The composition dependence of the ${\Gamma }_8$-${\Gamma }_6$ transition in mercury cadmium telluride (MCT) has long been classified apart among all disordered alloy systems. The nonconstant bowing parameter reported in this case could only be accounted for in terms of a finite departure from the virtual-crystal approximation but was not clearly reflected in the properties of carriers and/or the characteristics of the alloys. Moreover this nonconstant bowing appears to be a unique feature of MCT. First, it departs from all reports concerning standard II-VI and III-V materials which, even for structurally disordered systems like In-Ga-As, still exhibit a constant bowing and a parabolic dependence of the fundamental absorption edge on alloy composition. Second, it is not easily reflected in many band-structure calculations. Most of the time, even if they have been performed in the light of the coherent-potential approximation (CPA), they fail to reproduce such a complicated scheme.

As a matter of fact, because of the closer lattice matching achieved in Hg-Cd-Te than in Hg-Zn-Te, these calculations suggest that less deviation (or bowing) from the straight-line average between the two binary compounds should exist in Hg-Cd-Te than in Hg-Zn-Te. In both cases, they predict a parabolic (or nearly parabolic) behavior. This rather large discrepancy between the experimental results and the theoretical calculation is one of the characteristics of MCT. To clear up this point, we have investigated a series of samples with concentrations in the range x=0.5–1 (where x=1 corresponds to CdTe). We resolve a series of excitonic features and get a clear, unambiguous determination of the first interband transition at different alloy compositions. This gives a constant bowing parameter (i.e., a simple parabolic dependence) with a maximum departure from linearity of 33 meV at x=0.5. We discuss these results in the light of the virtual-crystal approximation, using the empirical pseudopotential method, and compare them with recent CPA predictions where both diagonal and off-diagonal disorder terms had been included. We find a surprisingly good agreement which demonstrates, on a purely experimental basis, the near-cancellation of disorder effects expected at the band edge in mercury cadmium telluride.

• Received 18 March 1988

DOI:https://doi.org/10.1103/PhysRevB.38.3948