Geometry optimization for weakly bound systems

[Dave Moore]

    I am trying to do geometry optimizations for weakly bound systems,
and I have noticed the following behavior.  When the Hessian element
becomes zero for a free parameter in the optimization, the value of the
parameter is set to zero for the next step, which then usually crashes.
A perhaps related phenomenon is that even for non-zero Hessian elements,
when they are small compared to other Hessian elements, often (but not
always), the corresponding geometry parameters are set to unphysical
values (overlapping atoms, etc.).  I have read the manual section on
geometry optimizations, and I have tried some of the METHOD options, but
nothing seems to help.  In the past I have always been able to overcome
these difficulties by choosing a better starting point for the
optimization, but in my current case I am stumped.  I would appreciate
any observations or suggestions concerning this matter.  Thanks,

Dave Moore
Chemistry Dept.
UNC Chapel Hill
USA
919-962-1579