PySCF interface retrieves incorrect atomic numbers if ECPs are used #247
Description
Describe the bug
Hej!
I wonder whether it is ok that ECP does affect a dispersion energy:
%cat b3lyp-d4.py
from pyscf import gto, scf, dft
import dftd4.pyscf as disp
mol = gto.M()
# 12 atom(s)
mol.atom = '''
C -1.207353289 -0.697065746 0.000000000
C -1.207353289 0.697065746 0.000000000
C 0.000000000 1.394131493 0.000000000
C 1.207353289 0.697065746 0.000000000
C 1.207353289 -0.697065746 0.000000000
C 0.000000000 -1.394131493 0.000000000
H -2.142871219 -1.237187275 0.000000000
H -2.142871219 1.237187275 0.000000000
H 0.000000000 2.474374550 0.000000000
H 2.142871219 1.237187275 0.000000000
H 2.142871219 -1.237187275 0.000000000
H 0.000000000 -2.474374550 0.000000000
'''
mol.charge = 0
mol.spin = 0
mol.verbose = 4
mol.build()
d4 = disp.DFTD4Dispersion(mol, xc="b3lyp")
print("w/o ECP:",d4.kernel()[0])
mol.ecp = 'ccecp'
mol.build()
d4 = disp.DFTD4Dispersion(mol, xc="b3lyp")
print("w/ ECP:",d4.kernel()[0])
%python b3lyp-d4.py
w/o ECP: -0.01734794021287107
w/ ECP: -0.021050072978489192
Package versions used:
Python 3.11.5
dftd4 3.6.0
pyscf 2.3.0
Thank you in advance!
With best regards,
Victor