Continuing DMFT calculations from previous runs

[danjuzam]

Hello,

I did some DFT+DMFT calculations with VASP (6.5.0) and solid_dmft, starting first with a low number of QMC steps (10^6) and 20 CSC steps. Once this was finished, I wanted to continue the calculation with a higher number of QMC steps (10^7), reading in the self energy from the first run. From what I understood from the documentation, this can be done by setting in the dmft_config.toml

load_sigma = true

and

path_to_sigma = "../01_1-Mio-MC/vasp.h5"

if that's the path to my vasp.h5 file from the first run. What I now don't understand is, that when I look into the observables_imp0.dat file generated by the continued run, all the initial values of chemical potential, G(beta/2), and the orbital occupations start again from the same values as in the first run. Shouldn't they be different now? Also, setting

mu_initial_guess

to some value (e.g., the last value in observables_imp0.dat of the first run) does not change the mu value the continued second run begins with. Am I missing something here? Any help is highly appreciated, thank you!

Best,
Daniel

[phibeck]

Hi @danjuzam,

per default, solid_dmft will always report in "iteration 0" the non-interacting initial values in the observables_*.dat file. So if iteration 0 is identical to your previous run, that's on purpose. The question is whether the follow-up iterations are continued from your previous run, but you should be able to see some statements regarding this in the output.

The easiest way to continue a calculations is to just continue the run as is - leave the jobname identical (making sure the folder exists in the current directory), don't load sigma etc. Then it will just append the following iterations directly after your first 20+ iterations.