Cheap but accurate calculation of chemical reaction rate constants from ab initio data, via system-specific, black-box force fields

Abstract

Building on the recently published quantum-mechanically derived force field (QMDFF) and its empirical valence bond extension, EVB-QMDFF, it is now possible to generate a reliable potential energy surface for any given elementary reaction step in an essentially black box manner. This requires a limited and pre-defined set of reference data near the reaction path and generates an accurate approximation of the reference potential energy surface, on and off the reaction path. This intermediate representation can be used to generate reaction rate data, with far better accuracy and reliability than with traditional approaches based on transition state theory (TST) or variational extensions thereof (VTST), even if those include sophisticated tunneling corrections. However, the additional expense at the reference level remains very modest. We demonstrate all this for three arbitrarily chosen example reactions.

Quantum chemical accuracy from density functional approximations via machine learning

“…Admittedly, employing the full electron density carries with it a considerable computational cost; nevertheless, it is useful to develop such frameworks, considering that more optimal algorithms could follow. Previously, we had shown that the electron density could be used in a self-consistent manner to train a system-specific density functional (akin to a system-specific force field 52 ) using a mapping from the external potential to the electron density and a second map of the density to the total energy 53 . Rather than delivering a solution to the KS equations, the first map (denoted the ML-HK map) bypasses the KS equations in a manner that is akin to solving the original Hohenberg-Kohn functional differential equation 54 .…”

Reactive Molecular Dynamics at Constant Pressure via Nonreactive Force Fields: Extending the Empirical Valence Bond Method to the Isothermal-Isobaric Ensemble

“…In contrast, if the choice was to use coupling terms that do not depend on mk but other degrees of freedom such as spatial coordinates (see refs. [40][41][42][43][44][45][46], we cannot discern a clear logic to derive an expression for σ EVB , which might explain the fact there is no evidence of any previous reported method to compute the stress tensor using EVB.…”

“…To overcome these limitations, a combination of Gaussian functions were proposed to model the coupling terms. 45 This methodology offers a promising route for calibration and development of EVB potentials, and provides a suitable framework for future applications of the EVB extension proposed in this work.…”

“…Several sophisticated EVB coupling recipes have been proposed over the years. 23,[40][41][42][43][44][45][46]51,52 Despite their proven success, these recipes use complex internal (spatial) coordinates to couple the force fields. Here, however, we aim to use functional forms C mk that depend on the energy…”

Theoretical evaluation of equilibrium constant for boron isotopes exchange between boric acid and borate in pure and saline water

“…Nevertheless, NMA with the explicit solvent approach cannot be conveniently employed for studying multicomponent solvents since achieving appropriate sampling of a canonical ensemble in this way is challenging. Consequently, theoretical evaluation of k 3−4 in saline water is achieved via NMA with the implicit solvent approach [21] as well as path integral molecular dynamics [22,23]. The latter method also allows studying the anharmonicity impacts on the calculated partition functions.…”