly measuring all pollutants in the traffic-mixture. These surrogates have been widely employed in national-scale analyses to assess TRAP exposure risks and highlight disparities in exposure across racial, ethnic, and socioeconomic groups . While these proximity models are simple to implement, they fail to account for crucial factors that influence emission rates, pollutant dispersion, and chemical transformations. To address these limitations, researchers have developed more sophisticated methods using atmospheric chemical transport models (CTMs) and dispersion models to estimate vehicle emission concentrations along roadways, particularly in urban areas . These models provide more precise estimates of pollution concentrations but are computationally intensive and challenging to scale for national-level analyses. The difficulty of applying these methods on a broad scale while maintaining high spatial precision limits their use in national-scale TRAP exposure studies.
|
6787f5eb81d2151a024c0f07
| 3 |
A significant body of research has documented disparities in TRAP exposure across racial, ethnic, and income groups in the U.S., consistently finding that people of color and lowerincome populations are disproportionately exposed to higher levels of air pollution. Studies have shown that Black, Hispanic, and Asian communities experience both higher levels of trafficrelated pollution and worse health outcomes , such as respiratory and cardiovascular diseases, when compared to White populations. A recent review by , which examined 55 studies on air pollution exposure by ethnic groups in high-income countries, reported that U.S. minority ethnic groups are more exposed to air pollutants like PM2.5 and NO2 than the majoritized ethnic group (i.e., White) . However, despite the clear evidence of exposure disparities, few national-scale studies have quantified the contributions of different vehicle types-such as light-duty vehicles (LDVs), medium-duty vehicles (MDVs), and heavy-duty vehicles (HDVs)-to exposure and equity separately . Furthermore, these studies often focus on a single pollutant, typically PM2.5, and rely on spatially coarse resolution data, such as county level aggregates from the EPA National Emissions Inventory (NEI). While useful for identifying broad trends, this resolution can obscure localized exposure patterns, especially near transportation corridors. Additionally, many studies lack the statistical analysis needed to assess exposure disparities across racial, ethnic, and socioeconomic groups.
|
6787f5eb81d2151a024c0f07
| 4 |
(2) evaluate the equity of exposure to near-roadway emission hotspots, with a particular focus on historically underserved communities; and (3) quantify the contributions of LDVs, MDVs, and HDVs to both overall exposure levels and disparities in exposure equity. These objectives aim to provide a comprehensive understanding of the spatial distribution of TRAP and its disproportionate impact on historically underserved populations.
|
6787f5eb81d2151a024c0f07
| 5 |
In this study, we introduce a novel surrogate for vehicle emission exposure, termed "vehicle emission density," which offers a continuous, fine-scale assessment with nationwide coverage. This approach overcomes key limitations of existing methods, providing a more comprehensive analysis of emissions exposure while remaining computationally feasible. By leveraging a national traffic dataset and the EPA's MOVES model, we estimate emission density values for every census block across the U.S., enabling a more precise evaluation of exposure to emissions from LDVs, MDVs, and HDVs. This method improves upon previous models, which often rely on coarser proximity-based or traffic density approaches, by offering greater spatial resolution in exposure assessments. Given the significant health impacts of PM2.5 and NOX emissions, our focus on these pollutants is particularly relevant, especially considering the disproportionately high emissions from medium-and heavy-duty trucks, which are major contributors to both particulate matter and nitrogen oxide concentrations in urban and freight corridor environments. This research advances the refinement of near-roadway emission exposure surrogates, striking a balance between spatial precision and computational efficiency. The vehicle emission density method more accurately captures variations in emission rates driven by factors like climate, vehicle maintenance, and traffic patterns. It also breaks down emissions from LDV, MDV, and HDV traffic on individual roadway segments. By calculating emission densities at the census block level, our study offers a nuanced understanding of near-roadway exposure and highlights disparities across national, state, and local levels. This work provides key data and insights to assess equity in traffic-related air pollution exposure and inform policies to reduce these disparities.
|
6787f5eb81d2151a024c0f07
| 6 |
To assess exposure to near-roadway hotspots and evaluate exposure equity, we first estimate the emission density for PM2.5 and NOX at the census block level, stratified by vehicle type (LDV, MDV, and HDV). This approach builds upon and refines the traffic density methodologies used in our previous studies . Emission factors are derived from the EPA's Motor Vehicle Emission Simulator (MOVES), while vehicle activity data is obtained from the Federal Highway Administration's (FHWA) Highway Performance Monitoring System (HPMS).
|
6787f5eb81d2151a024c0f07
| 7 |
The emission density metric quantifies the total vehicle emissions within each U.S. census block, including a 250-meter buffer zone surrounding the block. These emissions are the normalized by the area of the census block to obtain a density value in grams per square kilometer (g/km²). We compute separate emission densities for LDVs, MDVs, and HDVs to capture the distinct contributions of each vehicle class to local air quality.
|
6787f5eb81d2151a024c0f07
| 8 |
Next, we analyze the relationship between demographic characteristics (e.g., racial, ethnic, and socioeconomic variables) at the block level and emission density. This analysis enables us to evaluate exposure equity and identify disparities in traffic-related air pollution across different communities throughout the U.S. By using detailed spatial data and integrating community-level demographic information, we provide a comprehensive assessment of both exposure levels and the equity of exposure to near-roadway pollution.
|
6787f5eb81d2151a024c0f07
| 9 |
The 2018 HPMS dataset encompasses the U.S. public highway system as of the 2018 calendar year, offering comprehensive insights into its extent, condition, performance, usage, and operational characteristics. Of particular significance to this research, it includes traffic volume estimates derived from state-collected traffic counts conducted annually, covering all highways, arterials, and collectors (classified as functional classifications 1-6). Traffic count data within the HPMS are gathered through a stratified, random sampling method aimed at achieving traffic count estimates meeting minimum precision standards, typically less than 10% error for at least 80% of roadway links (U.S. . The sampling rate varies by roadway functional classification and vehicle type, accounting for traffic volume variability across different regions, including states, rural, small urban, and urbanized areas. The HPMS does not encompass traffic volume data for low-volume roads in rural regions and minor residential streets in urban areas (classified as functional classification 7), which collectively represent an estimated 14.9% of the total U.S. traffic volume (U.S. . Table provides a comprehensive breakdown of total lane-kilometers and total vehicle-kilometers traveled (VKT) categorized by vehicle type and functional classification. The extensive dataset includes approximately 6.5 million rows (representing road links) and covers around 401,000 lane-kilometers, accounting for a total of 12.5 billion annual daily vehicle-kilometers traveled across the six roadway functional classifications. The HPMS provides an estimate of traffic volume from all vehicles as a single figure of AADT for each roadway link, also known as a roadway segment. On higher functional system roadways (generally lower volume roads), the HPMS requires less detailed data for MDV and HDV traffic. Tabulations of MDV and HDV AADT are provided for all interstate highway and national highway system links. For other roadways, HDV and MDV AADT estimates are available for a stratified random sample of roadway links, adhering to FHWA requirements (U.S. FHWA, 2016). These are designed to offer sufficient information to estimate MDV and HDV traffic volumes for each state, rather than for each individual roadway segment. The availability of MDV and HDV traffic volume estimates varies among states due to differences in the capabilities of their counting equipment and specific data requirements.
