1.4k+

# Heteroepitaxial Growth

Authored by: John E. Ayers , Tedi Kujofsa , Paul Rago , Johanna E. Raphael

# Heteroepitaxy of Semiconductors

Print publication date:  October  2016
Online publication date:  October  2016

Print ISBN: 9781482254358
eBook ISBN: 9781315372440

10.1201/9781315372440-4

#### Abstract

Of the many available epitaxial growth techniques, molecular beam epitaxy (MBE) and metalorganic vapor phase epitaxy (MOVPE) have emerged as general-purpose tools for heteroepitaxial research and commercial production. This is because these methods afford tremendous flexibility and the ability to deposit thin layers and complex multilayered structures with precise control and excellent uniformity. Together, MBE and MOVPE account for virtually all production of compound semiconductor devices today.

#### 3.1  Introduction

Of the many available epitaxial growth techniques, molecular beam epitaxy (MBE) and metalorganic vapor phase epitaxy (MOVPE) have emerged as general-purpose tools for heteroepitaxial research and commercial production. This is because these methods afford tremendous flexibility and the ability to deposit thin layers and complex multilayered structures with precise control and excellent uniformity. Together, MBE and MOVPE account for virtually all production of compound semiconductor devices today.

MBE is an ultra-high-vacuum (UHV) technique that involves the impingement of atomic or molecular beams onto a heated single-crystal substrate where the epitaxial layers grow. The source beams originate from Knudsen evaporation cells or gas source crackers. These can be turned on and off very abruptly by shutters and valves, respectively, providing atomic layer abruptness. Because MBE takes place in a UHV environment, it is possible to employ a number of in situ characterization tools based on electron or ion beams. These provide the crystal grower with immediate feedback, and improved control of the growth process. Another advantage of MBE is flexibility; nearly all semiconductors can be grown, including III-V and II-VI semiconductors; Si, Ge, and Si1−xGex alloys; and SiC and Si1−x−yGexC alloys. However, III-phosphides are difficult to grow by MBE and alloys involving As and P are especially troublesome. Other drawbacks of MBE are the initial high cost and maintenance requirements of the UHV system, as well as limited throughput.

MOVPE is a vapor phase epitaxial (VPE) process that is carried out at atmospheric or reduced (e.g., 0.1 atm) pressure using metalorganic precursors. Often, hydride sources are used in conjunction with the metalorganic chemicals; occasionally, even elemental sources are used. Like MBE, MOVPE provides excellent control over the growth of thin layers and multilayered structures, including quantum well devices and superlattices. However, the lack of UHV conditions precludes the in situ use of electron or ion beam characterization tools. Nonetheless, optical in situ characterization methods have been utilized to some extent. Another disadvantage of MOVPE is the use of highly toxic source chemicals, especially arsine (AsH3) and phosphine (PH3). These flammable, explosive, and highly toxic gas sources are stored at high pressure in large quantities, raising a number of safety concerns in a production environment. In some cases, these hydride sources have been replaced with less toxic liquid sources (such as TBAs) contained in low-pressure bubblers.

VPE processes are also used for the growth of column IV semiconductors, including Si, Ge, and Si1−xGex alloys and SiC and Si1−x−yGexC alloys. These VPE processes utilize similar equipment and share some of the characteristics of MOVPE, but do not involve metalorganic precursors. Instead, hydride and halide sources are used. The use of all hydride sources leads to an irreversible process with abrupt interfaces. On the other hand, any involvement of halide precursors generally results in a reversible process that is generally less suitable for the growth of multilayered structures.

This chapter provides a brief overview of the important epitaxial processes of VPE and MBE. MOVPE is considered a special case of VPE. The remaining sections of the chapter describe the growth of particular materials, from the viewpoint of mismatched heteroepitaxy.

#### 3.2  Vapor Phase Epitaxy

VPE growth is accomplished by passing gaseous source chemicals over a heated single-crystal substrate, where epitaxial growth occurs. Atmospheric or reduced (~0.1 atm) pressure may be used. In either case, a carrier gas such as hydrogen usually makes up most of the flow (and therefore pressure) in the reactor.

VPE growth is extremely flexible, and allows the growth of nearly every semiconductor material of interest. The availability of ultrapure sources and careful reactor design allow the growth of materials with levels of purity matching all other epitaxial techniques. It has also proved possible to design reactors capable of handling multiple wafers in one run, while maintaining excellent uniformity, both across wafers and from wafer to wafer. This scalability has led to its wide commercial application.

#### 3.2.1  Vapor Phase Epitaxial Mechanisms and Growth Rates

The VPE growth of a crystal involves a series of basic steps: (1) the source chemicals are transported in the vapor phase to the heated substrate; (2) source molecules diffuse to the growing surface, where they are chemisorbed or physisorbed; (3) adsorbed species on the surface react to form the solid crystal; (4) reaction products diffuse from the surface; and (5) reaction products are carried away in the flowing gas stream. The slowest of these five steps will determine the growth rate. Typically, the rate limiter is either step 2 or step 3, and these two situations are called mass transfer limited (or diffusion limited) and reaction rate limited, respectively.

Consider the transport of a single reactant to the growing surface. (Usually, one reactant is provided in excess for the growth of a binary compound, so that this assumption remains useful.) The flux of this species to the surface at a particular point is given by Henry’s law:

3.1 $j = h ( N g − N 0 ) ,$

where:

• Ng is the concentration of the reactant in the gas phase
• N0 is the concentration of the reactant at the surface
• h is the gas phase mass transfer coefficient.

Suppose the reaction rate is linear; then

3.2 $j = k N 0 ,$

where k is the surface reaction rate constant. Usually, this rate is thermally activated so that

3.3 $k = k 0 exp ⁡ ( − E a / k T ) ,$

where Ea is the activation energy for the process. Typical activation energies are in the range of 25–100 kcal mole−1 (1.1–4.3 eV molecule−1).

Under steady-state conditions, the two fluxes above may be equated. Combining these equations, we can determine the growth rate as

3.4 $g = j n = N g n ( h k h + k ) ,$

where n is the number of atoms (or molecules) per unit volume in the growing crystal. For Si, n = 5 × 1022 cm−3 (Si), and for GaAs, n = 2.2 × 1022 cm−3 (Ga).

At low temperatures, kh, so that

3.5 $g ≈ k N g n .$

This is referred to as reaction-rate-limited growth. Under these conditions, the growth rate is a strong function of temperature. Here, the reaction rate is sensitive to the surface conditions and the growth rate depends on the orientation of the crystal substrate. This can result in faceted growth, which is usually undesirable, but can be advantageous in some situations, such as the implementation of epitaxial lateral overgrowth (ELO).

At high temperatures, hk, so that

3.6 $g ≈ h N g n .$

This situation is known as mass-transfer-limited growth (or diffusion-limited growth). Under mass-transfer-limited conditions, the growth rate is independent of the crystal orientation and nearly independent of temperature. There is a slight temperature variation (with an activation energy of 3–8 kcal mole−1) due to the temperature dependence of the diffusivity. Usually, VPE reactors are designed to operate in the mass-transfer-limited regime. However, this is not always possible due to constraints imposed by the substrate or source chemicals.

Thermodynamic considerations may also be important in determining the growth rates in the mass-transfer-limited regime. This is illustrated in Figure 3.1, which shows the general behavior for the cases of the (a) endothermic and (b) exothermic processes. For the endothermic process, which involves a positive heat of reaction, the growth rate increases monotonically with increasing temperature. In the reaction-rate-limited regime, the activation energy is large, typically 25–100 kcal mole−1, whereas in the mass-transfer-limited region, there is only a slight variation of the growth rate with temperature (3–8 kcal mole−1). In contrast, for the exothermic process, which is favored at lower temperatures, the growth rate decreases rapidly with increasing temperature under conditions of mass-transfer-limited growth. Many processes of interest for heteroepitaxy are either endothermic (chloride VPE of Si) or pyrolytic (hydride VPE of Si or SiC, MOVPE) in nature, and so display the general behavior shown in Figure 3.1a.

Figure 3.1   Growth rate (log scale) vs. reciprocal of temperature for (a) an endothermic process and (b) an exothermic process.

#### 3.2.2  Hydrodynamic Considerations

The simplified model of Section 3.2.1 fails to reveal many of the details associated with the fluid dynamics of vapor phase epitaxy. Nor does it give guidance in the determination of the mass transfer coefficient. However, in the typical case of laminar flow, simple analytical solutions exist for the determination of the growth rate under mass-transfer-limited conditions.

The nature of the gas flow in an epitaxial reactor can be understood based on a study of gas flow in simple pipes.1 This behavior may be characterized by the unitless Reynold’s number, given by

3.7 $N R = d v ρ μ ,$

where:

• d is the pipe diameter
• v is the gas velocity in the pipe
• μ is the absolute viscosity
• ρ is the gas density

The viscosity of hydrogen, the most commonly used carrier gas, varies from about 200 × 10−6 dyn cm−1 s−1 (200 μP) at 700°C to about 250 × 10−6 dyn cm−1 s−1 (250 μP) at 1200°C. The density of hydrogen at atmospheric pressure varies from about 2.5 × 10−5 g cm−3 at 700°C to 1.65 × 10−5 g cm−3 at 1200°C. Empirically, it is found that the transition from laminar flow occurs with the Reynold’s number in the range of 2000 < NR < 3000, and typical epitaxial reactors operate under laminar flow conditions, with NR ≈ 30.

Consider the idealized reactor shown in Figure 3.2, with a recessed susceptor and a constant cross-sectional area. The reactor height is h. Suppose that a single reactant contributes to the growth, and the concentration of this reactant is Ng at the entrance to the reactor. The growth is assumed to take place under mass-transfer-limited conditions so that N0 ≈ 0.

