About EDIFFG

[piaoxue001]

Dear everyone,
There is a problem confusing me. I am investigating the H adsorption-Dissociation process. During TS-searching by CI-NEB method,when the value for EDIFFG is set to 0.05eV/A, the result seems non-resonable.for example, OUTCAR in 01file:
POSITION TOTAL-FORCE (eV/Angst)
-----------------------------------------------------------------------------------
8.79906 13.59032 12.71903 0.094842 0.039910 0.071295
9.62675 13.86415 12.73074 -0.108965 -0.008318 0.089108
Why?
Does not I understand correct??

[graeme]

Are you setting ediffg to -0.05?

Does the OUTCAR report that the calculation has converged? Check that it has not terminated for another reason such as running out of time or reaching the specified NSW.

Are you using one of our optimizers, or a vasp built-in one? If ours, do you have FMAGFLAG set so that the convergence criterion considers the total RMS force instead of the max RMS force on any atom?

[piaoxue001]

Thank you for your reply .
I check that it has terminated normally, running in the time and not reaching the specified NSW.
But I do not use one of our optimizers.Is that reason?

[graeme]

And your ediffg is set negative?

Check the reported value of the RMS force in the OUTCAR and see why the calculation finished and if vasp considered it converged.

[graeme]

First, you do not have our code linked into vasp, so this is not using the climbing image NEB.

The force criteria is based upon projected forces. Look at the data under the:

CHAIN + TOTAL (eV/Angst)

to find the relevant forces for convergence.

[piaoxue001]

thank you for your help.
Please,why did you say that we do not have your code linked into vasp?? we have your code linked into vasp. What is the difference of two OUTCAR files which are availed by the NEB and CINEB methods ? If the OUTCAR contains the CHAIN + TOTAL (eV/Angst) data, the CINEB method is used? Am I right? If not please correct me.

[piaoxue001]

INCARfiles ：
system=##
ENCUT=300
ISTART=0
ICHARG=2
ISMEAR=0
SIGMA=0.1
NSW=1000
IBRION=3
EDIFFG=-0.02
ALGO=Fast
POTIM=0.1
IMAGES=3
SPRING=-5
LCLIMB=TRUE
LWAVE=FALSE
LCHARG=FALSE
PREC=Normal

[cenwanglai]

You should download vtstoolcode to your vasp src dirctory and recompile the codes again.

[piaoxue001]

Yes,you are right. Because of using different computers,I just find that a computer have VTST code linked into vasp,and the other do not have VTST code linked into vasp. Thank you for your help.Now I have another question.I find that the TOTAL-FORCE (eV/Angst) convergence in 02 05 is successful(6 images),the other is not coverged. The calculation ends normally in running time and NSW. 02 is saddle point. whether vasp only consider 02 converged？ thank you in advance.