[DFTB-Plus-User] Question regarding Brilloin-zone sampling

[Bálint Aradi]

Dear Thomas,

actually DFTB+ does not really knows, where the k-points come from. If 
you specify the SupercellFolding method, DFTB+ calls the k-point 
generator already in the parser, and passes only the generated k-points 
further, not the method, how those were generated. At the point, when 
the information about the k-points are printed, DFTB+ does not have any 
information, how those k-points were generated.

So, therefore, it is really-really unlikely, that the two calculations 
differ. The only thing I could imagine, that your MD is so chaotic, that 
minimal numerical inaccuries result in diverging paths. But even that 
seems to be very-very unlikely.

Is it possible to send me (off list) the input, where you have observed 
this behaviour? I could try to run it myself (if it is not too big or 
does not take too long) and check, whether I observe a similar 
behaviour. Alternatively, could you check, whether you get the same 
behaviour, if you repeat the calculation with the same settings?

Best regards,

Bálint



On 02.02.22 18:48, Hofer, Thomas wrote:
> Dear colleagues,
> 
> 
> I have got one curious question for which I could not find the answer in 
> the manual and the DFTB+ recipes.
> 
> 
> Currently we are testing DFTB+ for the treatment of a class of porous 
> solid-state systems, thereby varying the settings for the Brillouin-zone 
> sampling down until pure gamma point sampling. The weird behaviour is 
> observed for the settings with kmax=1 vs gamma-point sampling. For our 
> cubic systems we use the following setups:
> 
> 
> KPointsAndWeights = SupercellFolding {
>      1 0 0
>      0 1 0
>      0 0 1
>      0.0 0.0 0.0
> }
> 
> 
> and
> 
> 
> KPointsAndWeights = {
>    0.0 0.0 0.0 1.0
> 
> }
> 
> In both cases the output of dftb shows the following entry:
> 
> 
> [...]
> 
> K-points and weights:           1:  0.000000  0.000000  0.000000 
>      1.000000
> 
> K-points in absolute space:     1:  0.000000  0.000000  0.000000
> 
> [...]
> 
> 
> implying that Brillouin-zone sampling is identical in both cases.
> 
> 
> When executing a single point energy + force calculation of the same 
> input structure indeed the very same energy, forces and partial charges 
> are obtained, supporting this conclusion (possibly the stress tensor is 
> different, but I did not look into this).
> 
> 
> When executing a molecular dynamics simulation (with our in-house MD 
> code interfaced to dftb+) using 100% identical settings aside from 
> "KpointsAndWeights", the simulations initially show identical 
> progression as expected.
> 
> 
> However, after approx. 3000 MD steps (6 ps) the simulations show 
> divergence, although the initial assessment indicated that both levels 
> should yield identical results. In case of gamma-point sampling the 
> porous structure of the system collapses, whereas SupercellFolding at 
> kmax=1 provides good results compared to the experimental reference.
> 
> 
> Since the only difference in these simulations are the forces obtained 
> from dftb+ (only the forces enter the equations of motion), I am in need 
> for a explanation to describe the observed differences for an upcoming 
> manuscript. Is it possible, that the stress tensor calculated by dftb+, 
> which is used to determine the pressure in MD, is calculated differently 
> by these two settings? This would provide an explanation for the 
> observed differences.
> 
> 
> Even more curiously, we did the very same investigation for a different 
> system of the same material class. In this case kmax=1 and gamma-point 
> sampling both yielded effectively the same results down to the overall 
> computing time of the MD runs (as expected).
> 
> 
> Any suggestion or pointer in this matter is highly appreciated.
> 
> Many thanks and best wishes,
> Thomas
> 
> 
> /---------------------------------------------------------------------------------------------------------/
> assoc. Prof. Dr. Thomas Hofer
> Department for Theoretical Chemistry
> Institute of General, Inorganic and Theoretical Chemistry
> Innrain 80-82
> University of Innsbruck
> 6020 Innsbruck
> Austria
> Phone: +43-512-507-57111
> Email: t.hofer at uibk.ac.at
> /---------------------------------------------------------------------------------------------------------/
> 
> 
> 
> 
> _______________________________________________
> DFTB-Plus-User mailing list
> DFTB-Plus-User at mailman.zfn.uni-bremen.de
> https://mailman.zfn.uni-bremen.de/cgi-bin/mailman/listinfo/dftb-plus-user


-- 
Dr. Bálint Aradi
Bremen Center for Computational Materials Science, University of Bremen
http://www.bccms.uni-bremen.de/cms/people/b-aradi/

-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature
Type: application/pgp-signature
Size: 840 bytes
Desc: OpenPGP digital signature
URL: <http://mailman.zfn.uni-bremen.de/pipermail/dftb-plus-user/attachments/20220208/e328a939/attachment.sig>