Exploring thermal expansion of carbon-based nanosheets by machine-learning interatomic potentials

Property structure relationships in materials can be studied by a number of computational approaches such as ab initio quantum chemistry calculations, ab initio molecular dynamics (MD) simulations, classical MD and Monte-Carlo simulations, finite element modelling, etc. A choice of the computational method depends on the time and length scales and computational resources available. At the current stage of method and hardware development, ab initio quantum chemistry calculations are best suited for studying energy-structure relationships in relatively small systems consisting of tens of atoms. Ab initio MD simulations allow one to study dynamics of systems on a picosecond time scale for systems consisting of hundreds of atoms. Energy and forces in ab initio MD simulations are obtained from solving the electronic structure problem “on the fly”. Parameterization of the energy a system as a function of the relative atom positions, e. g., development of a classical force field, significantly speeds up calculations of the energies and forces in MD simulations, positioning classical MD simulations as the most suitable tool to obtain properties of the systems containing 10-10 atoms on the time scales from 10 to 10 s. However, the value of the property predictions using classical MD simulations is limited to the accuracy of the force field used, making a consistent derivation of a high quality classical force fields central to accurate prediction of the property-structure relationship from MD simulations.

High-temperature thermal stability, elastic moduli and anisotropy are among the key properties, which are used in selecting materials for cutting and machining applications. The high computational demand of ab initio molecular dynamics (AIMD) simulations in calculating elastic constants of alloys promotes the development of alternative approaches. Machine learning concept grasped as hybride classical molecular dynamics and static first principles calculations have several orders less computational costs. Here we prove the applicability of the concept considering the recently developed moment tensor potentials (MTP), where moment tensors are used as material’s descriptors which can be trained to predict the elastic constants of the prototypical hard coating alloy, Ti0.5Al0.5N at 900 K. We demonstrate excellent agreement between classical molecular dynamics simulations with MTPs and AIMD simulations. Moreover, we show that using MTPs one overcomes the inaccuracy issues present in approximate AIMD simulations of elastic constants of alloys.

The shock Hugoniot for full-density and porous ${\mathrm{CeO}}_{2}$ was investigated in the liquid regime using ab initio molecular dynamics (AIMD) simulations with Erpenbeck's approach based on the Rankine-Hugoniot jump conditions. The phase space was sampled by carrying out NVT simulations for isotherms between 6000 and 100 000 K and densities ranging from $\ensuremath{\rho}=2.5$ to $20\phantom{\rule{0.28em}{0ex}}\mathrm{g}/{\mathrm{cm}}^{3}$. The impact of on-site Coulomb interaction corrections $+U$ on the equation of state (EOS) obtained from AIMD simulations was assessed by direct comparison with results from standard density functional theory simulations. Classical molecular dynamics (CMD) simulations were also performed to model atomic-scale shock compression of larger porous ${\mathrm{CeO}}_{2}$ models. Results from AIMD and CMD compression simulations compare favorably with Z-machine shock data to 525 GPa and gas-gun data to 109 GPa for porous ${\mathrm{CeO}}_{2}$ samples. Using results from AIMD simulations, an accurate liquid-regime Mie-Gr\"uneisen EOS was built for ${\mathrm{CeO}}_{2}$. In addition, a revised multiphase SESAME-type EOS was constrained using AIMD results and experimental data generated in this work. This study demonstrates the necessity of acquiring data in the porous regime to increase the reliability of existing analytical EOS models.

Ab initio and classical molecular dynamics studies of electrode reactions

It is well-known that the calculation of thermal conductivity using classical molecular dynamics (MD) simulations strongly depends on the choice of the appropriate interatomic potentials. As proven for the case of graphene, while most of the available interatomic potentials estimate the structural and elastic constants with high accuracy, when employed to predict the lattice thermal conductivity they however lead to a variation of predictions by one order of magnitude. Here we present our results on using machine-learning interatomic potentials (MLIPs) passively fitted to computationally inexpensive ab-initio molecular dynamics trajectories without any tuning or optimizing of hyperparameters. These first-attempt potentials could reproduce the phononic properties of different two-dimensional (2D) materials obtained using density functional theory (DFT) simulations. To illustrate the efficiency of the trained MLIPs, we consider polyaniline C3N nanosheets. C3N monolayer was selected because the classical MD and different first-principles results contradict each other, resulting in a scientific dilemma. It is shown that the predicted thermal conductivity of 418 ± 20 W mK−1 for C3N monolayer by the non-equilibrium MD simulations on the basis of a first-attempt MLIP evidences an improved accuracy when compared with the commonly employed MD models. Moreover, MLIP-based prediction can be considered as a solution to the debated reports in the literature. This study highlights that passively fitted MLIPs can be effectively employed as versatile and efficient tools to obtain accurate estimations of thermal conductivities of complex materials using classical MD simulations. In response to remarkable growth of 2D materials family, the devised modeling methodology could play a fundamental role to predict the thermal conductivity.

