*From*: <qiang (+ at +) euch4e.chem.emory.edu>*Subject*: Re: CCL:G:SUMARY : G92 ECP and F functions*Date*: Thu, 14 Mar 1996 13:22:05 -0500 (EST)

Actually, I've extended the ECP integral routines to higher l(to h? if I remeber) values. Actually we have also implemented analytical hessian with ECP into Gaussian9x (x=2,4), which might be avaliable in the future. With this extension, u can calculate analytical hessian with f function (then of course gradient with f function) with DFT, HF, and MP2. (well, f function with MP2.., a little disk-demanding) However, the issume is, is f function crucial to ur geometry?! I guess one should always consider this before doing large calculations. For energetics, sure. But for geometry, according to our experince and literature, is not not critical. Of course, u know ur system the best. Any way, let's hope the code will be released in the near future. On Thu, 14 Mar 1996 hutschka (+ at +) quantix.u-strasbg.fr wrote: > > Hello, > > Two weeks ago I posted a question to the list. > Here is the sumary of the answers I got : > > My question was : > > I had troubles with trying to optimize a geometry > with g92 (RevE.2). > I'm trying to optimize a geometry at the HF level > using the Berny algorithm. > A RECP describe the core electrons of the metal and it contains > G,S-G,P-G,D-G,F-G components. > The basis set associated with this RECP is a [3s,3p,2d,2f] > GTF basis set. > I've no troubles to perform single point calculations but the > optimization lead to the following error message : > > ------ > RHF calculation of E > ------ > Compute integral first derivatives. > ... and contract with generalized density number 0. > Use density number 0. > RysSet: KIntrp= 2534 KCalc= 0 KAssym= 2170 > L702 exits ... SP integral derivatives will be done elsewhere. > Compute integral first derivatives. > Integral derivatives from FoFDir, PRISM(SPDF). > MinBra= 0 MaxBra= 3 MinLOS=-1 MaxLOS=-1 MinRaf= 0 MaxRaf= 3 MinLRy= 4. > IRaf= 0 NMat= 1 IRICut= 1 DoRegI=T DoRafI=F ISym2E=1 JSym2E=1. > Fock matrices symmetrized in FoFDir. > Use density number 0. > MaxL=3 MaxP=3 NDeriv=1 MaxT=77 > L709 cannot do derivatives in this basis. > Error termination in Lnk1e. > > ------ > > The error is arrising from the l709 link > I've looked at the code and it cames from the ECPGRD Subroutine > wich compute the derivatives of one electron integrals over effective > core potentials. > I looked at the manual and found nothing... > I've tried other tests and my believe is that it is not possible to > have F function describing the valence electrons of the metal > if you have a ECP (or RECP) for the core and if you intend to > optimize a geometry. > So my question is : > Does anybody know such limitations with g92 and ECP ? > More what are the limits in term of component (S,P,D,F) > of a basis set for an atom ?. > Does anybody know if this problem could be avoided with > g94 ?. > So any help or comment would be appreciated. > I will sumarize for the list. > > THE ANSWERS ARE : > > ***************************************************************** > >From Kazuo Teraishi : > ***************************************************************** > > I just encountered exactly the same problem with gaussian 94. > I think there are some limitations with ECP which are not mentioned > in the manual. Another limitation I came accross was the > calculation of analytical second derivatives with ECP. > > E-mail : JDA03546 (+ at +) niftyserve.or.jp > > ***************************************************************** > >From Dr. David Danovich > ***************************************************************** > > What you probably should do is to optimize by Fletcher-Powell (FP) > algorithm which does not require derivatives. > > E-mail : dodik (+ at +) yfaat.ch.huji.ac.il > > ****************************************************************** > >From Douglas J. Fox (Director of Technical Support) > ****************************************************************** > > The limit is on computing the derivatives of integrals using higher > than d functions with ECP's. This is not an aspect of Gaussian 92 that > was improved with G94. > > You can optimize these geometries with OPT=FP which requires only > energies and not gradients but it can be too slow if you have a large number > of degrees of freedom. > > E-mail : help (+ at +) gaussian.com > > ****************************************************************** > ****************************************************************** > > It's therefore clear that it's not possible to compute analytic > derivatives if the basis set of the metal atom describe by an ECP > contains higher than d functions. > The alternative way to optimize a geometry in this case is to use > the FP procedure and thus to choose as small as possible number of > degrees of freedom. > I thanks all those who answered . > Even those wich are not cited here because they send me MIME encoded > messages wich our mail did not understand. > I hope this sumary will be helpfull. > > HUTSCHKA Francois. > Laboratory of Quantum Chemistry > Universite Louis Pasteur > STRASBOURG - FRANCE > E-Mail : hutschka (+ at +) quantix.u-strasbg.fr > > > -------This is added Automatically by the Software-------- > -- Original Sender Envelope Address: hutschka (+ at +) quantix.u-strasbg.fr > -- Original Sender From: Address: hutschka (+ at +) quantix.u-strasbg.fr > CHEMISTRY (+ at +) www.ccl.net: Everybody | CHEMISTRY-REQUEST (+ at +) www.ccl.net: Coordinator > MAILSERV (+ at +) www.ccl.net: HELP CHEMISTRY or HELP SEARCH | Gopher: www.ccl.net 73 > Anon. ftp: www.ccl.net | CHEMISTRY-SEARCH (+ at +) www.ccl.net -- archive search > Web: http://www.ccl.net/chemistry.html > > ______________________________________________________________ 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 ______________________________________________________________