1,168 11 7MB
Pages 337 Page size 198.5 x 322.6 pts
Springer Series in
materials science
133
Springer Series in
materials science Editors: R. Hull C. Jagadish R.M. Osgood, Jr. J. Parisi Z. Wang H. Warlimont The Springer Series in Materials Science covers the complete spectrum of materials physics, including fundamental principles, physical properties, materials theory and design. Recognizing the increasing importance of materials science in future device technologies, the book titles in this series ref lect the state-of-the-art in understanding and controlling the structure and properties of all important classes of materials.
Please view available titles in Springer Series in Materials Science on series homepage http://www.springer.com/series/856
Dirk Ehrentraut Elke Meissner Michal Bockowski Editors
Technology of Gallium Nitride Crystal Growth With 190 Figures
13
Editors
Professor Dirk Ehrentraut
Dr. Elke Meissner
Tohoku University World Premier International Research Center - Advanced Institute for Materials Research (WPI - AIMR) Katahira 2-1-1, Aoba-ku 980-8577 Sendai, Japan E-mail: [email protected]
Fraunhofer-Institut Integrierte Systeme und Bauelementetechnologie (IISB) Abteilung Kristallz¨uchtung Schottkystr. 10, 91058 Erlangen, Germany E-mail: [email protected]
Dr. Michal Bockowski Polish Academy of Sciences, Institute of High-Pressure Physics Sokolowska 29/37, 01-142 Warszaw, Poland E-mail: [email protected]
Series Editors:
Professor Robert Hull
Professor J¨urgen Parisi
University of Virginia Dept. of Materials Science and Engineering Thornton Hall Charlottesville, VA 22903-2442, USA
Universit¨at Oldenburg, Fachbereich Physik Abt. Energie- und Halbleiterforschung Carl-von-Ossietzky-Straße 9–11 26129 Oldenburg, Germany
Professor Chennupati Jagadish
Dr. Zhiming Wang
Australian National University Research School of Physics and Engineering J4-22, Carver Building Canberra ACT 0200, Australia
University of Arkansas Department of Physics 835 W. Dicknson St. Fayetteville, AR 72701, USA
Professor R. M. Osgood, Jr.
Professor Hans Warlimont
Microelectronics Science Laboratory Department of Electrical Engineering Columbia University Seeley W. Mudd Building New York, NY 10027, USA
DSL Dresden Material-Innovation GmbH Pirnaer Landstr. 176 01257 Dresden, Germany
Springer Series in Materials Science ISSN 0933-033X ISBN 978-3-642-04828-9 e-ISBN 978-3-642-04830-2 DOI 10.1007/978-3-642-04830-2 Springer Heidelberg Dordrecht London New York Library of Congress Control Number: 2010920285 © Springer-Verlag Berlin Heidelberg 2010 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specif ically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microf ilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer-Verlag. Violations are liable to prosecution under the German Copyright Law. The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specif ic statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Cover design: eStudio Calamar Steinen Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)
Foreword
Semiconductor materials have been studied intensively since the birth of silicon technology more than 50 years ago. The ability to physically and chemically tailor their properties with precision is the key factor responsible for the electronic revolution in our society over the past few decades. Semiconductor material systems (like silicon and GaAs-related materials) have now matured and found well established applications in electronics, optoelectronics, and several other fields. Other materials such as III-Nitrides were developed later, in response to needs that the above mentioned semiconductors were unable to fulfill. The properties of IIInitrides (AlN, GaN InN, and related alloy systems) make them an excellent choice for efficient light emitters in the visible as well as the UV region, UV detectors, and for a variety of electronic device such as high frequency unipolar power devices. There was a major upsurge in the research of the GaN material system around 1970. Its prime focus was to grow bulk materials. The failure to produce p-type materials at the time discouraged most research groups, and their activities faded after a few years. Several important developments in growth procedures in the mid to late 1980s led to a resurgence and revival of the research interest in GaN and related materials. A method to grow smooth single crystalline GaN layers on a foreign substrate (i.e., heteroepitaxy on sapphire) using a thin (about 30 nm) low temperature grown AlN buffer layer was suggested by Akasaki et al. They also discovered a method to produce low Ohmic p-type GaN and demonstrated a pn-junction based light emission device. This resulted in the rapid development and commercialization of III-Nitride light emitters (LEDs and lasers). Using heteroepitaxy over sapphire (later SiC), Khan et al. also demonstrated the two-dimensional electron gas at the interface of high quality AlGaN-GaN layers and high electron mobility transistors, in the late 1980s. This led to a rapid development and improvement of electronic microwave power devices with impressive power handling capacity. These are now being employed in commercial applications. To date, a major part of the GaN research has used heteroepitaxy because IIIN bulk substrates are not yet sufficiently developed. Large size, high quality bulk GaN substrates are still much more expensive than sapphire or SiC, which are a preferred choice for GaN heteroepitaxy. The future need for bulk GaN substrates is currently under debate. These substrates provide a clear advantage of increased lifetime in continuous operation of lasers because of reduced dislocations (threading v
vi
Foreword
dislocation density 4.5 GPa. Experimentally, if GaN powder is compressed at a pressure of 6 GPa and heated to 2,400ı C maintaining that pressure, small GaN single crystals about 100 m in size have been produced upon cooling. This is the closest analog of growth from a melt such as for silicon. The need for these extremely high pressures and temperatures (and low growth rates as well) makes the melt growth an impractical technique for realizing large single crystals of GaN. Growth techniques for bulk GaN at or near atmospheric pressure, restricting the temperatures to the range 1 cm has been reported on 2" substrates. Si-doping during growth produces n-type material; p-type can be obtained by Mg doping. By elaboration with different patterning of the substrate or on a buffer layer deposited on the substrate, a pattern of cavities can be produced near the substrate, acting as a weak mechanical link after (or during) cool down. This can be optimized as a self separation technique where the substrate is automatically removed during cool down from growth. Wafers are otherwise cut from the boule by standard sawing, and polished. Obviously, such a polished bulk wafer can be used as a high quality substrate in subsequent bulk growth runs.
