Multiwfn forum

sblau

Member

Registered: 2024-06-12

Posts: 1

NBO in ORCA versus NBO run on 0.47 file generated by Multiwfn

(This is a question about Multiwfn, but please be patient as I provide relevant context)

I have encountered an issue in ORCA where running NBO analysis takes a shockingly long amount of time. In this example NBO analysis took almost 22 hours! Far longer than the actual SCF, which should never be the case in my experience. In order to debug this, I want to run NBO externally with Gennbo. Thus, I have converted my gbw file into a molden input file with orca_2mkl and read this file into Multiwfn and converted it into a .47 file. In the process, I got the message stating that "This function cannot be used if any spherical-harmonic basis function has >=f angular moment!", so I set iloadasCart=1 in settings.ini. I understand that this means that the Fock matrix will be flawed, and thus I will not be able to get the correct E2PERT info from NBO, but life goes on. I then ran Gennbo.sh on the .47 file and obtained this output.

Unfortunately, the external NBO output and the ORCA NBO output look completely different. The differences begin immediately, with the external NBO reporting:

Alpha MOs are non-orthogonal (max Sij=0.011); continue without MOs
Beta MOs are non-orthogonal (max Sij=0.011); continue without MOs

Then more linear dependencies are found than were reported in the ORCA NBO output, then negative population(s) and Pauli-violating population(s) are found and fixed, which wasn't the case in the ORCA NBO output. Then the Al S electrons are in the 5s and 6s Rydberg orbitals (clearly wrong) rather than the 1s and 2s core orbitals, as is the case in the ORCA NBO output.

My only conclusion is that something has gone very wrong in the conversion of the .gbw to .47. Am I just doing something very foolish? Help would be much appreciated!

Sincerely,
Sam

sobereva

Tian Lu (Multiwfn developer)

From: Beijing

Registered: 2017-09-11

Posts: 1,777

Website

Re: NBO in ORCA versus NBO run on 0.47 file generated by Multiwfn

Dear Sam,

Cartesian basis functions have much stronger linear depenency problem than spherical-harmonic basis functions (perhaps this is severer when Cartesian basis function coefficients were not varitionally optimized in SCF procedure). I suggest using a basis set without angular moment >=f for present purpose. NBO analysis is insensitive to basis set, so a relatively small basis set is adequate to obtain a reasonable result.

Best,

Tian