Up to what stresses can PBE / R2SCAN results be trusted?
Moderators: Global Moderator, Moderator
-
matthewkuner
- Newbie
- Posts: 16
- Joined: Wed Oct 12, 2022 8:17 pm
Up to what stresses can PBE / R2SCAN results be trusted?
As a general rule (i.e. for applying to most/all of the periodic table), above what stress does the frozen core approximation start to significantly affect the accuracy of a VASP calculation? Does this depend on the external pressure, or individual components of the stress tensor? And I assume that this only noteworthy for high pressures (i.e. highly compressed systems)--let me know if that assumption is incorrect. Thanks in advance!
-
jonathan_lahnsteiner2
- Global Moderator
- Posts: 207
- Joined: Fri Jul 01, 2022 2:17 pm
Re: Up to what stresses can PBE / R2SCAN results be trusted?
Dear matthewkuner,
This question cannot be answered in general because the answer will strongly depend on the material under consideration. You could compare your vasp PAW results to an all-electron calculation to be sure that the frozen core approximation is still applicable under the stress of interest.
You are right in your assumption that the applicability of the frozen core approximation will depend on the external pressure since it will decrease the volume of your simulation box. The applied external pressure will also influence the individual components of the stress tensor. And yes this will only be noteworthy for high external pressures.
Maybe there will be some user in the user-for-users forum who has more insight into this topic.
All the best
Jonathan
This question cannot be answered in general because the answer will strongly depend on the material under consideration. You could compare your vasp PAW results to an all-electron calculation to be sure that the frozen core approximation is still applicable under the stress of interest.
You are right in your assumption that the applicability of the frozen core approximation will depend on the external pressure since it will decrease the volume of your simulation box. The applied external pressure will also influence the individual components of the stress tensor. And yes this will only be noteworthy for high external pressures.
Maybe there will be some user in the user-for-users forum who has more insight into this topic.
All the best
Jonathan
-
marvin_poul
- Newbie
- Posts: 4
- Joined: Tue Sep 07, 2021 8:14 am
Re: Up to what stresses can PBE / R2SCAN results be trusted?
I've been pondering this question for some time. I'm trying to create p-T phase diagrams with data from VASP calculations, but this of course depends strongly on the applicability of the frozen core approximation. I've seen works in the literature that use VASP pseudopotentials for pressures as high as 50-100GPa in Mg e.g., but the potential error is rarely discussed. There is a post in this forum from some years back that asked this question before, but sadly didn't receive any answers: forum/viewtopic.php?p=15062&hilit=rcore#p15062.
Because I do not have an all electron code at hand, I did the next best thing, which is to compare two pseudopotentials for some polymorphs of Calcium, in this case the p- and s-valent potentials. The results are shown below, giving both E-V curves for both potentials, their difference as a function of "PAW overlap" as well as the overlap as a function of volume for the tested polymorphs.
I define "PAW overlap" as
\rho = 1 - \frac{r_nn}{2 r_core}
where r_nn is the smallest nearest neighbor distance in a structure and r_core is RCORE as read from POTCAR.
It seems that until an overlap of about 0.1 results can be trusted, giving about 1meV/atom error. For the s-valent potential of Ca this would translate to about 10-15A^3/atom volume or 50-ish GPa.
I'd very happy if anyone else can share their experience, I'm not even sure if RCORE is the correct radius to read from the POTCAR. As the linked forum posts mentions it would be extremely restrictive, especially for high Z elements.
Because I do not have an all electron code at hand, I did the next best thing, which is to compare two pseudopotentials for some polymorphs of Calcium, in this case the p- and s-valent potentials. The results are shown below, giving both E-V curves for both potentials, their difference as a function of "PAW overlap" as well as the overlap as a function of volume for the tested polymorphs.
I define "PAW overlap" as
\rho = 1 - \frac{r_nn}{2 r_core}
where r_nn is the smallest nearest neighbor distance in a structure and r_core is RCORE as read from POTCAR.
It seems that until an overlap of about 0.1 results can be trusted, giving about 1meV/atom error. For the s-valent potential of Ca this would translate to about 10-15A^3/atom volume or 50-ish GPa.
I'd very happy if anyone else can share their experience, I'm not even sure if RCORE is the correct radius to read from the POTCAR. As the linked forum posts mentions it would be extremely restrictive, especially for high Z elements.