[CP2K-user] [CP2K:18941] Re: k-points vs. supercell approach

[Krack Matthias]

Dear Léon

The timings appear reasonable to me. The run time for a small unit cell with just 12 atoms using a moderate number of k points can be much faster than a Gamma point calculation for a system with 600 atoms. The timings might get a bit closer when a wavefunction restart file is available.

Best

Matthias

From: cp2k at googlegroups.com <cp2k at googlegroups.com> on behalf of Léon Luntadila Lufungula <Leon.luntadilalufungula at uantwerpen.be>
Date: Tuesday, 20 June 2023 at 15:09
To: cp2k <cp2k at googlegroups.com>
Subject: [CP2K:18941] Re: k-points vs. supercell approach
Dear all,

I am currently continuing with k-point sampling as this method is much faster than the supercell calculations. However, I still don't quite understand why/how this is the case... On most of the posts I read on this forum, people state that the supercell approach should preferably be used as this is the standard in CP2K, so I don't understand why there is such a large difference in favor of k-point sampling in my case. Could anyone confirm/deny that this could be normal behaviour? I still would prefer to use the supercell approach as it has more functionalities, but I can't consider it a valid option if the computational cost is so much larger than the k-points approach...

Kind regards,
Léon
On Monday, 19 June 2023 at 14:18:48 UTC+2 Léon Luntadila Lufungula wrote:
Dear CP2K community,

I'm quite new to CP2K and I am trying to build my input files step-by-step to eventually do calculations on modified surfaces. Currently I'm still doing single point calculations on my bulk structure (anatase TiO2) to optimize the calculation parameters. I have already converged the values of the CUTOFF and REL_CUTOFF parameters and was now looking at comparing k-point sampling to gamma point calculations within the supercell approach.

I was initially thinking about using the supercell approach as some things are not yet implemented for k-points, such as the OT method for example, but I noticed that supercell calculations are considerably slower than k-point calculations. I have attached the input and output files for both approaches for the largest cell/k-mesh I have used, which is 5x5x2. The k-point calculation only took 14.342s, whereas the supercell calculation took a whopping 649.442s...

So my question is if I should just continue with k-points sampling or if there is something wrong with my inputs which results in the large difference in calculation time? The k-point calculation also throws a warning "*** WARNING in cryssym.F:165 :: Symmetry library SPGLIB not available ***", but I also don't know if this may be the reason for the discrepency or if I may ignore this warning.

Any help would be greatly appreciated!

Kind regards,
Léon
--
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+unsubscribe at googlegroups.com<mailto:cp2k+unsubscribe at googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/3e2ebd58-9a77-45b2-8d4a-204fd049bae3n%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/3e2ebd58-9a77-45b2-8d4a-204fd049bae3n%40googlegroups.com?utm_medium=email&utm_source=footer>.

-- 
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+unsubscribe at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/ZRAP278MB082769A4309F4FABFB382D3EF45CA%40ZRAP278MB0827.CHEP278.PROD.OUTLOOK.COM.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20230620/b9452614/attachment-0001.htm>