SCF error with Lu atom

Message

okuno · #1 Post by **okuno** » Tue Nov 14, 2006 2:01 am

Dear Vasp Users.

I'm tryning to calculate the electric structure whici has Lu ( rare-earth
atom) atom.

In the SCF cycle of the calculation, there there message with

WARNING: Sub-Space-Matrix is not hermitian in DAV 8
1.941943090033175E-006

and calculation is not conversed.

I had taken the paw-GGA method, and my INCAR file is geven below.

What is the problem with the Lu atom? ( With not Lu atom,
the calculation is conversed smoothly.)

PREC=High
ISMEAR=-5
SIGMA=0.1
EDIFF=1e-6
IALGO = 38

NELMIN=8

GGA = 91

LORBIT=11

#2 Post by **admin** » Wed Nov 15, 2006 1:59 pm

1) please increase LMAXMIX to 6
2) have a look whether the input geometry of that ionic step is reasonable (XDATCAR)

mattis · #3 Post by **mattis** » Wed Dec 05, 2007 7:53 am

I have same problem in my calculations. I am using PAW PBE potentials

In my case everything works very well if I use the '+3' version of the Lu potential. That is, nice and smooth convergence.
But if I run the same system (using same POSCAR) with the '+4' potential everything crashes.
This is what it looks like in the very first eletronic relaxation:

N E dE d eps ncg rms rms(c)
DAV: 1 0.322171344693E+03 0.32217E+03 -0.23663E+03553152 0.153E+03
WARNING: Sub-Space-Matrix is not hermitian in DAV 18 -1.87644267972666631E-2
.... then it crashes
I am having LMAXMIX = 6

Any suggestions on how to solve this problem would be really appreciated!
Can it be so that it is the potential itself that is wrong? The one I am using was created september 6th 2000. Are there any later ones available?
best regards
M.