question about LQUANT

[Irikura, Karl K. Dr. (Fed)]

Dear Molpro experts,

 

I have a state-averaged, diatomic situation in which the LQUANT directive appears to be ignored.  The problem state is 3-Phi, which rises above a second 3-Sigma when the bond is stretched.  Despite the LQUANT directive, the second 3-Sigma is returned.  I have two questions.

(1) Is there a way to get this to work?

(2) Even if it does work, will the subsequent icMRCI simply converge to the (lower) second 3-Sigma anyway?

 

I have been using the C2 point group because of the following icMRCI (and then spin-orbit CI).  In C2v, degenerate pairs are split between different irreps.  Because the icMRCI calculations are usually of different size for the different irreps, the degeneracies become broken.  Broken degeneracies also make the Omega values in the SO-CI difficult to assign.

 

Thanks for any suggestions,

Karl

 

***,PtH+ cation C2 point group, trouble using LQUANT to get 3-Phi

memory,1000,M;

 

angstrom

symmetry,xy;

geometry={

Pt

H 1 R

}

 

R=2.6

!R=2.5  ! 3-Phi is lower than (2)3-Sigma for short R, so no problem

 

core;

 

basis,default=aug-cc-pVTZ,Pt=aug-cc-pwCVTZ-PP

 

! PtH+, closed-shell

{rhf;wf,elec=18,sym=1,spin=0}

 

! CASSCF

{multi

        ! singlets

    ! OK only when singlets have weight=0

    wf,sym=1,spin=0; state,3; lquant,0,2,2;

    wf,sym=2,spin=0; state,2; lquant,1,1;

        ! triplets to include 3-Phi

    wf,sym=1,spin=2; state,3; lquant,2,2,0;

    wf,sym=2,spin=2; state,4; lquant,1,1,3,3;

    expec2,lzz;

}

table, energy, lzlz

title, Energies and Lz**2 values

text,Last two Lz**2 values should be 9

 

-------------------------------------------------------------------

Dr. Karl K. Irikura

National Institute of Standards and Technology

100 Bureau Drive, mail stop 8320

Gaithersburg, MD 20899-8320  USA

voice:  301-975-2510   fax:  301-975-3670

email:  karl.i...@nist.gov

http://www.nist.gov/mml/csd/informatics_research/index.cfm

-------------------------------------------------------------------

 

[Hans-Joachim Werner]

Dear Karl,

I reproduced the problem, and unfortunately there seems still to be a bug in the lquant treatment of the new multi program. It works correctly with the old program (multi,wmk_old). Sorry for this problem, we will try to fix it asap.

Note that lquant can only work correctly if the pix and piy orbitals are equivalent, both for the starting guess and the final orbitals. This requires to include always all degenerate states (e.g. pix and piy). But this is all correct in your input.

Regarding degeneracies in mrci you are right, if the number of included mrci states is not the same in different irreps (e.g. symmetries 1 and 4 in c2v). But if the states in a given molpro symmetry have different symmetries (e.g. pi and phi in symmetry 2, or sigma and delta in symmetry 1), one can usually get correct degeneracies by doing single-state mrci calculations. For example, with state,1,3 the mrci will use the third reference state and use a root following mechanism to converge to the state which has largest overlap with the specified reference. Convergence of this procedure is not guaranteed, but it works fine for your example, see attached. In case that the root following fails, it sometimes helps to specify option,maxdav=20. To make sure that the correct state is obtained, compare the reference energy in mrci (variable energr) with the corresponding multi energy. In the attached example these energies agree for all states. Unfortunately, there is no lquant procedure in mrci, and therefore this check is important.

Best regards
Joachim Werner