Dear Eduardo (and José), I don't think I can imagine some way in which incomplete scf convergence could cause the deviation, since the DM is supposed to be normalized at each step, correct me if I'm wrong... The electronic temperature might be able to cause such thing, but I'm definitely getting these small deviations even with it set to zero. I converge SCF to 1.d-5 by density. So, should I still blame the normalization?
2008/4/3, Eduardo Anglada <[EMAIL PROTECTED]>: > > > Hi! > > This is José Soler answer: > > "Without discarding the effect of imperfect density matrix > normalization, it seems to me that the most likely cause > of that small deviation from 1 is an incomplete scf > convergence, or an excessive electronic temperature." > > Regards, > Eduardo > > > On 02/04/2008, at 11:59, Vasilii Artyukhov wrote: > > That's how many spin-up electrons than spin-down there are in your system. > > It seems to me that these figures long after comma should probably be > discarded, they must be due to slightly incomplete normalization of the > density. Hence, a question to the developers: > > Is the above correct & how many significant figures should we keep in this > number & how small is the effect of such normalization on the accuracy? > > 2008/4/2, Jiaye, Li <[EMAIL PROTECTED]>: > > > > Dear Users > > > > This is the copied line from the output file of siesta: > > > > *siesta: Total spin polarization (Qup-Qdown) = 1.000035* > > > > What does "Qup-Qdown' mean? 1.000035 is the value of unpaired electrons? > > Thanks in advance! > > > > ps. I performed calculations with Pt metal, SpinPolarized flag was set > > .true. > > > > -- > > Sincerely > > > > James Li > > > > > > > > > > >
