First-principles NMR of oxide glasses boosted by machine learning.

Solid-state NMR has established itself as a cutting-edge spectroscopy for elucidating the structure of oxide glasses thanks to several decades of methodological and instrumental progress. First-principles calculations of NMR properties combined with molecular-dynamics (MD) simulations provides a powerful complementary approach for the interpretation of NMR data, although they still suffer from limitations in terms of size, time and high consumption of computational resources. We address this challenge by developing a machine-learning framework to boost predictive modelling of NMR spectra. We use kernel ridge regression techniques (least-squares support vector regression and linear ridge regression) combined with smooth overlap of atomic position (SOAP) atom-centered descriptors to efficiently predict NMR interactions: the isotropic magnetic shielding and the electric field gradient (EFG) tensor. As illustrated in this work, this approach enables the simulation of magic-angle spinning (MAS) and multiple-quantum magic-angle spinning (MQMAS) NMR spectra of very large models (more than 10 000 atoms) and an efficient averaging of NMR properties over MD trajectories of nanoseconds for incorporating finite-temperature effects, at the computational cost of classical MD simulations. We illustrate these advances for sodium silicate glasses (SiO2-Na2O). NMR parameters (isotropic chemical shift and electric field gradient) could be predicted with an accuracy of 1 to 2% in terms of the total span of the NMR parameter values. To include vibrational effects, an approach is proposed of scaling the EFG tensor in NMR simulations with a factor obtained from the time auto-correlation functions computed on MD trajectories.

Accelerating NMR Shielding Calculations Through Machine Learning Methods: Application to Magnesium Sodium Silicate Glasses.

In this work, we have applied the Kernel Ridge Regression (KRR) method using a Least Square Support Vector Regression (LSSVR) approach for the prediction of the NMR isotropic magnetic shielding (σiso) of active nuclei (17O, 23Na, 25Mg, and 29Si) in a series of (Mg, Na) - silicate glasses. The Machine Learning (ML) algorithm has been trained by mapping the local environment of each atom described by the Smooth Overlap of Atomic Position (SOAP) descriptor with isotropic chemical shielding values computed with DFT using the Gauge-Included-Projector-Augmented-Wave (GIPAW) approach. The influence of different training datasets generated through molecular dynamics simulations at various temperatures and with different inter-atomic potentials has been tested and we demonstrate the importance of a wide exploration of the configurational space to enhance the transferability of the ML-regressor.  Finally, the trained ML-regressor has been used to simulate the 29Si MAS NMR spectra of systems containing up to 20000 atoms by averaging hundreds of configurations extracted from classical MD simulations to account for thermal vibrations. This ML approach is a powerful tool for the interpretation of NMR spectra using relatively large systems at a fraction of the computational time required by quantum mechanical calculations which are of high computational cost.

Correlated Anion Disorder in Heteroanionic Cubic TiOF2.

Resolving anion configurations in heteroanionic materials is crucial for understanding and controlling their properties. For anion-disordered oxyfluorides, conventional Bragg diffraction cannot fully resolve the anionic structure, necessitating alternative structure determination methods. We have investigated the anionic structure of anion-disordered cubic (ReO3-type) TiOF2 using X-ray pair distribution function (PDF), 19F MAS NMR analysis, density functional theory (DFT), cluster expansion modeling, and genetic-algorithm structure prediction. Our computational data predict short-range anion ordering in TiOF2, characterized by predominant cis-[O2F4] titanium coordination, resulting in correlated anion disorder at longer ranges. To validate our predictions, we generated partially disordered supercells using genetic-algorithm structure prediction and computed simulated X-ray PDF data and 19F MAS NMR spectra, which we compared directly to experimental data. To construct our simulated 19F NMR spectra, we derived new transformation functions for mapping calculated magnetic shieldings to predicted magnetic chemical shifts in titanium (oxy)fluorides, obtained by fitting DFT-calculated magnetic shieldings to previously published experimental chemical shift data for TiF4. We find good agreement between our simulated and experimental data, which supports our computationally predicted structural model and demonstrates the effectiveness of complementary experimental and computational techniques in resolving anionic structure in anion-disordered oxyfluorides. From additional DFT calculations, we predict that increasing anion disorder makes lithium intercalation more favorable by, on average, up to 2 eV, highlighting the significant effect of variations in short-range order on the intercalation properties of anion-disordered materials.

