1 Overview of the OMVPE Process 1.1 Introduction The last twenty-five years have brought a true electronic revolution of enormous significance to our everyday lives. As an economic phenomenon, this revolution has resulted in an increase in the cost effectiveness of electronic functions at a rate unparalleled by any other technologies. At the beginning of this period, the cost of a single transistor device was of the order of $5. Today we have sixty- four-megabit integrated circuits that cost a fraction of this amount in inflation- adjusted dollars, with gigabit circuits in the laboratory. During the same time pe- riod, the performance—for example, the speed—has also increased by orders of magnitude. This decrease of more than eight orders of magnitude (10^) in the price per unit function is astonishing and has significandy affected the national and world economies. By the year 2000, the microelectronics industry is projected to total $10'^, fully 10% of world trade [1]. This behavior contrasts sharply with other elements in our economic life. In a semiserious but provocative essay, Stephen Jay Gould [2] suggested in 1984 that some manufactured goods follow the rate of phyletic size decrease. An extrapolation of past trends suggested that Hershey's would introduce in December 1998, at a price of $0,475, the amazing weighdess chocolate bar, only a slight exaggeration. General experience suggests that while the quality and performance of many products have increased dramatically in the
16
Embed
1 Overview of the OMVPE Process - University of California, San …iebl.ucsd.edu/sites/iebl.ucsd.edu/files/users/user5/1.pdf · 2017-10-27 · 1 Overview of the OMVPE Process 1.1
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1 Overview of the OMVPE Process
1.1 Introduction The last twenty-five years have brought a true electronic revolution of enormous significance to our everyday lives. As an economic phenomenon, this revolution has resulted in an increase in the cost effectiveness of electronic functions at a rate unparalleled by any other technologies. At the beginning of this period, the cost of a single transistor device was of the order of $5. Today we have sixty-four-megabit integrated circuits that cost a fraction of this amount in inflation-adjusted dollars, with gigabit circuits in the laboratory. During the same time period, the performance—for example, the speed—has also increased by orders of magnitude.
This decrease of more than eight orders of magnitude (10^) in the price per unit function is astonishing and has significandy affected the national and world economies. By the year 2000, the microelectronics industry is projected to total $10'^, fully 10% of world trade [1]. This behavior contrasts sharply with other elements in our economic life. In a semiserious but provocative essay, Stephen Jay Gould [2] suggested in 1984 that some manufactured goods follow the rate of phyletic size decrease. An extrapolation of past trends suggested that Hershey's would introduce in December 1998, at a price of $0,475, the amazing weighdess chocolate bar, only a slight exaggeration. General experience suggests that while the quality and performance of many products have increased dramatically in the
2 1 Overview of the OMVPE Process
last three decades, consider for example automobiles and California wines, the price has also invariably increased as well.
A question on the minds of those concerned with the semiconductor revolution has been at what level, and when, the progress will taper off and finally cease. Using Si in conventional structures, the ultimate limits for the processing and storage of data appear to be on the horizon [I, 3]. Beyond those limits, progress seems likely only using unconventional approaches including the use of photons and the photonlike characteristics of electrons. Photons themselves can be used for the logic functions in circuits, although the first application would appear to be their use in interconnects, where performance in conventional structures is limited by the need to move electrons from device to device in wires. An example of use of the photonlike properties of electrons is the switching between "on" and "off" states by tunneling between devices in a superlattice structure. This would allow more rapid switching of elements that could be packed more tightly, perhaps even in three-dimensional arrays. Applications such as artificial intelligence will require such advances.
In this information age we also demand the transmission of information at extremely rapid rates, in addition to the reading/writing of data to/from storage media and the display and printing of this information. In advanced applications, these all require the use of a variety of photonic devices that are fabricated mainly in compound semiconductor materials. This demand for a wide variety of materials and structures for high-performance devices and circuits requires epitaxial processes for their growth with the desired qualities.
The fabrication of superlattice structures with dimensions of the order often to several hundred angstrom units (A) fabricated in Si and Si/Ge alloys as well as in compound semiconductor materials will certainly require epitaxial growth processes with exquisite control, including the ability to change composition within a period of a few angstrom in at least one dimension, and ultimately in two or all three dimensions. The novel concept of self-assembly of quantum dot structures is currently being pursued with great vigor.
Other requirements imposed even on current epitaxial growth processes include the ability to grow high-purity layers (foreign impurity levels of a few parts per billion) as well as to intentionally introduce impurities for n-type, p-type, and semi-insulating behavior. An important requirement will certainly be the versatility to grow the widest possible range of materials, including alloys consisting of combinations of four or more elements. All of this will have to be performed economically, which probably means with large areas processed per run in an efficient (i.e., with little waste of time or the expensive starting materials) automated batch process, with a high degree of control guaranteeing uniformity and reproducibility. This must all be done in a safe, nonpoUuting environment. Together, these requirements represent a tall order, indeed.
