On Wednesday 31 May 2006 12:23, Andrea Marini wrote: > So, unless I am doing something wrong in my calculations > (perfectly possible) there could be something wrong in the code.
there is. If you perform the calculation of the Delta V and the calculation of electron-phonon coefficients in two separate steps, you get correct results (or, more exactly, results that are not obviously wrong). If however you perform the two calculations in a single step, you get incorrect results. This is true for all versions from 2.1 to 3.0 I think: version 2.0 didn't allow the single-step calculation, version 3.1 is correct in this respect (no warranty that it is correct in other respects, though). I think I know why this happen: the small group of q and the crystal group are stored in the same set of variables, recalculated and overwritten 2n+1 times. Since with the old algorithm all serious electron-phonon calculations needed a two-step calculation (phonons and Delta V with a smaller k-point grid, electron-phonon coefficients with a much larger grid) I hope that nobody published wrong results obtained with this code. As a side note: for some reason the Methefessel-Paxton smearing (ngauss1=1, hardcoded in PH/elphon.f90) was used, but I don't see any good reason to prefer it to simple gaussian (ngauss1=0). Other questions about the symmetrization requires some more thought. Thank you Andrea for noticing the above problem. Paolo -- Paolo Giannozzi e-mail: giannozz at nest.sns.it Scuola Normale Superiore Phone: +39/050-509876, Fax:-563513 Piazza dei Cavalieri 7 I-56126 Pisa, Italy