Accurate and Transferable Machine Learning Potential for Molecular Dynamics Simulation of Sodium Silicate Glasses.

An accurate and transferable machine learning (ML) potential for the simulation of binary sodium silicate glasses over a wide range of compositions (from 0 to 50% Na2O) was developed. The potential energy surface is approximated by the sum of atomic energy contributions mapped by a neural network algorithm from the local geometry comprising information on atomic distances and angles with neighboring atoms using the DeePMD code [Wang, H. Comput. Phys. Commun. 2018, 228, 178-184]. Our model was trained on a large data set of total energies and atomic forces computed at the density functional theory level on structures extracted from classical molecular dynamics (MD) simulations performed at several temperatures from 300 to 3000 K. This allows for the generation of a robust and transferable ML potential applicable over the full compositional range of glass formability at different temperatures that outperforms the empirical potentials available in the literature in reproducing structures and properties such as bond angle distribution, total distribution functions, and vibrational density of state. The generality of the approach enables the future training of a potential with other or more elements allowing for simulations of structures, properties, and behavior of ternary and multicomponent oxide glasses with nearly ab initio accuracy at a fraction of the computational cost.

Structural Modeling of O/F Correlated Disorder in TaOF3 and NbOF3-x(OH)x by Coupling Solid-State NMR and DFT Calculations.

The structure of MOF3 (M = Nb, Ta) compounds was precisely modeled by combining powder X-ray diffraction, solid-state NMR spectroscopy, and semiempirical dispersion-corrected DFT calculations. It consists of stacked ∞(MOF3) layers along the c⃗ direction formed by heteroleptic corner-connected MX6 (X = O, F) octahedra. 19F NMR resonance assignments and occupancy rates of the anionic crystallographic sites have been revised. The bridging site is shared equally by the anions, and the terminal site is occupied by F only. An O/F correlated disorder is expected since cis-MO2F4 octahedra are favored, resulting in one-dimensional -F-M-O-M- strings along the <100> and <010> directions. Ten different 2 × 2 × 1 supercells per compound, fulfilling these characteristics, were built. Using DFT calculations and the GIPAW approach, the supercells were relaxed and the 19F isotropic chemical shift values were determined. The agreement between the experimental and calculated 19F spectra is excellent for TaOF3. The 1H and 19F experimental NMR spectra revealed that some of the bridging F atoms are substituted by OH groups, especially in NbOF3. New supercells involving OH groups were generated. Remarkably, the best agreement is obtained for the supercells with the composition closest to that estimated from the 19F NMR spectra, i.e., NbOF2.85(OH)0.15.

Behaviors of sodium and calcium ions at the borosilicate glass-water interface: Gaining new insights through an ab initio molecular dynamics study.

We study reactivity and leaching at the calcium sodium borosilicate (CNBS)-water interface by means of a Car-Parrinello ab initio molecular dynamics simulation over a simulation time of 100 ps. With an emphasis on the comparison between the behaviors of Ca2+ and Na+ cations at the CNBS glass-water interface, different mechanism events during the trajectory are revealed, discussed, and correlated with other density functional theory calculations. We show that Na+ ions can be released in solution, while Ca2+ cannot leave the surface of CNBS glass. This release is correlated with the vacancy energy of Ca2+ and Na+ cations. Here, we found that the CNBS structure with the Na+ cation vacancy is energetically more favorable than the structure with the Ca2+ cation vacancy. The calcium adsorption site has been shown to have a greater affinity for water than can be found in the case of the sodium site, demonstrating that affinity may not be considered a major factor controlling the release of cations from the glass to the solution.

Systematic Method for the Exploration, Representation, and Classification of the Diphenylalanine Solvatomorphic Space.

