CCL:G: opt=Z-matrix convergence problems in Gaussian

 Sent to CCL by: "George B Fitzgerald" [gfitzgerald : accelrys.com]
 I dont know specifically about Gaussian, but I've seen energy bouncing between 2
 values in other applications. In the cases I've seen, it was oscillating between
 2 nearly degenerate states. It sounds like you were ok when you started the
 optimization, and then the geometry found a spot on the potential energy surface
 where these 2 states are near degenerate. I solved my problem by going to an
 open-shell calculation, which resulted in a lower energy and no more
 oscillation. If you're currently running closed shell, you might try that.
 -george
 -----Original Message-----
 Sent to CCL by: Tanja van Mourik [tanja.vanmourik]=[st-andrews.ac.uk]
 Hi All,
 We are trying to optimise a cluster consisting of ~50 molecules with M06-L with
 Gaussian 09. We are using Opt=Z-matrix, as optimisation in redundant internal
 coordinates does not work ("Error in internal coordinate system").
 With Opt=Z-matrix, the energy appears to converge nicely up to 100 cycles, but
 after that the energy starts to oscillate wildly. This happens always after 100
 cycles. Does anyone know why this is, and how to solve this problem? Currently
 we restrict MaxCycle to 100, and restart the calculation repeatedly (it needs
 many hundreds of optimisation cycles to converge, particularly with
 Opt=Z-matrix), but I am hoping for a better solution. Any ideas?
 Best wishes,
 Tanja