Probing the force field sensitivity of entropy and enthalpy differences in organic polymorphs using classical potentials

TL;DR

This study assesses classical potentials' ability to predict thermodynamic differences in organic polymorphs, highlighting the polarizable AMOEBA potential's relative success and the limitations of point charge models.

Contribution

It systematically evaluates the accuracy of various classical potentials in predicting enthalpy and entropy differences in organic polymorphs using the quasi-harmonic approximation.

Findings

AMOEBA predicts correct enthalpy sign for ~71% of polymorphs

Point charge potentials perform no better than chance for enthalpy sign prediction

Entropy predictions are less sensitive, with 64-75% accuracy across force fields

Abstract

We evaluate the effectiveness of different classical potentials to predict the thermodynamics of a number of organic solid form polymorphs relative to experimentally reported values using the quasi-harmonic approximation. Using the polarizable potential AMOEBA we are able to predict the correct sign of the enthalpy difference for 71+/-12 % of the polymorphs. Alternatively, all point charge potentials perform on par with random chance of correcting the correct sign (50%) for enthalpy. We find that the entropy is less sensitive to the accuracy of the potential with all force fields, excluding CGenFF, reporting the correct sign of the entropy for 64+/-13 - 75+/-11 % of the systems. Predicting the correct sign of the enthalpy and entropy differences can help indicate the low and high temperature stability of the polymorphs, unfortunately the error relative to experiment in these predicted values can be as large as 1 - 2.5 kcal/mol at the transition temperature.