|
6787f5eb81d2151a024c0f07
| 10 |
The HPMS dataset, in its original state, is annotated with a five-digit UACE (Urban Area Census Code) based on the 2010 Decennial Census (U.S. Census Bureau, 2020). A UACE of '99999' denotes a rural road segment, whereas '99998' denotes a small urban road segment. These small urban road segments are defined as those supporting a population of at least 5,000 and situated outside the adjusted urbanized area. Any UACE code other than '99999' or '99998' designates an urban road segment. To address any inaccuracies in the existing UACE data, such as missing or erroneous entries, the HPMS road network is further aligned (geographically intersected) with the urban and rural boundaries delineated in the 2010 Census. As a result, approximately 14% of road links (roughly 16% of roadway lane-kilometers) undergo modification in the UACE dataset.
|
6787f5eb81d2151a024c0f07
| 11 |
Further preprocessing involves eliminating road links lacking total AADT estimates and excluding road links where the total AADT estimate falls below the combined sum of MDV and HDV AADT. Consequently, about 4.8% of road links (equivalent to approximately 11% of roadway lane-kilometers) are eliminated from the HPMS dataset. Any roads lacking an estimate for the number of through lanes are presumed to have two lanes.
|
6787f5eb81d2151a024c0f07
| 12 |
Approximately 28% of road links in the HPMS dataset, representing 46% of roadway lanekilometers, lack MDV and HDV AADT data. To address this, we apply random forest regression (RFR) to estimate traffic volumes in data-deficient areas. RFR improves upon linear regression by capturing complex, nonlinear relationships in the data, providing more accurate estimates where traditional models fall short.
|
6787f5eb81d2151a024c0f07
| 13 |
We develop two national-level RFR models to predict AADT for MDVs and HDVs. These models incorporate total AADT, FHWA roadway functional classification, number of through lanes, and indicator variables for state and county. Given the dataset's scale and the computational demands of training each model, we use Bayes Search Cross Validation for hyperparameter tuning. This approach optimizes each model based on negative root mean squared error over 48 iterations with 3 cross-validation folds. For the MDV AADT model, we determine that 98 estimators with no maximum depth provide optimal performance. Conversely, the HDV AADT model benefits from 95 estimators with a maximum depth limited to 40. Both models are configured with a minimum sample split of 2, a minimum sample leaf of 1, and no maximum features.
|
6787f5eb81d2151a024c0f07
| 14 |
Following model development, we evaluate their performance by examining residuals and using local regression (loess). The results indicate no significant bias in the predictions relative to the magnitude of predicted values. Furthermore, we conduct a 5-fold Cross Validation and evaluate the same model evaluation metrics, which yield similar results. This robust performance suggests that our RFR models effectively capture the complex relationships between predictor variables (such as total AADT, roadway classification, and lane configurations) and MDV and HDV traffic volumes, while minimizing prediction errors.
|
6787f5eb81d2151a024c0f07
| 15 |
We use version 3 of the MOVES (U.S. EPA, 2020a), developed by the U.S. Environmental Protection Agency (EPA), to construct country-specific emission factor look-up tables. Given the extensive computational requirements for modeling annual average daily emissions at the link level across the entire U.S. using MOVES, we develop a streamlined approach. This approach involves selectively running the software for counties that represent unique combinations of model parameters for January and July, similar to the methodology used in the 2017 National Emissions Inventory (NEI) by the U.S. EPA (U.S. EPA, 2021). January and July are chosen to represent winter and summer conditions, with the resulting values averaged. Our study focuses exclusively on primary emissions-those directly emitted-of NOX and PM2.5, and does not address emissions resulting from secondary processes, which involve the atmospheric formation of pollutants.
|
6787f5eb81d2151a024c0f07
| 16 |
We develop MOVES modeling scenarios tailored to representative counties by integrating multiple datasets. Initially, we source county-level average monthly temperature and humidity data for January and July from the MOVES default database. These values are derived from the National Climatic Data Center's average temperature records spanning 2001 to 2011, ensuring comprehensive coverage across U.S. counties. Additionally, we gather information on countyspecific MOVES fuel regions and vehicle inspection and maintenance (I/M) programs. These factors exhibit significant variability between urban and rural areas within states, particularly in regions addressing air quality concerns. Across the U.S., we identify 117 distinct combinations of fuel regions and I/M programs. Subsequently, we merge these datasets to identify distinct Climate-Fuel-I/M combinations, rounding temperature and humidity values to the nearest 10 degrees (or %), as appropriate.
|
6787f5eb81d2151a024c0f07
| 17 |
In total, we identify 444 unique Climate-Fuel-I/M combinations. For each combination, representative counties are selected based on county population data sourced from the 2020 Census (U.S. Census Bureau, 2021). For combinations encompassing only one or two counties, all counties are included. For combinations spanning more than two counties, the selection criteria involve identifying the county with the largest population and the county closest to the 25th percentile in population size. This approach aims to capture a spectrum of urbanization levels within each Climate-Fuel-I/M combination. The county with the largest population represents the most urban, while the county closest to the 25th percentile represents a more rural context within the same Climate-Fuel-I/M grouping. In total, we identify 701 unique countymonth combinations based on these Climate-Fuel-I/M scenarios, encompassing counties from 479 unique counties across the U.S. This selection represents approximately 15.2% of all U.S.
|
6787f5eb81d2151a024c0f07
| 18 |
We conduct emissions modeling using MOVES default data for representative county-month combinations at the county level, focusing on the 2018 calendar year. Our methodology involves using MOVES Inventory mode with the Distance Traveled activity option to estimate link-level total emissions (in grams) and the total distance traveled (in miles). .1 for details). We then map these emission factors to individual roadway links, considering the county, MOVES road type, roadway functional classification, and roadway urban-rural classification (see Table .2 for the relational table detailing these associations). Finally, we apply these emission factors to HPMS traffic data, calculating the annual average emission rates per roadway link (g/mi/day) by multiplying the emission factor by the VMT and dividing by the road link length, for each roadway segment, vehicle type, and pollutant.
