Growth of Bulk GaN Crystals

Growth of Bulk GaN Crystals

3.06 Growth of Bulk GaN Crystals B Feigelson, US Naval Research Laboratory, Washington, DC, USA T Paskova, Kyma Technologies, Inc., Raleigh, NC, USA ª...

2MB Sizes 0 Downloads 41 Views

3.06 Growth of Bulk GaN Crystals B Feigelson, US Naval Research Laboratory, Washington, DC, USA T Paskova, Kyma Technologies, Inc., Raleigh, NC, USA ª 2011 Elsevier B.V. All rights reserved.

3.06.1 3.06.1.1 3.06.1.2 3.06.1.3 3.06.1.4 3.06.2 3.06.3 3.06.3.1 3.06.3.2 3.06.3.2.1 3.06.3.2.2 3.06.3.2.3 3.06.3.2.4 3.06.3.2.5 3.06.4 3.06.4.1 3.06.4.1.1 3.06.4.1.2 3.06.4.2 3.06.4.2.1 3.06.5 References

Introduction Low Defect Density and Smooth Morphology Thermal Conductivity Electrical Conductivity Variation of Substrate Surface Orientation Challenges of the Bulk GaN Growth Bulk GaN Growth from Liquid Phase Melt Growth Solution Growth Direct N2 dissolution in Ga Chemically enhanced nitrogen dissolution Excited N2 dissolution in Ga Chemically enhanced nitrogen or GaN solubility Technical challenges and conclusion Bulk GaN Growth from Gas Phase Physical Vapor Transport Sublimation Molecular beam epitaxy Chemical Vapor Transport Hydride vapor-phase epitaxy Summary

233 234 235 235 235 235 236 236 237 238 246 255 255 263 264 264 264 265 266 266 276 276

Glossary Convection The heat and mass transfer through a bulk, macroscopic movement of fluids (i.e., liquids, gases, and rheids), as opposed to the microscopic transfer of heat (heat conduction) or mass (diffusion). Flux In crystal growth, flux is another word for solvent. Flux method is a method of crystal growth where the components of the desired substance are dissolved in a solvent (flux). In transport phenomena studies flux is the rate of transfer of fluid, particles, or energy across a given surface (transport phenomena). In metallurgy, flux is any substance introduced in the smelting of ores to promote fluidity and to remove objectionable impurities in the form of slag. In soldering, a flux is used to remove oxide films, promote wetting, and prevent reoxidation of the surfaces during heating. Heteroepitaxy A kind of epitaxy performed with materials that are different from each other. In heteroepitaxy, a crystalline film grows on a

232

crystalline substrate or film of a different material. This technology is often used to grow crystalline films of materials for which single crystals cannot otherwise be obtained and to fabricate integrated crystalline layers of different materials. Homoepitaxy A kind of epitaxy performed with only one material. In homoepitaxy, a crystalline film is grown on a substrate or film of the same material. Isotropic crystal Whether a crystalline solid is anisotropic or isotropic depends on the way the atoms in the crystal are arranged, that is, on the symmetry of the crystal lattice. If the spacing and arrangement of the atoms in a crystal appear the same in each of the three planes (X, Y, and Z), the crystal is isotropic. In this chapter, isotropic crystal refers to a crystal in which all three dimensions are similar. Metastable state Metastability is a general scientific concept which describes states of

Growth of Bulk GaN Crystals

delicate equilibrium. A system is in a metastable state when it is in equilibrium (not changing with time) but is susceptible to fall into lower-energy states with only slight interaction. A metastable state may thus be considered a kind of temporary energy trap or a somewhat stable intermediate stage of a system. Molecular beam A molecular beam is produced by allowing a gas at higher pressure to expand through a small orifice into a container at lower pressure. The result is a beam of particles (atoms, free radicals, molecules, or ions) moving at approximately equal velocities, with few collisions occurring between them. The term beam means that evaporated atoms do not interact with each other or vacuum-chamber gases until they reach the wafer, due to the long mean free path of the atoms. Mosaicity A measure of the long-range disorder of a crystal. Piezoelectric polarization Polarization is a property of the electromagnetic waves that describes the orientation of their oscillation. Piezoelectric polarization field is a result of the lattice-mismatch-induced strain in heteroepitaxial nitride device structures. Retrograde solubility A phenomenon when solubility of a solute decreases with increasing temperature, that is, the solute has a negative temperature coefficient of the solubility. Spontaneous nucleation The nucleation of a phase change of a substance without the benefit of any seeding nuclei within or otherwise in contact with that substance. Examples of such systems are a pure vapor condensing to its pure liquid state, a pure liquid freezing to its pure solid state, and a

3.06.1 Introduction III-nitride devices, such as ultraviolet (UV)–blue– green emitters, high-power transistors, and detectors, have become commercial realities relying completely on heteroepitaxial approaches by employing a variety of foreign substrates, including sapphire, silicon carbide, and silicon. However, their performance is significantly limited by the structural quality of these materials as a result of the well-known

233

pure solution crystallizing to yield pure solute crystals. Spontaneous polarization Polarization is a property of the electromagnetic waves that describes the orientation of their oscillation. Spontaneous polarization field is an inherent physical property of the wurtzite semiconductors, having polar atomic arrangement along the c-axis with ionic chemical bonds. Solubility In general, an ability of a substance to dissolve. In the process of dissolving, the substance which is being dissolved is called a solute and the substance in which the solute is dissolved is called a solvent. A mixture of solute and solvent is called a solution. Solubility is understood as a maximum amount of solute that dissolves in a solvent at so-called equilibrium. In chemistry, equilibrium is a state where reactants and products reach a balance – no more solute can be dissolved in the solvent in the set conditions (temperature and pressure). Such a solution is called a saturated solution. Supercritical ammonia Ammonia at a temperature and pressure above its critical point. The vapor–liquid critical point denotes the conditions above which distinct liquid and gas phases do not exist. Supersaturation In general, the degree of the system deviation from equilibrium and is defined by the difference between the chemical potentials of a substance at alternative states. The term refers to a solution that contains more of the dissolved material than could be dissolved by the solvent under normal circumstances or to a vapor of a compound that has a higher (partial) pressure than the vapor pressure of that compound.

disadvantages of heteroepitaxy, such as lattice mismatch, thermal expansion coefficient mismatch and chemical incompatibility leading to high-dislocation density, mosaic crystal structure, biaxial induced stress, and wafer bowing. In order to reduce these structural defects, special approaches have been developed, such as low-temperature (LT) buffer technologies or epitaxial lateral overgrowth (ELOG) techniques, which complicate and extend the device growth process. Additional drawbacks

234 Growth of Bulk GaN Crystals

are related to the lower electrical and/or thermal conductivity of most foreign substrates, which require more complicated and expensive device designs. Despite the fact that the current heteroepitaxial nitride industry works quite well for most of the light-emitting diode (LED) devices, especially considering the wide availability of cheap substrates such as sapphire, the potential availability of native GaN substrates is expected to have an impact in two major ways for any industry, namely by reducing device production cost and by improving device performance. Even assuming that GaN substrates are not likely to compete with sapphire substrate cost in the near future, the cost may still be significantly reduced by shortening the real epitaxial time and by circumventing the technological LT buffer and/or lateral epitaxial steps required in the current nitride device manufacturing. The impact of the native GaN substrates on the device performance is manifested by the following: (1) low defect density and smooth morphology homoepitaxy; (2) good thermal conductivity and lower operation temperature; and (3) good electrical conductivity and simplified device architecture with improved current spreading. These are discussed in the following. 3.06.1.1 Low Defect Density and Smooth Morphology In contrast to many material systems for which even a low defect density leads to a dramatic deterioration of the device performance, many of the nitride-based devices can successfully operate even in the presence of dislocations with density in the range of 108–109 cm2. This is particularly valid for the blue LED. The defect density, however, may have a significant impact on LEDs operating in the UV and green wavelength regions, which are characterized by lower efficiency. Although it is believed that several mechanisms play a role in reducing their efficiency, it is known that the high-dislocation density in material with higher In content contributes to nonradiative recombination, reducing the carrier locations and the internal quantum efficiency. The high-dislocation density is even more critical for laser diode (LD) performance. The dislocations cause deterioration of the LD performance by providing fast diffusion paths along their lines and thus smearing out the quantum-well (QW) and shorting the p–n junctions, as well as by serving as nonradiative recombination centers, thus leading to heat

generation instead of optical emission (Porowski et al., 2004). The latter fact leads to an increase of the threshold current density and to limitation of the lifetime of the devices. During the past few years, GaN-based LEDs grown on low-dislocationdensity substrates have been demonstrated by many groups. Compared to the same devices grown on sapphire, these LEDs show enhanced optical and electrical properties, such as reduction in current– voltage differential resistance, reduction in threshold voltage, and increase in output power slope efficiency. As for the LDs, practically all the GaN-based lasers have been demonstrated on bulk substrates. With improvement of the substrate and epitaxial quality, as well as the device optimizations, the green range was approached. The longest wavelength achieved is 500–514 nm at 5 mW by Nichia (Miyoshi et al., 2009). The power electronic devices also suffer from the high-dislocation densities in the heteroepitaxial nitride structures. For example, pure screw dislocations have been identified as a path for reverse-bias leakage currents in high-power Schottky barrier diodes (Hsu et al., 2002; Simpkins et al., 2003). The dislocations also have a negative impact on devices, such as UV-blind avalanche photodetectors and solar cells by influencing the dark current and photoresponse (Parish et al., 1999). By using a free-standing GaN substrate, a vertical geometry Schottky diode with a full backside ohmic contact can be fabricated, which enables higher current conduction than lateral devices. Recent research shows that vertical mobility in GaN Schottky diodes is about 6 times higher than the lateral mobility (Zhou et al., 2006). Improved reverse-breakdown voltage performance, manifested by high breakdown field and its relatively low slope versus diode diameter, for vertical device geometries indicates that the vertical depletion mode has a more uniform field distribution than lateral mode, which minimizes premature breakdown caused by field crowding at contact edges. The reverse recovery characteristics of the Schottky rectifier, when the diode is switched from forward to reverse bias, shows an ultrafast switching time of less than 20 ns, similar for both types of device structures. Schottky barrier diodes have been fabricated that have shown some benefits of using bulk GaN as substrate instead of sapphire. Improvements in the reverse recovery time, on-resistance, reverse breakdown, and reverse leakage current were reported (Baik et al., 2003; Lu et al., 2007).

Growth of Bulk GaN Crystals

3.06.1.2

Thermal Conductivity

An additional drawback of most of the foreign substrates used for nitride-based devices is their low thermal conductivity, which is especially true for the sapphire substrates. High-power devices often require high output power and high current density, leading to high temperatures in the devices via selfheating, which reduce the device performance. A proper thermal management considers a substrate with high thermal conductivity such as SiC and AlN, but the need for a buffer and misfit dislocation generation may compromise their benefits. There are several reports in the literature, presenting finite element calculations of the heat dissipation in nitride devices grown on thermal insulating and thermal conductive GaN substrates (Narendran et al., 2004; Figge et al., 2005). The thermal resistance was shown to be 4–5 times higher when using sapphire. In the typical device structure on insulating substrate, the current is spread over a much smaller area leading to even more heating in the active region which deteriorates the device performance over time (Senawiratne et al., 2008). In the case of using GaN substrate, the wider area over which the current is spread additionally allows better heat extraction and leads to a reduction of the thermal resistance (Nam et al., 2004). 3.06.1.3

Electrical Conductivity

The advantage of the GaN substrates with good electrical conductivity over insulating substrates is multifaceted. Apparently, the use of GaN substrates allows simplified device design, which reduces the device production cost. The vertical design also allows larger contact area and larger area for spreading the current. In addition, some devices can benefit from anisotropic characteristics of some material parameters, being higher along the c-axis. 3.06.1.4 Variation of Substrate Surface Orientation Conventional device structures with QW active regions aligned parallel to the basal c-plane – in both cases grown on sapphire and on GaN substrate – suffer from the undesirable quantum-confined Stark effect (QCSE), due to the existence of strong piezoelectric and spontaneous polarizations. The strong built-in electric fields along the c-direction cause spatial separation of electron and holes that in

235

turn give rise to lower carrier recombination efficiency, reduced oscillator strength, and blue-shifted emission. The QCSE becomes more pronounced with increasing indium content in InGaN QWs needed for longer-wavelength LEDs and a noticeable blue shift of the emission occurs with increasing the current, while the effect is less pronounced at shorter wavelength with less In in the QW. The LED and LD devices can ultimately benefit from the use of designs that enable the active region to be parallel to nonpolar planes and thereby reduce the built-in electric fields as discussed above. However, the use of foreign substrates for the heteroepitaxial growth of GaN in a direction different from the [0001] axis favors the generation of stacking faults in addition to high-dislocation density, which act as self-formed QWs at certain wavelengths and thus prevent emission from intentionally formed QW at shorter wavelengths. LEDs at different wavelengths have been produced during the last several years with different surface orientations, showing no emission shift with increasing current and confirming the absence of polarization-induced field in nonpolar QWs for violet (Chen et al., 2003; Chakraborty et al., 2005b), blue (Chakraborty et al., 2005a, 2006), and green LEDs (Wetzel et al., 2008; Sharma et al., 2005; Schmidt et al., 2007). However, the external quantum efficiency in these devices was disappointingly low. Only recently, with the availability of bulk GaN nonpolar and semipolar substrates sliced from boules, the performance of the LED devices grown on them was significantly improved. With the current increased efficiencies and output power levels (Zhong et al., 2007; Iso et al., 2007; Sato et al., 2007; Li et al., 2009), LEDs are now considered more promising for solid state lightening (SSL) due to reduced power consumption, high durability, and lower maintenance cost, not to mention reduced carbon footprint on the environment. In addition, the availability of low-defect-density bulk GaN substrates sliced from boules with nonpolar and semipolar orientations allowed the production of LDs at longer wavelengths approaching the green gap during the past few years.

3.06.2 Challenges of the Bulk GaN Growth In order to take full advantage of unique electronic and optoelectronic GaN properties for the wide range of electronic and optoelectronic applications,

236 Growth of Bulk GaN Crystals

it is necessary to produce monocrystalline GaN substrates. Why until now, after many years of GaN device development, there are no high-quality and cost-effective GaN substrates available on the market? There is only one answer to this question. It is a challenge to grow bulk, high crystalline quality, large single crystal of GaN. What makes a given material easy or difficult to grow as single crystal? Crystals are normally grown from liquid or from gas phase. If a material is thermodynamically stable and melts congruently, then it is possible to grow a single crystal from its melt by using various melt-growth techniques. Unfortunately, GaN starts to decompose at temperatures above 840  C and at N2 pressure of 0.1 MPa (Unland et al., 2003). According to the P–T phase diagram of GaN, published by Karpinski et al. (1984), GaN decomposition at high temperatures is suppressed by applying pressure of nitrogen. In fact, GaN melts congruently at temperatures above 2200  C and, according to Utsumi et al. (2003), at that temperature the dissociation pressure is above 6.0 GPa. If growth from the melt faces difficulties, the second option is to grow a single crystal from solution. Unfortunately, GaN is not an easy material to grow from solution as well. The solubility of nitrogen in liquid gallium is negligible at atmospheric pressure and at temperatures below 1000  C. GaN solubility increases with temperature, but remains very low, about 1 at.%, even at temperatures of 1600 C and pressures of 2.0 GPa (Grzegory et al., 1995, 2002a,b). Low decomposition temperature of GaN at atmospheric pressures leads to the difficulties in finding a good solvent for GaN, which would provide reasonable concentration of Ga-N species in solution. In order to overcome the limitations of nitrogen solubility in gallium, a variety of approaches were developed over the years to grow GaN from solution. Deposition from vapor phase is, in most cases, the preferred technique for fabrication of thin layers of metals, dielectrics, and semiconductors. According to a general classification (Stringfellow, 1980), both

major variations of the vapor-phase growth, namely physical-vapor deposition (PVD) and chemicalvapor deposition (CVD), are widely used for thinfilm growth of semiconductors and only in special cases are used for bulk crystal growth because of their inherently slow growth rate. As described in detail in the following sections, GaN bulk crystal growth is probably the biggest exception. Practically, all vaporphase growth techniques have been explored to grow thick films on a variety of materials with the goal of separating these films and further use them as GaN native substrates for homoepitaxial nitride device growth. Currently, the bulk GaN growth development is going through an interesting phase. Why? First, the development of different bulk GaN growth techniques has taken place already for more than 10–20 years, and the science and technology behind some of these techniques are mature enough to start using them in industry. Second, the number of bulk growth methods in development is quite large at the moment (Figure 1). Currently, there are still many uncertainties in determining which GaN bulk growth techniques will prevail. Third, the development of the GaN solution growth methods reached a point which makes it possible, on the one hand, to compete between gas-phase growth and solution growth techniques and, on the other hand, to mutually complement and combine these techniques to produce GaN substrates for different applications.

3.06.3 Bulk GaN Growth from Liquid Phase 3.06.3.1

Melt Growth

As mentioned above, in principle, it is possible to grow GaN crystals from melt. Utsumi et al. (2003) studied the decomposition and melting behavior of GaN at different pressures and temperatures. Authors conducted in situ X-ray diffraction (XRD) experiments by using a multi-anvil high-pressure

Bulk single-crystal GaN growth

Gas-phase growth

Sublimation

MBE

HVPE

Figure 1 Bulk GaN growth methods.

Liquid-phase growth

Solution growth

Melt growth

Growth of Bulk GaN Crystals

(HP) apparatus for synchrotron radiation, installed on beamline at the SPring-8 site in Harima Science Garden City, Japan. They demonstrated that at pressures below 5.5 GPa, GaN decomposed into Ga and N2, and the decomposition temperature increased almost linearly with pressure. At 6.4 and 6.8 GPa, congruent melting occurred at 2225  C. It was shown that the decomposition temperature increases rapidly with pressure, whereas the pressure dependence of the melting temperature is negligible. Based on the results of GaN melting, an experiment for single-crystal growth from melt was performed at 6.8 GPa. GaN powder was placed in a boron nitride (BN) capsule and compressed in the multi-anvil apparatus. Temperature was first increased to 2300  C, which was high enough for melting; it was then slowly decreased to 2100  C at a constant rate of 1  C min1. The BN capsule recovered at ambient conditions was filled with many pieces of transparent GaN single crystals with an average size of about 100 mm, which exhibited a slightly yellowish color. Growth from melt is always considered as the preferred choice to grow bulk crystals of the material because of its simplicity and ability to produce large

237

quantities of material. Unfortunately, the growth of large GaN crystals from the melt (with diameter of 20 and bigger) is not technologically and economically feasible since the necessary pressures and temperatures are even higher than those used to produce synthetic diamonds. 3.06.3.2

Solution Growth

While there is a strict limitation to grow large GaN crystals from melt because of extremely high decomposition pressure of nitrogen at GaN melting point, a variety of approaches to grow GaN crystals from solution are being developed. The main motivation for GaN solution growth development is in the expected perfection of the bulk material grown from solution at near-equilibrium conditions. The key differences between the bulk GaN solution growth techniques in development are given by differences between the combinations of solute and solvent which are used by each technique. This combination determines the GaN growth conditions and distinctive features of each particular technique (Figure 2).

GaN solution growth

Nitrogen

Basic solutes

Solvent base

Ga

Driving force of crystal growth

The way solution is formed

Ga+Na

Ga+Li

Direct N2 dissolution

5–20 kbar

Temperature

1300–1600 °C

Pressure control solution growth

GaN dissolution

5–50 torr

600–1000 °C

RF, MW plasma

HPNS – high−pressure solution growth LPSG – low−pressure solution growth NAP – near−atmospheric pressure

Ga free solvents

Supercritical NH3

Temperature gradient: difference of GaN solubility at different T

Difference of gas (N) and solid-phase (GaN) solubility

Pressure

HPNS growth

GaN or Ga

Excited N2 dissolution

≤1 atm

1–2 atm

800–1000 °C

DC or AC electric field

LPSG

Chemically enhanced GaN dissolution

Chemically enhanced N2 dissolution

20–100 atm

1–3 atm

20–100 atm

1–5 kbar

900–1000 °C

750–900 °C

700–850 °C

400–600 °C

NH3 dissolution

Sodium flux

NAP solution growth

Ga free solvents

Na flux LPE

Ammonobasic (NH2–)

Ammonoacidic (NH4+)

5–20 kbar

600–1000 °C

Ammono thermal

High-pressure ammonothermal

AMMONO-bulk growth

Figure 2 Bulk GaN growth from solution. HPNS, high-pressure solution growth; LPSG, low-pressure solution growth; NAP, near-atmospheric pressure.