Other reasons for epitaxial growth of semiconductor layers are related to the
1.2 Comparison of Epitaxial Techniques 3
lower growth temperatures. The III/V and 11/VI binary compounds can be grown from the melt. However, the use of the high temperatures required often gives rise to unacceptably high concentrations of both native defects and foreign impurities. The high vapor pressures of the constituents often cause additional problems. The thirty-five-atmosphere phosphorus pressure over molten GaP requires a special crystal growth ''bomb." Alloy semiconductors do not melt congruently; that is, the solid is not in equilibrium with a liquid of the same composition. Thus, the growth of bulk crystals of ternary and quaternary semiconductor alloys is practically impossible; epitaxial growth is required.
1.2 Comparison of Epitaxial Techniques Several epitaxial techniques are currently available for the growth of semiconductor materials, including the oldest techniques, liquid-phase epitaxy (LPE) and chloride vapor-phase epitaxy (CIVPE), as well as hydride vapor-phase epitaxy (HVPE), molecular-beam epitaxy (MBE), chemical-beam epitaxy (CBE), and or-ganometallic vapor-phase epitaxy (OMVPE). Each technique has strengths and weaknesses, summarized in Table 1.1 [4]. The technique(s) best able to meet the requirements briefly summarized here will play an important role in the continued advance of semiconductor electronics.
1.2.1 Liquid-Phase Epitaxy (LPE) Liquid-phase epitaxy was used for much of the early research on III/V and 11/VI semiconductors. The apparatus required for the growth of excellent quality layers is extremely simple, and the low impurity and point defect levels achieved are impressive [4]. This is partly due to the stoichiometry of LPE material. For example, the growth of GaAs from a Ga-rich melt always produces material with the most Ga-rich stoichiometry. Thus, defects such as Ga vacancies and As atoms on Ga sites (the As antisite) are virtually nonexistent in LPE material. The As antisite defect is believed to be related to the deep electron trap denoted EL2, which is known to have a deleterious effect on several materials properties.
Freedom from background elemental impurities in LPE-grown material is partly due to the availability of high-purity metals, which are typically used as solvents, and the inherent purification process that occurs during the liquid-to-solid phase transition for solutes with distribution coefficients of less than unity. Very important for the LPE growth of Al-containing materials, such as AlGaAs, is the purification process where oxygen in the system forms highly stable AI2O3 on the surface of the liquid, thus preventing oxygen incorporation into the epitaxial layer. This allowed the early AlGaAs layers grown by LPE to be far superior to layers grown by any other technique. The problem with LPE is that the very
1 Overview of the OMVPE Process
Table 1.1 Overview of epitaxy techniques
Technique
LPE
CIVPE
HVPE
MBE
OMVPE
CBE
Strengths
Simple High purity
Simple High purity
Well developed Large scale
Simple process Uniform Abrupt interfaces In situ monitoring
Most flexible Abrupt interfaces High purity Simple reactor Robust process Uniform Large scale High growth rates Selective growth In situ monitoring
Uniform Abrupt interfaces Direct control of fluxes In situ monitoring Selective growth
Weaknesses
Scale economics Inflexible
No Al alloys Sb alloys difficult >20-A interface widths
No Al alloys Sb alloys difficult Complex process/reactor Control difficult Hazardous precursors
Expensive reactants Most parameters to control accurately Hazardous precursors
Low throughput No large-scale reactors Expensive (capital) Expensive reactants Hazardous precursors N materials difficult
simplicity that makes the process so attractive for laboratory appHcations ultimately limits its flexibility in the commercial production of elaborate modern device structures. For instance, the growth of multilayer structures with extremely abrupt interfaces is difficult by LPE. In addition, the thickness uniformity of epitaxial layers grown by LPE is generally poor, with both short- and long-range variations. This results in difficulties with both yield, for sensitive devices fabricated in LPE-grown material, and scaling the process to the large size required for production operations.
Another limitation is the difficulty in growing certain materials. Particularly
1.2 Comparison of Epitaxial Techniques 5
important examples are the alloys containing both Al and In. The high Al distribution coefficients lead to nearly insurmountable difficulties for LPE growth. These alloy systems will be discussed in Chapter 8.
In spite of these difficulties, LPE is still in use for the production of several simple devices such as GaAs light-emitting diodes (LEDs) due to the superior properties of the material produced and the cost-effective production of very thick (10-100 yLtm) epitaxial layers. For some materials and applications, LPE remains the only technique that gives adequate results. However, the domain in which LPE is used commercially is being steadily eroded due to the need for advanced materials and more elaborate structures.
1.2.2 Vapor-Phase Epitaxy (VPE) To simplify notation, we will classify all vapor-phase epitaxial processes using halide transport of either element as VPE. This includes what is traditionally called hydride vapor-phase epitaxy, where the group V (for III IV materials) or group VI (for 11/VI materials) element is transported to the growth interface using the hydrides. It also includes traditional halide VPE, where both the cation and anion elements are transported using the halide (typically the chloride). For the growth of GaAs, this typically occurs by using ASCI3 flowing over liquid Ga.
