Construction of a Gaussian Process Regression Model of Formamide for Use in Molecular Simulations

Abstract

FFLUX, a novel force field based on quantum chemical topology, can perform molecular dynamics simulations with flexible multipole moments that change with geometry. This is enabled by Gaussian process regression machine learning models, which accurately predict atomic energies and multipole moments up to the hexadecapole. We have constructed a model of the formamide monomer at the B3LYP/aug-cc-pVTZ level of theory capable of sub-kJ mol −1 accuracy, with the maximum prediction error for the molecule being 0.8 kJ mol −1 . This model was used in FFLUX simulations along with Lennard-Jones parameters to successfully optimize the geometry of formamide dimers with errors smaller than 0.1 Å compared to those obtained with D3corrected B3LYP/aug-cc-pVTZ. Comparisons were also made to a force field constructed with static multipole moments and Lennard-Jones parameters. FFLUX recovers the expected energy ranking of dimers compared to the literature, and changes in C�O and C−N bond lengths associated with hydrogen bonding were found to be consistent with density functional theory.

Application of the FFLUX Force Field to Molecular Crystals: A Study of Formamide

“…The similarity suggests that the differences in the IR intensities are due to either the way the Born charges have been calculated or to the predicted charges and dipole moments used to compute the polarization. Given the previously demonstrated accuracy of the formamide model, we suggest that the former is more likely, and we note that the alternative MD-based approach used in our previous study may give better results than the finite-difference method employed here …”

“…The similarity suggests that the differences in the IR intensities are due to either the way the Born charges have been calculated or to the predicted charges and dipole moments used to compute the polarization. Given the previously demonstrated accuracy of the formamide model, 24 we suggest that the former is more likely, and we note that the alternative MD-based approach used in our previous study may give better results than the finite-difference method employed here. 23 Whichever method is chosen, the ability to straightforwardly predict the IR spectra is another advantage of FFLUX, particularly in the context of polymorphism, where the simulated spectra could be used to experimentally verify predictions.…”

“…The GPR models for the FFLUX simulations were trained using our in-house Python pipeline, ICHOR . Details of the creation of the 1506-point formamide GPR model and a test of its accuracy in monomer and dimer calculations are presented in ref , but a brief overview is provided here. A 1 ns AMBER simulation of a formamide monomer at 300 K was performed to generate a set of geometries.…”

“…FFLUX 21,22 is a new force field that has previously been used to accurately predict the properties of liquid water 23 and the geometries of gaseous formamide dimers. 24 The dimer work showed that FFLUX effectively "sees" the electrons, with small changes in C�O and C−N bond lengths due to hydrogen bonding captured accurately relative to the training level of theory. The correct energy ranking of the dimeric minima was also obtained, suggesting that an accurate ranking of crystal polymorphs could be possible.…”

Application of the FFLUX Force Field to Molecular Crystals: A Study of Formamide

“…FFLUX [21][22] is a new force field that has previously been used to accurately predict the properties of liquid water 23 and the geometries of gaseous formamide dimers 24 . The dimer work showed that FFLUX effectively "sees" the electrons, with small changes in C=O and C-N bond lengths due to hydrogen bonding captured accurately relative to the training level of theory.…”