238 Growth of Bulk GaN Crystals

The driving force for GaN crystallization in any solution is determined by the degree of the solution deviation from thermodynamic equilibrium. This deviation is governed by the difference between free energies of the system in the initial and final stages, and the process proceeds in the direction which decreases the free energy of the system. The free energy of the mole of a substance is its chemical potential, so the difference  ¼ (Ga-N)  (GaN) between the chemical potential (Ga-N) of Ga-N species in the solution and the chemical potential (GaN) of crystallizing GaN at given conditions defines the driving force of GaN growth from solution. For convenience, the difference in chemical potential (also called supersaturation) is associated with measurable physical quantities. The difference between the actual concentration and the threshold concentration of solutes that can be dissolved into liquid (i.e., the concentration C0 of a saturated solution and solubility) is taken as criterion. If solution is saturated by the solute at a concentration C0, that is, if solution is in equilibrium with a crystal, the chemical potential of the solute is equal to the chemical potential of the crystal. If some characteristic of the system (P, T, or C) is altered, the chemical potentials of the crystal and of the solute change as well. The solute concentration C may become higher than C0 for the new conditions. The level of supersaturation  ¼ (C  C0)/C0, which can be calculated from C0 and the actual solute concentration C, characterizes the driving force of crystallization. The presence of a thermodynamic driving force itself does not mean that a process will take place. The particles in the solution should have enough energy to overcome the energy barrier of the new phase formation – crystallization. The crystallization starts with the creation of the nucleus. To continue its growth the nucleus should exceed the critical size, otherwise it will dissolve back to liquid. The size of the critical nucleus, which can survive, and nucleation rate depend on supersaturation. The bigger is the supersaturation, the smaller is the critical nuclei size; the higher is their nucleation rate, more of the nuclei can survive and continue to grow. The activation energy of the formation of the supercritical nuclei in the volume of the solution (homogeneous nucleation) is higher than on the solid surface (heterogeneous nucleation). The activation energy of the heterogeneous nucleation also depends on the atomic structure of the surface. The lowest activation energy is for epitaxial nucleation on the surface of the native

crystal. Thus, when the supersaturation exceeds some level and solution becomes labile (unstable), GaN crystals nucleate homogeneously and continue to grow in the volume of the solution. Below this level of supersaturation, all homogeneously forming nuclei dissolve back into solution and the solution remains metastable. When the solution is metastable, crystals can nucleate and start to grow only on available surfaces – walls of the crucible, dust particles in the solution, etc. The lowest degree of supersaturation is required to grow GaN on GaN seed. Therefore, to prevent spontaneous nucleation and growth, it is necessary to be inside the solution metastable region, which is also called the Ostwald–Miers region. In the following sections, a variety of solution growth techniques are discussed. Each of these techniques has specific features but, at the same time, there are common basic elements that unite all of them. These basic elements are discussed in more detail in the first few sections and in less detail in the successive sections. 3.06.3.2.1

Direct N 2 dissolution in Ga The simplest system to grow GaN from solution is to dissolve nitrogen in liquid gallium. In this system, the solute is nitrogen and the solvent is gallium. For the successful implementation of the solution growth, it is necessary to know P–T phase diagram of Ga–N2–GaN system. This phase diagram was studied at HPs and published by Karpinski and Porowski (1984) (Figure 3(a)). The latest and the most advanced data on the Ga–N2–GaN phase diagram at low pressures (Figure 3(c)) and nitrogen solubility in gallium are presented in the work by Unland et al. (2003). GaN decomposes at 1 atm of N2 pressure and temperatures higher than 840  C. At these temperatures, the molecular nitrogen will not dissolve in the liquid gallium because the system does not have enough energy for N2 adsorption and dissociation on the gallium surface, and also the solubility of nitrogen in gallium at these parameters is approaching 0 at.%. One way to overcome the potential barrier for the N2 molecule dissolution in Ga is to increase temperature and simultaneously suppress GaN decomposition by applying pressure. The first experimental results of nitrogen dissolution in liquid gallium at HPs and the first GaN crystals obtained by HP solution growth were presented by Madar et al. (1975). Their experiments were limited at pressures of 1.0 GPa, and temperatures of 1200  C. At these conditions, nitrogen solubility became evident,

Growth of Bulk GaN Crystals

(a) 30

(c) 80

25 GaN (s)

20

60

15

P (bar)

P (kbar)

10

GaN (s)

40

Ga (l) + N2 (g)

5

20

0 1200

1600 T (K)

0 800

0.8 0.6

Ga (l) + N2 (g)

2000

(b) 1.0

N (at.%)

239

840

880

920

960

1000

1040

T (°C) GaN (s)

0.4 0.2 0.0 1200

Ga (l)

1600

2000

T (K) Figure 3 Ga(l)–N2(g)–GaN(s) phase diagram: (a) P–T coordinates; (b) liquidus line for Ga–GaN system; (c) P–T coordinates. (a) Based on Karpinski J et al. (1984). (b) Courtesy of I. Grzegory. (c) Based on data from Unland et al. (2003)

about 0.1 at.%, but still very low and growth rates were much lower than 1 mm h1. One important result of this work was the evidence that N2 can be dissolved directly in the liquid Ga at high temperatures and pressures. HP nitrogen solution growth technique. Based on the work intended to determine Ga–N2–GaN phase diagram at HPs, GaN HP nitrogen solution (HPNS or HPS) growth technique was further developed at the Institute of High Pressure Physics of the Polish Academy of Sciences. Here are the most advanced facilities for developing bulk GaN growth from solution at HPs. Two stages of HPNS growth have to be distinguished for a better understanding of the growth mechanism. The first stage is the process at the gas– liquid interface, and the second stage takes place in the liquid. Under HP and temperature, the molecular nitrogen N2 is adsorbed on the gallium surface. Adsorption of nitrogen on liquid Ga surface leads to dissociation of N2 molecule (Krukowski, 1999). The energy barrier for this process is in the order of 4.2 eV, which is less than half of the N2 bonding energy. After the dissociation at the gallium surface, nitrogen dissolves in the melt in the atomic form. Figures 3(a) and 3(b) combined with each other

show how nitrogen solubility in gallium increases with temperature and pressure, and how small is the equilibrium nitrogen concentration in liquid gallium even at HPs and high temperatures. At a given temperature, the concentration of nitrogen in the melt at equilibrium with diatomic N2 depends on the N2 partial pressure and is governed by Sievert’s law, which states that the nitrogen solubility in the melt is proportional to the square root of the partial pressure of N2 above the melt: Neq ¼ KeqPN2, where Neq is the equilibrium nitrogen concentration or solubility, Keq is the equilibrium constant for the reaction ½ N2 ðgÞ $ N ðLÞ and PN2 is the partial pressure of N2. In return, the solubility of GaN in the melt refers to the equilibrium nitrogen concentration over GaN at the same given temperature. At the equilibrium P–T conditions, the nitrogen concentration in the solution is at equilibrium with N2 above the melt and with GaN in the melt. If the pressure is increased, the N2 solubility increases and nitrogen concentration in the solution exceeds the GaN solubility. The solution becomes supersaturated and the driving force for the GaN growth is established. In other words, the nitrogen supersaturation, which formed during N2 dissolution, is defined by the difference between

240 Growth of Bulk GaN Crystals

gaseous nitrogen and GaN solubility at the given P and T. Typical HPNS growth is conducted at P and T of GaN stability with reasonable overpressure. It causes relatively high rates of N2 dissolution and quickly creates in the liquid Ga a highly supersaturated subsurface layer. As a result, intensive nucleation of GaN in this layer occurs, and a polycrystalline surface crust of GaN is usually formed. The size of the crystals in this crust depends on supersaturation, which in return depends on the combination of temperature and nitrogen pressure. The next stage of the conventional HPNS growth strongly depends on the temperature distribution in the crucible. At the P–T conditions corresponding to the GaN stability, the solubility of GaN in gallium is an increasing function of temperature. Therefore, if proper temperature gradients (TGs) are maintained in the system, they create the driving force for the growth by creating the supersaturation in the part of the solution with lower temperature. As a result, GaN is dissolved in the hotter part of the solution (dissolution region), transported by diffusion and convection to the cooler part, where GaN crystallizes from the supersaturated solution. The supersaturation in different parts of the solution due to GaN dissolution depends on TGs in the solution. Obviously, if N2 still has an access to the melt surface, it continues to dissolve into solution. It makes the formation and distribution of the supersaturation more complex. N2 dissolution creates supersaturation in the subsurface layer of the melt due to either N2 overpressure or deviation of the P–T conditions from equilibrium. Supersaturation inside the melt also depends on the TGs, because nitrogen solubility depends on temperature. Nitrogen from both, N2 and GaN sources, moves from the dissolution region by diffusion and convection, and creates excess of nitrogen concentration over the equilibrium concentration in the parts of the solution at lower temperature. Convection of the solution is crucial for the solute transport to the growing crystal, and depends on the arrangement of axial and radial TGs in combination with gravity. The development of HPNS growth requires the development of a special equipment which allows one to create gas pressures up to 2 GPa, temperatures up to 2000 K, and favorable TGs in the Ga melt volume, sufficient to grow crystals of 1 cm or bigger. The high gas pressure system, designed and built in the Institute of High Pressure Physics (Poland) for the growth of GaN crystals (Figure 4),

Figure 4 The high-pressure apparatus, constructed in Institute of High Pressure Physics (Poland) for crystallization of GaN. The maximum working pressure is 15 kbar, the maximum temperature –1600  C, the internal diameter – 100 mm. Photo is courtesy of M. Bockowski.

allows one to conduct growth under controlled temperature and pressure for more than 200 h (Krukowski et al., 1999). At present, GaN crystals are grown in gas pressure chambers with volume up to 1500 cm3, allowing crucibles with the working volume of 50–100 cm3. The HP–HT reactor consisting of the pressure chamber and the multizone furnace is equipped with additional systems necessary for in situ annealing in vacuum, electronic stabilization and programming of pressure and temperature, and cooling of the pressure chamber. Pressure in the chamber is stabilized with precision better than 10 bar. The temperature is measured by a set of thermocouples arranged along the furnace and coupled with the standard input power control electronic systems. This allows stabilization of temperature 0.2 and programmable changes of temperature distribution in the crucible. Crucibles are made from BN or graphite. HPNS spontaneous growth. The best GaN crystals, grown by the HPNS method at pressures in the range

Growth of Bulk GaN Crystals

of 10–20 kbar and temperatures of 1400–1600  C, are self-nucleated, that is, no intentional seeding is used to grow these crystals. The crystals usually nucleate spontaneously on the internal surface of the polycrystalline GaN crust covering liquid Ga. Grown crystals are of wurzite structure, mainly in the form of hexagonal platelets (Porowski, 1999), shown in Figures 5(b) and 5(c). The large hexagonal surfaces correspond to {0001} polar crystallographic planes. The side faces of the crystals are mainly the semipolar {1011} and also nonpolar {1010} m-planes. The growth is strongly anisotropic, being much faster (about 100 times) in nonpolar directions perpendicular to the c-axis. The crystals grow therefore as hexagonal platelets with rates of about 1 mm h1 in c-direction and <100 mm h1 in ‹1010› directions (perpendicular to the c-axis). The HP solutiongrown GaN single crystals are transparent, of good morphology, with flat mirror-like faces, suggesting stable layer-by-layer growth. The maximum lateral size of the platelets for 100–150 h processes is 10–14 mm, whereas the thickness is 80–120 mm. The average size of the self-seeded crystals corresponds to the diameter of the HP reactor. According to Grzegory et al. (2002a), the significant increase of the temperature difference between the hot and cold ends of the crucible, which increases the supersaturation, does not lead to a considerable increase of (a)

z

241

the growth rates but rather to growth instabilities on the {0001} polar surfaces of the crystals. The most frequently observed features of the unstable growth are cellular structures, accelerated growth at the edges exposed to the N-flow, and, in the extreme case, hollow needles elongated along the c-directions (Figure 5(d)). The morphology of the two principal surfaces of spontaneously grown GaN platelets is different: one side is completely flat and the other is corrugated, with large number of multi-atomic steps, ridges, and even growth hillocks. It has been found that surface morphology and quality of the GaN crystals crucially depend on the doping of the growth solution (Grzegory et al., 2001b). For undoped Ga solution the corrugated surface corresponds to Ga face (0001) face, and the flat surface to N face (0001). The addition of Mg to liquid Ga changes the surface morphology of the plate – Ga face becomes completely flat and N face is corrugated. The Mg or Be impurity introduced into the growth solution does not change the general habit of the crystals significantly. In both cases, they are hexagonal platelets with c-axis perpendicular to the hexagonal faces. The crystals with Mg are perfectly transparent and colorless, whereas the undoped conductive samples are usually slightly yellowish in color. Crystals with Mg are generally thicker and require longer time to (c)

[0001]

c-plane (0001)

(1

m-plane

ne la ) r-p 120

e

a-plan

(1100)

Ga face

) (1210

[1210

]

y [2110]

N face

x

]

20

[11

(d)

1 mm

u (b)

[0001]

c-plane Figure 5 (a) Schematic of polar, nonpolar, and semipolar planes and crystallographic directions in GaN crystal; (b) schematic of GaN platelet crystal; (c) sample of the GaN platelet crystal, spontaneously grown by HPNS; (d) sample of the spontaneous GaN needle elongated along the c-direction. Photographs are courtesy of I. Grzegory.

242 Growth of Bulk GaN Crystals

reach the same linear dimensions as the crystals grown without Mg. The size difference is related to reduction of the growth rate in the plane perpendicular to c-axis for Mg-doped GaN crystals. GaN crystals grown from the solution in pure gallium show metallic behavior in the temperature range of 4.2–300 K (Grzegory, 2002b). High free electron concentration of (3–6)  1019 cm3 with mobility of 30–90 cm2 V1 s1 has been found. The main residual impurity detected in the crystals by secondary ion mass spectrometry (SIMS) is oxygen. Its concentration was estimated to be in the range of 1018–1019 cm3. It is well established that oxygen is a single donor in GaN. However, since the concentration of free electrons in the crystals is higher than the estimated concentration of the impurity, the presence of an additional donor cannot be excluded. The N vacancy as the additional source of free electrons was often proposed to justify the high concentration of free electrons. The addition of Mg into the growth solution drastically changes the electrical properties of GaN crystals. Mg incorporation leads to drastic decrease of free electron concentration and increase of electrical resistance in GaN. The room temperature (RT) resistivity of GaN crystals doped with Mg increases by orders of magnitude so that, at RT, these crystals are semi-insulating. The temperature dependence of resistivity for these samples at low temperatures up to about 250 K suggests hopping type of conductivity. At higher temperatures, the conductivity starts to be governed by the activation process, leading to the creation of free holes in the valence band. The GaN crystals grown by HPNS are of high structural quality as determined by XRD, transmission electron microscopy (TEM), and defectselective wet etching (etch-pit density (EPD)) techniques. The X-ray rocking curves for symmetrical reflection are as narrow as 18–25 arcsec for almost all Mg-doped crystals and for best crystals grown without doping (Porowski, 1999). For some of the conductive crystals, the rocking curves split onto few peaks indicating low- angle (about 1 arcmin) grain boundaries. This is probably due to some inhomogeneity in the distribution of residual donor impurities. The rocking curves for in-plane reflections are always very narrow, indicating that there is no twist mosaicity in all investigated crystals. HPNS seeded growth. The spontaneous nucleation has evident disadvantages for the crystal production. The number of the nucleated crystals is not controlled, their array is not reproducible, self-

nucleated crystals are not optimally oriented relatively to TGs, and crystals, which are close to each other, compete with neighbors and deteriorate the growth conditions. It is well known that seeding is a vital element of crystal growth. Spontaneous crystals usually have the highest crystalline quality because they start to grow from the very small self-nucleated seed. Crystals grown from a small seed inherit very little structural imperfections in small area around the interface between the crystal and seed. Structural perfection of the seed and lattice match of the seed and growing crystal become increasingly important when the seed size increases. If crystal grows in one direction from big area seed (substrate), the crystalline quality of the seed determines, to a large extent, the quality of the grown crystal. The preparation of the seed surface is also very important for the growth follow-up. Besides the standard procedures used for GaN substrate surface preparation, removing of the thin surface layer of GaN seed by dissolving (etching and wetting) in gallium is necessary before growing from solution. Bockowski et al. (2002) note that otherwise, at high temperature and high N2 pressure, a polycrystalline GaN layer covers the gallium surface adjacent to the seed, which makes the employment of the GaN substrate useless for further crystallization. The strategy for etching is to overheat the liquid gallium and the seed above the equilibrium temperature for the Ga-N2/GaN system at a given nitrogen pressure. The etching procedure may be done in two ways: in high nitrogen pressure and corresponding high temperature or in low N2 pressure at much lower temperature. Ga-polar surface of the seed crystal wetted at 1800 K for 30 min in nitrogen pressure of 1 GPa strongly roughened with hexagonal figures represents features of decomposition. The depth of the etched material is few tens of microns. The Ga-polar surface obtained during the same period of 30 min at very low N2 pressure (about 4 MPa) in 1400 K is only slightly changed due to dissolution and decomposition. The depth of the etched material is of the order of a few microns. The development of the HPNS seeded growth requires intensive experimentation and consumption of the seed material. Spontaneously grown GaN crystals are used as seed crystals. However, lack of large high-quality HPNS-grown crystals requires to use different materials for seeds as well: sapphire/ GaN metal-organic chemical-vapor deposition (MOCVD) templates (Bockowski et al., 2004),

Growth of Bulk GaN Crystals

hydride vapor-phase epitaxy (HVPE) GaN. Both approaches – TG growth (Bockowski, 2001; Grzegory, 2001a, 2001b)) and growth, based on continuous N2 dissolution (Bockowski et al., 2008) along with different growth setups – have been developed to grow GaN on different types of seeds. For the HPNS TG seeded growth, a large TG of the order of 20–50 K cm1 is applied along the crucible at an N2 pressure of 1 GPa for 50–250 h (Bockowski et al., 2002). The experimental configuration is presented schematically in Figure 6(a). Direct axial temperature gradients do not create convection and applying the small radial temperature gradients cause some convection in the melt, which is beneficial for nitrogen transport to the seed. The typical GaN substrate with the new grown material on the Ga-polar surface is shown in Figure 6(b), and the growth on N-polar surface is shown in Figure 6(c). The material deposited on the Ga-polar surfaces is transparent, colorless, and grown as a single hillock. The growth features on N-polar surface are quite different (Bockowski et al., 2002). The presence of several growth centers has been observed. In addition, simultaneous nucleation and

243

growth of randomly oriented crystals can be noticed. The material grown on the N-polar surfaces is also not free from structural defects. The presence of dislocations and domains of orientation different from the substrates has been observed (Grzegory, 2001a). Therefore, the HPNS TG growth on the Ga-polar surface seems to be more beneficial. The normal growth rate in c-direction on Ga-polar surface is in the range from 2 to 10 mm h1 and depends on the supersaturation at the growth front. This supersaturation is mainly a function of growth temperature, TG, and the height of the gallium droplet over the substrate. Inclusions of the solvent or voids have not been observed in the material grown on the Ga-polar surface. The central part of the surface is practically free from dislocations. Another configuration of the HPNS seeded growth ((Bockowski et al., 2008) exploits the supersaturation, which forms when N2 dissolves into Ga melt. This supersaturation is a result of different nitrogen and GaN solubility at given P and T. A small overpressure in the reactor creates small supersaturation and allows one to grow GaN on a seed. It is necessary to create only a small overpressure for the

(a) N2

T(r )

z 1723 K

1673 K

Ga

1623 K

1623 K 1593 K

(b)

T(r )

T(z ) Substrate

(c)

Figure 6 (a) Schematic of the configuration used for directional TG growth of GaN on HPNS grown substrates; radial and vertical T distribution are shown. SEM images of GaN substrate with the grown material on (b) a Ga-polar surface; (c) an N-polar surface. Courtesy of M. Bockowski.

244 Growth of Bulk GaN Crystals

given temperature and accurately maintain pressure and temperature during the long growth run. Otherwise, with higher overpressure, supersaturation in the solution becomes high enough for spontaneous nucleation and growth of parasitic crystals. Small opposite axial and radial TGs are applied to cause convection of the solution and transport of the continuously dissolving nitrogen to the HVPE seed (Figure 7). The authors conducted three-dimensional computer modeling of the experimental setup to optimize heat and mass transfer by convection. It was reported that GaN was grown on HVPE seeds with growth rates up to 3 mm h1 on Ga and N faces of the seeds, at N2 pressure of 0.75 GPa and temperatures of 1350–1400  C. In general, it is more difficult to maintain and to optimize supersaturation in the solution during growth by continuous N2 dissolution than in the TG growth. In the case of TG growth, the supersaturation is controlled by the temperature difference in the solution. Small changes in temperature and pressure do not affect the supersaturation. If temperature or pressure in the system requires substantial change to optimize growth, it is relatively easy to adjust supersaturation by changing temperature gradients in the solution. In the case of growth by continuous N2 dissolution, supersaturation is much more sensitive to any changes of the pressure or temperature in the system. Both, temperature and pressure, directly govern the supersaturation and, as a result, nucleation and growth. (a)

(b)

The low GaN solubility is the main hurdle for the GaN HPNS TG growth on seeds. Recently, new efforts in development of the HPNS TG growth were reported by Grzegory et al. in the 6th International Workshop on Bulk Nitride Semiconductors, 2009. The primary goal of this work was to increase the GaN solubility in the solution. Ga–In composition with more than 50 mol.% of In was studied as a possible path to improve GaN solubility. Pressure-controlled solution growth technique. In the conventional HPNS method, the N2 pressure always exceeds the equilibrium value corresponding to the highest temperature of the growth run. At the beginning, diatomic N2 under HP dissolves in the liquid gallium according to the reaction: ½ N2 ðgÞ ! NðlÞ and saturates gallium with nitrogen at the liquid–gas interface. If nitrogen concentration exceeds the GaN solubility at given P and T, the solution becomes supersaturated. The supersaturation in the solution increases with the deviation of the P–T conditions from the equilibrium inside the region of GaN stability. The more the pressure exceeds the equilibrium pressure for a given temperature (or the more the temperature is below the equilibrium T for that given pressure), the higher is the supersaturation. As already mentioned above, the high supersaturation gives rises to a polycrystalline GaN crust underneath the surface of the liquid Ga during the first stage of the HPNS growth. During the next (c)

(d) 1623 K

GaN seed 1673 K

18 cm

1623 K 1673 K

1673 K Baffle

1673 K 1678 K

1643 K 1673 K

Figure 7 (a, c) Photographs of the graphite crucible with the baffle and GaNseeds; (b, d) scheme of the experimental setup for the seeded growth by continuous N2dissolution. Figures are courtesy of M. Bockowski.

Growth of Bulk GaN Crystals

stage, the formed polycrystalline GaN crust dissolves in higher temperature region and GaN crystal(s) grow in the lower temperature region of the solution. The supersaturation at this stage is governed by the applied TGs and depends only on T. If the nitrogen supersaturation is well controlled during nitrogen dissolution in liquid gallium, then GaN nucleation and crystal growth can be controlled at this stage as well. A group from Japan, Inoue et al. (2000), developed an original approach to exploit this possibility. It is called pressure-controlled solution growth (PS-SG) method. The idea of the method is to gradually approach the low supersaturation in the solution while dissolving N2 and to suppress rapid spontaneous nucleation in order to grow GaN crystals just on one or few developed nuclei. Prior to the growth runs, the equilibrium N2 pressure was experimentally evaluated (about 0.88 GPa) at the temperature of 1450  C (Inoue et al., 2000). It was also found that at pressure of 0.98 GPa and temperatures below 1350  C no GaN crystal growth occurred because the rate of N2 dissociation at the liquid Ga surface and nitrogen dissolution are too low. During typical growth runs (Inoue et al., 2001a) the crucible with Ga is first heated up in the HP reactor to the temperature of 1475  C while it is pressurized with nitrogen to the condition ‘a’ (see Figure 8), when GaN is not stable N2 dissolves into Ga and saturates the solution to some degree. After keeping the crucible at this condition, the nitrogen pressure is increased slowly up to 0.98 GPa, state ‘c’ in the Figure 8, which is

16.0 GaN (s) P (kbar)

12.0 c b 8.0 a Ga (l) + N2 (g)

4.0

0 1000

1200

1400

1600

T (°C) Figure 8 Ga–N pressure–temperature phase diagram. In the PC-SG method, pressure is increased as a ! b ! c and maintained at a constant temperature during crystal growth.

245

relatively close to the equilibrium pressure. After passing the state ‘b’ (Figure 8) the rate of increasing pressure (P/t) corresponds to the rate of the growing supersaturation. GaN starts to nucleate and to grow at certain overpressure (the state ‘b’ in Figure 8) above the equilibrium pressure related to a certain supersaturation. If the increase in rate of nitrogen pressure (P/t) is low, the supersaturation slowly reaches the point when a single GaN nucleus forms and continues to grow. The nucleation rate at this supersaturation is still very low, so only one or few nuclei are developed and grown. The crystal is growing and consumes the dissolving nitrogen. If P/t is high, the supersaturation may quickly reach the value that causes high nucleation rates and leads to the formation and development of many nuclei. The size of the resulting crystals would be in this case small, because they compete with each other during growth. Inoue et al. (2001a) found that it is possible to grow GaN crystals as hexagonal platelets with length in nonpolar directions of 10–15 mm. Platelets are transparent and yellowish. Crystals were grown at different P/ t : 293 MPa h1, 49 MPa h1, 12 MPa h1, and 8 MPa h1. The lower the P/t is, the lower the growth rates are and crystals of good quality may be grown. No information is given on the thickness of the grown crystals. The GaN single crystals grown at pressure increase the rate below 49MPa h1, have full width at half maximum (FWHM) of the XRD rocking curve of about 120 arcsec, without any low-angle grain boundaries. The dislocation density is estimated to be less than 105 cm2 by TEM observations, the intensity of the yellowish band becomes very weak, and the photoluminescence (PL) intensity ratio of the bandedge band to the yellowish band is greatly improved if compared to crystals grown at higher rates of N2 pressure increase. Inoue et al. (2001b) found that single crystals can reproducibly be obtained near the surface of the Ga melt. The limiting factor of the growth is given by nitrogen transport to the growing crystal. It is known that nitrogen solubility is less than 0.5 at.% under these growth conditions and that, with such low nitrogen concentration in the solution, sufficient transport of nitrogen is created only in the relatively thin layer of gallium between the surface and the growing crystal. Within the thin layer, it is possible to form a concentration gradient sufficient to provide adequate nitrogen flux by diffusion to the surface of the growing crystal.