The VPE techniques have also played a major role in the development of compound (particularly IIIIV) semiconductors. The first ultra-high-purity GaAs was produced by the ASCI3 process, and hydride VPE is still used today in large-scale commercial operations for the production of the relatively simple GaAsP homojunction LEDs. These VPE techniques suffer from some of the same limitations as LPE. While the growth of superlattice structures is not impossible, it is rather difficult for both techniques. Alternating layers are normally obtained by physically moving the substrate back and forth between two, or more, reactor tubes [5]. Such a crude approach is not attractive as compared with later techniques such as MBE, CBE, and OMVPE, where the transport of source materials, rather than the substrate, is manipulated. Another distinct problem with chloride transport is the extreme difficulty of growing Al-containing materials due to the chemistry of Al chlorides [4]. Again, the commercial use of these techniques is steadily shrinking.
1.2.3 Molecular-Beam Epitaxy (MBE) Molecular-beam epitaxy is the technique mainly responsible for the revolution in device physics that has occurred due to the use of superlattice structures. For several years MBE was the only technique capable of producing perfectly abrupt interfaces (i.e., with no graded transition region). In contrast with the techniques already discussed, MBE is elegantly simple in concept. Elemental sources are
6 1 Overview of the OMVPE Process
evaporated at a controlled rate onto a heated substrate under ultrahigh-vacuum (UHV) conditions. At low growth rates, the resultant layer is indeed epitaxial. In the UHV environment, the growth process can be monitored as the crystal is built up one atomic layer at a time [6], as discussed in Chapter 3.
While MBE may be the ultimate research tool for the production of complex and varied structures, it has limitations for commercial applications. The need for UHV apparatus is expensive in terms of both capital outlay and operating expense. Frequent shutdowns are required to replenish the source materials, and opening the UHV apparatus requires bake-out before returning to the growth of very high-purity materials. This not only wastes valuable production time but also introduces a degree of nonreproducibility into the process; for example, the material quality may be affected by opening the reactor to change the sources. This is, of course, a factor in the overall economics of the use of MBE for production. A second problem limiting throughput for devices requiring thick layers is the relatively low growth rate.
Morphological defects thought to be due to the elemental cation sources have proven nearly impossible to eliminate. This is a problem resulting in a reduced yield of devices fabricated in material grown by MBE. Another major problem is the difficulty in growing phosphorus-containing materials. Phosphorus bounces around in the system, ultimately collecting in the vacuum pumps. In addition, the growth of alloys containing both As and P is particularly difficult. Growth of the semiconductors containing the still more volatile N (AlGaInN) is even more difficult by MBE. This seriously limits the versatility of the MBE process.
1.2.4 Organometallic Vapor-Phase Epitaxy (OMVPE)
Organometallic vapor-phase epitaxy is often referred to as metal-organic chemical vapor deposition (MOCVD) and by other permutations of these same letters (MOVPE and OMCVD). We will use the term organometallic to describe the precursor molecules, since it agrees with general chemical nomenclature. CVD is the most general term describing the growth process, since it implies nothing about whether the resultant layer is single crystalline, polycrystalline, or amorphous. We will be concerned exclusively with single-crystalline epitaxial layers; thus, the technique will be referred to as OMVPE.
The beginnings of OMVPE research are often attributed solely to the pioneering work of Manasevit and coworkers [7] in the late 1960s. However, recent litigation has brought to light patents describing earlier forms of OMVPE for the growth of III/V semiconductors [8-10]. Because this work was not published in the scientific literature, it has been inadvertently ignored by the technical community. However, it clearly predates the work of Manasevit. The Miederer et al. U.S. patent was filed in September 1963 [8], with a similar German patent [9] filed a year earlier. Perhaps even more surprising, because of the early filing date of
1.2 Comparison of Epitaxial Techniques 7
September 1954, is the Scott et al. patent for the growth of InSb using triethyl-indium and stibine [10]. Nevertheless, without doubt the early publications of Manasevit are of enormous significance and represent a critical factor leading to the rapid development of OMVPE in the early 1970s. The early work described mainly the production of single-crystalline layers of a wide range of III/V, 11/VI, and IV/VI semiconductors. Early doubts about the ultimate purity of these semiconductors were dispelled by the 1975 demonstration, in a key paper, of extremely high-purity GaAs, with low-temperature mobilities exceeding 100,000 cm W s [11]. The appearance of reports of high performance minority carrier devices in the late 1970s and early 1980s [12] was the final factor leading to the explosion of OMVPE activity in the 1980s and 1990s. Today, much of the OMVPE effort centers around development activities associated with the increasing use of OMVPE for commercial production operations.
