Data-driven dynamical coarse-graining for condensed matter systems

TL;DR

This paper introduces a data-driven stochastic coarse-graining method inspired by the Mori-Zwanzig formalism, enabling efficient simulation of complex condensed matter systems by capturing both equilibrium and dynamic properties.

Contribution

The authors develop a novel coarse-graining approach that incorporates noise and memory effects from data, improving dynamic accuracy over traditional methods.

Findings

Accurately reproduces equilibrium distributions.

Captures temporal correlations and memory effects.

Effective on multiple condensed matter systems.

Abstract

Simulations of condensed matter systems often focus on the dynamics of a few distinguished components but require integrating the dynamics of the full system. A prime example is a molecular dynamics simulation of a (macro)molecule in solution, where both the molecules(s) and the solvent dynamics needs to be integrated. This renders the simulations computationally costly and often unfeasible for physically or biologically relevant time scales. Standard coarse graining approaches are capable of reproducing equilibrium distributions and structural features but do not properly include the dynamics. In this work, we develop a stochastic data-driven coarse-graining method inspired by the Mori-Zwanzig formalism. This formalism shows that macroscopic systems with a large number of degrees of freedom can in principle be well described by a small number of relevant variables plus additional noise and memory terms. Our coarse-graining method consists of numerical integrators for the distinguished components of the system, where the noise and interaction terms with other system components are substituted by a random variable sampled from a data-driven model. Applying our methodology on three different systems -- a distinguished particle under a harmonic potential and under a bistable potential; and a dimer with two metastable configurations -- we show that the resulting coarse-grained models are not only capable of reproducing the correct equilibrium distributions but also the dynamic behavior due to temporal correlations and memory effects. Our coarse-graining method requires data from full-scale simulations to be parametrized, and can in principle be extended to different types of models beyond Langevin dynamics.