|
6787f5eb81d2151a024c0f07
| 19 |
Next, we estimate emission densities (in g/km²) for each U.S. census block. This process begins by creating a 250 m spatial buffer around each census block using GIS software. The buffer is intended to capture nearby roadways that might contribute to emissions exposure, even if they do not directly intersect the census block. Roadway segments that intersect the buffer are divided into two categories: those entirely within the buffered area and those partially crossing the buffer. For segments fully within the buffer, we aggregate the emissions data. We then compute the emission density for each type of vehicle traffic and pollutant by dividing the total emissions by the area of the census block. The overall emission density for all vehicle traffic, known as the total emission density, is calculated by summing the emission densities for LDV, MDV, and HDV traffic.
|
6787f5eb81d2151a024c0f07
| 20 |
We use 2020 Census block-level population data concerning race and ethnicity (U.S. Census Bureau, 2021), along with 2016-2020 American Community Survey (ACS) 5-Year income data at the block-group-level and poverty status data at the tract-level (U.S. Census Bureau, 2022). In our analysis, individuals identifying as "White" indicate those who reported their race as White alone in the 2020 Census, while "non-White" encompasses all other racial identities. The "other race" category includes individuals who self-identify as American Indian and Alaska Native alone, Native Hawaiian and Other Pacific Islander alone, or some other race category that does not fall under Black or Asian alone categories. This classification is used to encompass diverse racial identities that do not fit within the broader Black or Asian racial groups. "Hispanic or Latino" includes individuals of Hispanic or Latino origin, regardless of race. Median household income data for individual census blocks are derived from the encompassing block group, and poverty status (i.e., population living below the federal poverty level) is sourced from the encompassing census tract.
|
6787f5eb81d2151a024c0f07
| 21 |
To assess exposure equity, we classify each census block into ten emission density exposure groups (deciles) for each pollutant. These deciles range from blocks with the lowest emission density (decile 1) to those with the highest (decile 10), each containing an equal number of census blocks. For each decile, we calculate the proportion of the population by race, ethnicity, and poverty level, along with the population-weighted average median household income. This comparative analysis allows us to examine disparities in exposure to vehicle emissions across different demographic groups.
|
6787f5eb81d2151a024c0f07
| 22 |
To evaluate the direction, strength, and statistical significance of the relationship between emission density (exposure) and demographic factors such as the percentage of the population that is non-White and median household income, we develop ordinary least squares (OLS) regression models. These models are stratified by U.S. state/county, vehicle type (All, LDV, MDV, and HDV), and pollutant (NOX and PM2.5). We estimated 50,162 models for each county in the U.S., excluding several very rural counties with too few data points. The mean R 2 values were 0.028 (sd = 0.052) for NOX emission density and 0.029 (sd = 0.053) for PM2.5 emission density, indicating the variability explained by the models in predicting emission density based on demographic predictors across different geographic and pollutant-specific contexts.
|
6787f5eb81d2151a024c0f07
| 23 |
Figure illustrates the geographic distribution of NOX and PM2.5 census block emission densities across the United States in 2020, categorized by type of vehicle traffic. The data reveal substantial variations in average daily vehicle emissions nationwide, particularly highlighting significant emissions from truck traffic in certain regions. Specifically, the map delineates distinct patterns in NOX and PM2.5 emissions originating from LDV, MDV, and HDV traffic across the U.S. Emissions from LDVs are concentrated in urban areas, reflecting higher population densities and greater vehicle usage. Conversely, MDV and HDV emissions are more dispersed, prominently extending along interstate highways and arterial roads outside urban centers. These diverse emission patterns underscore the necessity of separately evaluating LDV, MDV, and HDV traffic emissions due to both their differing emission rates and spatial distributions. To provide a comprehensive assessment, we calculate the population-weighted mean average daily census block emission density, referred to as average emissions exposure, across various geographic scales, including counties and nationwide.
|
6787f5eb81d2151a024c0f07
| 24 |
At the national level, average exposure to NOX and PM2.5 vehicle emissions varies significantly across different types of vehicle traffic. On an average day, the average person is exposed to 102,000 g/km² of NOX from LDVs, 37,400 g/km² from MDVs, and 50,900 g/km² from HDVs (Figure and Table .3). Similarly, for PM2.5 emissions, the respective exposures are 3,010 g/km² from LDVs, 1,500 g/km² from MDVs, and 1,550 g/km² from HDVs (Figure S.2 and Table .3). These exposure values are depicted in Figure (for NOX) and Figure S.2 (for PM2.5), with detailed numerical data available in Table .3.
|
6787f5eb81d2151a024c0f07
| 25 |
Findings at the national level reveal that MDV and HDV traffic have a disproportionately large impact on emissions exposure relative to their traffic volume. For instance, despite MDVs and HDVs accounting for only 4.7% and 7.4% of total average daily VKT in the U.S., respectively, they contribute 19.6% and 26.7% of average NOX exposure, and 24.8% and 25.6% of average PM2.5 exposure (Table ). This indicates that emissions from MDV and HDV traffic collectively contribute to approximately half of the total vehicle emissions exposure in the U.S.
|
6787f5eb81d2151a024c0f07
| 26 |
Furthermore, MDVs and HDVs are responsible for a large proportion of PM2.5 exposure from exhaust emissions, accounting for 63.1% (30.4% + 32.7%) of the average daily exposure (Table ). Similarly, they are responsible for a significant proportion of PM2.5 exposure from nonexhaust emissions, accounting for 21.4% (12.7% + 8.7%) from brake wear and 15.2% (7.2% + 8.0%) from tire wear of the average daily exposure Table . These results highlight the significant role both exhaust and non-exhaust emissions play in overall air quality, particularly for MDVs and HDVs. At the county level, there is considerable variation in emissions exposure based on vehicle type and regional characteristics. While LDVs generally contribute more significantly to emissions exposure compared to MDVs and HDVs, certain counties-particularly those along interstate highways and major arterial roads-show a higher percentage of emissions from MDVs and HDVs (Figure ). Notably, counties in states such as Utah, Oregon, Oklahoma, and South Carolina exhibit a considerably larger proportion of emissions from MDVs. In some areas, emissions exposure from MDVs is more than double that from LDVs. This can be attributed to the high volume of MDVs on long-distance freight corridors and the relatively lower urban density in these regions. Similarly, HDVs contribute disproportionately to emissions exposure in many counties in the central U.S., further emphasizing the need to evaluate emissions at the county level and consider the role of specific vehicle types in shaping regional air quality. Additionally, we examine the relative contributions of different emission processes to PM2.5 exposure (See Table ). HDVs and MDVs are overwhelmingly responsible for PM2.5 exposure from exhaust emissions, with 90.2% of the total daily PM2.5 exposure from HDVs and 86.7% from MDVs. LDVs also contribute significantly to exhaust emissions, but to a lesser extent (52.5%), with a larger share of their PM2.5 exposure coming from non-exhaust sources like brake and tire wear. Brake wear is a major contributor to PM2.5 exposure for LDVs, accounting for 34.4% of their total daily exposure. This indicates that for LDVs, non-exhaust emissions (particularly from braking) play a larger role compared to MDVs and HDVs, where brake wear is a relatively smaller contributor (11.1% for MDVs and 7.4% for HDVs). Tire wear is a minor source of PM2.5 exposure across all vehicle types, contributing a smaller share compared to exhaust and brake wear (13.1% for LDVs, 2.2% for MDVs, and 2.4% for HDVs).
