Graphene problems with hybrid functional

[Lucas Lodeiro]

Hi all,

I am trying to compute a hybrid functional calculation (PBE0-T_C_LR) over a graphene structure, and at the second step the calculations go to very bad "solutions".

I used the same approach and basis sets to compute the same hybrid calculation over different systems without problems, as: metallic aluminium, organic molecules, semiconductors like Cu2O, slabs of semiconductors, and so on.

I do not know why the second step goes to a weird energy... I tried HSE, HF, and other functional... all have the same problem. Only GGA-DFT behaves well.

For hybrid calculations I used the GGA converged wfc, but the result is similar if I start it using atomic density...

I adjunct the input, output and coordinate file. I use a huge vacuum space because I am trying to insert graphene into an interface, and I use the same cell. But I tried using lower vacuum width, and changing the periodicity and psolver, but nothing works...

Some advice? Graphene is a problematic system for hybrid calculations as diamond?

Regards - Lucas Lodeiro

[fabia...@gmail.com]

Hi Lucas,

When I reduced the cutoff to 4.0 and eps_schwarz to 1e-8 it converged nicely. With a larger cutoff you probably have to decrease eps_schwarz even more. You should also check the impact of the aux basis, cFIT3 might not be the best choice.

Cheers,

Fabian

[Lucas Lodeiro]

Hi Fabian,

I tried using a reduced cutoff of 4.0, and it works well with eps_schwarz 1e-6. The problem appears when the cutoff is higher than 6.0, in that case, I reduced the  eps_schwarz until 1e-12, but the problem was the same, no differences. I try reducing the eps_pgf_orb from 1e-6 to 1-e-8 and it work without problem, and value of eps_schwarz did not change the result (a difference of 1e-5 in energy or less). Thanks for your advice, it impulses me to get insight of the eps keywords.

About the auxiliary basis set, I agree, cFIT3 is the smallest one and is not the best choice, but I have to deal with very big systems where the RAM is not enough, and I need to speed up the calculation using small basis sets, as the DZVP-SR-MOLOPT... If tis calculations converges in a day or a couple I will test with bigger basis sets.

Regards

Lucas Lodeiro

--
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.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/4efdfda1-3c59-48ce-bff1-c096a667143en%40googlegroups.com.