From owner-chemistry@ccl.net Sat Jul 8 00:07:52 1995 Received: from utsc.s.u-tokyo.ac.jp for smori@utsc.s.u-tokyo.ac.jp by www.ccl.net (8.6.10/930601.1506) id XAA14399; Fri, 7 Jul 1995 23:57:02 -0400 Received: by utsc.s.u-tokyo.ac.jp (5.67+1.6W/TISN-1.3/R2) id AA06997; Sat, 8 Jul 95 12:50:01 JST Date: Sat, 8 Jul 95 12:50:01 JST From: Seiji Mori(nakamura lab) Message-Id: <9507080350.AA06997@utsc.s.u-tokyo.ac.jp> To: chemistry@ccl.net Subject: CCL:Re: ab-initio and TM's J. Pollard wrote: >I was wondering if get peoples opinion on the best (or most accepted) >method to perform ab-initio calculations with second and third row TM >complexes. I am familiar with all the semi-empirical approaches but am >not with any ab-initio. What basis sets work best? Relativistic effects? >Thank you and I look forward to responses. J. Pollard Dear Pollard; I am investigating the reaction pathway in the reaction with organometallics with ab initio method. As you know, recently, Semiemperical MNDO/d and PM3(tm) hamiltonian is work well in the TM complex, for example Cr,Pt(II), Fe and so on. Basically, Pople and Hehre's 3-21G and 6-31G* basis sets( combined Hartree-Fock theory) provide good results for main group element such as C,H,N,O,Cl,Li,Mg..... I devided the section ,first row transition metals , heavier transition metals: (1) First-row transiiton metals: 3-21G basis set is also avalable for Sc-Zn , however, especially in the first transition metals, the description of d orbital(3d) is cheep because 3d fitting by Gaussian-type function as compared to Slater type function is difficult. So, we use the rich basis sets contained basis functions, for example, Wachters basis set with Hay's d function :Wachters, et al.JCP, 1970, 52,1033. Hay, JCP, 66,4377(1977) The problem of these basis sets is very huge, so the costs of computation is higher than 3-21G basis set .Number of basis functions of Zn (for example) is 27 at the 3-21G basis set and 39 the Wachters+Hay basis set . As to calculation of Zn atom,computer time is 42.1 sec by Gaussian94 in my Workstation in former,53.3sec in latter. Then we also use the effctive core potential method(ECP), that is, the core orbital is fitting by potential function, the valence orbital are used basis functions. The most famous ECP is editted by Hay&Wadt,Hay and Wadt,JCP,82,270,299(1982) These functions are relavistic ECP. In the Moller-Plesset perturbation theory combined with these basis set ,whether you use all electron basis sets or basis sets including ECP, the result is good agreement with the experiment ,In the Hartree-Fock method, the results sometimes are poorer than MP2 theory. In the Fe,Ni compunds (the system including the elements we have to take high-spin state into account),even if we use MP2 theory, the results are sometimes poor.Higher-cost CI(configration-interaction) method may be better. As to the structual reliability of methylmetalchloride(M= Ti, Ge,Sn,Pb) compounds , you can see: G. Frenking, Marburg, reported in J. Comp.Chem.13,919,935(1992). (2) Heavier transition metals And The element bigger than 4-th row element, the relavistic effect is affected to the chemical behaviors.We use ECP such as Hay&Wadt ECP because in the ECP me thod, we can take the relavistic effect into account. Recently, Stevens also reported the relavistic ECP in Can.J.Chem. 70,612(1992) (K-La,Hf-Rn) or JCP, 98, 5555(1993)(Ce-Lu) about from third-row to Lanthanide elements, however, I don't know papers about the reliability of Stevens ECP. Sincerely yours: Seiji Mori Department of Chemistry The University of Tokyo email:smori@chem.s.u-tokyo.ac.jp From owner-chemistry@ccl.net Sat Jul 8 22:37:28 1995 Received: from gusun.acc.georgetown.edu for choic@gusun.acc.georgetown.edu by www.ccl.net (8.6.10/930601.1506) id WAA23583; Sat, 8 Jul 1995 22:27:10 -0400 Received: by gusun.acc.georgetown.edu; (5.0/1.1.8.2/6Jan9518:19pm) id AA04337; Sat, 8 Jul 1995 22:25:52 +0500 Date: Sat, 8 Jul 1995 22:25:51 -0400 (EDT) From: Cheol Choi To: CHEMISTRY@ccl.net Subject: Calc. of Frequency ? Message-Id: Mime-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII content-length: 906 Hi! Everyone. I'm trying to calculate the frequencies using Cartesian Force Field which comes from Gaussian output. If one use Cartesian Coordinate , the Wilson's G matrix is simply diagonal matrix of triple of each atomic mass. Finally, one get the following eigenvalue equation: F X = lamda G X where lamda is eigenvalue, F is force field, G is Wilson's Gmatrix and X is eigenvector(s). Since G matrix is diaginal matrix(in Cartesian Coordinate), one can easily get inverse of G. Then the equation is Inverse(G) F X =lamda X where the product of inverse(G) and F still symmetric matrix. Finally, if one diagonalize the inverse(G)*F,one will get eigenvalue (lamda). Now here is my problem. I used m dyne/Angstrom for F and amu for G matrix element. But I coundn't get the correct frequencies(eigenvalue). Please help my solving this problem. Thanks in advance. Cheol-Ho Choi Georgetown Univ.