Enrol a graphite surface, lap material off it a micron at a time, and watch whether the identity survives. The maths underneath is the surface model from pebbleid/surface.py: a depth-correlated microstructure, three-light slope recovery, batch mould subtraction, band correlation. Surfaces are simulated to published graphite roughness and wear rates.
One pass of the Xe-100 is about 1 µm by mass balance, up to 6.7 at the top of the bracket. Re-enrol here does what the plant does on every pass: rewrites the template from the current read, so the next lap is measured from here.
κ is the one number nobody has measured. Published pore shapes point below 0.5. Drag it to 0.1 and lap: that is the pessimistic case.