From chemistry-request@ccl.net Thu Jun 11 18:53:10 1992 Date: 11 Jun 1992 12:09:11 -0400 (EDT) From: "DR. DOUGLAS A. SMITH, UNIVERSITY OF TOLEDO" Subject: programming languages really are languages To: chemistry@ccl.net Status: RO Just thought I would forward this interesting note to the net, since it came to me directly. I like the debate - it really is interesting to see and hear what others think about this issue. I will have some more and final comments in a day or two. Doug Smith ========================================================================== From: IN%"ornitz@Kodak.COM" 11-JUN-1992 06:06:10.00 To: IN%"dsmith@uoft02.utoledo.edu" CC: Subj: Programming "languages" Return-path: Date: 10 Jun 1992 10:10:43 -0400 (EDT) From: ornitz@Kodak.COM (Barry Ornitz) Subject: Programming "languages" To: dsmith@uoft02.utoledo.edu Date: Thu, 11 Jun 1992 16:22 -0500 From: Katrina Werpetinski Subject: 3-D Integration grids (again) To: chemistry@ccl.net Status: RO [ Please pardon my ignorance. I'm just a lowly chemical engineer pretending to be a computational chemist. :) ] I'm struggling with a problem in an LCGTO-LDF-SCF code concerning the numerical integration for the fitting of the exchange potential. The present code uses 26 angular points, randomly rotated(1) at each of the Gauss-Legendre radial points. This is insufficient for the problems I'm interested in studying (dihedral angles and torsional energy barriers). I've bumped it up to 110 points thanks to a couple of people who sent me the code to generate the additional angular points and weights. This takes care of the accuracy problem, but the program now takes significantly longer to run. I know I can get rid of many of the angular points in the core region. I'm also thinking of changing the radial grid. Does anyone have a favorite method of forming a grid (ie which radial quadrature to use, what criterion to use for when to increase the # of angular points, how many angular points to use, etc) and any reasoning behind it? Katrina werpetin@ecs.umass.edu Delly(2) doesn't specify what order angular grids or what sort of radial grid Dmol uses. Andzelm and Wimmer(3) use up to 302 angular points (randomly rotated) and Gauss-Chebyshev quadrature for the radial points in DGauss. Becke(4) uses up to 194 angular points and Gauss-Chebyshev for the radial points. Fournier and DePristo(5) say deMon uses a grid like Becke's with rotation of the angular points. Dunlap(6) uses 26 angular points and a logaritmically increasing grid that starts at the first Herman-Skillman point. 1 R.S.Jones, J.W.Mintmire, and B.I.Dunlap, IJQCS 22, 77-84 (1988) 2 B.Delly, J. Chem. Phys. 92, 508-517 (1990) 3 J.Andzelm and E. Wimmer, J. Chem. Phys. 96, 1280-1303 (1992) 4 A.D.Becke, J. Chem. Phys. 88, 2547-2553 (1988) 5 Rene Fournier and Andrew E. DePristo, J. Chem. Phys. 96, 1183-1193 (1992) 6 Brett Dunlap, J. Phys. Chem 90, 5524-5529 (1986)