NVE ensemble
The NVE ensemble (micro-canonical ensemble) is a statistical ensemble that is used to study material properties under the conditions of a constant particle number N, constant volume V and an internal energy E fluctuating around an equilibrium value E. This page describes how to sample the NVE ensemble from a molecular-dynamics run.
Instructions for setting up a NVE ensemble
There are two choices of thermostats which can be adjusted to set up a molecular-dynamics run. The Andersen thermostat can be used by setting the collision probability (ANDERSEN_PROB) with the fictitious heat bath to zero and the Nose-Hoover thermostat by setting the mass of the virtual degree of freedom (SMASS) to minus three. Both settings will switch the thermostat of such that the velocities are determined by the Hellmann-Feynman forces or Machine-learned force fields only. See table for the corresponding MDALGO setting and related tags.
| NVE ensemble | Andersen | Nose-Hoover |
|---|---|---|
| MDALGO | 1 | 0 or 2 |
| additional tags to set | ANDERSEN_PROB=0.0 | SMASS=-3 |
The additional tags in the column for every thermostat have to be set to the given values. Otherwise the NVE ensemble will not be realized. There are two implementations of the Nose-Hoover thermostat in VASP which will give the same results. The MDALGO=0 version can be used even if the code was compiled without the precompiler option -Dtbdyn. Other tags related to molecular-dynamics simulations can be found here.
An example INCAR file for the Andersen thermostat could look like
#INCAR molecular-dynamics tags NVE ensemble IBRION = 0 # choose molecular-dynamics MDALGO = 1 # using Andersen thermostat ISIF = 2 # compute stress tensor but do not change box volume/shape TEBEG = 300 # set temperature NSW = 10000 # number of time steps POTIM = 1.0 # time step in femto seconds ANDERSEN_PROB = 0.0 # setting Andersen collision probability to zero to get NVE enseble
| Mind: This INCAR file only contains the parameters for the molecular-dynamics part. The electronic minimization or the machine learning tags have to be added. |