|
6787f5eb81d2151a024c0f07
| 27 |
These findings further underscore the importance of addressing both exhaust and non-exhaust emissions when formulating policies to mitigate vehicle-related air pollution, particularly for HDVs and MDVs, which contribute significantly to overall exposure. Note. LDV = light-duty vehicle; MDV = medium-duty vehicle; HDV = heavy-duty vehicle. Average PM2.5 exposure represents the sum of PM2.5 exposure from exhaust, brake wear particulate, and tire ware particulate.
|
6787f5eb81d2151a024c0f07
| 28 |
Significant disparities in emissions exposure are observed across different population groups (Figure ). People of color and individuals living in poverty face disproportionately higher levels of traffic-related air pollution. On average, people of color are exposed to approximately twice the NOX and PM2.5 emissions from all vehicle types compared to White individuals. This trend is consistent across all non-White populations examined. Specifically, Asian individuals experience the highest exposure to emissions from LDVs and MDVs, while Black individuals face the greatest exposure to emissions from HDVs. Individuals living in poverty are also exposed to higher emissions levels, with roughly twice the NOX and PM2.5 emissions from LDV and MDV traffic compared to the general population. This disparity is particularly pronounced for HDV emissions, with individuals in poverty experiencing 3.4 times the NOX and 3.5 times the PM2.5 emissions compared to wealthier individuals. These findings highlight the urgent need for targeted interventions aimed at reducing emissions exposure in vulnerable communities, addressing environmental justice concerns, and improving air quality for disadvantaged populations. .4. The differences in emission densities across the deciles are striking, with several orders of magnitude separating the median emission density levels between the lowest and highest deciles, as detailed in Table .4. Next, we examine the relationships between emission density, race, and income at the countylevel. Figure displays regression slope coefficients from county-level linear regression models, illustrating NOX emission density as a function of the non-white population share and median household income. The slope coefficients indicate the change in emission density (g/km 2 ) for every 1% increase in the non-white population share or $1 increase in median household income.
|
6787f5eb81d2151a024c0f07
| 29 |
For instance, the deepest red shade in Figure .A signifies that a 1% rise in the non-white population share correlates with an increase in NOX emission density exceeding 1,700 g/km 2 . Conversely, the darkest blue shade in Figure .D indicates that a $1 increment in median household income is associated with a reduction in NOX emission density by at least 2.3 g/km 2 . This translates to a decrease of at least 2,300 g/km 2 for every $1,000 rise in median household income. Counties with statistically insignificant slope estimates (p > 0.05, based on a 95% confidence interval) are hatched in Figure and excluded from the analysis. For the corresponding figure on PM2.5 emissions, refer to Figure S.4.
|
6787f5eb81d2151a024c0f07
| 30 |
We find that 78.8% of counties exhibit a statistically significant relationship between total NOX emission density and the percentage of the population that is non-white (LDV 79.6%, MDV 78.9%, and HDV 74.2%). Similarly, 77.7% of counties demonstrate a statistically significant relationship between total NOX emission density and median household income (LDV 79.4%, MDV 79.2%, and HDV 71.8%). Similar patterns are observed for PM2.5 emission density, nonwhite population, and median household income. Furthermore, our analysis reveals a pervasive presence of higher TRAP emissions in areas where people of color and lower-income individuals reside. Figure illustrates that people of color and lower-income groups are disproportionately represented in regions with elevated emission density across the entire U.S., encompassing both urban and rural areas. This indicates that people of color and those with lower incomes face heightened exposure to NOX and PM2.5 vehicle traffic emissions across the nation.
|
6787f5eb81d2151a024c0f07
| 31 |
In this study, we introduce a refined method for assessing near-roadway emission exposure, termed emission density, focusing on NOX and PM2.5 at the census block level across the U.S. in 2020. Our findings reveal significant disparities in air pollution exposure by race, ethnicity, and socioeconomic status, as well as across different types of vehicle traffic. Specifically, people of color experience approximately twice the exposure to traffic emissions compared to White individuals, while lower-income groups face exposure levels two to four times higher than the national average. This disparity is pervasive nationwide, highlighting a substantial public health and environmental justice issue beyond urban areas.
|
6787f5eb81d2151a024c0f07
| 32 |
Moreover, our analysis identifies medium-and heavy-duty vehicles (MDV and HDV) as substantial contributors to traffic-related air pollution, despite their relatively small share of total traffic volume. Notably, emissions from brake and tire wear, particularly from LDV traffic, emerge as significant sources of PM2.5 exposure. These findings underscore the need for targeted strategies to mitigate emissions from MDV and HDV traffic, potentially through electrification of vehicle fleets, though this poses challenges that require careful consideration. While electrification offers promise for improving near-roadway air quality, our results emphasize the urgency of broader efforts to reduce emissions and address the underlying reasons why communities of color and lower-income populations are disproportionately exposed to high levels of vehicle traffic emissions. These efforts should encompass comprehensive approaches beyond vehicle electrification, including demand reduction strategies and addressing systemic factors contributing to disparate exposure patterns.
|
6787f5eb81d2151a024c0f07
| 33 |
Looking ahead, future research directions include forecasting future exposure levels, scenario analyses such as electrification and demand reduction simulations, and quantifying health impacts associated with traffic-related air pollution. However, our study acknowledges several limitations, including the exclusion of emissions from roadways not covered by the HPMS dataset, which includes residential streets, rural roads, and potentially high-traffic routes like those leading to warehouses and ports. Additionally, our exposure estimates do not consider secondary emissions formed in the atmosphere, which pose broader regional air quality challenges. Addressing these complexities and limitations will be crucial for advancing effective policies and interventions aimed at improving air quality and achieving environmental justice in communities across the United States.
