On Jul 1, 2009, at 16:49 , Lorenzo Paulatto wrote: >> (2) pot | wfc_extrapolation >> please clarify if 'first_order', 'second_order' can be used in >> calculations other than 'md' or 'vc-md'. I used these options in >> 'relax' and 'vc-relax' without any errors, too. Of course, I >> don't know >> if these were really in effect. > > Both can be used for non-dynamical optimization (such as BFGS), > although > they don't strictly make very much sense. Nevertheless, they can help > speeding up convergence, especially when the minimization steps get > smaller. I would need to consult Paolo Giannozzi (who implemented the > extrapolation) to be 100% sure.
I didn't implemented the extrapolation, I just tried to clean up the mess that it used to be. First- and second-order extrapolation are devised to work in conjunction with a MD sequence of time steps. There is no point in using them for structural optimization, in my opinion. >> (3) ecfixed, qcutz and q2sigma >> in the original paper, the framework is an extension of PR >> dynamics to >> CP. Please clarify [...] please experiment a little bit and report what you found P. --- Paolo Giannozzi, Dept of Physics, University of Udine via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222