246 Growth of Bulk GaN Crystals

A new larger HP furnace was built in order to grow larger GaN crystals (Inoue et al., 2001b). The processing volume was increased from 40 to 60 mm in the diameter and from 100 to 150 mm in the height. Ten pyrolytic boron nitride (PBN) crucibles (52 mm diameter), each charged with 80 g of 6N–Ga, were placed inside the new HP reactor. Growth runs were conducted at 1475  C and a rate of N2 pressure increase of 69 MPa h1 to grow self-seeded crystals. In addition, 20 diameter sapphire substrates with (0001) orientation were used in some of the crucibles as seeds. As a result of 16-h-growth-run selfnucleated crystals of 25-mm length and area over 300 mm2 were grown (Figure 9). The growth rate in nonpolar direction is in the range of 1 mm h1 for this crystal. There is no information about the thickness of this crystal. GaN crystals with columnar structure were heteroepitaxially grown on the sapphire substrate, with the (0001) plane of the crystals parallel to the (0001) plane of the substrate. There are no new publications on PC-SG method after 2001. Some other efforts to grow GaN from solution under HP were reported in the literature as well, but these efforts were too sporadic to produce valuable results. 3.06.3.2.2 Chemically enhanced nitrogen dissolution

Applying HP allows suppressing GaN decomposition while increasing the solution temperature in order to overcome the potential barrier for the N2 molecule dissolution in gallium. Moreover, it is

10 mm Figure 9 Sample of GaN crystals spontaneously grown by pressure-controlled solution growth (PC-SG): P ¼ 9–10 kbar, T ¼ 1400–1500  C. Central Research and Development Laboratory, Japan Energy Corporation. Yamaguchi University.

possible to increase the thermodynamic potential of gaseous nitrogen and to enhance the creation of the atomic nitrogen at the gallium surface by other methods, such as the use of species containing nitrogen atoms bonded more weakly than in the N2 molecule or to have species at the gallium surface which can promote the cracking of N2 molecule at lower temperatures, or the excitement of N2 plasma as well. Sodium flux growth. It was discovered by Yamane et al. (1997) that sodium addition to gallium dramatically decreases the pressure and temperature required to dissolve molecular N2 and to grow GaN crystals from the created solution. This approach is called sodium flux growth. GaN growth takes place inside of the GaN stability region at temperatures in the range of 650–850  C and pressures of 10–100 bar (1–10 MPa). Different aspects of the sodium flux method were studied in numerous works, and various modifications of this method were developed over the past 10 years in order to improve growth conditions. Molecular nitrogen is not dissolved in pure Ga below 950  C, and the main role of sodium is to promote the dissolution of the molecular N2 in the Na–Ga melt (Yamane et al., 2000). Sodium is one of the alkali metals which release electrons easily because of the low electron work functions. It is speculated that a N2 molecule adsorbed on the surface of Na–Ga melt could receive electrons from Na. This could weaken the bonding and the N2 molecule might dissociate into negatively charged N ions. As there is no sodium nitride and no binary nitride containing Na and Ga, in addition to the extremely low solubility of nitrogen in liquid sodium (7.1  109 mol.% N at 600  C), Ga is the only element which can react with N and form a nitride. Na flux spontaneous growth. After the dissolution, nitrogen can move inside the melt or react with Ga to form GaN at the surface of the melt; and which process will prevail depends on the rate of dissolution, nitrogen solubility, and velocity of the nitrogen transport. If the nitrogen concentration in a certain place of the melt exceeds the solubility and nitrogen supersaturation reaches the level sufficient for GaN nucleation and growth, GaN crystals will grow. It is commonly observed during sodium flux growth that GaN crystals readily form near the liquid–gas interface. It means that nitrogen dissolves and saturates the subsurface layer much faster than it can be removed by diffusion and convection. As a result, GaN polycrystalline layer may cover the melt surface or individual crystals grow near the liquid–gas

Growth of Bulk GaN Crystals

interface. According to studies by Yamane et al. (2000, 2005) and Aoki et al. (2000, 2002c), the morphology and yield of spontaneously grown GaN crystals strongly depend on the molar ratio rNa ¼ Na/ (NaþGa). As reported by Aoki et al. (2000), the yield of the crystals increases when rNa increases from 0.5 till 0.67 under the same growth conditions: pressure was maintained at 5 MPa during the heating at 750  C for 200 h. Evidently, N2 dissolves faster into the Ga–Na melt with higher Na content at the melt– gas interface. Crystal growth under these conditions takes place at the surface of the melt, on the wall, and at the bottom of the BN crucible. At lower molar ratio rNa, crystals have prismatic shape elongated in c-direction, and at higher rNa crystals acquire a platelet-type hexagonal habit. Yamane et al. (2005) studied how the yield and morphology of spontaneously nucleated crystals depend on the temperature and N2 gas pressure. The results of this study for the initial rNa ¼ 0.67 and 200 h of growth are summarized in Figure 10(a). The boundary of the GaN crystal formation region is shifted from equilibrium GaN/ GaþN2 line toward the higher pressures. GaN crystals do not form below the boundary line. This deviation is observed because it is necessary to have enough supersaturation to start spontaneous crystal growth. As shown in Figure 10(a), if a higher pressure is applied for a given temperature, a higher yield

of the crystals results. This is in agreement with an increase of supersaturation, when the deviation of the P–T conditions from equilibrium increases. This deviation from equilibrium conditions leads to the raise of both rates: nitrogen dissolution and GaN formation. When temperature is increased, under a given pressure, the yield of the obtained crystals increases because the dissolution rate becomes higher at higher temperatures. When supersaturation increases, the habit of the crystals changes from the hexagonal prisms elongated in c-direction to platelets. While GaN crystals grow from the sodium–gallium melt, gallium is consumed by these growing crystals. The Ga content decreases and the molar ratio rNa increases. In parallel with Ga consumption during GaN growth Na evaporates from the melt. The equilibrium vapor pressures of Na at 720 and 800  C are 18 and 45 kPa, respectively (Yamada et al., 2005). The resulting rNa during growth depends on the ratio between Ga consumption and Na evaporation rates. If the growth rate is faster than that of Na evaporation from the melt, the rNa increases along with the supersaturation, and the morphology of the crystals changes to thin platelets as reported by Yamane et al. (2005) for the samples prepared by using a premixed Ga–Na melt with rNa ¼ 0.67. Sodium vapor supply. Yamada et al. (2005) developed an interesting approach to maintain the molar ratio

(a) 5

68%

100%

100%

100%

Platelets

Pyramids

Prisms

Not formed

(b)

4 N2 pressure/(MPa)

247

35%

61%

100%

100%

25%

35%

82%

95%

19%

24%

3

2 68% (II)

(I)

(c)

1 6%

13% 1 mm grid

700

750 800 Temperature (°C)

850

Figure 10 (a) Temperature and N2 gas pressure conditions, and the yield and morphology of GaN crystallized on the crucible wall. (b) Photograph of a sample prepared in PBN crucible using 99.95% Na purified by distillation. (c) GaN platelet single crystal spontaneously grown in the crucible. Figures are courtesy of H. Yamane.

248 Growth of Bulk GaN Crystals

rNa constant during the long growth run. Instead of using premixed Ga–Na melt, they loaded gallium and sodium separately into BN crucibles (16-mm inner diameter, 12-mm depth). Then, the crucibles with Ga and Na were placed at the upper and lower positions within a 21-mm inner diameter, 400-mmlength stainless-steel container, respectively. The sample was heated at 720–800  C for 200 h with an electric furnace and the N2 pressure was maintained at 5 MPa during the growth. It was found that the Ga melt heated in Na vapor absorbs Na from the vapor and transforms into a Ga–Na melt. Colorless transparent GaN single crystals grew from this melt. It was verified by Yamada et al. (2006) that the growth of the prismatic single crystals started and continued from 75 to 200 h. During first 75 h, the Na fraction in the melt gradually increased and then rNa remained almost constant at 0.38–0.45. The authors explain such quasi-constant melt composition by existence of Na5Ga8 and other similar stable species in the melt. To maintain conditions which provide a low supersaturation during nitrogen dissolution, Aoki et al. (2002a) improved the growth setup and used Na purified by distillation and PBN crucibles. As a result, they demonstrated growth of spontaneous GaN platelet crystals of 5–10 mm in the longest direction (Figures 10(b) and 10(c)) at 750–775  C and 5 MPa of N2 pressure. Spontaneous crystals grown by sodium flux method are transparent, colorless, or slightly brown-colored with well-defined hexagonal shape. These crystals possess high crystalline quality. X-ray rocking curves measured for the 0002 reflection may have FWHM of 24–25 arcsec for small crystals, 45 arcsec for the 6-mm-size platelet, and 55 arcsec for the 10-mm-size platelet. Aoki et al. (2002c) determined the polarity of the spontaneously grown GaN single crystals by using X-ray anomalous dispersion and investigated the relation between the {0001} surface morphology and the direction of the polar axis. In the colorless transparent platelet crystals, the smooth basal plane face is the (0001) N-face and the other side with many step edges and hexagonal pits is the (0001) Ga-face. Seeded Na flux growth. Spontaneous nucleation cannot provide consistent conditions for the development of industrial crystal production. Manufacturing of large crystals is based on the seeded growth. The development and study of the spontaneous growth are only the initial stage for the development of the growth method. Growth on a

seed requires an accurate control of the supersaturation in the volume of the flux to suppress parasitic spontaneous nucleation and to grow GaN only on a seed. Similar to the HP solution growth from the Ga melt, the sodium flux growth can be performed in two ways. The first is already described for spontaneous GaN growth from the sodium flux. It is conducted by continuous N2 dissolution into the flux, resulting in a supersaturated nitrogen solution. The driving force for this growth is produced by the difference of the nitrogen chemical potentials in the gas and in GaN or, in other words, by the difference of nitrogen and GaN solubility at given P and T. Supersaturation in the solution depends simultaneously on both pressure and temperature in the system. The presence of sodium in the system makes possible to dissolve the molecular N2 at much lower temperature and pressure than in the absence thereof, and the sodium concentration varies the rate of the N2 dissolution. The nitrogen and GaN solubility in the Ga–Na melts is also higher than in pure Ga, and it gives more room to control the supersaturation during dissolution of nitrogen. Therefore, the possibility of growing large crystals by dissolving N2 gas into the flux strongly depends on both P and T parameters and the combination of few major factors, including the nitrogen dissolution rate, the GaN solubility, and transport of the nitrogen species in the solution to the crystals. The second approach exploits the difference of the GaN solubility in Ga–Na flux at different temperatures. The main parameter in this case is the temperature difference between GaN dissolution area and the area where GaN crystallizes on seed. The GaN solubility and dependence of the solubility on temperature are the major factors defining the TG growth. Nitrogen solubility in the Na flux. It is important to know the nitrogen and GaN solubility in the Ga–Na melt in order to clarify the limitations in the growth rates of both described approaches of the sodium flux method. It is also important to know how the solubility depends on temperature and on the flux composition. Aoki et al. (2003) estimated the nitrogen solubility in Ga–Na (Ga ¼ 0.5 mol.%, Na ¼ 99.5 mol.%) melt at 800  C and 5 MPa of N2 pressure as 0.25 mol.%. This is eight orders of magnitude larger than the solubility in pure Na. The authors hypothesize that the increase in nitrogen solubility is due to the formation of soluble species such as GaxNy(Na), whose structure and composition

Growth of Bulk GaN Crystals

Solubility of GaN (mol.%)

depend on the N2 pressure and the Ga concentration in the solution. A thorough study of the GaN and nitrogen solubility was conducted by Kawamura et al. (2005) and Morishita et al. (2005). Nitrogen solubility is referred to the nitrogen concentration in the solution at equilibrium with N2 above the melt, and GaN solubility is referred to the nitrogen concentration in the solution at equilibrium with GaN in the melt. GaN solubility, measured by Kawamura et al. (2005), is shown in Figure 11. These measurements differ, however, from new results published in a subsequent paper by the same group (Morishita et al., 2005). The authors consider that the new measurements are more accurate, because they used larger amount of the starting materials. According to the new measurements, the GaN solubility for the same Ga–Na flux composition (rNa ¼ 0.73) is much lower, and it is only in the order of 102 mol.% at 800  C in Ga0.27Na0.73 alloy (Figure 11). It is worth noting that the presence of Na in Ga–Na melt not only promotes the N2 dissolution, but also, evidently, increases the GaN solubility as well, which is almost zero in pure Ga at temperatures below 1000  C. Moreover, the dissolution rates depend on the Na molar ratio in the Ga–Na melt, and the nitrogen solubility also depends on the GaxNa1x composition (Kawahara et al., 2007). To clarify the roles of Ga and Na and the origin of the strong dependence of the GaN crystal yield on the flux composition, Kawahara et al. (2007) employed first-principles theoretical calculations. The nitrogen solubility is calculated at T ¼ 800  C as a function of the Ga fraction and presented along with the experimental values. The calculated nitrogen solubility agrees well with the experimental results, which were mentioned above, and shows the strong dependence of the solubility on Na/Ga ratio. There is a steep increase in the N solubility in the far Na-rich side. The N solubility

0.4 Kawamura et al. (2005)

0.3

Morishita et al. (2005)

0.2 0.1 0 600

650

700

750

800

850

T (°C) Figure 11 The solubility of GaN in Ga–Na (rNa ¼ 0.73) melt. (figures are courtesy of Y. Mori-adapted with permission).

249

starts to decrease, and its proportionality with Ga fraction is broken somewhere below the Ga fraction of 0.27. The authors consider that it is probably related to Ga aggregation phenomena appearing above a certain Ga fraction. The low nitrogen solubility in Ga–Na flux creates difficulties for the growth of large GaN crystals with reasonable growth rates. It is known that solutes, having greater affinity for nitrogen than Ga–Na, should decrease the activity coefficient of nitrogen and increase the nitrogen solubility. In the search to improve growth conditions Kawamura et al. (2003) and Morishita et al. (2003) studied the growth of GaN crystals from Ga–Na melt with Ca and Li additives to the flux. They found that both Ca and Li increase growth rates of GaN and at the same time make possible to grow transparent crystals. To clarify the origin of these observations, Morishita et al. (2005) studied how Ca and Li added to Ga–Na flux change the nitrogen and GaN solubility, and how these additives change the threshold pressure for the spontaneous nucleation and growth. The measurements of the solubility and threshold pressures were conducted for the melts with 27 mol.% of Ga with respect to the flux and 5 mol.% of Ca or Li with respect to Na. The authors found that the addition of Ca and Li significantly enhances N and GaN solubility in the melt. Additionally, authors provided very valuable results showing how GaN solubility depends on temperature and how N2 solubility depends on temperature and pressure (Figure 12). These results can be used, for example, to estimate the absolute value of the supersaturation at different temperatures and pressures for a particular flux. This value of the supersaturation can be obtained from the plots presented in Figure 12, as the difference between the nitrogen solubility at the specific T and P and the GaN solubility at the same T. Measurements of the threshold pressures for different compositions at different temperatures show that the threshold pressure changes depending on the relation between nitrogen solubility and GaN solubility. The addition of Ca and Li to the Ga–Na melt promotes the N2 dissolution along with the increase of the GaN solubility, and can reduce the threshold pressure if the supersaturation – based on the resulting difference between nitrogen and GaN solubility – is sufficient for GaN growth at lower pressure. Seeded growth by continuous N2 dissolution. The successful growth of GaN on a seed from sodium flux solution was reported by Aoki et al. (2002b). Growth on GaN seeds was performed by continuous

0.6 0.5

850 °C

0.4

800 °C 750 °C

0.3

Solubility of nitrogen in the Ga–Na melt

0.2

700 °C 650 °C

0.1 0 0

Solubility of nitrogen (at.%)

900 °C

(a)

1

3 2 4 Pressure (MPa)

5

6

Solubility of GaN (mol.%)

Solubility of nitrogen (at.%)

250 Growth of Bulk GaN Crystals

0.5

(c)

Ga–Na–Li Ga–Na–Ca Ga–Na

0.4 0.3 0.2 0.1 0 600

650

700

750

800

850

900

T (°C)

Solubility of nitrogen in the Ga–Na melt

(b) 0.5 0.4

900 °C 850 °C 800 °C 750 °C 700 °C 650 °C

0.3 0.2 0.1 0

0

1

2 3 4 Pressure (MPa)

5

6

Figure 12 (a) Solubility of gaseous nitrogen in the Ga–Na–Li and Ga–Na melt. (b) Solubility of gaseous nitrogen in the Ga–Na–Ca and Ga–Na melt. (c) Solubility of solid GaN in Ga–Na, Ga–Na–Ca, and Ga–Na–Li melt. (figures are courtesy of Y. Mori-adapted with permission).

dissolving N2 without applying a TG. Ga–Na flux with rN ¼ 0.86 was used for the growth. The authors determined three regions on P–T phase diagram, which represent the conditions for (III) spontaneous GaN nucleation and growth, (II) seeded growth, and region (I) where the seed crystals decompose (Figure 13(a)). It was found that the supersaturation at 850  C and at 2 MPa N2 partial pressure was not high enough to grow spontaneous crystals, and GaN was grown only on a seed. Colorless transparent platelets of GaN single crystals (1–3 mm size, 0.025–0.20 mm thick) prepared by the Na flux method were used as seeds. The seed crystals were put at the bottom of a sintered BN crucible. The polarity of seed crystals significantly affected the crystal growth along the c-axis. When the rough Ga-polar surface of the seed crystals faced up to the melt, the crystals did actually grow on seeds. On the other hand, crystal growth was barely observed in the case of exposure of the smooth N-polar surface to the melt. As shown in Figures 13(b) and 13(c), the quality of the GaN black crystals grown from seeds (200-h growth run) is much worse than that of the original transparent seeds.

Yamada et al. (2005) reported that Ga melt heated in Na vapor at 720–800  C and PN2 of 5.0 MPa absorbs Na from the vapor and transforms into a Ga–Na melt, which has an almost constant composition. Colorless transparent prismatic GaN single crystals grew on the wall of a BN crucible from this melt. Yamada et al. (2009) have also attempted seeded growth of GaN by heating a Ga melt with Na vapor. The schematic drawing of the setup used for the Na flux growth exploiting Na vapor is shown in Figure 14(a). The growth was conducted for 72 h at 900  C and N2 pressure of 1.6 MPa. Transparent colorless crystals were grown on seeds (Figures 14(b) and 14(e)). The growth rate of these crystals was about 1.2 mm h1. The FWHM of the XRD peak of 1010 from the m-plane of the seeded grown crystal shown in Figure 14 is 113 arcsec. It is broader than that of the seed crystals (25 arcsec). The authors consider that the main advantage of using Na vapor is the gradual saturation of Ga with sodium during long time before the growth begins. During this time, nitrogen dissolves slowly and is uniformly distributed in the melt. The supersaturation gradually increases while the Na molar ratio also increases.

Growth of Bulk GaN Crystals

Temperature (T/ °C)

(a) 850

800

251

(b) 700

750

6 5 /MPa

2

N2

3

N2 pressure P

4

(III)

(II)

(c)

(I) 1 Platelets Prisms Seeded Growth Decomposition

0.5

1 mm grid 0.85

0.90

0.95

1.00

1.05

Temperature–1 (T –1/10–3 k–1 ) Figure 13 (a) Nucleation regions and morphology of GaN by the Na flux method. GaN seed crystals decompose in the region (I), GaN grows only on seeds in the region (II), GaN nucleates on both seed crystals and crucible wall in the region (III). (b) Photographs of GaN seed crystals. (c) Crystals grown on seeds by continuous N2 dissolution. Figures are courtesy of H. Yamane.

(a) Ni-cover

(b)

(c)

Ga

Ni-cap

Upper BN crucible Seed crystals

Ni-stage Na

Lower BN crucible Outer BN crucible

Heater

1 mm (d)

(e)

N2 /Ar gas SUS-container Thermocouple Figure 14 (a) Schematic of the apparatus used for GaN seeded growth by the Na flux method with Na vapor supply; (b) seed crystal with c (N–polar) direction up to the melt, (c) GaN grown on the seed (b) by continuous N2dissolution, (d) seed crystal with þc (Ga-polar) direction up to the melt, (e) GaN grown on the seed (d) by continuous N2dissolution. Figures are courtesy of H. Yamane.

252 Growth of Bulk GaN Crystals

Growth on a seed starts with very low growth rates from the uniformly saturated solution, which is beneficial for the initial stage of growth. Na flux liquid-phase epitaxy. Based on thorough research and intensive experimental trials Kawamura et al. (2009a) developed the most advanced approach for the sodium flux method up to now. It is called Na flux liquid-phase epitaxy (LPE) method. LPE growth allows one to grow single-crystalline layer on a big area substrate (Figure 15(a)), and may shorten the way to the industrial production of large GaN substrates. Up to date, the authors demonstrated 3-mm-thick GaN layer grown from Na flux solution on a 20 GaN thin film prepared on a sapphire substrate by MOCVD (Figure 15(b)). The Na flux LPE GaN growth is driven by the supersaturation created by the difference between nitrogen and GaN solubility during the continuous dissolution of N2 under pressure. LPE growth was performed with a Ga0.27Na0.73 flux composition, at 4.5 MPa (45 bar) for 192 h. The estimated growth rate in c-direction is about 20 mm h1. Due to limitations inherent to low GaN solubility, Kawamura et al. (2009a) use a relatively thin layer of the Ga–Na melt to shorten the distance between the substrate and the melt/N2 interface. Nitrogen continuously dissolving into the subsurface of the Ga–Na melt needs to be delivered to the substrate with a velocity which provides the sufficient amount and even distribution of nitrogen near the substrate. Otherwise, the growing crystal traps impurities and N2 Gas (50 atm)

(a)

inclusions from the solution, and also loses its stoichiometry due to nitrogen deficiency in the grown crystal. Nitrogen transport in the solution is provided by both the nitrogen diffusion and the solution convection. For sufficient convection of the melt, Kawamura et al. (2009a) exploited the combination of the mechanical stirring and natural thermal convection. Kawamura et al. (2003, 2009c) introduced mechanical stirring to the Na flux LPE growth method by swinging the whole chamber back and forth (Figure 16(a)). For the natural thermal convection, an inverse axial thermal gradient was applied. The temperature at the bottom of the alumina crucible is maintained at 890  C and at the top is at 860  C. Such temperature distribution combined with the force of gravity creates convection in the melt. The inverse to the regular GaN solubility TG is used because the driving force of the growth (supersaturation) is created by continuous N2 dissolution and not by the difference of the GaN solubility at different temperatures. The intense motion of the melt by convection makes the nitrogen distribution in the solution more uniform and increases nitrogen concentration near the substrate. The delivery of nitrogen by convection boosts the supersaturation near the substrate and provides higher GaN growth rates on the substrate. The higher solubility of GaN at higher temperature near the substrate is surpassed by the continuous nitrogen dissolution and delivery from the surface of the melt to the substrate. The combination of both thermal convection and stirring creates the best conditions for

(b)

750 °C or 800 °C Stainless steel container Alumina crucible N2 Gas Polycrystal

860 °C N

N

N N

Ga–Na melt 890 °C

3 mm LPE-GaN MOCVD-GaN

Figure 15 (a) Schematic of the Na flux LPE setup and growth. (b) Photo of the 20 in diameter, 3-mm-thick Na flux LPE GaN crystal. (figures are courtesy of Y. Mori-adapted with permission)

Growth of Bulk GaN Crystals

(a)

253

Temperature-and pressure-resistant chamber

(b)

Exhaust gas Nitrogen gas Ga+Na Motor

Heater Electric furnace Crucible

Figure 16 (a) Schematic view of the chamber with the mechanical stirring, (b) Photo of the 20 GaN substrate grown with combination of thermal convection and mechanical stirring. (figures are courtesy of Y. Mori-adapted with permission).

