On Wed, 23 Jan 2008, X. Feng wrote: | Hi Andrei, | | Thanks a lot for the explainations and the example for Cr. | | SIESTA manuel mentioned that to generate a polarization | orbital one needs to use "P" in PAO.Basis.
Right | However, I saw | from Siesta website that in some of the optimized basis sets, | for example Ti and Mn, only a normal single zeta (4p) is used | as a polarization orbital. Then it is probably not a polarization orbital in a strict sence, derived from an s-orbital distorted by electric field. Rather, this is an orbital of p-angular symmetry, with radial dependence obtained or tuned I do not know how. | It is mentioned in the notes for these | bases that this normal single zeta is intended for polarization. What fo you mean by "intended for polarization"? Who intends to polarize whom? Can you give a reference? | Of course, this is not the definition of | a polarization orbital in the SIESTA manuel. No, it isn't. But from the point of view of Siesta, there are no fixed rules as for how to construct basis. You can draw you basis function by hand (, difitalize it) and go ahead with calculations. The question is, how good this basis would be. The "standard" scheme (DZP, Energy shift) is merely a reasonably good and fool-proof algorithm. | I don't know the reason | why people use a normal single zeta rather than following the manuel. Because they either believe that their tuned basis is more efficient, or they specifically want to use minimal basis, and they need to tune its functions as good as they can. | Is it better in some cases? Or, it is not a justified method? Depends on your definition of "better" and on cases in question. What do you want to achieve and at which price? The optimization of basis sets in any method using atom-type orbitals is a bit problematic. | By the way, do you have a tested basis sets for Cr, V? I did not tune them myself, if that's what you mean. However, I test basis sets (either standard, or borrowed from people) with respect to the tasks I have in mind for these elements. | I have pseudopotential for Cr and V without semicore states. | The magnetic moments are too high, so semicore states are necessary. This is a wrong argumentation. 3p semicore in Cr, V is not likely to affect your magnetic moment in a noticeable way, because 3p remains fully occupied. You'll see this if you compare total magnetic moments. So if your magnetic moments are too high it must be due to a different rason. Of course you may see that the figures for local magnetic moments, according to Mulliken populations, vary a bit as you vary basis functions, but this is due to ambigous definition of local magnetic moment, and has nothing to do with semicore as such. On the contrary, the presence of semicore may be quite important for comparing energies of different phases (e.g., magnetic phases), for getting right relaxation lattice constant (say, you may be 2-5% off the exp. volume with semicore and 15% off without it). | I'm currently testing PP's for Cr and V from a publication with | DZP, which will be generated by the method I learned today. | I'll be very grateful if you can send me PP's and basis, if you | have some. There is a PP + basis repository at the Siesta web site, and a lot of previous enquiries in the mailing list. Moreover, usually in their publications people describe what basis and PP they used. So if you want to profit from somebody's particular tuning for a particular system, you can write to people directly. You see, even for your Cr and V, the basis (and particularily whether you need semicore or not) depends so much on what you need (band sructure? phase diagram? phonons? transport?) and in which systems (bulk metals? clusters? organometallic molecules?) Good luck Andrei +-- Dr. Andrei Postnikov ---- Tel. +33-387315873 ----- mobile +33-666784053 ---+ | Paul Verlaine University - Institute de Physique Electronique et Chimie, | | Laboratoire de Physique des Milieux Denses, 1 Bd Arago, F-57078 Metz, France | +-- [EMAIL PROTECTED] ------ http://www.home.uni-osnabrueck.de/apostnik/ --+
