Catastrophic failure in diagonalization(hsdel)???

[Yann Tambouret]

Dear Molpro-users,

I'm trying to calculated the potential energy surfaces of O2 ground and 
first few excited states. I run Multi, Ci (Natorb), and CI (several 
states, lowest lying roots). For certain bond lengths, this final 
program execution returns an error:

Catastrophic failure in diagonalization(hsdel)
The expansion set has become singular
This difficulty can arise for many reasons
Sometimes it helps to redefine P space
Otherwise, try increasing or decreasing reference space or nstati

Can anyone explain to me why this happens?

If it would help, please email me for the full input/output files:
yannpaul at chem.bu.edu

Thanks,

Yann Tambouret

(question 1)