My Community

Greetings all, here's my predicament:
I'm trying to do a (gallium) dimer calculation and I'm following the VASP guide (dated 23 April 2009), which advises using constant velocity MD (i.e. IBRION=0, SMASS=-2) for this purpose. I'm using VASP 5.2 and it consistently crashes after the 10th MD step with the following error:

--------------------------------------------------------------------------------
positions are collinear
bond charge predicted
prediction of wavefunctions
wavefunctions rotated
writing wavefunctions
wavefunctions on file TMPCAR rotated
ERROR: while reading WAVECAR, header is corrupt
--------------------------------------------------------------------------------

Obviously, something goes wrong during the wavefunction rotation, it seems this rotation is hard-coded to be every 10 MD steps. Now, the punchline is: I DO NOT get this error in VASP 4.6 with exactly the same INCAR, POTCAR, POSCAR and KPOINTS.

I tried changing the mesh in KPOINTS, the constant velocity in POSCAR and tried two different POTCARs for gallium, but the error persists. I can't make much sense out of WAVECAR because it's a binary, so I don't have a clue what 'corrupt' could mean. I'm also not very familiar with the source code, and am a complete newbie in DFT, so I though I'd ask if anybody else has encountered / resolved the same problem. Let me know if I should post more info. Any help would be appreciated.

please show your complete INCAR file, and please have a look which IWAVPR was used during the calculation.
do you use a serial of a parallel executable?

I'm using a serial executable for this calculation. In the OUTCAR file it says IWAVPR = 2. The used INCAR file is:

-------------------------------------------------------------------
System = Ga dimer
IBRION = 0 ! Ab-inition MD
SMASS = -2 ! Constant velocity MD
POTIM = 1 ! Timestep (1fs)
NSW = 20 ! Run for 20 MD steps
ISYM = 0 ! no symmetry
ISPIN = 2 ! 2 -> allow for spin polarisation
VOSKOWN = 1 ! this is important(?), in particular for GGA (PW91)
ISMEAR = 0 ! Gaussian smearing (finite T), -tive occupancies?
SIGMA = 0.1 ! intermid. smearing width
AMIX = 0.2 ! mixing set manually
BMIX = 0.0001
! Electronic relaxation
PREC = Accurat !
EDIFF = 1E-6 ! stopping-criterion for electronic upd.
-------------------------------------------------------------------

Same problem. Crash after 10 MD steps with following message:
----------------------------------------
A bond charge predicted
prediction of wavefunctions
wavefunctions rotated
writing wavefunctions
wavefunctions on file TMPCAR rotated
ERROR: while reading WAVECAR, header is corrupt
--------------------------------------
This is for serial Vasp5.2.2 build with gfortran on Linux machine.
INCAR is:
----------------------------------------
PREC = Normal ! standard precision
ENMAX = 400 ! cutoff should be set manually
ISMEAR = 0 ; SIGMA = 0.1

IBRION = 0 ! molecular dynamics
NSW = 100 ! 100 steps
POTIM = 1.0 ! timestep 1 fs

SMASS = -3 ! Nose Hoover thermostat
TEBEG = 2000 ; TEEND = 2000 ! temperature

----------------------------------------
in OUTCAR file:
IWAVPR = 2

/Jakob Blomquist

I get the same problem too. After 9 steps of constant velocity dynamics the run crashes just after it said

positions are collinear
Information: wavefunction orthogonal band 200 0.8861

followed by one line for each core saying the same:

ERROR WAVPRE_NOIO: Call to routine ZHEEV failed! 1

My INCAR is as follows:

Pd32H, 2x2x2 fcc cells

ISTART = 0

ENCUT = 350.0
IALGO = 38
PREC = high
ISMEAR = 1
SIGMA = 0.1

IBRION = 0
ISYM = 0
SMASS = -2
POTIM = 1.0
NSW = 21

I get the same problem too