Convergence issue

[ma455...@gmail.com]

Hi all,

I'm new in using CP2K.

I'm currently calculating pure silicon with one vacancy, but the SCF calcualtion is very hard to converge. Could you please provide me some suggestions on improving this issue?

Best,

Hongyang

[Martin Konôpka]

Dear Hongyang,

One reason of convergence difficulties may be the geometry that is far from a relaxed one. In such cases I try to set a very loose SCF criterion (in the initial stage of the relaxation). After some number of geometry steps, I tighten the SCF criterion. 

Best regards

Martin.

pi 3. 12. 2021 o 1:25 ma455...@gmail.com <ma455...@gmail.com> napísal(a):

--
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/a2d9af27-175a-48d6-aaec-b63d708992a5n%40googlegroups.com.

[ma455...@gmail.com]

Dear Martin,

Thanks for your suggestion.

But what I want to do here actually is just do a single point non-sc calculation. Because this geometry is extracted from P doped system and I want to calculate the charge density difference so I need to obtain the charge density cube files seperately for both Si and P extracted from P doped Si.

Best,

Hongyang

[Martin Konôpka]

Dear Hongyang,

O yes. I such cases, if nothing else helped, I applied an extremely high electronic smearing, for example about 5000 K. Once the electronic structure was converged at such a temperature, I restarted from the converged wavefunctions using a lower temperature (smearing). Not too much lower. Then repeat it again setting even a lower temperature. So it is a manually driven simulated annealing - a rather time-consuming procedure requiring several steps. Usually it was not possible to scale down to temperatures close to zero but for the purpose of just obtaining a charge-density difference, this may be an acceptable way.

Best regards

Martin.

so 4. 12. 2021 o 4:03 ma455...@gmail.com <ma455...@gmail.com> napísal(a):

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/8e5b5892-6a0d-45b9-94ce-c5dd287676ean%40googlegroups.com.

[ma455...@gmail.com]

Dear Martin,

Thank you very much for your help! I will give it a try on this approach.

Best,

Hongyang

[ma455...@gmail.com]

Dear Martin,

This method works! I used smearing 300, 100, 50, 10, 5, and 2. Although it still doesn't converge after removing smearing, as you said it is good enough for plotting the charge density difference map. Thanks!

Best,

Hongyang

在2021年12月4日星期六 UTC+11 18:10:14<konop...@gmail.com> 写道：

[Martin Konôpka]

Dear Hongyang,

I have just one thing to add: In principle, if you use a smearing to obtain the charge density of a subsystem, then you should use the same samearing to obtain the additional two (or perhaps more) charge densities (used to calculate the difference); otherwise the density difference might not be meaningful. However, if the smearing is 2 Kelvin only (a really very low value, almost a limitingly low one), this is for your purpose like a zero-smearing calculation (most likely).

Best

Martin.

ut 7. 12. 2021 o 1:40 ma455...@gmail.com <ma455...@gmail.com> napísal(a):

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/b235ab7f-5a34-4bab-b9f5-e5abe43d8fden%40googlegroups.com.

[ma455...@gmail.com]

Dear Martin,

Got it. Thanks for your information!

Best,

Hongyang