id
stringlengths 24
24
| idx
int64 0
402
| paragraph
stringlengths 106
17.2k
|
|---|---|---|
60c753a5f96a001b412884da
| 11 |
Retrospective Identification of Epigenetic Targets. As a proof of concept on the practical applicability of the herein developed consensus model, we employed it in the retrospective identification of the epigenetic targets for two external and recently reported compounds (Figure ): (1) compound 17, an inhibitor of EP300 and CREBBP, as targets representative of the less populated compound datasets, and (2) compound 43a, an inhibitor of HDACs 1, 3, 6, 8 and BRD4, representing targets with the most populated compound datasets. These results
|
60c753a5f96a001b412884da
| 12 |
The full list of predictions is available as Tables S5-S15 in the Supporting Information. The number of targets predicted by each of the web tools is fixed, being 10 for HitPickV2, 20 for PPB2 and SEA, and 100 for SwissTargetPrediction, while the herein reported consensus model was re-fitted using the entire datasets and set up to perform the predictions for the 55 epigenetic targets, with only those involving an "active" outcome considered as the predicted targets.
|
60c753a5f96a001b412884da
| 13 |
predicted correctly the four HDACs but not BRD4, while SwissTargetPrediction predicted correctly only HDACs 1, 6 and 8. Although a more exhaustive external validation is needed, these results suggests that epigenetic targets with large amounts of chemogenomic data associated (such as HDACs) are generally well represented in current target prediction tools, while those with fewer data are not well covered. Moreover, it should be noted that the known epigenetic targets were not always among the top predictions for the available tools. For instance, HDACs 3, 6 and 8 were ranked 12, 13 and 15 by SEA, and HDACs 1, 6 and 8 from SwissTargetPrediction were ranked in positions 42, 43 and 44, so in a practical application, these targets would be hardly prioritized. Although the experimental validation of the predictions from all models would be needed to provide better means of comparison, these findings reinforce the potential usefulness of a tool focused on epigenetic targets for medicinal chemistry applications in drug discovery.
|
60c753a5f96a001b412884da
| 14 |
work is available free of charge at figshare repository (10.6084/m9.figshare.13519580). To encourage the medicinal chemistry community to apply the predictive consensus model developed in this work, the model was re-fitted using the entire datasets and has been implemented as a freely accessible and easy-to-use web application described in a separate work and available at .
|
60c753a5f96a001b412884da
| 15 |
Epigenetic drug discovery is increasingly important across different therapeutic areas. Despite the large amount of SAR data stored in public data sets, that information has not been used on a large scale to develop predictive models that support the medicinal chemistry community's efforts working on these cutting-edge targets. To fill this gap, in this study, we developed and evaluated the performance of five state-of-the-art machine learning algorithms built on three molecular fingerprints of different designs to predict 55 epigenetic targets of small molecules. To the best of our knowledge, this is the first study covering epigenetic targets on a large-scale basis. The performance of the herein reported models was validated using two different approaches, involving their performance estimation for binary classifications in 10-fold cross-validations in the context of each target, as well as the performance of their combination in the epigenetic target prediction task evaluated over 10 balanced samples of compounds containing an equal number of known active compounds for each target.
|
60c753a5f96a001b412884da
| 16 |
Although none of the herein reported models was identified as the best performing one for all the 55 targets, our results suggested Morgan and RDK fingerprints as the best representations for the derivation of binary classifiers for the studied targets, particularly when derived using SVM, where no significative difference was found for their performance. This cannot be generalized for other, or even for these targets, since it could be associated with the hyperparameter space employed to optimize the models. Moreover, a model's performance is also dependent on the dataset composition, so the trends herein presented could change as more bioactivity data is published and different sets of hyperparameters are studied.
|
60c753a5f96a001b412884da
| 17 |
A consensus model was built by combining the predictions of the best models derived from Morgan and RDK fingerprints (Morgan::SVM and RDK::SVM), also supported on the fact that predictions between models relying on the same fingerprint are more closely related than those relying on different representations as demonstrated by the hierarchical clustering analysis of their cross-validated predictions. The consensus models' performance and the two source models were analyzed on a DM basis, categorizing the predictions according to the Jaccard distance of the compounds in the test set to those in the training set. For the single target binary classification, the consensus model showed a significantly higher precision for identifying active compounds than those obtained by the individual models regardless of the DM. This trend was preserved when the models were evaluated to predict epigenetic targets.
|
60c753a5f96a001b412884da
| 18 |
The consensus model showed a mean BA of 0.835 considering the cross-validated predictions of the 55 target-associated binary classifiers, with mean precisions for identifying active compounds ranging from 0.923 for those compounds closer to the training set, to 0.810 for those farther from the training set. For the epigenetic target prediction task, mean precisions ranged from 0.952 to 0.773 under the same scheme.
|
60c753a5f96a001b412884da
| 19 |
Data Sets. Our primary source of SAR data was ChEMBL 27, we collected all the quantitative compound-protein associations from single protein assays, related to the 136 epigenetic targets identified in our previous work (biological activity reported as IC 50 , EC 50 , K i or K d ). In the context of each target, compounds were labeled as "active" when they had unequivocally assigned activities lower than or equal to 10 µM, and as "inactive" in the opposite case. Compounds whose label could not be unequivocally assigned (e.g., activity < 100 µM or activity > 1 µM) were removed from the data set. The remaining compounds were curated using the open-source cheminformatics toolkit RDKit, version 2020.03.1 and the functions Standardizer, LargestFragmentChoser, Uncharger, Reionizer and
|
60c753a5f96a001b412884da
| 20 |
TautomerCanonicalizer implemented in the molecule validation and standardization tool MolVS, as described in previous works. In short, the Simplified Molecular Input Line Entry System 56 (SMILES) of each compound was standardized, those compounds consisting of multiple components were split and the largest component was retained. Compounds containing any element other than H, B, C, N, O, F, Si, P, S, Cl, Se, Br and I, as well as compounds with valence errors, were removed from the data set. The remaining compounds were neutralized and reionized to subsequently generate a canonical tautomer without preserved stereochemistry. Once all compounds were standardized, those with molecular weight higher than 800 Da as well as duplicated compounds with contradictory labels were removed. We preserved compound-protein associations only for those targets with at least 30 compounds labeled as "active," corresponding to 72 different targets. Since chemogenomic data for these epigenetic targets include a higher proportion of associations for "active" compounds (64% on average), we extended our initial data with "inactive" compounds from PubChem. We included only compounds with annotated quantitative data (IC 50 ), all these compounds were curated using the same procedure described above and added only if they were not already included in the datasets. Finally, we kept 58 target-associated datasets containing at least 30 compounds labeled as "inactive."
|
60c753a5f96a001b412884da
| 21 |
Molecular Representations. To develop the machine learning models, we selected three molecular fingerprints of different design: (a) Molecular ACCess System (MACCS) Keys (166bit) as a dictionary based fingerprint where each position indicates presence or absence of a predefined structure, (b) Morgan fingerprint with radius 2 (2048-bit) Data Modelability. To a priori estimate the feasibility to obtain predictive binary classification models for each target, we calculated the modelability index (MODI) for each targetassociated dataset. MODI is defined as the proportion of compounds in a dataset for which its nearest neighbor belongs to the same class in a given feature space. For its calculation, we selected as compound representation the three different fingerprints described above and as metric to identify the nearest neighbors the Jaccard distance, defined as:
|
60c753a5f96a001b412884da
| 22 |
where J (A, B) is the Jaccard distance between compounds A and B in a given fingerprint representation, with a and b being the number of "on" bits for compound A and B, respectively, and c being the number of "on" bits for both compounds. Further modeling was performed only for 55 datasets with a MODI higher or equal than 0.7 for at least one molecular representation.
|
60c753a5f96a001b412884da
| 23 |
Machine Learning Methods. Binary classification models for each target were generated using five different machine learning algorithms: k-nearest neighbors(k-NN) , Random Forest (RF) , Gradient Boosting Trees(GBT) , Support Vector Machines(SVM) , and Feed-Forward Neural Networks (FFNN) . All machine learning methods were implemented using the Scikit-learn Python library (0.22.1). For model building, training instances were represented by a feature vector (fingerprint) and associated to a class label ("active" / "inactive"). To avoid hyperparameter bias when comparing different models, the hyperparameters for each model were optimized using stratified 10-fold cross-validation in an exhaustive search over a limited hyperparameter space. To keep the search space small, only selected hyperparameters on each algorithm were optimized. Hereunder, we provide brief explanations on each algorithm and the hyperparameters considered for its optimization; all hyperparameters not explicitly indicated in the text were set as default.
|
60c753a5f96a001b412884da
| 24 |
RF is one of the so-called ensemble methods relying on decision trees. In RF classification, a fixed number of decision tree classifiers are fitted on various bootstrapped subsamples of the training dataset. For a given sample, each decision tree predicts a label, and the final prediction of the sample is the label predicted by most of the trees. For this algorithm, the number of decision trees was fixed to 1000 and the number of features to consider when searching for the best splits in the individual trees was optimized in a representationdependent manner using candidate values of 1, 2, 3, 4 and 5 times the square root of the number of features in the fingerprint representation.
|
60c753a5f96a001b412884da
| 25 |
GBT is another ensemble method relying on decision trees. In this case, the decision tree classifiers are fitted in stages for the whole training dataset, where each subsequent tree is intended to "correct" the errors made by the previous one in terms of a loss function, usually the deviance of the fitted model with respect to a perfect model. For this algorithm, the number of decision trees was fixed to 1000, the number of features to consider when looking for the best splits in the individual trees was optimized in a representation-dependent manner using candidate values of 1, 2, 3, 4 and 5 times the square root of the number of features in the fingerprint representation, for the maximum depth of the individual trees we used candidate values of 4, 6, 8 and 10, and for the minimum number of samples to split an internal node in the individual trees we used candidate values of 2, 3, 4, and 5.
|
60c753a5f96a001b412884da
| 26 |
In SVM classification, the hyper-plane that best separate the two classes in the training dataset is constructed by maximizing the distance between training instances belonging to different classes (margin). As this hyper-plane does not always exist, a limited number of errors is allowed using a "cost" hyperparameter to control the relation between the training errors and the margin size. If linear separation of training classes is not possible in a given feature space, kernel functions are applied to project the data into a higher dimensional space where linear separation is possible. For this algorithm, "cost" was optimized using candidate values of 0.01, 0.1, 1.0, 10.0 and 100.0, and the kernel type to be used was selected from three options being non-kernel ("linear"), radial basis functions ("rbf"), and hyperbolic tangent ("sigmoid").
|
60c753a5f96a001b412884da
| 27 |
A FFNN is composed by different layers of computational neurons: an input layer, one or more hidden layers, and an output layer. Neurons in the input layer are associated to the features describing the data, each neuron in the hidden layer accepts the inputs of all neurons in the input layer and transform them to a weighted sum of the original inputs, then a nonlinear activation function is applied to this weighted sum and the result is passed to the neurons in the output layer, where the prediction is performed. The weights from the network are iteratively adjusted during the training stage on the basis of a cost function to minimize, typically cross entropy. For this algorithm, the solver for weight optimization was set as "lbgfs", the maximum number of iterations (how many times a training data point is passed to the network) was set to 1000, and the number of hidden layers was fixed to 1. The number of neurons in the hidden layer was optimized in a representation-dependent manner using candidate values of 0.1, 0.2, 0.3, 0.4, 0.5, 0.6 and 0.7 times the number of features in the fingerprint representation, and the activation function was selected from three different options, being logistic sigmoid function ("logistic"), hyperbolic tangent function ("tanh") and rectified linear unit function ("relu"). where TP means "true positives", TN "true negatives", FP "false positives", and FN "false negatives", with "positive" and "negative" refereeing to "active" and "inactive" compound labels, respectively.