|
6540c15748dad23120c35c07
| 0 |
Substoichiometric tungsten oxides WO3-x, especially nanoscopic forms of W18O49, are promising materials due to their many exciting properties, resulting in potential applications in various technologies, like electrochromic devices, sensors, etc. W18O49 exhibits near-infrared absorption, which could be exploited for heat shielding, heat generation, and water evaporation closely connected to recently demonstrated water desalinization. The tungsten suboxides could also be used in the photoreduction of carbon dioxide, photocatalysis, electrocatalysis, and photoluminescence. The tungsten suboxides' characteristics, including their distorted inner structure, their highly oxygen-deficient structure, and their nanoscopic size, were examined in many studies. The structure containing triangular and hexagonal channels allows W18O49 to act as an anodic material for lithium-ion batteries. Intercalation of lithium ions into the W18O49 structure was studied for triggered lattice contraction, which result in macroscopic material flexibility and electrochromic color change. W18O49 was successfully exploited as a precursor in the large-scale synthesis of WS2 nanotubes. Another feature of this material is its sensory properties for NOx and ammonia gases. There are several different approaches to preparing nanoscopic W18O49 and other tungsten suboxides in nanowhisker or nanofibrous forms that differ from each other in a synthetic method or in a different tungsten-containing precursor. The most frequently used method is based on the solvothermal reaction of WCl6 or W(CO)6 with aliphatic alcohols in an autoclave or just at elevated temperature in high boiling point solvents. This route results in uniform nanowhiskers of various lengths and thicknesses. Other possible preparation procedures involve controlled oxidation of metallic tungsten with water vapor or flame synthesis based on the deposition of tungsten oxides on various surfaces by heating a tungsten mesh with a high-temperature heater. Tungsten disulfide WS2 could also be used as a tungsten source for these nanofibers using the high-temperature, low-pressure reactor in the presence of water vapor. This study was conducted in a chamber of an electron microscope with a heating extension serving as a sample holder. A similar study has been done based on the oxidation of tungsten filament to substoichiometric tungsten oxide whiskers with characterization and material properties testing performed entirely in-situ in the electron microscope chamber.
|
6540c15748dad23120c35c07
| 1 |
Electrospinning is a versatile and accessible method for the production of submicron and nanoscopic fibers from organic polymers and inorganic compounds with a valuable extension to the industrial production of many high-end materials and products. There are suitable routes for preparing various inorganic oxides, carbides, sulfides, or even metallic nanofibers based on multi-step processes involving electrospinning. For electrospinning of the inorganic nanofibers, appropriate solutions must be prepared first. Usually, the inorganic precursors containing desired elements are dissolved with an organic, supporting polymer in a suitable solvent. The prepared solution is electrospuned into the so-called green composite nanofiber web. The following step is a high-temperature calcination in air, which removes the organic part and obtains pure inorganic oxide fibers. Nanofibers containing the W18O49 phase were prepared in the past by a combination of electrospinning of suitable substrate and solvothermal synthesis of desired W18O49 nanowhiskers from WCl6. So far, materials based on carbon, TiO2, and polyacrylonitrile nanofibers have been exploited as substrates for decoration by W18O49 nanowhiskers.
|
6540c15748dad23120c35c07
| 2 |
Recently, we have described a multigram preparation of tungsten trioxide/amorphous silica (for brevity -WO3/a-SiO2) fibers, which served as a precursor material for the preparation of polycrystalline tungsten metal nanofibers and WS2 microfibers. In both cases, amorphous SiO2 acted as a binder of the material nanograins. During the reduction of WO3 fibers to metallic tungsten, several reduction processes have been observed with varying compositions and morphologies.
|
6540c15748dad23120c35c07
| 3 |
To further this study, we have focused in the present work on the growth of the W18O49 nanowhiskers on the silica fibers using a new technique, i.e., the Reactor, which is an in-situ reactor in the scanning electron microscope (SEM) utilizing microelectromechanical systems (MEMS) heating chip (referred in the text as MEMS chip or MEMS heating chip) reaching also to in-situ transmission electron microscopy (TEM). This reactor was briefly described in a previous study. Here, the new Reactor and the growth process, i.e., the in-situ reduction of the -WO3/a-SiO2 fibers in the SEM and TEM, are described in great detail.
|
6540c15748dad23120c35c07
| 4 |
The growth of W18O49 nanowhiskers from heated and partially oxidized tungsten filament in TEM was described by Hashimoto in 1960. The mechanism of the process was profoundly studied by Zhang et al. , who described mainly the mass transport during the oxidative W18O49 nanowhisker growth from tungsten filament. Here, the reaction was carried out in an environmental transmission electron microscope (ETEM). Alternatively, a direct reduction of tungsten oxide was observed in-situ in TEM , similarly in references. These results found that W18O49 nanowhiskers grow in an anisotropic fashion involving volatile tungsten oxides from oxidized metallic tungsten or by e-beam reduction of tungsten oxide.
|
6540c15748dad23120c35c07
| 5 |
Following the detailed structural changes occurring during the W18O49 nanowhisker growth using in-situ experiments in the TEM, furnished with the same MEMS heating chip previously used in the SEM. Moreover, fundamental insight into the structure of the nanowhiskers and their characteristics are provided by analysis of nanowhisker cross-section using advanced microscopy techniques in synergy with theoretical calculations gained through the DTF method.
|
6540c15748dad23120c35c07
| 6 |
Polyvinyl alcohol (PVA, Mowiol 18-88) and silicotungstic acid hydrate H4SiW12O40 (HSiW, purum) were purchased from Merck and used as received. Deionized water was used as a solvent. For electrospinning, a Nanospider NS LAB500S instrument (Elmarco, Czech Republic) equipped with a cylindrical electrode with micro blades for allocation of the electrical charge and solution droplets was used (shown schematically in Figure ). Electrospinning preparation of green composite fibers consisting of PVA and HSiW was described in a previous study. The experimental procedure used here is described in Supplementary Experimental
|
6540c15748dad23120c35c07
| 7 |
The green fibers of PVA and HSiW were calcined in air at 600 °C in a muffle furnace. The furnace was heated within 1 h to the final temperature followed by 2 h dwell time. After heat treatment, the sample was left to cool down spontaneously to ambient temperature. The prepared -WO3/a-SiO2 nanofibrous material was analyzed by SEM and X-ray powder diffraction (XRD) (Supplementary Experimental Part 1).