We will see in later chapters that the OMVPE process is complex; thus, development initially proceeded somewhat more slowly than for the much simpler MBE technique. During the early 1980s a burning question was whether MBE or OMVPE would ultimately be the dominant technique for production of compound semiconductor materials for commercial device processing. At that time, questions related to purity and the inherent limits on interface abruptness still troubled the OMVPE community. Today those obstacles have been overcome. As we will see in Chapter 8, OMVPE has produced the highest-purity InP grown by any technique and GaAs as pure as by any other technique. More recently, it has been shown to be capable of producing high-quality nitrides for the highest-performance short-wavelength photonic devices. OMVPE is clearly the leader in this area. The ability to produce nearly atomically abrupt interfaces has also been demonstrated, as will be discussed in Chapter 9. Device results are summarized in Chapter 10. It will be seen that, in general, devices produced by MBE, CBE, and OMVPE have very similar performance characteristics. The major attractions of OMVPE relative to other techniques are the versatility and the demonstrated suitability for large-scale production. OMVPE is unquestionably the most versatile technique, suitable for the production of virtually all III/V and 11/VI semiconductor compounds and alloys. It has also proven to be the most economical technique, particularly for the production of devices requiring large areas, such as LEDs, photocathodes, and solar cells. In general, favorable economics require large-scale, high growth rates, and a high yield of suitable material (i.e., uniformity and reproducibility).
For these reasons, OMVPE research, development, and production efforts have grown geometrically during the last two decades. This research effort has paid off: a number of commercial OMVPE facilities around the world are used for the production of devices such as LEDs in the range from red to blue, injection lasers in several wavelength ranges from the near IR (including 1.55 /mm) to the blue, detectors, the highest-efficiency solar cells, and ultrahigh-speed transistors and integrated circuits. Commercial OMVPE equipment has been developed that now
8 1 Overview of the OMVPE Process
provides virtually turnkey operation of reactors for the growth of materials for these commercially important devices.
Nevertheless, several problems remain, including the need for expensive reac-tants and the large number of parameters that must be precisely controlled to obtain the necessary uniformity and reproducibility. An additional problem, addressed in detail in Chapter 4, is the use of hazardous materials such as the group V hydrides.
1.2.5 Chemical-Beam Epitaxy (CBE) Organometallic or metal-organic MBE (OMMBE or MOMBE) and CBE are hybrid techniques combining features of OMVPE and MBE. To be consistent, the term OMMBE will be used, even though MOMBE is much more common. In these techniques, organometallic or elemental group III elements and organometallic, hydride, or elemental group V elements are injected into a UHV system. OMMBE is often considered as distinct from CBE due to the use of elemental group V sources. Today the distinction has faded somewhat. For the purposes of this book, these techniques will be considered as special forms of OMVPE. In fact, the first OMMBE paper [13] described the use of typical OMVPE reactants in a UHV system as a means of clarifying the fundamental aspects of the OMVPE growth process. Another variation of the MBE technique, commonly referred to as gas source MBE (GSMBE), uses elemental group III sources combined with hydride or organometallic group V sources. Panish [14] was the first to use gaseous reactants in an MBE system. The advantages and disadvantages of OMMBE, GSMBE, and CBE will be considered in the context of the effect of reactor pressure on the OMVPE process. In general, the advantages and disadvantages of CBE, listed in Table 1.1, are similar to those for OMVPE. However, carbon contamination has been much more of a problem in CBE. Today, the quality of materials produced by CBE is comparable to that produced by MBE and OMVPE. As for MBE, the high vapor pressure of N over the III/V nitrides presents a problem for CBE growth. This, the absence of a large-scale reactor for production applications and the lack of clear and powerful motivating factors favoring CBE have probably relegated CBE mainly to the status of a research tool rather than a production technique.
1.3 Overview of the OMVPE Growth Process
All crystal growth processes, including OMVPE, are highly complex. Indeed, early crystal growth studies were invariably largely empirical, giving crystal growth, in general, the appearance of an art. Until fairly recently, a typical ap-
1.3 Overview of the OMVPE Growth Process 9
preach had been to treat the total OMVPE process as a "black box" that can be controlled by turning knobs in a systematic fashion until the desired materials properties are obtained, and then moving on to device fabrication and characterization. The standard measure of success has been device performance. The need for this black-box approach is partly because of the complex, multicomponent, multiphase systems that are normally of interest and partly because the process is dynamic and inhomogeneous phases are inherent. However, this approach is not necessary today. It would be inaccurate to claim that all aspects of the OMVPE process are thoroughly understood. Nevertheless, enough information is available to formulate simple, accurate models that are often quite useful in understanding and predicting the performance of OMVPE growth systems. The last few years have witnessed the introduction of optical techniques for the in situ monitoring and control of the OMVPE growth process. This has led not only to an improved understanding of the process but also to better control of the final product in commercial systems.
In developing an overall growth model, the fundamental processes occurring during crystal growth are commonly subdivided into thermodynamic and kinetic components. Table 1.2 gives a brief summary of the various processes affecting OMVPE growth. Thermodynamics determines the driving force for the overall growth process, and kinetics defines the rates at which the various steps occur. Hydrodynamics and mass transport, which are intimately linked, control the rate of transport of material to the growing solid/vapor interface. The rates of the chemical reactions occurring during growth, either homogeneously in the gas phase or heterogeneously at the growing interface, also play a role. The rates of the heterogeneous processes depend on the details of the surface structure, such as the reconstruction and step structure. Each of these factors will dominate some aspect of the overall growth process.
