[CP2K-user] [CP2K:21296] Re: out-of-memory issue in hybrid DFT TDDFT calculation in CP2K 2025.01
Augustin Bussy
augustin.bussy at gmail.com
Tue Mar 18 09:08:37 UTC 2025
Hello,
To the best of my knowledge, TDDFT is not implemented with k-point sampling
in CP2K. That being said, your calculation fails at the SCF stage, in the
RI-HFXk calculation.
The RI-HFXk method for k-point is optimized for small unit cells, and does
not scale well with unit cell size. 200 atoms is out of range for this
method.
To get the equivalent of k-point HF with large systems, you should use
supercell calculations. This will end up using less computational resources
since the standard HFX
implementation of CP2K scales linearly with system size.
Best,
Augustin
On Tuesday, 18 March 2025 at 09:36:34 UTC+1 Guozhen Zhang wrote:
> Dear CP2K developers and users,
>
> In an earlier test of hybrid DFT TDDFT calculation using a
> workable CP2K 2025.01on a different server, I encountered a memory issue
> that prevent me from running the job.
> In the test, I ran PBE single point calculation to obtain the
> wavefunction file, then read this file to start a PBE0 TDDFT single point
> calculation. The system of interest is an pure organic molecular cyrstal
> with 200 atoms in the unit cell. I used a 3*4*1 MP mesh for the k-points
> sampling in the first Brillouin zone. However, in the TDDFT calculation,
> the program has some memory overflow (as told by the administrator staff of
> the server) that kills the job.
> Has anyone seen similar issue? Any clue to fix it?
>
> Please find the input and output in the attachment.
>
> Thanks!
>
> Guozhen
>
>
>
>
>
--
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 visit https://groups.google.com/d/msgid/cp2k/d549a2f6-3129-4b52-a3a9-651443c6a4cfn%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20250318/680dd808/attachment.htm>
More information about the CP2K-user
mailing list