In this work, we
have quantitatively elucidated the source of the
hydrogen content in the atomic layer deposition of Al2O3 at different temperatures (80–220 °C), by replacing
the H2O precursor with heavy water (D2O) to
use as a tracer and discern between the H coming from the unreacted
metal precursor ligands and that from the unreacted −OD (hydroxyl)
groups coming from the (heavy) water. The main source of impurities
arises from the unreacted hydroxyl groups (−OD), reaching ∼18
atom % of deuterium at a deposition temperature of 80 °C. Reconsidering
carefully our own and literature experimental data, we concluded that
the generally accepted mechanism of steric hindering by monodentate
Al(CH3)2 adsorbates (dimethylaluminum) cannot
be solely responsible for the retention of hydroxyls during atomic
layer deposition (ALD). On this regard, we propose two additional
mechanisms that can lead to sterically hinder hydroxyl groups which
will then remain unreacted in the film: surface rehydroxylation resulting
in the reconfiguration of bidentate or tridentate adsorbates into
monodentate adsorbates and hindered subsurface hydroxyl groups during
the (heavy) water pulse and the hydroxylation of sterically hindered
dissociated methyl chemisorbed species. Based on these three steric
hindrance mechanisms, we constructed a growth model that consists
of the initial chemisorption configurations of trimethyl-aluminum
molecules with the alumina surface and the subsequent reconfiguration
of the resulting adsorbates into a monodentate configuration that
consequently leads to sterically hindered hydroxyl groups. The fraction
of AlOx adsorbates arranged in monodentate and bidentate configurations
entails a specific number of O/Al atoms and unreacted hydroxyl groups
inside the film. This model was able to explain the deuterium content,
the O/Al ratio, and the density obtained from Rutherford back-scattering
and heavy ion elastic recoil detection analysis measurements. Furthermore,
this model was able to predict more accurately the growth per cycle
to what has been reported to be the ALD window of alumina. Our findings
will spur further detailed investigations of the reaction and growth
modes in ALD films.