[molpro-user] question on Cu example in the manual

[Kirk Peterson]

Jussi,

at the CBS limit the d10s1 -> d9s2 excitation energy is still negative  
by about 1.4 kcal/mol.  Once one includes
correlation at the CCSD(T) or MRCI level (even with small basis sets)  
however, the situation is greatly improved and with a systematic
treatment of valence and core-valence correlation effects, one can  
essentially reproduce the experimental value.  You might
check out:  J. Chem. Phys. 125, 074110 (2006) as well as the original  
basis set paper cited in there. Recent calcs with new
PP-based correlation consistent basis sets (same PP as below) gives  
essentially the same quality of results.  So I wouldn't say
it's all that difficult, you just need a balanced treatment of  
electron correlation.

-Kirk

On Apr 3, 2009, at 10:03 PM, Jussi Eloranta wrote:

> Hi,
>
> I noticed the following example in molpro manual:
>
> ----
>
> ! $Revision: 2006.3 $
> ***,CU
> ! SCF d10s1 -> d9s2 excitation energy of the Cu atom
> !  using the relativistic Ne-core pseudopotential
> ! and basis of the Stuttgart/Koeln group.
> gprint,basis,orbitals
> geometry={cu}
> basis
> ECP,1,10,3;   ! ECP input
>  1; ! NO LOCAL POTENTIAL
>  2,1.,0.;
>  2; ! S POTENTIAL
>  2,30.22,355.770158;2,13.19,70.865357;
>  2; ! P POTENTIAL
>  2,33.13,233.891976;2,13.22,53.947299;
>  2; ! D POTENTIAL
>  2,38.42,-31.272165;2,13.26,-2.741104;
> ! (8s7p6d)/[6s5p3d] BASIS SET
> s,1,27.69632,13.50535,8.815355,2.380805,.952616,.112662,.040486,.01;
> c,1.3,.231132,-.656811,-.545875;
> p,1,93.504327,16.285464,5.994236,2.536875,.897934,.131729,.030878;
> c,1.2,.022829,-1.009513;C,3.4,.24645,.792024;
> d,1,41.225006,12.34325,4.20192,1.379825,.383453,.1;
> c,1.4,.044694,.212106,.453423,.533465;
> end
> rhf;
> e1=energy
> {rhf;occ,4,1,1,1,1,1,1;closed,4,1,1,1,1,1;wf,19,7,1;}
> e2=energy
> de=(e2-e1)*toev  ! Delta E = -0.075 eV
>
> ---
>
> This looks a bit odd since it claims that the ^2D is lower than ^2S
> (which should be the ground state). Further more, I got a bit curious
> because of this example and ran mcscf with 3d and 4s in the active  
> space
> and that gives completely wrong answers as well as it also claims that
> the ^2D is lower than ^2S. If one adds CISD on top of this
> the situation is corrected somewhat (the energy order becomes correct
> but the gap is now too high by over 0.5 eV). This behavior shows up  
> all
> the way to av5z (-> not a basis set issue). Including SO is not  
> going to
> help out (tried that).
>
> So, two things: a) the example in the manual is perhaps not a good one
> and b) any idea why Cu atom seems to be such a tough case?
>
>
> Sincerely,
>
> Jussi Eloranta
> Department of chemistry & Biochemistry
> Cal State Northridge
>
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user