How can we predict the geometry of a coordination compound?

[cncbmb]

This isn't a homework question, and I've always wondered about this. My chemistry teacher told me to look up "coordinate chemistry" in a textbook, but the textbook did not have the answer.

How can we predict the geometry of a coordination compound?

I googled my question first, and I found that Lewis and VSEPR don't work for transition metal coordination compounds: http://answers.yahoo.com/question/index?qid=20080405102937AAOh7Gd
"If your teacher thinks you can rationalize that using a Lewis structure and VSEPR principles, I'd like to see it. Pt2+ is 8 electrons, NH3 and Cl- each donate two to the bonds, that gives you 16e total around the Pt, 8 involved in bond pairs. So, that's what, AX4E4? You certainly can't use that to predict square planar, which is the observed structure. Neither can you explain why [PtCl4]2- is square planar but [NiCl4]2- is tetrahedral, even though they have the same number of valence electrons and (presumably) the same Lewis structure. Lewis and VSEPR don't work for TM compounds."

I guess does that mean we can somehow predict the geometries of non-transition metal coordination complexes?

 

[Borek]

I am not sure if we do have a universal theory that can be used to predict structure of every coordination compound.

 

[alxm]

Sure we do, Borek - (relativistic) quantum theory. Every other model is just an approximation of that, anyway.

Which is the problem, really. Any simple model is at best an approximation, but the effects that govern transition-metal coordination (primarily the relative energy and splitting of d-orbital energy levels) are relatively small. Which doesn't mean that they're unpredictable. Just that you need a pretty exact model to do it. In the case of platinum and other heavy metals, you also have to take into account relativistic effects and the f-orbitals.

 

[Borek]

Sure we do, Borek - (relativistic) quantum theory.

Honestly - I was going to write something like that, but decided against