Please look in the UG at clmextrapol, and also what is done in the optimize.job script created by x optimize. The clmextrapol utility is designed for this type of issue.
___________________________ Professor Laurence Marks Department of Materials Science and Engineering Northwestern University www.numis.northwestern.edu MURI4D.numis.northwestern.edu Co-Editor, Acta Cryst A "Research is to see what everybody else has seen, and to think what nobody else has thought" Albert Szent-Gyorgi On Apr 23, 2015 4:26 AM, "Madesis Ioannis(John)" <imade...@physics.uoc.gr> wrote: > First of all, I want to apologize for sending yesterdays answer to a new > topic.As it probably needs to be erased, you will find it as > "msg12195.html". Hopefully, I am not doing the same mistake now. > Continuing, I will copy that text here for proper continuity of the > discussion. > > This is why I specified that I work with phonopy. Phonopy, (if you don't > already know) requires all generated structures to be of P symmetry. To > be honest, this is not always possible since I cannot override Wien2K > algorith, which detects possible symmetries (initialization runs in > batch mode, inside a home-made script). For phonopy, all atoms must be > inequivalent, for proper results. Since I run FIRST all the > initializations, and keep a log file, I am in position to tell if the > spacegroup of all structures is P1, which again, is a primary target for > a good phonopy calculation. > > So, your suggestion to decrease the duration of the convergence for the > n+1 cases is to > > cp case_n/case_n.clmsum case_n+1/case_n+1.clmsum for 1<n<=Number of > displacements > > after the initialization of the cases and the run of case_1. > Are there any other files that could be transfered, therefor, enhance > this idea even more? > > Today, I came to the realization, that the idea worked! I am currently > running a calculation with 78 displacements. > At #15 and greater, I replaced the recommended file, with the one > proposed. The results so far, are the following: > The first 14 structures required required 20-21 cycles for -cc 0.001 -ec > 0.001 -fc 1 convergence. So, at this point I regard them as the "rule" > for the duration. > Of following 15 structures that have run since the replacement, 11 > required only 12 cycles, with the same convergence criteria. Checking > the time of the last modification, I can see that the 12 cycles require > 54 minutes, whereas the 20 cycles require 86. If one combines the > "success rate" (so far) with the improvement in time for convergence > (For the particular sample with 52 atoms), the result is 25% > improvement, that could probably go up to 37%, depending on the > following "success rate". > > -- > Ioannis Madesis (Μαδέσης Ιωάννης) > PhD Student > Atomic and Molecular Physics > Department of Physics > University of Crete > (0030)-210-6503598 > _______________________________________________ > Wien mailing list > Wien@zeus.theochem.tuwien.ac.at > http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien > SEARCH the MAILING-LIST at: > http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html >
_______________________________________________ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