Foreword
vii
Another vapor phase growth technique is the Metal Organic Vapor Phase Epitaxy (MOVPE) technique, very important for growth of thin epitaxial layers in device structures. Metalorganic precursors (such as triethyl or trimethyl-gallium) are used together with NH3 in the growth chemistry. For growth of thin epi-layers the growth rate is typically adjusted to about 1–2 m/h, to get high quality layers. For the purpose of bulk growth, the growth rate can be optimized to much higher values, up to about 50 m/h. Solution growth of bulk crystals at moderate temperatures (800% improvement over incandescent lighting, by far the most prevalent form of lighting in the world today. As can be seen from the summary of luminous efficacies for different light sources shown in Table 1.1, significant advances over the current state-of-the-art technology are required to meet the US DOE’s requirements. In 2008, the GaN LED market was estimated to be around $7 billion and is estimated to grow to approximately $15 billion by 2012 [4]. In spite of the already significant market for GaN LEDs, their use in SSL today is quite limited to a small but growing number of applications, restricted both by the cost and the overall technology readiness level of GaN-based LED lighting solutions compared to the existing approaches. As improved device technologies are developed, the market for LEDs for SSL is forecast to grow to $1.2 billion in 2012 [4] and will reach $20B or more over the long term (Table 1.2).
1.2.2 Electrical Systems and Power Electronics in the III-Ns Improving energy efficiency in electrical systems reduces the total energy requirements for many applications. For example, standby electrical power for consumer electronics can account for 5–10% of annual energy consumption in the home, adding up to more than $3 billion in annual energy costs [5, 6]. According to the
6
A.D. Hanser and K.R. Evans
Table 1.2 Luminous efficacy ranges for different light sources [3] Light Typical luminous efficacy source range (lm/W) Incandescent 10–18 Halogen 15–20 Compact fluorescent (incl. ballast) 35–60 Linear fluorescent (incl. ballast) 50–100 Metal halide (incl. ballast) 50–90 Current SSL products for general illumination Up to 62 SSL system luminous efficacy target (2025) 160
US DOE, this standby electricity consumption was 64.5 million megawatt hours in 2004 [7]. This amount of wasted energy is equivalent to the output of 18 typical power stations. This standby power consumption is a result of the use of low-cost yet inefficient power supplies that consume a significant amount of power when the electronic device is not in use. Additionally, the implementation of SSL will require high-efficiency power suppliers and controllers to maintain the overall highefficiency of the lighting system. Electrical systems are also finding increased use in automotive applications, where hybrid electric vehicles (HEVs) are growing in importance as a way to improve efficiency in transport systems. Using an electric motor in combination with an internal combustion engine boosts vehicle energy efficiency and enables reduction in transportation related oil demand and associated atmospheric emissions. One particular study shows that increased HEV use offers the possibility of reducing the total oil demand for light-duty vehicles by 2030 relative to current levels of consumption, in spite of an increase in both the number of vehicles and the average annual vehicle mileage [8]. Using an electric motor alone or in combination with an internal combustion engine boosts vehicle energy efficiency and enables a reduction in CO2 emissions, along with a reduction in transportation related oil demand. Finally, the electrical grid that includes the traditional generation and transmission of electrical power from power plants to users is growing to include new sources of power generation, including distributed and renewable power networks, such as wind, solar, biomass, and others. Current generation systems can benefit from more efficient transformation and switching of power, and smarter systems improve the efficiency of renewable energy sources through monitoring and management of the varying cycles of energy generation and load requirements. Improvements in the transformation of voltages and enhanced dynamic control of the voltage, impedance, and phase angle of high-voltage ac transmission lines can be achieved through solid-state power control devices [9]. Energy efficiency in power conversion for industrial and consumer electronics, electronics for fully electric and hybrid electric vehicles, and power generation and distribution for the electric grid, can all benefit from higher efficiency electrical components. Improving efficiency in these systems starts with the high power electronic (HPE) components that make up the power inverters and converters in power supplies. For modern systems these components are semiconductor devices made
1
Development of the Bulk GaN Substrate Market
7
primarily from silicon. Due to intrinsic materials limitations, silicon devices are limited in power handling capabilities and temperature of operation, which negatively impacts the overall electrical system efficiency. Gallium nitride has several properties that enable improved power performance with higher efficiency when compared to silicon. Gallium nitride has a wide bandgap (3.4 eV), a large critical breakdown electric field .3 MV=cm/, and a high electron mobility (1500 cm2=V s at RT). These properties result in low on-state resistance, low switching losses, high-temperature performance capabilities, and high power conversion efficiencies for electronic devices such as Schottky diodes and field effect transistors (FETs). Employing these devices would improve the efficiency of electrical systems and could have a significant impact in reducing overall electricity consumption worldwide, impacting virtually every aspect of electrical usage with potential savings in excess of $30 billion per year [10]. In 2008, the total available market for GaN power electronics was greater than $350 million, possibly growing to a large portion of the $22 billion power electronics market over the long term [11].
1.2.3 Positioning GaN Substrates for SSL and Power Electronics Markets Considerable technology development is needed to enable penetration of GaN devices into the SSL and power electronics markets. Accessing these markets requires the maturation of the processing technologies for fabricating the materials and devices that are used in the lighting and power components systems. Highly manufacturable, low-cost processes for fabricating high-performance devices are needed. Two significant elements will contribute to the successful implementation of GaN-based devices in these applications. First, the performance of the GaN devices must improve. For example, luminous efficacy needs to increase by more than a factor of three in order to meet and exceed the performance of competing approaches. GaN-based LEDs have steadily improved their efficacy over a period of time through improvements in several areas such as the device design; however, it remains to be seen if the current technology approach will be able to continue on its current performance trajectory. Most of the standard LEDs today are grown on foreign (non-nitride) substrates, which results in an extensive number of materials defects due to the imperfect match of materials properties between the substrate and epitaxial layers. Such an approach results in an imperfect match of materials properties between the substrate and epitaxial layers, which generates defects and degrades device performance. Researchers are investigating the effect of these defects on the performance of LEDs and whether device performance remains fundamentally limited due to material quality. This is of particular importance for high current density devices that are used for high total light output. As with lighting applications, device performance and demonstrated reliability remain as significant barriers to realizing the full potential of GaN-based power electronics devices.
8
A.D. Hanser and K.R. Evans
The second element is cost reduction of GaN-based device technology, where substrate cost can be a significant component of the total cost of the device. Substrate cost reduction is accomplished through increasing production volume. Navigating the cost–volume curve is one of the main challenges in semiconductor manufacturing and market demand drives the need for increased volume production. Demand is strong for GaN-based devices, but for devices based on GaN substrates, current markets such as those for LEDs are only beginning to use GaN substrates and only for high end applications, due to incumbent device technologies with low price points. Devices on foreign substrates have adequate performance for current LED applications, and the price point for these GaN epiwafers and devices is low due in part to the high volumes and low cost of substrates. As a result, emerging LED technologies utilizing bulk GaN substrates cannot take advantage of these markets yet, due to the high costs of the substrates that result from low production volumes and low manufacturing yields. Thus, new markets need to be identified that can accommodate these near-term limitations for bulk GaN substrates. Over the long term, GaN substrate manufacturing costs should approach that of GaAs substrates since the cost of the growth equipment, the starting materials, and the growth rates are similar; however, it must also be recognized that substrate cost is only one ingredient in determining the manufacturing cost of a given device for a given application’s requirement in terms of device performance, reliability, and form factor (Fig. 1.1).
