In this case I recommend reaction path -like searching algorithms. There are quite a few available (the NEB- nudge elastic band or the string algorithm). Also, Gaussian has a built in and quite powerful "Synchronous Transit-Guided Quasi-Newton (STQN) Method" implemented. Have a look at the documentation and try it out for your example:

http://www.gaussian.com/g_whitepap/qst2.htm

In this case I recommend reaction path -like searching algorithms. There are quite a few available (the NEB- nudge elastic band or the string algorithm). Also, Gaussian has a built in and quite powerful "Synchronous Transit-Guided Quasi-Newton (STQN) Method" implemented. Have a look at the documentation and try it out for your example:

http://www.gaussian.com/g_whitepap/qst2.htm

I would try

1) to reduce the number of degrees of freedom by feezing the less important parts of the system.

2) to do a guided (manual) scan for the ligand exchange by using the "chemical intuition" for identifiing the most suitable initial geometries for the TS search.

3) in such indentified suitable structures I would iteratively try to find a structure with the proper Hessian structure and

4) finaly I would start with one of the eigenvalue following TS search mechanisms

Its tedious, but usualy it helps

PS flexibility of the basis set and of the method is required. The Cu electronic structure will change in course of the reaction. Check the (s-d)excitation

Dear Lucas, thinking naively - would not the high stability of the six coordinated adduct point towards other mechanism than association? I agree with Jan that "chemical intuition" should help here. I would expect that with nitro groups in the aromatic ring this phosphate is a rather poor donor? Have you considered the a dissociative mechanism, with the first transition state being one with four coordinated copper and a mono coordinated phosphate?

I am agree with Jan, if you have some troubles to find the TS you may try to perform a "scan" of the H2O-Cu distance and for sure you will find the distance when the TS takes place. I also recommend to fix some atomic positions that you think that doesn't change in the optimization process. When you get the TS distance you only have to perform the frequency calculations on that point.

pseudo Newton-Raphson methods are in this case and in my opinion the best choice to try solving such an issue (and in general, once a given chemical intuition is present along with a detailed knowledge of the peculiar system under investigation). So you should drop automatic algorithm and start considering any possible reaction coordinate your system could go through to react.

This means that associative mechanism is surely a possibility, but this could be partially concerted, so you should attempt a progressive income of the first water and a concomitant elongation of the Cu-OP bond.

Alternatively, it cannot be excluded that a partial elongation comes first thus activating the water attack to copper.The method is actually the one already suggested by other researchers who answered the question:

1-choice of the proper reaction coordinate (which changes with the type of mechanism)

2-freeze the degrees of freedom you deem to correspond to the maximum extent of variation during the reaction

3-do an optimization (with the frozen DoF)

4-check the Hessian structure: if it correctly displays only one imaginary frequency associated to the desired vibration mode, then you just have to relax what was kept frozen previously, if not you have to go back to point one and select a new set of DoF to freeze