[molpro-user] Full CI convergence problems with hydrogen

[Peter Knowles]

This works fine (after correcting the benign syntax error in the  
input) with both versions 2002.6 and 2006.1 here. I can't reproduce  
your aug-cc-pVQZ number (I get -1.01901806) but perhaps this is for a  
different bond length. Tatiana's comment about integral thresholds is  
of course relevant, and should be investigated, particularly for  
short bond lengths.
Peter
On 20 Mar 2007, at 12:06, Lorenzo Lodi wrote:

> I was trying as a test some FCI calculation for the hydrogen  
> molecule (I am using molpro v2002.6), but when I try with the aug- 
> cc-pV5Z basis set I have convergence problems.
> This is the input file:
>
> memory, 80, M
> basis=aug-cc-pV5Z
> r = 0.8
> geometry={h1;h2,h1,r}
> rhf, WF, ELEC=2, SYM=1, SPIN=0
> FCI
>
> I get a FCI energy of -3.057081385992 hartrees, which is wrong.
> On the other hand, if I try with a smaller basis set like aug-cc- 
> pVQZ I get -1.082735905256 hartrees, which I belive is the correct  
> result for this basis set.
> For values of r greater than 0.9 bohrs everything is OK even with  
> the A5Z basis set.
> Incidentally, I've also tried other methods like MRCI or CCSD but I  
> always get some errors for some combinations of bond lenghts/basis  
> set.
> Is there anything that can be done about this?
>
> N.B.: the calulation FCI(A5Z) takes about 5min of a PC.
>
> Thank you,
>
> Lorenzo Lodi

--
Prof. Peter J. Knowles
School of Chemistry, Cardiff University, Main Building, Park Place,  
Cardiff CF10 3AT, UK
Telephone: +44 29208 79182 Fax: +44 2920874030
Email KnowlesPJ at Cardiff.ac.uk  WWW http://www.cardiff.ac.uk/chemy/ 
staff/knowles.html