GaN
Substrate Cost ($ / cm2)
$100.00
$10.00 SiC
InP
Sapphire $1.00 GaAs
$0.10 Silicon $0.01 1.0E+05
1.0E+06
1.0E+07
1.0E+08
1.0E+09
1.0E+10
1.0E+11
Estimated Annual Wafer Production (cm2)
Fig. 1.1 Estimated substrate cost per square centimeter vs. annual wafer production for semiconductor substrates
1
Development of the Bulk GaN Substrate Market
9
Table 1.3 Applications for GaN-based LED, power, and laser devices LED applications Power applications Laser diode applications Backlighting (mobile RF power amplifiers for Optical data storage devices, LCD displays) communications and radar Full-color, large-format Switch mode power Projectors video screens supplies Interior and exterior Motor controls Displays automotive lighting Signals and signage Satellite power systems Commercial printing; Lighting for general High-temperature Testing and measurement illumination electronics (spectroscopy, sensing)
Bulk GaN substrate manufacturers have recognized that intermediate and niche opportunities exist and can be leveraged to help meet the needs performance and cost requirements of the larger SSL and power electronics markets. It is these near and intermediate term markets that present the chance to establish manufacturing capacities, grow commercial volumes, increase manufacturing yield, reduce cost, and thus enable access to new and larger markets through growth and continuous improvements (see Table 1.3 for applications for GaN-based devices). Initial target market niches are those that will command a premium for higher performance devices, or benefit from some other aspect of the device, such as higher electrical efficiency, smaller size, and/or less temperature-sensitive performance. These initial markets will rapidly adopt devices that utilize a bulk GaN substrate as they address a need unfilled or poorly met by competitive approaches. We will now examine some of the factors that will contribute to bulk GaN substrates meeting market requirements and how these drivers will facilitate their entry into additional markets.
1.2.4 Key Drivers for Bulk GaN Substrate Commercialization Success One of the areas where GaN substrates provide a clear and immediate benefit is in GaN laser diodes (LDs). The applications for GaN-based LDs are primarily for next-generation, high-definition DVD players and recorders, although other applications include high definition video projectors and displays, commercial printing, and testing and measurement applications, such as spectroscopy and bio-sensing. The high-definition DVD systems utilize a 405 nm laser diode to read and write data on an optical disc, and GaN is uniquely suited to address this wavelength for semiconductor lasers. The format war between Blu-ray and HD DVD in the early 2000s is thought to have potentially hindered the development of the market for these next generation players [12], but in early 2008 the industry settled on the Bluray format, paving the way for increased volumes of laser diodes and optical drives. Sales of Blu-ray DVD players and recorders were expected to reach 5 million units
10
A.D. Hanser and K.R. Evans
in 2008 and were predicted to more than double to 12 million units in 2009 [13], while the GaN-based laser diode market is expected to grow to $1.2 billion by 2011 [14]. Read-only LDs for Blu-ray disc players typically operate at 20 mW [15] and have been developed with device lifetimes exceeding 10,000 h under continuous wave (CW) operation, but developing higher power lasers will enable faster write speeds in DVD recorders. Read/write powers for laser diodes vary from 125 mW for 2 standard write speeds to 170 mW (pulsed) for 4 standard write speeds [16] and up to 250 mW for 8 write speeds [17]. For LDs with higher output power the amount of heat generated during operation becomes significant. Extending the reliable performance of LDs to higher output powers requires improvements in material quality [18]. Manufacturing of GaN-based lasers remains focused in Japan. In 2008, Nichia Corporation, the leading producer of GaN LDs, was reported to hold an 80% market share of the GaN-based LD market [19]. Nichia is reported to use both a sapphirebased process and a bulk GaN substrate process [20], and has worked with Sony in developing and licensing GaN LD technology. Sony, along with Sharp and Sanyo, are developing LDs using a GaN substrate-based manufacturing process [21]. Significant challenges in the manufacturing of GaN-based LDs slowed the ramp to volume production and delayed the delivery of the Sony Playstation 3 and Blu-ray players. The main problem in manufacturing GaN LDs is getting a high production yield, which is reportedly due in part to the limited availability of a high-quality bulk GaN substrate [22]. Manufacturing approaches for GaN LDs based on sapphire have higher dislocation densities and require careful placement and alignment of the devices on the substrate. While virtually no data has been publicly released regarding the manufacturing yields for GaN-based LDs on sapphire or bulk GaN substrates, it is expected that the use of a high quality bulk GaN substrate will improve the yield of an optimized epitaxial growth process due to lower defects, improved uniformity, and simplified epitaxial growth processes. Empirical evidence of this is seen in the move from sapphire to GaN substrates by all LD manufacturers, particularly for high-power LDs. Markets for the GaN LDs capture the several key drivers for success in bulk GaN substrate commercialization: Strong and growing commercial demand for advanced III-N semiconductor
devices for a broad range of applications Improvements in bulk GaN materials properties that enable the device structural
quality and thermal properties not attainable through other approaches Demonstration of high-performance devices with an advantage derived from the
properties of the bulk GaN substrate GaN substrate manufacturers will use these market drivers to justify the deployment of resources for increasing the manufacturing capacity that results in the increased availability of high-quality bulk GaN substrates. This supply of substrates enables growth in the commercial device market, which motivates further substrate capacity expansion and makes possible manufacturing yield improvements. As this GaN substrate manufacturing sequence progresses, other new device applications can benefit
1
Development of the Bulk GaN Substrate Market
11
700 000 # of 4" GaN wafers to process per year for power electronics applications in units
600 000
4" wafers (Units)
500 000 400 000 300 000 200 000 100 000 0 2007
2008
2009
2010
2011
2012
2013
2014
2015
Fig. 1.2 400 GaN wafer forecast for power electronics applications Source: Yole Développement
from the increased availability of bulk GaN substrates through the reduction of R&D costs and a stable supply of substrates. When new devices are developed and demonstrated and new areas of commercial demand are generated, this can provide further motivation for substrate manufacturers to increase manufacturing capacities. For example, Fig. 1.2 shows the forecast for 400 GaN wafer volume for power electronics applications, with unit volumes at nearly 600,000 in 2015. This volume forecast represents a tremendous opportunity for GaN substrates, but will require significant cost reduction in order to meet expected cost requirements.
