Population analysis of HF & CCSD

Questions
#1

hello!

I am a user of QChem 6.0.2.

I am doing a comparison between Hartree-Fock, CCSD and DFT for partial charges of a molecule using population analysis.

When I used the Hirshfeld method, I checked the output file and found that the charges of HF and CCSD are almost the same.

I’m wondering why these values are the same. Is it possible that QChem outputs the charges based on HF even though it does CCSD calculations?

I wonder if I can trust these results.

Below is my codes.

#Code1
$molecule
0,1
C       0.000000        0.000000        0.489758
H       0.000000        0.000000        1.576776
H       0.000000        0.000000        -1.018930
O       0.000000        1.047477        -0.218525
O       0.000000        -1.047477       -0.218525
$end

$rem
   METHOD          ccsd
   BASIS           def2-tzvppd
   POP_MULLIKEN    TRUE
   HIRSHFELD       TRUE
   WRITE_WFN       ccsd
$end

#Code2
$molecule
0,1
C       0.000000        0.000000        0.489758
H       0.000000        0.000000        1.576776
H       0.000000        0.000000        -1.018930
O       0.000000        1.047477        -0.218525
O       0.000000        -1.047477       -0.218525
$end

$rem
   METHOD          hf
   BASIS           def2-tzvppd
   POP_MULLIKEN    TRUE
   HIRSHFELD       TRUE
   WRITE_WFN       hf
$end

For CCSD, charge was

          Hirshfeld Atomic Charges

     Atom                 Charge (a.u.)
  ----------------------------------------
      1 C                     0.270998
      2 H                     0.093885
      3 H                     0.194615
      4 O                    -0.279735
      5 O                    -0.279735
  ----------------------------------------
  Sum of atomic charges =     0.000028

and for HF, charge was

          Hirshfeld Atomic Charges

     Atom                 Charge (a.u.)
  ----------------------------------------
      1 C                     0.270923
      2 H                     0.093901
      3 H                     0.194651
      4 O                    -0.279723
      5 O                    -0.279723
  ----------------------------------------
  Sum of atomic charges =     0.000028
#2

These are HF charges in both cases, any differences you see are likely the result of thresholds being set differently for HF vs. CCSD. For CCSD, you would need charges based on the “relaxed” density. Those are available for some models (MP2 I think, if you use JOBTYPE = FORCE) but probably not for CCSD. In any case it would need to be a force (gradient) job because that’s what triggers computation of the relaxed density.

1 Like
#3

Add a rem “cc_ref_prop = true” will change the Mulliken charges to CC’s result. The dipole moment gets updated too. However, the Hirshfeld charges still remain to be the HF results.

1 Like
#4

Thanks!

Can you tell me why the Hirshfeld result doesn’t change?

#5

It is not implemented for use with the relaxed density. If you feel that you really need this, please contact Q-Chem support and make a feature request. However, I have trouble believing the results would be very different (that’s probably why no one has bothered to program it), so this is unlikely to be a priority. It would also be slightly more annoying to implement as compared to other wavefunction-derived charges, due to the promolecule density that would presumably need to be evaluated at the correlated level of theory also (i.e., doing atomic CCSD calculations). Are you expecting to see some significant difference when post-Hartree-Fock correlation is added?

#6

Yes, based my impression, the Hirshfeld population analysis is currently only implemented for HF/DFT calculations.