[molpro-user] CI

[Jacek Koput]

Hi there,

in the calculation with the CI program, I am facing the following 
convergence problem:

   ITER. STATE  ROOT     SQ.NORM     CORR.ENERGY   TOTAL ENERGY   ENERGY 
CHANGE       DEN1      VAR(S)    VAR(P)      TIME
     1     1     1     1.00000000     0.00000000   -55.08381912     
0.00000000    -0.10329888  0.76D-02  0.19D-02    92.77
     2     1     1     1.01025886    -0.10952326   -55.19334238    
-0.10952326    -0.00204259  0.20D-03  0.13D-03   523.90
     3     1     1     1.01213335    -0.11213381   -55.19595293    
-0.00261055    -0.00026112  0.41D-04  0.12D-04   966.12
     4     1     1     1.01339244    -0.11244918   -55.19626830    
-0.00031537    -0.00004888  0.39D-05  0.24D-05  1407.53
     5     1     1     1.01354611    -0.11249778   -55.19631690    
-0.00004860    -0.00003127  0.90D-06  0.12D-05  1829.66
     6     1     1     1.01352297    -0.11254369   -55.19636282    
-0.00004591    -0.00008028  0.21D-05  0.94D-06  2244.20

  Requested states
    1-  1     1

  Current NROOT
    1-  1     1

  CI VECTOR FOR STATE 1 DOES NOT OVERLAP SUFFICIENTLY WITH REFERENCE 
VECTORS

  It may help to increase NSTATI, e.g., use OPTION,NSTATI=2

  Rotated S(REF|INT)
                    1
          1     0.0023982
  ? Error
  ? Insufficient overlap
  ? The problem occurs in cihdia

Adding 'OPTION,NSTATI=2' results in a huge number of messages 'NON ZERO 
P-SP. GR.', ending with '** WARNING, NO CONVERGENCE CIACPF'.

Surprisingly, for the bond length being 0.02 A apart (diatomic), the CI 
iterations converge fine.

Is it possible to overcome this problem?

I would greatly appreciate any hint,
Jacek Koput