PBE energies of GAMESS vs QP2
Closed this issue · 5 comments
Hello,
I think there might be something off in PBE implemented in QP2.
I have attached two runs that use GAMESS for SCF versus QP2 for SCF to do CIS/PBE.
One can see that both PBE energies, as well as CIS energies, differ for these two.
Both cases use the same ECP, basis, etc.
My QP2 version:
Git Date : Tue Jun 16 18:00:02 2020 +0200
Git SHA1 : 470addc
Thank you.
Hi,
have you tried with another software to check that the Gamess result is correct?
If you do the CIS using PBE orbitals, and if the PBE calculations differ between the 2 codes, you will get different MOs and it is normal that the CIS/PBE energies differ.
Are the Hartree-Fock energies compatible between the 2 codes?
Hi,
have you tried with another software to check that the Gamess result is correct?
GAMESS: -6.2568293956 Ha
PySCF : -6.2568777711 Ha
QP2 : -6.2510636245 Ha
If you do the CIS using PBE orbitals, and if the PBE calculations differ between the 2 codes, you will get different MOs and it is normal that the CIS/PBE energies differ.
Are the Hartree-Fock energies compatible between the 2 codes?
Agreed, I tried CIS/PBE to see if the energy discrepancy between PBE was just an arbitrary shift. But apparently MOs are different. I don't have QP HF energy at the moment (Molpro gives: -6.08674468 Ha) but I don't recall seeing any discrepancies from other runs.
Thank you for looking at this.
Gani.
Hi,
I found the problem: The basis set for Si is different between your two calculations. If you run the PBE calculation starting from the GAMESS converted directory, you get:
-6.2570735 with grid_type = 2 (default) and
-6.2568561 with grid_type = 3
So the PBE is OK
Not sure how I missed that. Good to know PBE is fine. Thanks for looking again!