[CP2K:536] Re: Problem when adding LOCALIZE section

[Teodoro Laino]

Ciao Marcel,

Thanks for having triggered the commit. This kind of help is highly  
appreciated. I post this final comment on the group so that
the thread continues to be documented and in case people more into  
the LOCALIZATION part than me can contribute to the
solution of the problem.

Regarding this problem, the point is that if you run that input file  
with version before the -D2007-12-20T13:55:00 you will not get
any localization since that commit is changing the way the  
localization works.
Let me explain better. After that commit if you want a localization  
you can just specify (as you did):

&LOCALIZE
...
...

Before that commit this section alone has no effects, i.e. you don't  
get any localization (as you can see in your output file).
If you want a localization you've to specify for older versions  
(older than -D2007-12-20T13:55:00) also the keyword:
QS%LOCALIZE_WFN

If you do so the two version give the same number.

Now the question is why the two input files (with and without  
LOCALIZATION) give different numbers (when they should
possibly give the same numbers)?

This is a summary of the numbers I get:

teo at XXXXXXX:/data03/teo/tmp/cp2k/tests/QS> grep 'POTENTIAL E' out.old

WITHOUT LOCALIZATION

  INITIAL POTENTIAL ENERGY[hartree]     =                      
-0.171648104463E+02
  POTENTIAL ENERGY[hartree]    =         -0.171647032051E+02   
-0.171647032051E+02
  POTENTIAL ENERGY[hartree]    =         -0.171648070232E+02   
-0.171647551142E+02
  POTENTIAL ENERGY[hartree]    =         -0.171650339365E+02   
-0.171648480549E+02
  POTENTIAL ENERGY[hartree]    =         -0.171652387783E+02   
-0.171649457358E+02
  POTENTIAL ENERGY[hartree]    =         -0.171652753495E+02   
-0.171650116585E+02
teo at XXXXXXX:/data03/teo/tmp/cp2k/tests/QS> grep 'POTENTIAL E' out.new

WITH LOCALIZATION

  INITIAL POTENTIAL ENERGY[hartree]     =                      
-0.171648104463E+02
  POTENTIAL ENERGY[hartree]    =         -0.171650398555E+02   
-0.171650398555E+02
  POTENTIAL ENERGY[hartree]    =         -0.171654500287E+02   
-0.171652449421E+02
  POTENTIAL ENERGY[hartree]    =         -0.171637618288E+02   
-0.171647505710E+02
  POTENTIAL ENERGY[hartree]    =         -0.171590921092E+02   
-0.171633359555E+02
  POTENTIAL ENERGY[hartree]    =         -0.171537133877E+02   
-0.171614114420E+02

Are these the same numbers you were comparing? The conserved quantity  
is reflecting the same drift..

I guess that there's an additional bug in the LOCALIZATION (that was  
always there, unless you don't have an
example that is really doing localization with an older version in  
which you don't see this drift) possibly associated
with dirtying/overwriting structures used in the next SCF cycles or  
God knows what they do in that module...

Were you really  able to do localization+MD with this input file and  
an older version without observing this drift?

I noticed that the &LOCALIZE section into the QS section has no  
regtests at all (of course regtests of that section would be
highly appreciated, but this comment is for the couple of developers  
working on the localization.. long story..) and more
important this would not print the wannier centers..

Anyway If you want just print the wannier centers during an MD the  
way to go is:
&DFT
...
     &PRINT
        &LOCALIZATION
           IONS+CENTERS
           FILENAME =MYFILE
           ADD_LAST NUMERIC
        &END
     &END
&END

Hope this helps Marcel!

Ciao,
Teo


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20080109/91b92ee2/attachment.htm>