The growth rate will be proportional to the flux of reactant species arriving at the substrate surface. If this flux is controlled by diffusion, then

3.8 $j = − D ∂ N ∂ y ,$

Figure 3.2   (a–c) Concentration of reactants in a horizontal reactor, and the growth rate. (Adapted from S. K. Ghandhi, VLSI Fabrication Principles, 2nd ed. [New York: John Wiley & Sons, 1994]. Used with permission.)

where:

• N is the actual concentration of the reactant in the gas phase at a point above the susceptor
• D is the diffusivity of the reactant species in the carrier gas
• y is the distance from the substrate interface

If the gas velocity is assumed to be constant above the susceptor, then under steady-state conditions (with all time derivatives equal to zero) the two-dimensional (2-D) continuity equation for the reactant species is

3.9 $∂ N ∂ t = D ∂ 2 N ∂ x 2 + D ∂ 2 N ∂ y 2 − v ∂ N ∂ x = 0 ,$

where v is the gas velocity and x is the horizontal distance from the edge of the wafer. If diffusion is neglected in the flow direction, then

3.10 $0 = D ∂ 2 N ∂ y 2 − v ∂ N ∂ x .$

The boundary conditions are

3.11 $N = N g ; x = 0 ; 0 < y < h ; N = 0 ; 0 < x , y = 0 ; ∂ N ∂ y = 0 ; 0 < x , y = h .$

Solving, the flux of reactant is found to be

3.12 $j = − 2 D N g h ∑ r = 0 ∞ exp ⁡ ( − π 2 D x ( 2 r + 1 ) 2 4 v h 2 ) .$

If the flux of reactant species is envisioned to flow through a “diffusion boundary layer” in which its concentration varies linearly from Ng to 0, then this boundary layer thickness is given by

3.13 $δ D ( x ) = h 2 [ ∑ r = 0 ∞ exp ⁡ ( − π 2 D x ( 2 r + 1 ) 2 4 v h 2 ) ] − 1 .$

Under the simplifying assumption that (h2v/πD)>x, we obtain

3.14 $j ≈ − N g D v π x ,$

so that the diffusion boundary layer thickness is given by

3.15

and the growth rate may be estimated from

3.16 $g ≈ N g D n δ D .$

It is apparent from this analysis that the achievement of uniform growth over a large wafer requires a uniform diffusion boundary layer thickness. This can be achieved by tilting the susceptor, so that the cross section of the reactor decreases with x. Sometimes, this is done in conjunction with adjustments to the total flow and reactor pressure in order to achieve the desired result.

The increased performance of computers has enabled detailed numerical computations of the mass and heat flows in epitaxial reactors. These calculations take into account continuity, conservation of momentum (Navier–Stokes equations), conservation of energy, and conservation of mass of diffusing species. Thus, the growth rate and uniformity (in thickness and composition) can be predicted for a reactor design before it is built and tested. Nonetheless, the simple boundary layer picture allows one to construct a starting point for the reactor design, which can then be fine-tuned by the use of computation.

#### 3.2.3  Vapor Phase Epitaxial Reactors

A VPE reactor comprises a gas delivery system, a reaction chamber, and an effluent handling system; a basic setup is shown schematically in Figure 3.3. Welded stainless steel construction with metal gasket fittings is used to achieve the necessary leak integrity.

Typically, ultra-high-purity (UHP) H2 with 7N purity (seven nines purity, or 99.99999% pure) is used as the carrier gas in a VPE reactor. This gas may be purchased in UHP form, or commercial-grade hydrogen (3N5, or 99.95% pure) may be purified by diffusion through a palladium-silver membrane or by the use of a purifying resin. Other carrier gases may also be used, such as UHP N2, He, or Ar. Although these gases preclude the use of a palladium cell, resin purifiers are available for use with them.

Figure 3.3   VPE reactor.

The source chemicals may be gaseous, liquid, or solid. Gas sources (such as SiH4 and AsH3) may be obtained in pure form, or diluted in hydrogen, in high-pressure cylinders. Liquid sources (such as SiH2Cl2 or TMGa) are typically obtained in UHP form, in stainless steel bubblers. Solid sources (such as TEIn) may also be obtained in bubbler vessels. However, it is difficult to obtain good run-to-run repeatability with these. Occasionally, vapor phase sources are created in situ, as in the case of GaCl, which was used as the Ga source in the hydride and halide epitaxial processes for GaAs.

For a gaseous source, the flow is metered precisely by an electronic mass flow controller (MFC). The MFC has a built-in heated capillary. The measurement of the temperature difference across the capillary allows the determination of the mass flow with a precision of ±0.5%. MFCs have built-in closed-loop control systems and metering valves, so they can maintain the flow at a desired set point. Typically, MFCs are calibrated for use with pure H2 or N2. Hydrogen calibration is entirely adequate for a dopant gas that is diluted to a few 100 ppm in H2. In the case of pure gaseous sources such as AsH3, the MFC must be calibrated specifically for each source.

Liquid sources are transported to the reactor by a carrier gas, usually H2, which is passed through a stainless steel bubbler arrangement like the one shown in Figure 3.4. The carrier gas is metered precisely by an MFC placed upstream of the bubbler. If the mass flow of the carrier gas is $F H 2$

, the vapor pressure of the liquid source is PS, and the total pressure in the bubbler is PTOT, then the mass flow of the source is given by
3.17 $F S = F H 2 ( P S P T O T − P S ) .$

Figure 3.4   Liquid source bubbler.

This expression assumes that the carrier gas bubbles have sufficient residence time in the liquid to become saturated with its vapor, and closely approximates a real bubbler application. Typically, a three-valve arrangement is used around the bubbler so that the carrier gas can be made to bypass the bubbler when the source is not in use.

The vapor pressures of liquid sources are usually fit by the expression

3.18 $log ⁡ 10 P S = A − B T ,$

where:

• PS is the vapor pressure over the liquid (in torr)
• T is the absolute temperature
• A, B are empirical constants

For convenience, the temperature of a bubbler source is usually set to yield a source vapor pressure of 5–50 torr by means of a temperature-controlled bath. In some cases, the bubbler temperature must be kept above room temperature, which necessitates heating of the downstream lines to prevent condensation of the source.

The carrier gas and vapor phase sources are brought to a mixing manifold prior to their injection into the reactor chamber. In the mixing manifold, gas flows are switched between pressure-balanced vent and reactor lines. Thus, each flow may be stabilized to the vent line before being switched to the reactor. Precise pressure balancing, using a differential pressure transducer and a control valve, avoids unwanted flow transients.

VPE reactors are of the horizontal, vertical, or barrel types. The horizontal configuration is the simplest, and often used in research. The reaction chamber is a quartz tube, flanged at one end to facilitate loading and unloading of wafers. The substrate wafers are held by recesses in a graphite susceptor, which is usually tilted at a slope of 7°–10° to improve the thickness uniformity. The configuration is so named because of the horizontal gas flow in the tube.

The vertical reactor utilizes a vertical flow of gases, perpendicular to the surface of the wafers. Inherent in this geometry is a stagnation point at the center of the susceptor, where the gas velocity is zero. This tends to promote recirculation unless the susceptor is rotated at a high speed (>1000 rpm). Rotation also serves to improve the radial uniformity of growth. In addition, the flow, pressure, and rotation rate must be optimized for radial uniformity. Sometimes, complex planetary rotation systems are employed as well.

The barrel reactor can handle many wafers in a single run and achieves high throughput. Wafers are held in shallow depressions within the steeply sloped susceptors, and the gas flow is nearly parallel to their surfaces. Thus, the barrel reactor geometry is similar to that of the horizontal reactor, but rotated 90°.

Heating of an epitaxial reactor may be accomplished by radio frequency (RF) induction, infrared lamps, or resistive heaters. In cold-wall reactors used for endothermic and pyrolytic reactions, RF induction or infrared lamps are typically used. Internal resistive heaters are occasionally used, but the materials must be chosen carefully to avoid metallic contamination. External resistive heaters avoid this problem but are only applicable to hot-wall reactors used for exothermic processes. Temperature control with ±2°C precision is normally adequate if the growth is mass transfer limited. Measurement of the temperature can be achieved using optical pyrometry, or for low-temperature reactors (<900°C), thermocouples may be embedded in the susceptor.

Susceptors are usually made from machined graphite. At temperatures above 1300°C, however, the hydrogen carrier gas will react with graphite and etch its surface. For this reason, SiC-coated susceptors are often employed in reactors intended for high-temperature operation. Coatings of SiC or pyrolytic BN are sometimes used in lower-temperature reactors as well, to avoid the outgassing affects associated with the porosity of uncoated graphite.

Epitaxial reactors may operate at atmospheric or reduced (~0.1 atm) pressure. Low-pressure operation reduces the surface coverage of adsorbed species, increasing their mobility and allowing high-quality growth at reduced temperatures (50°C–100°C lower than for atmospheric growth). Reduced pressure can also decrease the tendency for recirculation under otherwise similar conditions. In low-pressure reactors, pressure control is achieved using a mechanical vacuum pump and a butterfly valve. Pressure measurement may be by capacitance or optical manometers.

The effluent is typically treated by activated charcoal absorption units or liquid scrubbers before being vented to the outside. These systems require frequent service as well as expensive waste disposal.

#### 3.2.4  Metalorganic Vapor Phase Epitaxy

The MOVPE process was developed in the late 1960s by Manesevit,24 who first demonstrated its use for the epitaxy of Ga-V compounds. Subsequently, the process has been adapted to nearly all III-V and II-VI semiconductors, including the antimonides, arsenides, phosphides, nitrides, sulfides, selenides, and tellurides, and also ternary and quaternary alloys. MOVPE-grown material is of extremely high purity: this epitaxial method has produced the highest-purity InP produced by any method and GaAs, which is as pure as that grown by any technique. Specially designed reactors have also made possible the growth of very abrupt interfaces and multilayered structures of the type necessary for quantum layer devices such as laser diodes and high-speed transistors. These developments and the ease of scaling the MOVPE process to high throughput have made it important for commercial production, as well as laboratory research.

MOVPE goes by a number of names, including organometallic vapor phase epitaxy (OMVPE), metalorganic chemical vapor deposition (MOCVD), organometallic chemical vapor deposition (OMCVD), and occasionally organometallic epitaxy (OME). Chemical vapor deposition (CVD) is a more general term that applies to noncrystalline films; as such, the more specific term vapor phase epitaxy should be used to refer to epitaxy. OMVPE is preferred by many researchers because it is consistent with the normal chemical nomenclature. MOVPE is the name chosen by the international conference and is used throughout this book. It is important to realize that these terms are used interchangeably in the literature, and are not meant to refer to different processes.

MOVPE is carried out in a reactor of the type shown schematically in Figure 3.3. Source chemicals are transported to the reactor by a carrier gas, where they react heterogeneously at the surface of a heated single-crystal substrate. In the growth of a binary semiconductor, one or both source chemicals may be metalorganic compounds. These are typically liquids at room temperature and are transported to the reactor by flow of carrier gas through a stainless steel bubbler. Gaseous sources may also be used; these are contained in high-pressure cylinders in either pure or diluted form. Ternary or quaternary alloys may be grown by introducing additional source chemicals. Changes in composition can be realized by ramping or switching the source flows. Layers may also be doped by the introduction of small concentrations of the appropriate sources.