An understanding of the conditions that govern the self-assembly process of peptides is a fundamental step toward the design of new nanostructures that possess interesting properties. In this work, we first synthesize and explore extensively diphenylalanine (FF) self-assembling crystals formed in different solvents (i.e., solvatomorphs) using polarized optical microscopy and transmission electron microscopy. Then, we develop a numerical method that allows an unambiguous classification of the solvatomorphs through a K-means automatic clustering method. In addition, we generate a two-dimensional (2D) representation of the solvatomorphic space together with the clustering results via a principal component analysis (PCA). The classification is based on structural similarities of solvatomorphs as revealed by the analysis of their respective infrared spectra. Among the 20 samples considered, 4 clear clusters are extracted within which the compounds show very similar crystalline structures. The information extracted allows us to assign many of the peaks that appear in the complex IR spectra of the samples considered. The implementation of the overall procedure we propose, i.e., "GAULOIS" and "REFRACT-R", is transferable to other types of spectra and paves the way for a systematic, fast, and accurate classification method applicable to various types of experimental spectroscopic data.

Confined water radiolysis in aluminosilicate nanotubes: the importance of charge separation effects.

Imogolite nanotubes are potentially promising co-photocatalysts because they are predicted to have curvature-induced, efficient electron-hole pair separation. This prediction has however not yet been experimentally proven. Here, we investigated the behavior upon irradiation of these inorganic nanotubes as a function of their water content to understand the fate of the generated electrons and holes. Two types of aluminosilicate nanotubes were studied: one was hydrophilic on its external and internal surfaces (IMO-OH) and the other had a hydrophobic internal cavity due to Si-CH3 bonds (IMO-CH3), with the external surface remaining hydrophilic. Picosecond pulse radiolysis experiments demonstrated that the electrons are efficiently driven outward. For imogolite samples with very few external water molecules (around 1% of the total mass), quasi-free electrons were formed. They were able to attach to a water molecule, generating a water radical anion, which ultimately led to dihydrogen. When more external water molecules were present, solvated electrons, precursors of dihydrogen, were formed. In contrast, holes moved towards the internal surface of the tubes. They mainly led to the formation of dihydrogen and of methane in irradiated IMO-CH3. The attachment of the quasi-free electron to water was a very efficient process and accounted for the high dihydrogen production at low relative humidity values. When the water content increased, electron solvation dominated over attachment to water molecules. Electron solvation led to dihydrogen production, albeit to a lesser extent than quasi-free electrons. Our experiments demonstrated the spontaneous curvature-induced charge separation in these inorganic nanotubes, making them very interesting potential co-photocatalysts.

Bleeding risk of ticagrelor compared to clopidogrel in intensive care unit patients with acute coronary syndrome: A propensity-score matching analysis.

BACKGROUND: Intensive care unit (ICU) patients with the most severe forms of acute coronary syndrome (ACS) require invasive therapies such as extracorporeal life support. The risk of bleeding in ICU patients with ACS treated with a dual antiplatelet therapy of aspirin and ticagrelor is unknown. The primary objective of this study was to compare the bleeding risk of ticagrelor and clopidogrel in ICU patients with ACS. METHODS AND FINDINGS: We conducted a retrospective study based on a propensity score and a proportional hazards model. All patients with ACS hospitalized in the ICU of a French university hospital between January 2013 and January 2017 were included in the study. Bleeding during ICU stay was defined as all Thrombolysis in myocardial infarction (TIMI) major or minor events. A total of 155 patients were included in the study. According to propensity score matching, 57 patients treated with aspirin and ticagrelor were matched with 57 patients treated with aspirin and clopidogrel. Median (first-third quartile) Simplified Acute Physiology Score II was 61.5 (41.0-85.0). Bleeding during ICU stay occurred in 12 patients (21.1%) treated with clopidogrel and in 35 patients (61.4%) treated with ticagrelor (p<0.0001). This significant association was found for both TIMI major bleeding (12.3% vs. 35.1%, p = 0.004) and TIMI minor bleeding (8.8% vs. 26.3%, p = 0.01). The relative risk of bleeding occurrence during ICU stay was 2.60 (confidence interval 95%: 1.55-4.35) for ticagrelor compared to clopidogrel. No significant difference in ICU mortality was found between the two groups (45.6% in the clopidogrel group vs. 29.8% in the ticagrelor group, p = 0.08). CONCLUSIONS: Bleeding complications are frequent and serious in ICU patients with ACS. A dual antiplatelet therapy of aspirin and ticagrelor is associated with a higher risk of bleeding compared to a dual antiplatelet therapy of aspirin and clopidogrel.

31P MAS NMR and DFT study of crystalline phosphate matrices.

