Dear Prof. Andrei Postnikov and Sonu, Thank you very much! I learnt a lot from the discussion. I think I know how to solve my problems.
Best wishes! Yours sincerely, Leila -----邮件原件----- 发件人: [email protected] [mailto:[email protected]] 发送时间: 2011年5月17日 17:41 收件人: [email protected] 主题: Re: [SIESTA-L] is this result reliable? > Respected Prof. Andrei Postnikov and all siesta users, > > I have few queries.. > >> >> However, you should know what you are doing, because the structure of >> wave atomic function inside the cutoff radius indeed starts to >> differ, and the error increases as you come closer. >> > >> Instead of paying to much attention to wave-functions, shouldn't we >> attend > to norm-conservation condition (to potentials generated by ATOM code > in siesta). Because satisfying norm-conservation (equal charge density > inside cutoff radii for both all electron and pseudo-wave functions) > implies that logarithmic derivatives and first energy derivatives of > logarithmic derivatives corresponding to all electron and pseudo-wave > functions agree at cutoff radii. > Which means better scattering properties of the ion core. > > or may be this condition is best satisfied only by best choice of > cut-off radii for each angular momentum components at particular > energy or at multiple energy references. > Sorry Sonu, we enter here a sensitive issues of pseudopotential construction; I don't think I am a right person to give you an enlightening answer... >> Moreover I think this is basically difficult to cover in Siesta, >> using the same fixed basis, a large range of varying interatomic >> distances with the same accuracy. > > since the basis is fixed, but if i increase the size of basis say DZ > --> DZP ---> TZP ---> TZP and diffuse functions, can i expect better > behaviour of basis over a range of interatomic distances, say from 0.9 > Ang to 2.0 Ang in system under consideration ? The short answer: yes you can. However, a priori you don't know how exactly. This is the standard problem of fixed bases. As you, say, change volume and your wave functions get more compressed, with the planewave basis, or with numerical adjustable basis (LMTO, LAPW) this will be taken care of automatically. Whereas with a fixed basis, you basis quality may be good either here, or there, but hardly everywhere. Unless your basis is "very" large, or you were very smart in constructing it. But then it becomes an issue of tests and human decision which (in)accuracy you are ready to tolerate, in your particular problem. >> > If not, shall I reduce the cutoff radii of my input file for the >> > pseudopotential generation? >> >> This is good for transferability but results in harder >> pseudopotential and introduce other kind of problems. >> >> will you please elaborate, what kind of problems in addition to > computational > time ? Weirdly behaving pseudopotential and pseudofunctions -> their long-range extension in reciprocal space -> high cutoff needed -> memory and time... Nothing more, indeed Best regards Andrei >> Thank you very much. > > With regards, > > Sonu Kumar > Phd Student > IITD >
