Additive fabrication doesn’t necessarily have to be done in 2D layers; a robot arm could possibly deposit material in 3D space in alternative geometrical schemes, maybe following the volumes of individual parts. But in any deposition model, the concept of resolution is critical to the quality and function of the thing being made. With layers, as they get thinner, the surface steps (the layer edges) become smaller; which is to say they become less rough.
In engineering the measurement Ra describes the roughness (or in a more beer glass half-full mood, smoothness) of a surface. Ra is surface Roughness average and is the average difference between peaks and valleys in a given surface region, measured in microns (or micro-inches in US units). A micron is a thousandth of a millimetre. At larger scales roughness is an everyday dimension for us humans — we can feel it and see it — but as Ra gets smaller it can still have a huge impact on the engineering and visual properties of a surface — in bearings, electrodes, mirrors and lenses, and in cosmetic finishes. A rough-turned steel bar or mill finish steel sheet may have an Ra of 25-50 microns, whereas for a hard steel bearing it should be between 0.2 and 0.8 microns. The Ra of window glass is 0.003, at which point it makes sense to give the measurement in nanometres (a thousandth of a micrometer, or a millionth of a millimetre), ie 3.0nm. The super-polished surface of a silicon wafer or hard-drive disk has Ra 0.1nm.
With conventional manufacturing it’s possible to achieve these kinds of surfaces by giving them special attention — grinding, lapping, polishing, and by coating and plating (or by moulding in tools treated these ways). But in an additive process where assemblies are made in-situ (an ambitious goal, we know) these high-function surfaces would ideally need to be made directly, at the ‘print head’. I’ve already described a hypothetical additive car built from 1micron layers, but it’s clear now that a range of resolutions would be needed, from 50 microns for some bulky parts to 1.0 nanometres for electronics (and that’s today’s electronic chips, by the way). One end of this range is 50,000 times courser than the other. You could print the whole car at 1.0nm resolution but that could take not ten years but ten thousand! Exactly. But it’s probable that at these small scales the surface roughness, the RA, is not in fact equivalent to the layer thickness. The droplets of material will merge together to form a wavy rather than ziggurat-stepped surface, so the Ra may be a fraction of the layer thickness, maybe more like a fifth. Perhaps? If true, the layer thickness range required becomes 250 microns down to 5.0 nanometres. A little better.
Julian Vincent — celebrated biomimetics pioneer — has already proposed here in his comment that a car-printing machine should have multiple print heads with a range of resolutions and material ‘specialisations’. Soon I plan to re-visit car printing, an impractical idea most likely, but extreme cases can sometimes help sort out ideas.