WARNING: Sub-Space-Matrix is not hermitian in DAV, but not with ICHARG=12

Message

SchrodingersCat · #1 Post by **SchrodingersCat** » Mon Jun 25, 2012 5:18 am

I am having a lot of trouble obtaining electronic convergence of surfaces. I tried to debug the convergence problem using the steps outlined in the "Electronic Optimization" presentation included with the VASP tutorials.

I am able to obtain electronic convergence of the slab with the system using ICHARGE=12 but if I try to use the resulting wave-functions after self-consistency is reached I still get the dreaded "not hermitian" error. I have already tried changing the mixing parameters as indicated in the tutorial, but I still am not having any success. What else can I do?

What else can cause the "not hermitian" error? I have seen lots of forum posts regarding the problem, but few actual solutions. I am already using the built-in lapack, and I have even tried the radical suggestion in the post which involves commenting out the offending routine, to no avail. Is there any solution to this problem?
<span class='smallblacktext'>[ Edited ]</span>

#2 Post by **admin** » Wed Jul 11, 2012 4:14 pm

usually this non-hermitian error is caused by a charge density (--> V) which
leads to a hamiltonian that cannot be diagonalized.
this may eg be due to an inconsistent WAVECAR read from a previous run:
please check if ALL parameters (FFT grids, lattice parameters, k-meshes,
NBANDS) are kept the same for the writing and the reading vasp-run