[molpro-user] QSD optimization error

[Nicholas Marshall]

Dear Molpro users,

I have a strange problem with a reaction-coordinate following
optimization that I am attempting.

***,title
memory,270,m
file,2,leu_mp2qsd.wfn,new

gthresh,energy=1.d-10,optgrad=1.d-6,optstep=1.d-4

(definitions of z-matrix variables)

   geometry={angstrom;nosymm;

    (z-matrix goes here)

    }
   basis=STO-3G
    hf
    wf,72,1,0
    frequencies,analytical
   save,5300.2
   basis=aug-cc-pVTZ
    hf
    wf,72,1,0

direct
    mp2
    core,9
    optg,maxit=200
    method,qsdpath
    numhes,-1
    hstart,5300.2
    print,opt
    put,molden,atz_mp2_opt.mol

This optimization goes completely wrong, producing a nonphysical
geometry with far-too-short bond lengths. The energy actually increases
during the optimization. The same optimization works normally with the
cc-pVTZ and STO-3G basis sets.

Note that in the above calculation, I have used a starting STO-3G SCF
Hessian, but the problem still occurs if the default Hessian from the
force field is used.

Note that this is a large molecule, with 22 total atoms.

Does anyone know what could be happening?

Thanks very much,

Nicholas Marshall
2nd year graduate student
Center for Computational Chemistry, University of Georgia