Dear all, Still waiting for some good advice on that topic!
Thank you all V Sent from my iPhone > On 27/ago/2014, at 10:56, Vincenzo Verdolino <vincenzo.verdolino at > gmail.com> wrote: > > Dear all, > > while doing my BO molecular dynamic simulations I realized that sometimes the > scf iteration doesn't converge in 100 or even 200 steps. My job approach is > to resubmit every n time steps restarting from the previous well completed > one. Of course if the previous one did not converged also the restart will > not and so on. i also realized as expected that if instead of restarting my > job from the previous restart file I do a 1 time step job > "restart=from_scratch" exporting manually last converged geometry and > temperature it dose converge and then I can do other several time steps > without any problem. I believe this is due to the extrapolation of wave > function and potential at the second order that i am using along the dynamics > > There should be a way to automatically force the pw.x to reoptimize from > scratch the wave function let's say every M steps. > > Could any of you give me some advice on that? > > thank you in advance > > Vincenzo