Here's a silicon nanocube with unreconstructed surfaces (i.e. no dimer rows). Although we haven't figured out yet how to index zones (viewing directions) or lattice planes in jmol, as you can see we have figured out how to highlight selected structures like the individual red, chartreuse, cyan and indigo dimer backbones in this model. If the silicon surface is not quite parallel to the lattice, these backbones necessarily form plateaus separated by "steps" on the surface. The "social behavior" of steps is another story. For example, steps whose edge runs parallel to exposed dimer backbones are most stable.
Can you measure the vertical distance from one backbone (e.g. the red zig-zag on the surface) to the perpendicular backbone below it (like the chartreuse zig-zag)? Steps of this height are seen routinely in atomic force microscope images of such silicon. Note also that the sequence of dimer rows doesn't repeat until one reaches the pink atom four backbones beneath the crystal's surface e.g by climbing down the pink-orange-green-blue-pink spiral staircase. Formation of the 2x1 Si{100} reconstruction involves the displacement and pairing up of dimer backbones (e.g. the red backbone with one of the two grey ones adjacent) to create dimer rows. Tip: Double-clicking on two atoms draws a scale bar between them, and using this trick you can sketch in "stretched" dimer row bonds on all six surfaces of the cube if you like.
You can bond and unbond one of the dimer rows by clicking on the buttons above. This "surface reconstruction" process in practice seems to stablize exposed silicon surfaces with this {100} or square orientation. Click here for more indexable version of the same nanosilicube model using a different applet. For clues to how the various colors are related, click here. We may eventually add dimer-row reconstructions to the two larger "off-{100}" silicon models here and here.