1.3 Benefits and Importance of Bulk GaN Substrates Within the different families of semiconductor materials, we observe that essentially all silicon, SiC, GaAs, and InP devices are fabricated on their associated bulk substrate, in spite of many attempts to develop SiC and especially GaAs and InP devices on cheaper substrates such as silicon and sapphire. The situation for nitride devices is different: most commercial nitride semiconductor devices are currently fabricated on foreign substrates such as Si, SiC, and sapphire, which reflects the slow rate of progress in the worldwide effort to develop a source of costeffective, high-quality bulk GaN substrates, and the resulting limited availability of these substrates. The selection of foreign substrates for the development and commercialization of GaN devices represents a compromise in the suitability of the substrate as a starting material for GaN epitaxial growth. In spite of some advantages of foreign substrates, which include readily available substrates of large size,
12
A.D. Hanser and K.R. Evans
1µm
0.25µm
Fig. 1.3 Atomic force microscopy images of epitaxial GaN films on sapphire (left) and on a bulk GaN substrate (right). The dislocation density of the film grown on sapphire is >109 cm2 and the steps on the surface are pinned by threading dislocations, resulting in nonparallel growth steps. The dislocation density of the film grown on bulk GaN is approximately 106 cm2 and the monolayer steps are nearly parallel and not pinned Source: A. Allerman, Sandia National Laboratories, and R. Dupuis, Georgia Institute of Technology
high quality, and low cost, there are a number of well-known disadvantages to this heteroepitaxial approach: lattice mismatch, thermal expansion mismatch, and chemical incompatibility between the substrate and the epilayers. Nevertheless, due to years of development with these foreign substrate approaches, epitaxial growth processes have been developed that have allowed the commercialization of GaN-based LEDs for many applications. This development path has also led to a number of novel epitaxial growth approaches and substrate configurations that reduce somewhat the dislocation density of the heteroepitaxially grown GaN layers. Examples of these techniques include epitaxial lateral overgrowth (ELOG) and pendeo- or cantilever-epitaxy, but each of these adds to the cost and complexity of the epitaxial process. Many LED manufacturers are transitioning to patterned sapphire substrates (PSS) which improves performance with only a modest cost increase (Fig. 1.3). The main benefits of the bulk GaN substrate specifically address the weaknesses of the foreign substrate approaches. Indeed, the driving force for bringing GaN substrates to the market is strong because of the inherent benefits that most device applications will realize through the application of a bulk substrate. How a bulk GaN substrate influences the cost of making a device is more complex and we discuss that in some detail below.
1.3.1 Device Performance For device performance optimization, there is a simple rule: the best devices come from the highest quality epitaxial device layers. High quality in this case means several things, but most notably that the epitaxial layer structure matches very well to the original design, and that structural imperfections are not present, at least
1
Development of the Bulk GaN Substrate Market
13
significantly. High-quality device epitaxy, in turn, depends on the combination of substrate choice plus the detailed and optimized epitaxial growth approach. As discussed above, a high-quality device epiwafer means high structural quality: accurate thicknesses and compositions of the various device layers, flat heterointerfaces, abrupt doping profiles, and low defect densities. While the primary benefit of a bulk GaN substrate for essentially all GaN devices could simply be stated as that of device epilayer structural quality, it is also sometimes difficult to understand how improved structural quality might impact the performance or reliability of a specific device and application. The answer lies to a large extent in what are the most important properties of the epitaxial device structure and then how the bulk GaN substrate affects those properties. For example, for blue/near-UV laser diodes for optical storage, device output power and reliability must meet certain threshold values for commercial acceptance. Systematic studies have shown that laser diode reliability depends strongly on defect density in the device active regions for a given output power and is accordingly benefited by use of a bulk GaN substrate. For many other devices there exists a limited amount of evidence that a bulk GaN substrate is beneficial. Here, we present in Table 1.4 a summary of demonstrated and predicted benefits of bulk GaN substrates for the more common device applications today, and we examine the benefits of bulk GaN substrates for several specific applications.
Laser Diodes Significant progress has been made in GaN-based LDs since the first demonstration by Nakamura et al. [23]. Much of this work was developed and demonstrated through crystal growth using sapphire as the growth substrate. As a consequence of this heteroepitaxial development approach, the high dislocation density and biaxial strain in the GaN layers has hindered the improvement in the performance of these nitride-based laser diodes. Dislocations can cause deterioration in the operation of quantum well based LDs mainly by three mechanisms [24]: (a) by serving as nonradiative recombination centers for electrons and holes leading to heat generation instead of optical emission; (b) by introducing fast diffusion along the dislocation lines, smearing out quantum wells and shorting p–n junctions; and (c) by disturbing the epitaxial growth front, so that atomically flat structures cannot be obtained. Dislocations also cause an increase in the threshold current density of the laser diode and ultimately limit the lifetime of the devices [25, 26]. It is well known that the device lifetime is increased when LDs are fabricated on the lower dislocation regions realized by the epitaxial lateral overgrowth (ELOG) technique, which can produce GaN with the dislocation density down to the mid-106 cm2 level [27, 28]. Improved results for LDs fabricated on hydride vapor phase epitaxy (HVPE)-grown bulk GaN substrates with dislocation density of less than 105 cm2 have been demonstrated [29]. Also, as discussed previously, for high power LDs, the thermal conductivity of the device affects performance and reliability. The failure of LDs is closely related to the heat generated by joule heating or nonradiative recombination [28]. Thermal issues in LDs are discussed further below.
