Epitaxy

Two ancient Greek words eiu (epi—placed or resting upon) and 7a£if (taxis —arrangement) are the root of the modern word epitaxy, which describes an extremely important phenomenon exhibited by thin films. Epitaxy refers to extended single-crystal film formation on top of a crystalline substrate. It was probably first observed to occur in alkali halide crystals over a century ago, but the actual word epitaxy was apparently introduced into the literature by the French mineralogist L. Royer in 1928 (Ref. 1). For many years the phenomenon of epitaxy continued to be of scientific interest to numerous investigators employing vacuum evaporation, sputtering, and electrodeposi-tion. A sense of much of this early work on island growth systems (e.g., metal films on alkali halide substrates) was given in Chapter 5. Over the past two decades, epitaxy has left the laboratory and assumed crucial importance in solid-state device processing. Interest has centered on epitaxial films exhibiting layer growth. This chapter focuses primarily on such films as well as on several broader issues related to epitaxy.

Two types of epitaxy can be distinguished and each has important scientific and technological implications. Homoepitaxy refers to the case where the film and substrate are the same material. Epitaxial (epi) Si deposited on Si wafers is the most significant example of homoepitaxy. In fact, one of the first steps in the fabrication of bipolar and some MOS transistors is the CVD vapor phase epitaxy (VPE) of Si on Si (see Chapter 4). The reader may well ask why the underlying Si wafer is not sufficient; why must the single-crystal Si be extended by means of the epi film layer? The reason is that the epilayer is generally freer of defects, purer than the substrate, and can be doped indepen-dently of the wafer. A dramatic improvement in the yield of early bipolar transistors was the result of incorporating the epi-Si deposition step. The second type of epitaxy is known as heteroepitaxy and refers to films and substrates composed of different materials, e.g., AI As deposited on GaAs. Heteroepitaxy is, of course, the more common phenomenon. Optoelectronic devices such as light-emitting diodes and lasers are based on compound semiconductor heteroepitaxial film structures.

The differences between the two basic types of epitaxy are schematically illustrated in Fig. 7-1. When the epilayer and substrate crystal are identical, the lattice parameters are perfectly matched and there is no interfacial bond straining. In heteroepitaxy the lattice parameters are necessarily unmatched, and, depending on the extent of the mismatch, we can envision three distinct epitaxial regimes. If the lattice mismatch is very small, then the heterojunction interfacial structure is essentially like that for homoepitaxy. However, differ­ences in film and substrate chemistry and coefficient of thermal expansión can strongly influence the electronic properties and perfection of the interface. Small lattice mismatch is universally desired and actually achieved in a number of important applications through careful composition control of the materials involved. Section 7.4 deals with such epitaxial interfaces in compound semi­conductor and the devices based on them.

When the film and substrate lattice parameters differ more substantially, we may imagine the other cases in Fig. 7-1. Either edge dislocation defects form at the interface, or the two lattices strain to accommodate their crystal-lographic differences. The former situation (relaxed epitaxy) generally prevails during later film formation stages irrespective of crystal structure or lattice parameter differences. The latter case is the basis of strained-layer heteroepi-taxy. This phenomenon occurs between film-substrate pairs composed of different materials that have the same crystal structure. Lattice parameter differences are an order of magnitude larger than in the case of lattice-matched heteroepitaxy. Structures consísting of Ge^Si,^ films grown on Si, currently under active research study, are important examples of strained layer epitaxy, and will be discussed further in Section 7.3.

Recently, there has been a great deal of exciting research devoted to both the basic science of epitaxy and its engineering applications. One área has ad-dressed the dream of creating three-dimensional integrated circuits possessing intrinsically high device packing densities. Rather than a single level of processed devices, a vertical multifloor structure can be imagined with each level of devices separated from neighboring ones by insulating films. What is crucial is the ability to grow an epitaxial semiconductor film on top of an amorphous substrate—e.g., Si on Si0₂ as shown in Fig. 7-2 . This will require selective nucleation of epi Si at existing crystalline Si, and nowhere else, followed by lateral growth across surfaces that are ill suited to epitaxy.

