Page 2 of 2

Re: DynMat calculations

Posted: Fri Aug 01, 2008 2:25 pm
by graeme
Goodness, this is way more difficult than it should be.

I was very happy with my ISYM solution. Well, actually, I never had vasp quit after 1 iteration, but with ISYM=1 (the default) they symmetry constraints set many components of the force to be zero and the resulting frequencies are nonsense. With ISYM=0, my results look reasonable. I'll put both calculations at

http://theory.cm.utexas.edu/prefactor/

I did reduce the number of kpoints so my calculation would be faster, but I doubt that this has anything to do with the problem that you are seeing.

What if you do a plain minimization using the same IBRION=3 and POTIM=0 settings, but turn off ? Does it quit after 1 iteration, or does it do the specified 40 iterations.

I have a feeling that there is some fundamental problem here, rather than something specific with our code. I'm hoping that you see something in these runs that is different from your setup that gets to the bottom of this.

Re: DynMat calculations

Posted: Thu Jan 15, 2009 12:55 pm
by dekoning
Hi,

I am also trying to use the VTSTTools to compute the dynamical matrix. Running vasp 4.6.34 I have precisely the problem mentioned earlier here. The job stops after two steps. Then I used an earlier version, vasp 4.6.28 using precisely the same input files (INCAR, DISPLACECAR etc) and to my surprise the job does NOT stop this time. It is still running and will take a few days to finish.

Re: DynMat calculations

Posted: Tue Jan 20, 2009 7:00 am
by graeme
Check to make sure that your force criteria (ediffg) is smaller than the force on your initial image. Alternatively, you can also set ediffg very small (1e-5 or so).

I don't understand how a different version of vasp would change the behavior of our dynamical matrix calculation. Perhaps check to see if the VTST version is the same; this is printed in the OUTCAR.

Re: DynMat calculations

Posted: Fri Jan 30, 2009 3:42 pm
by dekoning
I checked the VTST versions in both runs, and indeed they are different. The newer version, 2.03c stops after 2 steps. Version 2.01b, on the other hand, seems to run OK, in that it continues beyond these 2 steps. Yet, the job (NSW=433 in my case) just mysteriously dies after 100 or so steps. I checked the OUTCAR
and there is nothing wrong there. I ran the job twice and it happened both times.