|
60c753a5f96a001b412884da
| 28 |
We built a consensus model by combining the predictions of the best performing models showing the lower relation among their predictions. For that, we performed a hierarchical clustering with average linkage of the models relying on Morgan and RDK fingerprints (the best performing fingerprints), being described by their cross-validated predictions across all targets. As the distance metric for the construction of the hierarchical clustering, we selected de Jaccard distance defined in the Data Modelability section, where in this case J (A, B)
|
60c753a5f96a001b412884da
| 29 |
We compared the consensus model and the single models of which it is composed using precision (positive predictive value -PPV), sensitivity (true positive rate -TPR), negative predictive value (NPV), and specificity (true negative rate -TNR), defined as: In order to estimate the applicability domain of the models, these metrics were computed on a distance-to-model (DM) basis. For that, the mean Jaccard distance from each compound in the test sets to all compounds in the training sets was calculated as the DM metric, using the three different fingerprints employed as molecular representation. These average distances were categorized in four quartiles considering all the cross-validated predictions, and all four metrics were calculated for each target and quartile, when predictions on the corresponding quartile were available. compared for the epigenetic target prediction problem. To assess the global performance of the combination of the single-target binary classifiers in the epigenetic target prediction task, ten samples of compounds containing the same number of active compounds for each target were assembled. To reduce the target-bias associated to the different sizes on the targetassociated compound datasets, each of the compound samples was constructed by iteratively sampling one compound labeled as "active" from the less populated dataset in the sample (or in alphabetical order according to its gene code when there was more than one less populated sample). This process was performed until the sample contained exactly 6 active compounds (20% of the active compounds for the smaller dataset) for each target. If the addition of a compound yields a target containing more than 6 active compounds, the compound was discarded, and if the equal number of active compounds for each target was not satisfied after 1000 iterations, 10% of the sample was randomly discarded, and the iterative sampling continued. These ten samples were used as validation sets so that compounds in the sample were removed from the original target-associated datasets. The single target binary classifiers were refitted using the hyperparameters selected in the Single Target Validation strategy. The performance for the combination of the single-target binary classifiers was assessed by its capability of identifying the known active compounds among the known compound target associations, using PPV and TPR as metrics in the same DM basis described in the Single Target Validation strategy.
|
62569956742e9f1b495b88a2
| 0 |
Most chemical information is published in text and images in the primary scientific literature. The automated conversion of these unstructured, human-readable data formats into structured, machine-readable representations is essential to make the information available in publicly accessible databases. The reliable extraction of information from the depictions of the chemical structures is an ongoing challenge that still has not been fully solved yet. Chemical structure depictions are converted into computer-readable representations using optical chemical structure recognition (OCSR) systems .
|
62569956742e9f1b495b88a2
| 1 |
In order to evaluate the performance of the available OCSR tools, realistic benchmark datasets are necessary. At present, there are four real-world datasets available that contain chemical structure depictions that were collected and curated from publications and patents. The evaluation of the performance on realistic data is crucial to demonstrate whether the tools are robust enough to be used in an automated chemical literature mining process.
|
62569956742e9f1b495b88a2
| 2 |
The resolution of hand-drawn chemical structures is a more challenging task than the resolution of automatically generated depictions. In addition to the varying depiction features which are present anyway, the individual, unique way of drawing the structure adds an increased level of complexity. In 2021, the deep learning-based OCSR tool ChemPix demonstrated its capability to interpret simple hand-drawn hydrocarbon structures with high accuracy. There also are a few closed-source methods and commercial systems available that claim to be capable of resolving hand-drawn chemical structures . The authors of the deep-learning-based OCSR tool img2mol demonstrated the capability of their tool to recognise some hand-drawn chemical structures that they had picked themselves and noted the lack of a standardised benchmark set .
|
62569956742e9f1b495b88a2
| 3 |
Here we present DECIMER -Hand-drawn molecule images, a set of 5088 hand-drawn chemical structures depictions. Every image is mapped to a machine-readable representation of the underlying molecule. The diversely picked molecules represent a wide variety of small molecules. The dataset was created to facilitate the ongoing development in the field of OCSR and is openly accessible.
|
62569956742e9f1b495b88a2
| 4 |
The dataset consists of 5088 PNG images of unique hand-drawn chemical structures depictions (Figure ) which are mapped to their corresponding SMILES string as well as an SD file. The structures have been drawn by 24 volunteers from the Westphalian University of Applied Sciences, Campus Recklinghausen, Germany, who have graciously offered to use their free time to contribute to the generation of this dataset. The molecules have been picked from all structures in PubChem [23] using RDKit's implementation of the MaxMin algorithm based on Morgan fingerprints to ensure a diverse coverage of the chemical space. The only filtering rule that has been applied is a molecular weight maximum of 1500 Da. As a consequence, features like stereochemical information, charged groups as well as different types of isotopes are present in the dataset.
|
62569956742e9f1b495b88a2
| 5 |
In total, 6000 diverse molecules were selected from PubChem using RDKit's implementation of the MaxMin algorithm based on Morgan fingerprints. Subsequently, CDK Depict , a structure depiction generator based on the Chemistry Development Kit (CDK) , was used to create production-quality 2D images in batches. Each batch of images was then converted into PDF files and they were distributed among the volunteers. Using the chemical structures depictions generated by CDK as a visual template, each volunteer drew the structures on a piece of paper using a black or blue pen or on their tablet using an input device.
|
62569956742e9f1b495b88a2
| 6 |
Each volunteer sent back the scanned images or the images generated using their device after completing a batch. The curators reviewed the drawings, manually confirmed the correctness of the molecules, cropped the scanned images and stored them in separate image files. As part of the curation, structures that weren't correct due to human error were discarded. A total of 568 images out of 6000 were rejected due to issues with the depicted structure. Another 344 structures were not returned by the volunteers. This resulted in the final dataset of 5088 images in total.
|
62569956742e9f1b495b88a2
| 7 |
The following steps were taken in order to make the dataset findable, accessible, interoperable and reusable (FAIR) . The dataset was deposited in a publicly accessible data repository, in this case, Zenodo. This ensures that the dataset is easily findable. Furthermore, Zenodo provides a digital object identifier (DOI) that can be used to locate the dataset and it can also easily be integrated into Github as well. With Zenodo being an open, public repository, the dataset can be accessed from any part of the globe. To make it as interoperable as possible, the generated images use PNG as the final image format, which can be used across a variety of operating systems. Additionally, SMILES and SDF are representations of chemical structures which can be read by every cheminformatics toolkit. The dataset has been published under the CC-BY 4.0 licence. This licence includes that every user can redistribute or change the data as much as they want as long as they refer to the original authors when publishing results based on it. It is possible to use the data for non-commercial or commercial purposes without further obligations.
|
62569956742e9f1b495b88a2
| 8 |
No restrictions or limitations apply to using and reusing the dataset. Everyone can this dataset as a standard benchmark set for the evaluation of the performance of their OCSR tools. The dataset includes a wide range of chemical structures and represents a much larger chemical space. The structures were drawn by various individuals to ensure the diversity of drawing styles. The main limitation is caused by the molecular weight filter (< 1500 Da) as it excludes certain molecules like big macrocycles, proteins or artificial polymers. Additionally, markush structures are not represented.
|
65576851dbd7c8b54b745ce1
| 0 |
achieved through such methods. The cross coupling between two different redox active esters has also been reported but is not generally applicable to quaternary center synthesis. Here we show how a single catalyst can mediate regioselective quaternization of carbon using ubiquitous chemical feedstocks: olefins and carboxylic acids (Figure ). An inexpensive Fe-based catalyst in the presence of simple reducing agents can effectively achieve this powerful transformation in a chemoselective fashion with broad substrate scope through a similar mechanistic pathway. It is particularly striking that a single catalyst can both activate distinct functional groups (RAEs and olefins) and mediate bond formation between them. An illustrative demonstration of the power of such a reaction is depicted in Figure wherein the 10-step synthesis of a seemingly trivial structure (4) 10 can be truncated into a single step by coupling a primary tosylate-containing RAE 3 with either RAE 1 or olefin 2 with high chemoselectivity in 47% and 42% yield, respectively.
|
65576851dbd7c8b54b745ce1
| 1 |
The mechanistic framework outlined in Figure was inspired by key literature precedents outlined in Table . Seminal studies of Hirobe and co-workers showed that Drago-Mukaiyamatype olefin hydration could be achieved using Fe-porphyrins. Indeed, the pioneering studies of Setsune et al. identified sec-alkyl iron porphyrins formed upon exposure of 1-alkenes to Fe-TPPCl and NaBH4. Fe-H formation and metal hydride hydrogen atom transfer (MHAT) explains the branched-selective hydrometalation. Neta depicted alkyl-Fe 3+ (TPP) complexes as mesomers with single-electron density localized on carbon, which explained its ferrous phorphyrin-like behavior. Shenvi and co-workers recently utilized this same iron complex along with a silane to catalyze C-C bond formation between benzyl bromides and olefins to form quaternary centers via SH2 reactions. The Baran group demonstrated that Fe-catalysis could also enable decarboxylative Negishi couplings of RAEs and various aryl-organometallic reagents. Taken together, it seemed reasonable that RAEs might quaternize radicals through catalytic cycles in which Fe(II)porphyrins both form and trap open-shell carbon intermediates generated from olefins or other RAEs. These two related catalytic cycles would rely on overall reducing conditions (Zn or silane) to produce a pool of Fe(II) that might cleave N-hydroxy-phthalimide esters to generate primary and/or tertiary radicals. Due to the instability of tert-alkyliron complexes above 0 °C, 17 capture of the primary radical would be preferred, allowing its interception by the tertiary radical in a putative SH2 reaction. This same tertiary radical may be formed from an alkene with an iron hydride that can regenerate Fe(II) via MHAT or an HER reaction. In contrast to many MHAT reactions, this catalytic cycle would not require turnover by an exogenous oxidant; the RAEs partners would reoxidize Fe(II) to (III). Below we identify conditions to reduce this design to practice and achieve a versatile carbon quaternization that unites ubiquitous bench-stable starting materials (olefins and acids) in a simple and easily scalable fashion.
|
65576851dbd7c8b54b745ce1
| 2 |
The realization of this plan is summarized in Table using RAE 5 with either olefin 7 or RAE 8 leading to the same product 6 (see SI for the comprehensive optimization). As illustrated by entries 1-8,15 (olefin-RAE coupling) and 9-14 (RAE-RAE coupling), each reaction variant required simple and similar conditions: the Fe(TPP)Cl complex, a base, a reductant and a 1,2dichloromethane/ acetone solvent mixture. The choice of base proved crucial as exemplified in entries 2-3 with CsOAc emerging optimal. Increased yields are observed after a simple degassing (bubbling Ar) of the reaction mixture, although the presence of air is not detrimental (entry 4). Exclusion of acetone led to lower yields presumably due to catalyst insolubility (entry 5). Addition of other Fe-sources such as Fe(acac)3 led to diminished conversion (entry 6) as well. As observed in other MHAT transformations, the use of pre-formed Ruben's silane proved beneficial (entries 7-8). Turning to the RAE-RAE coupling, the optimized conditions above only led to 9% yield of the desired adduct 6 (entry 9). Increasing the Fe-loading to 20% had a measurable positive effect on conversion (entry 10). The most significant improvement emerged when switching the reductant from a silane to Zn metal (entry 11). A slight change in solvent composition to DCE/acetone 7:4 along with KOAc in place of CsOAc led to a further increase in yield (entries 12-13). The inclusion of a weak acid (Et3N•HCl) likely serves to activate Zn(0) and results in a 61% isolated yield of 6 (entry 14). Notably, the final optimized conditions for RAE-RAE coupling cannot be applied to the closely related olefin-RAE coupling (entry 15) and vice versa (entry 9). Nevertheless, both conditions employ a simple combination of commercially available reagents and the same Fe(TPP)Cl catalyst.
