Ab Initio Generalized Langevin Equation

Pinchen Xie,Roberto Car,Weinan E

DOI: https://doi.org/10.1073/pnas.2308668121

2024-02-15

Abstract:We introduce a machine learning-based approach called ab initio generalized Langevin equation (AIGLE) to model the dynamics of slow collective variables in materials and molecules. In this scheme, the parameters are learned from atomistic simulations based on ab initio quantum mechanical models. Force field, memory kernel, and noise generator are constructed in the context of the Mori-Zwanzig formalism, under the constraint of the fluctuation-dissipation theorem. Combined with deep potential molecular dynamics and electronic density functional theory, this approach opens the way to multi-scale modeling in a variety of situations. Here, we demonstrate this capability with a study of two mesoscale processes in crystalline lead titanate, namely the field-driven dynamics of a planar ferroelectric domain wall, and the dynamics of an extensive lattice of coarse-grained electric dipoles. In the first case, AIGLE extends the reach of ab initio simulations to a regime of noise-driven motions not accessible to molecular dynamics. In the second case, AIGLE deals with an extensive set of collective variables by adopting a local approximation for the memory kernel and retaining only short-range noise correlations. The scheme is computationally more efficient than molecular dynamics by several orders of magnitude, and mimics the microscopic dynamics at low frequencies where it reproduces accurately the dominant far-infrared absorption frequency.

Computational Physics,Mesoscale and Nanoscale Physics,Materials Science

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the dynamic modeling of slow collective variables in materials and molecules. Specifically, the authors introduce a machine - learning - based method - Ab initio Generalized Langevin Equation (AIGLE) for simulating the dynamics of these slow collective variables. The core of this method is to learn parameters from atom - scale simulations, which are based on ab initio quantum - mechanical models. Force fields, memory kernels and noise generators are constructed within the framework of the Mori - Zwanzig formalism and are constrained by the fluctuation - dissipation theorem.

The AIGLE method combines deep - potential molecular dynamics and electronic density functional theory, paving the way for multi - scale modeling in a variety of situations. The paper demonstrates this ability by studying two mesoscopic processes in crystalline lead titanate: one is the field - driven dynamics of planar ferroelectric domain walls; the other is the dynamics of a coarsened electric - dipole - extended lattice. In these two cases, AIGLE shows its advantages in noise - driven motion and handling a large number of collective variables respectively, especially the ability to accurately reproduce the dominant far - infrared absorption frequencies in the low - frequency band.

In summary, this paper aims to develop a new method that can efficiently and accurately simulate mesoscopic dynamic processes in complex material systems, especially non - Markovian processes that are difficult to reach by traditional molecular dynamics. By combining machine learning with ab initio quantum mechanics, AIGLE provides a powerful tool that can significantly improve the understanding and prediction ability of these systems.