Methods to achieve Si on insulator (SOI) epitaxy are mentioned in Section 7.5. A second área involves the fabrication of multilayer heterojunction composites. These remarkable epitaxial film structures include superlattices and quantum wells. Some of their simple properties together with applications involving incorporation into actual devices will be deferred until Chapter 14. The remainder of this chapter is divided into the following major sections:

Structural Aspects of Epitaxial Films

Lattice Misfit and Imperfections in Epitaxial Films

Epitaxy of Compound Semiconductors

Methods for Depositing Epitaxial Semiconductor Films

Epitaxial Film Growth and Characterization

Structural Aspects of Epitaxial Films

Single-Crystal Surfaces

Prior to consideraron of epitaxial films, it is instructive to examine the nature of the topmost surface layers of a crystalline solid film. The reason the surface will generally have different properties than the interior of the film can be understood by a schematic cross-sectional view as shown in Fig. 7-3. If the surface structure is the predictable extensión of the underlying lattice, we have the case shown in Fig. 7-3a. The loss of periodicity in one direction will tend to alter surface electronic properties and leave dangling bonds to promote chemical reactivity. It is more likely though that the structure shown in Fig. 7-3b will prevail. The absence of bonding forces to underlying atoms results in new equilibrium positions that deviate from those in the bulk lattice. A disturbed surface layer known as the "selvedge" may then be imagined. Within this layer the atoms relax in such a way as to preserve the symmetry of the bulk lattice parallel to the surface but not normal to it. One result of this, for example, could be a surface electric dipole moment in the selvedge. A more extreme structural disturbance is depicted in Fig. 7-3c. Here surface atoms rearrange into a structure with a symmetry that is quite different from the bulk solid. This phenomenon is known as reconstrucüon and can signifi-cantly alter many surface structure-sensitive properties, e.g., chemical, atomic vibrations, electrical, optical, and mechanical.

Surface reconstruction is quite common on semiconductor surfaces; perhaps the most famous example occurs on the (111) surface of Si. A cut through the covalent bonds of the bulk (111) plañe to créate two exposed surfaces leaves covalent bonds dangling normal to the surface into the vacuum. Dangling bonds are energetically unfavorable, and the surface reduces its overall energy by reconstructing in a manner that reduces the number and/or energy of the bonds.

A direct atomic image of the reconstructed surface of (111) Si, obtained by scanning tunneling microscopy (STM), is shown in Fig. 7-4. The Nobel Prize in physics was awarded to the developers of STM, G. Binnig and H. Rohrer, in 1986. The technique involves a highly controlled raster-fashion translation of a metal tip possessing an extremely small radius of curvature (< 1000 A), over the atomic terrain of a surface. Because the tip is only tens of angstroms from the surface, a tunneling current inversely proportional to this distance, and varying directly with the topography of the surface atoms, flows. This signal is recorded and ultimately converted into an image. STM may be thought or as the quantum mechanical analog of the stylus method for measuring film thickness.

Surface Crystallography

Just as there are 14 Bravais lattices in three dimensions, so there are just five unit meshes or nets, corresponding to a two-dimensional surface as shown in Fig. 7-5. Points representing atoms may be arranged to outline (1) squares, (2) rectangles, (3) centered rectangles, (4) hexagons and (5) arbitrary parallelo-grams. Miller-type índices are used to denote atom coordinates, directions, and distances between lines within the surface.

Consider the mesh of substrate atoms in Fig. 7-6a with an array of adatoms situated on the surface as indicated. This combination could, for example, correspond to the early growth of an epitaxial layer on the (100) plañe of a BCC crystalline substrate surface. The adsórbate atoms (shaded in) form a rectangular monolayer lattice or overgrowth above the substrate atom positions (dots). Unit dimensión vectors describing the monolayer lattice (b¡) are simply related to those of the substrate lattice (a¡). In the x direction ¿j, = 2 a,, and in the y direction b₂ = \a₂- We, therefore, speak of a P(2 X 1) overlayer where P indicates that the unit cell is primitive. Similarly, for an overlayer of the same geometry but oriented at 90° with respect to the first case, the notation is P(\ X 2). Other examples are shown in Fig. 7-6b, c. Note that in Fig. 7-6c the overgrowth layer is rotated with respect to the substrate coordi-nates and is identified by the letter R. Similarly, C is used to denote the centered lattice.

It is the geometry of the reciprocal lattice, however, and not that of the real lattice, that appears as the visible image in diffraction patterns. As the ñame implies, the reciprocal lattice has dimensions and features that are "inverse" to those of the real lattice. Long dimensions in real space appear at right angles and are shortened in inverse proportion, within the reciprocal space.

 Epitaxial Interface Crystallography

To address issues dealing with the structure of epitaxial interfaces it is first necessary to identify the crystallographic orientation relationships betweeh the film and the substrate. Unlike the notation used to describe the two-dimen-sional surfaces of the previous section, the traditional (3-D) Miller índices are employed here. For this purpose, the Índices of the overgrowth plañe are written as (HKL), and those of the parallel substrate plañe at the common interface are taken as (hkl). The corresponding parallel directions in the overgrowth and substrate planes, denoted by [UVW] and [uvw], respec-tively, must also be specified.

Ronellys Flores---CRF---libro the materials science of thin films

---