In data 31 luglio 2009 alle ore 12:42:02, Paolo Giannozzi <giannozz at democritos.it> ha scritto: > the remaining 1.3 electron difference could well > be due to the finite cutoff.
That specific pseudopotential is lacking 1.3 electrons even if you test it alone, with a huge cutoff.. > Atomic charges and wavefunctions are > normalized, but when you project over a finite PW basis set, you > lose some normalization. Also note that the atomic charge is > typically the charge of the electronic configuration used in > pseudopotential generation, not necessarily a neutral state. Yes, and it has a net charge of +0.5 abs(e). Although, I see that the code takes valence charge (7.00) as the initial charge, not the sum of the orbital occupations (6.50). Nevertheless, if I construct a pseudopotential with valence charge = 6.5 (e.g. exactly equal to the original one), its initial charge is ~6.5, not 5.7. My guess is that its PP_RHOATOM section of the UPF file has some problem. In fact if I regenerate the new potential with the same grid as the old one (xmin=-4.0, dx=0.0625, rmax=60.0) I can replace the atomic rho of the old pseudo with the one of the new, getting the (almost) correct starting charge. Convergence is still a bit flaky at the beginning. Which is normal, due to renormalization of charge inflating 6.5 electron to 7.0 without reshaping them. If this would be enough to guarantee convergence in the more complex case, I cannot say without trying. cheers -- Lorenzo Paulatto SISSA & DEMOCRITOS (Trieste) phone: +39 040 3787 511 skype: paulatz www: http://people.sissa.it/~paulatto/ *** save italian brains *** http://saveitalianbrains.wordpress.com/
