Trying Giussepe's advice on the system I am working on ATM, I could only produce small differences between plot_num=0 and 17, and negligible differences with increasing grid density. That would explain why I haven't noticed any issues yet. However, it is clear that, should I have to deal with a more problematic system, I would waste a lot of time to realize there is a problem at all, and I would spend a lot of foul language before finding a solution. So a big thank you to Axel, Giuseppe and Tiana for your advice.
Best regards, Jure On Mon, 27 Jul 2015 15:34:13 +0200 Giuseppe Mattioli <[email protected]> wrote: > Dear Jure and Axel > Axel is right. A smooth charge density including core electrons > yields best results, specially if the pseudopotential core region is > somehow entangled with the assignment of a given set of grid points > to the Bader's volume of a given atom. Notwithstanding (at least in > my experience): > > a) if you use PAW pseudopotentials you can extract a reconstructed > and smooth all-electron charge density by using the plot_num=17 > option of pp.x, together with a *very dense* real space grid (2 or 3 > times the grid you find in the pw.x output). In this way you > generally crunch reliable Bader's numbers. > > b) the same holds for *reasonably hard* NC pseudopotentials, where > any overlap between core regions and density minima is unlikely. > > Remember then to perform a last scf calculation after, e.g., a > standard "relax" calculation, using high values of nr1, nr2 and nr3 > (beware of memory usage...), before post-processing the charge > density to produce the cube file. _______________________________________________ Pw_forum mailing list [email protected] http://pwscf.org/mailman/listinfo/pw_forum
