The ABINIT Users Mailing List ( CLOSED )

[Damien Caliste <damien.caliste@cea.fr>]

Hello,

Le 31/01/2008, Matthieu Verstraete <mjv500@york.ac.uk> a écrit :
> I've tried out the Poisson solver for isolated systems, on a single
> water molecule (attched stuff) and it crashes abinit 5.5.3 on an 
> opteron/linux/pgi. Density gets NaN in it and code stops.
Ok, my fault, I didn't take care of the case where rhor may has
negative values. Then, the computation of the XC potential may crash
(note that on my machine it stops with a clean error from invcb).

I correct this, using mkdenpos(). See my revision 87 from my public
5.5.3 or the following patch:
=== modified file 'src/14poisson/psolver_rhohxc.F90'
--- src/14poisson/psolver_rhohxc.F90    2007-11-14 06:20:39 +0000
+++ src/14poisson/psolver_rhohxc.F90    2008-01-31 15:27:35 +0000
@@ -131,6 +131,10 @@
      end if

      hgrid = rprimd(1, 1) / dtset%ngfft(1)
+
+     ! Make the density positive everywhere (but do not care about
gradients)
+     call mkdenpos(dtset%nfft, 1, dtset%nspden, rhor)
+
      ! This may not work with MPI in the planewave code...
      call PSolver('F', 'G', mpi_enreg%me, mpi_enreg%nproc,
dtset%ngfft(1), & & dtset%ngfft(2), dtset%ngfft(3), dtset%ixc, hgrid,
hgrid, hgrid, &

Thanks for pointing this. I've added it as a test case in the 5.6
branch.

Damien.