We present the study of the phosphorus local environment by using 31P MAS NMR in a series of seven double monophosphates MIIMIV(PO4)2 (MII and MIV being divalent and tetravalent cations, respectively) of yavapaiite and low-yavapaiite type crystal structures. Solid-state and cluster DFT calculations were found to be efficient for predicting the 31P isotropic chemical shift and chemical shift anisotropy. To achieve this performance, however, a proper computational optimisation of the experimental structural data was required. From the three optimisation methods tested, the full optimisation provided the best reference structure for the calculation of the NMR parameters of the studied phosphates. Also, a better prediction of the chemical shifts was possible by using a correction to the GIPAW calculated shielding.

NMR shifts in aluminosilicate glasses via machine learning.

Machine learning (ML) approaches are investigated for the prediction of nuclear magnetic resonance (NMR) parameters in aluminosilicate glasses, for which NMR has proven to be a cutting-edge method over the last decade. DFT computations have emerged as a new dimension for complementing these NMR methods although suffering from severe limitations in terms of size, time and computational resources consumption. While previous approaches tend to use DFT-GIPAW calculations for the prediction of NMR parameters in glassy systems, we propose to employ ML methods, characterized by a speed similar to that of classical molecular dynamics while the accuracy of ab initio methods can be reached. We design ML procedures to predict the isotropic magnetic shielding (σiso) for different multicomponent relevant glass compositions. The ML predictions of σiso deviate from DFT-GIPAW calculations, when including relaxed and room-temperature structures, by 0.7 ppm for 29Si (1.0% of the total span of the calculated ) and 1.5 ppm for 17O (1.9%) in SiO2 glasses, 1.4 ppm for 23Na (1.5%) in Na2O-SiO2 and 1.5 ppm for 27Al (2.1%) in Al2O3-Na2O-SiO2 systems. We compare the performances obtained for a set of three descriptors suitable for encoding atomic local environments information (atom-centered representations) together with seven popular ML algorithms with a focus on the simple (but robust) linear ridge regression (LRR) and the popular smooth overlap of atomic positions (SOAP) descriptor.

Using hydrogen isotope incorporation as a tool to unravel the surfaces of hydrogen-treated nanodiamonds.

We report here on a robust and easy-to-implement method for the labelling of detonation nanodiamonds (DND) with hydrogen isotopes (deuterium and tritium), using thermal annealing performed in a closed system. With this method, we have synthesized and fully characterized (FTIR, Raman, DLS, 3H/2H/1H and 13C MAS NMR) deuterium-treated and tritium-treated DND and demonstrated the usefulness of isotope incorporation in investigating the surface chemistry of such nanomaterials. For instance, surface treatment with deuterium coupled to FTIR spectroscopy allowed us to discriminate the origin of C-H terminations at the DND surface after the hydrogenation process. As a complementary, tritium appeared very useful for quantification purposes, while 1,2,3H NMR confirmed the nature of the C-1,2,3H bonds created. This isotopic study provides new insights into the characteristics of hydrogen-treated DND.

Partial Transformation of Imogolite by Decylphosphonic Acid Yields an Interface Active Composite Material.

The phosphonic acid moiety is commonly used as an anchoring group for the surface modification of imogolite. However, the impact of the reaction on its structure has never been clearly analyzed before. We study the reaction of imogolite and decylphosphonic acid by combining infrared spectroscopy, X-ray scattering, scanning electron microscopy, transmission electron microscopy, and solid-state nuclear magnetic resonance spectroscopy. Instead of a surface functionalization, we observe the formation of a lamellar phase interconnected with imogolite bundles. Although we find no evidence for grafted imogolite tubes, we observe the expected dispersion characteristics and stabilization of water in toluene emulsions described in the literature. Based on the surface chemistry of imogolite, we propose an explanation for the observed reactivity and link the structural features of the obtained composite material to its dispersibility in toluene and its observed properties at the toluene-water interface.

Insight into the Crystalline Structure of ThF4 with the Combined Use of Neutron Diffraction, 19F Magic-Angle Spinning-NMR, and Density Functional Theory Calculations.

