Vasp.4.x is now available. It is a major upgrade of vasp, and large parts of the code have been rewritten to support F90 (Fortran 90) features. In vasp.4.x. the param.inc file is no longer required and the corresponding parameters are now read from the INCAR file (this means that one excecutable can be used for all calculations):
|NGX, NGY, NGZ||FFT mesh for wavefunctions|
|NGXF,NGYF,NGZF||FFT mesh for charges|
|NBANDS||number of bands included in the calculation|
|NEDOS||number of grid points for DOS|
|NBLK||blocking for some BLAS calls|
eigenvalues of (default mixing * dielectric matrix)The parameters for the mixing are optimal if the mean eigenvalue is 1, and if the width of the eigenvalue spectrum is minimal. For an initial linear mixing an optimal setting for A (AMIX) can be found easily by setting . For the Kerker scheme either A or (i.e. AMIX or BMIX) can be optimized, but we recommend to change only BMIX and keep AMIX fixed (you must decrease BMIX if the mean eigenvalue is larger than one, and increase BMIX if the mean eigenvalue is smaller than one).
In vasp.4.x the prediction of wavefunctions (used for molecular dynamics) can be done without the external file TMPCAR. Just set IWAVPR in the INCAR file to 10. In this case no TMPCAR file is created and all required quantities are stored in the main memory (see section 7.23).
In vasp.4.2 the augmentation charges are forced to be zero at the boundary of the augmentation sphere, therefore results are slightly different from vasp.3.2 (usually differences are smaller than 0.01 meV). The old behavior can be restored by setting
LCOMPAT = .TRUE.in the INCAR file.
The online real space optimization has be rewritten carefully. This means that calculations with
LREAL = Onwill give results which differ from those obtained with vasp.3.2. To obtain full compatibility to vasp.3.2 set
LCOMPAT = .TRUE.in the INCAR file. The main difference concerning real space projectors between vasp.3.2 and vasp.4.2 is that the real space projectors in vasp.4.2 are forced to be zero for the outermost radius. This was not done in vasp.3.2 (and is not done in vasp.4.2 with LCOMPAT= .TRUE.) and this results in problems in the stress calculation (there is an approximately 1 in 5000/atoms chance that the stress is wrong in vasp.3.2).
VASP.4.2 is fully parallelized please read section 4 for details on how parallelization was done, and section 8 for details concerning the new switches in the INCAR file.
Section 3.9 contains a list with performance profiles of several machines. Have a look at it!
A small improvement was made to ICHARG=1, please refer to the corresponding section 7.11. This should allow a faster startup with the flag ICHARG=1 (probably nice for phonon calculations).
The algorithm IBRION=1 (Quasi-Newton update of ions) is completely rewritten. The algorithm is now much faster, and significantly more stable, because possible linear dependencies screwing up convergence of the Quasi-Newton scheme are automatically removed from the iteration history (see section 7.19).
The elastic band method is now supported by vasp to allow the calculation of energy barriers (see Sec. 8.3). This is only possible on parallel computers or clusters of workstations.
Vasp is now able to correct the 1/L and errors introduced which exists for supercells with a finite monopole (charged cells) and dipole (see Sec. 8.6).