1
Introduction to Metalorganic Vapor Phase Epitaxy

S.J.C. Irvine1 and P. Capper2

1Centre for Solar Energy Research, College of Engineering, Swansea University, OpTIC Centre, St. Asaph, UK

2Ex-Leonardo MW Ltd, Southampton, UK

1.1 Historical Background of MOVPE

The technique of metalorganic chemical vapor deposition (MOCVD) was first introduced in the late 1960s for the deposition of compound semiconductors from the vapor phase, a variant of chemical vapor deposition (CVD) but with the advantage over the then-existing methods for compound semiconductor epitaxy of only requiring a single temperature for reaction and film deposition. The pioneers of the techniques, Manasevit and Simpson [1], were seeking a method for depositing optoelectronic semiconductors such as GaAs onto different, nonlattice-matched substrates, including spinel and sapphire. The near-equilibrium techniques such as liquid phase epitaxy (LPE) and chloride vapor phase epitaxy (VPE) were not suitable for nucleation onto a chemically very different surface than the compound being deposited. The first paper [2] reported on the single-crystal growth of GaAs on various oxide substrates. The process was based on a combination of a volatile alkyl organometallic for the Group III element and a hydride gas for the Group V element. This basic approach has remained for all the III-V compounds, with a few exceptions where arsine or phosphine are replaced by liquid sources, tertiarybutylyarsine (TBAs) and tertiarybutylphosphine (TBP) [3]. For the Group II-VI compounds, the hydrides were less useful where low-temperature growth was considered to be important to control native defect concentrations. However, early work using H2Se and H2S for sources to grow ZnSe and ZnS was successful at higher temperatures [4]. For the tellurides, the hydride route was not practical as H2Te is not stable under ambient conditions, so alkyl tellurium was used from the outset.

It was more than a decade after the first reports of MOCVD that the international conference series started with the conference in Ajaccio in 1981, when the "father" of MOCVD, Hal Manasevit gave an invited talk that covered a wide range of compound semiconductor materials that had been successfully grown by MOCVD, including III-V, II-VI, IV-VI, and II-IV-V2 [5]. Perhaps the most significant developments over that first decade, represented at that conference, were the progress with improved purity of the organometallic precursors and low-defect-density, lattice-matched epitaxy, which led to the first demonstration of electron mobility >100 000?cm2/V?s [6]. This made MOCVD competitive with other epitaxial techniques for GaAs, such as the emerging molecular beam epitaxy (MBE), and paved the way for high-performance optoelectronic devices, including quantum-well lasers and high-mobility transistors. The issues of purity of the early organometallics is addressed again in Section 14.1, which takes a historical perspective before looking forward to future prospects for MOCVD. The improved purity of the organometallic sources led to the demonstration of low-threshold lasers that opened the way for commercialization of MOCVD. Purification of precursors is covered in more detail in Chapter 13, where the challenges with traditional distillation methods were overcome using adduct purification.

The high-quality epitaxial nature of the films was emphasized by more commonly adopting the name of the growth method to be metalorganic vapor phase epitaxy (MOVPE) or organometallic VPE (OMVPE). There was some debate at the first international conference on what it should be called, which may seem strange now; but in those formative years, even the name was a point of discussion. The title of the conference was "IC-MOVPE I," which settled the issue. For the benefit of newcomers to MOVPE, all of these variants of the name can be found in the literature and, in most cases, can be used interchangeably. MOCVD can be considered broader, as it includes polycrystalline growth that is appropriate to the photovoltaic thin films covered in Chapter 10. The early niche applications of MOVPE were with GaAs photocathodes [7], GaAs heterojunction bipolar transistor (HBT) lasers [8], and GaInAsP lasers and detectors for 1.3-µm optical fibre-optic communications [9].