MOVPE processes are pyrolytic in nature. As a consequence, cold-wall reactors are used almost exclusively. Also, the irreversible nature of MOVPE allows the growth of extremely abrupt interfaces.

The variety of metalorganic source chemicals has increased greatly over the years due to the success of the method and resulting market demand. In general, the sources are molecules of the type MRn, where M represents a metal atom and R represents an organic radical. It is common practice to refer to the organic groups using M, E, NP, IP, NB, IB, TB, A, and Cp for methyl, ethyl, n-propyl, i-propyl, n-butyl, i-butyl, t-butyl, allyl, and cyclopentadienal, respectively. M, D, and T are used to denote mono-, di-, and tri-, respectively. Thus, TMGa represents trimethylgallium and DETe refers to diethyltelluride.

The metalorganic precursors are generally liquids at room temperature, contained in stainless steel bubbler vessels. A few are solid at room temperature, but these can be used with a bubbler arrangement as well. The melting points, boiling points, and vapor pressure parameters A and B are given in Tables 3.1 through 3.4 for sources of elements from columns II, III, V, and VI, respectively. As a general rule, the vapor pressures are highest for the lightest molecules.

The decomposition characteristics of the alkyl source molecules are determined in part by the strength of the metal–carbon bond. This bond energy determines the stability of the molecule with respect to decomposition by the removal of organic radicals (free radical homolysis). Therefore, it often determines the activation energy for reaction-rate-limited growth. In general, the metal–carbon bond strength decreases with the number of carbons bonded to the central carbon of the molecule (methyl > ethyl > isopropyl > tertiary butyl > allyl). In some situations, this means that a lower growth temperature may be used with an isopropyl source than with a methyl source. Bond strengths for some of the common alkyl precursors are provided in Table 3.5.

### Table 3.1   Melting Points, Boiling Points, and Vapor Pressure Data for Metalorganic Sources of Column II Elements (log10P(torr) = A−B/T)

Vapor Pressure

Precursor

Melting Point (°C)

Boiling Point (°C)

A

B (K)

P (torr) @ T (°C)

DMZn

–42

46

7.802

1560

124 @ 0°C

DEZn

–28

118

8.280

2109

3.6 @ 0°C

DMCd

–2

106

7.764

1850

9.7 @ 0°C

### Table 3.2   Melting Points, Boiling Points, and Vapor Pressure Data for Metalorganic Sources of Column III Elements (log10P(torr) = A−B/T))

Vapor Pressure

Precursor

Melting Point (°C)

Boiling Point (°C)

A

B (K)

P (torr) @ T (°C)

TMAl

15

126

8.224

2134.83

2.2 @ 0°C

TEAl

–52.5

186

10.784

3625

0.5 @ 55°C

TMGa

–15.8

55.8

8.501

1824

66 @ 0°C

TEGa

–82.5

143

9.172

2532

3.4 @ 20°C

DEGaCl

–7

8.78

2815

0.5 @ 60°C

TMIn

88

135.8

10.520

3014

0.3 @ 0°C

TEIn

–32

184

1.2 @ 40°C

### Table 3.3   Melting Points, Boiling Points, and Vapor Pressure Data for Metalorganic Sources of Column V Elements (log10P(torr) = A−B/T)

Vapor Pressure

Precursor

Melting Point (°C)

Boiling Point (°C)

A

B (K)

P (torr) @ T (°C)

TMP

–84

38

7.7627

1518

381 @ 20°C

TEP

–88

127

8.035

2065

46.5 @ 50°C

TBP

4

54

7.586

1539

141 @ 10°C

TMAs

–87.3

50

7.3936

1456

238 @ 20°C

TEAs

140

5 @ 20°C

DMAs

36.3

7.532

1443

176 @ 0°C

DEAs

102

7.339

1680

40 @ 20°C

TBAs

–1

65

7.243

1509

32 @ –10°C

TMSb

–87.6

80.6

7.73

1709

48.9 @ 10°C

TESb

–98

160

7.90

2183

4 @ 25°C

TMBi

–107.7

110

7.628

1816

27 @ 20°C

### Table 3.4   Melting Points, Boiling Points, and Vapor Pressure Data for Metalorganic Sources of Column VI Elements (log10P(torr) = A−B/T)

Vapor Pressure

Precursor

Melting Point (°C)

Boiling Point (°C)

A

B (K)

P (torr) @ T (°C)

DES

–100

91 ± 1

8.184

1907

47 @ 20°C

DTBS

149 ± 2

DMSe

57

DESe

108

7.905

1924

7.2 @ 0°C

DMTe

–10

92 (82)

7.97

1865

65 @ 30°C

DMDTe

220

6.94

2200

0.26 @ 23°C

DETe

137

7.99

2093

7.1 @ 20°C

DIPTe

8.29

2309

2.6 @ 20°C

### Table 3.5   Bond Strengths for Common Alkyl Precursor Molecules

Precursor

D 1 (kcal mole−1)

D 2 (kcal mole−1)

Dave (kcal mole−1)

DMZn

51 (54)

47

42

DMCd

53

46

33

TMAl

65

66, 61

TEAl

58

TMGa

59.5

35.4

59

TEGa

57

TMIn

47

TMP

66, 63

TMAs

62.8

55

TMSb

57

57

52, 47

DES

65

Generally, the alkyls of column II and column III elements are Lewis acids (electron acceptors), whereas the alkyls of column V and column VI atoms are Lewis bases (electron donors). It is possible for a gas phase reaction to occur between alkyls with Lewis acid and Lewis base character, resulting in an adduct. If the adduct so produced is a low-vapor-pressure molecule, it may not contribute to epitaxial growth, and in fact, it may give rise to fouling of the reactor. Such a parasitic reaction is highly undesirable. On the other hand, some adducts can contribute to growth, and these (such as TMIn-TEP) are sometimes used intentionally as sources.

#### 3.3  Molecular Beam Epitaxy

MBE is a UHV technique that involves the impingement of atomic or molecular beams onto a heated single-crystal substrate where the epitaxial layers grow.5 The source beams originate from Knudsen evaporation cells or gas source crackers. These can be turned on and off very abruptly by shutters and valves, respectively, providing atomic layer abruptness.

Because MBE takes place in a UHV environment, it is possible to employ a number of in situ characterization tools based on electron or ion beams. These provide the crystal grower with immediate feedback, and improved control of the growth process.

MBE has been developed to the point where nearly every semiconductor of interest may be grown using the technique, including III-V and II-VI semiconductors; Si, Ge, and Si1−xGex alloys; and SiC and Si1−x−yGexC alloys. However, III-phosphides are difficult to grow by MBE, and alloys involving As and P are especially challenging. Other drawbacks of MBE are the initial high cost and maintenance requirements of the UHV system and low growth rates. These drawbacks are offset to a large extent by the precise control and in situ characterization so that MBE is used extensively for commercial device production at this time.

An MBE reactor involves a number of source cells arranged radially in front of a heated substrate holder, as shown in Figure 3.5. The source cells supply all atoms necessary for the growth and doping of the required semiconductor layers; six or more cells may be required. The simplest type of source cell is a thermal evaporator (Knudsen cell), but other more elaborate schemes have been developed for some atoms. A basic requirement for MBE growth is line-of-sight source impingement. This means that the evaporated source atoms must have mean free paths greater than the source-to-substrate distance, which is typically 5–30 cm. This requirement places an upper limit on the operating pressure for an MBE reactor.

Figure 3.5   MBE reactor. (Reprinted from M. Henini, Thin Solid Films, 306, 331 [1997]. With permission. Copyright 1997, Elsevier.)

The mean free path for an evaporated particle (atom or molecule) may be estimated if it is assumed that all other particles in the system are at rest. Suppose the evaporated particle is moving at a velocity c, and all particles have a round cross section with diameter σ. Two particles that pass at a distance of σ or less will collide. Therefore, each particle can be considered to have a collision cross section of πσ2, and the collision volume swept out by a particle in time dt is πσ2cdt. If N is the volume concentration of particles, then the collision frequency will be

3.19 $f = N π σ 2 c ,$

and the mean free path will be

3.20 $λ = c f = ( N π σ 2 ) − 1 .$

A more accurate calculation of the mean free path may be made assuming that all of the particles are in motion. Based on this, the mean free path for an evaporated particle is

3.21 $λ = ( N π σ 2 2 ) − 1 = k T 2 π σ 2 P ,$

where P is the pressure.

Typical values of the cross section diameter σ range from 2 to 5 Å, so that the mean free path is about 103 cm at a pressure of 10−5 torr. This pressure therefore represents an approximate upper limit for the system pressure during growth, if the beam nature of the sources is to be maintained.

The requirement on the base pressure is considerably more stringent, and is set by purity requirements. If the grown films are to have no more than 10−5 (10 ppm) contaminants, then the base pressure should be no more than 10−10 torr. Achievement of the necessary UHV requires the use of a stainless steel chamber with metal gaskets. The system must be load locked, so that it is opened to the atmosphere only for maintenance. Any exposure of the chamber to air must be followed by a long bakeout to remove adsorbed contaminants. During growth, the chamber walls must be cooled to cryogenic temperatures by means of a liquid nitrogen shroud, in order to further reduce evaporation from this large surface area.

Growth of pure layers by MBE also requires the use of oil-free pumping in the UHV system. Cryogenic sorption pumps, titanium sublimation ion pumps, and turbomolecular pumps are used for this reason.

The simplest source cells are thermal evaporators, called effusion cells or Knudsen cells. High-purity elemental sources are used, and one cell is needed for each element. Typically, the effusion cells are made of pyrolytic boron nitride with tantalum heat shields. The source temperatures are maintained precisely (±0.1°C) to control the flux of evaporating atoms. Due to the inability to rapidly ramp up or down the cell temperature, a shutter is used to turn each beam on and off.

The flux of atoms from such an effusion cell may be calculated using the kinetic theory of gases.6 From this treatment, it can be shown that the evaporation rate from a surface area Ae is given by

3.22 $d N e d t = A e P 2 π k T m ,$

where:

• P is the equilibrium vapor pressure of the source at the effusion cell temperature T
• m is the mass of the evaporant

In terms of the molecular mass of the species, M, the effusion rate is

3.23 $d N e d t = A e P 2 π k T M / N A ,$

where NA is Avogadro’s number (6.022 × 1023 mole−1). Simplifying,

3.24

where P is the pressure in torr. Because the equilibrium vapor pressure P varies exponentially with temperature, the effusion cell temperature must be controlled to within ±0.1°C in order to keep the effusion rate within a ±1% tolerance.

