Moderators: Global Moderator, Moderator
Hello
I used the ZG configuration (one-shot sampling) for 2x2x2 of Y2O2S and got several POSCAR files from 0K to 700K. Then I used the 0K POSCAR to create a phonon dispersion. However, this looks very different than the phonon dispersion from the ZG calculations, which is supposed to be at 0K. Data and figures are attached.
Thanks-Nick
To build the ZG configuration one first computes the harmonic phonons in a supercell (gamma only) and then generates displaced structures by populating the phonon modes at different temperatures.
But if one simply calculates the phonons at these displaced structures one will get negative phonon modes because the structures are not at equilibrium.
Note that this is true for the structures generated at 0K due to zero-point vibrations.
This is a common misconception due to an abuse of language: when computing the phonons from DFT finite differences of DFPT one often says the system is at 0K but more correct would be to say that it is computed within the Born-Opennheimer approximation i.e. the atoms are frozen in their equilibrium positions. Then going beyond the Born-Opennheimer aprpoximation implies considering that the ions are moving, even if we are at 0K.
So in a sense the results you obtained are not surprising.
But perhaps I can help a little bit more by first asking what property are you interested in?
Thank you for your reply. I was using ZG to get POSCAR files per temperature in order to subsequently use them to calculate the phonon dispersion as a function of temperature. I got similar results using the ZG POSCAR file at T = 700K.
Nick
As far as I know, the the phonon disperion computed at one of the ZG configurations generated for temperature T does not correspond to the phonon dispersion at a that temperature T.
At least the connection between these two is not obivous to me. Is there some reference in the literature where this approach is used?
There are many ways to compute a temperature dependent phonon dispersion, see for example (and the relevant references):
phono3py: https://phonopy.github.io/phono3py/
scaild: https://www.uquantchem.com/scaild.html
tdep: https://tdep-developers.github.io/tdep/program/
sscha: https://sscha.eu/
The easiest (but very approximated) way is to simply compute the phonons at T=0, compute the lattice thermal expansion at the desired T and then computing the phonon dispersion with those new lattice parameters.
Note that depending on your crystal you might need to relax the ionic positions for T=0 at this new lattice.