[CP2K-user] Constrained DFT (CDFT) Errors

[Nico Holmberg]

Hi Brian, 

I'm not 100 % sure but I think that "EXIT CODE: 9" means that your 
calculation ran out of memory. Try increasing the number of nodes or 
initially for testing purposes decrease the amount of memory required by 
using: smaller cutoff and/or smaller cell (your should be able to get away 
with a smaller cell with wavelet, but do test this).

By the way, I noticed you're using Diagonalization instead of OT as the 
solver. If at all possible for your system (converges, produces reasonable 
results), switch to OT because in my experience CDFT converges much more 
nicely with OT than diagonalization.

To your questions, 

1) Do you want the charge to be positive or negative? If in the neutral 
case the CDFT constraint group has X electrons, then X+1 electrons will 
give it a negative -1 charge, and X-1 electrons will give it a positive +1 
charge. The total number of electrons must be conserved in both states if 
you subsequently plan on calculating the electronic coupling. You can also 
calculate the energy difference between CDFT states with a different total 
net charge as long as you take into account the changes in charge properly 
(but not any properties called with MIXED_CDFT section). 
2) I guess you mean the "TARGET" keyword not "STRENGTH". Use the number of 
valence electrons as determined by the pseudo and basis set if you're doing 
GPW calculations (not all-electron GAPW). So a neutral Cu atom would have 
11 valence electrons (target 11). 


BR, 

Nico



maanantai 22. lokakuuta 2018 17.44.59 UTC+3 Brian Day kirjoitti:
>
> Hi everyone,
>
> I am currently trying to run a CDFT calculation to determine the 
> electronic coupling integral of a charged system in CP2K. I have 
> successfully run the Zn dimer tutorial, but am running into issues when I 
> switch to my system; specifically, I get the following error: 
>
>
> ===================================================================================
> =   BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
> =   PID 21797 RUNNING AT opa-n68
> =   EXIT CODE: 9
> =   CLEANING UP REMAINING PROCESSES
> =   YOU CAN IGNORE THE BELOW CLEANUP MESSAGES
>
> ===================================================================================
>    Intel(R) MPI Library troubleshooting guide:
>       https://software.intel.com/node/561764
>
> ===================================================================================
>
> I have attached all files used to submit and run this job, including the 
> two outputs (Ni3HITP2-Mcent-cdft.out = cluster outputs, 
> Ni3HITP2-Mcent-state1-run.inp = cp2k outputs). 
>
> Note about the system: I am studying is a bonded system, and I am 
> localizing charge onto a certain part of the molecule, it is not two 
> distinct molecules (technically, this is a periodic system, but I am using 
> an H-terminated unit to approximate the framework). 
>
> Additional questions in case anyone happens to know the answers:
> 1. If the net charge on the a region of the neutral framework is non-zero, 
> should the charged state contain one more electron than the neutral-state 
> charge in that region, or simply one extra valence electron total? 
> (Although I do plan to test this once I get the code running)
> 2. If I am using an atom like copper, and a basis set which only contains 
> 11 valence electrons, should the 'strength' term in CDFT reflect the number 
> of valence electrons in the basis set, or the number of electrons total 
> (i.e. 11 vs 29)?
>
> Thanks!
>
> -Brian
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20181022/32a7d738/attachment.htm>