Boron

Dear Forum Members,

some more data points:

- tested exciting versions 0.9.057, 0.9.074, 0.9.093, 0.9.141
- only libraries supplied with the tgz from sourceforge, e.g. no MKL or fftw
- intel ifort 10.1.008, flags "-O0 -fltconsistency" for F90 and F77
- input as in original post, species file from exciting 0.9.074

Result:

- insulating state with density of states at the fermi level about
  0.2086886557E-01 for 0.9.057 and 0.9.074
- metallic state with density of states at the fermi level about
  13.19284707 for 0.9.094 and newer

Same for difference in interstitial charges, muffin-tin populations etc., also warnings
of the type given in the original post starting with 0.9.093.

So the Boron example (and my own input) break between 0.9.074 and 0.9.093,
and I am unable to reproduce the INFO.OUT from the web page anyway (has different
number of lo's ?).

Does this ring a bell ?

Best Regards

Christof

Dear Forum Members,

here is just another data point. By setting nosym=.true. in 0.9.093 the
problem goes away. However, switching symmetrie off in 0.9.150 has no effect,
e.g. the electron count is changing as described above.

For me this looks likely to be a bug with the symmetrie routines,
especially considering the changelog entry between 0.9.074 and 0.9.093.

Best Regards

Christof

Dear Kay,

thank you very much for fixing this ! Apart from the issue below it seems
to work fine now.

However, I still have a question. I was trying to do a quick and dirty (e.g.
don't try to make too much sense of it) calculation of the boron atom
(input below) and still get a single warning for the initial electron number:

Warning(charge): total charge density incorrect for s.c. loop     0
Calculated :    4.991653004   
Required   :    5.000000000

This does not repeat in later scf cycles. On the other hand there is no
warning at all with 0.9.074.

The same holds for my solid input: a single warning of the above type in
cycle 0 but no more visible problems after that and not even a single warning
with 0.9.074.

This still looks strange to me, because I would expect that especially the
atomic starting density should be quite good in an APW type scheme ?

Best Regards

Christof

tasks
0

sppath
'./'

avec
   14.956348800       0.000000000       0.000000000   
   6.900744994      -10.109542068       0.000000000   
   0.000000000       7.903713658      -10.608679221   

atoms
   1                                    : nspecies
'B.in'                                 : spfname
  1                                    : natoms; atpos, bfcmt below
    0.0 0.0 0.0    0.00000000  0.00000000  0.00000000

ngridk
1 1 1

rgkmax
8.0

xctype
20

nwrite
2

epspot
1e-7

nosym
.true.

Dear Kay,

the warning goes way when setting lradstp to 2 from the default 4, so the
electrons are lost somewhere in the interpolation from the atomic grid.

Still, this is something I am unsure about. I will try to find out in which
version this warning occurs first later.

Best Regards

Christof

Hello everybody.
To finish this thread: the occurence of this warning is
the result of a change to rhoinit.f90 between 0.9.093 and
0.9.114. In versions prior to the change the charge
difference during initialization is slightly below
the warning threshhold, in later versions slightly above.

There is some influence on box shape and size as well
as on lradstp and the contents of the species file
(number of points etc.). The warning is also visible for other
atom types, i.e. C and H.

Best Regards

Christof