incorrect crystal generation from seed

Dear developers, I have been trying to perform some analysis on a structure and apparently pyxtal fails to import it correctly. Here is the structure: https://gist.github.com/gpetretto/8f9849426b8eb465f89cf883b036fd98 and here is a piece of code demonstrating the issue.

from pymatgen import Structure
from pymatgen.analysis.structure_matcher import StructureMatcher
from pyxtal import pyxtal


s = Structure.from_file("FAU.cif")

c = pyxtal()
c.from_seed(s)
sc = c.to_pymatgen()
sc.to(filename="conv.cif")

sm = StructureMatcher()
print(sm.fit(s, sc))

Aside from the StructureMatcher failing, just looking at the final Structure is enough to tell that it is completely different from the original one.

I tried to check the code, but it seems that there is no problem in the conversion to/from pymatgen and the issue seems to happen when the instances of atom_site are generated. More precisely when the sym_ops are applied in the search_position and update method, but I could not identify the exact source of the issue. I am not sure if this will be of much help…

[qzhu2017]

@gpetretto I have fixed most of the issues as discussed in (spglib/spglib#114).

In my test, all zeolites return the same structures after symmetrization. Most of mp structures work as well. You can try the following code

from pymatgen import Structure
from pymatgen.analysis.structure_matcher import StructureMatcher
from pyxtal import pyxtal
from glob import glob

path = "cif_files/wrong_cif_zeolites/"
path = "cif_files/wrong_cif_mp/"

for f in glob(path+'/*.cif'):
    try:
        s = Structure.from_file(f)
        c = pyxtal()
        c.from_seed(s, tol=1e-6)
        sc = c.to_pymatgen()
        sm = StructureMatcher()
        if not sm.fit(s, sc):
            print(f, c.group.number, sm.fit(s, sc))
    except:
        print(f, "wrong")

I checked some structures which return mismatch. However, it seems that they are somewhat due to numerical issues.

One remaining issue, spglib sometime detects a wrong symmetry. For instance, the structure LTN.cif has a space group symmetry Fd-3m, but spglib returns a P2 symmetry.

This happens to only a few cases. Will check it later.

[qzhu2017]

Checked the failed mp structures. It is highly to be due to decimal issue.