[Pw_forum] charge-density.dat file missing after SCF calculations

Kiss, Ioan kissi at uni-mainz.de
Tue Jan 10 14:38:49 CET 2012


Dear PWSCF users and developers,

I would like to perform a series of geometry optimizations and band-structure
calculations for a slab (hexagonal surface slab model with 30 atoms in the unit cell)
and during my testing I have encountered a strange issue with espresso 4.3.2.
Namely, the SCF calculations and NSCF calculations end without any problem
and pw.x exits properly.
However, after the NSCF calculation is finished there is no charge-density.dat
file present in the prefix.save directory. All other files are there, including the
K000** directories, the UPF files, the data-file.xml, and the gvectors.dat.
Inside the gvectors.dat there is clearly and entry:
<CHARGE-DENSITY iotk_link="./charge-density.dat">
but the charge-density.dat is missing. When I intend to reuse this
.save directory for obtaining the band-structure, the bands.x executable
will of course crash with the error message:

 from read_rho_xml : error #        10
     searching for ./prefix.save/charge-density.xml

For the NSCF calculation I use k-point parallelisation over 4 pools of CPUs
and in the G-space over 8-to-16 CPU cores with the following options:
&CONTROL
      calculation = 'nscf' ,
           outdir = './' ,
           wfcdir = './',
       wf_collect = .true.
       pseudo_dir = './' ,
           wfcdir = './',
           prefix = '30Atom_BAND',
        verbosity = 'high'
            nstep = 3000
      max_seconds = 335600
/
&SYSTEM
   ibrav=  4, celldm(1) =7.788236356893399213, celldm(3)=21.28350100219813388,
    nat=  30, ntyp= 2, ecutwfc = 32.0, ecutrho =450.0,
 noncolin=.true. ,
 lspinorb=.true. ,
 starting_magnetization=0.0,
 occupations='smearing', smearing='methfessel-paxton', degauss=0.00735d0


I have the abovementioned issue both on a Xeon based Linux cluster running LSF with OpenMPI
and on a completely different IBM Power6 architecture with AIX, Loadleveler and MPI.
If I slightly change the cell parameters (expand about 1% along a and c) than I get a charge-density.dat.
In some other case if I use 16 instead of 8 CPUs for the R and G-space parallelisation
there will be also a charge-density.dat file generated without any issues.
Could you please give me a hint whether I have overlooked something very simple,
or if not, what could I do to be able to calculate the band structure for my system?

Best regards,
Janos Kiss.

==========================================
  Dr. Janos Kiss      e-mail: kissi at uni-mainz.de
  Johannes Gutenberg-Universitaet
  Institut f. Anorg. u. Analyt. Chemie
  AK Prof. Dr. Claudia Felser
  Staudinger Weg 9 / Raum 01-230
  55128 Mainz/ Germany
  Phone: +49-(0)6131-39-22703
  Fax:     +49-(0)6131-39-26267
  Web:     http://www.superconductivity.de/
 =========================================



More information about the users mailing list