From jeremy@med.su.oz.au Mon Sep 2 05:52:38 1996 Received: from blackburn.med.su.oz.au for jeremy@med.su.oz.au by www.ccl.net (8.7.5/950822.1) id FAA11616; Mon, 2 Sep 1996 05:11:31 -0400 (EDT) Received: (jeremy@localhost) by blackburn.med.su.oz.au (8.6.11/8.6.4) id TAA11126 for CHEMISTRY@www.ccl.net Mon, 2 Sep 1996 19:10:48 +1000 From: Jeremy R Greenwood Message-Id: <199609020910.TAA11126@blackburn.med.su.oz.au> Subject: summary: compound methods / G2 theory To: CHEMISTRY@www.ccl.net Date: Mon, 2 Sep 1996 19:10:48 +1000 (EST) X-Mailer: ELM [version 2.4 PL23] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Greetings all, I recently asked a question about cheaper compound methods than G2(MP2) and was inundated with excellent advice. Thanks to all who replied; a short trip to the library and I'd found a good solution within a day. Suggestions were generally of the following nature: * Try DFT or hybrid DFT methods instead * Try CBS methods instead * Try simplifying the problem by only performing high level correlation calculations on the most important part of the molecules * Try the G2(MP2,SVP) method of Smith, Radom, et al. The solution I've chosen is a modification of the latter; G2 energies include an empirical, basis-set-specific deltaE-HLC term, which is a linear combination of the number of alpha and beta electrons. Since all my systems have the same number of these, and since I'm most interested in relative energies, this term cancels, and I'm free to use the best basis sets I can afford. G2(MP2) = ZPE*.8929 + HLC(G2) + QCISD(T)/6-311g(d,p) + MP2/6-311+G(3df,2p) - MP2/6-311g(d,p) // MP2/6-31g(d) G2(MP2,SVP) = ZPE*.8929 + HLC(G2,SVP) + QCISD(T)/6-31g(d) + MP2/6-311+G(3df,2p) - MP2/6-31g(d) // MP2/6-31g(d) my method of choice: deltaE = delta(ZPE*.8929 + QCISD(T)/6-31g(d,p) + MP2/6-311++G(3df,3pd) - MP2/6-31g(d,p) ) // MP2/6-31g(d,p) Jeremy ****************************************************************************** Original Question: I'm attempting to calculate high accuracy energies for a series of species, and would like to use a compound method. G2 energies, or failing that G2(MP2) would be the obvious method of choice. Unfortunately, G2 is just out of reach; QCISD(T)/6-311G(d,p) is too expensive, as is MP4. However, QCISD(T)/6-31G(d,p) is not, nor are MP2 calculations. Questions: Are there somewhat less expensive compound methods which are recommended? Would substituting QCISD(T)/6-31G(d,p) for QCISD(T)/6-311G(d,p) make a complete nonsense of G2(MP2)? Is delta-E-HLC specific to the G-theory basis sets and therefore not portable? I'm considering the following simple alternative: at MP2/6-31g(d,p) geom E = (HF)ZPE*0.9 + MP2/6-311G++(3df,3pd) + QCISD(T)/6-31G(d,p) - MP2/6-31g(d,p) Thanks, Jeremy ***************************************************************************** Answers: >From KGRAFTON@aardvark.ucs.ou.edu Tue Aug 27 23:35:27 1996 In regards to your question about a less expensive method to calculate molecular energies, you might care to check out our recent paper: J. Phys. Chem. v100, p10083. Here we calculated the energy differences (electron affinites) between a number of quinones and their respective radical anions. We compare several methods and find that the best is B3LYP hybrid HF/DF method with a 6-311G(dp) or 6-311G(3dp) basis set. We found that the G2 method, which you mention specifically, gives very bad results for parabenzoquinone. I would very much like to hear about you work as it progresses. Sincerely Anthony K. Grafton Department of Chemistry and Biochemistry University of Oklahoma Norman, OK >From KGRAFTON@aardvark.ucs.ou.edu Thu Aug 29 00:51:15 1996 Whoops. In the mail I sent you earlier, I said we had tried G2 on quinone energies. That was a bit wrong. We sis try it, but did not have the computing resources to complete the jobs (which includes an 8 processor IBM/SP2 with 512MB/processor). The method that we did use and get poor results from was the CBS-4 calculation. Sorry about the confusion. Anthony Grafton Department of Chemistry University of Oklahoma -- From: "Frederick R. Bennett" Subject: CCL:G:G2 with reduced basis sets Hi, someone was asking about attempting G2 calculations with reduced basis sets. Well I found a paper addressing this issue exactly Gaussian-2 (G2) theory: Reduced basis set requirements, Curtis, Redfern, Smith, and Radom: JCP, 104 (13) 5184 Hope this helps Ciao =============================================================================== Frederick R. Bennett Papernet Address: Institut Fur anorganische, analytische und physikalische Chemie Freiestrasse 3 CH-3000 Bern 9 Switzerland Mouthnet Address: [41] (031) 631 4231 Faxnet Address [41] (031) 631 3994 Internet Address: bennett@ubeclu.unibe.ch =============================================================================== From: Anthony P Scott (also thanks to James W Gauld ) Jeremy, I understand that James has already sent you two references for the G2(MP2,SVP) method. Here are the two leading references for G2(MP2,SVP). Smith, B. J. and Radom, L. J. Phys. Chem. 99 (1995) 6468. - where it is first introduced Curtiss, L. A., Redfern, P. C., Smith, B. J. and Radom, L. J. Chem. Phys. 104 (1996) 5148. - a detailed description of G2(MP2,SVP), including the new HLC to use. In your email you didn't say why the qcisd(t)/6-311g** calculataions are not possible. Is it disk space or cpu time that limits you? If you are trying to do these calculations with Gaussian you are really beating your head against a brick wall. Two other programs are substantially more efficient that G94 for these type of calculations. Molpro96 ( limited to closed shell however) or ACES2 will result in (usually) much better cpu times that G94. They both make explicit use of symmetry, however ACES2 in C1 symmetry will often take longer than gaussian!! If it is disk space that is the problem then try to get more! Hope this is of some use. Kind Regards, Tony -- >From johnm@lorentzian.com Wed Aug 28 00:57:14 1996 If you have access to Gaussian 94, you might consider the CBS model chemistries - see J. Chem. Phys. 104 (1996) 2598-2619 for details. John Montgomery Lorentzian, Inc. -- >From aefrisch@lorentzian.com Wed Aug 28 13:45:26 1996 The Complete Basis Set methods (CBS) of Petersson and coworkers are an attractive alternative, especially CBS-4 and CBS-Q. They are also compound models. They are automated in G94 and we discuss them in chapter 7 of Exploring Chemistry. -- From: Elmer Valderrama Just to call your attention on the work "Combining MCSCF and DFT for dyn. electron correlation" N.O.J. Malcom and J.J.W.McDouall, J. Phys. Chem 100(24) 10131-10134 (1996) where an interesting proposal is advanced concerning compound methods. This possibility (MCSCF (or CASSCF) + DFT) was probably first noted by Clementi in 1974, but a way to do it was not advanced (see ref in work cited aboved) Elmer -- From: bernd@rs5.thch.uni-bonn.de (Bernd Engels) hello Jeremy R Greenwood asked for cheaper methods than QCISD or MP4. I would try to use DFT methods. There are also some extensions to the normal G2 which uses the DFT. bernd engels bernd@rs5.thch.uni-bonn.de -- Reply-To: qiang@euch4e.chem.emory.edu I'm not particularly experienced in this issue, but there are several related works in our group (K. Morokuma), which might provide some useful info. First of all, there are always those scaling approaches, which works rather well in many cases, including transition metal chemistry. One of them is the so-called PCI-80, proposed by M. Svensson and P. Sigbahn. Secondly, some of our group memebrs are working on the so-called IMOMO method, which is basically treating the active part with high accuracy, and the rest with relatively low accuracy. Indeed, in most chemical reactions, correlation effects are rather localized, the inactive part are either correlation-insensitive, or effects simply cancels out if one considers the relative reaction profile. In those cases, one can do really high level calculations such as G2 on the active site. For example, people in our group did some benchmark calculations on a series D-A reactions( with