The flux of evaporant arriving at the substrate surface can be calculated from the evaporation rate at the effusion cell by

3.25

where:

• l is the distance from the effusion cell to the substrate
• θ is the angle between the beam axis and the normal to the substrate

The model outlined above assumes a full effusion cell so that evaporation occurs at its mouth. In practice, the cell depletes with time, and this causes a falloff of the impingement rate, and a change in the beam profile,7 at the substrate. This effect can be mitigated to some extent by the use of tapered effusion cells.

Usually, the evaporation crucibles have a 1 cm2 evaporation surface and are located 5–20 cm from the substrate. Typical source pressures are 10−3–10−2 torr, resulting in the delivery of 1015–1016 molecules cm−2 s−1. This corresponds to a growth rate on the order of one monolayer per second, assuming a unity sticking coefficient for the impinging atoms.

Thermal effusion sources are switched on and off by means of pneumatically controlled shutters. A problem associated with this scheme is the change in thermal loading on the cell upon opening or closing the shutter. This causes unwanted temperature transients in the cell, which result in rather large variations (up to 50%) in the beam flux immediately after the shutter is opened.

Another disadvantage of thermal effusion sources is the inability to ramp the beam flux rapidly with time. Here, the limitation is due to the thermal mass of the effusion cell. This places an upper limit on the rate at which the composition may be ramped in a ternary or quaternary alloy. Whereas this restriction is important in the growth of multilayered or rapidly graded device structures, it is usually not a problem in graded buffer layers. In gas source MBE (GSMBE), the sources are controlled using MFCs that allow much more rapid ramping.

Electron beam evaporation sources have been used with some elements such as Si. Here, the elemental source is contained in a water-cooled crucible and evaporation occurs locally at the surface by the impingement of an electron beam. On and off control can be achieved by blanking of the electron beam. Scanning of the beam allows the realization of an extended-area source with characteristics similar to those of the thermal effusion cells.

#### 3.4  Silicon, Germanium, and Si1−xGex Alloys

Si, Ge, and their alloys may be grown by either VPE or MBE. Si(001) substrates are used almost exclusively for the epitaxy of these materials. Therefore, the in situ removal of the native oxide is a critical step prior to epitaxy. In the case of MBE, this can be achieved by flashing to a temperature up to 1200°C in the high vacuum. Prior to VPE growth, the oxide layer can be removed by a bakeout in hydrogen.

A number of sources can be used for Si VPE, including silicon tetrachloride (SiCl4), trichlorosilane (SiHCl3), dichlorosilane (SiH2Cl2), and silane (SiH4); however, only dichlorosilane and silane are in common use at this time. This dichlorosilane process is heterogeneous (it requires two molecules of SiCl2) and surface catalyzed (it occurs only in the presence of the silicon surface). It is also reversible and is accompanied by etch-back and autodoping processes, whereby atoms from the grown crystal are etched and returned to the gas phase. These processes are undesirable in multilayered epitaxial device structures, because they compromise the abruptness of heterojunctions and also lead to nonideal doping profiles in p-n junctions. However, they can be suppressed by a reduction of the growth temperature.

The silane process is irreversible due to the absence of chlorine. Compared with the chlorosilanes, SiH4 epitaxy can be carried out at a lower temperature but is extremely sensitive to oxidizing impurities. Silane epitaxy therefore mandates the use of load locks and careful bakeout procedures to avoid the formation of silica dust, which is detrimental to layer morphology. A unique aspect of the silane process is that homogeneous, gas phase nucleation is possible with this source.8 The dusting that results from homogeneous nucleation can also deteriorate layer quality. However, this problem can be minimized by the use of low-pressure, high gas velocities, and reduced temperature.

The VPE growth of Ge has been achieved using a number of halogenic sources,9,10,11 including germanium tetrabromide (GeBr4), germanium tetrachloride (GeCl4), and germanium diiodide (GeI2), as well as the hydride, germane (GeH4). The halide processes are reversible, and therefore accompanied by undesirable autodoping effects, which limit the abruptness of junctions. In addition, the iodide process is complex, requiring three separate temperature zones. For these reasons, the germane process is the preferred method of growth for germanium today and is used for the realization of Ge/Si and germanium-on-insulator (GOI).

Si1−xGex epitaxy may be carried out using a mixture of silicon and germanium sources in the vapor phase. The gas phase mole fraction is used to control the resulting solid phase mole fraction x. Practical systems for Si1−xGex VPE utilize SiH2Cl2 + GeH4 or any combination of the sources SiH4, Si2H6, GeH4, and Ge2H6. In the case of SiH2Cl2 + GeH4, the solid composition x depends on the ratio of the gas source flows by12

3.26 $x 2 1 − x = 2.66 X GeH 4 X SiH 2 Cl 2 .$

Selective growth of Si1−xGex may be achieved by the use of SiH2Cl2 + GeH4 + HCl; growth proceeds on bare silicon surfaces but not on dielectric films such as SiO2 or silicon nitride. This can be utilized in patterned or nanoheteroepitaxial growth schemes. Commercial Si1−xGex VPE reactors provide for the use of either combination of sources, to allow either nonselective (blanket) or selective growth. However, growth over a dielectric film is polycrystalline and should be properly referred to as chemical vapor deposition, not vapor phase epitaxy.

In addition to Si1−xGex, the carbon-containing alloys Si1−yCy and Si1−x−yCyGex are of interest for bandgap engineering of heteroepitaxial devices on Si wafers. These materials may be grown by the addition of a carbon precursor to the growth chemistry, and practical VPE systems employ monomethylsilane (SiCH6) for this purpose. Due to the extremely low solubility of C in Si (<10−6), all practical carbon-containing alloys are necessarily metastable13 and must be grown at low temperatures.

UHV vapor phase epitaxy14 has also been used to grow Si1−xGex alloys, with a growth pressure of ~10−3 atm. Under UHV conditions, good-quality layers may be grown with a cold-wall reactor and the homogeneous nucleation is suppressed. Any combination of the sources SiH4, Si2H6, GeH4, and Ge2H6 may be used, but the use of Si2H6 has been reported to provide better surface morphology.

Si1−xGex alloys across the entire compositional range may be grown by MBE using electron beam sources. Typical temperatures range from 500°C to 900°C. Usually, films with higher Ge content are grown at lower temperatures, keeping the growth temperature at approximately 60%–70% of the melting temperature. Temperature ramping may be employed during the growth of a graded layer. A unique aspect of MBE growth is the ability to grow Si1−xGex films at very low temperatures15 (300°C–400°C); the altered kinetics of lattice relaxation appears to enable the growth of layers with reduced threading dislocation densities. Kinetic models for lattice relaxation are described in Chapter 5.

In the fabrication of multilayered device structures, there is a tendency for three-dimensional (3-D) growth of Si1−xGex on Si, except at low values of x. However, the islanding can be suppressed by reduction of the growth temperature (<600°C for MBE), thus allowing the growth of strained-layer superlattices.16 Equilibrium modes of growth for SiGe/Si (111) and SiGe/Si (100) are discussed in Chapter 4.

#### 3.5  Silicon Carbide

Epitaxial SiC may be grown using various combinations of precursors. The most commonly used silicon source is SiH4,17 but Si2H618 and SiCl419 have also been used. The most popular carbon source is C3H8, but the sources C2H2,20 CH3Cl,21 CH4,22 CCl4, C7H8, and C6H14 have been used as well. Some researchers have even demonstrated the growth of SiC from a single precursor. Sources of this type include CH3SiCl323 and (CH3)2SiCl2.22

Usually SiC epitaxy is carried out in the system SiH4 + C3H8 + H2 in the temperature range of 1200°C–1800°C, with growth rates of 1–5 μm h−1.24 The quality of the epitaxial SiC is strongly dependent on the C/Si ratio in the gas phase. Typically, this ratio is 3:1, corresponding to a C3H8/SiH4 ratio of 1:1, although this depends on the reactor.

SiC may also be grown by GSMBE using the sources SiH4 + C2H425 or Si2H6 + C2H4.26,27 Using the sources SiH4 + C2H4, and 0.75 sccm flow of each, very low growth rates are obtained: 3 nm h−1 at 1000°C to ~50 nm h−1 at 1500°C. The addition of H2 increases the growth rates dramatically (0.2 μm h−1 at 1500°C). The growth rate depends on both source flows because neither is in strong excess.

The most common substrate for heteroepitaxial growth of SiC is Si. Usually, Si (111) is used for the heteroepitaxy of 4H-SiC or 6H-SiC, due to the threefold symmetry of its surface. On the other hand, the cubic polytype 3C-SiC may be grown heteroepitaxially on Si (001).

Misoriented substrates are often employed for SiC epitaxy. For heteroepitaxy of 3C-SiC on Si (001), it is necessary to use misoriented substrates in order to eliminate inversion domain boundaries. For homoepitaxy of 6H-SiC (0001), the growth temperature may be lowered significantly (e.g., from 1800°C to 1500°C) by the use of substrates that are misoriented by a few degrees from the (0001) plane toward a $⟨ 11 2 ¯ 0 ⟩$

direction. Growth on exact (0001) substrates at low temperatures is characterized by mixed 3C and 6H phases, but misorientation of the substrate by 1° or more toward the $⟨ 11 2 ¯ 0 ⟩$ direction eliminates this problem and allows growth of single-phase 6H-SiC at 1500°C. This technique is called step-controlled epitaxy28,29 and is now commonly employed for the fabrication of SiC devices.

#### 3.6  III-Arsenides, III-Phosphides, and III-Antimonides

GaAs and the ternary AlxGa1−xAs may be grown by MOVPE or MBE, and commercial production of AlxGa1−xAs lasers is split between these two techniques. These materials are most often grown on GaAs substrates. However, heteroepitaxy on Si and InP substrates has been investigated extensively with the goal of integrating AlxGa1−xAs devices with those from these other material systems.

MOVPE growth is carried out with the sources TMGa + TMAl + AsH3. Typically, a growth rate of ~10 μm h−1 is achieved using a mole fraction XTMGa = 10−4 and a growth temperature of 650°C. The growth rates for GaAs and AlAs are proportional to the respective organometallic source mole fractions. The V/III ratio is 5–30 for atmospheric pressure, but higher values may be used for reduced pressure growth.