|
65576851dbd7c8b54b745ce1
| 3 |
Significantly, neither Fe-catalyzed decarboxylative coupling (conditions A or B) necessitated O2 to turn over the catalytic cycle (Fe 2+ to Fe 3+ ), whereas recent work from our group(s) required an air atmosphere in conjunction with an Fe β-diketonate MHAT catalyst or cocatalyst. In the current work, the single iron(II) porphyrin turns over to iron(III) in the absence of oxygen by reaction with the RAE or its resultant radical: 1 H NMR studies demonstrated that Fe(TPP) will complex n-pentyl-CO(NHPI) prior to formation of n-pentyl-Fe(TPP) (see SI). Fe(TPP) can be formed from its corresponding hydride by either hydrogen evolution or alkene MHAT (see SI). As a result, Fe(TPP)Cl mimics the polyfunctionality of precious metals in innersphere (coordinative) alkene functionalization catalysis, but does so in a unique outer-sphere (noncoordinative) series of elementary steps. In the case of RAE-RAE coupling, the use of Zn metal is well-precedented to reductively decompose NHPI esters into carbon centered radicals.
|
65576851dbd7c8b54b745ce1
| 4 |
The different optimal bases in conditions A/B were determined empirically and likely fulfill different roles: silanes can require Lewis base addition to increase hydridic strength and/or accelerate metal hydride formation 20 en route to MHAT with alkenes. Bases can presumably sequester Lewis acidic metals, e.g. ZnCl2, that might accrue as an RAE is reduced by Zn(0).
|
65576851dbd7c8b54b745ce1
| 5 |
With a streamlined set of conditions in hand for the radical quaternization of olefins and carboxylates, the scope was examined as illustrated in Table . In general, these highly chemoselective conditions tolerate a range of functional groups including carbamates amides, alkyl halides, epoxides, aryl halides, esters, alcohols, nitriles, tertiary amines, ketones, ureas, and electron-rich/deficient heterocycles. It is particularly noteworthy that alkyl bromides, terminal alkynes as well as an alkyl boronate, remain intact during this quartenization despite their notorious reactivity in both transition-metal catalyzed couplings and radical chemistry. Similarly, aryl iodides and electron-deficient aryl chlorides are left untouched despite their common use in radical cross couplings. In the case of olefins, preferential reactivity can be achieved during the RAE-RAE coupling whereas less reactive olefins are spared when employing the RAE-olefin coupling. Stabilized radicals can also be coupled (α-heteroatom, benzylic) derived from a-heteroatom containing acids. In the case of RAE-RAE couplings of such substrates, FeOEP led to a dramatically increased yield. Complex pharmaceutically relevant structures can be coupled efficiently resulting in quaternary center containing analogs 36-40. The direct installation of allcarbon quaternary centers within saturated heterocycles is a particularly challenging motif to construct by conventional polar bond analysis; this quaternization allows an intuitive coupling from readily accessible heterocyclic fragments (10, 11, 12, 15, 16). The reactions described above are extremely simple and practical to perform using inexpensive reagents, a rapid setup time, and are amenable to gram-scale preparation (see 41 and 34). The absence of oxygen in conditions A/B (Schlenk or glovebox conditions are not required, just a simple Ar balloon) allowed facile scaling to gram quantities with no change in yield, and pointed to internal oxidative turnover, i.e. an overall redox neutral process (see Table ). These features bode well for adoption in a parallel/combinatorial synthesis campaign wherein ubiquitous, bench-stable precursors can combine delivering long sought-after complex libraries rich in sp 3 content including quaternary carbons.
|
65576851dbd7c8b54b745ce1
| 6 |
The capacity for these new reactions to dramatically simplify the way quaternary-center containing molecules has been exemplified through 11 different case studies (see SI for complete listing), of which six examples are illustrated in Figure . Thematically what all of these examples have in common is a near exclusive reliance on polar bond analysis during the planning stages, especially carbonyl chemistry. As a consequence, the conventional routes to these mostly simple structures rely on extensive redox manipulations, function group interconversions, and a variety of pyrophoric reagents for various carbonyl reductions. In contrast, direct carbon quaternarization sidesteps these inefficient tactics through a modular LEGO-like transformations of commercially available building blocks (olefins and acids).
|
65576851dbd7c8b54b745ce1
| 7 |
For instance, the simple alkyne 53 was previously prepared (en route to prostaglandin analog synthesis) in an eight-step route of which only one step forges a C-C bond and wherein the key quaternary center was purchased in the form of a dimethyl cyclohexanone 52 whose skeleton was tediously edited. In contrast, commercially available olefin 54 can be coupled with the RAE derived from the simple commercial alkyne-containing acid (55) under radical quaternarization conditions to deliver 53 in only 2 steps (48% isolated yield). Ketoaldehyde 56, a useful building block for steroid analog synthesis, was previously accessed relying on classic carbonyl chemistry, C-C homologation, and a variety of undesirable reagents (Br2, Mg, OsO4). Alternatively, the same structure was accessed directly following radical quaternarization of olefin 57 with RAE 58 followed by acidic workup in 45% isolated yield. Substrate 60, containing two quaternary centers was previously accessed through a laborious 11-step sequence with low ideality commencing from glutaric anhydride 59 again relying on polar bond analysis and carbonyl chemistry. Remarkably, diene 61 could be enlisted in a sequential radical cross coupling first by employing olefin-olefin Fe-catalyzed MHAT cross coupling with methylacrylate (72% yield) followed by radical quaternarization of 62 with RAE 63 to deliver the same product 60 in 58% isolated yield obviating the need for 9 inconvenient steps and toxic/pyrophoric reagents (LiAlH4, HBr, KCN). Enyne 65, used as a substrate in a cycloisomerization study, was previously accessed through a 15-step sequence relying on alkyne hydrometallation, Wittig, carbonyl chemistry, and acetylide addition as the key C-C bond forming steps. This lengthy sequence required the use of protecting groups and extensive redox manipulations (3 instances of LiAlH4 reduction). A more intuitive LEGO-like approach was enabled through radical retrosynthesis. Thus, an Ag-Ni-facilitated decarboxylative alkenylation of commercial acid 66 with vinyl iodide 67, followed by radical quaternarization with alkyne-containing RAE 68 afforded 65 in only 5-steps without any redox manipulations. Radical quaternarization could also be used to construct scaffolds appearing in natural products. For example, piperidine 70, featuring a quaternary carbon at the C-3 position, serves as a pivotal scaffold within the madangamine alkaloids. In the previous 7-step synthesis, the construction of the quaternary center again relied on carbonyl chemistry commencing from 2piperidone. Of those seven steps, two forged C-C bonds with the remainder comprising concession steps (redox, FG, and PG manipulations). Conversely, starting from commercial acid 71, a simple alkylation followed by radical quaternarization with free alcohol containing RAE 70 enables a 3-step synthesis. Finally, α-tocopherol analog 74, useful for the modulation of microglial activation, required a 10 step route reliant on classic polar bond analysis. In a complete departure from this strategy, commercially available Trolox (75) could be used to efficiently incorporate the desired alkanol side chain through radical quaternarization (58% yield) with RAE 72 to deliver analog 74 in only 3 steps.
|
65576851dbd7c8b54b745ce1
| 8 |
Quaternary carbons can now be dissected into feedstock precursors, alkenes and carboxylic acids, via radical quaternization. Numerous high fraction sp 3 materials are now available in a fraction of the prior synthetic burden. Indeed, the approach outlined herein represents a complete departure from the mostly carbonyl-focused retrosynthetic analysis (Table ) historically utilized to forge such systems. In this simple protocol, two substrates classes converge into the same catalytic cycle in which an iron(III) catalyst undergoes reduction to iron(II) by either Zn or a silane/alkene combination, and oxidative capture by a redox-active ester returns the iron(III). Notably, the alkene cross-coupling involves iterative outer-sphere steps to form new C-C bonds using a single catalyst; the RAE cross-coupling exhibits perfect heteroselectivity. In contrast, inner sphere cross couplings that involve weak alkene-metal center complexes to form sp 3 -sp 3 linkages (e.g. Heck, Kulinkovich, 1,4-addition) often do not allow chemoselective merger of complex fragments. It has been widely recognized that fragment coupling simplifies retrosynthetic analysis by increasing convergency. Simple tools such as the reactions described herein point to a unique conceptual platform to decisively solve this vexing problem in organic synthesis. A sequence of FG/PG manupulations was inevitable due to the necessity of a carbonyl group to construct the 4º-carbon
|
63c92b365ab3137db9ad0e7f
| 0 |
Phosphorus, one of the Group V elements, has been known since 17 th century. Phosphorus crystallizes in a number of allotropic forms: white, red, violet, black, etc. Since 1980s the main interest has been attracted to the black phosphorus with orthorhombic (Cmca) structure (Fig. ) which was found to be the most stable allotrope. At room temperature and pressures up to 260 GPa black phosphorus exhibits a sequence of pressure-induced phase transitions: orthorhombic (Cmca)rhombohedral (R-3m) -simple cubic (Pm-3m) -simple hexagonal (P6/mmm) -bodycentered cubic . However, in spite of efforts devoted to high-pressure investigation of phosphorus at room temperature, there were only few high-pressure studies at high temperatures .
|
63c92b365ab3137db9ad0e7f
| 1 |
Based on available experimental data on orthorhombic-to-rhombohedral phase transition and melting curve of black phosphorus , a tentative phase diagram of phosphorus was proposed in 1987 (Fig. ). This phase diagram is limited by 5 GPa and 1000°C and describes only two solid phosphorus allotropes i.e. orthorhombic and rhombohedral. Later the first-order liquid-liquid phase transition in phosphorus has been discovered which was attributed to the pressure-induced transition of a dense molecular fluid (liquid 1) into a polymeric liquid (liquid 2) . The simplified version of the phase diagram of phosphorus up to 5 GPa has been proposed , however it differs from the previous one only by the presence of the liquid 1 -liquid 2 transition line with negative slope, without specifying the position of the liquid 1 -liquid 2 -orthorhombic solid triple point.
|
63c92b365ab3137db9ad0e7f
| 2 |
In the present work, the melting and phase transitions of three phosphorus allotropesorthorhombic, rhombohedral and simple cubic (corresponding crystal structures are presented in Fig. ) -have been studied up to 10 GPa, and the data obtained were used for thermodynamic calculations of the phase relations in phosphorus. The thermodynamic analysis based on our experimental and literature data allowed us to construct equilibrium p-T phase diagram of phosphorus at pressures up to 12 GPa.
|
63c92b365ab3137db9ad0e7f
| 3 |
Melting of phosphorus in the 2-8 GPa pressure range was studied in situ by electrical resistivity measurements in a toroid-type high-pressure apparatus with a specially designed high-temperature cell . The cell was pressure-calibrated using room-temperature phase transitions in Bi (2.55 and 7.7 GPa), PbSe (4.2 GPa), and PbTe (5.2 GPa). The temperature calibration under pressure was performed using well-established melting points of silicon, sodium and caesium chlorides, platinum and Ni-Mn-C ternary eutectic. A sample was placed directly into a graphite heater, and electrical resistivity of the cell was measured at different pressures upon stepwise heating up to the melting using the method described earlier . The onset of melting was detected in situ by an apparent drop in electrical resistance of the cell due to appearance of the conducting polymeric liquid phosphorus. In a special set of experiments it was found that phosphorus does not react with graphite and graphite-like hexagonal boron nitride (hBN) over the whole pressure -temperature range under study.
|
63c92b365ab3137db9ad0e7f
| 4 |
Melting of phosphorus at pressures 4-6 GPa have been in-situ studied by energy-dispersive synchrotron X-ray diffraction using MAX80 multianvil system at F2.1 beamline, DORIS III (DESY). Standard assemblies with hBN pressure medium were used. Sample pressure at different temperatures was determined from thermal equation of state of hBN ; temperature was measured by a Pt-10%Rh/Pt thermocouple. The experimental details are described elsewhere .