The ease with which GaAs could be grown using trimethylgallium (TMG) and arsine was not readily replicated by all the semiconductor materials of interest. Indium phosphide had proved to be particularly challenging due to a polymerization prereaction that occurred between trimethylindium (TMI) and phosphine [10]. This was overcome, initially, by going to low-pressure MOVPE and replacing TMI with triethylindium (TEI). Later improvements in the purity of TMI and in overcoming issues with it being a solid source at room temperature have led to TMI now being the precursor of choice. Atmospheric-pressure MOVPE of InP was achieved by Moss [11] using an elegant solution that took advantage of the adduct formation of In alkyls. A room-temperature stable adduct, TMIn.TEP (where TEP is triethylphosphine), was formed that prevented the polymerization reaction with PH3 but the adduct decomposed over the hot substrate to yield TMIn and TEP. The TEP was stable and was exhausted from the reactor without further decomposition, while the TMIn and PH3 or AsH3 for the arsenic compounds reacted to form the compound semiconductor on the substrate. Adducts have also been used in the purification of the organometallic precursors, as will be described in Chapter 13, and for preventing prereactions with ZnO, as described in Chapter 11. The significance of growing InP was to be able to then grow ternaries and quaternaries for infrared lasers and detectors, lattice matched to InP. This enabled access to the growing market for 1.5-µm wavelength devices for long-range optic fibre-optic communications [12]. The purification of Group III precursors is discussed in Chapter 3 with regard to the device application of these materials. The decomposition chemistry and role of the Group V hydride is important to minimize the incorporation of carbon. In some devices, this is now used as an intentional P-type dopant, substituting on the Group V site.

The antimonides (InSb, GaSb, AlSb) are an important class of narrow-bandgap semiconductors for infrared detectors, long-wavelength lasers, and thermoelectric devices. Unlike GaAs- and InP-based semiconductors, where hydrides are normally used as the Group V source, for the antimonides it is necessary to use alkyl sources such as TMSb. The growth of the antimonides is described in Section 3.5.

The II-VI alloy mercury cadmium telluride (MCT) had also proved to be difficult to produce by MOVPE, and the growth in the interest in MCT for long-wavelength thermal imagers (operating around 10?µm) was stimulating research in different epitaxial techniques, including LPE, MBE, and MOVPE, all of which are used in production today. The challenge was from the very-high-equilibrium vapor pressure of mercury in MCT at growth temperatures ranging from 200?°C to 500?°C. Early success in the 1980s with an MOVPE approach using a liquid Hg source ensured a very fertile two decades of research that will be described in more detail in Chapter 9 [13]. This helped to demonstrate at an early stage that MOVPE can be a very versatile technique that has been proven again many times for different compound semiconductors over the intervening years.

Thus, the technique of MOVPE was born, but it was not until the late 1980s that MOVPE became a production technique of any significance. This success depended on painstaking work on improving the impurity of the organometallic precursors and development of MOVPE equipment to improve uniformity and deposit onto multiple substrates in the reaction chamber. The emergence of commercial equipment suppliers took over from the "home-built" reactors and provided much-needed standards to which production of epi-wafers could be benchmarked. By this time, the focus was on high-quality epitaxial layers on lattice-matched substrates, in contrast with the early work. One exception was the growth of CdTe epitaxial layers onto C-plane sapphire as a substrate for LPE growth of MCT for mid-wave infrared (MWIR) detectors. Ironically this was manufactured by Rockwell International, where Hal Manasevit carried out his original research. The topic of homoepitaxy versus heteroepitaxy has been a continual balance between achieving very high-quality epitaxial layers and materials functionality. To this day, the growth of high-quality epitaxy on silicon substrates has remained the ultimate challenge, through not only providing a ready supply of high-quality and cheap substrates but also the integration of optoelectronic and electronic devices. By the late 1980s and early 1990s, there was a commercial supply of GaAs on Si substrates to be used as a substitute for GaAs substrates. Improvements in the quality and size of GaAs substrates made this approach uncompetitive. Today, this is attracting a new generation of research activity; the topic of monolithic III-V integration on Si(001) substrates is covered in Chapter 8. New understanding of...