If you use DFT and the OptimizeGeometry function, we have pseudopotenitals and basis sets for pretty much all elements in the periodic table. With ATKClassical you also have a Tersoff potential for Si+O (), so the QuickOptimizer can indeed be used (as well as OptimizeGeometry). I suspect for a system of the type and size you indicate that will be your best option - the classical potential should be fairly accurate, and very fast. Then it depends on what properties you want calculate for the structure of course, if your choice moving forward is DFT or stick with classical (for MD, and thermal or mechanical studies for instance).
