Fast Decorrelating Monte Carlo Moves for Efficient Path Sampling.

Area of Science:

• Computational chemistry
• Molecular dynamics simulations
• Statistical physics

Background:

• Rare events in molecular dynamics (MD) simulations, such as chemical reactions or biomolecular conformational changes, occur beyond accessible timescales.
• Path sampling methods, using Monte Carlo (MC) sampling of MD trajectories, aim to overcome this limitation.
• Accurate prediction of rate constants for these rare events requires extensive sampling, often demanding a large number of trajectories.

Purpose of the Study:

• To develop novel Monte Carlo (MC) moves to enhance the efficiency of path sampling for rare events.
• To improve the speed of convergence in calculating rate constants for computationally challenging processes.
• To introduce methods that reduce trajectory correlation and accelerate sampling.

Main Methods:

• Proposed two new MC moves: stone skipping and web throwing.
• Constructed trajectories using a sequence of subpaths that obey superdetailed balance.
• Employed a reweighting procedure to accept a high proportion of generated paths.

Main Results:

• The new MC moves, while increasing the computational cost per trajectory, significantly reduce correlation between trajectories.
• This reduction in correlation leads to a substantial speedup in the overall simulation process.
• A factor of 12 speedup was observed in a study of DNA denaturation.

Conclusions:

• Stone skipping and web throwing are effective MC moves for accelerating rare event simulations.
• These methods improve the convergence rate for predicting rate constants in molecular dynamics.
• The developed techniques offer a significant advancement for studying complex chemical and biological processes computationally.