Finite temperature phonon dispersion in VASP 5.4?

[topojit]

Hello everyone,
Is it possible to calculate phonon dispersion at finite temperature (such as at 300k) using VASP (version 5.4)? I recently saw some works that calculate phonon at room temperature using first principle method (such as: PhysRevB.105.014304) or this package (https://tdep-developers.github.io/tdep/).
So, my question is:
1. Can VASP calculate phonon dispersion at non-zero temperature?
2. If not, does anyone have any recommended method or software to calculate this reliably?

[martin.schlipf]

We have some features for temperature dependent properties in newer versions of VASP e.g. the MC method for electron phonon. Of course it is also possible of running an MD simulation with VASP.5 and get the phonons from there.

But there is no native way to get the phonon dispersion at finite temperature with VASP. So I moved this to from users for users where other users may contribute their favorite tools

[topojit]

Hi, thank you for the information. I have some questions about these method you mentioned:

1. Can you explain the phonons from Molecular Dynamics method a bit more? I know you can check structural stability at different temperatures using molecular dynamics. So, is the process something like this: do a MD calculation at XXX temperature up to like 10ps, take the CONTCAR file from the end of MD and use that to do the phonon dispersion using the usual method and parameters?

2. The link you provided talks about Monte-Carlo sampling. Can VASP do Monte-Carlo simulations? We were looking into the curie temperature calculation and had to use Vampire and UPPASD software for the Monte-Carlo method. Can VASP calculate the curie/Neel temperature by itself (if so, which version and how)?

Thank you.

[martin.schlipf]

I don't have a script to analyze an MD trajectory but if you think of harmonic oscillator you know that if I would give you x(t) data then you could extract the frequency of the oscillator from it. For phonons it is a bit more complicated because multiple modes exist at the same time. Still a frequency analysis (Fourier transform) of the trajectory should give you the data but this is as much as I know about that. I would not be able to do this in practice without some research.

Sorry the Monte-Carlo comment was a bit short and lead to confusion. Monte Carlo is a method that can be used for many purposes. One use case is spin dynamics you mentioned; this is not available in VASP. What is available in VASP is a MC snapshot for the ion dynamics. Instead of averaging e.g. the DOS over multiple snapshots of a MD simulation, you construct one or more MC snapshots and average over those. This avoids the computational cost of running the dynamics between the MD snapshots.