Multiwfn forum

Hi Tian Lu,
                I would like to seek advice on a particular dilemma im having.

I am probing the electronic state/structure of a Ni cofactor that in literature is proposed to be 5 coordinate with a lower axial amino acid residue. (some suggestion is there for 6 coordinate with a potential weak coordination with the nature substrate(S)).

This Ni is proposed to be in the (I) state, which gives uncertainty on the 5 or 6 coordinate.

I was testing out several functionals for this system (annoyingly people were using B3LYP which im still using for comparsion)

I would like to discuss the use of the 4 functionals i would find pretty reliable here

PBE0-D4
R2-SCAN-D4 0 or 10 HF% exchange
WB97X-D4
TPSSH-D4

I have interestingly found that they converged to different geometries/wavefunctions

R2-SCAN and TPSSH gave a antiferromagnetically doublet with primariliy 2 alpha on Ni and a beta on the neighbouring carbon cofactor
Spin contamination is about ~ 1
Ni-S bond length of 2.48 (TPSSH) and 2.50 (R2-SCAN)
Mayer bond order of 0.7-0.8

PBE0 and WB97X gave a pure Ni(I) singlet that was more like 4 coordinate (or 5 - the below bond length feels a little too far even for a weak coordination)
Ni-S bond length of 3.07 (wB97XD) and 3.26 (PBE0)
Mayer bond order of 0.4-0.5 (I assumed this was a weak coordination but have been told to be careful in trusting this method)

From my understanding,
this is not a "problem" of overdelocalisation of DFT functionals (i.e self-interaction error) since the low or none HF exchange are giving a bond.
I am more afraid about the AF-doublet spin population i got from r2-scan and tpssh or im not sure if i can trust it?

What i have done now
1) re-Optimise r2-scan and tpssh with wb97x.gbw wavefunction
2) reoptimise tpssh with wb97x.gbw wavefunction
(I have tested stability of all wavefunctions with STABPerform true and they are all stable)

I wonder if the Mulliken spin populations might be unreliable and i should try a CASSCF single point to probe the spin populations on the R2-SCAN and TPSSH geometries.

I would like to seek advice on if my thought process thus far is fair
and any advice on which functional would be more reliable in this case or other methods i could use to probe the results thus far?

Thanks in advance ,
hehe

Hi Tian Lu,
                I would like some advice on optimising transition metal complex (or treatment of the electronic energies of transition metal complex). I have a metal center with a radical nearby that is antiferromagnetically coupled.

With hybrid DFT (10-25% HF exchange), spin-contamination will obviously pop up due to the use of single KS determinant. Due to this issue, your broken-symmetry energy can get "infiltrated" by higher spin states (spin contamination). Yamaguchi suggests using an alpha parameter to approximate the pure-spin state using your BS energy and HS energy (correcting the energy due to spin contamination). This alpha term is calculated using the spin of your system at HS and BS.

Introduced in orca (https://orca-manual.mpi-muelheim.mpg.de … ion-eq-eap),
They have presented a way to not only correct this spin contaminated energy but additionally, get an analytical gradient (from this spin-correction energy surface) to carry out geometry optimsation in orca.

I have couple of questions regarding this

1) If Hybrid DFT (or DFT for that matter)'s exchange correlation function is designed to work well for its parameters. How is using a single alpha parameter consistently for all functionals (I suspect the alpha parameter will not change significantly across the same spin system unless significantly large HF % for the hybrid DFT) suppose to be robust for all hybrid DFT functionals?

2) However, at the same time, the exchange correlation functional is parameterised for well-defined spin eigen-states (non spin contaminated states). So, using Yamaguchi's method to correct the errors associated with BS energies sounds fair as well. Although i have seen arguments that it needs more attention if core is more polarised/not closed shell (Phys. Chem. Chem. Phys., 2015,17, 14375-14382)

I have seen several papers that do geometry optimisation without Yamaguchi method applied and then do a single point energy with the Yamaguchi AP correction term. This has usually resulted in match with experimental (which definitely can be concidental...)

Coming to my 3rd question
3) Is it fair to apply to both geometry optimisation and single point energies or is it ok to apply to only single point energies ?
i guess it is difficult to say considering that it may or maynot change geometries significantly and when it does, it becomes a considerable task?

Appreciate your thoughts on this matter as im unsure on how to treat my system in the best manner.

Kind regards,
hehe