Atomic Structure of Heterojunctions

An ideal heterojunction consists of a semiconductor crystal (in the sense of a regular network of chemically bonded atoms) in which there exists a plane across which the identity of the atoms participating in the crystal changes abruptly. In practice, the ideal structure is approached quite closely in some systems. In high-quality AlGaAs-GaAs heterojunctions it has been found that the interface is essentially atomically abrupt [6]. There is an entire spectrum of departures from the ideal structure, in the form of crystalline defects. The most obvious cause of such defects is mismatch between the lattices of the participating semiconductors. The lattice constants of GaAs and AlAs are nearly equal, so these materials fit together quite well. In contrast, the lattice constants of Si and Ge differ significnatly, so that over a large area of the heterojunction plane, not every Si atom will find a Ge atom to which to bond. This situation produces defects in the form of dislocations in one or the other of the participating semiconductors, and such dislocations usually affect the electrical characteristics of the system by creating localized states which trap charge carriers. If the density of such interfacial traps is sufficiently large, they will dominate the electrical properties of the interface. This is what usually happens at poorly controlled interfaces such as the grain boundaries in polycrystalline materials. The term heterojunction is usually reserved for those interfaces in which traps play a negligible role.

From the above considerations one would logically conclude that closely matching the lattice constants of the participating semiconductors (good ``lattice matching'') is a necessary condition for the fabrication of high-quality heterojunctions. Indeed this was the generally held view for many years, but more recently high-quality heterojunctions have been demonstrated in ``strained-layer'' or pseudomorphic systems [7,8]. The essential idea is that if one of the semiconductors forming a heterojunction is made into a sufficiently thin layer, the lattice mismatch is accommodated by a deformation (strain) in the thin layer. With this approach it has proved possible to make high-quality heterojunctions between Si and GeSi alloys [4].

Heterostructures are generally fabricated by an epitaxial growth process. Most of the established epitaxial techniques have been applied to the growth of heterostructures. These include Molecular Beam Epitaxy (MBE) [6] and Metalorganic Chemical Vapor Deposition (MOCVD) [9]. Liquid Phase Epitaxy (LPE) is an older heterostructure technology, which has largely been supplanted by by MBE and MOCVD because it does not permit as precise control of the fabricated structure.

The examples of heterojunctions cited so far involve chemically similar materials, in the sense that both constituents contain elements from the same columns of the periodic table. It is possible to grow heterojunctions between chemically dissimilar semiconductors (those whose constituents come from different columns of the periodic table), such as Ge-GaAs and GaAs-ZnSe, and such junctions were widely studied early in the development of heterostructure technology [10]. There are, however, a number of problems with such junctions. Based upon simple models of the electronic structure of such junctions, one would expect a high density of localized interface states due to the under- or over-satisfied chemical bonds across such a junction [11,12] . More significantly, perhaps, the constituents of each semiconductor act as dopants when incorporated into the other material. Thus any interdiffusion across the junction produces electrical effects which are difficult to control. For these reasons, most recent work has focused upon chemically matched systems.

If a heterojunction is made between two materials for which there exists a continuum of solid solutions, such as between GaAs and AlAs (as AlGaAs exists for all x such that ), the chemical transition need not occur abruptly. Instead, the heterojunction may be ``graded'' over some specified distance. That is, the composition parameter x might be some continuous function of the position. Such heterojunctions have desirable properties for some applications.