BMAT error in CX search

[Chris Evenhuis]

I am just trying to learn how to find conical intersections.  The
lih2_D0D1.com runs fine.

I have modified the lih2_D0D1.com to treat H3.  This involved changing
the wf cards in the hf and the multi sections,
wf,4,1,0 > wf,2,1,0 and wf,5,1,1 > wf,3,1,1 and removing the occ card.

The CP-MSCF section completes without error.  The OPT sections fails
complaining about the B-matrix.

When theta is left at 30.57744006 the procedure fails with
> PROGRAM * OPT (Geometry optimization step)     Author: F. Eckert
>
> Max step sizes:     0.150 (TOTAL)     0.200 (DISTANCES)     2.000  
> (ANGLES)
>
> Entering BMAT geometry-optimization
> Using points  1 to 11 from record 701.1,  wavefunction type=MULTI
>
> Making model hessian for BMAT coordinates
>
> WARNING: THERE ARE    1 INT.COOR., 3N-6=    3
> BMAT using opt RF
>
> BFGS update of hessian using  5 points
>
> Entering Rational Function Geometry Optimization
>
> ERROR EXIT

If put theta to 35.57744006 the error changes slightly
> *********************************************************************** 
> *
> PROGRAM * OPT (Geometry optimization step)     Author: F. Eckert
>
> Max step sizes:     0.150 (TOTAL)     0.200 (DISTANCES)     2.000  
> (ANGLES)
> Doing optimization in 3N cartesian coordinates
> Cartesian Coordinates are in [bohr]
> Creating new Variables:
> H1X1              0.00000000
> H1Y1              0.00000000
> H1Z1              2.10440418
> H2X2              0.00000000
> H2Y2             -1.01278993
> H2Z2             -1.05220209
> H3X3              0.00000000
> H3Y3              1.01278993
> H3Z3             -1.05220209
> Symmetry will not be used
>
> Geometry written to block  1 of record 701
>
> Entering BMAT geometry-optimization
> Using points  1 to  1 from record 701.1,  wavefunction type=MULTI
>
> Making model hessian for BMAT coordinates
> bread: error found
> first coordinate must be defined
>
> undefined int.coordinate type at no.    0
> ****************************************
>
> ERROR EXIT

I have tried all sorts changes, such as replacing the hf with rhf or
that step removing completely, but nothing seems to work.

Cheers,

Chris Evenhuis