A study of the dependence of a macroscopic quantity, such as growth rate, on external parameters, such as substrate temperature and input precursor (source) flow rates, gave the first insights into the overall growth mechanism. In a classic paper, Shaw [15] showed that an examination of VPE growth rate versus temperature allows a general categorization of the process limiting the growth rate as mass transport, surface kinetics or thermodynamics. For an exothermic process such as OMVPE, increasing temperature results in a decrease in the thermodynamically limited growth rate. If the reaction rates limit the growth rate, termed the kineti-cally limited case, the growth rate increases with increasing temperature. Since gas phase diffusion is a nearly temperature-independent process, the growth rate is nearly independent of substrate temperature in the mass-transport-limited case.
Other factors such as the effects of substrate orientation and total flow rate, with the individual partial pressures kept constant, also help in the determination of the growth-rate-limiting process. Substrate orientation affects neither the thermodynamic driving force nor mass transport; thus, a dependence of growth rate on
10 1 Overview of the OMVPE Process
Table 1.2 Key processes in OMVPE growth
THERMODYNAMICS Driving force for epitaxy
Maximum growth rate Stoichiometry and dopant incorporation Solid composition of alloys Surface reconstruction
MASS AND HEAT TRANSPORT Hydrodynamics—boundary layers, recirculating flows, "dead" regions Temperature profile near the heated substrate Mass transport processes and rates—diffusion and convection
Nutrients to interface—products away from interface Wall effects
PHYSICAL SURFACE PROCESSES Surface reconstruction Step size and spacing Kink formation Surface diffusion—including anisotropy due to reconstruction Two-dimensional nucleation Three-dimensional nucleation—surface roughening
CHEMICAL REACTIONS Homogeneous—in the gas phase
Adduct formation Pyrolysis of precursors and adducts
Complex radical reactions
Heterogeneous—at the substrate surface Surface reconstruction Density and nature of steps, kinks, other "defects" Adsorption/desorption of precursors and intermediates Pyrolysis, including complex radical reactions Desorption of products
SUBSURFACE PROCESSES Diffusion in the bulk—extremely slow Diffusion in few layers near the surface—more rapid?
substrate orientation is a clear signal that the kinetics of surface reactions is the rate-Hmiting step. On the other hand, since neither thermodynamics nor reaction rates are dependent on total gas velocity, if the partial pressures are held constant, a dependence of growth rate on total flow rate is a clear indication that mass transport processes limit the growth rate.
This type of analysis can be applied to OMVPE growth. Consider, for example, the OMVPE growth of GaAs using trimethylgallium (TMGa) and ASH3. A consistent, general pattern for growth rate versus temperature has been reported in
1.3 Overview of the OMVPE Growth Process 11
10*
i • A •
A Krautleetal (1983) • Plassetal (1988) • Reep and Ghandhi (1984)
0.6 0.8 1.0 1.2 1.4 1000/T(1/K)
Figure 1.1. Growth efficiency (growth rate/TMGa molar flow rate) versus reciprocal temperature. The data, all for GaAs grown using TMGa and AsH^, are from Plass et al. [20] at a reactor pressure of 76 Torr; Krautle et al. [21] at atmospheric pressure; and Reep and Ghandhi [22], also at atmospheric pressure.
many studies. A collection of typical results is shown in Figure 1.1. The results of these and other studies in the same system show several characteristic features for OMVPE growth at normal operating pressures in the temperature range typically used, from 550° to 750°C:
1. The growth rate is nearly independent of temperature, indicative of mass-transport-limited growth.
2. Saxena et al. [16] report the growth rate to be independent of substrate orientation, which also suggests mass transport limited growth.
3. Frolov et al. [17] report that rotation of the pedestal on which the substrate sits increases the growth rate. Since this would decrease the mass transport boundary layer thickness, this finding is also consistent with the hypothesis that the growth rate is limited by mass transport.
4. Leys and Veenvliet [18] showed that increasing the flow velocity increases the growth rate, also due to a decrease in the mass transport boundary layer thickness.
These observations lead to an unambiguous determination that the growth-rate-limiting step is mass transport in the temperature range from approximately 550° to 750°C. At lower temperatures, the growth rate decreases with decreasing temperature, characteristic of a process limited by reaction kinetics. In this regime, the growth rate is also dependent on the orientation of the substrate. At temperatures above 750° C, the decrease in growth rate may be due to thermodynamic
12 1 Overview of the OMVPE Process
factors, such as the evaporation of the group III element, although alternative processes such as depletion of reactants on the reactor walls upstream from the substrate must also be considered.
