ABSTRACT
The ability to automatically extend and improve molecular force fields to accurately describe an ever wider range of compounds is a key enabler for the application of molecular modelling of chemicals and materials in industry. In this work we have developed a set of tools to process structural data available in well-known databases, to find deficiencies in the force field, and where necessary, automatically fit force- field parameters to on-the-fly calculated quantum data based on Density Functional Theory, supplemented by experimental data. The protocols have been applied to structures from the Maybridge, PoLyInfo and ILThermo databases, covering drug-like molecules, polymers and ionic liquids respectively. We demonstrate that the new version of the force field can type all structures in the extended data set without missing parameters, with a similar high-quality prediction of geometry (bonds, angles, torsions), energy and forces of earlier versions.
KEYWORDS:
Acknowledgments
The authors thank Cas van der Oord and Márk Jenei (University of Cambridge) for help in torsion parameterisation as part of their industrial CASE studentships. At Dassault Systèmes we thank Carsten Menke for parameterising titanium tetrachloride, Marine Ciantar for validation of ionic liquid properties, Michael Diedenhofen for providing heats of vapourisation for ionic liquids, Felix Hanke for advice on setting up DMol3, and Prabhu Raman for discussions on torsion fitting.
Disclosure statement
No potential conflict of interest was reported by the author(s).