Learning neural network potentials from experimental data via Differentiable Trajectory Reweighting

In molecular dynamics (MD), neural network (NN) potentials trained bottom-up on quantum mechanical data have seen tremendous success recently. Top-down approaches that learn NN potentials directly from experimental data have received less attention, typically facing numerical and computational challenges when backpropagating through MD simulations. We present the Differentiable Trajectory Reweighting (DiffTRe) method, which bypasses differentiation through the MD simulation for time-independent observables. Leveraging thermodynamic perturbation theory, we avoid exploding gradients and achieve around 2 orders of magnitude speed-up in gradient computation for top-down learning. We show effectiveness of DiffTRe in learning NN potentials for an atomistic model of diamond and a coarse-grained model of water based on diverse experimental observables including thermodynamic, structural and mechanical properties. Importantly, DiffTRe also generalizes bottom-up structural coarse-graining methods such as iterative Boltzmann inversion to arbitrary potentials. The presented method constitutes an important milestone towards enriching NN potentials with experimental data, particularly when accurate bottom-up data is unavailable.

Metal Hydride Technologies for Separation of Hydrogen Isotopes

The paper considers the issue of theoretical prediction of characteristics at which separation of hydrogen isotopes occurs in the “gas-metal” system. Mathematical modeling of sorption is based on the use of the lattice gas model often used for metal hydrides. In contrast to the Leicher ideal solution model, it is taken into account that the dissolution of hydrogen in the metal increases the volume of the crystal lattice. This leads to additional contributions to potential energy. The model also takes into account the interaction between the atoms of the incorporated (absorbed) hydrogen isotopes. These phenomena are described by the methods of thermodynamic perturbation theory. The different composition of the gas (protium and deuterium) in contact with the metal leads to the fact that sorption, accompanied by the formation of hydride, proceeds for the same temperature at different equilibrium pressures. This phenomenon characterizes the isotope effect. The temperature dependences of the pressure on the "plateau" for palladium hydride and deuteride are obtained. The differences in these pressures can be used for the practical use of metal hydrides in the separation of hydrogen isotopes.

Thermodynamic Properties of the Parabolic-Well Fluid

The thermodynamic properties of the parabolic-well fluid are considered. The intermolecular interaction potential of this model, which belongs to the class of the so-called van Hove potentials, shares with the square-well and the triangular well potentials the inclusion of a hard-core and an attractive well of relatively short range. The analytic second virial coefficient for this fluid is computed explicitly and an equation of state is derived with the aid of the second-order thermodynamic perturbation theory in the macroscopic compressibility approximation and taking the hard-sphere fluid as the reference system. For this latter, the fully analytical expression of the radial distribution function, consistent with the Carnahan-Starling equation of state as derived within the rational function approximation method, is employed. The results for the reduced pressure of the parabolic-well fluid as a function of the packing fraction and two values of the range of the parabolic-well potential at different temperatures are compared with Monte Carlo and Event‐driven molecular dynamics simulation data. Estimates of the values of the critical temperature are also provided.

Methanol Carbonylation over Acid Mordenite: Insights from Ab Initio Molecular Dynamics and Machine Learning Thermodynamic Perturbation Theory

<div>In this work we present a detailed \textit{ab initio} study of the carbonylation reaction of methoxy groups in the zeolite mordenite, as it is the rate determining step in a series of elementary reactions leading to ethanol. </div><div>For the first time we employ full molecular dynamics simulations to evaluate free energies of activation for the reactions in side pockets and main channels. Results show that the reaction in the side pocket is preferred and, when dispersion interactions are taken into account, this preference becomes even stronger. This conclusion is confirmed using multiple levels of density functional theory approximations with (PBE-D2, PBE-MBD, and vdW-DF2-B86R) or without (PBE, HSE06) dispersion corrections. These calculations, that in principle would require several demanding molecular dynamics simulations, were made possible at a minimal computational cost by using a newly developed approach that combines thermodynamic perturbation theory with machine learning.</div>

Methanol Carbonylation over Acid Mordenite: Insights from Ab Initio Molecular Dynamics and Machine Learning Thermodynamic Perturbation Theory

<div>In this work we present a detailed \textit{ab initio} study of the carbonylation reaction of methoxy groups in the zeolite mordenite, as it is the rate determining step in a series of elementary reactions leading to ethanol. </div><div>For the first time we employ full molecular dynamics simulations to evaluate free energies of activation for the reactions in side pockets and main channels. Results show that the reaction in the side pocket is preferred and, when dispersion interactions are taken into account, this preference becomes even stronger. This conclusion is confirmed using multiple levels of density functional theory approximations with (PBE-D2, PBE-MBD, and vdW-DF2-B86R) or without (PBE, HSE06) dispersion corrections. These calculations, that in principle would require several demanding molecular dynamics simulations, were made possible at a minimal computational cost by using a newly developed approach that combines thermodynamic perturbation theory with machine learning.</div>

A simple second order thermodynamic perturbation theory for associating fluids

An approximation within Wertheim's second order perturbation theory is proposed which allows for the development of a general solution for pure component fluids with an arbitrary number and functionality of association sites. The solution is closed, concise and general for all second order effects such as ring formation, steric hindrance and hydrogen bond cooperativity. The approach is validated by comparison to hydrogen bond structure data for liquid water.