Lithium thiophosphate electrolyte is a promising material for application in all-solid-state batteries. Ab initio molecular dynamics (AIMD) simulations have been used to investigate the ion conduction mechanisms in single-crystalline and glassy compounds. However, the complexity of real materials (e.g., materials with grain boundaries and multiphase glass-ceramics) causes AIMD simulations to have high computational cost. To overcome this computational limitation, we developed a new interatomic potential for classical molecular dynamics (CMD) simulations of Li solid-state electrolytes. The training datasets were generated from representative sulfide electrolytes (β-Li3PS4, γ-Li3PS4, Li4P2S6, Li7P3S11, and Li7PS6 crystals and 70Li2S-30P2S5 glass). Using the functional forms of the Class II and Stillinger-Weber potentials, all parameters were optimized by minimizing the differences in forces on atoms, stresses, and potential energies between the CMD and AIMD results. Subsequent validation showed that the optimized parameters can reproduce the dynamics of Li+ as well as the structures of the crystalline and glassy materials. The ionic conductivity of Li7P3S11 crystal was approximately five times that of the isostoichiometric 70Li2S-30P2S5 glass, indicating that CMD simulations using the developed force-field accurately reproduced the effective conduction path in Li7P3S11 from AIMD. The developed force-field parameters make it possible to simulate complex materials including amorphous-crystalline interfaces and multiphase glass-ceramics in the CMD framework.

Results of ab initio molecular dynamics (AIMD), quantum mechanics/molecular mechanics (QM/MM), and classical molecular dynamics (CMD) simulations of Cm(3+) in liquid water at a temperature of 300 K are reported. The AIMD simulation was based on the Car-Parrinello MD scheme and GGA-PBE formulation of density functional theory. Two QM/MM simulations were performed by treating Cm(3+) and the water molecules in the first shell quantum mechanically using the PBE (QM/MM-PBE) and the hybrid PBE0 density functionals (QM/MM-PBE0). Two CMD simulations were carried out using ab initio derived pair plus three-body potentials (CMD-3B) and empirical Lennard-Jones pair potential (CMD-LJ). The AIMD and QM/MM-PBE simulations predict average first shell hydration numbers of 8, both of which disagree with recent experimental EXAFS and TRLFS value of 9. On the other hand, the average first shell hydration numbers obtained in the QM/MM-PBE0 and CMD simulations was 9, which agrees with experiment. All the simulations predicted an average first shell and second shell Cm-O bond distance of 2.49-2.53 Å and 4.67-4.75 Å respectively, both of which are in fair agreement with corresponding experimental values of 2.45-2.48 and 4.65 Å. The geometric arrangement of the 8-fold and 9-fold coordinated first shell structures corresponded to the square antiprism and tricapped trigonal prisms, respectively. The second shell hydration number for AIMD QM/MM-PBE, QM/MM-PBE0, CMD-3B, and CMD-LJ, were 15.8, 17.2, 17.7, 17.4, and 16.4 respectively, which indicates second hydration shell overcoordination compared to a recent EXAFS experimental value of 13. Save the EXAFS spectra CMD-LJ simulation, all the computed EXAFS spectra agree fairly well with experiment and a clear distinction could not be made between configurations with 8-fold and 9-fold coordinated first shells. The mechanisms responsible for the first shell associative and dissociative ligand exchange in the classical simulations have been analyzed. The first shell mean residence time was predicted to be on the nanosecond time scale. The computed diffusion constants of Cm(3+) and water are in good agreement with experimental data.

Internal standard-assisted ab initio MD simulation for comparative thermal stability and decomposition mechanisms of energetic materials

Distribution of the mechanical properties of Ti–Cu combinatorial thin film evaluated using nanoindentation experiments and molecular dynamics with a neural network potential

We have performed extensive ab initio and classical molecular dynamics (MD) simulations of benzene in water in order to examine the unique solvation structures that are formed. Qualitative differences between classical and ab initio MD simulations are found and the importance of various technical simulation parameters is examined. Our comparison indicates that nonpolarizable classical models are not capable of describing the solute-water interface correctly if local interactions become energetically comparable to water hydrogen bonds. In addition, a comparison is made between a rigid water model and fully flexible water within ab initio MD simulations which shows that both models agree qualitatively for this challenging system.