14 Table 1.4 Summary of bulk GaN substrate demonstrated applications Device application Demonstrated benefits Laser Increased lifetime; reduced threshold current; higher output power; reduced leakage current; improved thermal conductivity; reduced dislocation enhanced diffusion LED Reduction of nonradiative recombination centers; reduced leakage current; improved thermal conductivity; improved efficiency; decreased In segregation associated with dislocations; reduced aging effects caused by electromigration of metals along dislocations RF FET Greater channel mobility, greater channel charge, greater channel mobility charge product; reduced buffer leakage; reduced surface states Power switches and High breakdown voltage; faster diodes switching speed; higher efficiency; reduced reverse recovery time Photodetectors, solar Low dark currents in cells photodetectors and solar cells; Geiger mode photodetector operation
A.D. Hanser and K.R. Evans and predicted benefits for device Predicted benefits Increased power capabilities and power density
Further improvements in IQE; reduced efficiency droop at longer wavelengths; increased output powers
Higher operating frequencies; higher operating powers; greater reliability
Higher reliability; lower on-resistance; greater power carrying capability; higher channel mobility Improved detectivity; improved photocurrent; improved efficiency
UV and Visible LEDs GaN-based LEDs cover a wide range of wavelengths and it has been shown that device performance (total lumen or power output, and device efficiency) depends to a large extent on the output wavelength and the current density of the device. Blue LEDs, which are combined with phosphors to produce the majority of white light LEDs produced today, perform at the highest efficiency level of GaN-based LEDs. Devices operating at shorter (UV) or longer (green) wavelengths have lower efficiencies. There is still extensive discussion on what mechanisms lead to the changing internal quantum efficiency (IQE) with different wavelengths and operating conditions, but these are the main issues that hinder LED performance. While there are several proposed mechanisms to explain the decrease in LED efficiency with increasing current, including carrier injection and confinement effects [30, 31] piezoelectric polarization effects [32], and Auger loss [33], there is also evidence that the material quality of the device may play a large role in governing the LED
1
Development of the Bulk GaN Substrate Market
15
performance. For visible wavelengths, the emission mechanisms in the InGaN active regions of the LED determine the IQE of the device. At blue wavelengths, the InGaN quantum wells require 15–20% indium content, while the indium content is in the range of 25–35% for green wavelengths. These compositional changes and associated material quality differences are believed to contribute to the efficiency drop seen at longer wavelengths [34]. InGaN is a notoriously difficult material to grow due to its low decomposition temperature, a reported miscibility gap [35], and increasing lattice mismatch with GaN as the indium content is increased. Material quality tends to decrease with increasing indium content, and compositional uniformity in quantum wells tends to decrease as well. Indium-rich regions form in quantum wells and result in carrier localization, which enhances efficiency at blue wavelengths but at longer wavelengths the effective localization is reduced, allowing higher indium content LEDs to be more susceptible to nonradiative recombination at dislocations in InGaN [36]. V-defect formation in the growth of InGaN has also been shown to be associated with threading dislocations [37], and it is believed that reducing the density of such defects will improve the performance of green LEDs [38]. Growth of InGaN via MBE on bulk GaN substrates has been shown to have a layer-by-layer growth mode when no dislocations are present [39]. Reducing the dislocation density for InGaN growth can improve its material quality and should improve LED performance. Nonradiative recombination is also an issue for UV LEDs, where the IQE is lower than visible wavelength LEDs due to the lack of carrier localization [34] and the typically high number of dislocations. UV devices operating near the GaN bandedge benefit significantly from a GaN substrate, while deep UV LEDs .200 nm < < 300 nm/ may benefit from an AlN substrate [40].
RF FETs RF transistors built of GaN on SiC show reliability issues that have been shown to be related to trapping of carriers in the devices [41], but presently there is only limited proof that a device fabricated on bulk GaN substrates has improved reliability. A notable study involving bulk GaN substrates showed improved GaN FET reliability when growing on bulk GaN vs. SiC substrates [42]. Recent studies by the U.S. Naval Research Laboratory researchers on semi-insulating GaN showed good X-band RF FET behavior and also showed that improved sheet resistance and higher carrier mobilities were generally observed in comparison to similar structures grown on SiC substrates [43]. Ongoing research at Kyma Technologies and elsewhere may provide additional evidence as more bulk semi-insulating GaN substrates are grown on and processed into transistors for high-speed, high-power RF applications.
High-Power Switches High-power switching devices based on GaN substrates such as GaN Schottky diodes show defect density dependent performance and therefore should benefit
16
A.D. Hanser and K.R. Evans
from the lowest possible defect density. Pure screw dislocations have been identified as a path for reverse-bias leakage current in GaN [44, 45]. Growing on a GaN substrate will reduce the total dislocation density and, therefore/hence, the screw dislocation density at the surface, reducing the gate leakage current. Schottky barrier diodes have been fabricated that have shown some benefits of using bulk GaN over sapphire as the substrate material, where improvements in the reverse recover time, on-resistance, reverse breakdown, and reverse leakage current were reported [46,47]. Recent activities by several groups in the power electronics industry, including Velox and International Rectifier, show strong motivation to build GaN power switching devices on Si and sapphire, yet bulk GaN may be required to significantly further advance the performance levels. Several papers by Japanese researchers over the past two years suggest an intense move to develop bulk GaN based power switching devices [48–50].
New Device Applications There is an ample amount of R&D into new device applications for GaN materials including terahertz applications and quantum well intersubband transition based devices such as the quantum cascade laser (QCL) initially developed at Bell Labs [51]. There are plenty of reasons to believe that bulk GaN will be important in these new applications. Indeed, past failed attempts to build a good GaN-based QCL were ascribed to the lack of a good GaN substrate [52]. Other applications include photovoltaics, where researchers have developed UV-blind avalanche photodetectors [53] and InGaN-based solar cells that offer the potential to develop ultra-high-efficiency devices [54]. Reduced dislocation densities enabled by GaN substrates are important for reduced dark currents and improved solar cell photoresponse.
1.3.2 Thermal Conductivity Many of the targeted high-end applications using GaN-based devices require high output power and high current density. Luminaires for SSL call for hundreds and thousands of lumens of light output, which require high current levels in highbrightness LEDs, while laser diodes operate at high current densities .kA=cm2 / for high power outputs. RF FETs in power amplifiers for wireless communications and radar target power densities of 5 W/mm and higher, while switches and diodes for power electronics will operate at high currents and current densities (up to hundreds of amps and several hundred A=cm2 ). All these operating conditions result in high temperatures in the devices through self-heating, and the performance of virtually all these devices is significantly affected by elevated temperatures. Changes in the device temperature can change the device performance, including dominant emission wavelength, intensity, and efficiency for optical devices, and on-resistance, gain, and efficiency in electronic devices.
