Jiayu Dai wrote: > Dear users, > > I have a question about the sample points in the Brillouin zone. We usaually > use > the MP point in our calculations, and it works well and fast in cubic crystal. > However, for disordered materials or liquids, we usually use Gamma point only, > such as point in CPMD. My question is how do we know the Gamma point is better > than the Baldereschi mean value point, or on the contrary??Which one gives > faster > convergence? > > Thanks a lot. > > Jiayu
My personal bias is that the Baldereschi point always works better, and that in an orthormombic cell using 1/4 1/4 1/4 (relative coordinates) is almost as good as using the full 2x2x2 monkhorst pack mesh, shifted, and, for a metal, much much better than gamma. Note that 1) if you use only 1/4 1/4 1/4 the forces will not be exactly symmetric even for a symmetric system 2) if you do not use nosym=.true. pwscf will in any case increase the k-points to recover symmetry 3) the wavefunction at 1/4 1/4 1/4 are complex, while at gamma can be chosen to be real, making the code twise as fast (I think) and the memory half. That's why e.g. the cp code uses gamma, even if it is actually the worst choice possible from the point of view of the sampling. nicola -- --------------------------------------------------------------------- Prof Nicola Marzari Department of Materials Science and Engineering 13-5066 MIT 77 Massachusetts Avenue Cambridge MA 02139-4307 USA tel 617.4522758 fax 2586534 marzari at mit.edu http://quasiamore.mit.edu