Truly selective area growth of GaAs is possible using the source combination DEGaCl + AsH3 and a SiO2 mask. The GaAs grows where windows have been opened in the oxide, but there is no deposition on the oxide itself. Moreover, this approach can be extended to AlxGa1−xAs by using DEGaCl + DEAlCl + AsH3.

In the case of MBE, elemental sources (7N Ga, 6N As, and 6N Al) are used in conventional Knudsen cells. A temperature of 550°C–600°C is used, with growth rates of 0.1–1 μm h−1.

GaAs heteroepitaxy on Si (001) substrates raises a number of challenging problems. The growth mode is 3-D (Volmer–Weber), so a low-temperature nucleation layer must be used in order to obtain a smooth device layer. Inversion domain boundaries (also known as antiphase domain boundaries) are produced if on-axis substrates are used, but this problem can be eliminated by the use of Si substrates that are misoriented by 2°–4°. Inversion domain boundaries are considered in Chapter 4. The large lattice mismatch (~−4%) results in large threading dislocation densities (~108−109 cm−2). Also, GaAs has about twice the thermal expansion coefficient of Si, so a large tensile strain is introduced in the GaAs during cooldown. This causes cracking in layers greater than about 4 μm thickness.

InxGa1−xAs is an important material for the channel regions of high-electron-mobility transistors (HEMTs) and also detectors for fiber optic communication systems operating in the range 1.3–1.55 μm. These materials can be grown by MOVPE using TEIn + TMGa + AsH3 or TMIn + TMGa + AsH3.30,31 The ethyl source participates in a parasitic reaction with arsine unless the growth pressure is reduced to ~0.1 atm. This problem is eliminated with TMIn so that high-quality material is obtained with atmospheric growth. Usually, InxGa1−xAs is grown with the methyl sources at 650°C with a growth rate of ~3 μm h−1. Other alloys involving Al can be grown by the addition of TMAl.

InxGa1−xAs is usually grown heteroepitaxially on InP or GaAs substrates. This material can be lattice matched to InP with x = 0.53. However, small processing variations in the composition result in the introduction of a large density of threading dislocations. InxGa1−xAs grown on GaAs is only lattice matched with x = 0. Linearly graded buffer layers (such as InxGa1−xP)32 are often employed to transition from the lattice constant of GaAs to that of the InxGa1−xAs device layer. Here, the threading dislocation is found to be proportional to the grading coefficient, and in practical layers, dislocation densities as low as ~105 cm−2 may be obtained.

AlxInyGa1−x−yP is an important material for high-brightness visible light-emitting diodes (LEDs) such as those used in street signs, traffic lights, and automotive applications, and for solar cells. This material can be lattice matched to GaAs and is usually grown heteroepitaxially on this substrate. This material is grown by MOVPE in the range 600°C–650°C. With methyl sources in the combination TMAl + TMIn + TMGa + PH3, growth pressures up to 1 atm can be used without parasitic reactions.

If the AlxInyGa1−x−yP material is constrained to lattice match the GaAs substrate, then the indium content must be fixed at approximately y = 0.5. The material compositions that match the lattice constant of GaAs may therefore be written as (AlxGa1−x)0.5In0.5P. The energy gap of this material lattice matched to GaAs is given by

3.27 $E g = 1.91 + 0.61 x .$

Even though the active layers of an AlxInyGa1−x−yP LED may be lattice matched to the GaAs substrate, commercial high-brightness devices make use of highly mismatched heteroepitaxial GaP window layers, which spread the current of the top contact and greatly improve the device efficiency. A GaP substrate could serve to further reduce substrate absorption, but this approach is not used due to the high threading dislocation density it would produce in the active layers of the LED.

In0.xGa1−xP may also be used for visible LEDs in the orange and red portion of the spectrum. These devices are usually fabricated by heteroepitaxy on GaAs (001) substrates. However, In0.5Ga0.5P LEDs have also been demonstrated on Si (001) substrates.33 These devices were grown using GaAs buffer layers and exhibited stable output at 660 nm despite the very high threading dislocation density (~107 cm−2).

The III-antimonides are of interest for applications as barrier layers in HEMTs,34 focal-plane detector arrays in the 3–5 μm atmospheric window, and the fabrication of thermophotovoltaic devices.35 These materials include InSb, AlSb, GaSb, and their alloys, and may be grown by MOVPE36 or MBE.37 InSb and GaSb substrates have relatively high threading dislocation densities, so GaAs38,39 or InP substrates are usually used. Typically, a growth rate of ~2.5 μm h−1 is obtained at 600°C using the methyl sources TMGa, TMAl, TMIn, and TMSb.

For InSb grown on GaAs (001), the extremely large lattice mismatch strain (|f| ~ 15%) gives rise to very high misfit and threading dislocation densities. This results in degradation of the carrier mobility near the heterointerfaces, which is associated with the misfit dislocations or a high-density tangle of threading dislocations, so that improved mobility is obtained away from the interface.

#### 3.7  III-Nitrides

GaN, InN, AlN, and their alloys exist in the wurtzite structure and are grown almost exclusively on sapphire (0001) or hexagonal 6H-SiC (0001) substrates. However, growth on semipolar and nonpolar orientations of sapphire, as well as Si(111) substrates,40,41 has been investigated as well.

The III-nitrides grow in a 3-D island mode on sapphire (0001) or 6H-SiC (0001) substrates. Therefore, low-temperature AlN nucleation layers42,43,44 are commonly used to achieve smooth layers free from large columnar islands. Typically, the AlN nucleation layer is grown at a temperature of 450°C–550°C, whereas single-crystal AlN is grown by MOVPE at ~1000°C. The low-temperature AlN grows as an amorphous layer but crystallizes during a subsequent heat treatment. The success of the recrystallization process depends on a thin nucleation layer, so typically this thickness is 50 nm or less. Low-temperature GaN nucleation layers,45,46 have also been used, with similar improvements in the overgrown GaN material. A discussion of low-temperature nucleation layers is given in Chapter 4.

For the heteroepitaxy of GaN on sapphire, nitridation of the sapphire surface prior to growth is a critical step for the attainment of good crystal quality.47 This step serves to replace O atoms by N to form a thin AlN layer. The change in the nucleation surface improves the final threading dislocation density in the overgrown material by a factor of 1/50.

The growth of III-nitrides on highly mismatched substrates gives rise to very high threading dislocation densities in the material. For GaN, the lattice mismatch is ~16% with sapphire (0001), which has the rhombohedral crystal structure with a = 4.7592 Å and c = 12.9916 Å.48 Therefore epitaxial lateral growth (ELO) has been applied to obtain material with low threading dislocation densities for LEDs and laser diodes. ELO, and the related technique of pendeo-epitaxy, are described in detail in Chapter 8. Both of these approaches depend on the large lateral-to-vertical growth rate ratio obtained using MOVPE with the [0001] growth direction.

The III-nitrides must be grown at relatively high temperatures by either MBE or MOVPE (1000°C–1100°C for MOVPE of GaN), so considerable thermal strain is introduced during temperature changes. Sapphire has a larger coefficient of thermal expansion than GaN at room temperature. However, this situation reverses at higher temperatures, so that a tensile strain is introduced in the GaN during the cooldown process. Thermal strain and cracking are described in Chapter 5.

#### 3.7.1  Vapor Phase Epitaxial Growth of III-Nitrides

In early work, Maruska and Tietjen85 demonstrated the VPE of GaN in the Ga + HCl + NH3 system. This approach has been replaced by MOVPE, using TMGa + NH3, TMAl + NH3, or TMIn + NH3. Ternary or quaternary layers may be grown by using the appropriate combinations of the metalorganic sources. Typically, a high V/III ratio is used, so the alkyl flows determine the growth rate and composition of the epitaxial layer. A relatively high substrate temperature must be used for the MOVPE growth of any of these III-Nitrides, due to the thermal stability of ammonia. Two problems in the MOVPE growth of III-nitrides involve the prereactions between NH3 and the metalorganic sources and the thermal convection of NH3, which can disturb the laminar flow conditions and result in recirculation. Several novel reactor designs have been developed to cope with these problems. Akasaki et al.49 demonstrated the growth of AlxGa1−xN on sapphire (0001) using a “dual-flow” reactor design, shown in Figure 3.6. In this cold-wall, RF-heated reactor, the NH3 and metalorganics are mixed a short distance before the reactor, and are directed at the slanted susceptor with high velocity (110 cm s−1) to minimize parasitic reactions.

Nakamura et al.50 designed a vertical, resistance-heated “two-flow” reactor for the growth of GaN, as shown in Figures 3.7 and 3.8. In this design, the main flow, comprising 5 slm NH3, 54 μmol min−1 TMGa, and 1 slm H2, was directed through a quartz nozzle in a direction parallel to the substrate surface. A subflow comprising 10 slm H2 and 10 slm N2 was directed perpendicular to the substrate surface, to bring the reactant gases in contact with the substrate. In their work, it was found that the perpendicular subflow was necessary in order to obtain a continuous film of GaN on a sapphire (0001) substrate at a growth temperature of 1000°C. Without it, only isolated islands of GaN were produced.

In the counterflow reactor of Lee et al.,51 illustrated in Figure 3.9, the NH3 is introduced through a counterflow while the organometallics are introduced through a front flow, thereby minimizing parasitic reactions and preheating the NH3. Mirror-smooth layers of GaN were grown on sapphire (0001) at 1070°C using a total flow of 6 slm.

Figure 3.6   Dual-flow MOVPE reactor for the growth of AlxGa1–xN. (Reprinted from I. Akasaki et al., J. Cryst. Growth, 98, 209 [1989]. With permission. Copyright 1989, Elsevier.)

Figure 3.7   Two-flow MOVPE reactor for GaN growth developed by Nakamura et al.50 (Reprinted from S. Nakamura et al., Appl. Phys. Lett., 58, 2021 [1991]. With permission. Copyright 1991, American Institute of Physics.)

In many of the early studies of MOVPE growth of III-nitrides, high temperatures (>1000°C), high gas velocities (>1 m s−1), and high ammonia flow rates (>5 slm) were required to obtain high-quality films. However, these conditions were deemed severe compared with those used in the growth of other III-V materials, leading to the development of other novel reactor designs using shrouding flows or NH3 counterflow. In the two-flow cold-wall reactor of Nishida et al.52 shown in Figure 3.10, an upper flow of H2 was used to suppress convective effects associated with the NH3. Mirror-smooth layers were obtained with reduced temperature (950°C) and reduced flows (1.5 slm of NH3, TMGa, and H2 as the carrier gas for the TMGa bubbler). An upper flow of 1 slm was used in their study.

