[CP2K-user] [CP2K:11082] comparing CP2K-DFTB (5.1) and DFTB+ (18.2)

[Wei Lai]

Dear Juerg,

Thanks for the effort in looking into the problems.  In addition, I found 
that the real-space part of 1/R contribution to gmcharge was done with a 
neighbor list, while the gamma_matrix contribution to gmcharge was not.  In 
my opinion they should be treated in the same way (perhaps both a neighbor 
list).  I'd like to bring this up to your attention.  I look forward to the 
fixes!

Thanks, Wei

On Friday, January 11, 2019 at 3:37:36 AM UTC-5, jgh wrote:
>
> Hi 
>
> I looked into these problems and I see that this behavior is 
> from a combination of bugs and hidden constraints on variables. 
>
> The bugs were introduced when the MIC was dropped after the implementation 
> of the k-point code. The Ewald parameters are partly set from the 
> input and partly set by using default values within the code. 
>
> To be on the save side I would suggest to only do DFTB calculations 
> with large enough computational cells. You can use 
> the MULTIPLE_UNIT_CELL keyword if you have a small cell. 
>    &TOPOLOGY 
>       MULTIPLE_UNIT_CELL ${NREP} ${NREP} ${NREP} 
>     &END 
>     &CELL 
>       ABC 4.0 4.0 4.0 
>       MULTIPLE_UNIT_CELL ${NREP} ${NREP} ${NREP} 
>     &END CELL 
>
> I guess a cell with 10 Angstrom box length should be ok. 
>
> Use an ALPHA value that is 1 or larger and adjust GMAX for 
> your box size to get converged values. 
>
> I am working on fixes, they should be ready in the TRUNK version soon 
> and will be in the next release. 
>
> best regards 
>
> Juerg 
>
> -------------------------------------------------------------- 
> Juerg Hutter                         Phone : ++41 44 635 4491 
> Institut für Chemie C                FAX   : ++41 44 635 6838 
> Universität Zürich                   E-mail: hut... at chem.uzh.ch 
> <javascript:> 
> Winterthurerstrasse 190 
> CH-8057 Zürich, Switzerland 
> --------------------------------------------------------------- 
>
> -----cp... at googlegroups.com <javascript:> wrote: ----- 
> To: "cp2k" <cp... at googlegroups.com <javascript:>> 
> From: "Maxime Van den Bossche" 
> Sent by: cp... at googlegroups.com <javascript:> 
> Date: 12/22/2018 10:45PM 
> Subject: Re: [CP2K:11082] comparing CP2K-DFTB (5.1) and DFTB+ (18.2) 
>
> Hi all, 
>
> I also did a quick comparison now between CP2K and DFTB+ 
> for a simple rocksalt MgO structure (8 atoms in total, 
> using the OT minimizer in CP2K). The input and output 
> files for both codes are attached. 
>
> What I found so far: 
>
> * Without periodicity, the two codes agree quite well 
>   in terms of the H0 and SCC energies (up to ~1e-4 Ha) 
>   as well as the Mulliken charges. 
>
> * In the periodic case, so far I haven't managed to 
>   get comparable results with the two codes. The CP2K 
>   results indeed appear sensitive to the choice of the 
>   Ewald alpha parameter, similar to what Wei observed. 
>   For each of the three ALPHA values I tried, the resulting 
>   energies and SCC charges are quite different from DFTB+. 
>
> * In contrast to Wei, though, I also get different results 
>   for both codes in non-self-consistent calculations 
>   (but at least here the CP2K results indeed no longer 
>   depend on ALPHA). 
>
> It would be great if you could check whether my CP2K 
> inputs have been appropriate. If so, I hope this example will 
> help to get to the bottom of this. 
>
> Best, 
> Maxime   
>   -- 
>  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+... at googlegroups.com <javascript:>. 
>  To post to this group, send email to cp... at googlegroups.com <javascript:>. 
>
>  Visit this group at https://groups.google.com/group/cp2k. 
>  For more options, visit https://groups.google.com/d/optout. 
>   
>
> [attachment "files.zip" removed by Jürg Hutter/at/UZH] 
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20190111/c36dbd7e/attachment.htm>