On Mon, Jul 23, 2012 at 1:57 PM, Davide Ceresoli < davide.ceresoli at istm.cnr.it> wrote:
> I can supply input coordinates and output files to whoever knows > how to fix this problem with PH in version 5.0. > > Hi Davide, the good news is that with espresso 4.x the calculation would run; the bad news is that it would produce random garbage. The error arises because of a newly introduced check which verifies that the irreducible representations are actually representations. The origin of the problem is the method used to find the irreducible representations: a random pseudo-dynamical matrix is generated, than it is symmetrized and diagonalized. The resulting eigenvectors are used as phonon displacement patters; if the eigenvalues (pseudo-frequencies) of two patterns are degenerate, also the real frequencies will be. The problem is that sometimes a bit of numerical error in the diagonalization makes two modes appear as not degenerate when they should be, which in turn causes the irreducible representation to not be a representation. We could increase the threshold a bit, but then there would be a higher risk of getting accidental degeneration, causing the irr.reps. to not be irreducible... As a temporary workaround you can fiddle a bit with the threshold at line 258 of PH/set_irr.f90 or with how the random matrix in generated in PH/random_matrix.f90 (i.e. increasing the diagonal component). I would also ask you to send me the coordinates, to use as a test case, as I'm trying to find a more reliable way to generate the matrix. bests -- Lorenzo Paulatto IdR @ IMPMC/CNRS & Universit? Paris 6 phone: +33 (0)1 44275 084 / skype: paulatz www: http://www-int.impmc.upmc.fr/~paulatto/ mail: 23-24/4?16 Bo?te courrier 115, 4 place Jussieu 75252 Paris C?dex 05 -------------- next part -------------- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20120723/bcc8c2be/attachment-0001.htm