[molpro-user] maxiter puzzle in multi

[Haitao Ma]

Dear Molpro-user,

    I got a puzzle as following, which had been posted here last year but 
no reply. So I re-sent it here to get more suggestion.
    *****from  CERMM, Concordia University*********
    We are experiencing trouble getting our CASSCF calculations to
converge. We know that the selected active space is appropriate for
our system (O4), and we employ equilibrium geometry for different
isomers.
We usually get the calculation to converge more
easily by doing successive calculations with increasing active space
and required number of states.
In such calculation,

     casscf;
       occ,   20;
       closed,8;
       wf,32,1,0; state,1;
       wf,32,1,2; state,7;
       wf,32,1,4; state,1;

     casscf;
       occ,   20;
       core,8;
       wf,32,1,0; state,5;
       wf,32,1,2; state,7;
       wf,32,1,4; state,1;

     casscf;
       maxiter,100;
       IPRINT,MICRO;            !print information on convergence problem
       occ,   20;
       closed,8;

the first casscf achieves convergence, but the next two won't, and
both print something like:

 ITER. MIC  NCI  NEG     ENERGY(VAR)     ENERGY(PROJ)   ENERGY CHANGE     
GRAD(0)  GRAD(ORB)   GRAD(CI)    
STEP       TIME

   1   40   29    0    -299.46358617    -299.46327296    0.00031321    
0.00605306 0.00000000 0.28965734 
0.16D-01    907.83

 *** NO CONVERGENCE ***

 We faile to understand what parameters will determine the number of
 interations, and what should be done to increse the number of
 iterations. As you can see on below the second calculation, the maxiter 
did
 not seem to help.

 CONVERGENCE PROGRESS
 --------------------

 ITERATION                 GRADIENT                       ENERGY                    
CHANGE IN ENERGY
     0                    0.001496911537             -299.463625399664
     1                    0.000000000000             -299.436598472010                
0.027026927654   
(PROJECTED)

 ** WVFN ****  START OF ITERATION  2 AT TIME  1240.64

 Entering internal optimisation at time  1254.91

            Time  Orb. grad.   Orb. change   CI gradient    CI change         
Energy      DIIS

   1  1  1298.62  0.00007543    0.00000000    1.00000000    0.00000000   
-299.43659916     0
   1  2  1298.66  0.00000020    0.00060323    1.00000000    0.00000000   
-299.43659919     0
 Too many P-space configurations for threshold      0.40
 Actual threshold used:      0.34
   2  1  1406.90  0.00086038    0.00060336    0.00391926    0.00084597   
-299.40184999     0
   2  2  1406.94  0.00000689    0.00873427    0.00391926    0.00000000   
-299.40185670     0
   3  1  1406.99  0.00000001    0.00874368    0.00000000    0.00000000   
-299.40185670     0

 CPU FOR TRAOPM:      1.33 ( 490.7 MFLOPS)  NOP_CORE=  40 NOP_FILE=   0 
MINWK=   612061  NWUSED=  6372060 
NWL= 89999500

 End of internal optimisation at time    1408.33

 ENERGY =   -299.4018566977  ENERGY CHANGE =   0.0617687020     
0.0347417743 (PROJECTED)  GRADIENT =  
0.72E-02  STEP =   0.25E-01  CONV = 1

 *** NO CONVERGENCE ***


Any insights appreciated.