Subject: How to reproduce Materials Project setup with missing POTCARs

[mathias_stokkebyenissen]

Hello everyone,

I am in the process of validating calculations that use the computational setup of the Materials Project. However, I am unable to obtain several of the required POTCARs, as they do not seem to be distributed anymore. Specifically, I cannot find the following pseudopotentials:

PAW_PBE Li_sv 23Jan2001
PAW_PBE Na_pv 05Jan2001
PAW_PBE Ge_d 06Sep2000
PAW_PBE Ce 28Sep2000
PAW_PBE Eu 08Apr2002
PAW_PBE Gd 08Apr2002

What is the recommended approach in this situation? Should I substitute these with the newer default pseudopotentials, or is there another way to best approximate the Materials Project setup?

Any guidance would be greatly appreciated.

[michael_wolloch]

Dear Mathias Stokkebyenissen,

Thanks for your question!

As far as I know, all POTCARs that were ever distributed are still available on the portal and are listed on the Available pseudopotentials wiki page. But I have trouble locating the ones you are describing as well.

What I did find was e.g. a PW91 Li potential with a fitting date:

Code: Select all

PAW_GGA Li_sv 23Jan2001

from the PAW POTCAR files: LDA & PW91 & PBE (outdated) set.

Can you double-check if you really are looking for PBE potentials and get back to me? Also a link to the appropriate MP website would be helpful, I think.
Thanks, Michael

[mathias_stokkebyenissen]

Dear Michael,

I am making use of the MP API, from which the potcar spec of a given calculation is displayed (marked in red below) e.g. for a given Gd structure one can pull:

mp-570087-GGA ComputedStructureEntry - Gd3 (Gd)
Energy (Uncorrected) = -42.1675 eV (-14.0558 eV/atom)
Correction = 0.0000 eV (0.0000 eV/atom)
Energy (Final) = -42.1675 eV (-14.0558 eV/atom)
Energy Adjustments:
None
Parameters:
potcar_spec = [{'titel': 'PAW_PBE Gd 08Apr2002', 'hash': '1f0d42b1e5f6769d319d3f247992aeb9', 'summary_stats': None}]
run_type = GGA
is_hubbard = False
hubbards = None
Data:
oxide_type = None
aspherical = True
last_updated = 2024-11-21 20:44:15.041237+00:00
task_id = mp-1926743
material_id = mp-570087
oxidation_states = {'Gd': 0.0}
license = BY-C
run_type = GGA

[michael_wolloch]

Dear Mathias,

I looped over all my available POTCAR files (a couple of which are experimental, unreleased, and actually not known to pymatgen), computed the md5_header_hash, and compared it to the one you posted. I could not find a match.
I have to say this is a bit puzzling to me. I know that some old and deprecated POTCARs are still in the MP database, but I assume that those would have been included in the standard releases.
Of course, there might be an issue with hash computations, but I could also not find the corresponding TITLE string.

If you really need the exact potentials, I would ask a question in the MP forums at matsci:
https://matsci.org/c/materials-project/8

Otherwise, I would use newer POTCARs. E.g., Gd PAW_PBE Gd 23Dec2003 from the potpaw_PBE.52 set.

Be advised that for most of the Lanthanides, old potentials are flawed significantly and have been updated in the 64 release. See the VASP specific sections in https://www.nature.com/articles/s42254-023-00655-3

Cheers, Michael

[mathias_stokkebyenissen]

Dear Michael,

Thanks for looking in to this so far.

Does this mean that the "VASP-team" no longer are in the possession of these POTCARs? As you can see on matsci: https://matsci.org/t/missing-potcars-fo ... ions/65576 I am not exactly sure that the Materials Project is willing or legally able to give out these POTCARs

just for good measure, here are the hash of all the missing POTCARs:
'titel': 'PAW_PBE Li_sv 23Jan2001', 'hash': '4799bab014a83a07c654d7196c8ecfa9'
'titel': 'PAW_PBE Na_pv 05Jan2001', 'hash': 'c71d0ed99a871c91fccae9347860d8ba'
'titel': 'PAW_PBE Ge_d 06Sep2000', 'hash': '648336e7599c2dd05b1ddacace0213eb'
'titel': 'PAW_PBE Ce 28Sep2000', 'hash': 'ff3a09f2ff91798e58eb4b9854e9be4a'
'titel': 'PAW_PBE Eu 08Apr2002', 'hash': 'd466d046adf21f6146ee9644049ea268'
'titel': 'PAW_PBE Gd 08Apr2002', 'hash': '1f0d42b1e5f6769d319d3f247992aeb9'

These POTCARs are found be querying the MP-API:

Code: Select all

from mp_api.client import MPRester
mpr = MPRester("your-API-key")
with mpr as m:
    entries = m.get_entries_in_chemsys('Gd')  # Or by material_id
entries[0]

# output:

Retrieving ThermoDoc documents: 100%|██████████| 5/5 [00:00<00:00, 93622.86it/s]

mp-570087-GGA ComputedStructureEntry - Gd3          (Gd)
Energy (Uncorrected)     = -42.1675  eV (-14.0558 eV/atom)
Correction               = 0.0000    eV (0.0000   eV/atom)
Energy (Final)           = -42.1675  eV (-14.0558 eV/atom)
Energy Adjustments:
  None
Parameters:
  potcar_spec            = [{'titel': 'PAW_PBE Gd 08Apr2002', 'hash': '1f0d42b1e5f6769d319d3f247992aeb9', 'summary_stats': None}]
  run_type               = GGA
  is_hubbard             = False
  hubbards               = None
Data:
  oxide_type             = None
  aspherical             = True
  last_updated           = 2024-11-21 20:44:15.041237+00:00
  task_id                = mp-1926743
  material_id            = mp-570087
  oxidation_states       = {'Gd': 0.0}
  license                = BY-C
  run_type               = GGA

[michael_wolloch]

Hi Mathias,

Does this mean that the "VASP-team" no longer are in the possession of these POTCARs?

No, it just means that I have not found POTCAR files with those title strings, or that match the hashes.

However, those strings and hashes are created by pymatgen from the raw strings of the files. Those are NOT the VASP computed hashes that are the preferred unique identifiers for POTCAR files.

At the moment, I assume that the confusion stems from the pymatgen processing, but I have no way to know for sure.
I commented on your post on matsci.org, and I am searching a couple of more obscure places for old and/or experimental pseudopotential files as well.

Thanks for the additional hashes. I will update you as soon as I find out more.
Michael

[michael_wolloch]

Dear Mathias,

With the Help of Aaron Kaplan from the Materials Project, I located the POTCAR files in question. I sent you copies in an email. I will see if we can provide the whole Tarball on the portal, but I cannot give any promises right now.

Let me know if you have another question on this topic, or if I can close it for now.

Cheers, Michael

[taoshuo950111]

Hi Michael,

Could you also send me a copy of these POTCARs, I also have a hard time to reproduce the total energy reported on MP for Gd2Fe17 (mp-1196805) and
Gd2Co17 (mp-1201816). Currently I'm using the (PAW_PBE Gd 23Dec2003) version from VASP PBE52 POTCAR set, which does not agree with MP.

Regards,
Shuo

[michael_wolloch]

Hi Shuo,

I am sorry, somehow, I did not get a notification about your post, and I am seeing it just now.

• Do you still need the Gd POTCAR?

• Is it the one from 08Apr2002 that MP uses?

• Are there more of these old POTCARs you need?

Cheers, Michael