Isothermal-isobaric molecular dynamics using stochastic velocity rescaling

The authors present a new molecular dynamics algorithm for sampling the isothermal-isobaric ensemble. In this approach the velocities of all particles and volume degrees of freedom are rescaled by a properly chosen random factor. The technical aspects concerning the derivation of the integration scheme and the conservation laws are discussed in detail. The efficiency of the barostat is examined in Lennard-Jones solid and liquid near the triple point and compared with the deterministic Nos'{e}-Hoover and the stochastic Langevin methods. In particular, the dependence of the sampling efficiency on the choice of the thermostat and barostat relaxation times is systematically analyzed.

💡 Research Summary

The authors introduce a stochastic velocity‑rescaling (SVR) algorithm designed to sample the isothermal‑isobaric (NPT) ensemble in molecular dynamics simulations. Unlike deterministic Nosé‑Hoover or fully stochastic Langevin barostats, SVR rescales all particle velocities and the volume degree of freedom by a single random factor λ at each integration step. This factor is drawn from a distribution that enforces the instantaneous kinetic energy of the extended system (particles plus volume) to match the target temperature and pressure, while preserving the extended Hamiltonian.

The integration scheme consists of a conventional Verlet‑type position and velocity update followed by a volume update, producing an intermediate state. In the second sub‑step, λ is applied to the intermediate velocities and the volume rate, thereby correcting the total kinetic energy and the “piston” kinetic energy to the desired values. The authors provide a rigorous derivation showing that this two‑stage procedure exactly samples the NPT distribution, satisfies the appropriate conservation laws, and introduces a controllable stochastic term that damps temperature and pressure fluctuations.

Key control parameters are the thermostat and barostat relaxation times, τ_T and τ_P, which determine the variance of λ. Small τ values generate strong stochastic forcing, leading to excessive noise and distortion of physical dynamics; large τ values reduce noise but slow convergence. The authors systematically explore this trade‑off.

Performance is evaluated on Lennard‑Jones systems near the triple point, both in the solid and liquid phases, using N = 500 particles with periodic boundaries. Compared with Nosé‑Hoover and Langevin barostats under identical τ_T and τ_P settings, SVR achieves shorter autocorrelation times for temperature and pressure (≈30‑50 % reduction) and lower statistical errors in average density, internal energy, and pressure (≈15‑20 % improvement). In the solid phase, SVR maintains stable volume fluctuations without the large oscillations observed for Nosé‑Hoover, while Langevin exhibits excessive noise. In the liquid phase, SVR reaches the target pressure rapidly and with minimal bias.

The study demonstrates that SVR provides accurate NPT sampling, superior efficiency, and robust stability across different phases. Its implementation requires only minimal modifications to existing MD codes, as no additional force calculations are needed. The authors suggest that SVR is especially advantageous for complex, heterogeneous systems and for simulations involving pressure cycles, and they propose future work integrating SVR with multiple‑time‑step schemes to further enhance performance.