Top Page > Browsing

Log Mesh of PAO Date: 2009/11/18 02:28 Name: Liao Chen   <liaochen@mail.ustc.edu.cn> Dear Prof.T.Ozaki and OPENMX users, I find that the pseudo-atomic basis functions are generated on Log-mesh, which has form r(i)=exp(xmin + i*dx), and that xmin are -7 for all elements while dx and number of mesh grids are different for each elements. I wonder: 1. why xmin are taken as -7? 2. how to determine dx and number of mesh grids for each element. Another Question: As far as I know, evaluating two-centre integral in reciprocal space needs to calculate spherical Bessel transform(SBT): g(k) = int r**2 jl(r) f(r) dr I performed SBT using Method proposed by Talman, which calculates SBT on log mesh kappa = ln(k) and the array of values returned also on log mesh: k(i) = exp(kappa_min + i * dkappa); Constraints are dkappa = dx. How to choose kappa_min in order to continue calculation using values of SBT accurately? I think it should be not too large or too small. Hope for your advice. Sincerely, Liao Chen

Page: [1]

Re: Log Mesh of PAO ( No.1 ) Date: 2009/12/01 13:19 Name: Masayuki Toyoda  <m-toyoda@jaist.ac.jp> Hi, As for the latter question, I can give you some infomation. I'm working on computation of two-electron integrals and also having difficulities in finding optimum parameters for SBT. Basically, I choose parameters simply by trial and error. Note that errors can be estimated by performinig direct and inverse SBT. From my experience, when using N=1024 points, rho0 = kappa0 = -10 (where r[i]=exp(rho0+drho*i) and k[j]=exp(kappa0+dkappa*j)) seems to yield good results. My articles are posted on the following page, a part of which relates to this issue and I hope helps you: http://www.openmx-square.org/exx/index.html Best regards, Masayuki Toyoda RCIS/JAIST

Page: [1]