GaN growth on the substrate (Kawamura et al., 2009a) (Figure 16(b)). As already discussed, it is possible to choose P and T conditions for the Na flux growth by which the supersaturation formed during N2 dissolution is not sufficient for spontaneous nucleation, and GaN grows only on seed (Aoki et al., 2002b). Unfortunately, growth rates under these conditions are very low. Higher supersaturation leads to higher growth rates, but causes spontaneous nucleation and growth of parasitic crystals near the gas–liquid interface. Kawamura et al. (2008) found that adding a small amount of carbon to the Ga–Na flux effectively inhibits parasitic nucleation and allows growing at higher rates. Parasitic nucleation was completely suppressed and GaN was effectively grown only on the substrate in Ga0.27Na0.73 solution at 800  C and 5.0 MPa of N2 pressure with the addition of 1 at.% of C. In the same growth run, but without adding C, almost all Ga was consumed by the parasitic crystals. Evidently, at 750  C temperature, the supersaturation is higher when the same N2 pressure of 5.0 MPa is applied. These conditions required adding 3 at.% of carbon to suppress parasitic GaN nucleation. The authors determined with SIMS that the presence of carbon in the solution does not change the C concentration in the grown crystals. It is about 2  1017 cm3 for the crystals grown in the solution without carbon and with carbon. It was also found that excess of carbon in the solution slows down growth rates in m-direction and develops m-faceting. The effects of the C additions are not fully clarified yet. Probably its presence in the Ga–Na solution increases the surface energy of GaN in the solution

and requires a higher supersaturation for the formation of the critical nuclei. The substrate for Na flux LPE growth is prepared by MOCVD growth of a thin GaN layer on sapphire wafer. The crystalline quality of such a GaN layer is quite poor for using this material as a seed in the bulk growth. Fortunately, it was found (Kawamura et al., 2009b) that GaN growth by the Na flux LPE technique reduces the dislocation density from 108 cm2 in the substrate to 104 cm2 in the grown GaN layer. The authors proposed the following mechanism of dislocation reduction during the solution growth: the first reduction by two orders of magnitude down to 106 cm2 is due to bundling of the dislocations during the first several microns of the LPE growth. After that, the bundled dislocations successively coalesce resulting in final dislocation densities of the order 104 cm2 after the crystal has grown to a thickness of over 1 mm. This mechanism works because growth rates in ‹0001› c-growth direction are much slower than growth rates in ‹1011› growth directions. C-faces develop faster than {1011} faces, and initial small GaN grains faceted with {1011} and {0001} become much bigger for increasing GaN layer thickness. When a 2-mm-thick GaN layer was grown on the 20 substrate, some areas have a dislocation density of the order of 102 cm2, while others of the order of 104 cm2. The high crystalline quality of the thick layer is confirmed by the X-ray rocking curve from the (0002) reflection, which has an FWHM of 28 arcsec. Na flux TG growth. Limited efforts have been applied so far for the development of the TG growth from Ga–Na flux. Aoki et al. (2004) studied Na flux

254 Growth of Bulk GaN Crystals

growth of GaN on seed by dissolving GaN feedstock in the higher-temperature region of the crucible and crystallizing on the seed in the lower-temperature region. GaN solubility in Ga–Na flux is very low, at 800  C. To increase GaN solubility, 17–24 mol.% of Li3N with respect to Ga were added to a flux with composition Ga03Na0.7. GaN powder was used as feedstock and small colorless transparent prismatic GaN single crystals prepared previously by the Na flux method were used as seeds. Crystal growth experiments were performed for 100–200 h under a TG condition from 800  C at the bottom of the crucible to 730–800  C at the top of the crucible. Only the growth of small spontaneously nucleated crystals at the cooler upper part of the crucible above the position of the seed was reported when the Ga–Na–Li3N flux was used. In some growth runs, seeds dissolved before solution became saturated, and there is no information about seeds in other runs. One of the tested solutions did not contain Ga. Only Na, Li3N, and NaN3 were used for the GaN dissolution and growth. A thin GaN layer was grown on the seed in this run. Growth rate in c-direction was about 0.2 mm?h1. This result is discussed later in the chapter. The main drawback of the TG growth from Na flux is the very low GaN solubility. It is necessary to find a way to increase this solubility in order to have reasonable nitrogen concentration in the solution for TG growth. NH3 dissolution. While molecular nitrogen requires a large amount of energy for N2 breakdown on the surface of the Ga melt and subsequent formation of GaN, ammonia reacts with Ga at reasonably low temperatures and atmospheric or below-than-atmospheric pressures. GaN forms by the following reaction 2Ga(l) þ2NH3(g) ¼ 2GaN(s) þ 3H2 (Balkas and Davis, 1996) with Gr ¼ 51 kJ  mol1 at 800  C. The equilibrium pressure PNH3 of this reaction is about 103 atm (100 Pa) at 1000  C. Synthesis of the high-purity GaN powder is conducted by flowing pure ammonia over the liquid Ga at temperatures in the range of 800–1000  C. To expand the simple synthesis of small GaN crystals or GaN polycrystalline material to the growth of large GaN crystals, one should take control on the nitrogen supersaturation in the Ga solution during NH3 dissolution. The use of NH3 does not change the GaN solubility in the molten gallium, and this solubility is very low at low pressure and at temperatures of 800–1100  C. One of the first successful attempts to accomplish epitaxial GaN growth

from solution on sapphire substrate using NH3 as a source of nitrogen was made by Logan and Thurmond (1972). Based on their experiments, the authors estimated the GaN solubility in Ga at 1150  C as 3  105 mole fraction. If the NH3 pressure is less than the equilibrium pressure of the reaction between ammonia and Ga at given T (100 Pa at 1000  C), no GaN is formed on the liquid Ga surface. With a small NH3 overpressure, GaN starts to form slowly. At temperatures higher than 950  C and at the proper PNH3, nitrogen may move inside the Ga melt from the Ga surface by convection and also by GaN dissolution faster than GaN forms at the surface. The authors determined balanced PNH3 – T parameters (T ¼ 1000  C, PNH3 ¼ 1.3  103 atm) when nitrogen is delivered through thin layer of Ga to the substrate and they achieved heteroepitaxial GaN growth. During 64 h, 300-mm-thick crystallites were grown on the sapphire substrate. The dissociation rate of GaN to Ga þ N2 at temperatures less than 1050  C is small compared to the growth rate. The addition of Bi to Ga decreased Ga activity and suppressed spontaneous GaN nucleation, which led to improved heteroepitaxial growth. Further development of the liquid-phase epitaxial GaN growth at low NH3 pressures was accomplished by another group (Meissner et al., 2005). In the literature, it is referred to as solution growth at lowpressure (LPSG) method. The goal of this development is to use NH3 as a source of nitrogen, to transport the dissolved nitrogen through a Ga-based melt and to deposit GaN on a big-area GaN substrate. The most challenging part in the development of this LPE type of growth is to keep nitrogen supersaturation in the solution below the level of the GaN spontaneous nucleation, and to obtain reasonable growth rates while maintaining high crystalline quality of the grown GaN. It is necessary to boost GaN solubility to be able to meet these requirements while dissolving NH3 in Ga. Hussy et al. (2008a) began to use compositions such as Ga75Ge25xAux and Ga75Ge25xAgx instead of pure Ga to enhance GaN solubility and to increase nitrogen concentration in the solution. The main parameter controlling the supersaturation in the solution is the NH3 partial pressure. The authors varied PNH3 between 1.5 and 3 mbar (150–300 Pa) at temperatures in the range of 930–1020  C. Hussy et al. (2008b) demonstrated a thin epitaxial GaN film grown on a sapphire substrate with MOCVD GaN film on top. T dislocation density

Growth of Bulk GaN Crystals

changed from a value of about 5  109 cm2 in the MOCVD seed down to 5  108 cm2 in the grown film. However, the LPSG method still suffers from parasitic GaN nucleation and very low growth rates. 3.06.3.2.3

Excited N 2 dissolution in Ga Plasma-assisted N2 dissolution. It was demonstrated that in microwave nitrogen plasma at 5 torr (670 Pa) and at temperature as low as 900  C an efficient GaN and even InN synthesis is possible (Angus et al., 1997). Similar results and growth of a thin GaN layer on sapphire substrate were reported by Ozawa et al. (2009) by using microwave plasma of the N2–H2 mixture at 7.5 torr (1000 Pa) and 700  C. The excited N2 gas was a very efficient source for the saturation of liquid Ga with atomic nitrogen and gave a GaN crust on the droplet surface. These works show that solubility of excited N2 increases significantly even at very low pressures. At the same time, the solubility of GaN in liquid Ga does not change, but remains very low at the typically applied temperatures. The Ga solution becomes supersaturated with nitrogen very fast, and GaN forms at the Ga surface or on the substrate, which is placed very close to the liquid–gas interface. High-voltage electric-field-enhanced N2 dissolution. An innovative approach to exploit excited nitrogen dissolution and to grow bulk GaN was reported by Ivantsov et al. (1997) and by Soukhoveev et al. (2001). This group demonstrated a 20-mm diameter and 15-mm-thick GaN ingot, and 2.50 diameter and up to 200-mm-thick GaN wafers from a Ga based melt at N2 pressure of 2 atm (0.2 MPa) and temperatures in the range of 800–1000  C. The ingot was grown for less than 10 h on a graphite seed and exhibited polycrystalline structure. Crystals grew spontaneously at the surface of the melt. Growth time varied from 5 to 24 h. The structural quality of the crystals was evaluated by X-ray measurements. Some areas of the crystals demonstrated singlecrystal structure, other areas presented a structure typical for textured material. The authors, in their papers, did not report the details of the growth setup neither explained what causes N2 dissolution in Ga at the applied pressures and temperatures. The explanation was given in the patent by Ivantzov et al. (2003). It is disclosed that N2 dissolves in the melt when a direct-current (DC) or alternatingcurrent (AC) electric field is applied near or at the melt surface. The strength of the applied electric field is about 200 V  cm1. Molecular nitrogen excited by the strong electrical field dissolves in

255

gallium and GaN growth takes place in the subsurface layer of the melt. For the continuous, uniform growth, the seed is rotated and pulled from the melt or pushed inside the melt. Different configurations of the electrodes are considered in the invention to vary electrical field in the vicinity of the growing crystal. A mixture of Ga with Bi was also used to optimize supersaturation in the solution. Growth of higher structural quality crystals requires a more accurate control of the supersaturation in the solution than it was achieved in the presented results. Unfortunately, the authors did not continue the development of this technique. Some features of this approach may be very useful when combined with other solution growth techniques. 3.06.3.2.4 Chemically enhanced nitrogen or GaN solubility

All previously discussed methods experience low GaN solubility in the solvent. This low solubility limits GaN growth rates in the solution. It would be very advantageous to find a solvent which can dissolve GaN and create a solution with high nitrogen and gallium concentration. Two general options may be considered for the solvent: Ga containing solvent or Ga free solvent. Ga-Li- or Ga-Li3N-based solvents. It is known that lithium nitride is the only alkali nitride and the only nitride among nitrides of group IA and IIA elements, which congruently melts below 850  C at ambient pressure. Li3N also creates binary nitrides with group IIIA elements such as Li3BN2, Li3GaN2, and Li3AlN2. Ternary Li3BN2 is used as a solvent catalyst in HP high-temperature synthesis and growth of cBN crystals. It is reasonable to suggest that Li3N in combination with Ga may be used for GaN growth as additive, which can increase the GaN solubility. Song et al. (2003) reported that 1–4-mm-size GaN platelet crystals were spontaneously grown from the solution formed from Ga and Li3N at N2 pressures as low as 2 atm (0.2 MPa) and temperatures of about 800  C. The grown crystals were colorless and transparent, with good crystalline quality as confirmed by X-ray measurements. To grow GaN crystals, the authors used solutions with Ga–Li3N molar ratio from 1:1 to 4:1. Crystals were grown by slow cooling from 800  C with rates of 2–3  C d1 for 120–180 h. To optimize the solution composition and P–T parameters of the growth, Wang et al. (2004) studied the Ga–Li–N ternary system using thermodynamic calculations. The Ga–Li–N phase diagram was calculated at 800  C and 1 bar (0.1 MPa) of gas

256 Growth of Bulk GaN Crystals

pressure. According to this phase diagram, GaN crystals can be grown from liquid phase in the area with reasonably high nitrogen concentration and without competition with another solid phase – ternary compound Li3GaN2. The authors proposed that at a certain temperature liquid Ga reacts with solid Li3N through the reaction Ga þ 2Li3N ¼ Li3GaN2 þ 3Li. Then Li3GaN2 dissolves in the Li–Ga melt to form a nitrogen-containing liquid phase in the Li–Ga–N ternary system. GaN crystals nucleate and grow from the melt as the temperature decreases and nitrogen supersaturation increases. Near-atmospheric pressure solution growth. As discussed above, the development and study of the spontaneous growth are only the initial stage for the development of the growth method. Large crystals suitable for commercial production of high-quality substrates require growth on a seed. In addition, industrial requirements in terms of crystal size and production cost make the growth at moderate temperature and pressure necessary. Feigelson et al. (2008) reported on a technical approach to dissolve a solid GaN source in the proper solvent to create a solution, and then to grow crystals on a seed from the supersaturated solution by applying a TG. The authors explored growth of GaN on the seed which was immersed in the solution from the top (top-seeded configuration), and also on the seed placed at the bottom of the crucible (bottom-seeded configuration). Axial TGs are opposite for each seed location. GaN source and solvent in the developed approach are created in situ (a)

during the first stage of the growth run. Initial components of the charge are Ga and a shaped piece of Li3N placed at the bottom of the crucible or on the top of Ga, depending on the seed location. Feigelson (2009, unpublished) found that at certain conditions Li3N exothermically reacts with Ga at temperatures around 680  C. The heat released by the reaction is estimated to be 8.1–8.6 kJ mol1. The product of this reaction is polycrystalline GaN, which forms in place of the Li3N and takes the original Li3N shape. Based on these findings the authors concluded that the exchange reaction Ga þ Li3N ¼ GaN þ Li takes place in the crucible at 680  C. As a result of the reaction, polycrystalline GaN feedstock is formed in place of Li3N. Li, released from Li3N, mixes with Ga and creates a solvent for the GaN (Figure 17(a)). It is possible to expect that during the reaction some amount of nitrogen is also released from Li3N and saturates the solution. GaN feedstock and solvent prepared via the reaction given above in the crucible are used to grow GaN on a seed by applying a TG. GaN has a positive temperature coefficient of the solubility in the formed solution. Feedstock of GaN dissolves at the higher temperature region and crystallizes at the lower temperature region. By the in situ prepared GaN feedstock and solvent, a single-crystal GaN layer was grown on the HVPE GaN seed, immersed in the solution from the top, at a temperature of 800  C and N2 pressure of 0.25 MPa (Figure 17(b)). GaN growth on HVPE GaN seed placed at the bottom of the crucible was also reported

(b) (f)

(d) N face 2 mm

Ga+Li+N GaN (c)

Grown layer

1 mm

Li3N

Immersed in solution

Ga

(g) (e)

100 μm

Figure 17 (a) Schematic of in situ GaN feedstock and solvent formation. Photographs of (b) HVPE seed partly immersed into solution with GaN grown layer; (c) grown GaN layer on HVPE GaN seed placed at the bottom of the crucible; (d) polycrystalline GaN seed with GaN crystals grown on immersed part; (e) near isotropic GaN crystals grown from solution; (f, g) GaN whiskers grown from solution.

Growth of Bulk GaN Crystals

(Figure 17(c)). To study the peculiarities of the growth on the seeds with different crystallographic orientation, polycrystalline GaN seeds were used as well. In this case, individual crystals were grown on each big crystallite of the polycrystalline seed (Figure 17(d)). Grown crystals are well shaped, transparent, and colorless. High crystalline quality of the grown crystals was confirmed by X-ray measurements. FWHM of about 16 arcsec was obtained for the (0004) rocking curve. It was shown that FWHM of X-ray rocking curve of the homoepitaxial GaN layer grown on HVPE seed has changed from 2.15 on the seed to 111 arcsec on the grown layer. It was also found that some additives to the solution dramatically change the ratio between growth rates in the c- and nonpolar directions. This last allows controlling the habit of the grown crystals and growing nearly isotropic crystals (Figure 17(e)) and GaN whiskers of millimeter length scale size (Figures 17(f) and 17(g)). Further development of this new method requires a thorough study of the GaN solubility in the Ga–Li-based solvents of different compositions. It also requires developing the right tools to induce and control proper thermal gradients and convection to improve nitrogen transport in the solution. Solution growth based on Ga free solvents. There are few reports in the literature regarding the possibility to grow GaN crystals from solution by dissolving GaN in a solvent which does not contain Ga. TG growth is used and the difference of the GaN solubility in the solution at different temperatures is the driving force for the GaN dissolution, transport, and crystallization. Aoki et al. (2004) studied the seeded growth of GaN from Na flux by the TG method. One of the tested solutions did not contain any Ga but just Na with Li3N and NaN3. GaN powder at the bottom of the crucible was used as feedstock and small GaN crystals spontaneously grown by Na flux method were used as seeds in the upper region of the crucible. A temperature of 800  C was applied at the bottom and 730  C at the top of the crucible under 10 MPa of N2 pressure for 200 h. A 20–25-mm GaN thick layer was epitaxially grown on seeds. Growth of GaN crystals also occurred by spontaneous nucleation at the cooler upper part of the crucible above the seed. Colorless transparent platelet single crystals (total mass 24 mg) were obtained by spontaneous nucleation. It was reported by Feigelson and Henry (2005) that it is possible to grow GaN crystals by using LiF– BaF2–Li3N salt-based solution. Sintered GaN powder

257

was used as feedstock and wurtzite-type rod-shaped GaN single crystals (0.5 mm length) elongated in [1102] r-direction were spontaneously grown using a 2.5-MPa N2 pressure and a temperature of 800  C. The use of Ga free solvents is still a very little explored area; and it is possible that this type of solvents might have certain advantages over traditional Ga-based solvents. In some way, Ga free solvents can be considered as a transition from Ga-based solvents, where nitrogen is the solute, to the supercritical ammonia-based solvent of the ammonothermal growth, where Ga becomes the solute, which is transported from the region of dissolution to the crystallization region. Ammonothermal growth. Supercritical water is known to be a good solvent (hydrothermal growth) for many inorganic compounds and it makes possible to grow crystals of many materials from their solutions in supercritical water. The ammonothermal growth technique for GaN bulk crystals was developed starting from analogies with the wellestablished hydrothermal growth technologies for quartz, emerald, and, more recently, for bulk ZnO. Ammonia (NH3) has many chemical analogies with water and exhibits a closer match to the physical– chemical properties of water than any other known solvent. As supercritical water is a polar solvent for oxides, the supercritical ammonia is a polar solvent for nitrides. Supercritical ammonia, as any other supercritical fluid, is a substance at a temperature and pressure above its critical point. Critical conditions for ammonia are Tc ¼ 132  C, Pc ¼ 112 bar (11.2 MPa). Supercritical ammonia-based solvents provide pathways to synthesize a range of compounds that are unstable in aqueous solvents. Wang and Callahan (2006) gives an excellent review on the chemistry of ammonothermal reactions, ammonothermal synthesis of nitrides, ternary amides, and ammonia adducts of aluminum and gallium. Different methods can be used to synthesize IIInitrides under ammonothermal conditions, but only two approaches are known to give well-crystallized III-nitrides at relatively low temperatures: (1) reaction of the pure metals of group IIIA with ammonia in the presence of alkali metals or acidic salts; and (2) thermal decomposition of ternary amides in melts of alkali metal amides. The knowledge of the formation and thermal stability of the ternary amides of IIIAgroup elements with alkali metals and the chemistry of ammonia adducts of group IIIA halides is essential for the development of ammonothermal crystal growth of III-nitrides, specifically GaN.

258 Growth of Bulk GaN Crystals

Ammonothermal research and development of GaN growth requires special autoclaves, which can withstand pressures of several kbar (1 kbar ¼ 1000 bar ¼ 0.1 GPa) at temperatures of 500–700  C. Typically, autoclaves are heated from outside. Pressure strongly depends on the filling the autoclave with ammonia and temperature inside the autoclave. Ammonia provides its own set of experimental and technical challenges and is much more difficult to handle than water. The general approach to grow GaN crystals by ammonothermal method exploits different GaN solubility in supercritical ammonia at different temperatures. GaN feedstock (or nutrient) dissolves in the solvent at one end of the autoclave kept at the dissolution temperature. The solvent, based on the supercritical ammonia, dissolves GaN and gives rise to soluble metastable Ga containing intermediate products. These intermediates are transported (chemical transport) to the crystallization end of the autoclave, where GaN has lower solubility at another temperature. As a result, GaN crystallizes on seeds or elsewhere from solutions, which becomes supersaturated in this region of the autoclave. It is worth mentioning that supercritical fluids have one order of magnitude lower viscosity than standard liquids. This is very beneficial for the transport of solutes from the dissolution region to the growth front of crystals. Solubility of GaN in the pure supercritical ammonia is very low. To increase solubility and to enhance the crystallization process, different additives, called mineralizers, may be added to the solvent. They play a crucial role in ammonothermal growth. Without mineralizers crystallization would be extremely slow or even impracticable. Ammonium NHþ cation and amide NH2 – 4 anion in ammonia (2NH3 ¼ NHþ 4 þ NH2) are exact analogs to hydronium H3Oþ cation and hydroxide OH– anion in water (2H2O ! H3Oþ þ OH). Mineralizers can be divided in two groups: ammonobasic and ammonoacidic mineralizers. Ammonobasic mineralizers create an ammonobasic environment by introducing an excess of amide NH2 anions, and ammonoacidic mineralizers create an ammonoacidic environment introducing an excess of ammonium NHþ 4 cations. Specific GaN growth features are very different in ammonobasic solution and ammonoacidic solution. Based on these differences, two different growth approaches were developed: ammonobasic and ammonoacidic. The ammonothermal growth has specific advantages over other GaN solution growth methods, regardless of the basic or acidic versions. Due to the

very low viscosity of the supercritical fluid, it is possible to create very uniform growth conditions in the big volume of the crystallization region of the autoclave. This in turn makes possible to use multiple seeds and to grow many large crystals in one single growth run. Practically, each crystal grows at near thermodynamic equilibrium with zero TG at the interface of the growing crystal. This gives the opportunity to grow many GaN crystals of high crystalline quality at once. There has been a continuous discussion on which of the two approaches, ammonobasic or ammonoacidic, is more suitable for GaN growth. Most groups and for a longer time have developed the ammonobasic growth technique. The present-day’s version of ammonothermal growth is more advanced than ammonoacidic. We make a step-by-step comparison of the features of both approaches to understand their technological differences and similarities. Ammonobasic version. Among the ammonobasic mineralizers are amides of alkali metals NaNH2, KNH2, LiNH2; alkali metals Li, Na, K, etc. and azides NaN3, KN3, and LiN3. They introduce NH2 ions to the solution. Whether amides, azides, or alkali metals are used as mineralizers in GaN crystal growth, the dissolution and crystallization mechanisms are all similar to each other. When alkali metals are used as mineralizers, they react with ammonia first to form amides and hydrogen, and then the amides subsequently attack the nitride nutrient. In the case of azides, such as KN3, NaN3, and LiN3, additional pressure will be generated due to azide breakdown when the temperature approaches the azide decomposition point. Amides are very sensitive to oxygen and moisture and require special handling precaution in preparation and storage to reduce the probability of oxygen contamination. Even storage in the drybox for a few weeks might cause oxygen contamination. Azides are less sensitive to the ambient and easier to handle and purify, so they may be preferable to amide because of lower solution-contamination risk. According to publications of different research groups (Wang et al., 2006; Hashimoto et al., 2008; Dwilinski et al., 2009b), amides are the most-used basic mineralizers. Ammonobasic solutions are to a certain extent chemically inert to the autoclave materials (Ni–Cr alloys). Autoclave walls are passivated after the first runs, and no further corrosion is observed afterward (Dwilinski et al., 2009a). This makes ammonobasic solutions more convenient to use than ammonoacidic

Growth of Bulk GaN Crystals

solutions which require Pt-liners to protect the autoclave from corrosion. The major characteristic of the supercritical solution with the basic mineralizer is nonstandard negative temperature coefficient of the GaN solubility (Dwilinski et al., 2003). The GaN solubility is higher at lower temperatures and lower at higher temperatures (Figure 18). Because of this phenomenon of the ammonobasic solution, the feedstock is placed in the LT dissolution zone and the hightemperature crystallization zone is located in the autoclave below the dissolution zone (Figure 19(a)). GaN solubility in the ammonobasic solution also depends on pressure (Figure 18(a)), specific basic mineralizer, and its concentration, and has always a negative temperature coefficient. The crucial part of the ammonothermal growth is the seed material. Like in the hydrothermal growth, the structural quality of the seed to a large extent determines the structural quality of the grown crystal. The majority of the groups developing GaN ammonothermal growth today exploit seeds prepared from HVPE GaN material. This choice is made because HVPE GaN wafers are the only seed material available for research in reasonable quantities. Unfortunately, there are no indications that ammonothermal GaN grown on GaN HVPE seeds improves the GaN structural quality of the seed, as it is observed in LPE Na flux growth. Such difference between solution growth techniques may be related to the different morphological development of the GaN crystals as a result of the substantially lower temperatures used in ammonothermal growth.