large substituents) and found really good agreement with several gas phase exp. results. They've also done some systematic studies on the S/T transition energies for a series of large molecules,and again found promising results. I guess the merit of those approach is, one does't really need any new code to do the calculation, (provided that u have a decent geometry, people in our group are also doing geom. optimization using IMOMO/IMOMM). All u have to do is, do a few single points calc. at different levels, and put them together. Sophisticated mathmatics doesn't always mean better results. (although personally I like rigorous math better:-) Any way, that's what I can think of and what I'm relatively familar with. If u r interested, u can contact me or froese@euch4g.chem.emory.edu for some references. Hope this will help in some way.. ______________________________________________________________ Qiang Cui Dept. of Chem. Emory Univ. 508 Webster Dr. Apt.#2 Atlanta, GA 30322. Decatur, GA 30033. (404)-727-2381 (404)-636-6149 http://tswww.cc.emory.edu/~qcui ______________________________________________________________ From tp@elptrs7.rug.ac.be Mon Sep 2 11:52:42 1996 Received: from elptrs7.rug.ac.be for tp@elptrs7.rug.ac.be by www.ccl.net (8.7.5/950822.1) id LAA13283; Mon, 2 Sep 1996 11:06:08 -0400 (EDT) Received: by elptrs7.rug.ac.be (AIX 4.1/UCB 5.64/4.03) id AA25686; Mon, 2 Sep 1996 17:07:19 +0200 Date: Mon, 2 Sep 1996 17:07:18 +0200 (DFT) From: "Park, Tae-Yun" To: Computational Chemistry List Cc: Ernie Chamot Subject: Successful reproduction with low level ab-initio calculation. Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Dear all, Several weeks ago, I posted the problems on calculating of heat of formation data using mopac93. After I posted this message, I received several suggestions how to calculate the heat of formation data as accurate as possible. I picked up one of the method suggested, which was based upon ab initio calculation. My final result was that the ab initio calculation was able to predict the experimental data very successfully, even if I didn't go up to very high level of basis. Therefore, my conclusion is that the ab initio method, not semiempirical one, should be used for the calculation heat of formation, at least for carbenium ions, if one try to get the same trends as experimental data. I truely thanks for all the CCL members who replied to my earlier question on calculating heat of formation for carbenium ions, and especially I'd like to express my deep thanks to Dr. Ernie Chamot who suggested me a correct way to solve the problem. Sincerely, Park, TAE-YUN =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= State University of Ghent Laboratorium voor Petrochemische Techniek Krijgslaan 281, Blok S5 9000 Gent, Belgium TEL:+(32)-0(9)-264-4527 FAX:+(32)-0(9)-264-4999 e-mail: tp@elptrs7.rug.ac.be =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= P.S. The ab initio calculation has been performed by GAMESS US version, and my lowest basis level that gives the successful reproduction of the experimental data was Pople's STO-NG minimal basis set. From owner-chemistry@ccl.net Mon Sep 2 17:52:43 1996 Received: from bedrock.ccl.net for owner-chemistry@ccl.net by www.ccl.net (8.7.5/950822.1) id RAA14679; Mon, 2 Sep 1996 17:42:15 -0400 (EDT) Received: from taurus1 for vladimir@taurus1.chapingo.mx by bedrock.ccl.net (8.7.5/950822.1) id RAA12244; Mon, 2 Sep 1996 17:41:18 -0400 (EDT) Received: by taurus1 (5.x/SMI-SVR4) id AA03584; Mon, 2 Sep 1996 16:33:29 -0600 Date: Mon, 2 Sep 1996 16:33:28 -0600 (CST) From: Vladimir Vesibe Naud To: CHEMISTRY@ccl.net Subject: HELP! Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Content-Transfer-Encoding: QUOTED-PRINTABLE =09Hello!, I'm a student, Mi name=B4s Vladimir Uribe, I would like know how make, or how I can use yhe program about the chem.ucla in the order of NETUSERS, My problem is: =09I need a models of atoms, or a periodic.zip for my computer. =09 =09I=B4m not well in English, sorry. =09=09=09=09See you soon =20