Figure 3.8   Details of the two-flow MOVPE reactor for GaN growth developed by Nakamura et al.50 (Reprinted from S. Nakamura et al., Appl. Phys. Lett., 58, 2021 [1991]. With permission. Copyright 1991, American Institute of Physics.)

Figure 3.9   Counterflow MOVPE reactor for the growth of III-nitrides. (Reprinted from C.-R. Lee et al., J. Cryst. Growth, 182, 11 [1997]. With permission. Copyright 1997, Elsevier.)

Figure 3.10   Two-flow GaN MOVPE reactor developed by Nishida et al.52 (Reprinted from K. Nishida et al., J. Cryst. Growth, 170, 312 [1997]. With permission. Copyright 1997, Elsevier.)

#### 3.7.2  Molecular Beam Epitaxy of III-Nitrides

MBE can also be used to grow the III-nitrides,53 and has the advantage of allowing lower growth temperatures. Either RF plasma cells or compact electron cyclotron resonance (ECR) microwave plasma sources of nitrogen are employed.54,55,56 The plasma source produces free radicals of nitrogen; without it, both NH3 and N2 are too stable to contribute to growth at typical temperatures employed for MBE growth. One such MBE system reported by Lin et al. is shown in Figure 3.11. In addition to the MBE chamber, this system has a UHV CVD system attached through a load lock, in which substrates may be cleaned in a hydrogen plasma prior to MBE growth. In such an MBE system the growth rate is limited by the supply of active nitrogen from the plasma source, so that operation with a high Ga flux results in the formation of Ga droplets on the surface.

Figure 3.11   GaN MBE system utilizing an ECR microwave plasma source of nitrogen. (Reprinted from M. E. Lin et al., J. Appl. Phys., 74, 5038 [1993]. With permission. Copyright 1993, American Institute of Physics.)

MBE-grown III-nitride materials may be doped n-type or p-type during growth through the use of appropriate dopants. Si, Ge, and Se have all been used as donor impurities, but Si is most commonly used, and electron concentrations as high as 1020 cm−3 have been achieved.57 p-Type doping is invariably achieved using the impurity Mg. As-grown films are p-type, without the need for thermal annealing. Nitrogen-rich conditions are favorable for Mg incorporation, and concentrations exceeding 1020 cm−3 have been demonstrated.58

The dilute nitrides GaInNAs and GaInNAsSb have potential applications in optoelectronics for high-bit-rate communications systems, such as 1.2–1.6 μm lasers and optical amplifiers. These materials have been grown heteroepitaxially on GaAs substrates by MBE.59 Due to the low solubility of N in these materials, they are susceptible to phase separation; therefore, metastable alloys must be grown at low temperatures (~425°C for MBE). The growth mode is SK (2-D growth of a wetting layer following by 3D island growth), but the 2-D to 3-D transition can be suppressed by the introduction of an Sb surfactant.

#### 3.8  II-VI Semiconductors

The II-VI semiconductors include all combinations of Zn, Cd, Hg, and S, Se, and Te, and may be grown by MOVPE or MBE. These materials are usually grown heteroepitaxially on GaAs, InP, or Si. ZnSe substrates are available, but with relatively small area and high defect densities. However, device structures can be designed to be lattice matched to either GaAs or InP substrates. Hg1−xCdxTe is of great interest for infrared devices, and may also be grown by MOVPE and MBE. Substrates such as CdTe, InSb, and even CdZnTe have been utilized, but GaAs or Si are used more commonly due to their larger area and better quality. ZnO is a direct bandgap semiconductor with Eg = 3.37eV and a large excition binding energy of 60 meV, so that room temperature excitonic UV emission should be possible.60 This has led to great interest in the epitaxial growth of this material, which has been achieved by plasma-assisted MBE,61 magnetron sputtering,62 and pulsed laser deposition (PLD)60 on a variety of substrates, including sapphire (0001), Si (001),60 and CaF2 (111).61

#### 3.8.1  ZnSe and Its Alloys

ZnSe and its alloys must be grown at low temperatures in order to minimize the influence of native defects and their complexes. MOVPE growth at relatively low temperatures is possible using the hydride sources H2Se, H2S, and H2Te. However, these hydrides give rise to gas phase prereactions with the metalorganics, degrading the layer quality and fouling the reactor. For this reason, the metalorganic sources such as DMSe, DES, and DETe are commonly used. These sources require an increase in the growth temperature unless photoirradiation from an ultraviolet lamp is employed. This photoassisted growth technique is complicated by the change of surface composition during heteroepitaxial growth. The mechanism of photoassisted growth appears to involve photogenerated carriers near the surface of the ZnSe, which promote the breaking of the alkyl source bonds. However, photoassisted carriers in a GaAs substrate do not participate in this process. Therefore, heteroepitaxial growth on GaAs substrates requires the growth of a high-temperature ZnSe buffer prior to the start of photoassisted epitaxy.

An important problem in the heteroepitaxy of wide-bandgap II-VI materials on GaAs substrates is the creation of stacking faults at the interface. It has been found that a single such defect can give rise to the rapid degradation and failure of an LED or laser diode. The nucleation of the stacking faults is related to the initial condition of the surface. In MBE growth, the formation of stacking faults can be suppressed by Zn stabilization (starting the Zn beam first).

#### 3.8.2  HgCdTe

The ternary material Hg1−xCdxTe, which is of great interest for infrared detectors in the 8–16 μm range, is usually grown on GaAs substrates by MOVPE63,64 or MBE.65,66 Occasionally, Si or sapphire67 substrates have also been utilized. Hg1−xCdxTe exhibits a very large lattice mismatch (~14%) with GaAs substrates over the entire compositional range. As a consequence, it has been reported that the epitaxial relationship can be either CdTe[001]||GaAs[001] or CdTe[111]||GaAs[001].68,69

Hg1−xCdxTe is typically grown on GaAs substrates by MOVPE with the sources Hg + DMCd + DETe.70 Typical growth temperatures range from 350°C to 420°C. It is found that these layers contain ~500 cm−2 hillocks, which may be associated with stacking faults at the interface. However, this problem can be prevented by the inclusion of a CdTe buffer layer. Hg1−xCdxTe can also be grown at a lower temperature (~175°C) using the source combination Hg + DMCd + DTBTe.71

In order to reduce the dislocation densities in Hg1−xCdxTe device layers grown on GaAs substrates, Cd1−xZnxTe buffer layers have been used, both the graded and constant composition types.72 (Uniform and graded buffer layer approaches are discussed in Chapter 6.) Also, wide-bandgap barrier layers are used to reduce the interface recombination velocity.

#### 3.8.3  ZnO

ZnO may be grown by plasma-enhanced MBE,7377 MOVPE,7881 magnetron sputtering,62 or PLD60,82,83 on a variety of substrates. Invariably, a low-temperature buffer layer is used to achieve smooth surface morphology and control the film orientation.61,62,84

MBE growth is conducted at a temperature of 550°C–600°C on a suitable buffer. Typically, the buffer layer is low-temperature ZnO deposited at a temperature of 300°C–350°C,60,61 but occasionally other buffers have been used. Park et al.74 demonstrated the epitaxy of smooth, high-quality ZnO on sapphire substrates using a low-temperature ZnO buffer on top of a Cr2O3 buffer created by the oxidation of an MBE-deposited chromium layer. MBE growth utilizes an RF plasma source74 or an ECR microwave plasma source76 for oxygen and a solid source for zinc.84 Nominally, undoped films are invariably n-type. Hirano et al.73 demonstrated that the use of flux-modulated MBE, in which the growth conditions are alternated between oxygen-rich and zinc-rich conditions, resulted in a lower residual electron concentration (9.7 × 1016 cm−2) compared with standard oxygen-rich conditions (2.4 × 1017 cm−2). p-Type ZnO has been reported using the dopants nitrogen,77 phosphorus, and arsenic.

ZnO growth by MOVPE may be carried out in a broad temperature range, but typically 300°C–400°C is used. The commonly used sources are diethylzinc and tertiary-butanol (t-BuOH), with Ar as the carrier gas,78 but these are not available in high purity. Dietylzinc and oxygen have also been used, but a parasitic gas phase reaction is problematic with this combination of sources. Ogata et al.80,81 have demonstrated MOVPE growth of ZnO using diethylzinc and nitrous oxide (N2O) at relatively high growth temperatures (600°C–700°C), but without gas phase reactions.

PLD may be conducted using a Nd:yttrium-aluminum-garnet (YAG) laser (355 nm, 5 Hz, 3.8 J cm−2) to ablate a ZnO target, typically in a partial oxygen pressure of 1 torr,60 in a system of the type shown in Figure 3.12. Under usual conditions it is necessary to use the oxygen ambient in order to obtain stoichiometric films. However, Yamaguchi et al.82 have demonstrated the growth of stoichiometric films without an oxygen ambient by applying a bias voltage to a molybdenum grid placed between the target and substrate, as shown in Figure 3.12. The bias voltage is applied on the molybdenum grid with respect to the target. Low deposition temperatures (~200°C) may be used for PLD growth of ZnO, but it is often necessary to anneal the films at a higher temperature (~700°C) to obtain good crystallinity.

Figure 3.12   PLD system for the growth of ZnO films.

#### 3.9  Conclusion

Molecular beam epitaxial and VPE techniques, especially MOVPE, have enabled the realization of a wide range of heteroepitaxial devices and structures. Both afford tremendous flexibility and the ability to deposit thin layers and complex multilayered structures with precise control and excellent uniformity. This enables the practical realization of advanced device structures such as heterojunction, quantum well, and quantum dot devices.

Many heteroepitaxial material combinations with diverse characteristics have been investigated. However, certain aspects of heteroepitaxy appear repeatedly among them, and will be covered in detail in the following chapters. These aspects include nucleation, growth modes, lattice mismatch and strain relaxation, crystal defects, thermal strain, and cracking.

