[molpro-user] Au UCCSD(T)-F12

[Grant Hill]

Dear Zongtang,

This is neither a bug or something particularly wrong with your input. If you look at the aug-cc-pVQZ-PP/MP2Fit auxiliary set on either the EMSL basis set exchange or in Turbomole’s online basis set library you will see that they contain k-type functions. These are not presently supported in Molpro.

What you can try is to simply remove the k-type functions and retain the rest of the auxiliary set. This will lead to substantial density fitting errors in the absolute correlation energy. But, these errors _tend_ to cancel for relative energies - see Hill and Peterson, J. Chem. Theory Comput. 8, 518 (2012).

Users should be aware of this potential problem, rather than running the calculation without realising what is happening. Hence, truncated auxiliary basis sets were not placed into the Molpro library. As I mentioned above, the calculation can be easily run by downloading the auxiliary basis and manually inserting it into your input file.

Best wishes,
Grant





> On 5 Oct 2016, at 18:04, Zongtang Fang <zfang at crimson.ua.edu> wrote:
> 
> Hello,
> 
> I am trying to do ccsd(t)-f12 for Au atom at the QZ level with molpro 2015.
> 
> I got an error message:
> 
> 
> " Basis set DEF2-QZVPP/JKFIT generated.   Number of basis functions:   247
>  Basis set AVQZ-PP/OPTRI generated.      Number of basis functions:   206
>  Cannot find default basis AVQZ-PP for atom Au
>  Type=FIT
>  Context=MP2FIT"
> 
> 
> Seems like MP2FIT generation messed up. 
> 
> I do not have problem with aD and aT. 
> 
> Is this a bug or something wrong with my input ?
> 
> Thanks. 
> 
> Zongtang
> 
> 
> 
> 
> 
> My input:
> 
> 
> ***,Cu, Doublet,
> 
>  geomtyp=xyz
>  geometry={
>  1
>  in angstroms
>    Au     0.000000    0.000000    0.000000
>  }
> 
>  basis=avqz-pp
> 
>  {hf;occ,4,1,1,1,1,1,1,0;closed,3,1,1,1,1,1,1,0;wf,19,1,1;}
>  {uccsd(t)-f12,ri_basis=avqz-pp/optri,df_basis=avqz-pp;}
> 
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user