Eliminating organic pollutant molecules in water is a world challenge, that non-thermal plasmas at atmospheric pressure are aiming to address. Non thermal plasmas are used for treating contaminated water due to their ability to produce reactive oxygen and nitrogen species, the so-called RONS, which diffuse into water, react with pollutant molecules, and degrade them. Many experimental works have been designed, using different plasma sources [1, 2], but in all cases HO● radicals are acknowledged to play the main role in the degradation process.Classical and ab-initio reactive molecular dynamics (MD) simulations are relevant techniques for describing interactions of HO● with pollutant molecules in water, leading to the knowledge of degradation products, chemical pathways and rates [3]. Basically, MD simulations are solving the Newton equations of motion of a set of species, say atoms, molecules, clusters, etc. It is able to describe system size up to billion of species. It only needs the knowledge of interactions (force fields) between all species as well as initial conditions, preferably consistent with experiments. So it is suitable for addressing chemical reactivity. Classical reactive MD simulations are using available semi-empirical force fields, such as reaxFF, while ab-initio MD are required when no such force field is available. In this case, force fields are calculated quantum mechanically at relevant time steps of MD simulations.In fact, MD simulations act as a predictive "virtual microscope" provided the force fields are enough precise.In this work, MD simulations of the interaction of HO● radicals with molecules representative of various emerging pollutant families, as pesticide, antibiotics, PFAS in water, are studied. For mimicking HO● radicals release into water by non-thermal plasma, HO● radicals are periodically injected into a simulation box containing a pollutant molecule surrounded by several tens of water molecules. For obtaining statistically significant results, at least 10 simulations, with different initial conditions are run. The products that are formed will be compared with experiments, especially HPLC, GCMS and MS/MS measurements of plasma treated water.Another approach, consists in introducing a temperature ramp during the simulation for determining at which temperature products are appearing and in which order. Such an approach is revealing the reaction barriers up to full decomposition of the pollutant molecule. It can thus, by back-analysis, help to define which plasma or other advanced oxidation processing will be relevant [3]. Acknowledgements Part of this work has been supported by Conseil Régional Centre-Val de Loire with grant #2021-00144786 for projectPerturb’Eau. Monica Magureanu, Corina Bradu, Florin Bilea, Olivier Aubry, Dunpin Hong, Manoj Panyamthatta-Tayaroth are gratefully acknowledged for stimulating discussions and for providing experimental results.

The dissociation mechanism and thermodynamic properties of HCl(aq) in hydrothermal fluids (to 700 °C, 60 kbar) by ab initio molecular dynamics simulations

We present the absolute enthalpy, entropy, heat capacity, and free energy of liquid water at ambient conditions calculated by the two-phase thermodynamic method applied to ab initio, reactive and classical molecular dynamics simulations. We find that the absolute entropy and heat capacity of liquid water from ab initio molecular dynamics (AIMD) is underestimated, but falls within the range of the flexible empirical as well as the reactive force fields. The origin of the low absolute entropy of liquid water from AIMD simulations is due to an underestimation of the translational entropy by 20% and the rotational entropy by 40% compared to the TIP3P classical water model, consistent with previous studies that reports low diffusivity and increased ordering of liquid water from AIMD simulations. Classical MD simulations with rigid water models tend to be in better agreement with experiment (in particular TIP3P yielding the best agreement), although the TIP4P-ice water model, the only empirical force field that reproduces the experimental melting temperature, has the lowest entropy, perhaps expectedly. This reiterates the limitations of existing empirical water models in simultaneously capturing the thermodynamics of solid and liquid phases. We find that the quantum corrections to heat capacity of water can be as large as 60%. Although certain water models are computed to yield good absolute free energies of water compared to experiments, they are often due to the fortuitous enthalpy-entropy cancellation, but not necessarily due to the correct descriptions of enthalpy and entropy separately.

Using ab initio molecular dynamics (AIMD) simulations, classical molecular dynamics (CMD) simulations, small‐angle X‐ray scattering (SAXS), and pulsed‐field gradient nuclear magnetic resonance (PFG‐NMR), the solvation structure and ion dynamics of magnesium bis(trifluoromethanesulfonyl)imide (Mg(TFSI)2) aqueous electrolyte at 1, 2, and 3 m concentrations are investigated. From AIMD and CMD simulations, the first solvation shell of an Mg2+ ion is found to be composed of six water molecules in an octahedral configuration and the solvation shell is rather rigid. The TFSI− ions prefer to stay in the second solvation shell and beyond. Meanwhile, the comparable diffusion coefficients of positive and negative ions in Mg(TFSI)2 aqueous electrolytes have been observed, which is mainly due to the formation of the stable [Mg(H2O)6]2+ complex, and, as a result, the increased effective Mg ion size. Finally, the calculated correlated transference numbers are lower than the uncorrelated ones even at the low concentration of 2 and 3 m, suggesting the enhanced correlations between ions in the multivalent electrolytes. This work provides a molecular‐level understanding of how the solvation structure and multivalency of the ion affect the dynamics and transport properties of the multivalent electrolyte, providing insight for rational designs of electrolytes for improved ion transport properties.

Using classical and ab initio molecular-dynamics (MD) simulations, we have studied a calcium aluminosilicate glass of composition (SiO2)0.67–(Al2O3)0.12–(CaO)0.21. Samples with 100 and 200 atoms were generated by classical MD simulations using a potential with 3-body interactions. Although we observe, for the model with 100 atoms, finite size effects for some structural properties, these effects are substantially reduced if the glass structure is refined by the ab initio MD simulations. In addition, some structural characteristics such as the Ca–O bond length and the angular distributions are improved by the ab initio description. The structural and vibrational characteristics of these glass samples are compared to that of a glass that has been quenched from the melt using first-principles simulations. The main differences are found on the SiOSi and SiOAl angular distributions and on the apparition of high-frequency bands in the partial Ca vibrational density of states for the classically generated glass samples.