The mechanism of thermal oxidation of silicon in dry oxygen was studied using 18O as a tracer. layers first grown in natural oxygen were grown further in 18O‐enriched oxygen up to thicknesses ranging from 5 to 26 nm at 930°C. 18O profiling was carried out using the nuclear reaction microanalysis. Fixed 18O is found near the interface (18O FNI) and near the external surface of the oxide (18O FNES). In this work we studied: (i) the possible origins of18O fixation near the external surface of the oxide (18O FNES) and (ii) the possible growth mechanisms of films of thicknesses ranging from 260 to 5 nm. It was shown that 18O FNES does not seem to be due to some interfering phenomena (network oxygen exchange with water vapor traces or effects connected with the experimental procedure) and is not due to a growth mechanism by Si transport. A step‐by‐step motion of network oxygen atoms seems to be the more likely to explain 18O FNES. The amount of 18O FNES becomes more important as the original thickness of the film becomes smaller. Even for thicknesses under 20 nm we always found large amounts of 18O fixed near the interface, which corresponds to growth by oxygen species migrating through the oxide network and without measurable interaction with it (very likely molecules for thick oxide samples). However for ∼5 nm thickness films, the amounts of 18O FNES can reach 50% and 80% of the total amount of 18O at pressures equal to 10 torr and 10−1 torr, respectively.