Michael G. Mauk
Newark, DE 19716-2000
In the standard technology of semiconductor epitaxy, electronic and optical device structures are made by the growth of multilayers of planar, crystalline semiconductor films on flat, highly-polished, oriented crystalline substrates. The substrate acts as a seed crystal for the growth of the semiconductor film, but otherwise, functions merely as a passive support for the deposited semiconductor layers. Many new device designs and improvements in material quality are possible, if instead, the semiconductor is grown on a substrate that has been first masked with a passive coating such as an oxide film, then patterned with openings in the mask that expose the underlying substrate. The openings serve as sites of preferential growth; there is virtually no nucleation on the mask itself. Typical lateral dimensions for the mask patternings range from 1 to 1000 microns; with crystal thicknesses of 1 to 100 microns. The resulting preferentially-seeded crystals have quite different shapes and properties compared to crystals grown on unmasked substrates. Such crystal growth on patterned, masked substrates has been termed selective epitaxy orepitaxial lateral overgrowth (ELOG).
From one point of view, the study of selective epitaxy characteristics (e.g., nucleation density, selectivity, aspect ratios, faceting, step coverage, and growth rates) is very useful in elucidating the growth mechanisms of a particular materials system and process. Here, we are interested in selective epitaxy as a diagnostic tool in characterizing and trouble shooting a crystal growth process. Using standard thin-film deposition, photolithography, and etching techniques, a wide range of mask materials and pattern geometries are at the disposal of the crystal growth experimentalist. As a specific example, a single selective epitaxy experiment can show the relative importance of surface versus volume diffusion, growth rate anisotropies, critical supersaturation for heterogeneous nucleation, and the mechanism of interface attachment kinetics.
Selective epitaxy and ELOG have been demonstrated in a number of semiconductor systems using several different growth techniques, including vapor-phase growth, solution growth, and electrodeposition. Our objective is to understand and predict the shape (crystal habit and morphology) of selectively-grown crystals based on a knowledge of the relevant phase diagram and other materials and transport properties; and control of process parameters such as growth temperature, cooling rates, dopant impurities, and mask design including mask materials, thicknesses, and pattern geometries. For example, a shape function that represents the cross-sectional profile of the selectively-grown crystal as a function of growth time would provide a convenient connection between experimental observations and mathematical models of the crystal growth process. In selective epitaxy, there is much latitude for control of the interplay and relative importance between two-dimensional surface diffusion and three-dimensional bulk diffusion, heterogeneous and homogeneous nucleation, interface attachment kinetics, and growth rate anisotropies. The general problem is probably intractable, but limiting cases can be illustrative and are seen in practice. At any rate, there are ample experimental results exhibiting many diverse features and a wide range of behavior, at least some of which should be fair representations of simplified models.
Ultimately, we would like a model of selective epitaxy, for which crystal growth processes can be tailored to achieve targeted device structures. We need to know what sort of crystal shapes are possible, and how to optimize the process for achieving the desired device structures. As a future endeavor, but perhaps worth at least considering at this stage, is an analysis of device physics for semiconductor structures made by selective epitaxy. There are significant similarities in the solute mass transport and boundary condition geometries involved in modeling the growth, and the optics and minority carrier charge transport operative in device. This suggests perhaps a more holistic approach to semiconductor technology where crystal growth, optics, and device operation are part of integrated model.
- cut some (100) and (111) GaSb substrates and (100) InAs substrates
- select mask with a pattern of holes
- photolith wafers for lift-off
- sputter deposit 1000 angstroms of TiN
- pre-epi cleaning