Simulation of quantum zero-point effects in water using a frequency-dependent thermostat

TL;DR

This paper introduces a computationally efficient thermostat method that accurately incorporates quantum zero-point effects in water simulations, matching results from more expensive path integral methods.

Contribution

The authors develop a generalized Langevin thermostat combined with Nose-Hoover dynamics to simulate quantum zero-point effects in water efficiently.

Findings

Accurately reproduces atomic pair correlation functions.

Matches results of path integral molecular dynamics.

Provides a simple, inexpensive alternative for quantum effects simulation.

Abstract

Molecules like water have vibrational modes with a zero-point energy well above room temperature. As a consequence, classical molecular dynamics simulations of their liquids largely underestimate the energy of modes with a higher zero-point temperature, which translates into an underestimation of covalent interatomic distances due to anharmonic effects. Zero-point effects can be recovered using path integral molecular dynamics simulations, but these are computationally expensive, making their combination with ab initio molecular dynamics simulations a challenge. As an alternative to path integral methods, from a computationally simple perspective, one would envision the design of a thermostat capable of equilibrating and maintaining the different vibrational modes at their corresponding zero-point temperatures. Recently, Ceriotti et al. (Phys. Rev. Lett. 102 020601 (2009)) introduced a framework to use a custom-tailored Langevin equation with correlated noise that can be used to include quantum fluctuations in classical molecular dynamics simulations. Here we show that it is possible to use the generalized Langevin equation with suppressed noise in combination with Nose-Hoover thermostats to efficiently impose a zero-point temperature on independent modes in liquid water. Using our simple and inexpensive method, we achieve excellent agreement for all atomic pair correlation functions compared to the path integral molecular dynamics simulation.