I’ve been working on an internal UFF implementation off-and-on over the course of the spring. (This term was busier than expected with work obligations unfortunately.)
The positives:
- there’s always an implementation available (beyond Lennard-Jones)
- should be faster than Open Babel (benchmarks pending)
- very helpful for the auto-optimize tool
- gradients are needed anyway for bond / distance / angle / torsion constraints
The energies were implemented back in February, but the analytical gradients were a pain until I figured out the problem over the weekend while comparing analytical to numeric results.
I’m inclined to merge this for further testing in the nightly builds. Presumably everyone will let me know if it distorts their geometries in weird ways. 