Because of its sensitivity to the atomic scale environment, solid-state NMR offers new perspectives in terms of structural characterization, especially when applied jointly with first-principles calculations. Particularly, challenging is the study of actinide-based materials because of the electronic complexity of the actinide cations and to the hazards due to their radioactivity. Consequently, very few studies have been published in this subfield. In the present paper, we report a joint experimental-theoretical analysis of thorium tetrafluoride, ThF4, containing a closed-shell actinide (5f0) cation. Its crystalline structure has been revisited in the present work using powder neutron diffraction experiments. The 19F NMR parameters of the seven F crystallographic sites have been modeled using an empirical superposition model, periodic first-principles calculations, and a cluster-based all-electron approach. On the basis of the atomic position optimized structure, a complete and unambiguous assignment of the 19F NMR resonances to the F sites has been obtained.

Structure of Strontium Aluminosilicate Glasses from Molecular Dynamics Simulation, Neutron Diffraction, and Nuclear Magnetic Resonance Studies.

The structure of strontium glasses with the composition (SiO2)1-2 x(Al2O3) x(SrO) x ( R = [SrO]/[Al2O3] = 1) and (SiO2)1-4 x(Al2O3) x(SrO)3 x ( R = 3) has been explored experimentally over both short- and intermediate-length scales using neutron diffraction, 27Al and 29Si nuclear magnetic resonance, and classical molecular dynamics simulations in model systems containing around 10 000 atoms. We aim at understanding the structural role of aluminum and strontium as a function of the chemical composition of these glasses. The short- and medium-range structure such as aluminum coordination, bond angle distribution, Q( n) distribution, and oxygen speciation have been systematically studied. Two potential forms of the repulsive short-range interactions have been investigated, namely, the Buckingham and Morse forms. The comparison of these forms allows us to derive general trends independent of the particular choice of the potential form. In both cases, it is found that aluminum ions are mainly fourfold coordinated and mix with the silicon network favoring the Al/Si mixing in terms of Al-O-Si linkages. For the R = 1 glass series, despite the full charge compensation ([SrO] = [Al2O3]), a small fraction of fivefold aluminum is observed both experimentally and in MD simulations, whereas the concentration of sixfold aluminum is negligible. MD shows that the fivefold aluminum units AlO5 preferentially adopt a small ring configuration and link to tricoordinated oxygen atoms whose population increases with the aluminum content and are mainly found in OAl3 and OAl2Si configurations. The modeled Sr speciation mainly involves SrO7 and SrO8 polyhedra, giving a range of average Sr2+ coordination numbers between 7 and 8 slightly dependent on the short-range repulsive potential form. A detailed statistical analysis of T-O-T' (T, T' = Al,Si), accounting for the population of the various oxygen speciations, reveals that both potentials predict a nearly identical Al/Si mixing.

Structural study of Na2O-B2O3-SiO2 glasses from molecular simulations using a polarizable force field.

Sodium borosilicate glasses Na2O-B2O3-SiO2 (NBS) are complex systems from a structural point of view. Three main building units are present: tetrahedral SiO4 and BO4 (BIV) and triangular BO3 (BIII). One of the salient features of these compounds is the change of the BIII/BIV ratio with the alkali concentration, which is very difficult to capture in force fields-based molecular dynamics simulations. In this work, we develop a polarizable force field that is able to reproduce the boron coordination and more generally the structure of several NBS systems in the glass and in the melt. The parameters of the potential are fitted from density functional theory calculations only, in contrast with the existing empirical potentials for NBS systems. This ensures a strong improvement on the transferability of the parameters from one composition to another. Using this new force field, the structure of NBS systems is validated against neutron diffraction and nuclear magnetic resonance experiments. A special focus is given to the distribution of BIII/BIV with respect to the composition and the temperature.

Structure and Chemical Durability of Lead Crystal Glass.

