[molpro-user] unreasonable norm in ccsd calculation reg

[Gerald Knizia]

On Tuesday 28 September 2010 11:25, Rajagopala Reddy seelam wrote:
> I am facing a problem regarding
> unreasonable norm in ccsd calculation. I am using Molpro 2002.6 to do
> this. I found one query regarding the same in molpro-users mail archive
> and there it was suggested that using shifts=value will work. I used it
> but no success. Can any one of you please help. 

As Kirk already pointed out, this usually happens when the HF reference 
calculation is somehow wrong or otherwise very bad (e.g., trying to use HF 
references for a strong multireference system where HF/CCSD is not a good 
method).

But if you are sure that your reference is okay, you can also tell Molpro to 
extend the unreasonable norms thresholds and increase the number of 
iterations. For example, the input
   _CC_NORM_MAX=2.0
   {uccsd(t); maxit,200; diis,1,1,8}
will be more tolerant to ill-conditioned systems.
-- 
Gerald Knizia