[molpro-user] Linear dependent basis

[Gerald Knizia]

On Monday 12 July 2010 19:23, John Keith wrote:
> AHA!  I can now see how that would cause serious problems and I'm
> lucky you caught this. I had used Cartesian functions to have
> continuity with earlier Pople basis set calculations I ran with
> Jaguar, the program I have the most experience with.  Lesson learned.
>
> I'll make this change and then write a follow-up if that resolved my
> linear dependency problem.

If you run into convergence problems, another thing you might want to do is to 
use tighter integral thresholds:

gthresh,oneint=1e-15,twoint=1e-14

This can circumvent some of the near-linear dependency problems, especially in 
correlated calculations of highly symmetric molecules (e.g., when hf 
converges, but ccsd does not).
-- 
Gerald Knizia