[molpro-user] NEVPT2 problem for C2 ground/excited states

[Alexander Sokolov]

Dear Molpro Users/Developers,

I have been trying to compute potential energy curves for the dissociation
of C2 molecule for the three low-lying singlet states. I run SA-CASSCF for
3 states and specify LQUANT,0,0,2, because I would like to compute the two
lowest Sigma^+_g states and one Delta_g state. The SA-CASSCF computation
runs fine and produces qualitatively correct potential energy curves.
However, when computing potential energy curves near conical intersection
in the region of 2.9-3.2 Angstrom, state-specific NEVPT2 computations for
the two lowest-energy states converge to the same root:

             1.1                 2.1                 3.1
2.8 -75.6722945 -75.6489181 -75.6359653
2.9 *-75.6433987 -75.6433986* -75.6299956
3.0 *-75.6360085 -75.6360085* -75.6217944
3.1 *-75.6274115 -75.6274115* -75.6114991
3.2 *-75.6181054 -75.6181054* -75.5990394

Similar situation is observed with state-specific CASPT2. As far as I
understand, the problem arises at the initial CI step, which produces the
same reference state in both state-specific computations, despite the fact
that the reference SA-CASSCF energies are different.

Is there a way to avoid this problem to generate correct NEVPT2 potential
energy curves near conical intersection?

I run MOLPRO 2010.1. Here's my input file:

 memory,500,m
 gprint,orbitals=5,basis,civector
 gthresh,energy=1.0d-10,orbital=1.0d-8

 do i=1,32

 BOHR

 Geometry = {
 C
 C 1 R
 }

 R = 1.8 + 0.1 * (i - 1)

 basis=cc-pvdz

 {rhf;orbital,2100.2}
 {multi,maxit=50;start,2100.2;closed,1,0,0,0,1,0,0,0;occ,3,1,1,0,3,1,1,0;wf,12,1,0;state,3;lquant,0,0,2}
 {nevpt2,maxiti=500;core,0;wf,12,1,0;state,1,1}
 {nevpt2,maxiti=500;core,0;wf,12,1,0;state,1,2}
 {nevpt2,maxiti=500;core,0;wf,12,1,0;state,1,3}

 e(i) = energy
 rvec(i) = R

 enddo

 table,rvec,e


Best wishes,

Alexander Sokolov
Postdoctoral Research Associate
Princeton University
Princeton, NJ
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20151116/3e894e05/attachment.html>