Unreasonable total energy

[forsdan]

Dear all,

I use the PAW-GGA-PBE potential for Fe (with 8 valence electrons) to calculate DOS for strained bcc-Fe slabs. I have two different slabs (different strains) and using the same INCAR, KPOINTS, POTCAR file for both (see below). For slab A corresponding to POSCAR_A everything works as expected and converges to a reasonable solution. For slab B (POSCAR_B) everything instead goes haywire and the energy goes way negative (order 10^7).

I've ran the simulations several times to check if it wasn't a temporary error. I've doublechecked the positions in different viewers and they are very close to the nearest neighbor distance in bcc-Fe. I don't get any warnings in the OUTCAR either concerning too short distances. Any suggestions to if I'm missing something trivial here?

Best regards,
/Dan Fors

Slab A:
-------------------

N E dE d eps ncg rms rms(c)
DAV: 1 0.551891628582E+03 0.55189E+03 -0.33583E+04 18432 0.192E+03
DAV: 2 -0.192729837913E+03 -0.74462E+03 -0.72909E+03 18464 0.535E+02
DAV: 3 -0.275235687319E+03 -0.82506E+02 -0.79830E+02 23680 0.213E+02
DAV: 4 -0.284353067310E+03 -0.91174E+01 -0.88475E+01 26000 0.678E+01
DAV: 5 -0.285288487635E+03 -0.93542E+00 -0.90635E+00 26912 0.156E+01
DAV: 6 -0.285450113382E+03 -0.16163E+00 -0.15983E+00 28880 0.352E+00
DAV: 7 -0.285483123861E+03 -0.33010E-01 -0.33147E-01 28128 0.114E+00
....

Slab B:
-------------------
N E dE d eps ncg rms rms(c)
DAV: 1 -0.149265736212E+07 -0.14927E+07 -0.11014E+07 36864 0.191E+03
DAV: 2 -0.218257509793E+07 -0.68992E+06 -0.21837E+07 36864 0.349E+03
DAV: 3 -0.424116855140E+07 -0.20586E+07 -0.42335E+07 36864 0.383E+03
DAV: 4 -0.611958599996E+07 -0.18784E+07 -0.61014E+07 36864 0.438E+03
DAV: 5 -0.742104644137E+07 -0.13015E+07 -0.73968E+07 36864 0.467E+03
DAV: 6 -0.843741588425E+07 -0.10164E+07 -0.84101E+07 36864 0.490E+03
DAV: 7 -0.923544646380E+07 -0.79803E+06 -0.92041E+07 36864 0.508E+03
....


POSCAR_A:
-------------------
Fe Surface (FM2)
1.0000000000000000
2.9007000000000001 0.0000000000000000 0.0000000000000000
0.0000000000000000 5.8014000000000001 0.0000000000000000
0.0000000000000000 0.0000000000000000 18.9386000000000010
10
Direct
0.5000000000000000 0.2500000000000000 0.5736538075676132
0.5000000000000000 0.7500000000000000 0.5736538075676132
0.0000000000000000 0.0000000000000000 0.6467109501230291
0.0000000000000000 0.5000000000000000 0.6467109501230291
0.5000000000000000 0.2500000000000000 0.7197680926784450
0.5000000000000000 0.7500000000000000 0.7197680926784450
0.0000000000000000 0.0000000000000000 0.7928252352338609
0.0000000000000000 0.5000000000000000 0.7928252352338609
0.5000000000000000 0.2500000000000000 0.8658823777892770
0.5000000000000000 0.7500000000000000 0.8658823777892770


POSCAR_B:
-------------------
Fe Surface (FM2)
1.0000000000000000
2.8715000000000002 0.0000000000000000 0.0000000000000000
0.0000000000000000 5.7430000000000003 0.0000000000000000
0.0000000000000000 0.0000000000000000 18.7291999999999987
10
Direct
0.5000000000000000 0.2500000000000000 0.5693996540161886
0.5000000000000000 0.7500000000000000 0.5693996540161886
0.0000000000000000 0.0000000000000000 0.6440798325609209
0.0000000000000000 0.5000000000000000 0.6440798325609209
0.5000000000000000 0.2500000000000000 0.7187600111056532
0.5000000000000000 0.7500000000000000 0.7187600111056532
0.0000000000000000 0.0000000000000000 0.7934401896503857
0.0000000000000000 0.5000000000000000 0.7934401896503857
0.5000000000000000 0.2500000000000000 0.8681203681951178
0.5000000000000000 0.7500000000000000 0.8681203681951178

INCAR
-------------------
SYSTEM = Fe Surface (FM2)
ISTART = 0
ICHARG = 2
EDIFF = 1E-5
ENCUT = 700.0000 eV
ENAUG = 700.0000 eV
PREC = Accurate
LREAL = .FALSE.
ISPIN = 2
LWAVE = .FALSE.
NELM = 400
NELMDL = -12
LMAXMIX = 6
ALGO = Normal
AMIX = 0.2
BMIX = 0.001
AMIX_MAG = 0.8
BMIX_MAG = 0.0001
NSW = 0
ISMEAR = -5
LORBIT = 12
NEDOS = 1000
MAGMOM = 10*-3


KPOINTS
-------------------
Monkhorst Pack
0
Monkhorst Pack
16 8 6
0 0 0



<span class='smallblacktext'>[ Edited ]</span>

[forsdan]

ok, I forgot to check the system out-file from the calculation and found "Sub-Space-Matrix is not hermitian in DAV", which have been well discussed before in the threads. I switched to another LAPACK-library and everything seems to be fine.

Best regards,
/Dan

<span class='smallblacktext'>[ Edited Sat May 17 2008, 11:21PM ]</span>