Wet oxidation for optoelectronic and MIS GaAs devices
molar volumes of the starting semiconductor and a dense, crys-
talline oxide. AlAs shrinks by about 13% while Al
0.98
Ga
0.02
As
shrinks only 6.7% upon oxidation. Analysis of oxidised films
has revealed the presence of appreciable quantities of hydrogen,
suggesting the presence of a considerable amount of hydrous
species such as AlO(OH) in the oxidised layer. This is consist-
ent with thermodynamic predictions. Since the relative amounts of
AlO(OH) and Al
2
O
3
are not well defined and may change depend-
ing on the specific process conditions, a priori prediction of film
volume is not strictly possible at this time. Similar behaviour is
expected for AlGaAs materials.
A serious mechanical problem can exist when oxidation is taken
to completion, i.e. when the oxidation fronts from opposing mesa
walls coalesce. The oxide/semiconductor interface is weak, and
the upper layers of the structure will delaminate from the oxide
unless a portion of unoxidised structure is present to hold the device
together. This is naturally the case for oxide-aperture VCSELs, but
use of wet oxidation for GOI (GaAs-on-insulator) devices will face
serious problems unless provision is made for holding the device
together. This problem is most pronounced for AlAs and is less
problematic with AlGaAs alloys.
Changes in composition mean changes in dimension, producing
stresses, especially in overlying layers. These can deform if they
are not too physically constrained by a thick semiconductor over-
layer. As seen in FIGURE 10.2, the oxidation-front profile tends to
be slightly deeper at the top interface than at the bottom interface.
This may be related to stress, although the origin of the effect is
not definitively established and the effect is rather small.
It is generally not possible to oxidise a certain distance into
a layer, pull the wafer from the furnace and then reinsert the
wafer to oxidise further for a controlled final depth. Wet oxida-
tion under these conditions is not a simple time-additive process.
There appears to be a change in the chemical or structural nature of
the oxide-semiconductor reaction front that discourages restarting
the wet oxidation process, perhaps by inhibiting the free access
of water to AlAs or AlGaAs for the generation of the H that is
essential for the rapid removal of As from the oxidising layer.
When inserting a wafer into a furnace with wet gas flow, there
may be a delay in the onset of rapid wet oxidation (FIGURE 10.6).
This is more of a problem for depth control of shallow oxidations,
since the apparent incubation time can be a substantial fraction
of the total time in the furnace. For such processes, it may be
especially important to load into dry gas, thermally equilibrate
to the final reaction temperature and then introduce the wet gas
flow. Since the heat capacities of dry nitrogen and water-saturated
nitrogen are different, it may be necessary to usedifferent flowrates
318