|
63c92b365ab3137db9ad0e7f
| 5 |
COMPRES assemblies with Re-foil or graphite heaters and hBN cylindrical capsules have been used. All assemblies were equipped with B-epoxy windows transparent to X-rays. The pressure and temperature have been either directly measured (thermocouples, pressure standards) or estimated from previously obtained calibration. The characteristic sequence of energy-dispersive X-ray diffraction patterns taken in the course of a melting experiment under pressure is presented in Fig. .
|
63c92b365ab3137db9ad0e7f
| 6 |
Melting of phosphorus in the 2.0-7.7 GPa pressure range has been studied by in situ electrical resistivity measurements (the results are shown in Fig. by red semi-filled stars). The experimental points at 2.0 and 2.4 GPa apparently reflect the melting of orthorhombic allotrope, while all others correspond to the melting of rhombohedral phosphorus. The melting curve of rhombohedral allotrope exhibits positive slope of 58 ( 2) K/GPa that points to a lower density of the melt as compared to the solid phase.
|
63c92b365ab3137db9ad0e7f
| 7 |
Melting of rhombohedral phosphorus was also studied by in-situ synchrotron X-ray diffraction using MAX80 (at 3.9 and 4.5 GPa) and SPEED-1500 (at 6.9 GPa) multianvil presses. The results are shown in Fig. (solid and open squares correspond to rhombohedral solid phase and liquid, respectively) and are in excellent agreement with the data obtained by the electrical resistivity measurements.
|
63c92b365ab3137db9ad0e7f
| 8 |
In experiments at pressures of 3.0 and 3.2 GPa we were able to detect solid-phase transition of orthorhombic phosphorus to rhombohedral allotrope by electrical resistivity measurements at a continuous temperature increase (T-scans); the results are represented by blue semi-filled diamonds in Fig. . These results are in excellent agreement with the data on this transition obtained earlier by Kikegawa et al. in p-scans (see Fig. ).
|
63c92b365ab3137db9ad0e7f
| 9 |
A quite different situation is observed for the phase transition of rhombohedral phosphorus to simple cubic allotrope. According to Kikegawa et al. , the pressure of this transition does not depend on temperature (at least up to 1100 K) and makes 10.3 GPa. However, in our experiments (T-scans at 8.9 and 9.4 GPa, and p-scan at 700 K; see Fig. ) we observed a significant dependence of the transition pressure on temperature, and the slope of this phase boundary was estimated as -420(70) K/GPa, which allowed us to located the rhombohedral-liquid 2-simple cubic triple point at about 8.3 GPa and 1490 K.
|
63c92b365ab3137db9ad0e7f
| 10 |
For orthorhombic, rhombohedral and simple cubic allotropes, the dependences of Gibbs energy on temperature and pressure were found based on the temperature dependence of Gibbs energy of white phosphorus in the form . Molar volumes (V 0 ), 300-K bulk moduli (B 0 ), their first pressure derivatives (B 0 ), and volume thermal expansion coefficients of solid phosphorus allotropes were taken from ; pressure dependencies of molar volumes were represented using the Murnaghan approximation .
|
63c92b365ab3137db9ad0e7f
| 11 |
Gibbs energy of polymeric liquid phosphorus (liquid 2) as well as molar volumes and bulk moduli of both liquid phosphorus phases were found by solving the inverse problem using literature data on melting of orthorhombic phosphorus . However, during the optimization process we used only slopes of two branches of the melting curve (Fig. ), and not the absolute temperature values because melting temperature of orthorhombic phosphorus at ambient pressure reported in (980 K) is overstated by 100 K compared to the generally accepted value of 883 K . The results obtained show that the low-pressure liquid phase (liquid 1) has 33% higher volume than the highpressure liquid phase (liquid 2) which is in agreement with results of ab initio quality molecular dynamics simulation of liquid phosphorus under pressure . Besides, bulk modulus of liquid 1 was found to be significantly lower than that of liquid 2 (see Table ).
|
63c92b365ab3137db9ad0e7f
| 12 |
The constructed p-T phase diagram of phosphorus is shown in Fig. . It should be noted that calculated melting curve of orthorhombic phase with appearance of low-pressure liquid phosphorus (liquid 1) and position of the liquid 1-orthorhombic-liquid 2 triple point (1.61 GPa, 1435 K) differ from the previously reported experimental data . At the same time, calculated melting temperature at ambient pressure corresponds to the generally accepted value for phosphorus , and the calculated melting curve of orthorhombic phase with appearance of high-pressure liquid phosphorus (liquid 2) perfectly agrees with the results of our in-situ experiments in the 2.0-2.6 GPa range.
|
63c92b365ab3137db9ad0e7f
| 13 |
In general, the proposed diagram accurately describes the observed growth of melting temperature of orthorhombic phosphorus up to 1.6 GPa and its decrease at higher pressures up to the orthorhombic-rhombohedral-liquid 2 triple point (2.65 GPa, 1153 K), as well as the stability of two liquid phosphorus phases and existence of the first-order phase transition between them in full accordance with previous publications . Also noteworthy is the fact that calculated melting line of simple cubic phosphorus with positive slope of 316 K/GPa is in good agreement with our experimental data at temperatures above the rhombohedral-liquid 2-simple cubic triple point (8.26 GPa, 1491 K).
|
63c92b365ab3137db9ad0e7f
| 14 |
Melting and solid-state phase transitions of phosphorus allotropes have been in situ studied in the 2-11 GPa pressure range using electrical resistivity measurements and synchrotron X-ray diffraction. While melting curves of rhombohedral and simple cubic allotropes exhibit positive slopes 58 ( 2) and 300 ( 90) K/GPa, respectively, the rhombohedral-simple cubic phase boundary has negative slope of -420(70) K/GPa), and the rhombohedral-polymeric liquid -simple cubic triple point is located at about 8.3 GPa and 1490 K. Calculations of phosphorus phase equilibria have been performed using our experimental and literature data in the framework of the
|
64fec0a8b338ec988a460e91
| 0 |
Nowadays, it is widely used for plasterboard and stucco manufacturing, construction, fire-proofing, biomedical applications, and even some forms of 3D printing. It is estimated that 100-200 million tonnes are used each year , making it one of the most widely consumed mineral resources. However, as there are no significant natural deposits of bassanite existing on the Earth's surface, it has to be synthesized. The main route to obtain hemihydrate is a thermal treatment of gypsum at temperatures typically between 150 °C and 200 °C to remove some of the structural water . Depending on the method used, αor β-hemihydrates are obtained either in humid/wet conditions or in dry air. Both types of hemihydrate will revert to gypsum when in contact with water, undergoing an exothermic hydration process that leads to the setting and hardening of the final gypsum-based material . α-and β-hemihydrates differ in reactivity due to different crystal morphologies and structural defects, but whether they have the same crystal structure is still debated . From an economic point of view, it is important to achieve hemihydrate synthesis in the shortest possible time with the lowest possible energy input is important since the mining of the raw material, gypsum, is relatively inexpensive and the energy expenditure is a significant portion of the synthesis or processing . If dehydration is carried out in kilns and the energy is produced from fossil fuels, the environmental costs are particularly high. Using solar energy to dehydrate gypsum to produce β-hemihydrate is one way to improve the carbon footprint . Alternatively, α-hemihydrate is produced from gypsum using an autoclave process . The widespread use and growing demand for calcium sulfates has led to the recognition that in some geographical regions, including the EU, calcium sulfate will be classified as a critical raw material within the next decade . Therefore, the search for alternative, more sustainable synthesis routes, which are not based on raw natural gypsum, and instead include calcium sulfate from waste, is of particular importance . Furthermore, considering the environmental impacts and monetary costs of gypsum disposal, targeted processing of calcium sulfate waste is crucial .
|
64fec0a8b338ec988a460e91
| 1 |
Over the past decade, several alternative approaches have been explored to overcome the high-energy consumption of dehydration and the unfavourable continuous operation of the autoclave process to produce calcium sulfate hemihydrate at lower temperatures with reduced waste heat. Many of these methods are based on "wet" dehydration of gypsum or direct precipitation from mixed solutions containing calcium and sulfate, combined with a mixture of salt/acid , alcohol/water , or additives . Other approaches include solvent-assisted grinding of gypsum and a solvothermal route using amorphous Ca-ethoxide as a precursor . While some of these methods show promise in reducing energy consumption and waste heat, they often rely on expensive additives, the use of large quantities of organic solvents or the need for extensive milling times. These factors limit their widespread application in industrial settings. Therefore, further research is warranted to find alternative approaches that strike a balance between efficiency, cost-effectiveness, and environmental sustainability in the production of calcium sulfate hemihydrate on a larger scale.
|
64fec0a8b338ec988a460e91
| 2 |
These hurdles have stimulated the development of a solution-based processing method for gypsum using hyper-saline brines (> 4 M), as a simple yet effective way to control the stability ranges of the different calcium sulfate phases. We demonstrate that in (hypersaline) brines, it is possible to produce calcium sulfate hemihydrate in a rapid and cyclic manner. Importantly, this approach allows for total flexibility in the choice of source materials, be it raw or waste gypsum , or through the direct mixing of solutions containing calcium and sulfate ions . Another major advantage of this method, compared to solid state and autoclave methods, is that all the transformations take place in an aqueous medium at temperatures well below the boiling point of water . Liquid water solutions inherently have a high heat capacity, which can help reduce energy losses during the bassanite synthesis processes. By using insulated, closed-loop reactors, we can further improve energy efficiency and optimize the production of calcium sulfate hemihydrate, making the process even more attractive from an economic and environmental perspective.
|
64fec0a8b338ec988a460e91
| 3 |
Bassanite is considered to be metastable in solution with respect to gypsum and anhydrite 1 (Fig. ). This explains why bassanite deposits do not persist on Earth, except in extreme arid areas, such as the desert of Atacama or volcanic deposits 1 . However, experimental observations indicate that the persistence of bassanite is dramatically increased in solution with high ionic strength , the fact of which has been used to produce bassanite . Yet, when looking at the solubility of the different CaSO 4 phases in a concentrated NaCl solution the stability region of bassanite is not expected (and sometimes even not observed ). The problem mainly lies in the quality of the thermodynamic data. In Fig. we show solubility curves of the three calcium sulfate phases calculated using the widely used geochemical modelling software PHREEQC 2 and the included Pitzer database (see also SI: Materials and Methods).
|
64fec0a8b338ec988a460e91
| 4 |
According to this calculation, in hypersaline brines (in our case 4.8 M, see SI: Materials and Methods), the transition point between gypsum and anhydrite is shifted to a temperature of T < 30 °C, but bassanite supposedly remains metastable for the entire temperature range. Thus, bassanite should not form from solution under these conditions, which is incorrect, as we show later. However, it is important to note that these curves are merely extrapolations from experimental data, where the solubility of CaSO 4 phases in high salinity solutions has been mainly determined experimentally for some temperature ranges. This becomes apparent when the solubility curves (Fig. ) are calculated using the latest version of PHREEQC 3 with an updated Pitzer database (SI: Methods, Eq. S1, Table ). Updated and improved data suggest that bassanite does become the stable phase at T > ~110 °C in 4.8 M brines. This would explain the experimental observation that bassanite forms at this temperature in high salinity experiments , but it contradicts the observation that anhydrite should eventually form at these conditions (Fig. and ref. ). These uncertainties regarding the phase diagram of CaSO 4 have hampered the search for a large-scale synthesis route for bassanite at a lower temperature, and thus lower environmental cost.
|
64fec0a8b338ec988a460e91
| 5 |
Here, guided by our previous experimental results, solubility calculations (Fig. ) and other studies we preselected an optimal window, indicated by the red rectangle Fig. , to produce bassanite via a simple and efficient cyclic synthesis method. Although various thermodynamic and applied aspects of this brine solution method have been reported in the literature , the actual conversion of gypsum to bassanite in brines has not been so far explored from the perspective of nano-scale mechanisms, which will be discussed in detail in the next sections.
