On Thu, 2012-11-01 at 13:05 -0700, Yongsheng Zhang wrote: > > There is a simply question about the smooth step function [G^2 + qcutz * (1 > + erf ( (G^2 - ecfixed)/q2sigma) )]. This function is recommended in the MD > simulations, like vc-md and vc-cp. But why not in the vc-relax?
you can do that as well. Not sure you will gain much, though. I htink that the smooth step function is more useful in MD calculations where it prevents unphysical "jumps" during the run P. -- Paolo Giannozzi, IOM-Democritos and University of Udine, Italy