1
Development of the Bulk GaN Substrate Market
17
High-power GaN devices cannot be operated continuously without proper thermal management and an integral part of managing the heat in a device structure is to determine how the heat propagates from the device action region through the epitaxial device layers. Exploring these issues, researchers have investigated the effect of the thermal conductivity of the substrate and the epitaxial layers on the device junction or channel temperature. Published experimental measurements show that the room temperature thermal conductivity (RT-TC) of GaN can be as low as 1:0 W=cm K for highly defective .1010 cm2 / GaN grown on sapphire or SiC; for high-quality .106 cm2 / bulk GaN substrates the RT-TC has been measured up to 2.3–2.5 W/cm K [55, 56]. As a result, the benefit of a highly thermally conductive substrate such as SiC may be reduced due to the high defect density of the buffer layer and device active region. The impact of the thermal conductivity of defective epitaxial AlGaN/GaN layers grown on SiC for RF applications have been investigated using micro-Raman thermography. The high spatial resolution of the micro-Raman technique (in the micron to submicron range) allows one to see temperatures that are significantly higher in the GaN epi than in the underlying SiC substrate, due to a high thermal impedance in the GaN layer. Additionally, a thermal boundary resistance is observed between the GaN and the substrate in AlGaN/GaN electronic devices [57]. These factors contribute to a high channel temperature in AlGaN/GaN FETs for RF applications, which may have an impact on device performance and reliability. Substrate thermal effect comparisons have been tested in AlGaN/GaN RF FET modeling studies for GaN and SiC substrates with experimental tests conducted in parallel [58]. Thermal models show that the introduction of defects in the epitaxial GaN and AlN layers grown on SiC influences the flow of heat to the substrate and increases the peak channel temperature in the device, and microRaman measurements show similar channel temperatures for the two substrate cases. For RF applications, bulk GaN substrates offer the possibility of similar thermal performance to SiC, while improving the material quality and the potential for improving device reliability. For GaN LDs, improved junction temperatures have been demonstrated using bulk GaN substrates relative to sapphire substrates, which are attributed to the improved thermal conductivity of the GaN substrate [28].
1.3.3 Thermally Activated Device Failure Failure in semiconductor devices due to a thermally induced mechanism with a first-order dependence on a particular defect concentration follows the rule: 1 D AŒD exp.Ea =kT /; where A is the preexponential factor, Ea is the activation energy associated with the failure mechanism, ŒD is the concentration of the defect associated with the rate limiting step in the failure mechanism, and T is the temperature in the region
18
A.D. Hanser and K.R. Evans
of failure. Such behavior has been observed in GaAs-based HBTs [59], solar cells [60], and GaAs-based laser diodes [61]. For GaN devices, researchers at Sony have demonstrated the clear first order dependence of GaN laser diode lifetime on threading defect density in the device active region [27]. For devices that fail due to thermal activation of a mechanism that involves a threading defect, bulk GaN offers a great advantage over foreign substrate based approaches. Other benefits may exist as well, based on increased efficiency and improved thermal conductivity. Reliability studies are underway for several other device types, particularly RF FETs [62], and associated informative publications should result over the next couple of years.
1.3.4 Device Cost The cost of manufacturing a device that has specific performance, reliability, and form factor requirements can be a complex function of the costs of starting materials, epitaxial growth equipment, device fabrication steps and device yield. (Packaging costs and yields and performance and reliability testing costs and yields can also be important but are not discussed here.) These cost factors need to be considered when evaluating the benefit of a bulk GaN substrate. For example, a low-performance LED can be made at low cost on sapphire substrates due to a thin epitaxial layer structure and relatively high yields. On the other hand, a high-performance LED structure generally requires a thicker epitaxial layer structure and may have lower yields in epitaxial growth and processing due to tighter specifications, such as lumen output or wavelength variation. Some increases in certain cost factors result in overall cost reductions: some LED manufacturers have moved to slightly more costly patterned sapphire substrates, and this has resulted in improved LED performance, effectively reducing the overall cost of a higher performance LED. Beyond materials cost considerations, process development costs need to be considered for new growth processes. In the development of GaN device technology, III-N epitaxial growth capabilities preceded the availability of bulk GaN substrates. Because of this, GaN device developers used foreign substrates exclusively, working primarily with sapphire, SiC, and silicon. For other semiconductor materials, such as GaAs, SiC, and InP, a native substrate was available during the device development stage. With these materials, a move away from the native substrate may offer the potential for cost savings but can also present challenges in terms of development cost and yield of epitaxy and device processing. In the case of GaN, device manufacturing processes have been optimized on foreign substrates. As GaN substrates become available, device manufacturers must contend with not only an initially more expensive starting substrate, but the additional cost of reoptimizing growth processes for the new substrate. The use of GaN substrates should simplify the growth process when compared to foreign substrates and may improve device yields, but the optimization process will depend on many factors, such as substrate surface preparation, off-cut angle and orientation, doping, etc. Device
1
Development of the Bulk GaN Substrate Market
19
manufacturers will make the switch to bulk GaN substrates when the benefits of the device performance outweigh the costs of the materials and process development. Improved device yield is potentially the biggest cost benefit for GaN substrates but also the least discussed and the most difficult to predict. GaN LD developers quickly discovered that defects in the epilayers dramatically affected yield through poor device performance and reliability, which provided strong impetus to accelerate the development of bulk GaN substrates [22, 27]. Current developers and manufacturers of LEDs, RF FETs, and power switching electronics utilize foreign substrates primarily, which indicates that device performance requirements are currently met with the lower cost manufacturing approaches. As GaN substrates are further developed and as the GaN substrate cost is reduced, we believe that essentially all GaN devices, except perhaps for the lowest performance LEDs, will eventually transition to GaN substrates, where the rate of changeover will depend on the specific device and the cost, quality, and diameter availability of bulk GaN. In some cases, additional cost is incurred when employing a foreign substrate. In LED manufacturing, there is a trend toward removing the substrate to improve light extraction and possibly to improve thermal conductivity [63]. RF FETs fabricated from GaN on Si have thermally conductive epoxies used to contact the backside of the epi nominally to improve heat removal [64]. As discussed earlier, bulk GaN enables good thermal conductivity in both the buffer layer and the device active region, which is to be contrasted with poor thermal conductivity in GaN buffers and device layers grown on foreign substrates, including SiC. History shows that all devices get pushed harder in terms of operating conditions as their respective markets mature, and thus all devices become thermally limited. As a result, while GaN on Si RF FETs operating at a moderate level of performance can be made cheaper today than on bulk GaN, that will change as GaN substrate costs come down and diameter availability increases. In some cases the number of RF FETs used can be reduced by increasing the power per FET, which will eventually be achievable by transitioning to bulk GaN substrates.
