WAVECAR: different cutoff or change in lattice found

Message

TAT · #1 Post by **TAT** » Fri Oct 24, 2008 3:39 am

Hi,
I am trying to run a simulation of system containing Carbon and Calcium. The input file looks as
LREAL = Auto
ENCUT = 350 ! Energy cutoff
ISTART = 1
ICHARG = 1
EDIFF = 1E-6
EDIFFG = -.001 ! Break condition for ionic
NSW = 1000 ! number of Ionic steps
IBRION = 2 ! Conjugate gradient algorithm
POTIM = .2 ! scaling constant
ISIF = 3
ISYM = 2
SYMPREC = 1E-6
ISMEAR = 0 ! Gaussian Smearing
SIGMA = .03 ! Width of smearing
NPAR = 1

The cell size is ~4.29 x 4.29 x 8.44 and I do K point sampling of 12 x 12 x 6

once the converging criteria is obtained and I restart the run, I get this even time

WAVECAR: different cutoff or change in lattice found
I tried to increase the accuracy but doesn't help
this I have with this system only.
please help.

#2 Post by **admin** » Wed Nov 12, 2008 2:42 pm

you are correct, if the lattice parameters are changed during automatic cell optimization, the number of basis functions (NXG,...) within the cutoff-sphere changes as well because the length of the unit vetctors in the reciprocal lattice changes according to a,b,c. The best procedure to do this automatic cell geometry optimization (especially if the initial choice of the lattice parameters is not a very good one) is
1) do a few NSW steps (3-5)
2) restart from CONTCAR without using WAVECAR from that run to continue, until the changes in cell volume and shape become very small.
(please mind that there may be basis set errors using a fixed basis set (NGX,...) if the cell volume/shape changes significantly during automatic relaxation.)
You can find a few very nice viewgraphs discussing this behaviour in the accuracy.pdf VASP-workshop tutorial (to be downloaded from http://cms.mpi.univie.ac.at/vasp-workshop, tutorials and session talks)