force field, energy function

The abinito application uses a number of different energy functions, depending on which stage it's in. For the final fullatom stages, the weekly releases should be using the talaris2013 scorefunction. The earlier stages use various centroid scorefunctions, such as score0, score1, score2, score3, and score5. These are very similar scorefunctions, but differ in turning up various scoreterms gradually.

There isn't yet a good comprehensive reference for the Rosetta energy function, instead, when it's described it's described in bits and pieces. Particularly, for the early stages of abinitio, you'll want to look at the various abinitio papers which have come out of the Baker Lab. The various CASP analysis papers typically have a good overview about how the protocol works. Note that you may need to go back quite a ways to the early (late 90's) papers about Rosetta to catch all the details.

Regarding zinc, there's a more fundamental issue in that Rosetta ab inito is mainly written around folding of single chain proteins. If you're interested in folding a protein containing zinc, your best bet is probably to fold it without the zinc, and then add the zinc in later. If you know which residues are involved in zinc binding, you may want to add in constraints between them during the folding process to make sure they end up in a position which can bind zinc. (That said, you may want to look at the recent papers about CASP from the Baker lab - there may be the beginnings of use of metal binding information in some of the more recent protocols.)

Even after, you'll probably need to have constraints to enforce the geometry. While talaris2013 can score zinc atoms, the knowledge about geometry and the like is still a little crude, so it may need to be supplemented with external information in the form of constraints.