Unable to parse `LOCPOT` with VASP 6.4.2

[abhardwaj73]

atomate2 fails to create TaskDoc as it cannot parse the LOCPOT file generated after the VASP calculation. This happens if we use newer PBE 54 pseudopotential with hash or just using VASP version 6.4.2. The issue is where the element's name is printed in the LOCPOT file, which would be further used in the POSCAR file for the next steps.

Steps to reproduce the behavior:

1. Use any atomate2 workflow. (I'm using MPMetaGGADoubleRelaxMaker)
2. Use newer PBE 54 pseudopotentials with hash
   OR
3. Use older PBE 54 pseudopotentials without hash but use VASP 6.4.2 version.
4. See error :

File "/home/ab6989/.conda/envs/my_pymatgen/lib/python3.12/site-packages/atomate2/vasp/jobs/base.py", line 292, in get_vasp_task_document
return TaskDoc.from_directory(path, **kwargs)
 .....

File "/home/ab6989/.conda/envs/my_pymatgen/lib/python3.12/site-packages/pymatgen/core/periodic_table.py", line 1051, in from_str
properties = {tokens[0]: ast.literal_eval(tokens[1])}

IndexError: list index out of range

This is because of an extra “/” coming after the element's name in the LOCPOT file. Or, an extra “/hashnumber” in case of hash pseudopotentials. I've made sure of this by removing this extra bit and successfully running the task_doc = get_vasp_task_document("/home/ab6989/MPScanRelaxSet/atomate2/Tests/job_2023-12-18-13-57-46-455621-69653")

Examples:
LOCPOT file when using hashless pseudopotential but 6.4.2 version VASP.

Mg1 O1
   1.00000000000000
    -0.0000000000000000    2.1041807445817637    2.1041807445817637
     2.1041807445817637   -0.0000000000000000    2.1041807445817637
     2.1041807445817637    2.1041807445817637    0.0000000000000000
  Mg/             O/
               1               1
Direct
  0.0000000000000000  0.0000000000000000 -0.0000000000000000
  0.5000000000000000  0.5000000000000000  0.5000000000000000

LOCPOT file which works in creating task_doc and doesn't fail.

Mg1 O1
1.0
   0.0000000000000000    2.1299999999999999    2.1299999999999999
   2.1299999999999999    0.0000000000000000    2.1299999999999999
   2.1299999999999999    2.1299999999999999    0.0000000000000000
Mg O
1 1
direct
   0.0000000000000000    0.0000000000000000    0.0000000000000000 Mg
   0.5000000000000000    0.5000000000000000    0.5000000000000000 O

[abhardwaj73]

An easy workaround for this issue is to use VASP version 6.4.0.
If using the newer VASP version 6.4.2, need to update INCAR with LVTOT = "False" (to not parse from LOCPOT file). This can be done by using update_user_incar_settings command.

[rkingsbury]

So if I understand correctly, VASP 6.4.2 adds the extra / to the LOCPOT, but earlier versions do not? I wonder if this is intended behavior or a bug in VASP 6.4.2.

Ultimately, to fix this we'll need to figure out whether it's an issue with VASP, something to be updated in TaskDocument, or something to updated in pymatgen. Interested to hear if @janosh or @utf have noticed anything like this.

[abhardwaj73]

Correct, VASP 6.4.2 adds the extra / to the LOCPOT. It might be because the newer version is intended to work with the newer hash pseudopotentials.

[janosh]

@abhardwaj73 Thanks for bringing this up! Just to add some context from @sudarshanv01 (VASP dev) who confirmed the added slash in 6.4.2 as expected behavior:

• The slash is part of the LOCPOT header formats, most possible other files that have the same header. They are supposed to be there for labeling the PAW with the SHA number of the potential
• A potential fix is to change pymatgen parsing to make sure that the atomic species are parsed even if they have the /

@esoteric-ephemera added

we should fix this on the PMG side - it's going to affect CHG*, AECCAR*, and LOCPOT. Anything with charge density parsing will break

[esoteric-ephemera]

Hey @abhardwaj73 , this was hopefully fixed in a pymatgen PR. Would you be able to install pymatgen from git (pip install git+https://github.com/materialsproject/pymatgen.git) and let us know if this is working now?

[janosh]

closing as fixed in #3542 after migrating issue from atomate2 to pymatgen. new pymatgen release coming later this week.

[abhardwaj73]

Hello @esoteric-ephemera and @janosh, Thanks for your help in resolving these atomate2 issues. I used the pip install git+https://github.com/materialsproject/pymatgen.git version to run the MPMetaGGADoubleRelaxMaker atomate2 workflow, and got the following error for a FCC Silicon system:

2024-01-25 16:50:31,434 INFO MP pre-relax 1 failed with exception:
Traceback (most recent call last):
  File "/home/ab6989/.conda/envs/myenv/lib/python3.12/site-packages/jobflow/man                                                                                                                                                                                                               agers/local.py", line 114, in _run_job
    response = job.run(store=store)
               ^^^^^^^^^^^^^^^^^^^^
  File "/home/ab6989/.conda/envs/myenv/lib/python3.12/site-packages/jobflow/cor                                                                                                                                                                                                               e/job.py", line 583, in run
    response = function(*self.function_args, **self.function_kwargs)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/ab6989/.conda/envs/myenv/lib/python3.12/site-packages/atomate2/va                                                                                                                                                                                                               sp/jobs/base.py", line 218, in make
    write_vasp_input_set(
  File "/home/ab6989/.conda/envs/myenv/lib/python3.12/site-packages/atomate2/va                                                                                                                                                                                                               sp/files.py", line 202, in write_vasp_input_set
    vis.write_input(directory, potcar_spec=potcar_spec, **kwargs)
  File "/home/ab6989/.conda/envs/myenv/lib/python3.12/site-packages/atomate2/va                                                                                                                                                                                                               sp/sets/base.py", line 114, in write_input
    f.write(v.get_string(significant_figures=16))
            ^^^^^^^^^^^^
AttributeError: 'Poscar' object has no attribute 'get_string'. Did you mean: 'get_str'?                                                                                                                                                                               

2024-01-25 16:50:31,435 INFO Finished executing jobs locally

It looks like there is some issue in get_string as get_str. The workflow creates the INCAR file but the POSCAR doesn't get created even in the first step.

I also checked using 2 ways of generating structure: using a .cif file and directly giving lattice, species, and coord. But, it had the same error in both of them.

Let me know if it is due to some mismatch in different codes, or if I should make any changes in my workflow. I created a new environment and used the GitHub pymatgen and pip install atomate2 for checking it.