Today, it is assumed that it is necessary to have very high crystalline quality seeds to grow truly bulk GaN crystals of high crystallinity using the ammonothermal method. Until now, Dwilinski et al. (2009a) reported the best structural quality: 1.50 diameter ammonothermal crystals grown on native ammonothermal high structural quality seeds. The seeds were grown and enlarged by the group during repeated growth runs for internal use only. Seeds become a very serious issue in the development of the ammonothermal GaN growth method and future industrial production. The expansion of the crystal production requires the development of adequate amount of large area and high-quality seeds. Seeds for the ammonothermal growth can be grown from solution by any available method, but currently only crystals produced by ammonothermal technique can be used as a proper seed material. Ammonothermal GaN growth starts from the dissolution of the Ga containing nutrient. It dissolves into solution by the reaction with supercritical ammonia and mineralizer and forms the soluble intermediates. For example, if GaN is used as feedstock and potassium amide as mineralizer, the reaction looks like GaN þ KNH2 þ 2NH3 $ KGa(NH2)4. The equilibrium of this reaction in the LT dissolution zone is shifted to the right, producing soluble Ga-containing amide intermediates. Pure Ga or gallium-containing intermediates can be also used as feedstock material. When pure Ga is used as nutrient, the possible reaction 2Ga þ 2KNH2 þ 6NH3 ¼ 2KGa(NH2)4 þ 3H2 forms intermediates and at the same time it is possible

(a) Solubility (% gram GaN/gram NH3)

Solubility of GaN (mol.%)

4 T = 400°C 500°C 3

2

1

0

100

200 300 Pressure (MPa)

400

259

9

(b)

8 7 6 5 4 3 2 1 0 400 420 440 460 480 500 520 540 560 580 600 Average temperature of the autoclave (°C)

Figure 18 (a) Solubility of GaN in the NH3 þ KNH2solution (molar ratio KNH2:NH3 ¼ 0.07). (b) Retrograde solubility of GaN measured in NH3þKNH2solutions. KNH2concentration is about 3.5  0.5 M, the temperature difference 10  C cm1, pressure 1.2–2.4 kbar. (a) Figure courtesy of R. Dwilinski, AmmonoSp. z o.o. (b) Figure courtesy of B. Wang.

260 Growth of Bulk GaN Crystals

(a)

(b)

T1

Heater 1

T1

Nutrient transport

Nutrient transport

Feedstock and mineralizers Baffle Seeds Heater 2 T2

T2 > T 1

T2

Figure 19 Schematic view of GaN ammonothermal temperature gradient growth: (a) ammonobasic version; (b) ammonoacidic version. (a) Figure courtesy of R. Dwilinski, AmmonoSp. z o.o.

to observe another reaction 2Ga þ 2KNH2 þ 2NH3 ¼ 2KNH2 þ 2GaN þ 3H2, which creates GaN in the dissolution region. Both reactions have different rates, depending on temperature. This makes the control the crystal growth process complicated when using gallium metal as the feedstock. Hashimoto et al. (2008) reported that by using Ga as nutrient, a maximum growth rate of 88 mm d1 for 12-h growth was achieved; however, the growth rate dropped after this time. This occurred because during growth the metallic Ga nutrient transformed into GaN in the feedstock crucible. Once the metallic Ga transformed to GaN, the dissolution of Ga-containing species into NH3 significantly decreased, resulting in reduction of the growth rates. When amide intermediates are used as nutrient, an excessively high supersaturation may be reached because the intermediates are very soluble in ammonia. It is advantageous to have a porous medium and to increase the surface area of the GaN feedstock in contact with the solution for enhanced GaN dissolution. Small pieces of polycrystalline GaN or GaN grit are most commonly used as a nutrient. Pure GaN for the feedstock may be produced as a by-product of HVPE GaN growth or by any other method producing high-purity GaN material, for example, the reaction between gaseous ammonia and Ga. After the dissolution in the supercritical fluid, the Ga containing soluble species are transported from the dissolution region to the crystallization region by convection. Due to the low viscosity of the

supercritical fluids, axial and radial TGs create intense convection inside the autoclave. To optimize the flow of the solution in the autoclave and between dissolution and crystallization regions, a special baffle with different openings is placed between these two regions (Figure 19). The baffle helps to correct the dissolution rate and transport of the intermediates from nutrient to the crystallization zone, to adjust the supersaturation in the crystallization zone, and to make supersaturation more uniform along the surfaces of the growing crystals by optimizing the flow inside the crystallization region. The baffle also reduces particle flow (secondary nucleation). Mass and heat transfer computer modeling helps to optimize the geometrical parameters of the autoclave, the baffle, and the growth setup along with the heater design and their heat distribution outside the autoclave. With an external heating system, the autoclave walls are always hotter than the seeds. As a consequence of the GaN negative solubility coefficient, the supersaturation at the autoclave wall is always higher than in the vicinity of the seeds. The threshold supersaturation for the heterogeneous nucleation is not high, which makes the window for the acceptable supersaturation near seeds quite narrow, because otherwise the supersaturation at the wall would exceed the degree for which GaN starts to nucleate and grow on the wall of the autoclave. Parasitic crystals growing on the wall compete with crystals growing on seeds and lower the process yield.

Growth of Bulk GaN Crystals

Because of the retrograde GaN solubility, the chemical transport of GaN is directed from the LT dissolution region (upper half of autoclave) to the high-temperature crystallization region at the lower half of the autoclave, where GaN crystallizes on seeds. The flux of the foreign phases and impurities is going in the opposite direction from the hot zone, where standard solubility of all other than GaN materials in the autoclave is higher, to the cold zone where their solubility is lower. Such counterflow may be useful to reduce the concentration of impurities in the growing crystals (Dwilinski et al., 2009b). Ammonia as a solvent may dissociate at high temperatures due to the reaction 2NH3 $ N2 þ 3H2. Uncontrolled ammonia decomposition could limit the length of growth runs and deteriorate the quality of crystals (Wang and Callahan 2006). As the quantity of ammonia in the system decreases, the concentration of species dissolved in ammonia solution increases, increasing the supersaturation. According to the Le Chatelier’s principle, raising the operating pressure shifts the equilibrium and reduces ammonia breakdown. Approximately 8% of ammonia decomposes at 600  C and 2 kbar (0.2 GPa). Lowering the growth temperatures or raising the operating pressure would further reduce ammonia decomposition. However, it is always a question of

(a)

261

the compromise between the more comfortable growth conditions at higher temperatures and maximum pressure tolerated by the autoclave. Different groups developing ammonobasic growth exploit temperatures in the range 400–600 C and pressures in the range from 2 kbar (0.2 GPa) up to 5 kbar (0.5 GPa). There is another benefit of the retrograde GaN solubility. Maximum temperature in the autoclave is defined by the available pressure, which is limited by the materials and design of the autoclave, and also by the safety and the cost issues. The negative temperature coefficient of the GaN solubility allows to grow GaN crystals at the hottest part of the autoclave and to take advantage of the maximum temperature available in the system. Crystals grown at relatively higher temperatures benefit from the higher surface diffusion rates and tend to have lower dislocation density (Dwilinski et al., 2009b). It was demonstrated for the first time by Dwilinski et al. (2008) that ammonobasic solution growth (AMMONO-bulk method) allows to produce up to 20 diameter single-crystal GaN of very high structural quality. The grown crystals have nice hexagonal habit, are more than 10 mm thick, and allow to produce c-face substrates up to 20 in diameter and m-face substrates up 10  24 mm2 (Figure 20). It takes (b)

(c)

(d) 1

0

1

2

Figure 20 Photographs of (a) as-grown GaN crystal grown by the ammonobasic technique at Ammono Sp. z o.o.; (b) an epi-ready (0001) GaN wafer prepared from an as-grown crystal; (c) thick AMMONO-GaNcrystal; (b) typical m-plane AMMONO-GaNsubstrate made of such a crystal. Figures are courtesy of R. Dwilinski, AmmonoSp. z o.o.

262 Growth of Bulk GaN Crystals

approximately 60–80 days (20–30 mm d1) to grow such crystals. There is no more detailed information about growth rates of these crystals. Crystals are grown on multiple seeds, which are also produced by ammonobasic solution growth. The material is confirmed to be the pure hexagonal phase of GaN. The FWHM of the X-ray rocking curve on (0002) plane of the 10 wafer has a value as low as 16 arcsec. It reflects the outstanding crystalline quality of the material grown by the ammonobasic method. The investigated crystals exhibit exceptionally flat crystal lattice, which shows that the grown crystals are stress-free. For the best-quality samples, radius of curvature calculated from X-ray rocking curves measurements is greater than 1000 m (Dwilinski et al., 2009b). For comparison, the crystal-lattice curvature radius of GaN grown by HVPE is in the range of 10–12 m. Dislocation density measured by the selective etching technique is in the order of 103 cm2. Undoped AMMONO-GaN shows n-type conductivity with carrier concentration up to 1019 cm3 and resistivity  ¼ 103–102 cm. The developed AMMONO-bulk method allows to control the conductivity by proper doping. It is shown by Dwilinski et al. (2009b) that p-type material with carrier concentration up to 1018 cm3 and resistivity  ¼ 101–102 cm, and semi-insulating material with resistivity  ¼ 106–1012 cm can also be produced by ammonothermal growth. These breakthrough results of the AMMONObulk method are very encouraging for the future development of the GaN solution growth techniques and production of single-crystal GaN substrates. Ammonoacidic version. Examples of acidic mineralizers include ammonoacids NH4I, NH4Cl, and NH4F; alkali halides KI, LiF, and NaF; and hydrazine hydrochloride NH2NH3Cl. These mineralizers introduce NHþ 4 ions into the solution. Ammonoacids NH4X (X ¼ Cl, Br, and I) are the most exploited in the GaN growth development. It is considered (Ehrentraut et al., 2008a) that acidic mineralizers are easier to handle than basic mineralizers due to their sufficient chemical stability. Ammonoacidic solutions cause corrosion of the HP autoclaves. They require the insertion of a protective platinum (Pt) inner liner, which must be firmly attached to the inner wall of the autoclave to avoid any leaking. Ammonobasic solutions do not require such protection. The solubility of GaN in supercritical NH3 with NH4X (X ¼ Cl, Br, and I) mineralizers was studied by Ehrentraut et al. (2008b). Contrary to

ammonobasic solutions, GaN in ammonoacidic solutions exhibits a regular positive temperature coefficient of the solubility. It means that the nutrient dissolves in the hotter dissolution region of the autoclave (at the bottom) and the Ga-containing metastable phase is transported to the LT crystallization region where it decomposes to deliver the Ga–N species to the growing crystal (Figure 19(b)). The authors determined that in the temperature range 250–600  C the log solubility is a linear function of reciprocal temperature in the Arrhenius plot. The GaN concentration in the solution normalized to 1 mol of the mineralizer NH4Cl is around 2.0 at 550  C. The activation energy of 15.9 kcal  mol1 was calculated from the slope of the linear fitting of the plot. The authors also found that GaN solubility increases with the increase of the molar ammonoacid concentration in the range 0.0127–0.032–0.127. The retrograde solubility has both advantages and disadvantages over the forward solubility. Ehrentraut et al. (2008a) consider that the main benefit of the forward solubility is that of avoiding an undersaturated state in the solution around the grown crystal during the slow cooling after growth. In ammonobasic solution, the GaN solubility increases during the cooling phase, which can lead to some surface dissolution of the grown crystal. From another point of view, this phenomenon may be considered as an advantage of the retrograde solubility, because it eliminates the formation of the thin nonuniform GaN film, which grows fast during slow cooling due to the supersaturation increases with the decreasing temperature in case of regular GaN solubility. In comparison with ammonobasic solutions, growth with acidic mineralizers is conducted at reduced temperature (<550  C) and pressure (<150 MPa). Typical conditions reported by Ehrentraut et al. (2008c) are T ¼ 425/525  C for the upper/lower zones of the autoclave and P ¼ 145 MPa. Like in ammonobasic growth, Ga or GaN may be used as nutrients in ammonoacidic growth. For short growth runs, shorter than 24 h, Ga is preferred because it dissolves faster. It leads to higher supersaturation in the crystallization region and higher growth rates. In contrast, for long growth runs, GaN with grain size in the low millimeter range was found to be a better choice for the feedstock, because of its slower dissolution rate. A mixed Ga/GaN feedstock may also be used to adjust growth rates.

Growth of Bulk GaN Crystals

It was found by Purdy (1999) that with the acidic mineralizers GaN can crystallize as cubic phase. Ehrentraut et al. (2008b) reported that increasing the acidity, that is, exchanging Cl with Br and I in NH4X leads to an increase of the lattice distortion in the grown GaN. The fraction of cubic GaN phase is highest when using NH4I and also increases toward lower process temperatures for all three NH4X (X ¼ Cl, Br, and I) mineralizers. On the contrary, GaN grown from ammonobasic solution shows always a pure hexagonal phase. All ammonoacidic growth experiments were conducted with HVPE GaN seeds. It revealed that better crystalline quality seed material is required to grow ammonoacidic GaN crystals of high crystalline quality. X-ray rocking curve measurements of 0.2-mm-thick ammonothermal GaN crystal grown on HVPE GaN substrate with growth rates of 20 mm d1 were presented in the paper by Ehrentraut et al. (2008a). These measurements revealed that the FWHM from (0004) reflection for the Ga (N)-terminated faces was 610 (480) and 310 (330) arcsec, respectively. The first 20 in diameter and 0.5-mmthick GaN crystal was grown on GaN HVPE substrate by acidic ammonothermal method. Ehrentraut et al. (2008c) describe the crystal quality still as unsatisfactory. HP ammonothermal growth. There is another approach to the ammonothermal growth development. D’Evelyn et al. (2004) have succeeded in growing millimeter-sized spontaneously nucleated GaN crystals and crystals on HVPE seeds by the TG HP ammonothermal method. In order to access more extreme conditions and accelerate the growth rate, this group utilized a sealed capsule placed in an HP cell with a solid pressure medium of the type that is well known in belt-press synthesis of diamond at HP and high temperature. HVPE GaN seeds, polycrystalline GaN feedstock, a mineralizer (unfortunately, it was not disclosed), and ammonia were sealed into a capsule. The capsule was incorporated into an HP cell capable of producing a TG and placed in a pressure apparatus. Crystals were grown at temperatures between 600 and 1000  C, the pressure was estimated as lying between about 5 and 20 kbar. GaN crystals grown by this HP ammonothermal method were transparent, yellow in color, and well faceted. Spontaneously nucleated crystals were of millimeter size. A transparent 10  13 mm crystal was grown on a clear HVPE GaN seed. The biggest crystal was 15  18 mm in diameter, nontransparent as a result of nontransparency of the seed. On

263

the macro-scale, the Ga surface was very smooth and the N surface somewhat rougher, which is contrary to observations in common Na flux growth (Aoki et al., 2002c). The dislocation density in the layer grown in the c-direction with respect to the HVPE seed was in the 106–107 cm2 range, similar to that observed in HVPE GaN seeds. GaN grown laterally from the seed in the [1120] and [1010] directions did not propagate dislocations from the seed, so that the dislocation density in these areas was below 100 cm2. Unfortunately, the authors did not provide data on growth rates or duration of the growth runs. Altogether, this work demonstrates the possible path in ammonothermal growth to increase growth rates. HP apparatuses with solid pressure medium most likely are not suitable equipment for bulk GaN industrial production. If special autoclaves with a working pressure up to 1 GPa were available, growth rates of GaN ammonothermal growth could be increased significantly. 3.06.3.2.5 Technical challenges and conclusion

Many technical aspects of the presented techniques are not extensively discussed because of the limited space of the chapter and also because part of the confidential know-how, which is a major part of the whole picture. Growth of bulk GaN from solution requires extreme conditions. One needs to find a balance between available resources and direction of the development. Any direction actually offers a serious challenge. HP methods require great efforts and resources for the development of special high gas pressure equipment and experimental procedures. Any increase of the reaction volume involves design and development of the new HP reactor. Solvents, which can be used for GaN growth at near-atmospheric pressure, are very aggressive, and materials for the crucible and reactor become daunting problems. The combination of pressure and chemistry, which allows preparing solution for GaN growth at intermediate pressures, raises many additional questions. All techniques deal with complex issues of oxygen contamination. The challenging nature of the GaN crystals requires systematic interdependent development of all components for any GaN solution growth technique in order to produce large high crystalline quality GaN crystals. A good illustration of such an approach has been demonstrated in the development of AMMONObulk growth. This method is the most advanced solution growth technique today. It is the result of

264 Growth of Bulk GaN Crystals

continuous research efforts during almost 20 years. These efforts included the development and test of several generations of custom-designed autoclaves, thousands of experiments to collect necessary feedback data for optimization of the growth parameters. The multistep preparation of the growth run had to be also developed to achieve the reported results. To reveal its real potential, any promising GaN solution growth technique will require efforts comparable to those applied for the AMMONO-bulk growth development.

3.06.4 Bulk GaN Growth from Gas Phase Vapor-phase techniques are generally grouped in two main categories: physical-vapor deposition (PVD) and chemical-vapor deposition (CVD). The PVD processes typically involve either a direct or a reactive evaporation of the source, achieved by heating the source or by sputtering of a target to provide the source material into the gas phase. The CVD processes typically include a reaction of the source material in gas phase in order to produce a desired material deposit. 3.06.4.1

Physical Vapor Transport

In a typical physical vapor transport (PVT) for nitrides, the nutrient material in form of powder is sublimed from the source region within a closed or semi-open crucible following a simple reaction: MN ! M þ 1=2N2

where MN identifies the nitride compound, with M the metal and N the nitrogen. The vapors are subsequently transported by a nitrogen flow at subatmospheric pressure to a seed held at a lower temperature than the source, where they recrystallize. 3.06.4.1.1

Sublimation Sublimation growth, a type of PVT growth of crystals, has several advantages over melt growth in that it allows lower growth temperatures, morphological stability, and relatively easy implementation. This process is a viable low-pressure growth that ensures high purity of the resulting material and, even more important, it provides truly high growth rates (>500 mm h1). The method was successfully utilized

for AlN growth due to congruent evaporation of this material at high temperatures providing nearly stoichiometric composition of the vapor phase. The first experiments have been started about three decades ago (Slack and McNelly, 1976, 1977). The pressure of the main species (Al and N2) over the AlN surface reaches 1 atm only at 2260  C, which is the reason why the temperature range of 1950–2250  C is normally used for sublimation growth of AlN. As a result, crystals up to 10 in diameter (Makarov, 2006) or up to 20 with usable area of 85% (Schujman et al., 2008) were recently reported. It should be mentioned, however, that due to the high reactivity of aluminum vapor at high sublimation temperatures, serious problems such as the need for high-purity AlN source and stability of the crucible for AlN growth still remain open. This method cannot be applied to GaN due to the excessively high vapor pressure of N over Ga. In contrast to the case of AlN, the nitrogen pressure over GaN under practical growth conditions (temperatures 1000  C) lies between 10 and 100 bar. Vaporization of GaN at high temperatures is congruent only under Langmuir conditions (free evaporation in vacuum) (Munir and Searcy, 1965). At higher pressures (i.e., closer to equilibrium conditions), evaporation of GaN becomes noncongruent and is accompanied by liquid Ga accumulation in the source. As a result, molecular nitrogen dominates in the vapor phase. However, because of extremely low chemical reactivity of N2 on GaN surfaces, growth of GaN under these conditions does not occur. Hence, external flow of nitrogen-containing reactive species is required to carry out the growth process. The best choice of such species in all respects is ammonia. The first free-standing GaN crystals of considerable size have been obtained by growth from the vapor phase (Zetterstrom, 1970). In this approach, synthesized GaN powder was heated in dry NH3 flow at 1150–1200  C for 24 h. This resulted in growth of GaN crystals with needle shape and size of about 1 mm. The growth of GaN occurs via sintering although the contribution of the PVT seems to be also important under these conditions. Growth of GaN by flowing ammonia over liquid gallium was carried out at 1000–1150  C in an open or partly closed quartz ampoule (Ejder, 1975; Elwell et al., 1984). In the latter case, NH3 was supplied to a boat with liquid Ga through a small orifice of 1 mm in diameter that provided diffusion-limited growth conditions. GaN was deposited on the walls of the ampoule in the form of whiskers, needles, platelets,

Growth of Bulk GaN Crystals

etc. It was concluded (Ejder, 1974) that the GaN crystals had been grown from the vapor phase containing atomic gallium and ammonia. In addition, it was also shown that raising the growth temperature made it possible to increase the size of grown GaN crystals (Matsumoto et al., 1974). Therefore, attempts to grow free-standing GaN crystals by the Ga–NH3 system without seed or substrate were not successful. In the best case, needle or platelet crystals were obtained but their lateral size was very far from desirable. An alternative was to prepare the source by coldpressing of GaN powder at pressures varied within the range of 1575–2200 kg cm2 (Balkas et al., 2000). This allowed a densification of the source, which could not be achieved via reactive Ga–NH3 synthesis. The growth was carried out in ammonia atmosphere at pressures close to 1 atm. The source temperature varied from 1100 to 1400  C, while the seed temperature was kept at about 1200  C. Under these conditions, GaN growth rates about 0.5 mm h1 were achieved and good-quality GaN crystals were obtained. In order to enhance the nucleation different approaches have been suggested, intentional introduction of impurity or use of different seeds. It was found that the introduction of Bi or Sn impurities into the solution increased the number of crystallites nuclei (Elwell et al., 1984). In particular, prismatic GaN crystals appeared in the reactor at a temperature of 825–1010  C under NH3 pressures of 0.3–2.5 mbar. The largest crystals (up to 2.5 mm long and 1 mm in diameter) have been formed on SiC or GaN seeds. It was observed that the crystals of plate-like habits were formed preferably at relatively low growth temperatures, while needle-shaped crystals grew at higher temperatures. A different scheme of growth process was proposed later and become known as sublimation sandwich technique (SST) (Vodakov et al., 1980; Vodakov and Mokhov, 2002). In this method, the source of Ga vapor and the substrate were separated by a distance of several millimeters forming a growth cell. GaN crystals are grown in a tubular quartz reactor with radio-frequency (RF) heating. In this growth process, the growth cell is placed inside the reactor on an RF-graphite susceptor while an ammonia flow passes over the substrate. Both horizontal and vertical reactor configurations were used. The horizontal configuration provided a more effective supply of ammonia in the growth cell, while the vertical one was more suitable for longer process. Both the Ga source and the substrate were placed

265

in the growth cell, forming a sandwich where transport of the main growth species (Ga and NH3) occurs. A TG zone was created in the growth cell by setting a certain temperature difference T between the source and the substrate. This temperature difference serves as a driving force for Ga transport inside the growth cell and GaN crystallization on the substrate. Normally, the distance between the source and the substrate is in the range 2–5 mm (Vodakov et al., 1997). A mixture of liquid Ga and GaN powder was usually used as a source of gallium. The experiments showed that a use of pure Ga ensured the maximum growth rates but resulted in poorer temporal and spatial stability of source operation. On the contrary, a use of pure GaN powder leaded to partial decomposition of the source material into liquid Ga and vapor N2 phases which usually took place since the beginning of the growth process. The most promising approach was found by employing a mixture of liquid gallium and GaN powder. 3.06.4.1.2

Molecular beam epitaxy A detailed description of the molecular beam epitaxy (MBE) method and its different modifications is subject of Chapter 3.12. For sake of completeness, we just include a short comment of nontraditional usage of this method for producing free-standing quasi-bulk GaN material for specific substrate application, namely producing nonpolar cubic GaN material of substantial thickness. Majority of wurtzite-type GaN films reported by MBE have been grown on basal plane of sapphire substrates. The growth optimizations aiming higher structural quality and smooth surfaces have been employing different approaches, that is, LT buffers, MOCVD template layers, N- or Ga-rich conditions, and different nitrogen pressures. It is understood that the mode of nucleation and growth of epitaxial films is strongly governed by the interfacial free energies between the deposit and the substrates, the deposit and the vapor, and the vapor and the substrate. The layer-by layer growth mode occurs for zero interfacial free energy between the deposit and the substrate. Thus, after a relatively smooth buffer layer is achieved the growth of the layer could be carried out at elevated temperature in order to promote high surface mobility of the adsorbed adatoms, but not too high in order to avoid a roughening transition. In case of using (001) Si substrates, the deposited GaN films possess zincblende structure (Lei et al.,

266 Growth of Bulk GaN Crystals

1991; Basu et al., 1994; Foxon et al., 2008). This approach has faced severe challenges over the years because the GaN films tend to form both wurtzite and cubic phases. Achieving a single-phase cubic GaN material has always been a challenge. However, recently, the interest in zincblende GaN has rapidly been increasing due to several reasons: the absence of spontaneous and piezoelectric polarization fields in (100) GaN, the ability to cleave cubic (100) GaN along the perpendicular {110} cleavage planes for device fabrication, and enhanced mobility of holes in the p-type cubic GaN, approaching the best values reported for wurtzite material. Attempts to grow (100) cubic GaN have been made by MBE, HVPE, and MOCVD techniques by employing (001) Si, or cubic GaAs, or cubic SiC. The thickest crack-free cubic GaN layers up to 60 mm so far have been reported by PA-MBE (Foxon, 2008). Undoped thick layers were grown on semi-insulating GaAs (001) substrates, using arsenic as a surfactant to initiate the growth, followed by a deposition of GaAs buffer in order to improve the properties of the cubic GaN material. This type of thick layers can afterwards be easily separated from the substrates by chemical etching of the GaAs substrates, thus producing free-standing cubic GaN for substrate applications.

