HF ESP slightly different as g09

1

Hi,

I just started using psi4 and was playing around a bit. I am trying to reproduce ESP values from g09 and cubegen and I somehow discovered that ESP values are approx. 5% higher in psi4 compared to g09. Not sure why. Probably a few default values differ between psi4 and gaussian?

my Psi4 Input:
import os, sys
import psi4

psi4.core.set_output_file(‘output.dat’, False)
psi4.set_num_threads(8)

#! Sample Computation

psi4.set_memory(‘12000 MB’)

h2o = psi4.geometry("""
0 1
C -1.20744 -0.23653 -0.01964
C 0.08998 0.55291 0.04457
O 1.22232 -0.26135 -0.10754
H -1.27162 -0.78375 -0.95443
H -2.06573 0.42573 0.05493
H -1.26862 -0.95308 0.79687
H 0.14342 1.11190 0.97807
H 0.13002 1.27160 -0.76457
H 1.25876 -0.87987 0.59988
“”")

psi4.set_module_options(‘scf’,{‘e_convergence’: ‘1e-8’})

E,wfn = psi4.energy(‘SCF/aug-cc-pVTZ’, return_wfn=True)
psi4.oeprop(wfn,‘GRID_ESP’)

and the g09 input:
%nprocshared=8
%mem=12GB
%chk=chk_mol0.chk
#HF/aug-cc-pVTZ pop=MK density=SCF iop(6/33=2,6/41=10,6/42=17,6/50=1)

MP2 Point

0 1
C -1.20744 -0.23653 -0.01964
C 0.08998 0.55291 0.04457
O 1.22232 -0.26135 -0.10754
H -1.27162 -0.78375 -0.95443
H -2.06573 0.42573 0.05493
H -1.26862 -0.95308 0.79687
H 0.14342 1.11190 0.97807
H 0.13002 1.27160 -0.76457
H 1.25876 -0.87987 0.59988

ethanol.gesp

Any thoughts?

Thanks for the help!
Michael

2

Could be a density fitting issue, worth trying:

psi4.set_options({"SCF_TYPE": "PK", "E_CONVERGENCE": 1.e-8})

I wouldn’t recommend messing with set_module_options too much in general unless you are after something very specific.

3

Thanks for the help!

Unfortunately no luck. The suggested settings changed almost nothing.

I also tried to run the same on an MP2 level and the calculated ESP is really similar to the psi4 HF ESPs and again off in comparison to g09:

My input:
import os, sys
import psi4

psi4.core.set_output_file(‘output.dat’, False)
psi4.set_num_threads(8)

#! Sample Computation

psi4.set_memory(‘10000 MB’)

h2o = psi4.geometry("""
0 1
C -1.20744 -0.23653 -0.01964
C 0.08998 0.55291 0.04457
O 1.22232 -0.26135 -0.10754
H -1.27162 -0.78375 -0.95443
H -2.06573 0.42573 0.05493
H -1.26862 -0.95308 0.79687
H 0.14342 1.11190 0.97807
H 0.13002 1.27160 -0.76457
H 1.25876 -0.87987 0.59988
“”")
E,wfn = psi4.energy(‘MP2/aug-cc-pVTZ’, return_wfn=True)
psi4.oeprop(wfn,‘GRID_ESP’)

4

An energy evaluate of MP2 does not evaluate a MP2 density. Please use the properties interface for one-electron computations:

properties("MP2/aug-cc-pVDZ", properties="GRID_ESP")

It was doubtful that DF vs non-DF would change much, especially in terms of densities. If you compute a simpler molecule like H2 do you see similar differences? Acetic acid could land on multiple minima with nearly identical energies.

5

It seems your feeling was correct for ethanol :grinning:.For H2 the ESPs agree.
Something else:
I can only run the property command in the psi executable. Is there also a way to access the property command via the python API? Is there a difference between properties and property and prop?

6

property was discontinued b/c it’s also a python keyword. prop is just an alias. psi4.properties() in PsiAPI should work just fine (https://github.com/psi4/psi4/blob/master/tests/pytest/test_psi4.py#L211).

7

Thanks. I think I messed up something during installation. My psi4 does not have an attribute properties. I installed psi4 using conda. I guess I forgot to set a variable.

8

possibly you’re using v1.1, the c. May 2017 release via -c psi4. you probably want the latest nightly build, -c psi4/label/dev. make a new conda env if you are in fact at 1.1.

9

That solved the problem. Thanks!
One more question: Is it possible to calculate and save the MP2 density to a file and use it later to calculate the ESPs with different grid.dat files?

10

Unfortunately not quite yet. It will be enabled by solving the following issue: https://github.com/psi4/psi4/issues/887

I would drop a line there saying that it would be useful to you and if you want to tackle it we can certainly help you through the process!