Page 1 of 1

ZPOTRF error - still trying

Posted: Fri Oct 14, 2005 9:36 pm
by bth20
Hi

As I posted the other day, I am running noncollinear calculations and I'm using the tetrahedron method (ISMEAR = -5). BUT, I'm experiencing crashes almost straight away with several occurrences of the message

magnetization density of overlapping atoms calculated
LAPACK: Routine ZPOTRF failed! 1

I've now tried the suggestion from a previous post of using the package lapack_double.o but I get the same error.

Any suggestions?

Thanks

Ben

PS recall my INCAR file is:

SYSTEM = Co monatomic chain (Noncollinear)
LSORBIT = .TRUE.
ISYM = 0
GGA_COMPAT = .FALSE.
ISMEAR = -5
MAGMOM = 4.0 0.05 -0.05
ICHARG = 1
PREC = High
NELM = 300
NELMIN = 4
NBANDS = 20
EDIFF = 1.0e-5
EDIFFG = 1.0e-4
ISPIN = 2
IBRION = 1
ISIF = 2
NSW = 100
LORBIT = 10
RWIGS = 1.302

ZPOTRF error - still trying

Posted: Mon Oct 24, 2005 7:51 am
by Veronika
You are NOT running noncolinear calculations with this INCAR file.
If you want to do so, you have to set LNONCOLLINEAR = .TRUE. -tag ...
read for example: http://cms.mpi.univie.ac.at/vasp/vasp/node142.html

ZPOTRF error - still trying

Posted: Thu Oct 27, 2005 9:07 pm
by bth20
I disagree. Check out this:

http://cms.mpi.univie.ac.at/vasp/vasp/node144.html

in particular the very first line which reads:

LSORBIT = .TRUE. switches on spin-orbit coupling and automatically sets LNONCOLLINEAR= .TRUE..

I rest my case

Ben

ZPOTRF error - still trying

Posted: Fri Nov 04, 2005 7:06 am
by Veronika
Sorry, I overlooked the LSORBIT tag ....

ZPOTRF error - still trying

Posted: Mon Dec 05, 2005 11:06 am
by admin
please note that
1) the appropriate k-mesh for a chain of atoms (not interacting with neighboring chains) probably is conflicting with ISMEAR=-5
2) depending on the geometry input / non-collinear setup: with your definition of MAGMOM, the magnetic moment for exactly one atom is specified. If there are more than 1 atoms in the chain, this may lead to a starting density which is not reasonable and causes the ZPOTRF error.