wB97X-3c energy is inconsistent with those with other methods

water-3c.txt (47.7 KB)
Dear Psi4 community,

I have tried 3c composite methods available from version 1.9.
For starters, I computed water energy with several 3c methods.
All obtained energies seemed reasonable except the wB97X-3c case.
The code is shown below.

import psi4


h2o = psi4.geometry("""
                    H 1 0.96
                    H 1 0.96 2 104.5

e1 = psi4.energy('hf3c', molecule=h2o)
print(f'energy of HF-3c: {e1} au.')
e2 = psi4.energy('pbeh3c', molecule=h2o)
print(f'energy of PBEh-3c: {e2} au.')
e3 = psi4.energy('b973c', molecule=h2o)
print(f'energy of B97-3c: {e3} au.')
e4 = psi4.energy('r2scan3c', molecule=h2o)
print(f'energy of r2SCAN-3c: {e4} au.')
e5 = psi4.energy('wb97x3c', molecule=h2o)
print(f'energy of wB97X-3c: {e5} au.')

After running the code, I got the following values.

energy of HF-3c: -75.51179001562714 au.
energy of PBEh-3c: -76.25794311950531 au.
energy of B97-3c: -76.397690483936 au.
energy of r2SCAN-3c: -76.41834613518611 au.
energy of wB97X-3c: -17.27093036339403 au.

My environment on my Mac (apple M3) is the followings:

dftd3-python              1.0.0           py312h75e9645_2    conda-forge
dftd4                     3.6.0                h39831be_0    conda-forge
dftd4-python              3.6.0           py312he784a9e_1    conda-forge
gcp-correction            2.3.1                hbf1e955_2    conda-forge
psi4                      1.9.1           py312hfc7ebee_1    conda-forge
python                    3.12.3          h4a7b5fc_0_cpython    conda-forge

The log file is also attached.
Is this a sort of bug or am I missing something?

I would appreciate your help.
Thanks in advance.

Best wishes,

wB97X-3c replaces the oxygen core orbitals with an ECP, so you should expect the computed energy to differ from all electron results.

1 Like

Aha, I forgot that the meaning of ā€œ3cā€ has slightly changed.
wB97X-c does not use gCP but does ECPs.

Thanks for the pointer.