From 9702029k "-at-" lv.levels.unisa.edu.au Mon Jul 13 20:55:22 1998 Received: from zorba.levels.unisa.edu.au (SYSTEM -AatT- Zorba.levels.unisa.edu.au [130.220.17.9]) by www.ccl.net (8.8.3/8.8.6/OSC/CCL 1.0) with ESMTP id UAA10028 Mon, 13 Jul 1998 20:55:20 -0400 (EDT) Received: from xiy205359.unisa.edu.au ("port 4854" %! at !% XIY205359.Levels.UniSA.Edu.Au) by Levels.UniSA.Edu.Au (PMDF V5.1-10 #27872) with SMTP id <01IZDYJ5NCY69878XC &$at$& Levels.UniSA.Edu.Au> for chemistry -x- at -x- www.ccl.net; Tue, 14 Jul 1998 10:25:16 +0930 Received: by xiy205359.unisa.edu.au with Microsoft Mail id <01BDAF12.DE23C440 %-% at %-% xiy205359.unisa.edu.au>; Tue, 14 Jul 1998 10:33:40 +0930 Date: Tue, 14 Jul 1998 10:33:37 +0930 From: Josh Bowden <9702029k ( ( at ) ) lv.levels.unisa.edu.au> Subject: Summary : Mixing basis sets. To: "'chemistry;at;www.ccl.net'" Message-id: <01BDAF12.DE23C440 ( ( at ) ) xiy205359.unisa.edu.au> MIME-version: 1.0 Content-type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by www.ccl.net id UAA10028 Hello all, Thanks to you who replied. I wrote : >As a continuation of recent discussion on this list about 3-21G vs 6-31G*, can I ask if anyone >has information on mixing basis sets in a calculation. Programs such as CRYSTAL 95 and >HyperChem (probably others as well) allow you to apply different basis to individual atoms. If >the lack of polarization functions ( as stated by cory (- at -) chem.ucalgary) is a problem can you >not apply a bigger basis set to 'problem' atoms and reduce the basis of others? >Has anyone looked at this systematically and is there any referencs on this topic? >Are there any problems in this metholodgy (does it have any major downfalls)? Dr. Cory C. Pye (cory at.at zinc.chem.ucalgary.ca) wrote : >The biggest problem with mixing basis sets is that you run the risk of >obtaining an 'unbalanced' overall basis set. This basically means that the >electronic distribution, compared to the true distribution, is lopsided. >A way to think of this is that if you have a homonuclear diatomic with a >minimal basis set on one atom (A) and a triple zeta valence + pol + dif on >the other atom (B), then, because B has more functions over which to distribute >electron density, you will basically have a dipole moment for this homonuclear >diatomic (more charge on B than A), which is physically unrealistic. > >One should therefore always try to balance the basis set by using similar >types on different atoms. For example (from my own work) in >J. Org. Chem., 60, 2328-9 (1995), J. Org. Chem., 63, 105-112 (1998), and >J. Phys. Chem B, 102, 3564-3573 (1998), we did not have available a >6-31G* basis set for atoms of interest beyond Ar (Br,I,Se,Te,As,Sb,Ge,Sn(x2);Cd) >so we spliced in a Huzinaga basis set (from his book --- >take the minimal basis set with the highest # of primitives in the innermost >shell; split the valence shell (n) to ( (n-1)(1) ); and add a polarization >function provided) to give a SVP type basis set (like 6-31G*). > >On a related note, there is a basis set called 3-21G(N) (by Hehre, I believe) >where the 3-21G* basis set is augmented by nitrogen polarization functions. Dr. Georg Schreckenbach (schrecke at.at t12.lanl.gov) wrote : >In any case, I think people have used such "locally dense" basis sets a >lot. Typically, you would use the high level basis set on, e.g., the metal >atom AND its nearest neighbors, and smaller basis sets elsewhere. One >reference is a review by Don Chesnut: D. B. Chesnut, in Annual Reports on >NMR Spectroscopy Vol. 29, Academic Press 1994, p. 71. He uses such locally >dense basis sets for NMR calculations, and I believe that he has newer >papers on this as well. > The more recent NMR review by de Dios has similar information (A. >C. de Dios, Prog. NMR Spectros. 29 (1996), 229.) You will find further >references in these reviews. > >Furthermore, and this is slightly off from your question, the idea of using >a higher level of theory for the interesting atoms, and a lower level >elsewhere, has been pursued in the QM/MM methods: In this case, you do >whatever quantum mechanical method you need at the reaction center, and >molecular mechanics at the bulky ligands or solvent or so. The Morokuma >version of this ("ONIOM") is present in GAUSSIAN98, as far as I know. >Others have done such things as well, e.g. T.K.Woo/T.Ziegler. > >Best regards, Georg Robert J. Zellmer (rzellmer { *at * } chemistry.ohio-state.edu) wrote : >I think most programs that allow you to put in general basis sets would be >able to do this. I know GAMESS is capable of this and I have done it using >GAMESS. I don't know of any systematic study of doing this, at least not one >that has been published. I know a few years back when I was doing some >calculations on perfluoroethers in conjuction with someone at Dupont they had >determined that d-type polarization functions could be left off the Fluorines >w/o any real affect on the calculations, but they were still needed on the >Oxygen atoms. This might be mentioned in some of the papers published by the >folks at Dupont and I can check if you like. > Stefan Konietzny (konietz { *at * } chemie.uni-kl.de) wrote : >Hi! > >I dont know which program you use, but in GAUSSIAN it is possible. > >Try the GEN Keyword instead of of general basis set. > ># hf/gen fopt .... > >After the molecule specification you have to apply the basis to the atoms. >You can do this like this > >H C 0 >6-31+g(d) >**** > >This means every H and C atom has 6-31+g(d). > >But if you use the numbers of the atoms, you can apply different basis >sets to different H (or C, or anything else) atom. > ># hf/gen fopt name=konietz > >h2o > >0 1 >h >o 1 1.1 >h 2 1.1 h 108. > >2 0 >6-31g >**** >1 3 0 >3-21g >**** > >Hope this helps. > Thanks again, Josh Josh Bowden Ian Wark Research Institute University of South Australia E-mail : 9702029k %! at !% lv.levels.unisa.edu.au