#### Problems

1. Calculate the Reynold’s number for an MOVPE reactor operating at 0.1 atm and 650°C with 10 slm of H2 carrier gas, if the reaction chamber is a round tube with a diameter of 10 cm. Determine if laminar or turbulent flow conditions prevail in the reactor.
2. Consider the MOVPE growth of GaAs using TMGa + AsH3 at 650°C in a round reactor tube with a diameter of 10 cm. Ten stand liters per minute of H2 carrier gas is used. The mole fraction of TMGa in the reactor is 10−4 and the total pressure is 1 atm. (a) Estimate the boundary layer thickness at a distance 1 cm down the susceptor. (b) Estimate the growth rate. Assume the diffusivity of TMGa in hydrogen is 0.31 cm2 s−1.
3. Repeat Problem 2 for the case of P = 0.1 atm. Assume that the diffusivity scales as 1/P.
4. Suppose 20 sccm of H2 is bubbled through a DMZn bubbler maintained at −10°C. (a) Assuming the bubbles become saturated with the vapor of DMZn, estimate the flow of DMZn to the reactor. (b) Calculate the mole fraction of DMZn in the reactor, if the total flow of H2 carrier gas is 5 slm.
5. Consider MBE growth of GaAs. The Ga effusion cell has a diameter of 2 cm, is located 25 cm from the substrate, and is held at a temperature of 1000°C. (a) Calculate the impingement rate for Ga at the substrate, in atoms per square centimeter per second. (b) Estimate the growth rate. Assume the vapor pressure of Ga at 1000°C is 4 × 10−3 torr.