|
6540c15748dad23120c35c07
| 8 |
The electrospun nanofibrous materials were characterized by SEM with a Versa 3D (Thermo Fischer Scientific) microscope and by scanning transmission electron microscopy (STEM) on an FEI Magellan 400 XHR microscope (Thermo Fischer Scientific). A Helios UC Focused-Ion-Beam (FIB)-SEM system (Thermo Fisher Scientific) was used for the in-situ SEM experiments inthe Reactor and lamella preparation from a W18O49 nanowhisker. For the lamella preparation, W18O49 nanowhiskers were dispersed in isopropyl alcohol, dropcasted on a Si substrate. A Helios 5 FX FIB microscope was used for localization of a specific nanowhisker, micromanipulation and final lamella preparation. A gas injection system (GIS) was utilized for creating of an amorphous carbon protection layer. First deposition step was performed by electron assisted deposition and later ion assisted deposition was used. Chunk thinning and final polishing operations were performed at FIB accelerating voltages ranging from 30 kV to 2 kV and FIB currents ranging from 2 nA to 25 pA.The SEM micrographs were analyzed by the ImageJ software to determine the fiber diameters and the size distribution.
|
6540c15748dad23120c35c07
| 9 |
Samples were drop-cast onto copper grids with a lacey carbon support film after sonicating the sample suspension in methanol. Energy dispersive X-ray spectroscopy (EDS) data were measured on a Thermo Fisher Scientific Talos F200i equipped with a Bruker Dual-X spectrometer, operated in the STEM regime at a high voltage of 200 kV and beam current of 0.5 nA. Spectra and images were post-processed by the Velox software.
|
6540c15748dad23120c35c07
| 10 |
Electron energy loss spectra (EELS) were measured with a CEFID (Ceos GmbH, Heidelberg, Germany) spectrometer on a double aberration-corrected Themis-Z microscope (Thermo Fisher Scientific Electron Microscopy Solutions, Hillsboro, USA, (TFS)) at an accelerating voltage of 200 kV. The EELS data was recorded in STEM mode at a beam current of 60 pA, a semi-convergence angle of 21 mrad, and a semi-collection angle of 60 mrad, using an ELA direct detection camera (Dectris AG, Baden, Switzerland). Radiation damage to the tungsten oxide was minimized by repeated rapid frame scanning in a focus frame while accumulating the energy-loss spectra. The fine calibration of the spectrometer's energy scale was performed against standards (Si, a-Al2O3, NiO, BN, C) and respective XANES reference data reported in the NIMS data base.
|
6540c15748dad23120c35c07
| 11 |
Scanning electron diffraction data were recorded with the CEFID on the ELA detector in zero-loss filtered mode. An electron probe with a convergence angle of 0.2 mrad was adjusted in STEM microprobe mode and further defocussed to reduce the electron flux by enlarging the probe size to about 10 nm. A primary beam current of less than 5 pA was used.
|
6540c15748dad23120c35c07
| 12 |
TheReactor is based on an SEM stage containing a MEMS chip with a micro-heating plate covered by a cap containing a pressure-limiting aperture and a gas inlet (Figure ). By covering the stage with the cap, theReactor body is closed permitting locally increased gas pressure (up to 500 Pa) of various inlet gases. A sample deposited on the micro-heating plate in the Reactor is heated by the Joule effect (up to 1200 °C with a maximal heating rate of 4.10 4 K s - 1 ). Very fast sample heating and cooling allows for a precise control of the reaction setup and kinetic measurements. The heated specimen eventually reacts with the gaseous atmosphere.
|
6540c15748dad23120c35c07
| 13 |
The reaction is continuously monitored using the signal of backscattered electrons (BSE) or secondary electrons (SE), which are detected inside the SEM column. Obviously, the electrons of the beam can intervene in the reaction directly and indirectly by heating the specimen. the electron beam (e-beam), the manipulator needle is retracted, and the cap of the µReactor is inserted (c). The reaction volume is thus closed and sealed. Overpressure inside the reactor (up to 500 Pa) is assured by the pressurelimiting aperture in the cap viewed by the SEM from the top (white area in (e)). Note that the SEM chamber remains under high vacuum conditions. In-situ SEM imaging is possible through the hole in the aperture (e). The gas inlet is incorporated into the cap. Gas escapes from the µReactor mainly through the aperture. The reactor can be opened/closed inside the SEM chamber without a need for chamber venting, which assures clean sample preparation (by FIB) and sample positioning (by the manipulator EasyLift needle) without exposure to air. The sample deposited on the heating segment of the MEMS chip reacts with the admitted gas or under vacuum at elevated temperatures.
|
6540c15748dad23120c35c07
| 14 |
A Helios UC FIB-SEM system (Thermo Fisher Scientific) with the µReactor was used for the in-situ SEM imaging of the growth of nanowhiskers (shown schematically in Figure ). The Reactor is described in detail in Figure . It allows forming of a local overpressure of the processing gas (up to 500 Pa) when placed inside an SEM chamber, while the SEM is operated in a standard high-vacuum mode (with a chamber pressure < 10 2 Pa). The sample heating inside the µReactor is provided by a MEMS-based micro-heating plate allowing maximum temperature up to 1200 °C (Figure ). The temperature of the heating plate is controlled from the SEM user interface. The processed samples were placed manually on the micro-heating plate (Figure ) by tweezers and shifted to the desired position using an EasyLift manipulator needle. Once the sample was placed on the heating plate and aligned under the electron beam, the cap of the µReactor was inserted to seal the reaction volume (Figure ). In-situ imaging was done by an in-lens (TLD) detector collecting signals of secondary (SE) and backscattered (BSE) electrons passing through an aperture hole in the reactor cap (Figure ). Pure hydrogen was flowing into the reaction volume through a capillary connecting the µReactor cap (Figure ) with a gas feedthrough on the SEM chamber. The inlet flow rate was controlled by a mass flow controller placed outside the SEM chamber, between the hydrogen cylinder and the gas feedthrough of the SEM chamber.
|
6540c15748dad23120c35c07
| 15 |
The temperature of the heating plate was calculated from the resistance of the heated wire. The pressure inside the reaction volume was estimated from the heating power, which depends on the pressure of the surrounding gas, employing the Pirani gauge principle. The advantage of this approach is that both temperature and pressure can be measured directly in the vicinity of the sample without the need for a bulky thermocouple or a large volume pressure gauge connected to the µReactor.
|
6540c15748dad23120c35c07
| 16 |
In another embodiment of the experiment (Experiment No. 2-see Sec. Supplementary Discussion 2 and its visualization in Video 2), the precursor (-WO3/a-SiO2 nanofibers) was heated in the µReactor without being exposed to hydrogen gas by heating the specimen under high vacuum. In this case, the cap of the reactor was retracted, and the reaction volume was opened to the SEM chamber as shown in Figure .