|
64fec0a8b338ec988a460e91
| 6 |
In our experiments, we followed the phase changes occurring in a solution containing 0.2 M CaSO 4 and 4.8 M NaCl (see SI: Materials and Methods). This choice of solution conditions was based on our previous work, in which we found that bassanite precipitated directly at T > 80 °C, while at T < 80 °C only gypsum was observed . Taking into account the updated solubility calculations (Fig. , SI: Materials and Methods) and our previous experimental results (Fig, ), we selected a narrow window of temperatures, around T = 90 °C, to evaluate the stability of calcium sulfate phases, which was done by performing in situ and time-resolved X-ray scattering/diffraction (Fig. , Fig. )
|
64fec0a8b338ec988a460e91
| 7 |
First, we monitored precipitation from a solution containing 0.2 M CaSO 4 and 4.8 M NaCl at 90 °C using in situ scattering (Fig. ). In this scenario, bassanite formed rapidly as the only crystalline phase. Subsequently, the solution was cooled down to T = 35 °C, which eventually induced the precipitation of gypsum. However, at this temperature the gypsum phase had started to form only after an extensive induction time of ~4800 s, and both phases still coexisted for the following 3600 s. This pointed to very slow kinetics of the hydration of bassanite to gypsum, thus explaining its persistence under high-salinity conditions. This stands in opposition to the typical behaviour of this phase in low salt environments, where a short induction time and rapid transformation is typically reported . On the other hand, increasing the temperature to T = 90 °C (> 13000 s), resulted in an almost instantaneous (< 300 s) reconversion of gypsum into bassanite. These cycles were repeated several times, showing a similar duration of the different phase transformations (Fig. ), thus demonstrating the full reversibility of the process.
|
64fec0a8b338ec988a460e91
| 8 |
Surprisingly, this temperature-dependent cyclic phase change confirms a range of physicochemical conditions where bassanite is thermodynamically more stable than gypsum. Additionally, it also shows that the formation of anhydrite, the CaSO 4 phase with the lowest solubility is kinetically hindered at T = 90 °C, . Importantly, the almost instantaneous gypsum-bassanite conversion occurs at a distinct temperature boundary. In a different variant of our scattering experiment, we precipitated gypsum at T = 35 °C and heated this slurry in 5 °C increments up to T = 90 °C, with each temperature being maintained for 900 s (15 min) (SI: Fig. ). The scattering data revealed that the rapid transition is viable only at a threshold temperature of T = 80 °C, which is significantly lower than the one predicted by the solubility calculations shown in Fig. (T = ~110 °C). Furthermore, in static experiments (i.e. without stirring), we obtained bassanite at T = 70 °C after 5 days (SI: Table ). Overall, these observations contradict the calculated solubility data shown in Fig. and, but they do confirm the previously reported stability behaviour of the calcium sulfate phases precipitating from brines solutions shown in Fig. . We further focused experimentally on a single cycle of gypsum-bassanite-gypsum conversion with extended reaction times for 35 °C < T < 90 °C (>25000 s with scattering and >70000 s with Raman spectroscopy, Fig. ). With these measurements, we intended to exclude a possibility that the occurrence of the cyclic transformations is driven by the residual seeds of the bassanite. Such seeds, hypothetically, might nucleate the hemihydrate phase, although it would be thermodynamically unfavourable . To fully exclude this scenario in Fig. , the experiments start with gypsum. Moreover, the equilibration periods, when only gypsum exists, are extended with respect to However, in Raman, the conversion of bassanite to gypsum was notably slower, with bassanite persisting for an extended period (~51000 s) after the solution commenced to cool down to T = 35 °C (T < 90 °C at ~24000 s, T = 35 °C at ~47000 s). Bassanite started to convert to gypsum at ~75000 s. The longer persistence of bassanite in the in situ Raman experiment compared to the in situ scattering experiment is most likely the result of the inherent differences in the hydrodynamic regimes between both setups . In the latter, the solution was vigorously stirred (1000 rpm) in the reactor vessel and continuously recirculated through a capillary using a peristaltic pump. In contrast, the Raman setup involved gentler stirring (400 rpm) of the solution in the reaction vessel.
|
64fec0a8b338ec988a460e91
| 9 |
The scattering data contain a plethora of information regarding phase, morphology and size evolution during the transformation reactions because we measured a continuous, gapless signal for 0.1 nm -1 < q < 35 nm -1 . Based on this data, we elucidated the structural changes taking place during the conversion reaction. One important aspect is that as the calcium sulfate crystals evolve, they exhibit orientation-dependent anisotropic 2D intensity profiles (c.f. Fig. ). This phenomenon is a characteristic feature of the calcium sulfate system when characterized using a flow-through capillary . Specifically, both gypsum and bassanite primarily form acicular single crystals . As the crystals reach sizes of a few hundreds nanometers to several micrometers, they align themselves with the flow along their long-axes as they pass through a horizontally mounted capillary. This leads to a texture in an otherwise polycrystalline system (SI: Fig. , SI: Fig. ). Additionally, at small-angles and at length-scales >>1 nm, a texture is also observable because calcium sulfate single crystals are internally mesostructured , meaning that they consist of slightly misaligned crystallographic domains, resembling brick-in-the-wall-like units. The direction-dependent intensity variations were clearly observable for the diffraction rings (wide-angle scattering, WAXS), as well as for the scattering around the beamstop in the center (small-angle scattering, SAXS) (SI: Fig. , SI: Fig. ). The extent of the anisotropy evolved with time following the temperature profile and is related to the changes in crystal phase, size, and morphology (SI: Fig. ). Therefore, the analysis of orientation-dependent changes in intensity and the crystal morphology of calcium sulfate provides valuable insights into the crystallisation mechanisms, which will be discussed in the following paragraphs. In Fig. we present the evolution of peak areas from the reflections at q bass ~10.42 nm -1 , characteristic of bassanite (110), and q gyp ~8.25 nm -1 , characteristic of gypsum (020), as shown in Fig. . The trend, as a function of T and time, reveals changes in anisotropy as the crystalline phases evolve.
|
64fec0a8b338ec988a460e91
| 10 |
Generally, an increase in anisotropy is correlated with larger crystal sizes and aspect ratios. Conversely, at T > 80 °C (> 4800 s), the decrease in anisotropy indicates a dissolution-reprecipitation process, since it correlates with a decrease in the crystal size of gypsum and concurrent nucleation and subsequent growth of isotropic nano-sized bassanite crystals. This effect becomes evident through analysis of the changes in the ⍴ parameter, calculated from the pseudo-invariant expressing q 2 -weighted integral intensity:
|
64fec0a8b338ec988a460e91
| 11 |
Fig. ). The value ⍴ = Q* meridian /Q* equatorial highlights that the nucleation of bassanite at T > 80 °C is associated with the formation of near-isotropic crystals/particles (⍴ approaches 1). Conversely, the growth of larger crystals of any phase results in ⍴ deviating from 1, where ⍴ < 1 implies that crystals' long axes are aligned horizontally. Interestingly, after the formation of small isotropic bassanite particles, upon cooling to T = 35 °C, this phase continues to grow, becoming more anisotropic. This is evidenced by the sharp drop in ⍴ after ~9600 s. Gypsum does not appear in diffraction until ~15900 s . This extended growth period of bassanite outside its stability window (Fig. ) offers the possibility to convert gypsum to phase-pure bassanite with specific sizes/morphologies. Our small-angle data analysis further supports this notion (SI: Materials and Methods).
|
64fec0a8b338ec988a460e91
| 12 |
The morphological and size-distribution changes of the crystals during the gypsum-bassanite phase transformation (Fig. , Fig. ) are also evident at low-angles (SI: Fig. , Fig. ). In Fig. , we present SAXS data from three characteristic moments: 3200 s when only well-developed gypsum is present at T = 35 °C; 5200 s, which marks the crossover between the dissolution of gypsum and the formation of bassanite; and 7200 s when bassanite exists at T = 90 °C. Our minimum q was limited to 0.1 nm -1 , which corresponds to a maximum feature size of ~60 nm. However, gypsum and bassanite crystals typically grow to many microns in size and beyond , and therefore, we observe them in scattering predominantly as Porod interfaces following approximately an I(q) ∝ q -4 dependence (dashed line in Fig. ). Such an interface is observed in all the curves for at least q < ~0.3 nm -1 , indicating that larger crystals are always present. A comparison of scattering features from 3200 s (53.3 min) and 7200 s, i.e., pure gypsum vs. bassanite, reveals that the initially precipitated gypsum crystals from the first cycle (Fig. ) exhibit a deviation from a Porod interface at q ~0.7 nm -1 (i.e., the 1st small "hump"-like deviation, Fig. ). The profile is anisotropic (SI: Fig. ) and implies the presence of oriented scattering features of ~10 nm in size, confirming the mesostructured nature of gypsum crystals . Noteworthy, these oriented nano-features partially disappear in developed bassanite crystals after 7200 s and in the re-precipitated gypsum after 15900 s (Fig. ).
|
64fec0a8b338ec988a460e91
| 13 |
nm -1 (compare 3200 s vs. 5200 s in Fig. ). The isotropic character and slight shift in q of this feature indicate the formation of small particles independently of the oriented ones (the 1 st hump) already present in the crystals. This supports the notion of the de novo nucleation of bassanite. In Fig. , a cryo-TEM image captured as close as possible after the onset of the gypsum-bassanite transformation (corresponding to ~4860 -5200 s in Fig. ) depicts aggregates of polydisperse crystalline nanoparticles of bassanite (electron diffraction, inset in Fig. ), with the smallest particles measuring around 10 nm in diameter. We further quantified the evolution in the small-angle scattering patterns by independently fitting the data using Eq. S3 (SI: Materials and Methods, and SI: Fig. ) for both the meridian-and equatorial-integrated curves. The trends, summarised in Fig. , provide additional insights into the mechanism of the transformations, including the presence of mesocrystallinity. These trends capture comprehensive structural features for q < 3 nm -1 and are not specific to any particular phase. However, when correlated with the diffraction peaks (Fig. and Fig. ), we can easily attribute the observed changes to the evolution of crystalline phases. In Fig. , we present a schematic illustration of the concepts resulting from the observed trends.
|
64fec0a8b338ec988a460e91
| 14 |
In Fig. , the parameter B, in general, is proportional to a specific surface of the solid phases when D > 3 (i.e. surface fractals, SI: Eq. S3). Initially, as gypsum crystals grow before ~4000 s (T = 35 °C), this parameter gradually decreases in both directions due to a reduction in surface-to-volume ratio. The decrease in surface-to-volume ratio suggests the development of larger crystals while smaller particles/crystals disappear, possibly through processes such as
|
64fec0a8b338ec988a460e91
| 15 |
Ostwald ripening or coalescence , or the coalescence of the internal mesocrystalline domains . Due to the shape of the crystals and their alignment, B evolves independently for the two integration directions. The proportional to the product of the volume fraction and the volume of smaller particles; R s is the gyration radius of small structural units. In addition, Fig. shows trends in B s and bkg parameters, which are not included in Fig. . These two sets of values affect the region in which small-angle scattering transitions into the wide-angles and the atomic-level structure. They contribute mostly to a relatively narrow part of the scattering curve, where uncertainties are high (see B s -dependent component in Fig. ), which makes them less representative or suitable for interpretation.
|
64fec0a8b338ec988a460e91
| 16 |
These two processes result in a peak in B at 4860 s (Fig. , 2-3 fold jump worth respect to ~4000 s). The system becomes briefly isotropic, due to the de novo nucleation of bassanite (see also Fig. ). The meridian and equatorial values of B overlap and gradually decrease until gypsum is fully dissolved at 5400 s.
|
64fec0a8b338ec988a460e91
| 17 |
After 5400 s, at T = 90 °C, the anisotropy begins to develop, and the equatorial values of B sharply decrease, indicating the continued evolution of bassanite crystals (Fig. , stage II and III). By 9600 s, when T = 90 °C, the meridian and equatorial B values reach their plateaus. After 9600 s, as the temperature gradually decreases to T = 35 °C, the meridian B remains relatively constant.
