[CP2K-user] [CP2K:22386] Re: NEB RESTART.kp files not detected

[tkuehne]

Dear Alex,

the restart files are valid. The warning is caused by the explicitly 
specified
WFN_RESTART_FILE_NAME.

During a BAND calculation, CP2K assigns a separate project name to every
replica. For example, with

  PROJECT projectname

the replica project names are projectname-BAND1, projectname-BAND2, and so 
on.
This is why CP2K writes

  projectname-BAND1-RESTART.kp
  projectname-BAND2-RESTART.kp
  projectname-BAND3-RESTART.kp
  ...

An explicitly specified

  WFN_RESTART_FILE_NAME projectname-RESTART.kp

is interpreted literally and is shared by all replicas. CP2K does not insert
the -BAND<N> part into an explicitly supplied filename. Since that literal
file does not exist in your case, CP2K falls back to an atomic guess.

To restart from the individual replica wavefunctions, keep the original
GLOBAL/PROJECT name, make sure that the BAND restart files are in the 
working
directory, and omit WFN_RESTART_FILE_NAME:

&FORCE_EVAL
  &DFT
    &SCF
      SCF_GUESS RESTART
    &END SCF
    ...
  &END DFT
  ...
&END FORCE_EVAL

CP2K will then select the appropriate file automatically for each replica:

  projectname-BAND<N>-RESTART.kp

For a Gamma-point calculation, the corresponding files have the .wfn
extension instead.

If you also want to continue the band coordinates and optimizer state, load
the global restart file separately, for example:

&EXT_RESTART
  RESTART_FILE_NAME projectname-1.restart
&END EXT_RESTART

These are two independent parts of the restart:

  EXT_RESTART        restores the band coordinates and optimizer state
  SCF_GUESS RESTART  restores the electronic wavefunction of each replica

I reproduced both cases with a CP2K 2026.2 development build: the explicit
base filename produces the warning you reported, whereas omitting
WFN_RESTART_FILE_NAME correctly uses the per-replica .kp files.

This behavior is now documented in:
https://github.com/cp2k/cp2k/pull/5710

Best regards,
Thomas
On Monday, June 8, 2026 at 5:32:33 PM UTC+2 Alex Espinoza wrote:

> Hello, 
>
> this is an old problem where I'm doing a NEB calculation for a V2O5 slab. 
> I'll get RESTART.kp files, but the program can't seem to find them. Or 
> maybe I'm inputting it wrong in the .inp file. Does anyone know the proper 
> format to consider RESTART.kp files into these band calculations? It's 
> frustrating always starting from scratch and using ATOMIC GUESS. 
>
> The error is shown below, cp2k version 2024.1:
>
> WARNING in qs_initial_guess.F:306 :: User requested to restart the    ***
>
>  *** wavefunction from the file named: projectname-RESTART.kp. This file 
> ***
>
>  *** does not exist. Please check the existence of the file or change      
> ***
>
>  *** properly the value of the keyword WFN_RESTART_FILE_NAME. Calculation  
> ***
>
>  *** continues using ATOMIC GUESS.
>

-- 
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/95e46f8c-8f16-49f6-bdbc-6a038b902c51n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20260811/af00fcf2/attachment.htm>