In nearly all cases when an epitaxial layer of HgCdTe is grown on a CdZnTe
substrate, there will be a finite lattice mismatch due to the lack of preci
se control over the ZnTe mole fraction. This leads to strains in the layer,
which can be manifested in one or more ways: (1) as misfit dislocations ne
ar the interface, (2) as threading dislocations, (3) as surface topographic
al textures, and (4) as crosshatch lines seen by x-ray topography. We have
found that much of the strain can be relieved by growing on a reticulated s
ubstrate. Specifically, when the substrate has been etched to form mesas pr
ior to growth of the layer, the resulting layer on the tops of the mesas sh
ows evidence of significantly reduced strain. CdZnTe substrates oriented (1
11)A were prepared with two sets of mes as on 125 mu m centers and 60 mu m
centers, and with other planar areas remaining for comparison. From a Hg me
lt, a layer of LWIR HgCdTe was grown about 16 mu m thick on each substrate.
Nomarski microscopy showed that the layers on the mesa tops were extremely
flat, showing no sign of curvature or surface texture. X-ray topography sh
owed no cross hatch on the mesa tops, while the usual cross hatch appeared
in the planar regions. The LPE layer extended laterally beyond the edges of
the original mesa because of faster growth in non-(111) directions. Sample
s were cleaved and examined in cross section. The linear density of etch pi
ts seen in the cross section near the substrate, which represent misfit dis
locations, was three times lower in the layer on the mesas than in the laye
r in the unpatterned region, although both regions have the same layer/subs
trate lattice mismatch. When an epilayer is grown on an unpatterned wafer (
the conventional approach), the growth in any small region is confined late
rally by the growing layer in the neighboring regions. However, when growth
occurs on a reticulated surface, the lateral confinement is removed, provi
ding strain relief and fewer defects.