|
64fec0a8b338ec988a460e91
| 18 |
However, the equatorial counterpart continues to slowly and monotonously decrease by ~20%, around 12000 s. This observation aligns with the evolution of the diffraction peaks and the ⍴ parameter shown in Fig. , indicating a continuous increase in the aspect ratio of the bassanite crystals (Fig. , stage III). This can be attributed to the slow dissolution of bassanite crystals, strongly supported by our in situ Raman spectroscopy data, which we will further discuss.
|
64fec0a8b338ec988a460e91
| 19 |
After 15000 s, bassanite gradually starts its reconversion to gypsum (Fig. ; Fig. , stage IV and V). This transformation is accompanied by a reduction in the surface-to-volume ratio in both the meridian and equatorial directions, leading to a decrease in the B parameter. This is because the crystals of both calcium sulfates change shape during the phase transition. Additionally, the re-precipitation of gypsum from bassanite after 15000 s is characterised by overall lower B values compared to the initial gypsum formed before 4000 s.
|
64fec0a8b338ec988a460e91
| 20 |
This suggests that these two entities of gypsum have different crystal size distributions and (meso)structures, as we infer from other trends observed in Fig. . The changes in exponent D (Fig. ) represent the evolution of interface roughness and the presence of surface fractals . These interfaces originate from large crystals "outside" our q-range. Smooth Porod interfaces exhibit D = 4, while rougher surfaces have 3 < D < 4, with lower D values indicating higher roughness . Additionally, this parameter tracks the anisotropic evolution of crystals, where the roughness of the gypsum surface is direction-dependent before ~4000 s (T = 35 °C). This phenomenon further supports the mesocrystalline nature of freshly precipitated gypsum crystals (Fig. , stage I).
|
64fec0a8b338ec988a460e91
| 21 |
The formation of bassanite nano-crystals leads to a decrease in D (Fig. In this regard, the two occurrences of gypsum, before and after the heating step at T = 90 °C, again appear to have different mesostructures, as also indicated by parameter B. This distinction is further reflected in the trends observed in the R s and G s parameters.
|
64fec0a8b338ec988a460e91
| 22 |
In Eq. S3 (SI: Materials and Methods), the mentioned variables parameterize a Guinier form factor, which is used to approximate any small units that may exist within the crystals, such as "bricks-in-the-wall" or "loose" particles with radii of gyration R s . Additionally, G s is proportional to the product of the volume fraction and the volume of these smaller particles. The initial as-precipitated gypsum, before 4000 s, contains small building units with a radius of ~4 nm. We interpret these units as components of a mesocrystal, consistent with our previous work (Fig. , stage I). The dissolution of gypsum with increasing temperature induces the rapid formation of a large population of nanoparticles at 4860 s, causing the radius of gyration of the small units to increase to ~5.5 nm (Fig. , stage II). Further crystallisation of bassanite results in a partial loss of the mesostructure. The constituent building blocks are only observable in the meridian direction (R s ~4 nm), while they vanish in the equatorial direction, as indicated by G s ~0 after >5400 s. This observation aligns with a decrease in the specific surface area expressed by parameter B, assuming crystal growth occurs through the coalescence of internal mesocrystal domains . Examining the meridian and equatorial trends in Fig. , the surviving mesostructure indicates that the crystals possess domains with shorter, nano-sized features aligned parallel to the long axes (c-axes) of the crystals. Such alignment correlates with the layer structure of gypsum and the channel structure of bassanite, where these anisotropic motifs align parallel to the long axes of the respective crystals (Fig. ). In the perpendicular directions, the domains fuse into a homogenous body, cancelling the mesostructure.
|
64fec0a8b338ec988a460e91
| 23 |
Importantly, this difference in the mesostructure is also observed for the re-precipitated gypsum crystals formed through the hydration of bassanite (Fig. , stage V). The question now arises, how can we rationalise the occurrence of these different mesostructures? It is known that the hydration of bassanite to gypsum involves a dissolution-reprecipitation process in contrast to solid-state diffusion , which is seemingly also the case for our gypsum-bassanite transformation. When we compare the two gypsum formation events (Fig. ), the first occurrence happens through mostly homogeneous nucleation upon mixing of the two stock solutions. However, all subsequent nucleation and crystallization events occur from heterogeneous solutions, as the growth of de novo crystalline phases occurs in the presence of dissolving preceding phase, and thus solid interfaces. This is the case for both bassanite and gypsum nucleation (Fig. , stages II-V). We propose that such heterogeneous surface-affected nucleation processes lead to the formation of less mesostructured crystals compared to those growing from homogeneous bulk solutions. Furthermore, as we discuss in the next section, the determined structural and morphological changes, as well as the structural disorder associated with phase transitions in brines, are driven by temperature-dependent changes in the solubility of bassanite and gypsum (Fig. ). These changes in solubility, in turn, control the rates of nucleation and crystal growth.
|
64fec0a8b338ec988a460e91
| 24 |
To specifically measure the solubility changes of calcium sulfate phases during temperature variations, we employed Raman spectroscopy to monitor the cyclic formation of gypsum and its subsequent transformation to bassanite (Fig. , SI: Materials and Methods). Through the analysis of time-resolved Raman spectra, we were able to determine the variations in the concentration of dissolved "free" sulfate in the solution. This was achieved by tracking the peak of the ν1 mode of aqueous sulfate at ~980 cm -1 (Fig. , Fig. ). Using our Raman setup we performed an experiment analogous to the one with scattering in Fig. , in which after mixing the stock solutions of CaCl 2 and Na 2 SO 4 (0.2 M CaSO 4 in 4.8 M NaCl), gypsum precipitated (Fig. , Fig. ). The time scales of the spectroscopic experiment differ from the scattering experiment due to specific constraints of each experimental setup (SI: Materials and Methods, Table ). However, the essential aspects of the gypsum-bassanite transformations are analogous and directly comparable between the two approaches (Fig. ). Upon the precipitation of gypsum, the sulfate concentration gradually decreased during the initial 1800 s to ~39 mM, after which it remained stable until the temperature was raised to T = 90 °C (Fig. ). The plateau in sulfate concentration at T = 35 °C indicated that the system had approached a state of thermodynamic equilibrium. However, from the scattering trends observed in Fig. , it is evident that gypsum crystals continued to evolve even before the temperature increase (Fig. , stage I). As the solution started to cool down to T = 35 °C, bassanite began to dissolve, increasing the sulfate concentration to ~41 mM. After a prolonged induction time (relative ~48000 s), gypsum started to crystallize, causing the sulfate concentration to decrease to ~37 mM. These two reduction steps in sulfate concentration are consistent with our scattering data (Fig. , Fig. and Fig. ), where we observed the morphological evolution of bassanite crystals induced by cooling before any gypsum crystals appeared. In fact, this dissolution of bassanite accompanied by an increase in sulfate concentration, is analogous to the reverse gypsum-bassanite transformation. The key difference lies in the kinetic rates: gypsum dissolves rapidly while bassanite dissolves slowly. This difference in dissolution rates may explain why secondary gypsum crystals exhibit a higher degree of mesoscale ordering compared to the initial crystals (Fig. ). Namely, the gradual increase in saturation in the presence of an interface (i.e. dissolving bassanite crystals) would promote the formation of fewer nuclei, slower growth, and consequently, larger and better-ordered gypsum crystals, which aligns with our observations. When gypsum became the dominant phase, the sulfate concentration increased slightly, and with only gypsum remaining, the concentration of sulfate equilibrated at ~41 mM. When the same solution was heated again to T = 90 °C, a very similar trend in the sulfate concentration profile as the one observed during the first heating stage was observed (Fig. ).
|
64fec0a8b338ec988a460e91
| 25 |
In Fig. , we compare the experimental solubility values of gypsum and bassanite in a 4.8 M NaCl brine at different temperatures, as extracted from the Raman spectra, with the solubility curves calculated using PHREEQC 3 (see also Fig. ). This graph reveals a notable discrepancy between the theoretical curves and the experimental data. At room temperature, the measured solubilities of gypsum and, especially, bassanite are lower than the calculated solubilities. This finding is remarkable because solubility measured from precipitation experiments typically overestimates the actual solubility due to the time required to reach true equilibrium. Noteworthy, when the temperature is rapidly increased to T = 90 °C, the observed increase in free sulfate concentration in the solution contradicts expectations, as it was commonly assumed that the solubility of gypsum significantly decreases with rising temperatures under low salinity conditions 1 . However, the revised PHREEQC 3 code predicts this increase in gypsum solubility with temperature for highly saline solutions. Overall, numerical solubility predictions in brines are (still) not accurate enough to determine the phase stability in the CaSO 4 system.
|
64fec0a8b338ec988a460e91
| 26 |
The cyclic transformations of gypsum-to-bassanite in high-salinity brines offer an efficient route to produce hemihydrate. To demonstrate the practicality of this method, we also processed natural gypsum into bassanite. We used 0.2 M CaSO 4 equivalent solutions/slurries, containing 34.4 g/L of dispersed natural gypsum powder (see SI: Materials and Methods). We reproduced the processes at high salinity with 4.8 M NaCl brines. A detailed plot of the Raman spectra in the first 60000 s after raising the temperature to T = 90 °C shows, as expected, the evolution of the transformation of gypsum to bassanite and back to gypsum (SI: Fig. ). After raising the temperature, gypsum completely transformed into bassanite within 1500 s and that the hemihydrate remained stable for >30000 s after the temperature was lowered to T = 35 °C, until gypsum eventually reprecipitated.
|
64fec0a8b338ec988a460e91
| 27 |
We highlight here the experiment performed at 0.2 M CaSO 4 and 0.4 M NaCl (SI: Fig. ). Compared to the high salinity experiment (4.8 M NaCl), the low salinity one showed three significant differences. Firstly, a much higher temperature, T > 115 °C, was required to facilitate the rapid transformation of gypsum to bassanite (SI: Fig. ). Secondly, during this transformation, there was no increase in the sulfate concentration; instead, it sharply dropped (SI: Fig. and SI: S9C) indicate that the solubility crossover occurs at a much lower temperature than predicted in the case of 4.8 M NaCl. This discrepancy is mainly because the calculated solubility of bassanite is much higher than the measured one at high salinity. In contrast, the experimental and calculated solubility values compare better for low salinity, although the experimental values are lower than the predicted ones. The increase in sulfate concentration with increasing temperature, during the high salinity experiment can be attributed to an increase in gypsum solubility, as evidenced by the trend in both calculated and experimental data. Lastly, the fast transformation from bassanite to gypsum at low salinity can also be attributed to the shift in the solubility crossover.
|
64fec0a8b338ec988a460e91
| 28 |
To establish the critical salt concentration for a rapid gypsum to bassanite transformation (assuming, arbitrarily, <1 h to be a practical threshold), a series of experiments at different salt concentrations were carried out at T = 90 °C using unstirred batch reactors. These experiments revealed that a NaCl concentration of >4 M is needed to induce rapid (<1 h) transformation times (SI: Table ). At lower salt concentrations, such as 1.4 M NaCl, the transformation rate is significantly reduced, with gypsum not transforming to bassanite even after 11 days. Additionally, the critical temperature for a rapid gypsum-to-bassanite transformation was established at 4.8 M NaCl. Table in SI, shows that only at T = 90 °C, the transformation is fast (<1 h) in unstirred reactors. Experiments conducted at T = 70 °C indicate that it takes up to 5 days to transform gypsum to bassanite, which from an industrial perspective is not interesting. But it does confirm again that in high salinity solutions, the solubility crossover is shifted to much lower temperatures than those predicted by the thermodynamic calculations.
