Variational control forces for enhanced sampling of nonequilibrium molecular dynamics simulations

Area of Science:

• * Statistical mechanics
• * Computational physics
• * Nonequilibrium systems

Background:

• * Estimating rare event probabilities is crucial in molecular dynamics simulations.
• * Nonequilibrium molecular dynamics (NEMD) simulations are essential for studying systems driven out of equilibrium.
• * Large deviation functions characterize the probabilities of time-integrated dynamical observables.

Purpose of the Study:

• * To introduce a novel variational algorithm for estimating rare event likelihoods in NEMD.
• * To optimize control forces for enhanced statistical efficiency in calculating large deviation functions.
• * To provide a method that is quantitatively accurate and computationally efficient.

Main Methods:

• * Development of a variational algorithm utilizing optimal control forces.
• * Calculation of gradients for cost function optimization based on trajectory susceptibility.
• * Benchmarking against numerically exact solutions and Monte Carlo methods.

Main Results:

• * The variational estimate for rare event likelihoods is quantitatively accurate in many cases.
• * The algorithm demonstrates fast convergence and significant statistical efficiency gains.
• * Singular changes in optimal control forces were observed during a dynamical phase transition.

Conclusions:

• * The developed variational algorithm offers a powerful and efficient tool for rare event estimation in NEMD.
• * The method accurately evaluates large deviation functions, outperforming traditional Monte Carlo approaches.
• * This approach has broad applicability in studying complex dynamical systems and phase transitions.