3.06.4.2

Hydride vapor-phase epitaxy

basis. Nowadays, the only commercially available GaN substrates are produced by the HVPE process. These epitaxial techniques allow high growth rates and thus the deposition of very thick layers (from about 300 mm to few millimeters), which are subsequently separated from the original substrate giving the so-called free-standing GaN wafers. An accurate thermodynamical analysis of the HVPE growth process requires the postulation of the chemical reaction pathway. All practical methods for synthesis of GaN at high temperature utilize an active source of nitrogen, such as ammonia, NH3, and a vapor source for Ga. In the HVPE growth, the Ga atoms can be obtained by letting a Ga source to react with pure HCl at about 800  C to form GaCl according to the following reaction:

Thermodynamic

xGað1Þ þ HClðgÞ ! xGaClðgÞ þ ð1 – xÞHClðgÞ þ x=2H2 ðgÞ

GaClðgÞ þ NH3 ðgÞ ! GaNðsÞ þ HClðgÞ þ H2 ðgÞ ð2Þ

where l ¼ liquid, g ¼ gas, s ¼ solid, and x ¼ the mole fraction of HCl reacting in the process. The GaN deposition is determined by the efficiency of both chemical reactions. Values of x in the reaction (1) were found to be in the range from 0.70 to 0.86, depending on the temperature, the position of the HCl inlet, the carrier gas ambient, and the extension of liquid Ga surface exposed to the HCl gas as well. The chemical reaction (2) depends on the fraction of ammonia not decomposed into nitrogen and hydrogen, because GaN cannot be formed by direct reaction between GaCl and N2. It is known that ammonia is a thermodynamically unstable gas at the temperatures usually employed in the GaN growth. Fortunately, the thermal decomposition of NH3 is a very slow reaction, and when no catalyst is present, no more than about 4% of the NH3 is typically decomposed at temperatures higher than 950  C (the deposition temperature of GaN lies in the 950–1150  C range). The main formation mechanism of GaN given by Equations (1) and (2) is accompanied by GaN decomposition in the sense that the reaction (2) may be reversed (etching of the growing crystal) and also thermal decomposition may take place: GANðsÞ þ HClðgÞ ! GaClðgÞ þ 1=2N2 ðgÞ þ 1=2H2 ðgÞ ð3Þ

Chemical Vapor Transport

3.06.4.2.1

the GaCl is then transferred to the growth zone to react with pure ammonia in order to give GaN according to

ð1Þ

GaNðsÞ ! GaðlÞ þ 1=2N2 ðgÞ

ð4Þ

However, in the temperature regime of interest (950–1150  C), reactions (3) and (4) are practically inhibited. The main gaseous species involved in the HVPE growth of GaN are GaCl, GaCl3, HCl, NH3, H2, and the carrier gas. Thermodynamical functions of the material, the enthalpy H(T), entropy S(T), and specific heat Cp(T), of all the species involved in the growth have been theoretically calculated (Przhevalskii, 1998) for the standard pressure of 1 atm and arbitrary temperature using a polynomial approximation of the Gibbs free energy. Detailed thermodynamical calculations of HVPE growth of GaN have been published by several authors (Ban, 1972; Seifert et al., 1983; Koukitu et al., 1998; Aujol et al., 2001). Based on their analysis, several important features can be pointed out, which make the gallium nitride growth different from the growth of other

Growth of Bulk GaN Crystals

III–V materials, and should be taken into account for successful HVPE-GaN growth: 1. An inert ambient atmosphere is needed and is more effective for the GaN growth than hydrogen atmosphere, the latter being widely used for other III–V systems, like GaAs. 2. In HVPE-GaN growth, NH3 should be used as the source of the group V element rather than a nitrogen halide (NCl3), which is highly explosive. Moreover, the thermal dissociation of NH3 results in the formation of N2 molecules which are extremely stable and nonreactive at the temperatures used, while the thermal dissociation of AsCl3, in GaAs growth, results in the formation of As2 and As4 molecules, which typically remain volatile and chemically reactive in the next step of the growth reaction. 3. Another problem comes from the fact that the GaN reaction tends to create huge amounts

267

of by-products such as NH3Cl, GaCl3, and GaCl3?NH3, leading to a massive condensation of these reaction species on the reactor walls. This fact makes the development of last-running process for very thick samples quite difficult. Reactor design. Typically, the HVPE growth of GaN layers is carried out in reactors based on the first concept reported over 30 years ago (Maruska and Tjetjen, 1969). There are different modifications and they can be summarized into two groups: horizontal and vertical reactor design (Paskova and Monemar, 2003). The horizontal reactor typically has five main temperature zones (Figure 21(a)). In the first upstream zone HCl reacts with metallic Ga forming GaCl and H2. The area of the liquid Ga source is increased as much as possible (typically 10–100 cm2) in order to achieve a large reactive Ga surface area for efficient GaCl production. The optimum

NH3/N2 N2 HCl/N2

(a)

Substrate

Thermocouple Closing hatch Quartz tubes (b)

N2

Ga-boat

Gas outlet

RF-coil

Substrate

Ga-boat

Quartz tubes

Gas inlet HCl/N2 NH3/N2 NH3/N2 Figure 21 Schematic drawing of: (a) horizontal reactor with five temperature zones; and (b) inverted vertical reactor with two temperature zones.

268 Growth of Bulk GaN Crystals

temperature in the first zone is about 850  C, but some authors give values between 820 and 900  C (Fremund, 1981). The second zone may be used for other metallic sources such as In or Al when needed, or for dopants. In the third zone, maintained in the range of temperatures 1000–1060  C, GaCl and NH3 are introduced and mixed. The substrate holder is placed in the fourth region of the reactor. We note that the reported substrate temperatures vary considerably, between 950 (Ohki et al., 1982) and up to 1150  C (Crouch et al., 1978) but some of the best-quality GaN films were reported as achieved at 1050–1080  C (Molnar et al., 1995). The most common method of heating is resistive, although RF-heating is also quite popular for some of the reactor zones. The horizontal reactors utilize a substrate holder that have a variety of designs and is situated parallel, or inclined, or perpendicular to the gas flow direction. More uniform growth can be achieved by tilting of the substrate holder to eliminate reactant depletion along the flow direction. Another approach used in some horizontal reactors is rotation of the substrate holder utilizing a gas foil rotation technique or other mechanisms. In most reported reactor designs, the active gases are delivered to the mixing point through separated parallel quartz liners. An alternative is the coaxial arrangement of the gas inlet tubes (Seifert et al., 1983; Safvi et al., 1996) to achieve a better mixing of the reaction gases prior to the deposition and to improve the uniformity of the layers. In this case, it was recommended to maintain a certain distance between the gas liner inlets and the substrate in order to avoid a polycrystalline patch in the centre of the wafer (Safvi et al., 1996). In the vertical design, the reactants are typically introduced through the top. The substrate is held flat on a susceptor that is perpendicular to the gas flow direction. The vertical reactor design usually facilitates easer substrate rotation during the growth to improve the film uniformity. Heating is accomplished by resistance or RF induction heating and temperature monitoring is accomplished by an infrared pyrometer or a thermocouple. Another important modification (Molnar et al., 1995) of the vertical reactor design is in an implementation of a technique of lowering the substrate holder isothermally into a dump tube, in which there is a counterflow of NH3 in a N2 carrier gas. This allows an abrupt interruption of the growth, and then either change and equilibrate the gas flows in the main tube, or slowly cool down the sample by further lowering the holder.

An alternative modification is an inverted vertical reactor where the process gases are supplied through the bottom inlet flange, while the top flange can be lifted for loading and unloading of substrates (Figure 21(b)). In this configuration, the substrates are placed in the upper part where the gases are mixed. The inverted reactor keeps all advantages of the vertical design and also provides the possibility of raising the substrate holder. An additional advantage of the inverted vertical reactor is the minimization of solid particle contamination of the growing surface. Growth kinetics. The kinetics of the process and growth mechanisms occurring at the solid/vapor interface during HVPE growth of GaN was simulated by several models (Cadoret, 1999; Aujol et al., 2001) assuming a surface process involving the following steps: (1) adsorption of NH3 molecules, (2) adsorption of N atoms coming from ammonia decomposition, (3) adsorption of GaCl molecules on the N atoms forming NGaCl, and (4) decomposition of the NGaCl via different desorption mechanisms (Figure 22). Two of them were suggested in analogy with the GaAs model: desorption forming HCl and desorption forming GaCl3 (Cadoret, 1999). One more GaCl2 desorption mechanism was suggested for the GaN growth based on the experimental results (Trassoudaine et al., 2001). Statistical treatment of the dynamic equilibrium between the adsorbed and gas-phase species allowed explicit expressions of the growth rates via the different pathways. An optimum HVPE growth process requires a good selection of the reactor geometry and operating conditions, to ensure a minimization of undesired parasitic reactions, and to provide a uniform reactant distribution across the substrate. Optimization of HVPE growth is typically done by empirical studies of external parameters such as temperature, flow rates of active gases, and substrate orientation. One of the most important questions in understanding the HVPE GaN growth mechanism is the kinetics of HVPE GaN growth process. There are several detailed kinetics considerations published in the literature (Seifert et al., 1983; Shintani and Minagawa, 1974; Liu and Stevenson, 1978; Paskova et al., 1999a) even in the initial stages of development of the technique. The regularities established in the fundamental dependence of growth rate upon the temperature generally agree among different researchers, although the reported details are partly in contradiction to each other. These studies have identified regions with a different character,

Growth of Bulk GaN Crystals

(iv) Desorption

GaCl3 mechanism

H2 mechanism Adsorption

H

Cl

Cl

Cl

Ga

Ga

Ga

(ii)

GaCl2 mechanism

Cl

Ga

H

(iii)

(i) NH3

Cl

Cl

Ga

Ga

Cl

Cl

Ga

Ga

Cl

Ga Cl

Cl

269

Cl

Ga

Ga

Ga

Cl

Ga

V N

N

N

N

N

N

N

N

N

N

N

N

N

N

Figure 22 Schematic drawing of possible reaction steps. Adsorption (i–NH3 adsorption, ii –N adsorption, iii –GaCl adsorption) and desorption (H2 desorption (Cadoret, 1999), GaCl3 desorption (Cadoret, 1999) and GaCl2 desorption (Trassoudaine et al., 2001)) mechanisms of the surface growth process of HVPE growth of GaN.

Temperature (°C)

Growth rate (μm min–1)

1200

1000

800

600 a-plane, He * c-plane, He * a-plane, H2 * a-plane, N2 c-plane, N2

10

1

0.1

0.7

0.9 1.0 1.1 × 10–3 0.8 Reciprocal temperature (K–1)

Figure 23 Growth temperature dependence of the GaNgrowth rate (Seifert et al., 1983).

illustrated in Figure 23. The growth rate increases with increasing temperature at low temperatures, after which the growth rate reaches a maximum value, and a further increase of the growth temperature results in a decrease in the growth rate. The LT region. The exponential dependence of the growth rate in the LT region is a characteristic of kinetically controlled growth and may be attributed to the surface reaction, adsorption, and surface diffusion. The monotonic increase of the growth rate with substrate temperature leads to a calculation of the activation energy of the epitaxial GaN growth from the Arrhenius plot. The reported values in the literature vary between 0.9 and 1.6 eV (21 and 37 kcal mol1) and the magnitude of the value suggests that

the main rate-determining step in this LT region is related to the surface reaction (Shintani and Minagawa, 1974; Liu and Stevenson, 1978). A thermodynamic calculation of the equilibrium partial pressures in this region allowed a proposal of a dominant mechanism for surface processes in different ambient (Seifert et al., 1983). The high-temperature region has been the subject of extensive experimental studies because of the larger practical interest and better material quality ensured. The growth rate reaches its maximum value at a temperature in the wide range of 950–1080  C, according to different reports (Molnar et al., 1995; Shintani and Minagawa, 1974; Liu and Stevenson, 1978). The difference in the temperature value for

270 Growth of Bulk GaN Crystals

104 a-plane, L = 50 mm a-plane, L = 52 mm a-plane, L = 60 mm c-plane, L = 50 mm

120

Tgr = 972 °C * * Tgr = 1080 °C, FHCl = 25ml min–1

Growth rate (μm h–1)

Growth rate (μm h–1)

160

80

40

0

0

10

20 30 40 HCl flow rate (ml min–1)

50

Figure 24 The dependence of GaN growth rate on the HCl flow rate. The growth temperature was 1080  C. The distance between the NH3outlet and the substrate is given in the inset, as well as the sapphire orientation.

maximum growth rate may be attributed to differences in reactor design and growth conditions. At the temperature of 1080  C, the dependence of GaN growth rate on HCl flow rate is shown in Figure 24. The growth rate increases linearly as a function of both HCl flow rate and GaCl partial pressure (Liu and Stevenson, 1978; Paskova et al., 1999). This fact combined with the weak temperature dependence of the growth rate in the higher growth temperature region indicates that the growth rate is limited by the transport of the gallium-containing species to the growth surface. The distance of the substrate from the gas mixing point also appeared to strongly affect the growth rate in this region. An exponential decrease of the growth rate with increasing distance from the mixing point at constant values of growth temperature and carrier gas flow can be seen in Figure 25 (Shintani and Minagawa, 1974; Paskova et al., 1999). This effect was discussed in detail from a chemical kinetics standpoint by (Shintani and Minagawa, 1974) and (Safvi et al., 1996). It is important to point out that increasing the mixing space can improve the uniformity of the growth rate along the substrate (Safvi et al., 1996) and can be helpful to avoid polycrystalline deposition; however, there is an optimum distance beyond which the crystal quality of the material deteriorates (Paskova et al., 1999). While the HCl flow rate and the distance of the substrate from the mixing point appeared to be the key growth parameters, other parameters such as

Tgr = 1090 °C, FHCl = 35ml min–1

103

102

10

1

0

50 100 Distance from mixing point (mm)

150

Figure 25 GaN growth rate vs. the substrate position from the gas mixing point (Shintani and Minagawa, 1974).

the NH3 flow rate or the substrate orientation do not seem to affect the growth rate in a critical way, although they can affect the layer quality to a large extent. The weak growth rate dependence on NH3 flow rate is reasonable, since the concentrations typically used in GaN HVPE is far larger than that of HCl, in order to suppress GaN dissociation. The growth rate of GaN deposited on the two commonly used sapphire orientations <1120> and <0001> was found to be different with a ratio of about 1.5 (Figure 24) (Paskova et al., 1999; Shintani and Minagawa, 1974). Studies (Seifert et al., 1983) of the sapphire orientation effect on the temperature dependence of the growth rate revealed a significant difference mainly in the range of low growth temperatures. In addition, the different carrier gases used were found to have different effect in the range of low growth temperatures, namely, the growth rate in case of using hydrogen was higher than that in inert gas carrier gas for the temperature interval from 800 to 1050  C (Figure 23). It was concluded that it was indicative of a growth rate limiting surface reaction involving hydrogen (Dam et al., 2006). The growthrate-determining steps in the high-temperature region, which are typically thought to be related to enhanced gas-phase pre-reactions leading to a depletion of the chemical reactants at the substrate surface and/or enhanced etching of the growth surface, are likely to involve the hydrogen as a critical component controlling the process chemistry. There are also some experimental findings which were observed to have a beneficial role in improving the film quality, although they are not fully clarified

Growth of Bulk GaN Crystals

UCSB - exp* Kyma - EPD+

1010

Kyma - CL+

Dislocation density (cm–2)

from the thermochemistry point of view (Jacob et al., 1978; Molnar, 1999), showing that additional HCl gas intentionally introduced downstream of the Ga boat can change the Cl/Ga ratio at the growth surface and thus enhance the lateral/vertical growth ratio, resulting in both a decrease of growth rate and an improvement of film quality. Main growth approaches. The optimization of the HVPE-GaN growth process requires detailed consideration of its multidisciplinary nature and complex multiparameter character. The latter obstacle is really very complicated because the large number of the growth parameters, typically of geometrical, thermal, chemical, and thermodynamical nature, has to be optimized at the same time, often finding a compromise between opposing requirements. The need for optimal nucleation is another issue that complicates the HVPE growth of GaN when initiating the growth on a foreign substrate. Several approaches gave good results, such as different single buffers (reactively sputtered ZnO; Detchprohm et al., 1992; Molnar et al., 1997), MOCVD-grown GaN thin layers (Paskova et al., 1999b, 2000b), reactively sputtered AlN (Paskova et al., 1999b, 2001), or TiN (Oshima et al., 2002), and also different modifications of growth forced in lateral in-plane directions (as ELOG around masks of different materials and shapes; Usui et al., 1997; Parillaud et al., 1999; Sone et al., 1999; Davis et al., 2002), or grooved substrates. The used nucleation scheme may significantly influence the characteristics of the HVPE GaN layers as well as their evolvement with GaN thickness. The dislocation density dependence on the thickness (Figure 26) is of particular practical interest. It is important to note that the growth process (including nucleation scheme, the growth rate, and growth recipes) as well as the material properties resulting from using different foreign substrates and optimizations are completely different for thin films with thickness from 1 to 50 mm (often called templates), for thick films in the range of 100–2000 mm, and for boule growth with thickness larger than several millimeters, the latter two being of interest for substrate applications after separation from the foreign substrate. Detailed description of the characteristics of the HVPE-grown GaN available on the market are not included in this review, but there are several thorough reviews in the literature (Molnar et al., 1997; Morkoc, 2001; Gibart et al., 2003; Paskova and Monemar, 2003; Paskova and Evans, 2009) that discuss the properties of the HVPE-grown GaN in

271

Kyma - CL m-plane++

109

MCC - CL#

- - - UCSB - model* 108

107

106

105

0.1

1

10

100

1000

10 000

Thickness (μm)

Figure 26 Dislocation density as a function of the thickness of GaN layers. (Mathis et al., 2000); þ(Hanser, 2004); þþ(Paskova and Evans, 2009); #(Fujimori, 2008).

detail, and that we would recommend to interested readers. In the following section, we only focus on the growth approaches developed over the years to achieve high-quality GaN layers of thickness varied from 1 mm to 10 mm. GaN template grown by HVPE. GaN templates is the term typically used for layers deposited and remaining on foreign substrates (usually sapphire), with thickness in the range of 1–150 mm, that can be used for direct growth of device structures. The advantages of this approach in the device growth is based on the fact that these template substrates eliminate the need for the LT buffer growth step and, even more important, the dislocation density in these layers is noticeably reduced down to 108–107 cm2, depending on the thickness. This ensures low defect density and higher structural quality in the subsequently grown multilayer structures of devices. The challenge in using these templates, however, lies in their inherent bowing, which increases with the layer thickness and negatively affects the following device technological step such as the photolithographic processing. In contrast to the MOCVD nitride technology, the early work on LT GaN buffers in the same chamber was not successful in the HVPE technology. Alternatively, buffers of different nature, deposited by different techniques, mostly reactive sputtering, have been proposed and demonstrated to ensure significant improvement of the crystal quality of the layers with thickness up to several tens micrometers without cracks. The ZnO buffer was first suggested as a buffer for HVPE growth of thick GaN films by

272 Growth of Bulk GaN Crystals

(Detchprohm et al., 1992), and was further developed and optimized at Lincoln laboratory (Molnar et al., 1997). It is known to be one of the most effective buffers for growth of thick GaN layers of very good quality, but thick large-area crack-free layers suitable for quasi-substrates applications have not been demonstrated by this buffer. Two more buffers grown typically at high growth temperature with single-crystalline charter were later suggested by Paskova et al. (1999b) and successfully employed in the HVPE growth to avoid the highly conductive region at the interface due to columnar structures in case of direct growth on sapphire (Figure 27). Reactively sputtered thin AlN buffer with thickness of about 2 mm have been proven to ensure smooth epilayer surface and low defect density GaN material (Valcheva et al., 2002). The use of MOCVD templates for GaN growth on sapphire has been first shown in MBE growth of thin GaN layers. Later on, it was successfully employed in the HVPEGaN growth and is currently used by many research groups. The report showed that the MOCVD GaN templates can provide good buffers for growth of thick HVPE GaN layers with very good structural and optical characteristics, although it is difficult to grow films with sufficient thickness without cracks. Single substrate growth by HVPE. In order to serve as good buffer for subsequent growth of thick HVPEGaN films for substrate applications, a specific

(a)

requirement needs to be fulfilled. Namely, the interface region between substrate and GaN layer has to be easily destroyed by a suitable lift-off process. There are several approaches for separation of the thick GaN layer from the substrate. The most popular is the self-separation, which can be achieved by the so-called void-assisted separation, which in turn requires a special buffer. Such a buffer could be reactively sputtered TiN, as suggested by Japanese researchers in 2002 (Oshima et al., 2002). This approach was developed and successfully implemented into production by Hitachi Cable, which recently announced the production of crack-free 30 freestanding GaN wafer (Figure 28(a)) (Oshima et al., 2007). Another high-temperature reactively sputtered AlN buffer was also found to be effective in suppressing defect formation in the GaN nucleation region (Hanser et al., 2004). This buffer approach also permits reducing of tensile stress during GaN growth by forming a weak interface which leads to optimized structural properties of films with thickness up to 1–2 mm (Hanser et al., 2007). In addition, the weak interface allows for the self-separation of the films. This approach was successfully developed by Kyma Technologies, which announced the production of free-standing GaN wafers of different size (Figure 28(b)) (Hanser et al., 2007). Crack-free thick HVPE GaN layers with thicknesses of about 300–500 mm have been demonstrated

(b)

HVPE GaN

HVPE GaN AlN

20 μm

Sapphire

(c)

HVPE GaN

10 μm

20 μm

Sapphire

(d)

,MOCVD GaN

Sapphire

HVPE GaN

10 μm

MOCVD GaN:Si

Sapphire

Figure 27 Panchromatic CL images of cross sections of GaN template layers residing on sapphire substrates grown: (a) without a buffer; (b) with reactively sputtered AlN buffer; (c ) with MOCVD grown undoped GaN buffer; (d) with MOCVD grown GaN doped with Si. From (Paskova et al., 1999b)).