An additional observation reported universally for the OMVPE growth of III/V alloys is the linear dependence of growth rate on the group III flow rate entering the reactor in the temperature range where mass transport is the growth-rate-determining step in the overall process. Representative data for GaAs growth using TMGa and ASH3, AlAs from TMAl and ASH3, and InP from TMIn and PH3 are plotted in Figure 1.2. The constant relating the growth rate to the group III source mole fraction in the input gas stream might be termed the mass-transport coefficient. A similar, but more common measure of reaction efficiency is the ratio of growth rate to the input molar flow rate of the minority component [19], which is plotted in Figure 1.1. This quantity has the advantage of being nearly independent of the reactor pressure. As discussed more fully in Chapter 6, the mass-transport coefficient increases with decreasing reactor pressure, leading to generally incorrect and misleading claims that low-pressure reactors are more efficient. Data for several combinations of reactants for a number of III/V semiconductors are listed in Table 1.3. We will return to a discussion of these numbers later, but generally, in a system with no parasitic gas-phase reactions leading to depletion of the nutrient upstream from the substrate, values of reaction efficiency in the vicinity of 10"̂ />t/mol are observed. Since the growth rate is limited by mass transport, the linear dependence on group III flow rate suggests that the group III molecule is completely depleted at the solid/vapor interface. In the simple case
o O
0.0
a GaAs :Mizutaetal (1984)
A AlAs iMizutaetal (1984)
• InP: Hsu etal (1983)
A a
^ .
A
10 20
AikyI Flow Rate (iimole/min)
30
Figure 1.2. Growth rate versus group III alkyl flow rate for GaAs, using TMGa and AsH,, from Mizuta et al. [23]; AlAs using trimethylaluminum (TMAl) and AsH^ from Mizuta et ai. [23]; and InP using trimethylindium (TMIn) and PH3 from Hsu et al. [24].
1.3 Overview of the OMVPE Growth Process 13
Table 1.3 Summary of OMVPE growth rates
System
GaAs
AlGaAs
GaSb
GaAsSb
InAs
InP
GalnAs
GalnP
InSb
GaN
rg//jii(pLm mol-
1.5 X 10^
1.6 X 10^ 2.9 X 10^
2.8 X 10-̂ 6.7 X 102
3.4 X 102
1.7 X 10^
4.5 X 10^
4.5 X 10^ >104
2.8 X 10^
0.7-1.6 X 10-̂
3 X 10^ 1.0 X 10^
2.0 X 10^
0.9-1.5 X 10-̂ 2.0 X 10-̂
4 - 6 X 10-̂
3.8 X 10-̂ 4.5 X 102
7.9 X 102
1.2 X 10^
6.9 X 10^
7.6 X 102
1.4 X 10^
1.3 X 10^
2.0 X 10^
8.8 X 102
5.2 X 10^
1.5 X 104
1.0 X 104
1.4 X 10^
Low Pressure
) or 1 atm
LP
LP
LP
LP
Source*
TMGa + AsH,
TEGa + AsH,
TMGa + TMAl + AsH,
TMGa + TMSb
TMGa + TBDMSb
TMGa + TMSb + AsH,
TMGa + TMSb +
TMAsCorAsH^)
EDMIn + AsH,
TMIn + AsH,
TEIn + AsH,
TEIn + TMAs
TIPIn + TBAs
TEIn + PH3
T M I n - T E P + PH3
TEIn 4- PH3 TMIn + PH3
TEIn + TEGa + ASH3
TMGa + TMIn + TMAs TMGa + TEIn + ASH3
TMGa + TMIn + ASH3
TEIn + TEGa + PH3
TMIn + TMGa + PH3
TMIn + TBDMSb
TMIn + TDMASb
TMGa + NH3
Trc) 600-775
600
6 0 0 - 6 2 5
5 7 5 - 6 0 0
650
700
700-750
680 -720
620
5 7 0 - 6 5 0
600
6 0 0 - 6 5 0
3 8 0 - 6 0 0
4 0 0 - 6 0 0
650-700
5 7 5 - 6 5 0
5 5 0 - 6 0 0
3 0 0 - 4 0 0
600
550
650
650
600
550
625 600
5 2 0 - 6 0 0
600
625
450
325 -425
1,000
Notes
a
b
c
c
d
e
f
g
h
i
J k
1
1
m
n
b
0
P
q r
s
t
q u V
b
w
X
y
y
z
* Notation defined in Chapter 4. ^H. M. Manasevit and W. I. Simpson, J. Ei ^C. P. Kuo, R. M. Cohen, and G. B. String; '̂ P. D. Dapkus, H. M. Manasevit, and K. L. 'I Y. Seki, K. Tanno, K. lida, and E. Ichiki,. ^G. B. Stringfellow and H. T Hall, / Crysi
^E. E. Wagner, G. Horn, and G. B. Stringfellow, 7. Electron. Mater. 10 239 (1981). e Y. Mori and N. Watanabe, J. AppL Phys. 52 2792 (1981). ''M. J. Ludowise and C. B. Cooper, Proc. Soc. Photoopt. Instrum. Eng. 323 117 (1982).
