Molcas Forum

andrewshyichuk

Member

Registered: 2020-02-13

Posts: 86

Custom (hybrid) functionals in SCF KS-DFT / Minor B3LYP issue

Dear Community,

Is there a way in Molcas SCF (KS-DFT) to:
- separately specify exchange and correlation functionals?
- modify the mixing constants in hybrids?

My aim is to get exact (HF) exchange with some correlation.
It can be done via setting the non-HF part of a hybrid functional to zero, or via setting exchange=HF, correlation=LDA (or whatnot).
Hence the question.

The only off-the-shelf option with pure HF exchange is TLYP, although the manual is a bit confusing.
The formula suggests the non-local part of the correlation only, while the description says "where the non-local correlation functional is the LYP functional".
Description for BLYP says, "Correlation functional of Lee, Yang, and Parr, which includes both local and non-local terms".
So, in TLYP, is it a LYP correlation as is, or only the non-local part of it?

The reference for B3LYP does not clarify that: it is a Becke's paper where he gives the idea of hybrid functional and uses the PW correlation.

LYP, however, is used (to make B3LYP) later in Gaussian 92 (according to Stephens et al, https://pubs.acs.org/doi/pdf/10.1021/j100096a001).
The latter paper also says that "LYP does not have an easily separable local component" and gives a formula for B3LYP with LYP correlation as is.

So, what is TLYP, exactly?

Thank you in advance.
Andrew

Last edited by andrewshyichuk (2021-08-30 11:48:02)

Ignacio

Administrator

From: Uppsala

Registered: 2015-11-03

Posts: 1,085

Re: Custom (hybrid) functionals in SCF KS-DFT / Minor B3LYP issue

There's the DFCF keyword, if it works. I'm not aware of any way to modify the hybrid mixing.

As far as I understand it, TLYP is HF+LYP. It says that as "non-local" correlation it uses LYP, the fact that LYP includes both local and non-local doesn't change anything. It doesn't say "the non-local part of LYP", it says just "LYP" (and the code seems to confirm it). Is it wrong? Maybe, but all functionals are wrong

andrewshyichuk

Member

Registered: 2020-02-13

Posts: 86

Re: Custom (hybrid) functionals in SCF KS-DFT / Minor B3LYP issue

Dear Ignacio,

Thanks for clarifying the things

As for DFCF - as I see from the code, it will only change the DFT part of hybrids, i.e. PBE0 exchange will be 25% HF exchange and 75%*DFCF PBE exchange, meaning it's not possible to make a 100% HF exchange PBE0.

Best regards.
Andrew