Flat bands of phonon dispersion calculation

[Ssingsen]

Hi TDEP developer,

I plotted the phonon dispersion, but the result shows flat bands between high-symmetry points. Could you please suggest to me how to fix this problem?

My details of calculation is below,

• Duo to a large unit cell (504 atoms, a = b = 18 A and c = 50 A), I used only unit cell without creating supercell to run AIMD simulation with NVT ensemble for 1000 fs.
• I extracted forces using the script as follow: extract_forceconstants -rc2 100 -s 50 > fc2.log
• Then, I used the following script to get .gnuplot: phonon_dispersion_relations --readpath --unit thz -nq 10 -qg 1 1 1 --dospoints 10 --temperature 300

I know my setting is too bias inaccurately due to the big system. Also, I got negative frequency as well, even if my system was optimized already.

Thank you very much.

[flokno]

Hi @Ssingsen , did you successfully run the TDEP tutorials and got fine results there?

Other comments

• [ ] please add your fc2.log output file.
• [ ] did you try smaller rc2 cutoffs? Starting with nearest neighbors only. We have written a tutorial on how to find good cutoffs and I think if you inspect your model parameters closely, you will find that there is a lot of room for improvement: https://github.com/tdep-developers/tdep-tutorials/tree/main/02_sampling/convergence/cutoff