Growth of Bulk GaN Crystals

(a)

(b)

Figure 28 (a) Currently the largest 30 GaN substrates announced by Hitachi Cable (CompoundSemiconductor.net) and (b) GaN substrates, currently available on the market, of different size and conductivity announced by Kyma Technologies.

by using the ELOG technique on 20 sapphire and on GaAs substrates. This growth procedure (Usui et al., 1997) consists principally of a HVPE selective homoepitaxial growth on a thin MOCVD-grown GaN layer. The MOCVD growth was performed using the conventional two-step procedure, consisting of a 20-nm-thick LT buffer layer and a hightemperature main layer. A SiO2 layer was then deposited by CVD and window stripes with a period of about 7 mm were opened using conventional photolithographic techniques. Alternatives of using W (Sone et al., 1999) and SiN (Davis et al., 2002) masks instead of SiO2, or using GaAs (Motoki et al., 2002), LiGaO2 (Kryliouk et al., 1999), and LiAlO2 (Richter et al., 2006) substrates instead of sapphire have also been demonstrated. In addition, in order to further reduce the dislocation density, the ELOG approach can be performed by the so-called twostep approach, which can employ either a second layer of mask usually with shifted period with respect to the first mask, or a change of the growth parameters, usually a reduced growth temperature (Gibart et al., 2003). These types of thick layers can be separated from the substrates by different methods, depending on the type of the substrate used and the

273

nucleation scheme. The easiest substrate removal is through a chemical etching as often done in the case of GaAs or LiGaO2 substrate. This approach is now routinely used at Sumitomo Electric Industries Ltd for removing the initial GaAs substrate. In the case of transparent sapphire substrates, a laserinduced lift-off process was developed in order to separate GaN from the sapphire using pulsed UV laser beams. The process works by irradiating the sapphire/GaN interface with intense laser pulses just at the absorption edge of GaN. It leads to a fast and strong local heating which causes thermal decomposition in that interface region of the film, yielding metallic Ga and nitrogen gas evaporation. This approach was particularly developed for twostep-ELOG nucleation and successfully used in wafer production by Lumilog Ltd. (Gibart et al., 2003). Boule growth by HVPE. The use of MOCVD templates for GaN growth on sapphire has been implemented by many research groups. Their results consistently showed that the MOCVD GaN templates can provide good buffers for growth of thick HVPE GaN layers with very good structural and optical characteristics, although it is difficult to grow films with sufficient thickness without cracks. However, when growing boules with thickness of several millimeters, the cracks predominantly occur in the sapphire, while the boule remains safe. Mitsubishi Chemical Co. reported the growth of 20-boules up to 5–6-mm thickness by using such a buffer approach (Figure 29(a); Fujito et al., 2008; Fujimori, 2008). Two-inch boule growth ranging from 2.5 up to 10 mm has been reported also by ATMI Inc., (Vaudo et al., 2002), Ferdinand-Braun-Institute (Weyers et al., 2008), and Kyma Technologies Inc. (Figure 29(b); Hanser et al., 2004; Paskova and Evans, 2009), respectively. Different buffers and high growth rates up to 300 mm h1 are likely to be used by the different groups. The boule growth is the most preferable and economical pathway to achieve GaN substrates with the necessary quality, size, and cost. At this time, however, GaN boule growth is still in its early stages of development and no mature technology is available yet. The boule growth approach is also the technique of choice for producing substrates with surfaces different from the (0001) for growth of device structures on crystallographic planes which exhibit lower or 0 polarization (the so-called nonpolar and semipolar planes). This is so far accomplished by

274 Growth of Bulk GaN Crystals

(a)

(b)

Figure 29 Currently, the largest 20 GaN boules reported by: (a) Mitsubishi Chemical and (b) Kyma Technologies. (a) Courtesy of Fujito.

slicing nonpolar or semipolar substrates perpendicularly or inclined with respect to the c-plane of the boules. Seeded regrowth. Another approach that recently attracted significant attention is the so-called seeded regrowth by the same HVPE process. This approach aims at combining the high growth rate and higher crystalline quality achievable by the homoepitaxial regrowth employing GaN seeds instead of using sapphire and also to avoid the initial highly defective interface region. Development of bulk regrowth on GaN seeds by HVPE has been somewhat limited, as the general growth focus has been more on the generation of large-area GaN single wafer production. Early work has demonstrated trends in growth and identified the main regrowth issues. Using HVPE regrowth, enlargement of GaN seed crystals by HVPE homoepitaxial regrowth has also been demonstrated in the past few years (Grzegory et al., 2006; Paskova et al., 2000). The preliminary HVPE regrowth results show that the material produced during regrowth possesses significantly

improved structural characteristics. In particular, the dislocation density was found to be noticeably reduced (Paskova et al., 2000). Another important aspect of seeded growth is the expansion of the seed crystal. In HVPE growth of GaN, crystal expansion can be controlled through physical–chemical gradients and concentrations, which result in the stabilization of crystal planes under the appropriate growth conditions. For inplane expansion, size expansion will depend on the anisotropy of the growth rates for different crystallographic planes. Growth parameters, such as the temperature, pressure, growth rate (source species flux), or the ratio between the nitrogen and the gallium species, can influence the stabilization of one crystal plane over another. Additionally, the geometric configuration of the seed in the system may contribute to the flux of the growth species from additional sources, such as surface migration. As shown in the ELOG growth of GaN (Hiramatsu, 2001), inclined crystalline facets can form under certain growth conditions that give undesirable crystalline morphologies. An important part of the growth process development is to identify growth parameters that give a planar (as opposed to prismatic) morphology, and then find the way to extend the growth in directions normal to those edge planes. One issue that has been observed with seed expansion is the variation in impurity content and carrier concentration. The physical and electrical properties of GaN grown by HVPE are very sensitive to the orientation of the crystallization front. This is a challenge for bulk growth, particularly for undoped and semi-insulating material, where a low background concentration of shallow impurities, such as oxygen, is desired. The crystalline quality of the regrown material as compared to the crystalline quality of the seeds is of particular importance. The analysis of the defect density in the regrown HVPE-GaN with different thicknesses has shown that the trend of decreasing the defect density with increasing thickness remains (Paskova et al., 2000). There is a clear improvement of the surface morphology in all overgrown layers as revealed by atomic force microscopy (AFM) imaging. Figures 30(a) and 30(b) show two images of a 40-mm-thick GaN layer as-grown on sapphire and of a regrown layer with a thickness of about 150 mm on the same free-standing layer separated from the sapphire. The surface of the as-grown layer on sapphire is quite smooth consisting of an array of terraces with a width of about 120 nm separated by 5–6-A˚ steps.

Growth of Bulk GaN Crystals

20 nm

0 (a)

3.0

20 nm

0

3.0

(b)

2.0

2.0

1.0

1.0

0.0 0.0

1.0

2.0

3.0

275

0.0

1.0

2.0

0.0 3.0 μm

Figure 30 AFM images of as-grown GaN film directly on sapphire (a) and regrown film on free-standing GaN(b).

However, one can also see that some of the steps are terminated tangentially at the edges of surface depressions due to dislocations of mixed type. The overgrown film on the free-standing GaN film has a lower surface roughness compared to that of the asgrown layer. No terrace depressions were observed and the density of dislocations intersected the surface is 2 times lower, indicating an annihilation of threading dislocations. A PL spectrum of as-grown 40-mm-thick GaN layer on sapphire, together with a spectrum of the same film after separation and two spectra of the regrown GaN films (up to thicknesses of 100 and 200 mm), is shown in Figure 31(a). The narrow (2 meV) exciton peaks in the spectrum of the asgrown film are fully reproduced in the overgrown films, indicating their good crystalline quality. These rather small variations in the PL peak (a)

positions correspond to a small variation in the corresponding lattice parameters in the seed and in the regrown films, and reflect the defect distribution in the material (both dislocations and point defects). The stress in the overgrown GaN layers was analyzed by Raman scattering measurements. Three Raman spectra in the region of the E2 mode are shown in Figure 31(b). The solid lines represent calculated spectra derived from a fitting assuming a Lorentzian line shape. The peak positions of the asgrown, the separated, and the overgrown films are observed to be very close to each other, indicating a very small change in the strain. The frequency shift to higher wave numbers, with respect to the reference value of 566.2 cm1 (Kisielowski et al., 1996) for homoepitaxial and bulk strain-free material, verifies the presence of some small compressive strain in the (b) z(xx)z E2 (high) Intensity (a.u.)

Intensity (a.u.)

Overgrown (200 μm) Overgrown (100 μm) Free-standing (40 μm)

DBE ABE

FA

Free-stand.

FB

As-grown (40 μm)

Overgrown

As-grown 3.40

3.42

3.44

3.46 3.48 Energy (eV)

3.50

3.52

555

560

565 570 Raman shift (cm–1)

575

580

Figure 31 (a) Low-temperature (2K) photoluminescence spectra, and (b) Raman scattering spectra of HVPE-GaN films: as-grown; free-standing; and overgrown HVPE-GaN films on the free-standing GaN films.

276 Growth of Bulk GaN Crystals

layers, in agreement with the assignment of the PL line shift. Based on both PL and RS experimental data, it is clear that both the seeds and the regrown crystals are under a small compressive strain and that the separation from the substrate does not lead to a complete removal of the stress. The residual strain along the c-axis was studied by high-resolution XRD. The FWHM values of the 2–! rocking curves were found to be less than 46 arcsec for all samples. The linewidths of the (002) reflection in the overgrown layers showed a small broadening which increases with increasing thickness together with some asymmetry indicative of a small stress gradient in these thicker regrown films (Paskova et al., 2000). The lattice constants measurements show the highest lattice constant c of 5.1857 A˚ and a of 3.1878 A˚ in the as-grown layers, while the free-standing and regrown layers have both parameters smaller than the ideal parameters c ¼ 5.1856 A˚ and a ¼ 3.188 A˚, reported for strain-free material. The strain in the seeds and in the regrown material is mainly due to residual impurities and native defects. The residual stress in thick GaN layers was tentatively attributed to the dislocationrelated inhomogeneously distributed stress along the growth axis. Since the regrown layers replicate mainly the defect distribution and the stress field of the underlying films, the defect and stress control in the regrown HVPE-GaN films are of great importance. In summary, the overall goal of seeded regrowth is to optimize the growth process to achieve large-area, high-quality bulk crystals for manufacturing GaN substrates. By doing so, this will result in a number of manufacturing benefits beyond the current heteroepitaxially based GaN substrate growth process. As explained above, the process optimization requires a control over the bulk crystal morphology, in order to achieve well-controlled material and electrical properties, and conditions that lead to possible seed crystal expansion.

3.06.5 Summary In contrast to classical semiconductors, such as silicon (Si) and gallium arsenide (GaAs), for which the development of substrates and thin-film device structure went hand in hand, the III-nitride thin-film device technology is by far advanced over the development of native substrate crystal growth. The bulk

nitride growth is still seriously underdeveloped, but there has been recently a considerable interest and the ongoing research activity may change this situation. Among several approaches such as different modifications of solution growth or sublimation growth, the HVPE has already demonstrated the production of 20 wafers of reasonable quality. Recent advances in solution growth, above all in ammonobasic solution growth, showed the feasibility to produce GaN crystals of the highest crystalline quality and size of 20. In any semiconductor technology there are two features which can make it successful, namely the lower fabrication costs using viable growth conditions and/or convincing material quality enhancements. Until now, the HVPE technique did not have a real competitor and seems to satisfy both of them: bulk GaN of sufficient size and quality has been demonstrated while using a relatively simple and efficient process. However, there are still many questions to be answered, problems to be solved, modifications of the reactors, and growth procedures to be optimized in order to establish a robust industrial manufacture technology for GaN. It is not clear which technique or which combination of techniques will finally be optimal for nitride industrial production; however, the present large activity for the development of bulk nitrides and some recent results lead to think that some important breakthrough will be obtained in a near future. (See Chapters 3.05, 3.07 and 3.10).

References Angus JC, Argoitia A, et al. (1997) Growth of Bulk, Polycrystalline Gallium and Indium Nitride at Sub-Atmospheric Pressures. San Francisco, CA: Materials Research Society. Aoki M, Yamane H, et al. (2000) Growth of GaN single crystals from a Na–Ga melt at 750 degrees C and 5MPa of N2. Journal of Crystal Growth 218(1): 7. Aoki M, Yamane H, Shimada M, Sarayama S, and DiSalvo FJ (2002a) GaN single crystal growth using high-purity Na as a flux. Journal of Crystal Growth 242: 70–76. Aoki M, Yamane H, et al. (2002b) Conditions for seeded growth of GaN crystals by the Na flux method. Materials Letters 56(5): 660–664. Aoki M, Yamane H, et al. (2002c) Morphology and polarity of GaN single crystals synthesized by the Na flux method. Crystal Growth and Design 2(1): 55–58. Aoki M, Yamane H, et al. (2003) Dissolution and recrystallization of GaN in molten Na. Japanese Journal of Applied Physics: Part 1 (Regular Papers, Short Notes and Review Papers) 42(12): 7272–7275. Aoki M, Yamane H, Shimada A, Sarayama S, Iwata H, and DiSalvo FJ (2004) Single crystal growth of GaN by the temperature gradient Na flux method. Journal of Crystal Growth 266: 461–466.

Growth of Bulk GaN Crystals Aujol E, Napierala J, Trassoudaine A, Gil-Lafon E, and Cadoret R (2001) Hydrogen and nitrogen ambient effects on epitaxial growth of GaN by hydride vapor phase epitaxy. Journal of Crystal Growth 222: 538–548. Baik KH, Irokawa Y, Kim J, et al. (2003) 160-A bulk GaN Schottky diode array. Applied Physics Letters 83: 3192. Balkas CM and Davis RE (1996) Synthesis routes and characterization of high-purity, single-phase gallium nitride powders. Journal of the American Ceramic Society 79(9): 2309–2312. Balkas CM, Zitar Z, Bergman L, et al. (2000) Growth and characterizations of GaN single crystals. Journal of Crystal Growth 208: 100. Ban VS (1972) Mass spectrometric studies of vapor phase crystal growth: II. GaN. Journal of Electrochemical Society 119: 761–765. Basu SN, Lei T, and Moustakas TD (1994) Microstructure of GaN films deposited on (001) and (111) Si substrates using electron-cyclotron-resonance assisted molecular beam epitaxy. Journal Materials Research 9: 2370. Bockowski M (2001) Growth and doping of GaN and AlN single crystals under high pressure of nitrogen. Crystal Research and Technology 36(8–10): 771–787. Bockowski M, Grzegory I, Krukowski S, et al. (2002) Directional crystallization of GaN on high-pressure solution grown substrates by growth from solution and HVPE. Journal of Crystal Growth 246: 194–206. Bockowski M, Grzegory I, Krukowski S, et al. (2004) Deposition of bulk GaN from solution in gallium under high N2 pressure on silicon carbide and sapphire substrates. Journal of Crystal Growth 270: 409–419. Bockowski M, Strak P, Grzegory I, Lucznik B, and Porowski S (2008) GaN crystallization by the high-pressure solution growth method on HVPE bulk seed. Journal of Crystal Growth 310: 3924–3933. Cadoret R (1999) Growth mechanisms of (00.1)GaN substrates in the hydride vapor-phase method: Surface diffusion, spiral growth, H2 and GaCl3 mechanisms. Journal of Crystal Growth 205: 123–135. Chakraborty A, Baker TB, Haskell BA, et al. (2005a) Milliwatt power blue InGaN/GaN light-emitting diodes on semipolar GaN templates. Japanese Journal of Applied Physics 44(30): L945–L947. Chakraborty A, Haskell BA, Masui H, et al. (2006) Nonpolar m-plane blue-light-emitting diode lamps with output powerof 23.5mW under pulsed operation. Japanese Journal of Applied Physics 45(2A): L739–L741. Chakraborty A, Keller S, Meier C, et al. (2005b) Properties of nonpolar a-plane InGaN/GaN multiple quantum wells grown on lateral epitaxially overgrown a-plane GaN. Applied Physics Letters 86: 031901. Chen C, Adivarahan V, Yang J, Shatalov M, Kuokstis E, and Kahn MA (2003) Ultraviolet light emitting diodes using non-polar a-plane GaN–AlGaN multiple quantum wells. Japanese Journal of Applied Physics 42(9): L11039–L11040. Crouch RK, Debnam WJ, and Fripp AL (1978) Properties of GaN grown on sapphire substrates. Journal of Materials Science 13: 2358. D’Evelyn MP, Narang KJ, Park D-S, et al. (2004) Growth and characterization of bulk GaN crystals at high pressure and high temperature. In: Min Ng H, Wraback M, Hiramatsu K, and Grandjean N (eds.) In: Materials Research Society Symposium Proceedings, vol. 798, 275. Warrendale, PA: Materials Research Society. Dam CEC, Hageman PR, and Larsen PK (2006) Carrier gas and position effects on GaN growth in a horizontal HVPE reactor: An experimental and numerical study. Journal of Crystal Growth 205: 123–135.

277

Davis RF, Roskowski AM, Preble EA, et al. (2002) Gallium nitride materials – progress, status and potential roadblocks. Proceedings of IEEE 90(6): 993–1004. Detchprohm T, Hiramatsu K, Amano H, and Akasaki I (1992) Hydride vapor phase epitaxial growth of a high quality GaN films using a ZnO buffer layers. Applied Physics Letters 61(22): 2688. Dwilinski R, Doradzinski R, et al. (2008) Excellent crystallinity of truly bulk ammonothermal GaN. Journal of Crystal Growth 310(17): 3911–3916. Dwilinski R, Doradzinski R, et al. (2009a) Homoepitaxy on bulk ammonothermal GaN. Journal of Crystal Growth 311(10): 3058–3062. Dwilinski R, Doradzinski R, et al. (2009b) Bulk ammonothermal GaN. Journal of Crystal Growth 311(10): 3015–3018. Dwilinski RT, Doradzinski RM, Garczynski JS, Sierzputowski LP, and Kanbara Y (2003) Bulk Monocrystalline Gallium Nitride. US Patent No. 6,656,615.B2, 2 December 2003. Ehrentraut D, Kagamitani Y, et al. (2008a) Ammonothermal synthesis of thick gallium nitride film employing acidic mineralizers. Journal of Materials Science 43(7): 2270–2275. Ehrentraut D, Kagamitani Y, et al. (2008b) Physico-chemical features of the acid ammonothermal growth of GaN. Journal of Crystal Growth 310(5): 891–895. Ehrentraut D, Kagamitani Y, et al. (2008c) Reviewing recent developments in the acid ammonothermal crystal growth of gallium nitride. Journal of Crystal Growth 310(17): 3902–3906. Ejder E (1975) Photoconductivity of Zn-doped GaN. Journal of Electrochemical Society 36: 289–292. Elwell D, Feigelson RS, Simkins MM, and Tiller WA (1984) Crystal growth of GaN by reaction between gallium and ammonia. Journal of Crystal Growth 66: 45. Feigelson BN and Henry RL (2005) Growth of GaN crystals from molten solution with Ga free solvent using a temperature gradient. Journal of Crystal Growth 281(1): 5–10. Feigelson BN, Frazier RM, et al. (2008) Seeded growth of GaN single crystals from solution at near atmospheric pressure. Journal of Crystal Growth 310(17): 3934–3940. Figge S, Bo¨ttcher T, Dennemarck J, et al. (2005) Optoelectronic devices on bulk GaN. Journal of Crystal Growth 281: 101–106. Foxon CT, Novikov SV, Stanton NM, Campion RP, and Kent AJ (2008) Free-standing zinc-blende (cubic) GaN substrates grown by a molecular beam epitaxy process. Physica Status Solidi (b) 245: 890–892. Fremund R, Cerny P, Kohout J, Rosicka V, and Burger A (1981) Optimized growth conditions of GaN epitaxial layers. Crystal Res. and Technol. 16: 1257–1266. Fujimori T (2008) HVPE and ammonothermal GaN substrates for high performance device. In: Plenary Talk at IW on Nitride Semiconductors. Montreux, Switzerland, 6–10 October. Fujito K, Kiyomi K, Mochizuki T, et al. (2008) High-quality nonpolar m-plane GaN substrate grown by HVPE. Physica Status Solidi (a) 5(5): 1056–1059. Gibart P, Beaumont B, and Vennegues P (2003) Epitaxial lateral overgrowth of GaN. In: Ruterana P, Albrecht M, and Neugebauer J (eds.) Nitride Semiconductors, Handbook on Materials and Devices, ch. 2, pp. 45–106, Berlin: Wiley-VCH. Grzegory I (2001a) High pressure growth of bulk GaN from solutions in gallium. Journal of Physics: Condensed Matter 13(32): 6875–6892. Grzegory I (2002b) High-pressure crystallization of GaN for electronic applications. Journal of Physics: Condensed Matter 14: 11055. Grzegory I, Bockowski M, Łucznik B, et al. (2002a) Mechanisms of crystallization of bulk GaN from the solution under high N2 pressure. Journal of Crystal Growth 246: 177–186.