1.4 GaN Device Trends for Bulk GaN Substrates 1.4.1 Lasers and LEDs The markets for GaN-based optoelectronic devices are large and will continue to generate new applications for LEDs and LDs. For example, many systems would greatly benefit from the availability of LDs with wavelengths >500 nm, including laser-based video projectors and analytical systems. Extrapolations of performance and material quality requirements toward longer wavelengths based on observed trends in blue/violet optical storage laser development are such that most experts agree that a low-defect-density free-standing GaN substrate is a necessary building block of such a laser diode. However, conventional c-plane quantum well structures
20
A.D. Hanser and K.R. Evans
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 turn give rise to lower carrier recombination efficiency, reduced oscillator strength, and red-shifted emission. The QCSE becomes more pronounced with increasing indium composition in InGaN QWs needed for longer wavelength LEDs. Avoiding the QCSE is important for high efficiency LDs and LEDs and has led to intense investigation of nonpolar and semipolar GaN orientations. Nonpolar orientations of GaN eliminate the spontaneous polarization fields, the absence of which has been confirmed experimentally for both violet laser diodes [65, 66] and blue LEDs [67] and suggest nonpolar substrate orientations may benefit longer wavelength GaN-based LDs. Semipolar orientations mostly eliminate the spontaneous polarization effect and may be easier to achieve high-quality epitaxy [68]. It is believed that a high-quality nonpolar and semipolar GaN substrates represent a potential enabler of GaN-based LDs with wavelengths >500 nm. Based on recent results, power LEDs for SSL will continue to be developed on sapphire substrates for the foreseeable future. Companies such as Nichia, Cree, and Osram continue to report improvements in LED efficacy at increasing output power [69], and in total lumen output [70]. Surface emission LED designs improve light extraction efficiency, and removal of the GaN epitaxial layers from the underlying substrate is simplified through the use of laser lift-off techniques with a transparent sapphire substrate [71]. While many groups are performing R&D on LEDs on GaN substrates, fundamental limits in LED performance due to material quality have yet to be definitively identified, and so the demand for a GaN substrate for LEDs in a production application has not yet materialized. Panasonic has introduced white LEDs based on blue LEDs grown on GaN substrates, which the company says have very high thermal and electrical conductivity that substantially improved LED performance at high current levels [72]. Also, the brightest white light LEDs will continue to be those based on blue LEDs with phosphors until the efficiency of green LEDs is improved to allow RGB white light sources to perform at comparable levels.
1.4.2 Power Switches Normally-off switches for HPE are of growing interest and many market applications will benefit from FET device designs that enable a zero bias off-state and high breakdown voltage and current handling capabilities. Vertical power Field effect transistors (FETs) device designs enable smaller chip sizes at higher voltages than lateral designs and are the preferred design approach in most cases. Vertical devices make a conductive substrate necessary, while low leakage currents and high breakdown voltage require low defect densities. As a result, high-quality conductive bulk GaN substrates are desirable for these devices. Several groups are investigating the use of an AlGaN/GaN high electron mobility transistor (HEMT) structure in power
1
Development of the Bulk GaN Substrate Market
21
FETs to take advantage of the structure’s high carrier mobility and low sheet resistance, although the HEMT structure is a normally-on device [50]. GaN MOSFET designs are also being investigated, although there are many challenges in developing a viable gate dielectric for GaN. Standard SiO2 gate dielectrics have generated devices that exhibit very good channel mobilities [73], and other dielectrics are being investigated for improved performance [74]. A fully integrated approach of high-quality GaN materials enabled by GaN substrates, along with gate dielectrics developed and optimized for the GaN materials system, would make possible highefficiency, high-performance, normally-off FET devices that meet standard design requirements for power electronic circuits.
1.4.3 High-Frequency, High-Power HEMTs GaN-based RF devices for wireless communications systems outperform Si-based devices in terms of operating voltage and power density, but cannot yet match Si in cost. As a result, Si remains very competitive in wireless base station applications in the 2.4–5.8 GHz frequency range. With cost being a large driver for the commercial applications, lower cost manufacturing approaches for GaN RF devices dominate, which include GaN devices fabricated on Si and SiC substrates. These device architectures have delivered the performance and reliability needed to enter the market and are expected to continue to improve with the ongoing development. Thus, the demand for GaN substrates in high-frequency applications is relegated to frequencies starting at 10 GHz. Since the exact relation between RF device performance and material quality is not clearly known, additional development is needed to identify benefits associated with bulk GaN substrates. These higher frequency applications are smaller in market size compared to the LED, LD, and HPE markets, and likely present only niche opportunities for bulk substrate manufacturers [75]. A key challenge for GaN RF FET developers is realization of a normally off device; nonpolar GaN offers clear potential for such an “enhancement mode” or emode FET to be developed. As nonpolar GaN becomes more available, emode GaN FET development will become important.
1.5 Bulk GaN Substrate Trends The bulk GaN substrate industry is just at the start of establishing and expanding techniques for manufacturing substrates for commercial applications. Researchers continue to explore methods of improving and optimizing existing crystal growth techniques for fabricating bulk GaN substrates, while exploring new methods for creating larger, high-quality crystals with the desired structural, optical and electrical properties at an even lower cost. Three main techniques are currently employed for the development and manufacturing of bulk GaN crystals for substrates. We will
22
A.D. Hanser and K.R. Evans
briefly examine these three crystal growth techniques and the trends for their use in producing substrates for GaN devices.
1.5.1 Hydride Vapor Phase Epitaxy The HVPE growth process has several benefits for manufacturing GaN substrates, including high growth rates (up to 300 m=h have been demonstrated for manufacturing), high purity, and doping control, which make bulk materials a possibility and have led to successful substrate commercialization. The HVPE manufacturing method currently dominates in the delivery of GaN substrates to the market in terms of volume. There are a number of challenges, however, that the HVPE process faces in achieving high quality bulk substrates. GaN substrates tend to have either an engineered variation of the quality and material structure of the substrate, or a domain-like structure with a number of subgrains and low angle grain boundaries within the crystal. Due to these defects in the GaN, the crystal quality is nonuniform and the dislocation density can vary across the substrate and from top to bottom. As a result, the materials uniformity of HVPE material has yet to approach that of higher structural quality, albeit smaller, crystals grown by other processes, such as the ammonothermal or solution growth techniques. Additionally, crystal lattice tilt, wafer bow, and precise crystalline orientation control over larger area substrates remain as challenges. Steady improvements to the GaN HVPE process has led recently to the demonstration of long length boules, on the order of 1 cm in the [0001] direction, enabling the development of nonpolar substrates to be sliced from c-plane oriented crystals. These substrates have greatly improved material quality relative to high defect density heteroepitaxial material grown in nonpolar directions, although the size of these nonpolar substrates is small (up to 10 mm10 mm [56]). This remains the primary limitation for this technology at this time. Increasing the size of bulk nonpolar substrates will occur through several developments, including extended c-direction growth or growth on nonpolar GaN seeds. Significant issues remain, however, in maturing both of these processes to achieve longer and larger crystals. Once optimized though, GaN seeded boule growth by HVPE is expected to produce excellent quality material at an attractive cost. More detail on the HVPE growth process for bulk GaN substrates is found in Sect. II of this book.
