Re: Hellmann–Feynman forces do not converge
Posted: Thu Jun 06, 2024 3:46 pm
Hi,
"Regarding the structure you find with IBRION=1 and NFREE=10 is this some meta-stable configuration?"
Unfortanetly I do not have the experimental data (yet) to confirm if there are really multiple configurations.
I know that for very similar compounds, DFT calculations have calculated multiple polaron configurations with slighltly different energy on the order of a few 10 meV, and that spectroscopic methods also hint at multiple excited states with energy differences on a similar range. So I can imagine that such metastable states can exist in this compound.
"To get accurate polaron binding energies, you probably need to make a convergence study with the size of the supercell."
I have considered this, I have not tried it yet. To keep a somewhat constant cubic cell shape, the supercell would need to be increased from 80 (2x2x1 unit cells) to about 640 atoms (4x4x2 unit cells).
I fear these calculations will take considerable time and can be more problematic for converging.
In my case, I am more interested in the trend among 5 different structures, rather than their absolute energy. I expect (perhaps erroneously) that the error among these different structures is very similar, as their supercell size and shape are more-or-less the same.
However, to get some estimate of the error caused by self-interaction of the polarons with their mirrorimages, I should/will do such size vs. energy calculations on at least on of the structures.
Kind regards,
Jeffrey
"Regarding the structure you find with IBRION=1 and NFREE=10 is this some meta-stable configuration?"
Unfortanetly I do not have the experimental data (yet) to confirm if there are really multiple configurations.
I know that for very similar compounds, DFT calculations have calculated multiple polaron configurations with slighltly different energy on the order of a few 10 meV, and that spectroscopic methods also hint at multiple excited states with energy differences on a similar range. So I can imagine that such metastable states can exist in this compound.
"To get accurate polaron binding energies, you probably need to make a convergence study with the size of the supercell."
I have considered this, I have not tried it yet. To keep a somewhat constant cubic cell shape, the supercell would need to be increased from 80 (2x2x1 unit cells) to about 640 atoms (4x4x2 unit cells).
I fear these calculations will take considerable time and can be more problematic for converging.
In my case, I am more interested in the trend among 5 different structures, rather than their absolute energy. I expect (perhaps erroneously) that the error among these different structures is very similar, as their supercell size and shape are more-or-less the same.
However, to get some estimate of the error caused by self-interaction of the polarons with their mirrorimages, I should/will do such size vs. energy calculations on at least on of the structures.
Kind regards,
Jeffrey