Cherry, It is rather strange that you have a hard time to converge with Pulay and it gets easier with linear mixing. Have you considered using the "kick" instead? Try using DM.NumberKick with a value that is apropriate for your case (eg., the number of the step where you realize that the system starts to oscillate in dDmax. Also, in the very beginning, it might be good to use a low value for the mixing weight until the forces have gone down to something reasonable, such as around 0.8 eV/Ang or lower. Then you can restart the calculation with a higher value of beta. I guess that, if you really want to do what you have in mind, youĺl have to hack into the code...
Marcos > Dear All, > > I wonder if anyone know how to do a linear mixing for the first a few steps > then do the Pulay or Broyden mixing for the rest? I know you can write two > .fdf > files, the first one does linear, then the second one does Pulay using the > .DM > file. But in a relaxation process, you cannot do that for each step. I > found > the Pulay is particularly efficient, however sometimes the first several > mixings by Pulay can make it never converge... > > Many thanks, > > Cherry > > > > ____________________________________________________________________________________ > Yahoo! oneSearch: Finally, mobile search > that gives answers, not web links. > http://mobile.yahoo.com/mobileweb/onesearch?refer=1ONXIC > -- Dr. Marcos Verissimo Alves Post-Doctoral Fellow Condensed Matter and Statistical Physics Sector International Centre for Theoretical Physics Trieste, Italy -------- I have become so addicted to vi that I try to exit OpenOffice by typing :wq!
