Energy increasing during geometry relaxation

[jcconesa]

Hi,
I'm calculating total electronic relaxation in a molybdenum bronze model (KMo3O9), starting from the experimental geometry. Using just GGA, spin polarized, relatively gook K mesh considering that according to experiment it a semiconductor. Surprisingly, during relaxation (IBRION=2) total energy continuously increases; with other settings it rather oscillates (period is [blockquote][/blockquote] about 8-10 geometry relaxation cycles). Increasing ENCUT to 520 eV does not solve the problem. Any hint on the possible cause of the problem? Is such behaviour known? If needed i can send all input files.
Thanks

[alex]

Hi,

please send the inputs (no POTCAR needed).

Thanks,

alex

[jcconesa]

These are the input files:

POSCAR:
....................
TITL KMo3O9
8.129000
1.000000 0.000000 0.000000
0.547059 0.837094 0.000000
-0.031623 -0.017111 0.785744
2 6 18
Direct
0.683000 0.683000 0.835100
0.317000 0.317000 0.164900
0.877000 0.877000 0.359400
0.123000 0.123000 0.640600
0.799700 0.320500 0.269000
0.200300 0.679500 0.731000
0.679500 0.200300 0.731000
0.320500 0.799700 0.269000
0.913800 0.431400 0.203700
0.845700 0.321300 0.591100
0.568600 0.086200 0.796300
0.678700 0.154300 0.408900
0.086200 0.568600 0.796300
0.154300 0.678700 0.408900
0.431400 0.913800 0.203700
0.321300 0.845700 0.591100
0.773900 0.773900 0.482400
0.847300 0.847300 0.096700
0.459900 0.459900 0.704900
0.961500 0.961500 0.689700
0.226100 0.226100 0.517600
0.152700 0.152700 0.903300
0.540100 0.540100 0.295100
0.038500 0.038500 0.310300
0.258500 0.741500 0.000000
0.741500 0.258500 0.000000
....................
(this is the primitive cell of the centered monoclinic structure given in
FA Schroder, W Schuckmann, Z. Naturforsch. B (1977) 32, 365)

KPOINTS:
....................
gamma-centered
0
G
6 6 8
....................

INCAR:
....................
SYSTEM = opt kmo3o9
NPAR = 4
LPLANE = .FALSE.
ALGO = Normal
LSCALAPACK = .FALSE.
ISMEAR = 0
SIGMA = 0.03
ENCUT=450
PREC = Accurate
LREAL=Auto
BMIX = 2.0
LASPH = .TRUE. ; ADDGRID = .TRUE. ; LMAXMIX = 4
ISPIN = 2
MAGMOM = 2*0 1 23*0
IBRION = 2
ISIF = 3
NSW = 100
EMIN = -10
EMAX = 10
NEDOS = 2001
LORBIT = 11
LWAVE=.FALSE.
....................

The POTCAR file contains:
PAW_PBE K_pv 17Jan2003
PAW_PBE Mo_pv 04Feb2005
PAW_PBE O 08Apr2002

The evolution of the relaxation can be seen using the command
grep "free energy" OUTCAR
which yields this output:
.....................
free energy TOTEN = -209.860906 eV
free energy TOTEN = -209.666404 eV
free energy TOTEN = -209.876452 eV
free energy TOTEN = -209.881599 eV
free energy TOTEN = -209.882807 eV
free energy TOTEN = -209.871554 eV
free energy TOTEN = -209.891482 eV
free energy TOTEN = -209.903745 eV
free energy TOTEN = -209.905472 eV
free energy TOTEN = -209.908044 eV
free energy TOTEN = -209.911707 eV
free energy TOTEN = -209.900578 eV
free energy TOTEN = -209.902301 eV
free energy TOTEN = -209.902550 eV
free energy TOTEN = -209.884955 eV
free energy TOTEN = -209.898649 eV
free energy TOTEN = -209.901050 eV
free energy TOTEN = -209.903351 eV
free energy TOTEN = -209.905679 eV
free energy TOTEN = -209.907614 eV
free energy TOTEN = -209.889286 eV
free energy TOTEN = -209.904717 eV
free energy TOTEN = -209.913145 eV
free energy TOTEN = -209.913069 eV
free energy TOTEN = -209.912896 eV
free energy TOTEN = -209.908107 eV
free energy TOTEN = -209.908469 eV
free energy TOTEN = -209.908686 eV
free energy TOTEN = -209.905396 eV
free energy TOTEN = -209.906708 eV
free energy TOTEN = -209.904487 eV
free energy TOTEN = -209.905909 eV
.....................

Although it stops well before NSW is reached, it does not seem like a normal relaxation convergence. In other related systems (containing also partialy reduced Mo) the energy slowly but steadily increases after having gone through a minimum.
Hopefully someone can help.
Thanks