Classical molecular dynamics (MD) simulations have been employed to model the fabrication of very small ( < 5 nm ) features in silicon. As scale-down continues in the semiconductor and other thin-film device industries, fundamental understanding of etch mechanisms at these smaller scales is becoming increasingly important for process design and control.  In our MD approach, confined beams of ions and radicals have been used to approximate a perfect masking layer, using either cylindrical (hole) or rectangular (trench) geometries with a nominal confinement size of ~2 nm.  Bombarding species (chosen from typical plasma processing gases) include Ar⁺ and CF_x⁺ ions, as well as F and CF_x radicals. The nano-feature evolution is tracked over several thousand impacts, allowing direct examination of the etch process as a function of ion/radical fluence. By varying the ratios of the incoming species, various physical and chemical contributions can be examined and their corresponding roles in the etch process and sidewall formation can be studied.  For example, we show that ion bombardment alone is sufficient to create a dense, mixed damaged/passivation layer along the sidewalls of the feature being etched that is ~1 nm thick.  However, for some bombarding chemistries, radical deposition along the feature also plays an important role in determining the final sidewall composition and subsequent etch characteristics. We elucidate which mechanisms are most critical under the various conditions studied.  Additional simulations have been carried out to examine the effects of an explicit masking layer (i.e. amorphous carbon deposited on top of the silicon) on both the final geometry of the feature and to overall feature size limitations. For example, ions impinging on the masking layer's sidewalls can lead to sputtering of the masking material into the feature, where it mixes into the substrate layer during subsequent bombardment, changing the effective etch rate.  Further, the mask can collapse at the very small length scales examined, limiting the transport of radicals and ions to the substrate material.  Mask erosion under certain chemistries can lead to loss of critical dimension control or mask impurity incorporation into the substrate. Future developments to capture more realistic phenomena (such as surface charging effects, longer-than-MD-timescale diffusion effects, etc.) are also discussed, including the parallelization of the code to allow for atomistic simulation of larger scale features.