UT theoretical chemistry code forum

Dear all,

I've been running a dynamical matrix calculation in a supercomputer. Unfortunately, the job reached its scheduled 72-hour walltime before all of the desired displacements present in the DISPLACECAR file were computed. I would like to know if there is there a way to restart the simulation without computing those displacements again. In other words, how do I know which displacements from which atoms have been already obtained?

Thanks a lot for any input!

Kind regards!

Roberto Iglesias

I think the best way to deal with queue time limits is to divide up the dynamical matrix calculation into sets of displacements which can be completed within the time provided. You can see instructions for how to do this here: http://theory.cm.utexas.edu/vtsttools/dymmat.html

The basic idea is to make a set of DISPLACECAR files, each containing a set of displacements which can be run. These calculations can be run simultaneously. Then, you can combine them into a single dynamical matrix using a command such as:

dymmatrix.pl n DISPLACECAR1 .. DISPLACECARn OUTCAR1 .. OUTCARn

You can also do this after-the-fact. If you have a calculation for which a set of displacements have been done and the force information is included in the OUTCAR, you can remove these displacements from the DISPLACECAR and run the calculation again with a new DISPLACECAR including the calculations which still need to be calculated.

Dear Graeme,

Thank you very much for your answer. I see what you mean, but still, and this may be a very naive question, I don't know how to find out to which atoms the forces included in the OUTCAR file are linked to, so as to be able to remove their displacements from the DISPLACECAR file.

Any hints?

Thanks again!

Roberto

Sure, you can just count the number of forces evaluated, and these will correspond to the displacements listed in order in your DISPLACECAR file. Remember that the calculation does one force evaluation at the minimum, so you need to subtract 1 from the force evaluations to get the number of displacements evaluated.

Thanks a lot, Graeme! I think I get it. I'll try and do as you say asap!

Kind regards!
Roberto