[Wien] Band character plotting and Linearization energy

Guo-ping Zhang gpzhang at nano.indstate.edu
Fri Dec 5 04:04:52 CET 2008


Dear Peter,

Thanks a lot for your quick response. I tested your suggestions as
below, but could not solve the problem of character plotting. In a
separate mail, I will ask about the linearization energy.


Here is what I did
(1)
x lapw1  -band -up -p
x lapw1  -band -dn -p
x lapwso -band -up -p


(2)
#copy both case.vectorsodn_XX and case.vectorsoup_XX to $SCRATCH of the
master node

x lapw2  -band  -so -qtl -p -up
x lapw2  -band  -so -qtl -p -dn

(x lapw2 -band -so -qtl -p failed)

x spaghetti -so -p

 1.1   2    4                      # line width, line switch, color switc
       ^________ no difference if 0,1,2,3.

There is no character plotting.

I look at file  case.bands.agr and find out  the file has
 # bandindex:  7
.......
   0.00000  -9.16053   0.07000
   0.02361  -9.15234   0.07000
   0.04723  -9.12780   0.07000
   0.07084  -9.08695   0.07000
........

(I also looked into the source code spag.f Line 865 and for some
reason charr did not get a value larger than 0.01. Since the output is
seven times of this number, which gives 0.07 in case.agr)

Since all the last columns are 0.07000, the symbol size is basically
zero in xmgrace.

However, if I use x spaghetti -so -p -up|dn then the character
plotting does work, but now the up and down band structures are
different, which is incorrent in a case with SOC. I doubt whether my
step 2 is correct (x lapw2  -band  -so -qtl -p -up)

I have tested character plotting on a single processor on a different
machine, but results are the same. I can verify that both WIEN2k_08
and _06 lead to the same problem.

My question is whether you could send me your case.band.agr for fccni
example with SOC (with 3000 k points), your :log file and your
case.insp. I can see what I did wrong. In the meantime, if you need
any additional information, please let me know.

Thank you so much!

Best wishes,

Guoping







On Thu, 4 Dec 2008, Peter Blaha wrote:

>
> > (1) Is it true that band character plotting is still not possible for
> > spin-polarized with SOC?
> >
> > I succeeded for the one without SOC, but not with
> > SOC. Note the band structure plotting is fine for either case, just the
> > characterer plotting is not working.
> >
> >  In my case.insp file,  I have
> > 1.1   0    4                      # line width, line switch, color switch
>
> A line-switch of 0 will NEVER do "character plotting". You need 2 or 3
>    2...lines and open circle
>    3...lines and filled circles
>
> And yes, I think character plotting works also for spinpolarized + SO cases.
>
> > (2) Linearization energy selection, and a failure  similar to
> > Stefaan's
> >
> > I followed Stefaan's excellent notes closely and tested it with C60 solid.
> >
> > -------------------First question--------------------
> >
> >  The default global linearization energy parameter in case.in1 is
> > very different from the real DOS, but the job runs fine and converges as
> > well.
> >
> > This is the default case.in1
> >
> > WFFIL        (WFPRI, SUPWF)
> >   7.00       10    4 (R-MT*K-MAX; MAX L IN WF, V-NMT
> >   0.30    3  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global APW/LAPW)
> >  0  -0.77605   0.01000CONT 1
> >  0   0.30000   0.00000CONT 1
> >  1   0.30000   0.00000CONT 1
> > ...
> > K-VECTORS FROM UNIT:4   -9.0       3.0   140   red   emin/emax/nband
> >
> > When I checked the DOS, I found the main DOS before the Fermi level is
> > around -0.3, but the default global energy is 0.30 (see above).
> >
> > Is this normal?
>
> It is not "normal", but it can happen for open structures, surfaces,...
>
> > I understand the sp states spread quite a big region. (-in1new was also
> > tested, but no too difference). I then changed
> > select.f's upper bound (in Lapw1) from 0.5 to 4.5 to cover those highly energy
> > states to get a  better linearization energy. This does not change the
> > default case.in1
>
> Why should select.f in lapw1 change the input file of lapw1 ???
> It may have an effect on the actual energies used (see the scf1 file!!!)
>
> If you use -in1new, this should give a very different case.in1 file!
>
> For sp states, often the energy parameters are quire insensitive, so you may not
> see any difference in the DOS.
>
>
> > After a first run, I plot total and partial dos and create a new case.in1
> > WFFIL        (WFPRI, SUPWF)
> >   7.00       10    4 (R-MT*K-MAX; MAX L IN WF, V-NMT
> >  -0.31    4  0      (GLOBAL E-PARAMETER WITH n OTHER CHOICES, global APW/LAPW)
> >  0  -0.36000   0.00500CONT 1 <----- 2 S state's center below Ef
> >  0   1.40000   0.00000CONT 1 <----- center of 2S* states I need
> >  1  -0.32000   0.00500CONT 1 <----- 2 P state's center below Ef
> >  1   1.50000   0.00000CONT 1 <----- center of empty 2P* states I need
> > ....
> > K-VECTORS FROM UNIT:4   -9.0       3.0   140   red   emin/emax/nband
> >
> > When I run my job with this input file, I found WIEN2K_06 version crashed
> > at lapw1 (due to Seclr4.F), and then I dumped overlap matrix. I found that
> > clearly some of basis functions appear to be linearly dependent. On a
> > different cluster (larger memory and disk, the matrix can be very large),
> > I ran WIEK2k_08 and found it is working fine but there are some states
> > below -9 for quite a few k points. These states must be wrong since free
> > carbon atom's 1s energy is around -9, but this error is very similar to
> > Stefaan's finding.
>
> First of all again: check the actual energies used in lapw1, which are printed
> in case.scf1. As soon as you "search" (the 0.005 just before the CONT) you may
> actually use quite different energies.
>
> Second: for small spheres (you said C60, and this must have very small spheres)
> the radial wavefunctions (inside the sphere) at expansion energies of -0.3 and 1.3
> might still be very similar leading to linear dependency. Whether the LAPACK detects
> linear dependencies and stops or it continues depends probably a lot on the BLAS/lapack
> software and small numerical details. But "eigenvalues below ..." always hints to
> linear dependency problems.
>
> Are you sure the center of 2s and 2p energies are that close together? (-0.36/-0.32)
> What does the scf2 file or case.in1 after in1new say ?
>
> --
>
>                                        P.Blaha
> --------------------------------------------------------------------------
> Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
> Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
> Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
> --------------------------------------------------------------------------
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>


More information about the Wien mailing list