The paper describes in the first part the formation of carbon flakes up to
10-20 mu m thickness (average growth rate 2 nm/s) on the graphite tiles of
the toroidal belt limiter. This occurred as a consequence of a slight chang
e of the geometry and turned parts of the surface area from net erosion int
o net deposition zones. The possible influence of the morphology on this be
haviour is discussed in the second part by means of an erosion experiment w
here the gradual disappearance of a boron substrate could be discriminated
from the simultaneous carbon deposition on the surface. The two counter-act
ing processes co-exist within 10-30 mu m distance and lead to an extremely
non-uniform carbon deposition even in net erosion zones. The carbon agglome
ration coincides with surface imperfections, e.g. grooves, but agglomeratio
n by temperature enhanced mobility is not excluded. The changeover from net
deposition to net erosion averaged over larger distances can still be obse
rved and is due to the hydrogen and carbon fluxes in the SOL. This is confi
rmed by Monte-Carlo code calculations.