PointGroupAnalyzer reports incorrect point group for C2v molecules

[bjmorgan]

Correction (2026-04-10): The 4-atom example below is actually D2d, not C2v — thanks to @newtontech for catching this in the linked PR. See this comment for a corrected 4-atom C2v example.

Python version

Python 3.12.3

Pymatgen version

2025.10.7

Operating system version

macOS (Darwin 24.6.0)

Current behavior

PointGroupAnalyzer misidentifies the point group for molecules with C2v symmetry.

For a 4-atom C2v molecule, it reports D2d at all tolerances tested (0.01, 0.1, 0.3).

For the 7-atom "Square-face capped trigonal prism" reference geometry from pymatgen's own chemenv module (AllCoordinationGeometries), results are tolerance-dependent:

• tol=0.01: Cs (misses the sigma_h mirror plane)
• tol=0.1: D2d (detects spurious symmetry)
• tol=0.3: D2d (detects spurious symmetry)

The sigma_h mirror plane in the 7-atom case is exact to machine precision: applying the permutation [3,4,5,0,1,2,6] produces coordinates identical to negating the z-coordinates, with zero residual.

Expected Behavior

Both molecules should be identified as C2v at all reasonable tolerances.

Minimal example

import numpy as np
from pymatgen.core import Molecule
from pymatgen.symmetry.analyzer import PointGroupAnalyzer

# Case 1: 4-atom C2v molecule
# Two atoms paired in z, two on the mirror plane offset in y
mol = Molecule(["H"] * 4, [
    [0, 0, +1],
    [0, 0, -1],
    [+1, -1, 0],
    [-1, -1, 0],
])

for tol in [0.01, 0.1, 0.3]:
    pga = PointGroupAnalyzer(mol, tolerance=tol)
    print(f"tol={tol}: {pga.get_pointgroup()}")
# All report D2d; expected C2v

# Case 2: pymatgen's own chemenv reference geometry
from pymatgen.analysis.chemenv.coordination_environments.coordination_geometries import AllCoordinationGeometries

acg = AllCoordinationGeometries()
pts = acg.get_geometry_from_name('Square-face capped trigonal prism').points
mol2 = Molecule(["H"] * 7, pts)

for tol in [0.01, 0.1, 0.3]:
    pga = PointGroupAnalyzer(mol2, tolerance=tol)
    print(f"tol={tol}: {pga.get_pointgroup()}")
# tol=0.01: Cs, tol=0.1: D2d, tol=0.3: D2d
# Expected: C2v

Relevant files to reproduce this bug

No response

[bjmorgan]

Thanks to @newtontech for catching this in the linked PR (newtontech#1) — the original 4-atom example is actually D2d, so pymatgen was right about it. Apologies for the bad example.

Here is a corrected 4-atom C2v example that triggers the same bug as the 7-atom case:

import numpy as np
from pymatgen.core import Molecule
from pymatgen.symmetry.analyzer import PointGroupAnalyzer

mol = Molecule(["H"] * 4, [
    [0, 0,  1],
    [0, 0,  3],
    [ 3, 0, -2],
    [-3, 0, -2],
])
for tol in (0.01, 0.1, 0.3):
    print(f"tol={tol}: {PointGroupAnalyzer(mol, tolerance=tol).get_pointgroup()}")
# All report Cs; expected C2v