[molpro-user] Bug?!!! Incorrect spin-orbit coupling values and scales linearly with the number of processors

[Andy May]

Ryan,

I've made a bug report about the problem:

https://www.molpro.net/bugzilla/show_bug.cgi?id=4412

Best wishes,

Andy

On 15/04/14 20:51, Ryan Zaari wrote:
> Dear all,
>
> I would be grateful for a response or for others to let me know if they
> encounter the same error (input file included at the bottom).
>
> I calculated the SiH2 spin orbit coupling for components (LSx, LSy, LSz)
> of the lowest energy singlet and triplet states.
> |S Ms>  = |00> for singlet and |S Ms>=|1 -10,1> for triplet.
>
> Now the problem occurs with the number of processors I use.  Below is
> the <00 | LSy | 1 1> output data for which I varied the number of
> processors and also included the lsint keyword or not:
>
> # of processors<00 | LSy | 1 1>, without lsint <00 | LSy | 1 1>, with lsint
> 2-44.7 cm-141.8 cm-1
> 4-138.9 cm-141.8 cm-1
> 8-327.5 cm-141.8 cm-1
> 12-516.2cm-141.8 cm-1
>
> A plot of the number of processors vs. without lsint <00 | LSy | 1
> 1> produces an exact linear plot.
> This scaling of the spin-orbit coupling occurs for all 9 components of
> LSx, LSy, LSz between triplet and singlet states.
> I have included the input file below.
>
> Thanks,
> Ryan Zaari
>
>
>
>
>
> ***,SiH2
> memory,100,m
>
> nosym
> geomtyp=xyz
> geometry={
> Si          0.0000006195        0.0232176634       -0.0512652632;
> H          -0.0000087894        0.8058770300        1.0930785117;
> H          -0.0000084752       -1.4528938516        0.3355529707;
> }
>
> basis = TZVP
>
> HF
>
> {casscf
> maxiter,40
> occ,11
>   closed,5
>   wf,16,1,0;state,2;
>   wf,16,1,2
>   dynw,10
>   cpmcscf,grad,1.1,ms2=0,record=5102.1
>   cpmcscf,grad,1.1,ms2=2,record=5101.1
>   }
>
> !lsint
>
> {mrci; maxiter, 100, 100; wf,16,1,0;noexc;save,4010.3}
> {mrci; maxiter, 100, 100; wf,16,1,2;noexc;save,4030.3}
>
>
> {mrci;TRANLS, 4010.3, 4030.3, 0, 2, LSY}
>
>
>
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user
>