From jkl@ccl.net Fri Aug 25 15:54:54 1995 Received: from bedrock.ccl.net for jkl@ccl.net by www.ccl.net (8.6.10/950822.1) id PAA05869; Fri, 25 Aug 1995 15:25:53 -0400 Received: from zinc.chem.ucalgary.ca for lohrenz@zinc.chem.ucalgary.ca by bedrock.ccl.net (8.6.10/950822.1) id PAA00491; Fri, 25 Aug 1995 15:25:48 -0400 Received: by zinc.chem.ucalgary.ca (AIX 3.2/UCB 5.64/4.03) id AA31915; Fri, 25 Aug 1995 12:21:51 -0600 From: lohrenz@zinc.chem.ucalgary.ca (John Lohrenz) Message-Id: <9508251821.AA31915@zinc.chem.ucalgary.ca> Subject: Re: DFT METHODS, COMPARISON To: CHEMISTRY@ccl.net (Everyone CCL) Date: Fri, 25 Aug 1995 12:21:50 -0600 (MDT) In-Reply-To: <199508251516.LAA27273@alchemy.chem.utoronto.ca> from "E. Lewars" at Aug 25, 95 11:16:43 am X-Mailer: ELM [version 2.4 PL23] Content-Type: text Content-Length: 2940 BLYP, B3LYP, Becke3LYP, LYP, .... Thorsten Koch wrote: > does anyone have any references about the performance and quality of the >diverse DFT-methods like LDSA, Becke, Becke3 ... ? >Up to now, this are only names to me and I'd like to know which method >should be preferred for certain purposes. we had this discussion before (see my summary earlier this year)... But what did we learn from it? Just a few comments: We should first of all not just refer to what is stated in the Gaussian manual. DFT was invented much earlier and there are many other programs (DMON, DeMOL, ADF, DeFT,... you name it) which have been applied for much longer time than GAUSSIAN/DFT. Unfortunately the latter somewhere stated that the Lee-Yang-Parr correction is the most reliable and since then almost everybody is using this as kind of a benchmark. There are other and to my knowledge sometimes (I explicitly say sometimes) more reliable functionals. Errol Lewars wrote: >There was a recent question about which DFT method is best. A quite extensive >compilation of geometries and relative energies from four methods (SVWN/6-31G, >SVWN/6-31G*, B3LYP/6-31G, and B3LYP/6-31G*) indicates the best of these is >B3LYP/6-31G*: *Practical Strategies for Electronic Structure Calculations*, >Warren J. Hehre, Wavefunction Inc., 1995 [sales@wavefun.com]. See also >J. E. Del Bene, W. B. Person and K. Szczepaniak, J. Phys. Chem., 99, 10705- >10707 (1995), especially refs 1f-h. This paper says that B3LYP fails for >H-bonded complexes with the 6-31G* basis. Second, if we really want to compare the quality of the functionals we should not mix the results of Local-Density Calculations (LDA) with Non-Local calculations (NL). Everybody should know that LDA has certain short-comings and compares to NL may like HF to MPx (no flames please). Thus the better performance of B3LYP with respect to SVWN doesn't say anything about the quality of B3LYP. Compare with Becke-Perdew, or the hybrid-methods and so on, and we will really learn something. Seiji Mori wrote: > Utility of DFT is fastly increasing in the application of DFT method . > Now Becke3LYP is known as the best method of DFT, but I think that the ^ok, who said that? Finally I'd like to state that one should always mention what kind of systems he is refering to. It is for example known, that the hybride- methods give excellent results for main-group (mainly C,H,O,N) molecules, but might as well fail for transtion metals. Cheers, John -- ========================================================================= Dr. John Lohrenz Dept. of Chemistry Phone: (403) 220 3232 University of Calgary FAX: (403) 289 9488 2500 University Drive, N.W. Calgary, Alberta, T2N 1N4 email: lohrenz@zinc.chem.ucalgary.ca Canada =========================================================================