(continues)
14 1 Overview of the OMVPE Process
Table 1.3 — Continued
'C. H. Chen, C. T. Chiu, L. C. Su, K. T. Huang, J. Shin, and G. B. Stringfellow, 7. Electron. Mater. 22 87 (1993). JC. B. Cooper, R. R. Saxena, and M. J Ludowise, /. Electron. Mater 11 1001 (1982). •̂ M. J. Cherng, G. B. Stringfellow, and R. M. Cohen, Appl. Phys. Lett. 44 677 (1984); M. J. Cherng, R. M. Cohen, and G. B. Stringfellow, J. Electron. Mater 13 799 (1984).
'K. Y. Ma, Z. M. Fang, R. M. Cohen, and G. B. Stringfellow,/ Appl. Phys. 70 3940 (1991). '"H. M. Manasevit and W. I. Simpson./ Electrochem. Soc. 120 135 (1973). "B. J. Baliga and S. K. Ghandhi, / Electrochem. Soc. 121 1642 (1974). «K. T. Huang, Y. Hsu, R. M. Cohen, and G. B. Stringfellow,/ Crystal Growth 156 311 (1995). PT. Fukui and Y. Horikoshi, ypn. / Appl. Phys. 19 L551 (1980). 4M. Razeghi, M. A. Poisson, J. P. Larivain, and J. R Duchemin, / Electron. Mater 12 371 (1983); M. Razeghi,
M. A. Poisson, and J. P. Duchemin, unpublished results (1983). ^R. H. Moss and J. S. Evans. / Cryst. Growth 55 129 (1981). ^M. Ogura, K. Inone, Y. Ban, T. Uno, M. Morisaka, and N. Hase, Jpn. / Appl. Phys. 21 L548 (1982). 'C. C. Hsu, R. M. Cohen, and G. B. Stringfellow, / Cryst. Growth 63 8 (1983). "C. B. Cooper, M. H. Ludowise, V. Aebi, and R. L. Moon, Electron. Lett. 16 20 (1980). ^J. P Noad and A. J. SpringThorpe, / Electron. Mater 9 601 (1980). *J. Yoshino, T. Iwamoto, and H. Kukimoto, / Cryst. Growth 55 74 (1981); J. Yoshino, T. Iwamoto, and H. Kuki-
moto, Jpn. J. Appl. Phys. 20 L290 (1981). "C. C. Hsu, R. M. Cohen, and G. B. Stringfellow,/ Cryst. Growth 62 648 (1983). yj. Shin, A. Verma, G. B. Stringfellow, and R. W. Gedridge, / Cryst. Growth 143 15 (1994). 'S. Nakamura and G. Fasol, 77?̂ Blue Laser Diode: GaN Based Light Emitters and Lasers (Springer, Berlin, 1997), p. 37.
of diffusion through a mass-transport boundary layer, the group III flux, would be [19]
-^^ ^ , (1.1)
where D is the diffusion coefficient, p* is the input partial pressure of the group III source, p' is the group III partial pressure at the interface, and 8^ is the thickness of the boundary layer. Since the intercept at / = 0 in Figure 1.2 occurs at/7* = 0, the group III partial pressure at the interface must be nearly zero. In the typical case where/7v»Pi*ji, the growth rate is independent of group V flow rate.
For 11/VI systems, the more volatile group II precursor is often present in excess. In this case the pressures in Equation (l.l) refer to the group VI precursor, and the growth rate is independent of the input group II partial pressure.
The approach taken in this book will be to examine each aspect of OMVPE separately and then to assemble the pieces into a coherent model of the OMVPE growth process in Chapter 7. As discussed earlier, thermodynamics, reaction kinetics, and mass transport play distinct and important roles in the OMVPE process. Thermodynamics, which defines the driving force for the epitaxial growth process, is discussed in Chapter 2. This chapter includes both traditional bulk thermodynamics as well as the thermodynamics of the surface, which has a profound importance for the OMVPE growth processes, since they occur largely at the solid/vapor interface. Chapter 3 deals with the physical processes occurring at
1.3 Overview of the OMVPE Growth Process 15
the surface during epitaxial growth. Chapter 4 adds the complexity due to the precursor molecules themselves. The bond strengths and configurations, seldom mentioned in discussions of the growth process, are important factors in the homogeneous and heterogeneous reactions occurring during OMVPE. We will see that the reactions observed can often be rationalized in terms of these considerations. Reaction kinetics and mass transport nearly always limit the overall reaction rate during OMVPE growth. Since these are subjects that are best dealt with separately, Chapters 5 and 6 treat homogeneous and heterogeneous kinetics and hydrodynamics and mass transport, respectively. Chapters 2 to 6 are organized with a general treatment of the topic, including a review of the basic concepts, preceding a discussion directed specifically toward understanding the OMVPE growth processes. The growth of GaAs using several combinations of precursor molecules is often used to illustrate the basic concepts, since the GaAs system has been studied significantly more than any other system. In Chapter 7, the various factors are brought together in a treatment of the overall process, with GaAs as the major example, but also considering other systems including the 11/VI semiconductors. The 11/VI semiconductors are frequently treated as an independent topic. However, as we will see, a study of the problems encountered in the OMVPE growth of these materials leads to a fuller understanding of the overall growth process for all materials. Another slight departure from the traditional treatment of OMVPE is the inclusion of growth using organometallic (and hydride) precursors in a UHV environment, as already discussed. The synergy gained by considering these dissimilar systems together provides important insights into the complex OMVPE process.