278 Growth of Bulk GaN Crystals Grzegory I, Bockowski M, Lucznik B, et al. (2001b) Seeded growth of GaN at high N2 pressure on (0001) polar surface of GaN single crystalline substrates. Materials Science in Semi-Conductor Processing 4: 535–541. Grzegory I, Jun J, Bockowski M, et al. (1995) III–V nitrides: thermodynamics and crystal growth at high N2 pressure. Journal of the Physics and Chemistry of Solids 56(3–4): 639–647. Grzegory I, Łucznik B, Boc´kowski M, et al. (2006) Growth of bulk GaN by HVPE on pressure grown seeds. Proceedings of SPIE 6121: 612107. Hanser D, Liu L, Preble EA, Thomas D, and Williams M (2004) Growth and fabrication of 2 inch free-standing GaN substrates via the boule growth method. In: Materials Research Society Symposium Proceedings, vol. 798, Y2.1.1. Warrendale, PA: Materials Research Society. Hanser D, Tutor M, Preble E, et al. (2007) Surface preparation of substrates from bulk GaN crystals. Journal of Crystal Growth 305: 372–376. Hashimoto T, Wu F, et al. (2008) Status and perspectives of the ammonothermal growth of GaN substrates. Journal of Crystal Growth 310(5): 876–880. Hiramatsu K (2001) Epitaxial lateral overgrowth techniques used in group III nitride epitaxy. J. Phys.: condens. Matter 13: 6961–6975. Hsu JWP, Manfra MJ, Molnar RJ, Heying B, and Speck JS (2002) Direct imaging of reverse-bias leakage through pure screw dislocations in GaN films grown by molecular beam epitaxy on GaN templates. Applied Physics Letters 81: 79. Hussy S, Berwian P, et al. (2008a) On the influence of solution density on the formation of macroscopic defects in the liquid phase epitaxy of GaN. Journal of Crystal Growth 311(1): 62–65. Hussy S, Meissner E, et al. (2008b) Low-pressure solution growth (LPSG) of GaN templates with diameters up to 3 inch. Journal of Crystal Growth 310(4): 738–747. Inoue T, Seki Y, Oda O, Kurai S, Yamada Y, and Taguchi T (2000) Growth of bulk GaN single crystals by the pressure controlled solution growth method. Japanese Journal of Applied Physics, Part 1 (Regular Papers, Short Notes and Review Papers) 39(4B): 2394–2398. Inoue T, Seki Y, Oda O, Kurai S, Yamada Y, and Taguchi T (2001a) Growth of bulk GaN single crystals by the pressure-controlled solution growth method. Journal of Crystal Growth 229: 35–40. Inoue T, Seki Y, Oda O, Kurai S, Yamada Y, and Taguchi T (2001b) Pressure-controlled solution growth of bulk GaN crystals under high pressure. Physica Status Solidi (b) 223(1): 15–27. Iso K, Yamada H, Hirasawa H, et al. (2007) High brightness blue InGaN/GaN light emitting diode on nonpolar m-plane bulk GaN substrate. Japanese Journal of Applied Physics 46: L960. Ivantzov V, Sukhoveev V, and Dmitriev V (2003) Method of Manufacturing GaN Ingots. US Patent 6,562,124, 13 May. Ivantsov VA, Sukhoveev VA, et al. (1997) GaN Crystals Grown from a Liquid Phase at Reduced Pressure. San Francisco, CA: Materials Research Society. Jacob G, Boulou M, and Bois D (1978) GaN electroluminescent devices: Preparation and studies. Journal of Luminescence 17: 263–282. Karpinski J, Jun J, and Porowski S (1984) Equilibrium pressure of N2 over GaN and high pressure solution growth of GaN. Journal of Crystal Growth 66(1): 1–10. Kawahara M, Kawamura F, et al. (2007) A first-principles investigation on the mechanism of nitrogen dissolution in the Na flux method. Journal of Applied Physics 101(6): 66106. Kawamura F, Imade M, et al. (2009a) Growth of high-quality large GaN crystal by Na flux LPE method. Proceedings of the

SPIE – The International Society for Optical Engineering 7216: 72160B (13pp.). Kawamura F, Iwahashi T, et al. (2003) Growth of transparent, large size GaN single crystal with low dislocations using Ca–Na flux system. Japanese Journal of Applied Physics, Part 2 (Letters) 42(7A): L729–L731. Kawamura F, Morishita M, et al. (2005) The effects of Na and some additives on nitrogen dissolution in the Ga–Na system: A growth mechanism of GaN in the Na flux method. Journal of Materials Science: Materials in Electronics 16(1): 29–34. Kawamura F, Morishita M, et al. (2008) Effect of carbon additive on increases in the growth rate of 2 in GaN single crystals in the Na flux method. Journal of Crystal Growth 310(17): 3946–3949. Kawamura F, Tanpo M, et al. (2009b) Growth of GaN single crystals with extremely low dislocation density by two-step dislocation reduction. Journal of Crystal Growth 311(10): 3019–3024. Kawamura F, Umeda H, et al. (2009c) LPE growth of bulk GaN crystal by alkali-metal flux method. Materials Science Forum 600–603: 1245–1250. Kisielowski C, Kruger J, and Ruvimov S (1996) Strain-related phenomena in GaN thin films. Physical Review B 54: 17745–17753. Koukitu A, Hama S, Taki T, and Seki H (1998) Thermodynamic analysis of hydride vapor phase epitaxy of GaN. Japanese Journal of Applied Physics 37: 762–765. Krukowski S (1999) Growth of GaN single crystals under high nitrogen pressures and their characterization. Crystal Research and Technology 34(5–6): 785. Kryliouk O, Reed M, Dann T, Anderson T, and Chai B (1999) Large area GaN substrates. Materials Science and Engineering B 66(1): 26–29. Lei T, Fanciulli M, Molnar RJ, Moustakas TD, Graham RJ, and Scanlon J (1991) Epitaxial growth of zinc blende and wurtzitic gallium nitride thin films on (001) silicon. Applied Physics Letters 59: 944. Li X, Ni X, Lee J, et al. (2009) Efficiency retention at high current injection levels in m-plane InGaN LEDs. Applied Physics Letters 95: 121107. Liu SS and Stevenson DA (1978) Growth kinetics and catalytic effects in the vapor phase epitaxy of gallium nitride. Journal of Electrochemical Society 125: 1161. Logan RA and Thurmond CD (1972) Heteroepitaxial thermal gradient solution growth of GaN. Journal of the Electrochemical Society 119(12): 1727–1735. Lu H, Zhang R, Xiu X, Xie Z, Zheng Y, and Li Z (2007) Low leakage Schottky rectifiers fabricated on homoepitaxial GaN. Applied Physics Letters 91: 172113. Madar R, Jacob G, Hallais J, and Fruchart R (1975) High pressure solution growth of GaN. Journal of Crystal Growth 31: 197. Maruska HP and Tjetjen JJ (1969) The preparation and properties of vapor deposited single crystal-line GaN. Applied Physics Letters 15: 327. Mathis SK, Romanov AE, Chen LF, Beltz GE, Pompe W, and Speck JS (2000) Modeling of threading dislocation reduction in growing GaN layers. Physica Status Solidi (a) 179: 125–145. Matsumoto T, Sano M, and Aoki M (1974) Pair luminescence from Zn-doped GaN. Japanese Journal of Applied Physics 13: 373–374. Meissner E, Birkmann B, et al. (2005) Characterisation of GaN crystals and epilayers grown from a solution at room pressure. Physica Status Solidi (c) 7: 2040–2043. Miyoshi T, Masui S, Okada T, et al. (2009) 510–515 nm InGaN-based green laser diodes on c-plane GaN substrates. Applied Physics Express 2: 062201. Molnar RG (1999) Hydride vapor phase epitaxial growth of III-nitrides. In: Pankove J and Moustakas TD (eds.) Galium

Growth of Bulk GaN Crystals Nitride (GaN) II, Semiconductors and Semimetals, vol. 57, ch. 1, pp. 1–32. London: Academic Press. Molnar RJ, Gotz W, Romano LT, and Johnson JM (1997) Growth of gallium nitride by hydride vapor epitaxy. Journal of Crystal Growth 178: 147–156. Molnar RJ, Nichols KB, Maki P, Brown ER, and Melngailis I (1995) The role of impurities on the hydride vapor phase epitaxially grown gallium nitride. In: Materials Research Society Symposium Proceedings, vol. 378, 479. Warrendale, PA: Materials Research Society. Morishita M, Kawamura F, et al. (2003) Growth of bulk GaN single crystals using Li–Na mixed flux system. Japanese Journal of Applied Physics, Part 2 (Letters) 42(6A): L565–L567. Morishita M, Kawamura F, et al. (2005) Promoted nitrogen dissolution due to the addition of Li or Ca to Ga–Na melt; some effects of additives on the growth of GaN single crystals using the sodium flux method. Journal of Crystal Growth 284(1–2): 91–99. Morkoc H (2001) Comprehensive characterization of hydride VPE grown GaN layers and templates. Materials Science and Engineering R33: 135–207. Motoki K, Okahisa T, Nakahata S, et al. (2002) Growth and characterization of freestanding GaN substrates. Journal of Crystal Growth 237–239: 912–921. Munir AA and Searcy AW (1965) Activation energy for the sublimation of gallium nitride. Journal of Chemical Physics 42: 4223. Nam OH, Ha KH, Kwak JS, et al. (2004) Characteristics of GaN-based laser diodes for post-DVD applications. Physica Status Solidi (a) 201(12): 2717–2720. Narendran N, Gu Y, Freyssinier JP, Yu H, and Deng L (2004) Solid-state lighting: Failure analysis of white LEDs. Journal of Crystal Growth 268: 449–456. Ohki Y, Toyoda Y, Kobayasi H, and Akasaki I (1982) Fabrication and properties of a practical blue-emitting GaN MIS diode. Paper Presented at International Symposium. GaAs and Related Compounds, Institute of PhysicsConference Series No. 63, ch. 10, pp. 479–484. Japan. Oshima Y, Eri T, Sunakawa H, and Usui A (2002) Fabrication of freestanding GaN wafers by hydride vapor phase epitaxy with void-assisted separation. Physica Status Solidi (a) 194(2): 554–558. Oshima Y, Yoshida T, and Eri T (2007) Fabrication of 3-inch freestanding GaN substrates by hydride. Vapor phase epitaxy with void-assisted separation. In: 49th Electronic Materials Conference. Notre Dame, Indiana, 20–22 June. Ozawa T, Dohi M, et al. (2009) Solution growth of GaN on sapphire substrate under nitrogen plasma. Journal of Crystal Growth 311(3): 440–442. Parillaud O, Wagner V, Buehlmann HJ, and Ilegems M (1999) Localized epitaxy of GaN by HVPE on SiC and sapphire substrates. Materials Research Society Internet Journal of Nitride Semiconductor Research 4S1: G4.3. Parish G, Keller S, Kozodoy P, et al. (1999) High-performance (Al,Ga)N-based solar-blind ultraviolet p–i–n detectors on laterally epitaxially overgrown GaN. Applied Physics Letters 75: 247–249. Paskova T and Evans KR (2009) GaN substrates – progress, status and prospects. IEEE Journal of Selected Topics in Quantum Electronics 15: 1041–1052. Paskova T and Monemar B (2003) Hydride vapour phase epitaxy growth of thick GaN layers. In: Manasreh O (ed.) III-Nitride Semiconductors: Growth, ch. 10, pp. 175–236. New York: Taylor and Francis Group. Paskova T, Birch J, Tungasmita S, et al. (1999b) Thick hydride vapor phase epitaxial GaN layers grown on sapphire with different buffers. Physica Status Solidi (a) 176: 415–419.

279

Paskova T, Paskov PP, Birch J, et al. (2000a) HVPE regrowth on free-standing GaN quasi-substrates. In: Proceedings of the IWN2000, Japanese IPAP Conference Series, vol. C1, pp. 19–22. Nagoya. Paskova T, Svedberg E, Henry A, Ivanov IG, Yakimova R, and Monemar B (1999a) Thick GaN layers grown on a-plane sapphire substrates by hydride vapour phase epitaxy. Physica Scripta T79: 67–70. Paskova T, Tungasmita S, Valcheva E, et al. (2000b) Hydride vapor phase homoepitaxial growth of GaN on MOCVD grown templates. Materials Research Society Internet Journal of Nitride Semiconductor Research 5S1: W3.14. Paskova T, Valcheva E, Birch J, et al. (2001) Defect and stress relaxation in HVPE-GaN films using high temperature reactively sputtered AlN buffer. Journal of Crystal Growth 230: 381–386. Porowski S (1999) Near defect free GaN substrates. Materials Research Society Internet Journal of Nitride Semiconductor Research 4S1: G 1.3. Porowski S, Grzegory I, Krukowski S, Leszczynski M, Perlin P, and Suski T (2004) Blue lasers on high pressure grown GaN single crystal substrates. Europhysics News 35: 3. Przhevalskii IN, Yu S, Karpov S, and Makarov YN (1998) Thermodynamic properties of group-III nitrides and related species. Materials Research Society Internet Journal of Nitride Semiconductor Research 3: article 30. Purdy AP (1999) Ammonothermal synthesis of cubic gallium nitride. Chemistry of Materials 11: 1648–1651. Richter E, Hnnig Ch, Zeimer U, et al. (2006) Freestanding two inch c-plane GaN layers grown on (100)-lithium aluminium oxide by hydride vapor phase epitaxy. Physica Status Solidi (c) 3: 1439–1443. Safvi SA, Perkins NR, Horton MN, Thon A, Zhi D, and Kuech TF (1996) Effect of reactor geometry and growth parameters on the uniformity and material properties of GaN/sapphire grown by hydride vapor-phase epitaxy. In: Materials Research Society Symposium Proceedings, vol. 423, 227. Warrendale, PA: Materials Research Society. Sato H, Tyagi A, Zhong H, et al. (2007) High power and high efficiency green light emitting diode on free-standing semipolar (10–22) bulk GaN substrate. Physica Status Solidi (RRL) 1(4): 162–164. Schmidt MC, Kim KC, Sato H, et al. (2007) High power and high external efficiency m-plane InGaN light emitting diodes. Japanese Journal of Applied Physics 46: L126. Schujman SB, Schovalter LJ, Bondokov RT, et al. (2008) Structural and surface characterization of large diameter, crystalline AlN substrates for device fabrication. Journal of Crystal Growth 310: 887. Seifert W, Franzheld R, Butter E, Sobotta H, and Reide V (1983) On the origin of free carriers in high-conducting n-GaN. Crystal Research and Technology 18: 383. Senawiratne J, Li Y, Zhu M, et al. (2008) Junction temperature measurements and thermal modeling of GaInN/GaN quantum well light-emitting diodes. Journal of Electronic Materials 37(5): 607–610. Sharma R, Pattison PM, Masui H, et al. (2005) Demonstration of a semipolar (10-1-3) InGaN/GaN green light emitting diode. Applied Physics Letters 87: 231110. Shintani A and Minagawa S (1974) Kinetics of the epitaxial growth of GaN using Ga, HCl and NH3. Journal of Crystal Growth 22: 1. Simpkins BB, Yu ET, Waltereit P, and Speck JS (2003) Correlated scanning Kelvin probe and conductive atomic force microscopy studies of dislocations in gallium nitride. Journal of Applied Physics 94: 1448. Slack GA and McNelly TF (1976) Growth of high purity AlN crystals. Journal of Crystal Growth 34(2): 263–279. Slack GA and McNelly TF (1977) AlN single-crystals. Journal of Crystal Growth 42: 560–563.

280 Growth of Bulk GaN Crystals Sone H, Nambu S, Kawaguchi Y, et al. (1999) Optical and crystalline properties of epitaxial-overgrown-GaN using tungsten mask by hydride vapor phase epitaxy. Japanese Journal of Applied Physics 38: L356–L359. Song Y, Wang W, et al. (2003) Bulk GaN single crystals: Growth conditions by flux method. Journal of Crystal Growth 247(3–4): 275–278. Soukhoveev V, Ivantsov V, et al. (2001) Characterization of 2.5-inch diameter bulk GaN grown from melt-solution. Physica Status Solidi (a) 188(1): 411–414. Stringfellow GB (1980) Vapor phase growth. In: Pamplin BR (ed.) Crystal Growth, ch.5, 181. Oxford: Pergamon. Trassoudaine A, Aujol E, Cadoret R, Paskova T, and Monemar B (2001) A new mechanism in the growth process of GaN by HVPE. In: Materials Research Society Symposium Proceedings, vol. 639, G.3.2. Warrendale, PA: Materials Research Society. Unland J, Onderka B, Davydov A, and Schmid-Fetzer R (2003) Thermodynamics and phase stability in the Ga–N system. Journal of Crystal Growth 256: 33–51. Usui A, Sunakawa H, Sakai A, and Yamaguchi AA (1997) Thick GaN epitaxial growth with low dislocation density by hydride vapor phase epitaxy. Japanese Journal of Applied Physics 36: L899–L902. Utsumi W, Saitoh H, Kaneko H, Watanuki T, Aoki K, and Shimomura O (2003) Congruent melting of gallium nitride at 6 GPa and its application to single-crystal growth. Nature Materials 2: 735–738. Valcheva E, Paskova T, Tungasmita S, et al. (2002) Interface structure of hydride vapor phase epitaxial GaN grown with high-temperature reactively sputtered AlN buffer. Applied Physics Letters 76: 1860–1962. Vaudo RP, Xu X, Loria C, Salant AD, Flynn JS, and Brandes GR (2002) GaN boule growth: A pathway to GaN wafer with improved material quality. Physica Status Solidi (a) 194(2): 494–497. Vodakov YuA, Karklina MI, Mokhov EN, and Roenkov AD (1980) GaN epitaxial growth by sublimation sandwich method. Izvestia Akademik Nauk USSR Inorganic Materials 16: 537. Vodakov YuA and Mokhov EN (2002) Growth of bulk GaN crystals by the sublimation sandwich method. In: Paskova T and Monemar B (eds.) Nitrides as Seen by Technology, ch. 4, pp. 59–78. Trivandrum, India: Research Signpost. Vodakov YuA, Mokhov EN, Roenkov AD, Boiko ME, and Baranov PG (1997) High rate GaN epitaxial growth by sublimation sandwich method. Journal of Crystal Growth 183: 10. Wang B and Callahan MJ (2006) Ammonothermal synthesis of III-nitride crystals. Crystal Growth and Design 6(6): 1227–1246. Wang BG, Callahan MJ, et al. (2006) Ammonothermal growth of GaN crystals in alkaline solutions. Journal of Crystal Growth 287(2): 376–380. Wang WJ, Chen XL, et al. (2004) Assessment of Li–Ga–N ternary system and GaN single crystal growth. Journal of Crystal Growth 264(1–3): 13–16. Wetzel C, Zhu M, Senawiratne J, et al. (2008) Light-emitting diode development on polar and non-polar GaN substrates. Journal of Crystal Growth 310: 3987–3991. Weyers M, Richter E, Hennig Ch, Hagedorn S, Wernicke T, and Trankle G (2008) GaN substrates by HVPE. Proceedings of SPIE 6910: 691001.1–691001.10. Yamada T, Yamane H, et al. (2005) Single crystal growth of GaN using a Ga melt in Na vapor. Journal of Crystal Growth 281(2–4): 242. Yamada T, Yamane H, et al. (2006) The process of GaN single crystal growth by the Na flux method with Na vapor. Journal of Crystal Growth 286(2): 494.

Yamada T, Yamane H, et al. (2009) Growth of colorless transparent GaN single crystals on prismatic GaN seeds using a Ga melt and Na vapor. Materials Research Bulletin 44(3): 594–599. Yamane H, Aoki M, Yamada T, et al. (2005) Time dependence of the growth morphology of GaN single crystals prepared in a Na–Ga melt. Japanese Journal of Applied Physics, Part 2 (Letters) 44(5A): 3157. Yamane H, Kinno D, Shimada M, Sekiguchi T, and DiSalvo FJ (2000) GaN single crystal growth from a Na–Ga melt. Journal of Materials Science 35: 801. Yamane H, Shimada M, Clarke SJ, and Disalvo FJ (1997) Preparation of GaN single crystals using a Na flux. Chemistry of Materials 9: 413. Zetterstrom RB (1970) Synthesis and growth of single crystals of gallium nitride. Journal of Materials Science 5: 1102. Zhong H, Tyagi A, Fellows NN, et al. (2007) High power and high efficiency green light emitting diode on free-standing semipolar (10–11) bulk GaN substrate. Applied Physics Letters 90: 233504. Zhou Y, Li M, Wang D, et al. (2006) Electrical characteristics of bulk GaN-based Schottky rectifiers with ultrafast reverse recovery. Applied Physics Letters 88: 113509.

Further Reading Abernathy CR, MacKenzie JD, and Donovan SM (1997) Growth of group III nitrides by metalorganic molecular beam epitaxy. Journal of Crystal Growth 178: 74–86. Avrutin V, Silversmith DJ, Mori Y, et al. (2010) Bulk GaN and AlN: Progress and challenges. IEEE Proceedings (in press). Davey JE and Pankey T (1968) Epitaxial GaAs films deposited by vacuum evaporation. Journal of Applied Physics 39: 1941. Denisa A, Goglioa G, and Demazeau G (2006) Gallium nitride bulk crystal growth processes: A review. Materials Science and Engineering R 50: 167–194. Chernov AA (1984) Modern crystallography III: Crystal growth. In: Val˘nshtel˘n BK (ed.) Volume 3 of Modern Crystallography, Volume 36 of Solid-State Sciences Series, 517. Berlin: Springer. Ehrentraut D, Meissner E, and Bockowski M, (eds.) (2010) Technology of Gallium Nitride Crystal Growth, Springer Series in Materials Science vol. 133 (ISBN: 978-3-642-04828-9). Berlin: Springer. Fornari R and Roth M (eds.) (2009) Recent advances in bulk crystal growth. Focused Issue MRS Bulletin 34(4): 239–283. Fremunt R, Cerny P, Kohout J, Rosicka V, and Burger A (1981) Optimized growth conditions of GaN epitaxial layers. Crystal Research and Technology 16: 1257. Grandjean N, Leroux M, Laugt M, and Massies J (1997) Gas source molecular beam epitaxy of wurtzite GaN on sapphire substrates using GaN buffer layers. Applied Physics Letters 71: 1240–1242. Gunther KZ (1958) Molecular beem epitaxy. Zeitschrift fuer Naturforschung 13a: 1081. Herman MA (1989) In: Panish MB (ed.) Molecular Beam Epiaxy. Springer Series in Material Science, vol. 7. Berlin– Heidelberg: Springer. Herro AG, Zhuang D, Schlesser R, Collazo R, and Sitar Z (2006) Growth of large AlN single crystals along the [0001] direction. In: Materials Research Society Symposium Proceedings, vol. 892, FF21.01.1. Warrendale, PA: Materials Research Society. Jain SC, Willander M, Narayan J, and Van Overstraeten R (2000) III-nitrides: Growth, characterization, and properties. Journal of Applied Physics 87: 965.

Growth of Bulk GaN Crystals Lebedev V, Jinschek J, Kra¨ußlich J, Kaiser U, Schro¨ter B, and Richter W (2001) Hexagonal AlN films grown on nominal and off-axis Si(0 0 1) substrates. Journal of Crystal Growth 230: 426–430. Mesrine M, Grandjean N, and Massies J (1998) Efficiency of NH3 as nitrogen source for GaN molecular beam epitaxy. Applied Physics Letters 72: 350. Nikishin SA, Antipov VG, Francoeur S, et al. (1999) High-quality AlN grown on Si(111) by gas-source molecular-beam epitaxy with ammonia. Applied Physics Letters 75: 484. Paskova T (ed.) (2008) Nitrides with Nonpolar Surfaces (ISBN: 978-3-527-40768-2). Berlin: Wiley-VCH. Paskova T, Becker L, Bo¨ttcher T, Hommel D, Paskov PP, and Monemar B (2007) Effect of sapphire-substrate thickness on the residual strain and bending in GaN films grown by hydride vapor phase epitaxy. Journal of Applied Physics 102: 123507. Paskova T and Evans KR (2009) GaN substrates – progress, status and prospects. IEEE Journal of Selected Topics in Quantum Electronics 15: 1041–1052. Paskova T, Hanser A, Preble E, et al. (2008) Defect and emission distributions in bulk GaN grown in polar and nonpolar directions: A comparative analysis. Proceedings of SPIE 6894: 68940D.1–68940D.7. Paskova T and Monemar B (eds.) (2002) Nitrides as Seen by Technology (ISBN: 81-7736-198-8) Trivandrum, India: Research Signpost. Sitar Z, Paislei M, Yan B, Ruan J, Choyke WJ, and Davis RF (1990) Growth of AlN/GaN layered structures by gas source molecular-beam epitaxy. Journal Vacuum Science and Technololy B 8: 316–322. Popovici G, Morkoc H, and Mohammad SN (1998) Deposition and properties of group III nitrides by molecular beam epitaxy. In: Gil B (ed.) Group III Nitride Semiconductor Compounds, p. 19. Oxford: Clarendon. Porowski S (2004) Blue lasers on high pressure grown GaN single crystal substrates. Europhysics News 35: 3. Vispute RD, Narayan J, Wu H, and Jagannadham K (1995) Epitaxial growth of AlN thin films on silicon (111) substrates by pulsed laser deposition. Journal of Applied Physics 77: 4724. Wang B and Callahan MJ (2006) Ammonothermal synthesis of III-nitride crystals. Crystal Growth and Design 6(6): 1227–1246.

281

Wickenden DK, Faulkner KR, Brander RW, and Isherwood BJ (1971) Growth of epitaxial layers of gallium nitride on silicon carbide and corundum substrates. Journal of Crystal Growth 9: 158. Yoshida S, Misawa S, and Itoh A (1975) Epitaxial growth of aluminum nitride films on sapphire by reactive evaporation. Applied Physics Letters 26: 461–463. Yoshida T, Oshima Y, Eri T, et al. (2008) Fabrication of 3-in GaN substrates by hydride vapor phase epitaxy using void assisted separation method. Journal of Crystal Growth 310: 5–7. Yoshimoto M, Hatanaka A, Itoh H, and Matsunami H (1998) GaN growth on sapphire and 6H-SiC by metalorganic molecular beam epitaxy. Journal of Crystal Growth 188: 92. Zhang W and Meyer BK (2003) Growth of GaN quasi-substrates by hydride vapor phase epitaxy. Physica Status Solidi (c) 0(6): 1571–1582.

Relevant Websites http://www.ammono.com – Ammono. http://www.hitachi-cable.co.jp – Hitachi Cable, Ltd., Optical Transceivers; Products and Services. http://www.inlustra.com – Inlustra Technologies, Inc. http://www.unipress.waw.pl – Institute of High Pressure Physics ‘‘UNIPRESS’’. http://www.kymatech.com – Kyma Technologies, Inc. http://www.lumilog.com – Lumilog. http://www.m-kagaku.co.jp – Mitsubishi Chemical Corporation. http://www.samsungcorning.com – Samsung Corning Precision Glass, Product, Gan Substrate. http://global-sei.com – Sumitomo Electric Industries, LTD. http://www.topgan.fr.pl – TopGaN Ltd.