Sub-Space-Matrix is not hermitian in DAV

[suddha]

Dear VASP Users,

Even though this problem has been discussed before in the VASP forum, I would like to discuss this again to get some answers from experienced vasp users.

I am getting the following error

bond charge predicted
N E dE d eps ncg rms
rms(c)
DAV: 1 -0.767144323689E+02 -0.60958E+00 -0.24088E-01128064 0.303E-01
0.846E+00
RMM: 2 -0.794331234876E+02 -0.27187E+01 -0.45271E-03120000 0.129E-02
0.250E+01
RMM: 3 -0.772415358411E+02 0.21916E+01 -0.79535E-04120000 0.534E-03
0.147E+01
RMM: 4 -0.763071360321E+02 0.93440E+00 -0.27991E-04120000 0.346E-03
0.444E+00
RMM: 5 -0.763992391792E+02 -0.92103E-01 -0.14888E-04120000 0.247E-03
0.757E+00
RMM: 6 -0.761854686246E+02 0.21377E+00 -0.10570E-04120000 0.199E-03
0.129E+00
RMM: 7 -0.761831806685E+02 0.22880E-02 -0.91569E-05120000 0.182E-03
0.706E-01
RMM: 8 -0.761868104793E+02 -0.36298E-02 -0.46987E-05120000 0.124E-03
0.729E-01
RMM: 9 -0.761872928288E+02 -0.48235E-03 -0.13613E-05120000 0.661E-04
0.411E-01
RMM: 10 -0.761885913648E+02 -0.12985E-02 -0.22308E-05119999 0.810E-04
0.163E-01
RMM: 11 -0.761889464785E+02 -0.35511E-03 -0.55940E-06120000 0.434E-04
0.107E-01
RMM: 12 -0.761890680683E+02 -0.12159E-03 -0.18996E-06119884 0.236E-04
0.610E-02
RMM: 13 -0.761891078646E+02 -0.39796E-04 -0.67992E-07117909 0.144E-04
0.458E-02
RMM: 14 -0.761891082755E+02 -0.41088E-06 -0.14019E-07113244 0.709E-05
6 F= -.76189108E+02 E0= -.76191918E+02 d E =-.174117E+00 mag= 7.4322
curvature: -25.44 expect dE=-0.587E-01 dE for cont linesearch -0.587E-01
ZBRENT: increasing intervall
opt : 94.0000 next Energy= -76.312053 (dE=-0.297E+00)
bond charge predicted
N E dE d eps ncg rms
rms(c)
WARNING: Sub-Space-Matrix is not hermitian in DAV 13
3.869115711473709E-005


My INCAR file is as given below

PREC = ACCURATE
ENCUT = 500
IBRION = 2
NBANDS = 3000
NSW = 100
ISIF = 3
ALGO = FAST
NELMAX = 100
NELMIN = 4
EDIFF = 1.0E-06
EDIFFG = -0.01
VOSKOWN= 1
NBLOCK = 1
ISPIN = 2
INIWAV = 1
ISTART = 0
ICHARG = 2
LWAVE = .FALSE.
LCHARG = .FALSE.
ADDGRID= .FALSE.
ISMEAR = 1
SIGMA = 0.2
POTIM = 0.002
LREAL = .FALSE.
NPAR = 6

The initial structure given in the POSCAR file is
title
1.
-7.276000000 -3.638000000 -3.638000000
-3.638000000 -3.638000000 -7.276000000
-3.638000000 -7.276000000 -3.638000000
4 4
Cartesian
-10.914000000 -10.914000000 -10.914000000 !A
-5.457000000 -5.457000000 -5.457000000 !A
-3.638000000 -3.638000000 -3.638000000 !A
-1.819000000 -1.819000000 -1.819000000 !A
-14.552000000 -14.552000000 -14.552000000 !B
-12.733000000 -12.733000000 -12.733000000 !B
-9.095000000 -9.095000000 -9.095000000 !B
-7.276000000 -7.276000000 -7.276000000 !B

The job was run over 48 processors (12 processors in each node) with a simple job file
#!/bin/bash

source /opt/intel/composerxe/bin/compilervars.sh intel64
source /opt/intel/impi/4.0.2.003/intel64/bin/mpivars.sh
VASP_DIR=/home/appl/vasp-5.3/bin
mpirun -f machines -nolocal -n 48 $VASP_DIR/vasp-nc



The structure was generated using the ATAT code.
As can be seen from the previous response from "ADMIN", the POTIM was supposed to be reduced. It has been reduced to 0.002. Even then, the problem persists...
It may also be mentioned that the default value of POTIM tag did not work either...It was gradually reduced from 0.1 to 0.008, 0.005 to 0.002.
The same error persists.
The ALGO tag also was changed....the same problem persists...

The IBRION tag also was tested...the same error persists..

Can I please get some response on this matter from vasp users as to what could be the possible solutions for this problem..

Thank you and regards
Suddhasattwa Ghosh

[admin]

as ZBRENT could not find a reasonable interpolation based on the forces and positions, maybe the extrapolated geometry is unreasonable. Please check it

[suddha]

Dear ADMIN and vasp users,
thanks for the response. The structures have been generated for doing cluster expansion calculations and their respective energies are obtained in a high-throughput way. (runstruct_vasp in ATAT)

I would like to know how to avoid this problem of unreasonable geometry..

This problem started appearing only for primitive cells with more than 6 atoms. The same INCAR file works well for the following structure.
title
1.
1.819000000 1.819000000 -5.457000000
5.457000000 1.819000000 -1.819000000
0.000000000 -7.276000000 0.000000000
2 6
Cartesian
5.457000000 1.819000000 -5.457000000 !A
7.276000000 0.000000000 -7.276000000 !A
1.819000000 -5.457000000 -1.819000000 !B
1.819000000 -1.819000000 -1.819000000 !B
3.638000000 0.000000000 -3.638000000 !B
3.638000000 -3.638000000 -3.638000000 !B
5.457000000 -1.819000000 -5.457000000 !B
7.276000000 -3.638000000 -7.276000000 !B

Can you please suggest on an alternative way of doing a full relaxation of these type of ATAT structures...

thanks
Suddhasattwa