[molpro-user] Convergence not reached for a CCSD iteration in CCSD(T) optimization

[Gerald Knizia]

Dear Yan-Bo,
this looks very strange. The numbers under VAR(P) and VAR(S) effectively
say that the equations are already solved, but Molpro somehow cannot
converge the remaining tiny bit on the order of 1e-10. What Molpro
version is this? 

Could you try putting the following:

  {gthresh,oneint=1e-18,twoint=1e-15}

before the first RHF calculation and see if this helps?

Btw: In order to run CCSD with non-standard thresholds, you can invoke
it like this:

  {ccsd(t),thrvar=1e-12,thrden=1e-8}

(thrvar is the threshold on the residuals/amplitude updates, and thrden
the threshold for the change in energy)

Best wishes,
Gerald


On Fri, 2015-08-21 at 00:32 +0800, wyb wrote:
> Dear MolPro users
> 
> 
> I encountered a problem in the CCSD(T)/aug-cc-pVTZ optimization [but
> CCSD(T)/cc-pVTZ optimization can be achieved].
> In one of the CCSD step, the "VAR (P)" can not be lowered below
> 1.0D-12 even I increase the maximum number of CCSD iterations to 300.
> ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>  ITER.   SQ.NORM   CORR.ENERGY TOTAL ENERGY  ENERGY CHANGE     DEN1
> VAR(S)    VAR(P)     DIIS     TIME
>    1      1.25279637    -0.50350353  -116.32933471    -0.03743526
> -0.02192105  0.19D-02  0.50D-02  1  1     8.30
>     .               .                        .                      .                         .                        .                    .               .        .    .        .
>     .               .                        .                      .                         .                        .                    .               .        .    .        .
>     .               .                        .                      .                         .                        .                    .               .        .    .        .
> 
>   97      1.28543234    -0.50769672  -116.33352790    -0.00000000
> 0.00000000  0.11D-14  0.17D-10  6  2   336.65
>   98      1.28543234    -0.50769672  -116.33352790    -0.00000000
> -0.00000000  0.94D-15  0.18D-10  6  6   340.07
>   99      1.28543234    -0.50769672  -116.33352790     0.00000000
> 0.00000000  0.94D-15  0.18D-10  6  2   343.52
> 
> 
> 
>  Norm of t1 vector:      0.07197772      S-energy:     0.00000000
> T1 diagnostic:  0.01138068
> 
> D1 diagnostic:  0.03667313
> 
> ?CONVERGENCE NOT REACHED AFTER MAX. ITERATIONS
> ------------------------------------------------------------------------------------------------------------------------------------------------------------
> 
> I think the possible way to solve this problem is to promote the
> convergence of the "VAR (P)" item using certain keyword(s) or to
> reduce the criterion a little bit (for example to 1.0D-11), but I can
> not find the related keyword(s) in the manual. Maybe I did not catch
> what the manual says.
> 
> 
> Can anyone tell me how to sovle this problem?
> 
> 
> Thank you in advance!
> 
> 
> Best,
> 
> 
> Yan-Bo Wu
> ---------------------------------------
> Institute of Molecular Science
> Shanxi University
> Taiyuan, 030006, Shanxi Province, China.
> 
> 
> 宝宝的进口奶粉纸尿裤，1元全包了
> 
> 
> 宝宝的进口奶粉纸尿裤，1元全包了
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user