|
66c0d0b3f3f4b05290336ed3
| 0 |
The modularity of metal-organic materials means that compounds with identical structural topologies but different ligands can be readily synthesised (they are 'isoreticular'). This in turn enables the synthesis of diverse and extensive ligand solid-solutions, which allows control of chemical function, e.g. methane separation and catalytic activity . The physical properties of metal-organic frameworks (MOFs), e.g. mechanical, magnetic, or electronic function, can equally be controlled through ligand solution. There remains a great deal to learn about the physical properties of mixed-ligand MOFs, especially the possibility of creating function that goes beyond the linear combination of stoichiometric end-members. Recent work has shown that ligand solid-solutions in zeolitic imidazolate frameworks (ZIFs) subtly modulate the magnetic ordering temperatures of sod topology ZIFs and control the pressure-induced pore closing ZIF-4 analogues, and that solid-solutions of terminal halide ligands in Cr(pyz) 2 Br x I 2-x produce temperature-induced valence tautomeric transitions not present in the stoichiometric phases. Ligand solid-solution control over mechanical and magnetic function in vdW magnets is of special interest because pressureand strain-control over magnetic function can be readily achieved in devices. This is particularly true for noncollinear magnets, where continuous evolution of magnetic order and properties is possible. We have recently reported a family of new layered MOMs with non-collinear magnetic structures, including the canted ferromagnet NiCl 2 (btd). This material consists of NiCl 2 chains coordinated by the nitrogens of the nonlinear btd ligand to form corrugated sheets [Fig. ]. The easy-axis ferromagnetic chains in combination with the tilting of chains induced by the ligand geometry, leads to noncollinear canted ferromagnetism with significant coercive field, µ 0 H c = 1.0(1) T. The modularity of this system, together with the promise of its magnetic function, prompted us to investigate how substitution of the btd ligand for bod will affect both the structure, and the magnetic and mechanical properties of this MOM.
|
66c0d0b3f3f4b05290336ed3
| 1 |
Our previous work established that NiCl 2 (btd) can be made phase pure and crystalline through the direct reaction of NiCl 2 • 6 H 2 O and btd, and thus we first explored this approach to create the solid-solutions NiCl 2 (btd) 1-x (bod) x , attempting syntheses with target bod fraction, x t = 0, 0.25, 0.5, 0.75 and 1.0
|
66c0d0b3f3f4b05290336ed3
| 2 |
Analysis of the powder X-ray diffraction (PXRD) data confirmed that we were able to produce the desired phase up to x t ≤ 0.75 [Fig. ESI Fig. ], however, we found that the the pure bod phase did not form. Consequently, we synthesised a series through the reaction of ethanolic solutions of nickel chloride and ligands over the same target range of x t , analogous to CoCl 2 (btd). We found by analysis of PXRD data that this again produced powders isostructural to NiCl 2 (btd) up to x t ≤ 0.75, but at x t = 1 we obtained a phase mixture for which primary phase was unknown. The purity of all other compounds was assessed using Pawley refinement of the PXRD data, which showed that the samples synthesised through direct reaction contained very small quantities of starting material, but that the solution-synthesised samples had broader diffraction peaks, likely due to small particle sizes [Fig. The phase mixture formed during solution synthesis with x t = 1 included a number of small single crystals (further details of the single crystal diffraction characterisation can be found in the ESI Sections S2.3, S2.4). We isolated a larger crystal (63 × 14 × 10 µm 3 ) which we found using single crystal X-ray diffraction to be a new 1D coordination polymer NiCl ], though we also were also able to find crystals of NiCl 2 • 2 H 2 O. Re-analysis of our PXRD data in the light of this new structure showed that it was primarily NiCl 2 (bod) 2 and a small quantity of nickel chloride hydrates. Further examination using single crystal electron diffraction of the remainder of the reaction mixture, which was dispersed as powder after being ]. Comparison of refinements with only bod, only btd and mixed ligand showed that the sample included nanocrystals of a monoclinic twinned NiCl 2 (bod) and an orthorhombic polymorph of NiCl 2 (btd), although we cannot exclude that this orthorhombic phase includes a low proportion of bod (<5%). We note these phases are not seen in the bulk PXRD and hence, we ascribe the formation of a small number of nanocrystals of NiCl 2 (btd) to the presence of adventitious btd, likely facilitated by its high vapour pressure. This highlights the capability of electron diffraction to find and solve the structures of even minor crystalline phases. To determine the btd:bod ratio, and hence x, we dissolved a portion of each sample in d 6 -DMSO and carried out solution state 1 H NMR [Fig. (a), Fig. ]. This revealed that the bod content determined by NMR, x NMR , was uniformly bod-poor compared to the target composition. This, together with the formation of NiCl 2 (bod) 2 in preference to NiCl 2 (bod), suggests that the more electron deficient bod ligand does not coordinate as readily as the btd ligand. This is further borne out by the lack of reported metal complexes containing bod as a ligand in the CSD. We found that the x NMR = 0.31 (x t = 0.75) sample synthesised through direct reaction was poorly crystalline and contained significant impurities, so has not been further analysed. We thus focussed on samples with x t ≤ 0.75 for solution state reaction and x t ≤ 0.50 for solid state samples.
|
66c0d0b3f3f4b05290336ed3
| 3 |
Comparison of the Pawley derived unit cell volume and lattice parameters with the composition determined from NMR shows linear, Vegard's law-type behaviour. We find that the interlayer spacing, c, expands on incorporation of bod, with the M-L-M distance, b, in turn shortening. The contraction along b can be explained by the shorter N-N distance in bod than btd, which would predict b btdb bod = 0.40 Å, which is in quantitative agreement with the fitted value of 0.391(7) Å. The significant interlayer expansion cannot be easily rationalised by differences in the size between btd and bod, but seem rather to reflect subtle differences in the angle between N-Ni-N axes of neighbouring Ni octahedra, though may also suggest that the interlayer vdW forces are weaker for bod than btd. The relative lack of change along the a axis suggests that the NiCl 2 chain is relatively unperturbed by the differences in Ni-N bonding and that changes in intermolecular forces between bod and btd are not a driving factor. We find no evidence of superlattice reflections indicative of long-range ordering of the bod and btd ligands, and no clear evidence of structured diffuse scattering that would be a signature of local ordering, though this can be challenging to detect in powder diffraction measurements of 2D compounds. Having developed this solid-solution series, we then investigated their physical properties, focussing on the mechanical compressibility and magnetic properties. We measured the compressibility of these materials using high pressure synchrotron X-ray powder diffraction at the I15 beamline of Diamond light source, using a hydraulic pressure cell to obtain the fine pressure resolution required [Fig. ]. This cell allows measurements from ambient to 0.4 GPa with pressure increments of ∆P = 0.02 GPa. We used silicone oil AP-100 as a pressure transmitting medium , which should be hydrostatic and non-penetrating in this pressure regime. We investigated here the doped samples synthesised directly using solid-state synthesis because they were more crystalline. The lattice parameters were refined using Pawley refinement. A limited number of impurity peaks were identified and fitted using additional structure free peaks [ESI Section S2.2, Figs. S4-12, Tables ].
|
66c0d0b3f3f4b05290336ed3
| 4 |
We found no evidence of pressure-induced framework degradation or phase transitions up to 0.4 GPa. Fitting of the pressure dependence of the volume using the second-order Birch-Murnaghan equation of state was carried out using the PAS-Cal Python package , and revealed that the bulk compressibility was B 0 = 18.7(3) GPa for the pure NiCl 2 (btd), with the two doped samples both slightly stiffer: x = 0.10 has B 0 = 20.6(0.3) GPa and x = 0.22 has B 0 = 19.96( ) GPa [Fig. ]. These values are comparable to those reported for other nickel(II) layered materials, e.g. Ni(NCS) 2 B 0 = 17.0(2) GPa, and NiI 2 B 0 = 27.7(9) GPa, and stiffer than ZnCl 2 (3,5-dichloropyridine) 2 , B 0 = 14.52 (8), which contains 1D ZnCl 2 chains. Our X-ray diffraction measurements allow us to determine not only the bulk modulus, but how the compressibility varies with direction. We find that the principal compressibilities do approximately coincide with the crystallographic axes, although in a monoclinic system the principal strains will not lie in general, along the unit cell axes. The compressibility is largest along the interlayer direction, X 1 (~c) K 1 = 27.3(3) TPa -1 . X 2 (~a) is next stiffest, corresponding to the Ni-Cl-Ni chain direction, K 2 = 14.8(4) TPa -1 , with the stiffest direction being the X 3 (b) along the Ni-N-(O/S)-N-Ni bonds direction, K 3 = 7.7(4) TPa -1 [Fig. ), ESI Table ]. The softness of X 1 is typical of vdW layered materials, e.g. Ni(NCS) 2 K vdW = 32.5(2) TPa -1 , where K vdW is the compressibility normal to the vdw layers. As inorganic materials tend to be less compressible, the direction with purely inorganic connectivity might be expected to be the stiffest, but in fact it is nearly twice as soft as the direction with purely metal-organic connectivity. This perhaps is due to the fact that the X 3 direction primarily involves direct bond compression, whereas the X 2 direction corresponds to bending of the Ni-Cl-Ni angle, although it is notable that DFT calculations suggest significant ππ-interactions between the organic ligands along this direction which may modulate the compressibility. This trend is consistent with previous investigations of metal organic materials, such as in [CuCl(pyrazine) 2 ]BF 4 where the Cu-pyrazine-Cu plane was significantly stiffer than the Cu-Cl-Cu chains, and the plastically deforming ZnCl 2 (3,5-dichloropyridine) 2 , where the ZnCl 2 chains are as soft as the vdW directions (K ZnCl2 ≈ 23 TPa -1 ).
|
66c0d0b3f3f4b05290336ed3
| 5 |
As the structure is anisotropic, doping with bod also changes the compressibility differently in different directions. The interlayer direction becomes notably stiffer, with compressibility dropping to 24.7(4) TPa -1 (x = 0.10) and 24.70(11) TPa -1 (x = 0.20). Within the plane, the inorganic X 2 axis becomes slightly stiffer, 12.6(2) TPa -1 (x = 0.10) and 14.1(3) TPa -1 (x = 0.20), whereas the organic X 3 axis in fact softens, 8.14( ) TPa -1 (x = 0.10) and 8.9(4) TPa -1 (x = 0.20) [ESI Table ]. This suggests that organic substitution can be used to subtly modify the compressibility of MOMs, as found for MOFs, and hence the efficacy of strain tuning, whether in bulk or on surface. ]. All the samples are canted (weak) ferromagnets, with a ferromagnetic ordering temperature T C = 17(1)) K, and substantial magnetic hysteresis [Fig. ]. By contrast, the field dependence of the magnetism systematically varies on doping with bod, with the more bod added the softer the magnet [Fig. (d), ESI Table ]. The drop in hysteresis can be seen most clearly in the coercive field, H c , which decreases by 60% on doping with 33% bod (i.e. x = 0.33). The reduction in hysteresis suggests that the O atom has weaker spin-orbit coupling and is more electronegative, reducing the ligand field, which will together reduce the single-ion anisotropy. It is also possible that the slight differences in tilt angles between NiCl 2 chains induced by the differences in internal bond angles between bod and btd change the degree of canting, though this is not clearly observed, and the changes in H c are much larger than predicted by geometry alone.
|
66c0d0b3f3f4b05290336ed3
| 6 |
The effect of isovalent substitution on magnetic function we observe is consistent with previous studies: replacing S with Se in NiPS 3 and Co(NCS) 2 (pyridine) 2 does not produce large changes in the ordering temperature, with a reduction in T c of 5% in NiPS 3 , and an increase in T c of 30% (1.5 K) for Co(NCS) 2 (pyridine) 2 ; 35 but this change does completely switch the single ion anisotropy from easy-plane anisotropy in NiPS 3 to easy-axis anisotropy for NiPSe 3 . The large effects on magnetic properties we observe is in contrast to the layered ZIF analogues where the organic substituents on the imidazolates had relatively small effects on both ordering temperature and fitted superexchange parameters, suggesting that substitution of atoms more directly on the superexchange pathway is critical to produce large effects. Previous work has shown that pressure can tune noncollinearity in MOMs, and so the combination of mechanical and magnetic tunability we demonstrate suggests that doping will be an effective method to modulate strain switchability.
