Semi-empirical calculations-basic question

[k.cerda]

Hi,

I'm just learning to use QuantumATK for my thesis work. I need to simulate organic molecules which contain 5500-15.000 atoms. Therefore DFT does not seem to be feasible due to the large number of atoms. I turned to Semi-Empirical method and it also does not seem to be capable of converging with such number of atoms.  Today I was reading the QuantumATK manual pages and came across the following paragraph:

"Note

The Hückel parameters have been fitted for non-self-consistent calculations. To use the parameters in self-consistent calculations, the self-consistent onsite shifts must be compensated by a reverse shift of the vacuum_levels."

on page: https://docs.quantumatk.com/manual/SE.html.

Combining this expression with the fact that the no SCF checkbox is checked by default in Semi-Empirical Calculator, can we say that the results of  non-self-consistent calculations with Semi Empirical Calculator in QuantumATK are valid and accurate? In other words, are the results of no SCF Semi-Empirical calculations valid? I understand that self-consistent field is no doubt more accurate but given the very large number of atoms, can the results of non-self-consistent Semi Empirical package be considered valid for a thesis or publication work?

Thanks,
K. Cerda

[nisreen]

Dear Cerda,

I'm facing the same issue with my thesis, and I can't get any converged results with self consistent option using the semi empirical approach, did you reach to any result if we can rely on the non self consistent results? I've been searching for so long for any answer but was unable to find anything

Regards,

[Anders Blom]

It's very hard to give a definite answer, but as you noted the Huckel model is actually designed to be non-selfconsistent, and is also originally developed for organic systems, so I think at the very least the results will be "representative" enough to use for real work. You can perhaps search the literature for other work on similar systems using the same model, and see how they treat SCF vs non-SCF.

Actually, if I recall correctly, we mainly added the SCF option for the semi-empirical models to use them in transport simulations, where definitely the effect of charge transfer is crucial, and a non-SCF solution would not be accurate.

Btw, it is not entirely impossible to use DFT for the system size  you mention, at least with a small basis set like SingleZetaPolarized. You might want to give it a try at least :-)