[molpro-user] Problem with ZnCl2

[Muller, Richard]

I'm having a problem computing geometries of ZnCl2. I certainly could 
be doing something wrong, because although I've used molpro quite a 
bit, this is the first time I've used either the Hay/Wadt ECP, or the 
B3LYP DFT functionalities.

Here's the jist of the problem: both Jaguar and GAMESS/US give a linear 
geometry for ZnCl2, but molpro gives a bent geometry. To make sure I 
didn't do something dumb in the optimization, I attach here a geometry 
scan using a fixed Zn-Cl bond length (2.165 \AA) and angles from 90-180 
degrees. I compare (left to right) Jaguar using the LACVP basis/ECP, 
Gamess using the 6-31G** basis set, Gamess using the Hay/Wadt ECP, and 
Molpro using the LANL2DZ (which I think is essentially the same basis 
set/ECP as the Gamess Hay/Wadt). All programs use the B3LYP functional:

Ang   Jag/LACVP      GMS/6-31G**   GMS/HW   MP/LANL2DZ
90	-986.0182	-2699.2983	-95.4352	-95.1170
100	-986.0356	-2699.3169	-95.4529	-95.0963
110	-986.0478	-2699.3302	-95.4654	-95.0812
120	-986.0575	-2699.3405	-95.4748	-95.0706
130	-986.0650	-2699.3485	-95.4820	-95.0584
140	-986.0704	-2699.3544	-95.4874	-95.0548
150	-986.0746	-2699.3585	-95.4913	-95.0524
160	-986.0776	-2699.3609	-95.4939	-95.0509
170	-986.0791	-2699.3621	-95.4953	-89.0280
180	-986.0794	                   	-95.4958	
				
(this will probably only look good in a fixed-width font).

Am I doing something dumb here? Here's an example of the input file I'm 
using:

***,ZnCl2
basis=LANL2DZ

geometry={,
  ANGSTROM
  Zn
  Cl1, Zn, 2.165
  Cl2, Zn, 2.165, Cl1 170
}
ks,b3lyp

I'd like to augment these calculations with some CC calculations, but I 
wanted to figure out what I'm doing wrong first. Thanks in advance for 
any help anyone can answer.


Rick Muller
rmuller at sandia.gov