Machine learning force fields based on local parametrization of dispersion interactions: Application to the phase diagram of C_60

摘要

We present a comprehensive methodology to enable the addition of van der Waals (vdW) corrections to machine learning (ML) atomistic force fields. Using a Gaussian approximation potential (GAP) [Bartok et ai, Phys. Rev. Lett. 104, 136403 (2010)] as a baseline, we accurately machine learn a local model of atomic polariz-abilities based on Hirshfeld volume partitioning of the charge density [Tkatchenko and Scheffier, Phys. Rev. Lett. 102, 073005 (2009)]. These environment-dependent polarizabililies are then used to parametrize a screened London-dispersion approximation to the vdW interactions. Our ML vdW model only needs to learn the charge density partitioning implicitly by learning the reference Hirshfeld volumes from density functional theory (DFT). In practice, we can predict accurate Hirshfeld volumes from the knowledge of the local atomic environment (atomic positions) alone, making the model highly computationally efficient. For additional efficiency, our ML model of atomic polarizabilities reuses the same many-body atomic descriptors used for the underlying GAP learning of bonded interatomic interactions. We also show how the method enables straightforward computation of gradients of the observables, even when these remain challenging for the reference method (e.g., calculating gradients of the Hirshfeld volumes in DFT). Finally, we demonstrate the approach by studying the phase diagram of C_(60), where vdW effects are important. The need for a highly accurate vdW-inclusive reactive force field is highlighted by modeling the decomposition of the C_(60) molecules taking place at high pressures and temperatures.