Thanks for your fast response. Even though I am not using matdyn.x, which is well documented, but dynmat.x, which lacks this info about eigenvectors vs. eigendisplacement, your answer revealed my mistake. I had already tried to multiply the vectors with the masses, as I suspected they might be eigendisplacements, but it still didn't give me orthogonal eigenvectors. Simply because I forgot to take the square root of the masses first. That fixed it.
Nevertheless it would be nice if the documentation could be a little clearer for dynmat.x, to specify that it prints non-orthogonal phonon displacements, like matdyn.x for flvec (in contrast to fleig). Currently it just states that it diagonalizes the dynamical matrix (which I assume results in orthogonal mode vectors), and prints out frequencies and modes to filout (dynmat.out). Thanks again, Sridhar! Flo On Wed, Aug 27, 2014 at 7:27 PM, Sridhar Sadasivam <sridhu88 at gmail.com> wrote: > If you are talking about the vectors printed in the matdyn.modes file, then > yes they are not orthogonal as those are actually eigendisplacements and > not the eigenvectors of the dynamical matrix. But if you multiply the > eigendisplacements of each atom by the square root of the corresponding > atomic mass, then the resulting eigenvectors will be orthogonal. > > Hope that helps. > Sridhar > > > On Wed, Aug 27, 2014 at 8:53 PM, Florian Altvater <altvater at berkeley.edu > <http://pwscf.org/mailman/listinfo/pw_forum>> > wrote: > > >/ Hi, > />/ I calculated the phonons for an isolated naphthalene molecule as well as > />/ naphthalene crystal at q = (0,0,0). After diagonalizing the dynamical > />/ matrix with dynmat.x and enforcing the appropriate acoustic sum rule - > />/ "zero-dim" and "crystal" respectively - I get 6 and 3 acoustic modes > />/ with frequencies of 0 cm^-1. So far so good. > />/ > />/ If I use the displacements printed to dynmat.out (filout) and calculate > />/ the various scalar products of the modes, i.e. c_ij = m_i . m_j, I find > />/ that they are normalized but not orthogonal (c_ij != 0 for some i != j). > />/ Some of the overlaps c_ij are as high as 0.75, so it is not just > />/ numerical noise. By trying to find more information in the source files, > />/ I discovered that matdyn.x prints two files, flvec and fleig, where > />/ flvec prints non-orthogonal normalized vectors. So I thought that this > />/ could be the issue, as it is not clearly stated if dynmat prints > />/ eigenvectors or normalized modes like flvec. However, multiplying with > />/ the masses didn't help either. > />/ > />/ Is dynmat.x not printing eigenvectors? If it is/should, what could be > />/ the problem here? How would I debug the problem? > />/ > />/ Thanks so much for you help! > />/ Flo > />/ _______________________________________________ > />/ Pw_forum mailing list > />/ Pw_forum at pwscf.org <http://pwscf.org/mailman/listinfo/pw_forum> > />/ http://pwscf.org/mailman/listinfo/pw_forum > /> -------------- next part -------------- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20140828/222c49d6/attachment.html