#### References

S. Whitaker, Introduction to Fluid Mechanics (Englewood Cliffs, NJ: Prentice-Hall, 1968).
H. M. Manasevit, Single-crystal gallium arsenide on insulating substrates, Appl. Phys. Lett., 12, 156 (1968).
H. M. Manasevit and W. I. Simpson, The use of metal-organics in the preparation of semiconductor materials. I. Epitaxial gallium-V compounds, J. Electrochem. Soc., 116, 1725 (1969).
H. M. Manasevit, The use of metalorganics in the preparation of semiconductor materials: Growth on insulating substrates, J. Cryst. Growth, 13–14, 306 (1972).
M. A. Herman and H. Sitter, Molecular Beam Epitaxy: Fundamentals and Current Status (New York: Springer-Verlag, 1988).
J. H. Jeans, An Introduction to the Kinetic Theory of Gases (Cambridge: University Press, 1967).
B. B. Dayton, Gas flow patterns at entrance and exit of cylindrical tubes, in 1956 National Symposium on Vacuum Technology Transactions, eds. E. S. Perry and T. H. Devant (Oxford: Pergamon Press, 1957), p. 5.
T. U. M. S. Murthy, N. Miyamoto, M. Shimbo, and J. Nishizawa, Gas-phase nucleation during the thermal decomposition of silane in hydrogen, J. Cryst. Growth, 33, 1 (1976).
R. P. Ruth, J. C. Marinace, and W. C. Dunlap Jr., Vapor-deposited single-crystal germanium, J. Appl. Phys., 31, 995 (1960).
E. F. Cave and B. R. Czorny, Epitaxial deposition of silicon and germanium layers by chloride reduction, RCA Rev., 523 (1963).
K. J. Miller and M. J. Grieco, Epitaxial P-type germanium films by the hydrogen reduction of GeBr4, SiBr4, and BBr3, J. Electrochem. Soc., 110, 1252 (1963).
J. M. Hartmann, Y. Bogumilowicz, F. Andrieu, P. Holliger, G. Rolland, and T. Billon, Reduced pressure-chemical vapor deposition of high Ge content Si1−xGex and high C content Si1−yCy layers for advanced metal oxide semiconductor transistors, J. Cryst. Growth, 277, 114 (2005).
S. S. Iyer, K. Eberl, A. R. Powell, and B. A. Ek, SiCGe ternary Alloys-Extending Si-based heterostructures, Microelectron. Eng., 19, 351 (1992).
C. Li, S. John, E. Quinones, and S. Banerjee, Cold-wall ultrahigh vacuum chemical vapor deposition of doped and undoped Si and Si1−xGex epitaxial films using SiH2 and Si2H6, J. Vac. Sci. Technol. A, 14, 170 (1996).
Yu. B. Bolkhovityanov, A. S. Deryabin, A. K. Gutakovskii, M. A. Revenko, and L. V. Sokolov, Heterostructures GexSi1−x/Si (001) (x=0.18–0.62) grown by molecular beam epitaxy at a low (350°C) temperature: Specific features of plastic relaxation, Thin Solid Films, 466, 69 (2004).
J. C. Bean, L. C. Feldman, A. T. Fiory, S. Nakahara, and I. K. Robinson, GexSi1−x/Si strained-layer superlattice grown by molecular beam epitaxy, J. Vac. Sci. Technol. A, 2, 436 (1984).
J. A. Powell, L. G. Matus, and M. A. Kuczmarski, Growth and characterization of cubic SiC single-crystal films on Si, J. Electrochem. Soc., 134, 1558 (1987).
S. Nishino and J. Saraie, Heteroepitaxial growth of cubic SiC on a Si substrate using the Si2H6-C2H2-H2 system, in Amorphous and Crystalline Silicon Carbide, eds. G. L. Harris and C. Y.-W. Yang (Berlin: Springer, 1989), p. 45.
W. Muench, W. Kurzinger, and I. Pfaffender, Epitaxial deposition of silicon carbide from silicon tetrachloride and hexane, Thin Solid Films, 31, 39 (1976).
P. Liaw and R. F. Davis, Epitaxial growth and characterization of β-SiC thin films, J. Electrochem. Soc., 132, 642 (1985).
K. Ikoma, M. Yamanaka, H. Yamaguchi, and Y. Shichi, Heteroepitaxial growth of β-SiC on Si(111) by CVD using a CH3Cl-SiH4-H2 gas system, J. Electrochem. Soc., 138, 3031 (1991).
P. Rai-Choudhury and N. P. Formigoni, β-Silicon carbide film, J. Electrochem. Soc., 116, 1440 (1969).
S. Nishino and J. Saraie, Heteroepitaxial growth of cubic SiC on a Si substrate using methyltrichlorosilane, in Springer Proceedings in Physics, eds. M. M. Rahman, C. Y. Yang, and G. L. Harris, Vol. 43 (Berlin: Springer, 1989), p. 8.
T. Chassagne, G. Ferro, D. Chaussnde, F. Cauwet, Y. Monteil, and J. Bouix, A comprehensive study of SiC growth processes in a VPE reactor, Thin Solid Films, 402, 83 (2002).
R. S. Kern, S. Tanaka, L. B. Rowland, and R. F. Davis, Reaction kinetics of silicon carbide deposition by gas-source molecular-beam epitaxy, J. Cryst. Growth, 183, 581 (1998).
R. F. Davis, S. Tanaka, L. B. Rowland, R. S. Kern, Z. Sitar, S. K. Ailey, and C. Wang, Growth of SiC and III-V nitride thin films via gas-source molecular beam epitaxy and their characterization, J. Cryst. Growth, 164, 132 (1996).
R. S. Kern, S. Tanaka, L. B. Rowland, and R. F. Davis, Reaction kinetics of silicon carbide deposition by gas-source molecular-beam epitaxy, J. Cryst. Growth, 183, 581 (1998).
N. Kuroda, K. Shibahara, W. S. Yoo, S. Nishino, and H. Matsunami, Extended Abstracts of the 34th Spring Meeting of Japan Society Applied Physics and Related Societies, Tokyo (1987), p. 135 (in Japanese).
N. Kuroda, K. Shibahara, W. S. Yoo, S. Nishino, and H. Matsunami, Extended Abstracts of the 19th Conference on Solid State Devices and Materials, Tokyo (1987), p. 227.
C. P. Kuo, R. M. Cohen, K. L. Fry, and G. B. Stringfellow, OMVPE growth of GaInAs, J. Cryst. Growth, 64, 461 (1983).
C. P. Kuo, J. S. Yuan, R. M. Cohen, J. Dunn, and G. B. Stringfellow, Organometallic vapor phase epitaxial growth of high purity GaInAs using trimethylindium, Appl. Phys. Lett., 44, 550 (1984).
K. Yuan, K. Radhakrishnan, H. Q. Zheng, and G. I. Ng, Metamorphic In0.5Al0.5As/In0.53Ga0.47As high electron mobility transistors on GaAs with InxGa1−xP graded buffer, J. Vac. Sci. Technol. B, 19, 2119 (2001).
S. Kondo, S.-I. Matsumoto, and H. Nagai, 660 nm In0.5Ga0.5P light-emitting diodes on Si substrates, Appl. Phys. Lett., 53, 279 (1988).
B. R. Bennett, R. Magno, J. B. Boos, W. Kruppa, and M. G. Ancona, Antimonide-based compound semiconductors for electronic devices: A review, Solid State Electron., 49, 1875 (2005).
L. M. Fraas, R. Ballantyne, J. Samaras, and M. Seal, Electric power production using new GaSb photovoltaic cells with extended infrared response, AIP Conf. Proc., 321, 44 (1994).
Y. Paltiel, A. Sher, A. Raizman, S. Shusterman, M. Katz, A. Zemel, Z. Calahorra, and M. Yassen, Metalorganic vapor phase epitaxy InSb p+nn+ photodiodes with low dark current, Appl. Phys. Lett., 84, 5419 (2004).
T. D. McLean, T. M. Kerr, D. I. Westwood, J. D. Grange, and I. J. Murgatroyd, Inst. Phys. Conf. Ser., 74, 145 (1984).
H. Ehsani, I. Bhat, C. Hitchcock, J. Borrego, and R. Gutmann, Characteristics of GaSb and GaInSb layers grown by metalorganic vapor phase epitaxy, AIP Conf. Proc., 358, 423 (1996).
H. Ehsani, I. Bhat, R. Gutmann, and G. Charache, p-Type GaSb and Ga0.8In0.2Sb layers grown by metalorganic vapor phase epitaxy using silane as the dopant source, Appl. Phys. Lett., 69, 3863 (1996).
Y. Koide, H. Itoh, N. Sawaki, I. Akasaki, and M. Hashimoto, Epitaxial growth and properties of AlxGa1−xN by MOVPE, J. Electrochem. Soc., 133, 1956 (1986).
Y. Dikme, A. Szymakowski, H. Kalisch, E. V. Lutsenko, V. N. Zubialevich, G. P. Yablonskii, H. M. Chern, C. Schaefer, R. Jansen, and M. Heuken, Investigation of GaN on Si(111) for optoelectronic applications, Proc. SPIE, 4996, 57 (2003).
H. Amano, N. Sawaki, I. Akasaki, and Y. Toyoda, Metalorganic vapor phase epitaxial growth of a high quality GaN film using an AlN buffer layer, Appl. Phys. Lett., 48, 353 (1986).
H. Amano, I. Akasaki, K. Hiramatsu, and Sawaki, Effects of the buffer layer in metalorganic vapour phase epitaxy of GaN on sapphire substrate, Thin Solid Films, 163, 415 (1988).
Y. Koide, N. Itoh, X. Itoh, N. Sawaki, and I. Akasaki, Effect of AlN buffer layer on AlGaN/α− Al2O3 heteroepitaxial growth by metalorganic vapor phase epitaxy, Jpn. J. Appl. Phys., 27, 1156 (1988).
S. Nakamura, GaN growth using GaN buffer layer, Jpn. J. Appl. Phys., 30, L1705 (1991).
N. Kuznia, M. A. Khan, D. T. Olsen, R. Kaplan, and J. Freitas, Influence of buffer layers on the deposition of high quality single crystal GaN over sapphire substrates, J. Appl. Phys., 73, 4700 (1993).
S. Keller, B. P. Keller, Y.-F. Wu, B. Heying, D. Kapolnek, J. S. Speck, U. K. Mishra, and S. P. DenBaars, Influence of sapphire nitridation on properties of gallium nitride grown by metalorganic chemical vapor deposition, Appl. Phys. Lett., 68, 1525 (1996).
Y. V. Shvyd’ko, M. Lucht, E. Gerdau, M. Lerche, E. E. Alp, W. Sturhahn, J. Sutter, and T. S. Toellner, Measuring wavelengths and lattice constants with the Mössbauer wavelength standard, J. Synchrotron Rad., 9, 17 (2002).
I. Akasaki, H. Amano, Y. Koide, K. Hiramatsu, and N. Sawaki, Effects of AlN buer layer on crystallographic structure and on electrical and optical properties of GaN and Ga1−xAlxN (0 <x ≤ 0.4) films grown on sapphire substrate by MOVPE, J. Cryst. Growth, 209 (1989).
S. Nakamura, Y. Harada, and M. Seno, Novel metalorganic chemical vapor deposition system for GaN growth, Appl. Phys. Lett., 58, 2021 (1991).
C.-R. Lee, S.-J. Son, I.-H. Lee, J.-Y. Leem, and S. K. Noh, High-quality GaN epilayers grown by newly designed horizontal counter-flow MOCVD reactor, J. Cryst. Growth, 182, 11 (1997).
K. Nishida, S. Haneda, K. Hara, H. Munekata, and H. Kukimoto, MOVPE of GaN using a specially designed two-flow horizontal reactor, J. Cryst. Growth, 170, 312 (1997).
X. Wang and A. Yoshikawa, Molecular beam epitaxy growth of GaN, AlN, and InN, Prog. Cryst. Growth Char. Mater., 48/49, 42 (2004).
J. T. Torvik, M. Leksono, J. I. Pankove, B. V. Zeghbroeck, H. M. Ng, and T. D. Moustakas, Electrical characterization of GaN/SiC n-p heterojunction diodes, Appl. Phys. Lett., 72, 1371 (1998).
N. Gogneau, E. Sarigiannidou, E. Monroy, S. Monnoye, H. Mank, and B. Daudin, Surfactant effect of gallium during the growth of GaN on AlN (0001) by plasma-assisted molecular beam epitaxy, Appl. Phys. Lett., 85, 1421 (2004).
E. Monroy, N. Gogneau, F. Enjalbert, F. Fossard, D. Jalabert, E. Bellet-Amalric, L. S. Dang, and B. Daudin, Molecular-beam epitaxial growth and characterization of quaternary III-nitride compounds, J. Appl. Phys., 94, 3121 (2003).
H. M. Ng, D. Doppalapudi, D. Korakakis, R. Singh, and T. D. Moustakas, MBE growth and doping of III-V nitrides, J. Cryst. Growth, 189, 349 (1998).
G. Namkoong, W. A. Doolittle, and A. S. Brown, Incorporation of Mg in GaN grown by plasma-assisted molecular beam epitaxy, Appl. Phys. Lett., 77, 4386 (2000).
J. S. Harris Jr., The opportunities, successes and challenges for GaInNAsSb, J. Cryst. Growth, 278, 3 (2005).
W. Zheng, Y. Liao, L. Li, Q. Yu, G. Wang, Y. Li, and Z. Fu, Structure and properties of ZnO films grown on Si substrates with low temperature buffer layers, Appl. Surf. Sci., 253, 2765 (2006).
H. J. Ko, Y. F. Chen, Z. Zhu, T. Hanada, and T. Yao, Effects of a low-temperature buffer layer on structural properties of ZnO epilayers grown on (111) CaF2 by two-step MBE, J. Cryst. Growth, 208, 389 (2000).
H. F. Liu and S. J. Chua, Efects of low-temperature-bufer, RF-power, and annealing on structural and optical properties of ZnO/Al2O3(0001) thin films grown by RF-magnetron sputtering, J. Appl. Phys., 106, 023511 (2009).
W. E. Hoke, P. J. Lemonias, and R. Taczewski, Metalorganic growth of high-purity HgCdTe films, Appl. Phys. Lett., 45, 1092 (1984).
I. B. Bhat, N. R. Taskar, and S. K. Ghandhi, The organometallic heteroepitaxy of CdTe and HgCdTe on GaAs substrates, J. Vac. Sci. Technol. A, 4, 2230 (1986).
J. P. Faurie, S. Sivanathan, M. Boukerche, and J. Reno, Molecular beam epitaxial growth of high quality HgTe and Hg1−xCdxTe onto GaAs (001) substrates, Appl. Phys. Lett., 45, 1307 (1984).
K. Nishitani, R. Okhata, and T. Murotani, Molecular beam epitaxy of CdTe and Hg1−xCdxTe on GaAs (100), J. Electron. Mater., 12, 619 (1983).
T. H. Myers, Y. Lo, R. N. Bicknell, and J. F. Schetzina, Growth of CdTe films on sapphire by molecular beam epitaxy, Appl. Phys. Lett., 42, 247 (1983).
J. T. Cheung and T. J. Magee, Recent progress on LADA growth of HgCdTe and CdTe epitaxial layers, J. Vac. Sci. Technol. A, 1, 1604 (1983).
H. A. Mar, K. T. Chee, and N. Salansky, CdTe films on (001) GaAs:Cr by molecular beam epitaxy, Appl. Phys. Lett., 44, 237 (1984).
S. K. Ghandhi, I. B. Bhat, and N. R. Taskar, Growth and properties of Hg1−xCdxTe on GaAs substrates by organometallic vapor-phase epitaxy, J. Appl. Phys., 59, 2253 (1986).
K. Yasuda, H. Hatano, T. Ferid, M. Minamide, T. Maejima, and K. Kawamoto, Growth characteristics of (100) HgCdTe layers in low-temperature MOVPE with ditertiarybutyltelluride, J. Cryst. Growth, 166, 612 (1996).
V. S. Varavin, S. A. Dvoretsky, V. I. Liberman, N. N. Mikhailov, and Yu. G. Sidorov, Molecular beam epitaxy of high quality Hg1−xCdxTe films with control of the composition distribution, J. Cryst. Growth, 159, 1161 (1996).
K. Hirano, M. Fujita, M. Sasajima, T. Kosaka, and Y. Horikoshi, ZnO epitaxial films grown by flux-modulated RF-MBE, J. Cryst. Growth, 301, 370 (2007).
J. S. Park, S. K. Hong, T. Minegishi, I. H. Im, S. H. Park, T. Hanada, J. H. Chang, M. W. Cho, and T. Yao, The high quality ZnO growth on c-Al2O3 substrate with Cr2O3 buffer layer using plasma-assisted molecular beam epitaxy, Appl. Surf. Sci., 254, 7786 (2008).
J. S. Park, S. K. Hong, I. H. Im, J. S. Ha, H. J. Lee, S. H. Park, J. H. Chang, M. W. Cho, and T. Yao, Growth of high-quality ZnO films on Al2O3 (0001) by plasma-assisted molecular beam epitaxy, J. Cryst. Growth, 311, 2163 (2009).
F. Xiu, Z. Yang, D. Zhao, J. Liu, K. A. Alim, A. A. Balandin, M. E. Itkis, and R. C. Haddon, ZnO growth on Si with low-temperature ZnO buffer layers by ECR-assisted MBE, J. Cryst. Growth, 286, 61 (2006).
D. C. Look, D. C. Reynolds, C. W. Litton, R. L. Jones, D. B. Eason, and G. Cantwell, Characterization of homoepitaxial p-type ZnO grown by molecular beam epitaxy, Appl. Phys. Lett., 81, 1830 (2002).
J. R. Botha, K. T. Roro, C. Weichsel, A. W. R. Leitch, and J. Weber, Arsenic-related recombination in MOVPE-grown ZnO/GaAs films, Superlat. Microstruct., 42, 26 (2007).
P. Kuznetsov, V. Lusanov, G. Yakushcheva, V. Jitov, L. Zakharov, I. Kotelyanskii, and V. Kozlovsky, MOVPE growth and study of ZnO, ZnMgO epilayers and ZnO/ZnMgO MQW structures, Phys. Status Solidi C, 7, 1568 (2010).
K. Ogata, K. Maejima, Sz. Fujita, and Sg. Fujita, ZnO growth toward optical devices by MOVPE using N2O, J. Electron. Mater., 30, 659 (2001).
K. Ogata, S.-W. Kim, Sz. Fujita, and Sg. Fujita, ZnO growth on Si substrates by metalorganic vapor phase epitaxy, J. Cryst. Growth, 240, 112 (2002).
H. Yamaguchi, T. Shitara, T. Komiyama, Y. Chonan, and T. Aoyama, ZnO films deposited by optimized PLD technique with bias voltages, Phys. Status Solidi C, 7, 326 (2010).
J.-H. Leem, D.-H. Lee, and S. Y. Lee, Properties of N-doped ZnO grown by DBD-PLD, Thin Solid Films, 518, 1238 (2009).
Y.-S. Jung, O. Kononenko, J.-S. Kim, and W.-K. Choi, Two-dimensional growth of ZnO epitaxial films on c-Al2O3 (0001) substrates with optimized growth temperature and low-temperature buffer layer by plasma-assisted molecular beam epitaxy, J. Cryst. Growth, 274, 418 (2005).
H. P. Maruska and J. J. Tietjen, The preparation and properties of vapor-deposited single-crystalline GaN, Appl. Phys. Lett., 15, 327 (1969).

## Use of cookies on this website

We are using cookies to provide statistics that help us give you the best experience of our site. You can find out more in our Privacy Policy. By continuing to use the site you are agreeing to our use of cookies.