ISTART = 1 does not restart from WAVECAR

[cpp6f]

vasp.4.6.27

When I set ISTART = 1 and provide a WAVECAR file, the program sets ISTART to 0 (printed in the OUTCAR file) and reinitializes the calculation anyway. The manual says that vasp 4.4 does not support this option for parallel versions, but that version 4.5 does not have this problem. Here is my INCAR file

Electronic minimization
PREC = NORMAL
GGA = 91 #Change ot "91" to run Perdew-Wang 91
VOSKOWN = 1
LREAL = .TRUE. #Use FALSE for small systems
IALGO = 48
ENCUT = 396
EDIFF = 1e-4 #Criteria for SCF convergence
ISYM = 0 #0=no symmetry, 1=symmetry on
ISPIN = 1 #1=closed shell systems, 2=spin polarized
NELM = 400
ISTART = 1
Ionic relaxation
ISIF = 0 #2=optimize ion positions, 3=optimize ions, shell shape and size
EDIFFG = -0.05 #Positive converges on energy, Negative converges on forces
NSW = 0
IBRION = 2 #1=Quasi Newton, 2=Conjugate Gradient, 0=MD, 5=Vibration
POTIM = 0.5 #Ionic step size scaling, reduce value if geom steps too big
DOS related values
ISMEAR = 2 #Smearing set for metals, change for others or molecules
SIGMA = 0.2 #Smaller values --> more accurate energies, harder to converge
Parallelization
NSIM = 4 #Use always on hydra, birch, aspen
NPAR = 2 #Set equal to number of nodes
LPLANE = .TRUE. #Use always on hydra, birch, aspen
LSCALU = .FALSE. #Use always on hydra, birch, aspen

[admin]

WAVECAR can only be read successfully if ALL parameters (NGX, NGY NGZ, the number of k-points and the number of bands) of the old and the new run match. Furthermore, one has to continue from CONTCAR from the previous run (if ionic relaxation was done then).
Please check these parameters of run (N-1) and N

[cpp6f]

Thanks. Is there any way that I can take the WAVECAR of a converged calculation and use it as an initial guess for the wavefunction of a slightly displaced structure (i.e. slightly displacing the stationary geometry along one of the normal vibrational modes)?

[Veronika]

No, since also the atomic positions have to match (as was written by the admin: ALL parameters of the old and the new run ...).