[molpro-user] problem with spin-orbit coupling calculation in molpro-2006

[Alexander Mitrushenkov]

Dear Yuri,

I looked at your input file. First of all, it was not possible to run such a 
calculation with molpro 2002.6, as  general contraction was not implemented 
in the previous version of spin-orbit integral program.
Using uncontracted basis, like e.g.

basis={spd,c,vdz;spd,o,vdz;sp,h,vdz;}

gives the same results both in 2002.6 and 2006.1. In particular in your 
example the strange splittings are caused by the symmetry breaking in CASSCF 
calculations (which is valid both for 2006.1 and 2002.6), i.e. three P states 
of oxygen are not completely equivalent, and energies are different. This 
clearly gives wrong results when diagonalizing spin-orbit matrix. If you 
force for three states to have exactly the same unperturbed energy, as 
suggested in the manual, i.e. using input like
================================
***, c2h4+o
memory,100,M

geomtyp=xyz

basis=vdz;

geometry={angstrom
nosym; noorient
7
rotational dynamics
C,         0.0000000000,       0.0000000000,      -0.6727675475
C,         0.0000000000,       0.0000000000,       0.6727675475
H,         0.0000000000,       0.9347675226,      -1.2426977645
H,         0.0000000000,      -0.9347675226,      -1.2426977645
H,         0.0000000000,       0.9347675226,       1.2426977645
H,         0.0000000000,      -0.9347675226,       1.2426977645
O, 5, 1, 1
}

{hf; wf, 24, 1, 2;}
{multi; occ, 13; closed, 10; state, 3;}

{ci; save, 3010.1; state, 3; noexc;}

clear,hlsdiag;
hlsdiag=[-152.825298376992,-152.825298376992,-152.825298376992]

{ci; hlsmat, ls, 3010.1;}
==================================

you will see that the splittings are perfectly OK which shows that spin-orbit 
integrals and matrix are perfectly correct.

The problem of CASSCF symmetry breaking is not easy to solve especially using 
lower symmetry calculations.

So please let me know if this answers your question or not. Moreover, in 
principle you were saying that this problem is not seen using z-matrix input, 
or old molpro 2002.6. Could you please then provide corresponding inputs and 
outputs?

Best regards,

Sasha

-- 
Dr. Alexander Mitrushchenkov
Universität Stuttgart
Institut für Theoretische Chemie
Pfaffenwaldring 55
D-70569 Stuttgart

Tel. +49 (0)711 / 685-64409
Fax  +49 (0)711 / 685-64442
e-mail: sasha at theochem.uni-stuttgart.de



On Friday 01 September 2006 21:29, Yuri Georgievski wrote:
> Hi,
> It seems that there is a problem when one tries to calculate the
> spin-orbit coupling with new version of molpro. This problem appears
> only when the geometry is given in cartesian coordinates. Here is the
> example input:
>
> ***, c2h4+o
> memory,100,M
>
> geomtyp=xyz
> geometry={angstrom
> nosym; noorient
> 7
> rotational dynamics
> C,         0.0000000000,       0.0000000000,      -0.6727675475
> C,         0.0000000000,       0.0000000000,       0.6727675475
> H,         0.0000000000,       0.9347675226,      -1.2426977645
> H,         0.0000000000,      -0.9347675226,      -1.2426977645
> H,         0.0000000000,       0.9347675226,       1.2426977645
> H,         0.0000000000,      -0.9347675226,       1.2426977645
> O, 5, 1, 1
> }
>
> basis=vdz
> {rhf; wf, 24, 1, 2}
> {multi; occ, 13; closed, 10; state, 3}
>
> geometry={angstrom
> nosym; noorient
> 7
> rotational dynamics
> C,         0.0000000000,       0.0000000000,      -0.6727675475
> C,         0.0000000000,       0.0000000000,       0.6727675475
> H,         0.0000000000,       0.9347675226,      -1.2426977645
> H,         0.0000000000,      -0.9347675226,      -1.2426977645
> H,         0.0000000000,       0.9347675226,       1.2426977645
> H,         0.0000000000,      -0.9347675226,       1.2426977645
> O, 15, 1, 1
> }
>
> {multi; occ, 13; closed, 10; state, 3}
> {ci; save, 3010.1; state, 3; noexc}
> lsint
> {ci; hlsmat, ls, 3010.1}
> ---
>
> As a result of this calculation one obtains spin-orbit splitting of 110
> 1/cm while for a stand-alone oxygen it should be 75 1/cm for this basis
> set. If one uses a z-matrix geometry representation the result seems to
> be ok. I tried both Pople and Dunning basis sets and this problem
> appears in all cases.  For previous version of molpro, 2002.6, this
> problem does not exist.
>
> Yuri