|
6540c15748dad23120c35c07
| 17 |
To prepare a sample for the heat treatment in the SEM µReactor, the calcined fibers in the form of flakes were placed on the MEMS heating chip and rubbed on the surface by a micromanipulator. Using this method, several individual nanofibers were deposited on the heating stage area in the center of the MEMS chip. For all SEM image acquisitions, the acceleration voltage was set to 10 kV and the current to 0.4 nA. After preparing of the sample, heating was set up to 700 °C (heating rate 2 K s 1 ). Once the sample reached this temperature, the SEM was found to have lost focus. In order to renew the analysis, the imaging setting was brought to focus again, and the temperature was raised to 800 °C or more. The raster scan speed was variably set to 20.5 or 41 s per image. After the reaction ended, the magnification was reduced to see the whole set of fibers on the MEMS chip.
|
6540c15748dad23120c35c07
| 18 |
Originally, the MEMS chip (Figure ) was fabricated and used for the in-situ TEM experiments. The TEM observation was conducted through a perforated thin amorphous silicon nitride membrane attached to the heating plate (Figures ). The MEMS chip with the deposited sample was placed into the Thermo Fischer Scientific NanoEx-i/v heating and biasing holder for in-situ STEM imaging and elemental analysis (EDS) at elevated temperatures (Figure ). The holder is similar to the standard single-tilt holder used for analysis of TEM grids, however, with a rectangular fitting for the MEMS chip and with contacts for an external power supply. In-situ experiments were performed on a Thermo Fisher Scientific Talos F200i microscope equipped with a Bruker Dual-X (EDS) spectrometer, operated in the TEM regime at the high voltage of 80 kV and beam current of 1 nA. The setup allows continual measurements with real-time video output of the process. The in-situ annealing was observed at 820 °C. The in-situ TEM experiment was acquired as real-time Video 3 and Video 4.
|
6540c15748dad23120c35c07
| 19 |
The same conditions used for obtaining the fibers in-situ in the SEM experiments were also tested in a custom-made tube furnace (Figure ). This furnace could optionally operate under vacuum or hydrogen atmosphere with a controllable pressure. The furnace allowed also shock heat treatment by moving the furnace, fixed on a rail towards the sample boat.
|
6540c15748dad23120c35c07
| 20 |
The prepared -WO3/a-SiO2 nanofibers (calcined at 600 °C) were placed in an alumina combustion boat and were covered with another one to prevent the material from spilling out during evacuation (see Figure ). The boat with the sample was placed into a quartz tube and initially evacuated to reach a high vacuum (1.10 4 Pa) followed by setting up a partial hydrogen atmosphere by allowing hydrogen from the gas cylinder into the quartz tube and partially closing the valve into the vacuum pump system. By careful optimization of the inlet and outlet valves, the pressure was set to a constant value of 100 ± 5 Pa. The furnace was preheated at 800 °C and after reaching the desired temperature, the furnace was moved to the region of the quartz tube with the sample. In this way, shock calcination was performed for 1 h, followed by spontaneous cool-down to ambient temperature, venting, and collecting the resulting material for structural and chemical evaluation.
|
6540c15748dad23120c35c07
| 21 |
The electrospinning setup (Nanospider) suitable for the multigram synthesis of W and WS2 fibers was described elsewhere. In the past, silicotungstic acid (HSiW) was used as a useful tungsten oxide precursor for electrospinning. The fabricated green nanofibers were calcined at 600 °C. Careful analysis showed that they consist of WO3 grains attached to an amorphous silica phase (for brevity, named -WO3/a-SiO2). The mean diameter of the -WO3/a-SiO2 fibers was 225 ± 88 nm. The preparation process of the precursor and material characterization are presented in Supplementary Experimental Part 1. The nanowhiskers have the tendency to form bundles during growth, which adversely affects their further processing into functional electronic devices or converting them into WS2 nanotubes of high crystalline order.
|
6540c15748dad23120c35c07
| 22 |
On the one hand, the reaction mechanism and the kinetics could be revealed by in-situ TEM experiments. Alternatively, experiments carried out here using in-situ SEM Reactor and under entirely different conditions permitted the visualization of the overall process as a function of the growth parameters. Here, the precise heating profile available by the MEMS chip and the hydrogen pressure tuning within the bell-jar-shaped µReactor made the tool ideal for optimization of the reduction reaction. Figure presents the µReactor with the MEMS chip within SEM used in the present study. Full details of the µReactor with its MEMS chip is provided in the Experimental section. It is important to realize that the pressure inside the Reactor is ~500 Pa (5 Torr), while the vacuum in the SEM chamber is 10 -2 Pa (10 -4 Torr). The following discussion describes the optimization of the W18O49 nanowhiskers growth from the -WO3/a-SiO2 nanofibers in a hydrogen atmosphere and under vacuum at elevated temperatures (up to 900°C) using the µReactor within the SEM. Of the several experiments carried out in the µReactor within the SEM, two are described in detail (Table in Experimental part). The two experiments differed in their reaction temperature and hydrogen pressure. The goal was to study the whole reduction process while observing the morphological changes. Presumably, the even surface formation was caused by tungsten oxide mass transport from the -WO3/a-SiO2 fiber's body into the W18O49 nanowhiskers. Consequently, nanowhiskers grown early-on seeded formation of neighboring whiskers further on during the reaction, thus collectively generating bundles (Figure treaction < 1054 s) on the surface contour of the -WO3/a-SiO2 nanofiber. The emergence of W18O49 nanowhiskers into bundles was reported previously. At this point, the original nanofiber was partially covered with multiple W18O49 nanowhisker forming a bundle, which protruded outside. In addition, surface coarsening of the e-beam irradiated nanofiber was observed (see the arrow in Figure at treaction = 1443 s). This surface coarsening of the nanofibers occurred on scars in the vicinity of the nanowhiskers. The nanowhiskers in the final stage of the reaction were shrunk and in the form of bundles.
|
6540c15748dad23120c35c07
| 23 |
Figure shows a fiber that reacted in the Reactor free of e-beam irradiation simultaneously with the irradiated one. The length of the nanowhiskers of the irradiated nanofiber (Figure ) are only a quarter of those which grew in the non-irradiated zone (Figure ). Statistical analysis of the nanowhiskers' length of the e-beam irradiated and non-irradiated shows a factor four difference between the length of the two families (391 ± 134 nm for the non-irradiated and 92 ± 36 nm for the irradiated W18O49 nanowhiskers). This intriguing finding is attributed to the interaction of the e-beam with the growing nanowhiskers. On the one hand, the e-beam has a strong chemical reducing nature, but on the other hand, the heating effect of the e-beam cannot be absolutely excluded. Indeed, the vapor pressure of tungsten dioxide, which may be formed, is appreciably smaller that of WO3 at elevated temperatures, which may explain the lower growth rate of the e-beam irradiated W18O49 nanowhiskers. Following Experiment No. 1, a kinetic study of the growth of WO3-x nanowhiskers (analyzed later as the W18O49 phase, see Figure ) on the surface of the e-beam-irradiated -WO3/a-SiO2 nanofiber in the Reactor within the SEM was undertaken. For a detailed discussion of the kinetic analysis see Supplementary Discussion 1. Experiment No. 2, dedicated to the partial reduction of -WO3/a-SiO2 nanofibers at 900 °C under vacuum (10 -3 Pa, i.e. 10 -5 Torr) is described in detail in Supplementary Discussion 2 and visualized in Video 2.
