At a heterojunction the constituents of which share neither common anion nor common cation, the interface can be considered as a very thin interlayer of a third material. Depending on the band alignments this built-in interface layer may be a thin quantum well for both electrons and holes. In this case it can trap carriers at the interface. We show why this effect is mainly significant at type-II heterojunctions. The effect is thus twofold: modification of the apparent band offset (related to the effective band gap), and enhancement of the overlap integrals between conduction and valence states, which leads to high transition probabilities. Here we first review a lot of heterojunctions likely to present this phenomenon. Then we show that this can be observed for ${\mathrm{Al}}_{\mathit{x}}$${\mathrm{Ga}}_{\mathit{y}\mathrm{\ensuremath{-}}\mathit{x}}$${\mathrm{In}}_{1\mathrm{\ensuremath{-}}\mathit{y}}$As/InP systems, and we detail the case of ${\mathrm{Al}}_{\mathit{x}}$${\mathrm{In}}_{1\mathrm{\ensuremath{-}}\mathit{x}}$As/InP. We provide a semiempirical tight-binding calculation which includes strain effects at the interface. The resulting local density of states clearly demonstrates whether electrons or holes are trapped at the heterojunction.

Energy-band lineups at several [100] heterojunctions of III-V semiconductors are calculated using a self-consistent tight-binding treatment. The calculations exhibit transitivity to within 0.2 eV for ${\mathrm{In}}_{1\mathrm{\ensuremath{-}}\mathit{x}}$${\mathrm{Ga}}_{\mathit{x}}$As/${\mathrm{In}}_{1\mathrm{\ensuremath{-}}\mathit{y}}$${\mathrm{Al}}_{\mathit{y}}$As/InP, GaAs/InAs/InP, GaAs/GaP/InP, and GaSb/GaAs/InAs. For ${\mathrm{In}}_{1\mathrm{\ensuremath{-}}\mathit{x}}$${\mathrm{Ga}}_{\mathit{x}}$As/${\mathrm{In}}_{1\mathrm{\ensuremath{-}}\mathit{y}}$${\mathrm{Al}}_{\mathit{y}}$As/InP, the results are in good agreement with experimental data. For heterojunctions where the constituents share neither a common anion nor cation, the two possible interfaces do not necessarily lead to a single-band offset. This, and also the strain configuration, has to be considered when applying the transitivity rule (that is, the fact that for three semiconductors, A, B, and C, the band offset at the heterojunction A/B can be deduced from the band offsets at the heterojunctions A/C and C/B), provided we take care that the material C corresponds to the material at the interface.

Without the introduction of any external atom, the heterojunction band offsets can be modified by the interface configuration. This phenomenon has been studied, for III-V (100) heterojunctions, from both theoretical and experimental points of view (experiments refer to the lattice-matched GaInAs/InP and GaInP/GaAs systems). The modulation of the band offset is of the range of about 100 meV and only appears when the constituents share neither common anion nor common cation. From a theoretical point of view, the variation of the band offset is related to the modification of interface dipoles due to the chemical rearrangement and to the strain redistribution around the interface.

The variations of the energy levels in a conduction quantum well of InP/InAs/InP versus its width are calculated using two different approaches: a self-consistent tight-binding calculation and an effective-mass description. The introduction of nonparabolicity appears to have a noticeable effect. The importance of taking into account the nonparabolicity becomes essential in the estimation of the equivalent parallel effective mass, which characterizes the parallel transport in such systems.