[molpro-user] UCCSD(T) calculation with DHF-QZVP-2C set does not converge for a Cs atom

[Kirk Peterson]

Dear Alexander,

looks to me like clearly the 2c basis set has some significant linear dependency problems from the p functions. This is shown in the bottom of the integral output:

 Eigenvalues of metric

         1 0.439E-07 0.617E-03 0.729E-02 0.114E-01 0.379E-01 0.508E-01 0.143E+00 0.181E+00
         2 0.439E-07 0.114E-01 0.508E-01 0.143E+00 0.188E+00 0.389E+00 0.509E+00 0.665E+00
         3 0.439E-07 0.114E-01 0.508E-01 0.143E+00 0.188E+00 0.389E+00 0.509E+00 0.665E+00
         4 0.188E+00 0.509E+00 0.665E+00 0.100E+01 0.100E+01 0.124E+01 0.134E+01 0.207E+01

Those 3 very small eigenvalues are the clue what the problem is.  The 1c basis set doesn't have this issue.  Notice in the 2c basis set that primitive #3 from the original p set has an exponent of 0.47666322385 (which is actually uncontracted) and the 2nd extra p function that is included in the 2c set is nearly identical, 0.4782. I would just remove that last extra p and then you should be fine.

best regards,

-Kirk


On Apr 23, 2014, at 1:02 PM, sasha medvedev <vinsanity305 at mail.ru> wrote:

> Hello, molpro users and developers.
> 
> I met a rather strange result: 
> uccsd(t) calculation for a Cs atom did not converge,
> when I used dhf-qzvp-2c basis set. 
> 
> So, the ending of the output file was as following:
> 
> ? Error
>  ? No convergence. This error exit can be avoided using the NOCHECK option
>  ? The problem occurs in ccsd
> 
> Here is the code (I also attached the appropriate input file to this mail):
> ***
> gthresh,energy=1.d-8,orbital=1.d-8,coeff=1.d-7
> ....
> symmetry,x,y
> geometry={Cs}
> {rhf;occ,3,1,1,0;core,0,0,0,0;wf,9,1,1;}
> {uccsd(t);core,0,0,0,0;wf,9,1,1;}
> ...
> 
> I should note, that the dhf-qzvp-1c set converges in such calculation (you can run the attached input file - it contains both sets).
> And, by the way, the difference between -1c and -2c set is only in two p-functions, which are added to take into account spin-orbit effects.
> Of course, I see that the result depends on global thresholds [so, calculations with dhf-qzvp-2c set converge for bigger thresholds; 
> for instance with:  gthresh,energy=1.d-6,orbital=1.d-5,coeff=1.d-5 ]. 
> But it still seems to be strange, that almost the same basis set doesn't converge in the 
> same routine.
> 
> I use molpro2010, so I take dhf basis sets here: http://www.cosmologic.de/basis-sets/basissets.php
> 
> Alexander Medvedev,
> student of the 5th course,
> MSU, Chemistry Department,
> Chair of laser chemistry
> <Cs_dhf-qzvp-1c_2c.inp>_______________________________________________
> 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/20140424/8bc116d9/attachment.html>