I’m working on calculating the diffusion coefficient of ethanol using the MSD plot method and have obtained the following values: Dx = 1.4e-9 m^2/s, Dy = 3.4e-10 m^2/s, and Dz = 2.5e-9 m^2/s. I’m puzzled because I expected these values to be approximately equal. Does this discrepancy suggest that the force field I’m using might not be accurate? Moreover, is there a specific procedure to accurately determine diffusion coefficients using LAMMPS? Are there any tutorials available on this subject?
Best,
It’s impossible to comment without knowing the size of your sample, the equilibration history, and the production time, to cite a few. A simulation is always affected by finite-size errors, therefore a slightly anisotropic behaviour is normal. Please refer to the abundant literature on this topic to work out the simulation parameters. I personally wouldn’t blame the force field choice (also, because I don’t have a clue of what you have used).
1 Like
They are about the same order of magnitude, which means this is not a complete disaster.
No. You are using (non-bonded) potentials with spherical symmetry, so how they be able to discriminate between directions unless you have a non-homogeneous system?
This is not LAMMPS specific, but the Einstein relation relates the self-diffusion to the slope of the mean squared distance in the limit of infinite time, so you need to determine whether your data is sufficiently converged. That is most easily done by plotting a graph of the data you use.
This is a very fundamental operation and not LAMMPS specific. It should be discussed in proper text books on molecular dynamics and review articles. You also may find tutorials, specifically with software packages that specialize in the analysis of MD simulation data. In the LAMMPS manual you have all LAMMPS specific information with the description of the compute md command.
1 Like
You can do it several times in identical systems but with different random seeds to see if the bias is constant.
2 Likes