|
6540c15748dad23120c35c07
| 24 |
From the carried out experiments, it is evident that the µReactor offers significant advantages for in-situ studies of heterogeneous reactions and vacuum annealing. Its small volume facilitates clear observations under pressure, which enhances safety, preserves the microscope's technical condition, and optimizes gas consumption. Chemically speaking, the reductive growth of W18O49 nanowhiskers is more efficient when using hydrogen as a reducing agent. Compared to vacuum annealing, the nanowhiskers exhibit a higher aspect ratio, and the reaction allows for more controlled nanofibrous morphology. In both scenarios described above, the W18O49 nanowhiskers formed bundled structures protruding from the nanofibrous backbone. In the follow-up experiment the W18O49 nanowhiskers were produced and visualized via in-situ W18O49 growth experiments in the TEM using MEMS chip for sample heating and direct observation through a SiN membrane.
|
6540c15748dad23120c35c07
| 25 |
The shear plane structure of tungsten suboxides is well-documented. The rearrangement of the corner-shared [WO6] octahedra array in WO3 (Figure ) produces uniform crystal planes with edge-sharing [WO6] octahedra. In the case of the W18O49 nanowhiskers, a unique pattern of pentagonal columns consisting of a [WO7] central cluster is surrounded by five [WO6] octahedra forming hexagonal channels (Figure ). Experiment No. 2 in the Reactor showed the growth of W18O49 nanowhiskers from the-WO3/a-SiO2 nanofibers under vacuum (1.10 4 Pa) at 900 °C (see Supplementary Discussion 2). The same MEMS chip used in the Reactor can be also used for in-situ TEM experiments utilizing the NanoEx-i/v holder. Therefore, the in-situ TEM (under high vacuum In-situ observations revealed discrepancies and misalignments in the shear planes during the growth, suggesting the complex structure of the W18O49 nanowhiskers. As a result, a crosssectioned lamella of a W18O49 nanowhisker was produced using focused ion beam lift-out procedure. This cross-section underwent analysis with high-resolution scanning transmission electron microscopy in high-angle annular dark field (HRSTEM-HAADF) mode, as displayed in Fig. -d and Fig. (which offers a comprehensive view of the lamella). Notably, the nanowhisker's structure was heterogeneous, encompassing multiple phases.
|
6540c15748dad23120c35c07
| 26 |
Previous studies have examined the bundled structure of W18O49 nanowhiskers with diameters exceeding 80 nm, noting that the space between individual domains was amorphous. However, in contrast to these findings , every region of the nanowhisker cross-section presented in this study (as seen in Fig. and) was distinctly crystalline and consistently oriented along the [010] axis. This cross-section (Fig. ) predominantly features the W18O49 structure, interwoven with other domains constituted of related W-O phases, specifically the structural motifs of fully oxidized WO3 (Fig. ), W20O58 (Fig. ), and W12O34 (Fig. ). The W18O49 lattice (Fig. ) embodies two quintessential structural motifs: hexagonal channels formed by six [WO6] octahedra and pentagonal columns assembled from a central pentagonal bipyramidal polyhedron [WO7], encircled by five [WO6] octahedra. These motifs in Fig. are highlighted with a blue hexagon and a yellow pentagon, respectively. The emergence of the hexagonal channel can be perceived as a compensation for the lattice distortion. This rearrangement stems from the formation of pentagonal column through the removal of oxygen from the WO3 lattice. This structural rearrangement produces a relatively stable phase compared to other Mágneli tungsten suboxide phases. Importantly, two edge-linked pentagonal columns found in the W18O49 are the structural element motif containing the tungsten in the oxidation state 5+. Contra intuitively, the W 5+ are not the ones in the pentagonal bipyramidal coordination but the central atoms in octahedra sharing edges. These W 5+ -W 5+ pairs are responsible for unique absorption properties in the near infrared regions and other electronic features, like bipolarons. Next to the W18O49 phase, a significant portion of the analyzed section is composed of a fully oxidized array of [WO6] octahedra, forming a WO3 lattice (Fig. ). Between these WO3 regions, other structural motifs are observed, notably a lattice resembling the W20O58 phase (Fig. ). In analogy to the W20O58 lattice, which consists of a repeating unit of three closely packed pairs of [WO6] octahedra and a distorted hexagonal channel, the observed motif features an extended pair line by an additional octahedral pair (indicated by a yellow stretched irregular hexagon). The shear plane stacking is counterbalanced by pairs of distorted hexagons (marked in blue), which are also characteristic of the W20O58 phase. Such W20O58 resembling structural motif is present in various places over the analyzed cross-section differing by size and orientation. Intriguingly, another arrangement of pentagonal columns was identified (Fig. ),
|
6540c15748dad23120c35c07
| 27 |
Further analysis of the multiphase area in the cross-section uncovers segments of other suboxides (e.g., W5O14) and variations in pentagonal column stacking and various defects which are not discussed any further in the current work. Notably, the cross-sectional analysis may provide insight into the growth discrepancies observed during the in-situ nanowhisker growth within the TEM (Fig. ). The growing structures, misaligned with previous layers, might have contained different phases which were observed in Fig. .
|
6540c15748dad23120c35c07
| 28 |
Followingly, another lamella from different nanowhisker was analyzed locally by electron nanodiffraction (Fig. ) and electron energy loss spectroscopy (EELS) (Fig. ). Two locations on the lamella consisted of W18O49 and WO3 phases were probed in the direction of the nanowhisker growth along the <010> axis. Indeed, the W18O49 lattice displayed in Figure and discussed below is appreciably more complex than the WO3 monoclinic lattice.
|
6540c15748dad23120c35c07
| 29 |
Interestingly, the arrays of octahedra forming both WO3 and W18O49 phases are aligned along the same axis <101>║<103>, respectively. This is direct indication that the oxidized WO3 phase is grown alongside with the W18O49 phase and not oxidized afterwards. This deduction is in close match with the in-situ growth in TEM as shown in Figures 3 and corresponding Therefore, during the growth multiple phase (W18O49; WO3 and other suboxides) are formed.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.