Possibly incorrect electron_configuration returned by vasp.inputs.PotcarSingle

[pranabdas]

Python version

3.10.16

Pymatgen version

2025.1.24

Operating system version

No response

Current behavior

It seems pymatgen returns incorrect valence configuration while parsing VASP POTCAR. Consider following code, where the POTCAR file corresponds to Hg (VASP-5.4, GGA, PBE, PAW, default):

from pymatgen.io.vasp.inputs import PotcarSingle

with open("POTCAR", "r", encoding="utf-8") as file:
    data = file.read()

print(PotcarSingle(data).electron_configuration)

It prints:

[(5, 'd', 10), (4, 'f', 14)]

Expected Behavior

Previously, I tested pymatgen==2023.8.10, it used to return:

[(6, 's', 2), (5, 'd', 10)]

The second line (valence) in POTCAR file is:

12.0000000000000 

Also, there is this line:

   POMASS =  XXX.XXX; ZVAL   =   12.000    mass and valenz

So, I suppose the previous behavior (i.e, [(6, 's', 2), (5, 'd', 10)], 12 valence electrons) was correct.

Atomic configuration block:

   Atomic configuration
   15 entries
     n  l   j            E        occ.
     1  0  0.50         XXX      2.0000
...
     4  3  3.50         XXX     14.0000
     5  0  0.50         XXX      2.0000
     5  1  1.50         XXX      6.0000
     5  2  2.50         XXX     10.0000
     6  0  0.50         XXX      2.0000
     6  1  1.50         XXX      0.0000

Minimal example

We can create a temporary directory and create virtualenv for testing:

mkdir test
cd test
python -m venv .venv
source .venv/bin/activate
pip install pymatgen

Then please use POTCAR of Hg and run the above python code.

To test older version of pymatgen, do the same instead install:

pip install pymatgen==2023.8.10 numpy==1.*

Relevant files to reproduce this bug

Sorry, but you need access to VASP POTCAR files, which cannot be shared due to copyright issues.

[DanielYang59]

Hi thanks for reporting this and looks like a bug to me, and I could recreate this issue :)

from pymatgen.core import Element


ele_hg = Element.from_Z(80)

print(ele_hg.full_electronic_structure)
# [(1, 's', 2), (2, 's', 2), (2, 'p', 6), (3, 's', 2), (3, 'p', 6), (4, 's', 2), (3, 'd', 10), (4, 'p', 6), (5, 's', 2), (4, 'd', 10), (5, 'p', 6), (6, 's', 2), (4, 'f', 14), (5, 'd', 10)]

pymatgen/src/pymatgen/core/periodic_table.py

Lines 42 to 62 in 5b997f7

	_madelung = [
	(1, "s"),
	(2, "s"),
	(2, "p"),
	(3, "s"),
	(3, "p"),
	(4, "s"),
	(3, "d"),
	(4, "p"),
	(5, "s"),
	(4, "d"),
	(5, "p"),
	(6, "s"),
	(4, "f"),
	(5, "d"),
	(6, "p"),
	(7, "s"),
	(5, "f"),
	(6, "d"),
	(7, "p"),
	]

---

I guess it's related to the change in #3944 and we would need to fix how the valence shell is determined