Silicate glasses containing lead, also called lead crystal glasses, are commonly used as food product containers, in particular for alcoholic beverages. Lead's health hazards require major attention, which can first be investigated through the understanding of Pb release mechanisms in solution. The behavior of a commercial crystal glass containing 10.6 mol % of PbO (28.3 wt %) was studied in a reference solution of 4% acetic acid at 22, 40, and 70 °C at early and advanced stages of reaction. High-resolution solid-state 17O and 29Si NMR was used to probe the local structure of the pristine and, for the first time, of the altered lead crystal glass. Inserted into the vitreous structure between the network formers as Si-O-Pb bonds, Pb does not form Pb-O-Pb clusters which are expected to be more easily leached. A part of K is located near Pb, forming mixed Si-O-(Pb,K) near the nonbridging oxygens. Pb is always released into the solution following a diffusion-controlled dissolution over various periods of time, at a rate between 1 and 2 orders of magnitude lower than the alkalis (K and Na). The preferential release of alkalis is followed by an in situ repolymerization of the silicate network. Pb is only depleted in the outermost part of the alteration layer. In the remaining part, it stays mainly surrounded by Si in a stable structural configuration similar to that of the pristine glass. A simple model is proposed to estimate the Pb concentration as a function of glass surface, solution volume, temperature, and contact time.

Collective water dynamics in the first solvation shell drive the NMR relaxation of aqueous quadrupolar cations.

Using molecular simulations, we analyze the microscopic processes driving the Nuclear Magnetic Resonance (NMR) relaxation of quadrupolar cations in water. The fluctuations of the Electric Field Gradient (EFG) experienced by alkaline and magnesium cations, which determine the NMR relaxation time, are mainly due to the dynamics of water molecules in their solvation shell. The dynamics of the ion plays a less important role, with the exception of the short-time dynamics in the lighter Li+ case, for which rattling in the solvent cage results in oscillations of the EFG autocorrelation function (ACF). Several microscopic mechanisms that may a priori contribute to the decay of the EFG-ACF occur in fact over too long time scales: entrance/exit of individual water molecules into/from the solvation shell, rotation of a molecule around the ion, or reorientation of the molecule. In contrast, the fluctuations of the ion-water distance are clearly correlated to that of the EFG. Nevertheless, it is not sufficient to consider a single molecule due to the cancellations arising from the symmetry of the solvation shell. The decay of the EFG-ACF, hence NMR relaxation, is in fact governed by the collective symmetry-breaking fluctuations of water in the first solvation shell.

Self-healing capacity of nuclear glass observed by NMR spectroscopy.

Safe management of high level nuclear waste is a worldwide significant issue for which vitrification has been selected by many countries. There exists a crucial need for improving our understanding of the ageing of the glass under irradiation. While external irradiation by ions provides a rapid simulation of damage induced by alpha decays, short lived actinide doping is more representative of the reality. Here, we report radiological NMR experiments to compare the damage in International Simplified Glass (ISG) when irradiated by these two methods. In the 0.1 mole percent (244)Cm doped glass, accumulation of high alpha decay only shows small modifications of the local structure, in sharp contrast to heavy ion irradiation. These results reveal the ability of the alpha particle to partially repair the damage generated by the heavy recoil nuclei highlighting the radiation resistance of nuclear glass and the difficulty to accurately simulate its behaviour by single ion beam irradiations.

Supercell program: a combinatorial structure-generation approach for the local-level modeling of atomic substitutions and partial occupancies in crystals.

BACKGROUND: Disordered compounds are crucially important for fundamental science and industrial applications. Yet most available methods to explore solid-state material properties require ideal periodicity, which, strictly speaking, does not exist in this type of materials. The supercell approximation is a way to imply periodicity to disordered systems while preserving "disordered" properties at the local level. Although this approach is very common, most of the reported research still uses supercells that are constructed "by hand" and ad-hoc. RESULTS: This paper describes a software named supercell, which has been designed to facilitate the construction of structural models for the description of vacancy or substitution defects in otherwise periodically-ordered (crystalline) materials. The presented software allows to apply the supercell approximation systematically with an all-in-one implementation of algorithms for structure manipulation, supercell generation, permutations of atoms and vacancies, charge balancing, detecting symmetry-equivalent structures, Coulomb energy calculations and sampling output configurations. The mathematical and physical backgrounds of the program are presented, along with an explanation of the main algorithms and relevant technical details of their implementation. Practical applications of the program to different types of solid-state materials are given to illustrate some of its potential fields of application. Comparisons of the various algorithms implemented within supercell with similar solutions are presented where possible. CONCLUSIONS: The all-in-one approach to process point disordered structures, powerful command line interface, excellent performance, flexibility and GNU GPL license make the supercell program a versatile set of tools for disordered structures manipulations.