|
66c0d0b3f3f4b05290336ed3
| 7 |
In conclusion, we report two methods for the synthesis of solidsolutions of NiCl 2 (btd) 1 -x (bod) x . We find that there is an approximately linear dependence of the lattice parameters on ligand substitution, with the expected contraction along the M-L-M chain on doping with the smaller bod ligand. The btd ligand is preferentially incorporated into the structure, likely as it is a more electron rich ligand. Investigation of the mechanical properties using high pressure synchrotron X-ray diffraction showed that incorporation of bod stiffens the framework, primarily due to a reduction in interlayer compressibility, as the layers themselves become slightly more compressible. The canted ferromagnetism is retained on doping but doping tunes the hysteresis, producing a significant reduction (up to 60%) in coercive field. These results demonstrate that functionalisation of organic ligands can be a valuable way to tune both the magnetic function and pressureresponsiveness of van der Waals metal-organic magnets. data and powder X-ray diffraction data. E.M. and M.J.C. analysed the bulk magnetic data. E.M., S.L. and S.P.A. measured and analysed the single crystal X-ray diffraction data. J.P.T. measured the electron diffraction data, J.P.T. and M.J.C. analysed the electron diffraction data. M.J.C. wrote the paper with input from all other authors.
|
61b1d32a0e35eb99589a8774
| 0 |
Tuberculosis is a difficult disease to treat; the standard regimen is four antibiotics, rifampicin, isoniazid, pyrazinamide and ethambutol, for six months. An infection that is resistant to both rifampicin and isoniazid is called multi-drug resistant tuberculosis (MDR-TB) and the treatment regimen recommended by the World Health Organization (WHO) is complex but always includes levofloxacin or moxifloxacin, which are fluoroquinolones .
|
61b1d32a0e35eb99589a8774
| 1 |
Rifampicin acts by binding to the β-subunit of the RNA polymerase (RNAP, encoded by the rpoB gene), preventing the extension of the RNA (Fig. ). The most common resistance-conferring mutation is rpoB S450L, however a wide range of mutations have been observed clinically . The majority of these are found in amino acids 428 to 452 which pack against the drug (usually known as the "rifampicin resistance determining region" or RRDR), enabling the development of nucleic acid amplification tests, such as the Cepheid GeneXpert MTB/RIF system which is endorsed by the WHO for diagnosis of MDR-TB . Not all non-synonymous mutations in the RRDR, however, confer resistance, for example rpoB L443F . Nor does resistance arise purely within the RRDR: rpoB I491F and V170F are proximal to S450L and the former was suspected to be behind an outbreak of MDR-TB in Eswatini since it is not detected by GeneXpert .
|
61b1d32a0e35eb99589a8774
| 2 |
The fluoroquinolones target the DNA gyrase (DNAG), a tetrameric enzyme which unwinds DNA by forming and re-ligating double stranded DNA breaks prior to transcription and replication (Fig. ). Specifically, two fluoroquinolone molecules intercalate into DNA breaks and bind specific gyrA residues via a coordinated Mg 2+ ion. This stabilizes DNA-DNA gyrase covalent linkages and prevents re-ligation of DNA double stranded breaks.
|
61b1d32a0e35eb99589a8774
| 3 |
The most common DNA gyrase mutations found in MDR-TB samples are gyrA D94G and gyrA A90V and these mutations are strongly associated with fluoroquinolone resistance . These residues are part of the gyrA "quinolone resistance determining region" (QRDR), defined as gyrA codons 74 to 113 . However, again, not all mutations in this region confer resistance, leading to false positive resistance results in genotypic assays . Rarely seen DNA gyrase mutations in gyrB are also associated with fluoroquinolone resistance, and a gyrB QRDR from residues 461 to 501 has also been proposed . The residues of the two QRDR regions make up the fluoroquinolone binding pocket, and gyrB A642P is the only mutation significantly associated with an increase in minimum inhibitory concentration (MIC) to fluoroquinolones that was found outside this region .
|
61b1d32a0e35eb99589a8774
| 4 |
We assume that mutations cause resistance by reducing the affinity of an antibiotic ligand for its target. Since we are only interested in whether a mutation increases or decreases the antibiotic's affinity for the target, the difference in binding free energy (DDG) between the wild type and mutant systems is calculated. This can be achieved by employing relative binding free energy (RBFE) methods, whereby a wild type amino acid is transmuted into the mutant along a nonphysical pathway defined by a progress coordinate, 0 ≤ l ≤ 1. For equilibrium-based methods, a series of short molecular dynamics (MD) simulations are performed at fixed values of l and the resulting DG values are related to the difference in binding free energy via a thermodynamic cycle (Fig. ). This approach has been shown to successfully predict if mutations in a relatively small protein, S. aureus DHFR (157 residues), confer resistance to trimethoprim, an antibiotic used to treat urinary-tract infections. In this paper we shall apply the same approach to two much larger protein complexes, the RNA polymerase (4,671 residues) and the DNA gyrase cleavage complex (1,473 residues), to assess how well we can predict the effect of seven and five mutations on the action of rifampicin and moxifloxacin, respectively.
|
61b1d32a0e35eb99589a8774
| 5 |
RNA polymerase and DNA gyrase system setup. The structure of the M. tuberculosis RNA polymerase (PDB:5UH6) , including a 14-base stretch of DNA, 2 RNA nucleotides, 2 zinc ions, a magnesium ion and a bound rifampicin molecule, was solvated with 114,838 waters and 127 sodium ions -the latter to ensure electrical neutrality -creating a cubic simulation unit cell of initial dimensions 20.1 x 15.2 x 13.1 nm. The flexible loop region of each gyrB protein that were not resolved in the structure of the M. tuberculosis DNA gyrase cleavage complex (PDB:5BS8) were modelled in using the ModLoop server . This structure, including the 19-base stretch of DNA, 4 Mg 2+ ions, 2 bound moxifloxacin molecules and 403 crystal waters was placed in a rhombic dodecahedron unit cell with dimensions 13.8 x 13.8 x 9.8 x 0.0 x 0.0 x 0.0 x 0.0 x 6.9 x 6.9 nm. The unit cell was solvated with 59,895 waters, 175 Na + and 112 Cl -ions providing electrical neutrality and a 100mM salt concentration. The generalized AMBER and AMBER ff99SB-idln forcefields were used throughout . To facilitate the covalent bond between gyrA Tyr129 and the phosphate backbone of DNA by GROMACS, two modified amino acids (TYX and TYY) were created. These 'hybrid' amino acids contained the parameters for Tyr, excluding the hydroxyl hydrogen, all nucleotides in the covalently bound DNA chain and the covalent bond between the Tyr hydroxyl oxygen and the corresponding DNA backbone phosphorus atom. The PDB file order and residue naming was adjusted to reflect the modified amino acids. The system was left with a non-integer charge due to the exclusion of the hydrogen atom from Tyr, so a solvent chloride ion was modified to provide a balancing charge.
|
61b1d32a0e35eb99589a8774
| 6 |
The energies of the resulting RNA polymerase and DNA gyrase unit cells of 396,776 atoms and 205,883 atoms, respectively, were then minimized by GROMACS 2016.3 and 2018.2 respectively, using a steepest descent algorithm for 1,000 steps before being gradually warmed from 100 K to 310 K over 500 ps. For comparison, the DHFR unit cell only contained 27,115 atoms . The resulting structure and velocities were used to seed three RNAP and five DNAG equilibration simulations, each 50 ns long. The temperature was maintained at 310K using a Langevin thermostat with a time constant of 2 ps. An isotropic Parrinello-Rahman barostat with a 1 ps time constant and a compressibility of 4.46 x 10 -5 bar -1 was applied to keep the pressure at 1 bar. Electrostatic forces were calculated using the particle mesh Ewald algorithm with a real space cutoff of 1.2 nm whilst van der Waals forces were only calculated between atoms less than 1.2 nm apart with a switching function applied from 0.9 nm. The lengths of all bonds involving a hydrogen were constrained using LINCS , permitting a timestep of 2 fs. For DNA gyrase, to prevent the moxifloxacin coordinated Mg 2+ from dissociating from moxifloxacin, we used a harmonic distance restraint of sufficient strength (100,000 kJ mol -1 nm -2 ) to maintain the distance observed in the crystal structure (0.209 nm) throughout all simulations, lower values were not sufficient. A series of assumed independent structures were obtained by saving the coordinates of the system every 10 ns from each of three RNAP equilibration simulations and each of five DNAG equilibration simulations.
|
61b1d32a0e35eb99589a8774
| 7 |
Mutations were then introduced into each of these structures using pmx . To reduce the likelihood of clashes between the 'new' sidechain and the remainder of the protein (i.e. in simulations with λ~1) we then applied a short Alchembed procedure to each structure -this involved a 1,000 step simulation where λ was increased from 0 to 1 using a soft-core van der Waals potential. This created a pool of presumed independent mutated structures that could be used to seed alchemical thermodynamic integration simulations.
|
61b1d32a0e35eb99589a8774
| 8 |
Following best practice , the free energies (Fig. ) required to remove the electrical charge on the perturbing atoms (DG qoff ), transmute the van der Waals parameters (DG vdW ) and recharge the remaining atoms (DG qon ) were separately calculated using GROMACS 2016.3 for RNAP and 2019.1 for DNAG. Each calculation required eight simulations at equally spaced values of the progress parameter, λ. To accelerate convergence, 10,000 replica exchanges were attempted between neighboring λ-simulations every 1,000 timesteps. The process was repeated for both apo and complexed forms of either the RNAP or DNAG, thereby resulting in six independent free energies (Fig. ). The timestep was reduced from 2 fs to 1 fs and LINCS constraints were removed for the vdW transitions for all DNA gyrase mutations and the qon transition of gyrA D94G to prevent crashing. To ensure the drug remained bound, a harmonic distance-based potential with spring constant 1,000 kJ mol -1 nm -2 was applied between the centers of mass of the drug and the RNAP beta subunit. Two additional free energies describing the cost of removing this restraint (Fig. ) were then also calculated.
|
61b1d32a0e35eb99589a8774
| 9 |
In previous studies of S. aureus DHFR all alchemical free energies were repeated the same number of times which, since n values of the final difference in binding free energy (ΔΔG) were then obtained, simplified the calculation of errors. Both simulation unit cells studied here were over an order of magnitude larger and we therefore instead calculated errors at the level of an individual alchemical free energy (e.g. DG vdW ), with the final error in ΔΔG estimated by adding these in quadrature. Throughout a 95% confidence limit was estimated by multiplying the standard error by the appropriate t-statistic. We arbitrarily decided that at least three independent values of each alchemical free energy would be calculated, and then additional repeats would be run with the aim of reducing the magnitude of the overall 95% confidence limit to less than 1 kcal/mol. Achieving the latter was not always possible even when large numbers of repeats were run (n ≥ 10, see Supplementary Information).
|
61b1d32a0e35eb99589a8774
| 10 |
To study the five DNA gyrase mutations, a total of 231 alchemical free energies were calculated (8x λ simulations 0.5 ns long) and including equilibration simulations, a total of 1.17 µs of molecular dynamics simulations were initially performed. As described later, for DNA gyrase, nine calculations were extended to 5 ns which increased the total molecular dynamics performed to 1.49 µs. To avoid equilibration effects, the first 0.25 ns of each λ simulation was discarded. removing the electrical charge from atoms being perturbed, followed by transforming their van der Waals parameter, before finally recharging the atoms being perturbed. Double headed arrows represent the restraint used to prevent rifampicin from leaving the binding pocket. In all cases we are making use of the fact that free energy is a state function and therefore we can write the difference binding free energy (ΔΔG binding ) as a sum of so-called alchemical free energies (e.g. ΔG 4 -ΔG 3 ).
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.