[Wien] regarding almblm problem

Peter Blaha pblaha at theochem.tuwien.ac.at
Wed Nov 5 09:57:24 CET 2014


It can't be that difficult to "guess" the meaning of these many numbers.

Run this analysis in serial with only few (one) k-points first.

I don't know about the aix-problem, so stick to the Linux computer.

x lapw2 -alm    will produce a case.almblm file for all occupied bands.
x lapw2 -all e1 e2   should do the same for bands within e1-e2. (the 
-all flag has an influence on which bands (eigenvalues) are listed, but 
not on the atoms)

The file may start out like:
   K-POINT:  0.0833333333  0.0833333333  0.2500000000  2734 122           1
            1           1          12  jatom,nemin,nemax
            1   ATOM
            1  0.111111111111111       NUM, weight
    0   0   1 -0.89065124E+00 -0.33802776E-01    0.18346099E+00 
0.69628664E-02  ...

So it gives the first k-point, then the FIRST atom, then the FIRST band, 
and then lists for l=0,m=0 the ALM, BLM,CLM (and LO-coefficients)
Then comes the next lm pair (1,-1):
    1  -1   2 -0.28642028E-05 -0.42607702E-07  ....
and so on.

It does this for all bands (NUM) and then for the other k-points, but 
when all is done, it repeats the sequence for jatom=2, ... (ignore the 
misleading ATOM label, the important number is in the line above ATOM 
and labeled jatom)

  K-POINT:  0.0833333333  0.0833333333  0.2500000000  2734 122           1
            2           1          12  jatom,nemin,nemax
            1   ATOM
            1  0.111111111111111       NUM, weight
    0   0   1  0.31776969E-01  0.60091844E-01  ....

(So in my case, this is an antibonding s-character.

Run also   x lapw2 -qtl, where case.output2 will contain the 
corresponding partial charges (the squares of the Alm,Blm,...). Of 
course, there you loose the sign, but still should be able to identify 
the corresponding quantitites.


On 11/04/2014 11:54 PM, Oliver Albertini wrote:
>     It seems that the problem I am encountering with 'x lapw2 -alm' did not occur
> when running on a linux cluster using version 12.1 of W2k. I ran TiC in
> serial to test this.
>    On the aix cluster, I believe there must be something wrong: I am even getting binary
> junk at the top of the  case.almblmup/dn_* files (in the line that begins
> with 'K-POINT', in place of the last field, I see binary output).
>     Also, the case.almblmup/dn files themselves are empty, I only have the numbered files 'k-point job'
> (case.almblm_1, etc.). I am guessing this is not another symptom of the
> problem?
> Most importantly, not all atoms are represented (when not using '-all
> e1 e2' flag, I expect to see all the atoms) and the atoms I am expecting to see
> in that energy window are not present when using said flag.
>
> Can you please guide me towards the relevant code, i.e. would this type
> of problem occur in the script lapw2para? Or perhaps SRC_lapw2/abc.f? For what
> it's worth, I have not had weird lapw2 problems before this.
>
> Thank you,
>
> Oliver Albertini
>

-- 

                                       P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-165300             FAX: +43-1-58801-165982
Email: blaha at theochem.tuwien.ac.at    WIEN2k: http://www.wien2k.at
WWW:   http://www.imc.tuwien.ac.at/staff/tc_group_e.php
--------------------------------------------------------------------------


More information about the Wien mailing list