The fundamental understanding of the OMVPE process, developed in Chapters 2-6, is used as the basis of a discussion of process design in Chapter 7. The design of the OMVPE process is considered in terms of choice of precursor molecules, design of the reactor hardware, and choice of growth parameters, including total system pressure, temperature, V/III ratio, and growth rate. The optimum process design is found to be related to the application (i.e., the materials, structures, and materials properties desired). Unfortunately, there can be no one set of parameters that will give optimum results for all applications.
Applications are discussed in Chapters 8-10. In Chapter 8 an effort is made to give guidance about the optimum set of growth parameters for each individual material, stressing III/V and 11/VI systems but also including a brief discussion of the OMVPE growth of oxides, for both dielectric and superconductor applications, and metals. The empirical effects of growth parameters on materials properties are emphasized, with an effort to tie the results to the understanding developed in Chapter 7. Next, special structures, including both superlattices, low-dimensional structures, and growth on dissimilar substrates—GaAs on Si, for example—are discussed in Chapter 9. This is followed by a summary, in Chapter 10, of device results achieved in materials and structures grown by OMVPE.
16 1 Overview of the OMVPE Process
References 1. M. J. Kelly, Adv. Mater. 9 857 (1997). 2. Stephen Jay Gould, "Phyletic Size Decrease in Hershey Bars," in Hen's Teeth and Horse's Toes
(Norton, New York, 1984), pp. 313-319. 3. A. Fowler, Physics Today 50 50 (1997). 4. G. B. Stringfellow, Reports on Progress in Physics 45 469 (1982). 5. G. Beuchet, in Semiconductors and Semimetals, Vol. 22A, ed. W. T. Tsang (Academic Press,
Orlando, 1985), pp. 261-298. 6. C. T. Foxon, in Handbook of Crystal Growth, Vol. 3a, ed. D. T. J. Hurle (Elsevier, Amsterdam,
1994), pp. 155-181. 7. H. M. Manasevit, AppL Phys. Lett. 116 1725 (1969); H. M. Manasevit and W. I. Simpson,
/. Electrochem. Soc. 12 156 (1968); H. M. Manasevit, J. Crys. Growth 13/14 306 (1972). 8. W. Miederer, G. Ziegler, and R. Dotzer, U.S. Patent 3,226,270 (1963). 9. W. Miederer, G. Ziegler, and R. Dotzer, German Patent 1,176,102 (1962).
10. T. R. Scott, G. King, and J. M. Wilson, U.K. Patent 778,383.8 (1954). 11. Y. Seki, K. Tanno, K. lida, and E. Ichiki, / Electrochem. Soc. 122 1108 (1975). 12. For example, see R. D. Dupuis and R D. Dapkus, Appl. Phys. Lett. 32 406 (1978) (lasers); R. R.
Saxena, V. Aebi, C. B. Cooper, M. J. Ludowise, H. A. van der Plas, B. R. Cairns, T. J. Maloney, P G. Bordern, and P E. Gregory, J. Appl. Phys. 51 4501 (1980) (solar cells); R. R. Bradley, / Crys. Growth 55 223 (1981) (LEDs); and J. P Andre, P Guittard, J. Hallais, and C. Piaget, J. Crys. Growth 55 235 (1981) (photocathodes).
13. E. Veuhoff, W. Pletchen. P Balk, and H. Luth, J. Crys. Growth 55 30 (1981). 14. M. B. Panish, / Electrochem. Soc. Ill 2729 (1980). 15. D.W. Shaw, / Crys. Growth 31 130 (1975). 16. R. R. Saxena, C. B. Cooper, M. J. Ludowise, S. Hikido, V. M. Sardi, and PG. Borden, / Crys.
Growth 55 5% {\9U). 17. I. A. Frolov, P. B. Boldyrevskii, B. L. Druz, and E. B. Sokolov, Inorg. Mater (Eng. trans.) 13
632(1977). 18. M. R. Leys and H. Veenvliet, J. Crys. Growth 55 145 (1981). 19. G. B. Stringfellow, in Semiconductors and Semimetals, Vol. 22A, ed. W. T. Tsang (Academic
Press, Orlando, 1985), pp. 209-259. 20. C. Plass, H. Heinecke, O. Kayser, H. Luth, and P Balk, J. Crys. Growth 88 455 (1988). 21. H. Krautle, H. Roehle, A. Escobosa, and H. Beneking, / Electron. Mater 12 215 (1983). 22. D.H. Reep and S.K. Ghandhi, J. Electrochem. Soc. 131 2697 (1984). 23. M. Mizuta, T. Iwamoto, F. Moriyama, S. Kawata, and H. Kukimoto, J. Crys. Growth 68 142
(1984). 24. C. C. Hsu, R. M. Cohen, and G. B. Stringfellow, J. Crys. Growth 63 8 (1983).