[CP2K-user] [CP2K:14733] Calculations stopping after initial scf guess

Hayden Brandt bran... at clarkson.edu
Tue Feb 9 14:17:39 UTC 2021


Thank you for the suggestions. When I run in PRINT_LEVEL medium the 
calculation stops at the same point. I have also tried running the 
calculation without the restart file. I will definitely try your 
suggestions regarding those few input parameters. Thanks for pointing out 
my mistake on the MAX_SCF line. We were originally using a similar input 
file for a much larger geometry optimization (706 atoms), so that's why I 
had the poor energy convergence threshold. When testing the system I just 
used a much smaller version of the system we were testing earlier, and I 
changed a minimal number of parameters, so the parameters used probably 
aren't ideal for this smaller system. 

Hayden
On Tuesday, February 9, 2021 at 9:01:13 AM UTC-5 tkuehne wrote:

> Dear Hayden, 
>
> just a wild guess, but what WF RESTART are you reading? Was it created 
> using exactly the same basis 
> and not a smaller one, in which case you may end up in a stall since CP2K 
> tries to read more entries 
> from the RESTART file than is present. So you may try to simply delete the 
> RESTART file if present. 
> But a few more general comments: There is no point in setting EPS_DEFAULT 
> to 10^-14 when only 
> converging the „SCF“ to 10^-5. Using OT, where the mean preconditioned 
> gradient deviation is used as 
> the convergence criterion, 10^-5 is on the very loose side in particular 
> when doing total energy or even 
> geometry optimization. Moreover, using MAX_SCF 2500 the OUTER_LOOP is 
> essentially ineffective, 
> except you were prepared to do many thousands of OT cycles ;)
>
> Cheers, 
> Thomas
>
> Am 09.02.2021 um 14:51 schrieb Hayden Brandt <br... at clarkson.edu>:
>
> The input for the calculation is included below. Unfortunately I don't 
> think I can post the full output because the file is too large. I doubt 
> that this problem is a result of the system size, as the calculation 
> involves just 88 atoms. I am running it on 256 cores, and the same 
> calculation has finished in 63 minutes on a different cluster with the same 
> number of cores. Thank you for your help.
>
> On Tuesday, February 9, 2021 at 8:42:19 AM UTC-5 ry... at gmail.com wrote:
>
>> Dear user,
>>
>> it may well be, that your calculation is too slow. Perhaps, it's a very 
>> large system on the small computer. In any case, it's difficult to judge 
>> without a relevant input.
>>
>> Yours,
>>
>> Vladimir
>>
>> вторник, 9 февраля 2021 г. в 13:38:21 UTC+1, br... at clarkson.edu: 
>>
>>> We are running CP2K 7.1 in parallel with an Intel installation. This is 
>>> a new installation, and every calculation that we try stops before the 
>>> doing any scf iterations. The last lines printed to the output file are:
>>>
>>>  ----------------------------------- OT 
>>> ---------------------------------------
>>>   Minimizer      : CG                  : conjugate gradient
>>>   Preconditioner : FULL_KINETIC        : inversion of T + eS
>>>   Precond_solver : DEFAULT
>>>   Line search    : 2PNT                : 2 energies, one gradient
>>>   stepsize       :    0.05000000                  energy_gap     :    
>>> 0.00100000
>>>   eps_taylor     :   0.10000E-15                  max_taylor     :      
>>>        4
>>>   ----------------------------------- OT 
>>> ---------------------------------------
>>>
>>>   Step     Update method      Time    Convergence         Total energy  
>>>   Change
>>>   
>>> ------------------------------------------------------------------------------
>>>
>>> And then no SCF iterations are done, or at least they are not printed to 
>>> the output file. SLURM does not report any errors and the job does not end 
>>> even after it has stopped updating the output file. There isn't much more 
>>> information available when I try running in debug mode. Is it possible that 
>>> this is related to a bug in version 7.1?
>>>
>>
> -- 
> 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 cp... at googlegroups.com.
> To view this discussion on the web visit 
> https://groups.google.com/d/msgid/cp2k/c9a08f54-adc5-46fa-ae78-216c35f80aabn%40googlegroups.com 
> <https://groups.google.com/d/msgid/cp2k/c9a08f54-adc5-46fa-ae78-216c35f80aabn%40googlegroups.com?utm_medium=email&utm_source=footer>
> .
>
> <rsmm.inp>
>
>
>
>
> ==============================
> Thomas D. Kühne
> Dynamics of Condensed Matter
> Chair of Theoretical Chemistry
> University of Paderborn
> Warburger Str. 100
> D-33098 Paderborn
> Germany
> td... at mail.upb.de
> +49/(0)5251/60-5726 <+49%205251%20605726>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20210209/12465867/attachment.htm>


More information about the CP2K-user mailing list