Incorrect Band gaps from simple calculations

[Thomas Kasel]

Hello All,

As a test I am trying to accurately (relatively) reproduce the band gaps of silicon and graphene structures using PBE. I have however, run into an issue where graphene gives a band gap of 1.2 eV while silicon gives a band gap of 0.18 eV, the exact opposite of what one would expect. I would like to ask what it is I am doing wrong in my calculations. Attached is a zip file containing all my input files including basis set and pseudopotentials.

Additionally, I cannot seem to get the SCF to converge unless I specify:

&OT

 PRECONDITIONER FULL_ALL

 MINIMIZER CG

&END OT

Why is this setting so critical to SCF convergence in these cases?

Cheers and thanks in advance,

Tom

[Jürg Hutter]

Hi

For silicon: if you are using the unit cell (with 2 atoms) you need a large number of k-points.
In this case SCF needs diagonalization and an appropriate mixing scheme. I would also add smearing
of occupation numbers. If you have this working, you can start investigating basis sets for convergence.

For graphene: The problem is the Dirac cone at the K special point. Selecting a matching cell and k-point
sampling is mandatory to get correct results. There is ample literature addressing this problem.

regards
JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Thomas Kasel <tomk...@gmail.com>
Sent: Monday, September 9, 2024 10:50 PM
To: cp2k
Subject: [CP2K:20677] Incorrect Band gaps from simple calculations

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/cf2a1d8d-c40c-4da4-9857-38cd2b7e7e91n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/cf2a1d8d-c40c-4da4-9857-38cd2b7e7e91n%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Thomas Kasel]

Thanks to your suggestions I managed to get the SCF to converge for PBE based upon playing with the mixing. The only problem is now that CP2K is not printing the HOMO-LUMO gap. Is there a setting I can specify to force print the gap?

Regards,

Tom

[Thomas Kasel]

I managed to find a way to print the HOMO - LUMO gap from the pdos. The only problem now is that for some reason I get a band gap of 85 eV for silicon when using a hybrid functional.

[Jürg Hutter]

Hi

Obviously, you have a serious problem in your (secrete) input.
Without further information it is impossible to help.

regards
JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Thomas Kasel <tomk...@gmail.com>

Sent: Wednesday, September 11, 2024 10:34 PM
To: cp2k
Subject: Re: [CP2K:20694] Incorrect Band gaps from simple calculations

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/519f408c-fb0c-4c48-9a70-a36fbdc02f23n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/519f408c-fb0c-4c48-9a70-a36fbdc02f23n%40googlegroups.com?utm_medium=email&utm_source=footer>.

[Thomas Kasel]

The original files are in the first posting. But I suppose those are no longer relevant. I have attached my input,  basis set, and pseudopotential for silicon. Thank you for all your help.

Cheers,

Tom

[Lucas Lodeiro]

Hi

there are many weird things from my point of view. 

the main ones are: 

*CUTOFF of 600 eV (44 Ry), which is less than the used RELCUTOFF of 60 Ry... you must use N times the RELCUTOFF for the CUTOFF.

*small cell without kpoints... you must use supercell (4x4x4 for example) or kpoint mesh to properly get the electronic structure.

*your functional definition does not make sense to me... which functional form are you pointing out?

*why add all virtual orbitals? it is not useful at all, and very expensive.

Regards - Lucas

To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/d83110d8-d806-4b3c-b35e-c3050ad192b1n%40googlegroups.com.

[Thomas Kasel]

I will admit I am new to CP2K and used a lot of setting I found online in various powerpoints. 

I am trying to call the HSE06 functional.

Regards,

Tom

[Lucas Lodeiro]

Here you can find detailed information on how to use a correct input for hybrid functionals in CP2K, particularly there is a detailed input for HSE06.

https://www.cp2k.org/_media/events:2018_summer_school:cp2k-uk-stfc-june-2018-sanliang-ling.pdf

https://github.com/cp2k/cp2k/tree/master/tests/QS/regtest-hfx-wfn-fitting

Regards - Lucas

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ca98c3ff-4af2-4797-822a-085d9e89c007n%40googlegroups.com.

[Thomas Kasel]

Much appreciated!

Regards,

Tom