Please see whether this bug-report has corrected your problems. The bug-report highlights that this is only a matter of printing the initial spin configuration, not how the actual spin-configuration is "seen" by siesta. https://bugs.launchpad.net/siesta/+bug/1648053
2017-03-06 17:27 GMT+01:00 Pouya Partovi-Azar <parto...@zedat.fu-berlin.de>: > Hello everyone, > > I’m doing a spin polarized calculation on a system consisting of hydrogen, > carbon an nitrogen. I just encountered a strange behavior when comparing > the outputs of SIESTA v3.2 and v4.1. I have attached the input files used > for the calculations and the outputs generated by the code. > > If I set “DM.InitSpinAF .false.” I expect to have 18 in "initdm: Initial > spin polarization (Qup-Qdown)” according to the order in the atomic > coordinates block (+3+5*(+2)+5*(+1)=18). I get the number 18 in v3.2 output > file, but not in v4.1, where I get 5! > > Now, when I set “DM.InitSpinAF .false.” I would expect to get 2 in > "initdm: Initial spin polarization (Qup-Qdown)” (+3-2+2-2+2-2+1-1+1-1+1=2), > which again I get correctly in v3.2. However, in v4.1 I get 1! > > Theses inconsistencies do not affect the final spin polarization of the > system I have considered in these calculations, but there are, of course, > systems where different initial spin configurations result in different > final spins, and consequently different band structures, density of states, > and transport properties for different spins. > > Does anyone know the main reason behind this difference in two SIESTA > versions? How to solve it? By the way, the corresponding manuals do not > specify anything about possible differences in two versions. > > Thanks in advance, > Pouya > > PS: In the case of v4.1, I sometimes even get fractional numbers as > initial spin configurations! > > > > -- Kind regards Nick