Dear Dr. Postnikov, I followed your suggestions about flipping spin of a particular atom (the right electrode in my case) and it worked fine. Thank you very much for your help and for providing the DMtune code.
On Tue, Jan 6, 2015 at 9:36 PM, <[email protected]> wrote: > Dear Parsa Ravan, > > I don't have experience with TranSiesta but it seems that > your questions concern plain Siesta definitions, so I'll try. > > > > Dear TranSiesta users, > > I am trying to do transport calculations for a > > ferromagnetic/molecule/ferromagnetic junction. I have three questions > > about > > the job and I would be very grateful if someone could help me: > > > > 1. By simply setting "SpinPolarized .true." the code runs very > > nicely > > but > > when I initialize the spin using the block: > > > > %block DM.InitSpin > > 1 - > > %endblock DM.InitSpin > > > > the scf loop DOESN'T converge. I need to add the DM.InitSpin block > > because for the antiparralel configuration of the electrodes it is > > required > > that magnetic moment direction of the right electrode be reversed. > > I don't know what is your whole calculation in this example but > please mind that the DM.InitSpin addresses atom by atom > and not all atoms of given species. So your above definition says > set maximal spin-down on atom 1 and leave the rest (if any) > non-magnetic. If this is indeed what you want it should work. > Otherwise, having already the spin-up case converged, > you can invert the spin density as you wish, using my tool > http://www.home.uni-osnabrueck.de/apostnik/Software/DMtune.tar.gz > (which may require counting basis functions in tricky cases). > > > > 2. What are the numbers in the 4th column of kgrid block? > > They are displacements of the k-points... > It is a long story; you can check, e.g., > Moreno and Soler http://dx.doi.org/10.1103/PhysRevB.45.13891 > > > %block kgrid_Monkhorst_Pack > > 1 0 0 0.0 > > 0 1 0 0.0 > > 0 0 10 0.5 > > %endblock Kgrid_Monkhorst_Pack > > The userguide says that they usually must be set 0.0 or 0.5. Why? > > As unshifted mesh explicitly contains Gamma and other symmetric points, > whereas the shifted mesh misses the band extremities, but it contains > less inequivalent k-points than in unshifted mesh, at the same mesh > density. > Why 0.0 and 0.5 ? - the reduced k-points ought to be symmetric somehow, > in order to be compact. Siesta does not use much of symmetry of k-points, > but at least time reversal symmetry allows to half the total number > of k-points. With an arbitrary displacement (not 0.0 or 0.5) > this won't be possible. > > > > 3. For the antiparallel alignment of the electrodes I need to > > differentiate > > between those atoms of the right and left electrodes that are > included > > in the > > scattering region. I tried many forms of chemical species > > declarations, > > for > > example this one: > > > > NumberOfSpecies 5 > > %block Chemical_Species_Label > > 1 26 Fe1 > > 2 16 S > > 3 6 C > > 4 1 H > > 5 26 Fe2 > > %endblock Chemical_Species_Label > > > > but the code does not accept the syntax. > > Sure; what is expected is just chemical label (I think). > It makes sense to differentiate chemically similar atoms only if you > attribute to them different pseudopots or different bases. Otherwise > different spin configurations etc. are managed addressing > the individual atom index, not the species index. > > Best regards > > Andrei Postnikov > > >
