[molpro-user] problems with contraction in RS2c

[H.-J. Werner]

Dear Ramon,
please try to reduce threshold thrdls. It is similar to thrdlp but used for the
contracted singles in rs2c. The default value for thrdls is 1.d-7, so
you might try 1.d-8. Of course I can't tell if this will lead to numerical
problems (which may show up in convergence problems).
Please let me know if this helps.
Best regards
Joachim Werner
On Do, 19 Jul 2007, ramon at buzon.uaem.mx wrote:

>
> Hi,
>
> we are doing RS2 type calculations on a weakly bound complex, when using
> the RS2c module we observe a jump or discontinuity in the potential at long
> range(see the table below). When looking at the output of rs2c we notice
> that there is a corresponding change in the number of contracted internal
> and N-1 configurations used in the neighborhood of the jump. If we use
> the RS2 module the problem disappears but we would like to use the more
> efficient RS2c.
>
> A similar problem had been already discussed in the mailing list(31/10/03)
> for acpf calculations and Prof. Werner suggested a solution. We first
> tried adding the option thresh,pnorm=0,thrdlp=1.d-8 but this had no
> effect on the calculated energies. Then we tried fixing the pair list
> using the option save,5000.2;start,5000.2 and in this case the problem
> that appears is that when RS2 reads the wavefunction from 5000.2(in the
> second point of the calculation) it complains that its overlap with the
> previous CI vector is nearly zero and exits(see below the table)
>
> Your help is appreciated
>
> Ramon Hernandez
>
>    R(ang)       R      ERS2C        INT-POT      (INT-POT)*27211.4
>   37.042390  70.0  -300.3143994   -0.00079168     0.0000
>   10.583540  20.0  -300.3144003   -0.00079257    -0.02440715
>    8.466832  16.0  -300.3144039   -0.00079514    -0.09422729
>    7.408478  14.0  -300.3144158   -0.00079960    -0.21562064
>    6.350124  12.0   -300.3151163  -0.00146832    -18.41250990
>    5.291770  10.0   -300.3152382  -0.00150677    -19.45864921
>
>
> Overlap between present and previous reference vectors
>                   1
>         1     -.0000017
>
> NO VECTOR OF INTERNAL CI OVERLAPS SUFFICIENTLY WITH REFERENCE VECTOR 1.
>INCREASE NSTATI!
>
> OVERLAP BETWEEN INITIAL AND PRESENT VECTORS
>                   1
>         1     -.0000017
>
> ovmax= 0.170665286226771260E-05   ovref= 0.170665286226771260E-05
>
> ? Error
> ? Insufficient overlap
> ? The problem occurs in cihdia
>
> ERROR EXIT
> CURRENT STACK:      CIPRO  MAIN
>
>
>
>
>

-- 
Prof. Hans-Joachim Werner
Institute for Theoretical Chemistry
University of Stuttgart
Pfaffenwaldring 55
D-70569 Stuttgart, Germany
Tel.: (0049) 711 / 685 64400
Fax.: (0049) 711 / 685 64442
e-mail: werner at theochem.uni-stuttgart.de