[molpro-user] weird SCF convergence/reproducibility problem

[Gregory Sandala]

Dear Jan,

I can reproduce the G03 RHF energy (HF=-262.3384479) by using the  
"MULTI" command in Molpro.  I've previously found this algorithm to  
converge more reliably for some systems, especially radical anions.  
Strangely, when I read the MCSCF (CSF=1) wave function into ROHF I  
continue to get the same (higher) energy you have found.

I'm certainly not a guru (and this doesn't really solve your problem,  
i.e., why isn't ROHF converging to the lower solution), but it may  
nonetheless be of some assistance.

With best wishes,
Greg


INPUT:

memory,8,m
gthresh,energy=1d-12,twoint=1d-18,prefac=1d-20
geomtyp=xyz
geom={
     4
E(UHF)=-262.355671204 au|Dpl=(0.5152,-0.4848,0) D||E(ROHF) 
=-262.338447756 au|Dpl=(0.9471,-1.9519,0) Debye
O    -0.832632    1.113558    0.000000
C     0.000000    0.281073    0.000000
O    -0.367117   -1.293021    0.000000
O     1.199748   -0.031341    0.000000
}
basis=vdz
int
{uhf;wf,30,2,2;occ,13,3;closed,12,2}
! yields the correct energy
!UHF STATE 1.2 ENERGY               -262.355671266903
!UHF STATE 1.2 DIPOLE MOMENT           0.20268671    -0.19074510      
0.00000000
{multi;occ,13,3;closed,12,2;wf,30,2,2}
{hf;wf,30,2,2;occ,13,3;closed,12,2}
! yields a different energy, and dipole vector looks very different too
!RHF STATE 1.2 ENERGY               -262.331395031327
!RHF STATE 1.2 DIPOLE MOMENT           0.10470623    -0.21621532      
0.00000000


SELECTED OUTPUT:

1PROGRAM * MULTI (Direct Multiconfiguration SCF)       Authors: P.J.  
Knowles, H.-J. Werner (1984)     S.T. Elbert (1988)


  Number of closed-shell orbitals: 14 ( 12  2 )
  Number of active  orbitals:       2 (  1  1 )
  Number of external orbitals:     40 ( 27 13 )

  State symmetry 1

  Number of electrons:     2    Spin symmetry=Triplet   Space symmetry=2
  Number of states:        1
  Number of CSFs:          1   (1 determinants, 1 intermediate states)


   
************************************************************************ 
**********************************************************

         RHF-SCF         MULTI           UHF-SCF
    -262.33139503   -262.33844792   -262.35567127
   
************************************************************************ 
**********************************************************


***
Gregory M. Sandala, PhD
School of Chemistry, Bldg F11				
University of Sydney				
NSW 2006, Australia
Ph: +61 2 9351 5361
sandala at chem.usyd.edu.au

On 27/04/2008, at 8:22 AM, Gershom (Jan M.L.) Martin wrote:

> Dear MOLPRO gurus:
>
> Very recently I ran into a peculiar example of a transition state  
> for which MOLPRO disagrees with every other code on the ROHF  
> reference energy. See the simple input below. I tried every SCF  
> convergence acceleration/coaxing trick in the book and a few that  
> aren't, to no avail.
>
> The other structures on the same PES do not exhibit the problem.  
> Does anybody have a clue what I am overlooking?
>
> (And yes, I know this is a highly multiref problem. Still, ROHF  
> should give the same answer as Gaussian and ACES?)memory,8,m
>
> Many thanks in advance!
> Jan Martin
>
>
>     4
> E(UHF)=-262.355671204 au|Dpl=(0.5152,-0.4848,0) D||E(ROHF) 
> =-262.338447756 au|Dpl=(0.9471,-1.9519,0) Debye
> O    -0.832632    1.113558    0.000000
> C     0.000000    0.281073    0.000000
> O    -0.367117   -1.293021    0.000000
> O     1.199748   -0.031341    0.000000
> }
> basis=vdz
> int
> {uhf;wf,30,2,2;occ,13,3;closed,12,2}
> ! yields the correct energy
> !UHF STATE 1.2 ENERGY               -262.355671266903
> !UHF STATE 1.2 DIPOLE MOMENT           0.20268671     
> -0.19074510     0.00000000
> {hf;wf,30,2,2;occ,13,3;closed,12,2}
> ! yields a different energy, and dipole vector looks very different  
> too
> !RHF STATE 1.2 ENERGY               -262.331395031327
> !RHF STATE 1.2 DIPOLE MOMENT           0.10470623     
> -0.21621532     0.00000000
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20080428/182a12aa/attachment.html>