From schw0531@compszrz.zrz.tu-berlin.de Sat Oct 31 13:57:39 1992 Date: Sat, 31 Oct 92 12:57:39 +0100 From: Prof. Dr. Helmut Schwarz To: chemistry@ccl.net, rafapa@obelix.cica.es Subject: Re: ECPs and 2nd derivatives again Recently the question of which programs have analytical 2nd derivatives for ECPs was posted. Someone mentiooned the GRADSCF program from Komornicki. Has GAMESS the same capability? Thanks Rafael R. Pappalardo Dept. of Physical Chemistry Univ. of Seville (SPAIN) e-mail: rafapa@obelix.cica.es From schw0531@compszrz.zrz.tu-berlin.de Sat Oct 31 14:50:49 1992 Date: Sat, 31 Oct 92 13:50:49 +0100 From: Prof. Dr. Helmut Schwarz To: chemistry@ccl.net, rafapa@obelix.cica.es Subject: Re: ECPs and 2nd derivatives again Acording to the question of Rafael R. Pappalardo about 2nd derivatives in GAMESS I have some remarks. We have used the british version of GAMESS developt and distributed by Martyn F. Guest for the CRAY and CONVEX computers and had a lot of troubles using even 1 derivatives for ECP. For a number of diatomic molecules the program fails to find a minimum. By performing a variation of the bond distance the one will end up with a optimized distance. I think there are also some bugs evaluating the integrals because the GAMESS-UK gives not the same result as the HONDO-8 program (which should be very simmilar). In my opinion its connected with the use of ECP`s, because for split valence basis sets (6-31G**) we obtain the same results. Moreover, we do not succied by performing MRD-CI-ECP calculations with GAMESS (the program is alway looking for the "core electrons", replaced by the ECP). As mentioned by Gernot Frenking, there is im principle a possibility to perform ECP calculations with the GAUSSIAN program series, but this use is limited. By calculating some TM containing ions we have recognized a disadvatage of the ECP`s in the parametrization of HAY. The description of the excitations (Ionizations) is more than poor. Indepentent of the level of inclusion of the correlation energy, the contraction scheme or the inclusion of inner core shells, we always endup with numbers, not comparable with all elctron values. In my opinion this is ralated to the parametrization procedure as well as to the analytical expresion of the ECP. The ECP works well for saturated low spin TM complexes, describing the geometries (and energies) suprisingly good, but calculations on such systems like M=O, M=N, M-NO, M-NHn completelly fails. The results for M-C=O, M-X (F,Cl,Br,I), M-Ethene are also not very satifying. To overcome this dificulties, we start to parametrize the ECP for ions using the analytical expression of DURAND and BARTHELAT (very simalar to HAY) for the pseudopotential. Now a other problem with the GAUSSIAN apear. To make the PP more flexible, we require powers of r in the range of -6 to 6 and by reading in this PP parameters some of this numbers are replaced by zero. This is the fact in the 88, 90 and also in the 92 program series. Have somebody an idea how to solve this problem ? Jan Hrusak