Page 1 of 1

A question of using NEB

Posted: Sun Mar 29, 2009 2:21 pm
by biliu
Hi,

I ran my NEB calculations using 5 images with one image for one cluster node. I found that only one image is generating output files, such as, OSZICAR, OUTCAR and so on. I want to make sure first that I set up the calculations correctly. So here is my input file (INCAR):

SYSTEM = NEB
PREC = Normal !Low; MEDIUM; High; Normal; Accurate
ENMAX = 280 !Cutoff set manually (preferred)
ISTART = 0 !0:from scratch; 1:restart w/ const E-cutoff; 2:restart w/ const basis-set; 3:full restart
ICHARG = 2 !0:from initial wf; 1:from file CHGCAR; 2:superposition of charg den
NELM = 60 !Maximum electronic SCF steps
EDIFF = 1E-5 !Allowed error in total energy
EDIFFG = -0.01 !Break condition for ion relaxation
ISMEAR = 2 !-5:tetra w/ Blochl; -4:tetra w/o Blochl; -3:loop; -2:fixed partial occu; -1:Fermi; 1(N):MP method(metal); 0:Gaussian
SIGMA = 0.2 !Smearing parameter 0.2 recommended for MP method
IBRION = 1 !-1:Ion not moved; 0:MD; 1:quais-Newton; 2:CG; 3:damping factor supplied
NSW = 200 !Ionic steps
POTIM = 0.05 !Step parameter: 0.02 A stepwidth
ALGO = V !Normal; VeryFast; Fast; All; Damped
LVTOT = .FALSE. !Write total local potential, default: FALSE
STRING = -5 !String strength
IMAGES = 5 !Number of NEB intermediate images, excluding the endpoint.
ICHAIN = 0 !TS search option.0:NEB
LCLIMB = .TRUE. !Flag for climbing image algorithm.

Also, the KPOINTS:

K-Points
0
Monkhorst-Pack
3 3 1
0 0 0

Thanks a lot for your prompt help.

B

Re: A question of using NEB

Posted: Sun Mar 29, 2009 6:40 pm
by graeme
Make sure that you are using an MPI version of vasp and have asked for a multiple of 5 processors for your mpi calculation.

Re: A question of using NEB

Posted: Sun Mar 29, 2009 8:38 pm
by biliu
Thanks graeme. I did use an MPI version of the VASP code. I also used a multiple number of processors based on the number of images used. If I set up the calculations right, could this be that the processors of my computer cluster are not communicating correctly? Thanks again.

Re: A question of using NEB

Posted: Sun Mar 29, 2009 10:09 pm
by graeme
Look in the top of the OUTCAR file for the lines like:
running on 8 nodes
each image running on 1 nodes
distr: one band on 1 nodes, 1 groups

The only tag in the INCAR that should matter here is the IMAGES tag

Also, just to be sure, you are running this in a directory where the 00-06 are subdirectories? If so, in which directory are the output files showing up in?

Re: A question of using NEB

Posted: Mon Mar 30, 2009 4:23 pm
by biliu
Hi Graeme,

Thanks for you input. I found where the problem came from. I changed my mpi version which is different from the one I used to compile vasp. Now the calculations are behaving well. Thanks again.

B