[molpro-user] (no subject)

Kirk Peterson kipeters at wsu.edu
Fri Aug 17 16:26:35 BST 2012


Gerry,

unfortunately in the context of automatic optimizations with optg, the correlation method will revert to using mostly default options, e.g., HF orbitals, etc.  There are two options in your case as I see it.  In the first place try giving your HF calculation an initial guess from your MCSCF calculation and then hope that all subsequent HF calcs are then ok.  This will probably work.  The 2nd option is to use the Minimize program for your optimization. It's not as efficient as optg for geometry optimizations but it will get the job done with all options respected.

best,

-Kirk

On Aug 17, 2012, at 6:30 AM, "Hoffman, Gerald" <GHOFFMAN at edinboro.edu> wrote:

> I am trying to run an optimization on a triplet van der Waals complex at the level of CCSD(T).  Apparently, the orbitals returned from the HF step were very poor, because the coupled cluster routine crashed at once with an unreasonable norm.  In an effort to improve the quality of the orbitals, I inserted an MCSCF calculation, and included the "ignore_error" option in the RCCSD(T) directive.  This strategy worked fine the first time through, but in the first calculation in the optimization step, RCCSD(T) crashed, requesting that the "ignore_error" option be invoke (even though it already was invoked).
> 
> So, I need either to improve the quality of the orbitals calculated in the RHF step without an MCSCF step, or I need to find a way to convince RCCSD(T) that I really do want the error to be ignored.
> 
> Any ideas?
> 
> Thanks.
> 
> Gerry
> 
> Gerald J. Hoffman
> 
> Assistant Professor of Chemistry
> 
> Edinboro University of Pennsylvania
> 
> 230 Scotland Road
> 
> Edinboro, PA 16444
> 
>  
> 814-732-2813
> 
>  
> ghoffman at edinboro.edu
> 
>  
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.molpro.net/pipermail/molpro-user/attachments/20120817/9fa7fe55/attachment.html>


More information about the Molpro-user mailing list