Virtual Molecular Dynamics

TL;DR

This paper introduces virtual molecular dynamics, a multiscale approach using temporal coarse-graining with Pade approximants, significantly improving efficiency and avoiding coarse-grained variables in simulating complex molecular systems.

Contribution

It presents a novel framework combining stochastic processes, Pade approximants, and energy annealing to enhance molecular dynamics simulations without coarse-grained variables.

Findings

Achieved increased efficiency over traditional MD methods.

Successfully modeled structural transitions in biological and physical systems.

Demonstrated applicability to diverse molecular systems.

Abstract

Molecular dynamics is based on solving Newton's equations for many-particle systems that evolve along complex, highly fluctuating trajectories. The orbital instability and short-time complexity of Newtonian orbits is in sharp contrast to the more coherent behavior of collective modes such as density profiles. The notion of virtual molecular dynamics is introduced here based on temporal coarse-graining via Pade approximants and the Ito formula for stochastic processes. It is demonstrated that this framework leads to significant efficiency over traditional molecular dynamics and avoids the need to introduce coarse-grained variables and phenomenological equations for their evolution. In this framework, an all-atom trajectory is represented by a Markov chain of virtual atomic states at a discrete sequence of timesteps, transitions between which are determined by an integration of conventional molecular dynamics with Pade approximants and a microstate energy annealing methodology. The latter is achieved by a conventional and an MD NVE energy minimization schemes. This multiscale framework is demonstrated for a pertussis toxin subunit undergoing a structural transition, a T=1 capsid-like structure of HPV16 L1 protein, and two coalescing argon droplets.