1.5.2 Ammonothermal Growth The ammonothermal (or solvothermal) growth process is similar to the hydrothermal growth process used for the growth of quartz crystals in supercritical water, where elevated heat and pressure is used to enhance the solubility of the solute. In the ammonothermal growth process, supercritical ammonia in conjunction with
1
Development of the Bulk GaN Substrate Market
23
special mineralizers is used as the solvent to dissolve and transport GaN to seed crystals where growth occurs through recrystallization. Solvothermal processes have been demonstrated on very large scales and with very high crystal quality for other materials, including quartz and zinc oxide (ZnO). The scale of the manufacturing process offers great potential for cost reduction, which is one of the strengths of the approach. Additionally, very high-quality GaN crystals with low dislocation densities and low lattice tilt have been demonstrated using the ammonothermal growth process [76]. Drawbacks to the growth technique include a slow growth rate .100 m=day/, the extended development time for developing the growth chemistry, and scaling the process to obtain the economies of scale needed to reduce the manufacturing cost. Details of the ammonothermal process are examined in Sect. III of this book.
1.5.3 Solution Growth Solution (or flux) growth techniques are liquid phase processes that involve solutions that dissolve the growth feedstock and provide transport of the growth species for recrystallization and growth on seed crystals. In high-pressure solution growth using a gallium melt as the solvent, nitrogen is dissolved at high pressures (1 GPa or 10 kbar) and temperatures .1500 ıC/ to form crystals that are either thin hexagonal platelets or elongated hexagonal needles [77]. In low-pressure solution growth, a solvent containing gallium is used to dissolve nitrogen and transport it to seed crystals where GaN crystallizes and grows. These techniques are typically conducted at low to moderate pressures .50 bar/ and temperatures .750 ıC/ [78], with several variations on the process. The driving force for growth from the solution is a thermal gradient within the growth system and growth occurs through precipitation and recrystallization due to a change in solubility resulting from the change in temperature. While the high-pressure process appears to have significant barriers to mass production, the low-pressure processes show promise. Strengths of the solution growth process are the near-equilibrium growth conditions under which the crystal grows, a higher growth rate than the ammonothermal process (up to 240 m=day [79]), and the low pressure and temperature are beneficial. There is also the potential for scaling the process, as solution growth processes have been demonstrated in other materials systems for high-quality, low-cost, large crystals. Challenges in the development of a GaN solution growth process include a relatively low solubility of nitrogen, complex solvent chemistries, and impurity control. An overview of flux growth processes is presented in Sect. IV of this book.
1.5.4 Combined Growth Techniques While MOCVD templates on foreign substrates have been used extensively as seeds for crystal growth development, researchers are beginning to recognize the possible
24
A.D. Hanser and K.R. Evans
benefits of combining features from different growth techniques to enhance the growth of bulk GaN crystals for substrates. For example, the use of high-quality seeds grown by either the solution growth process or the ammonothermal process can be used to supply seeds for the HVPE process, where higher growth rates and larger existing manufacturing capacities can produce high-quality substrates more rapidly. Additionally, the HVPE process can generate high-purity GaN as a source material for the solution and ammonothermal processes. These methods for crystal growth are just now beginning to be explored.
1.6 Summary The III-N materials enable new semiconductor devices with previously unobtainable performance capabilities in terms of light output, power handling, and efficiency, and these attributes will make possible the reinvention of existing technologies in ways that benefit many facets of our lives. While there are many potential applications for these materials, the biggest applications appear to be light generation and the control of electrical power. Bulk GaN substrates are expected to be used for many devices addressing these market opportunities, but there are numerous challenges facing the adoption of bulk GaN substrate technology. Considerable technology development is needed to enable penetration of GaN devices into the SSL and power electronics markets. The performance of the GaN devices must improve in many areas: LED luminous efficacy at high lumen output; FET and diode breakdown voltage and on-resistance; and power switching reliability, among others. Substrate cost must also be significantly reduced in order to become economically practical in the device applications, which is particularly challenging due to incumbent GaN device technology based on foreign substrates. Bulk GaN substrates provide clear and immediate benefits for GaN-based LD applications over sapphire-based approaches, particularly at higher operating current density, and GaN substrate manufacturers are currently focused on addressing this market. Other benefits for LEDs, RF FETs, power devices, and photodetectors have been shown, but commercialization of these devices requires high volumes of low cost substrates. These improvements will occur through the continued development of bulk growth techniques, such as HVPE, ammonothermal, solution growth, and combinations thereof. As these improvements are made, the potential for III-N devices to do more (more light, more power) with less (higher energy efficiency, less electricity, less waste heat) will be enabled. Acknowledgements The authors would like to thank their colleagues Tanya Paskova and Edward Preble at Kyma Technologies for their valuable contributions to the research and writing of this chapter. We would also like to thank Philippe Roussel of Yole Développement for the power electronics market data and Andrew Allerman of Sandia National Laboratories and Russell Dupuis of the Georgia Institute of Technology for the AFM data.
1
Development of the Bulk GaN Substrate Market
25
References 1. Some Facts about the Lighting Industry Today and Its Potential for Change in the Future, http://www.nglia.org/documents/SSL-Benefits.pdf (October 2008) 2. http://www.netl.doe.gov/ssl/usingLeds/general_illumination_efficiency_luminous.htm (October 2008) 3. http://www.netl.doe.gov/ssl/comm_testing.htm (October 2008) 4. Strategies Unlimited, The Market for High-Brightness LEDs in Lighting: Application Analysis and Forecast – 2008, August 2008 5. http://www.lbl.gov/today/2006/Mar/14-Tue/standby.html (October 2008) 6. J.P. Ross, A. Meier, Whole-house measurements of standby power consumption. In Proceedings of the Second International Conference on Energy Efficiency in Household Appliances, Naples (Italy). (Association of Italian Energy Economics, Rome, 2000). Also published as Lawrence Berkeley National Laboratory Report LBNL-45967, (Rome), 7. Pulling the plug on standby power, The Economist, March 9, 2006 8. A. Raskin, S. Shah, The emergence of hybrid vehicles. AllianceBernstein, Research on Strategic Change, June 2006 9. M. Amin, J. Stringer, The electric power grid: today and tomorrow. MRS Bulletin 33, 399 (2008) 10. J.W. Palmour, In IEEE Compound Semiconductor IC Symposium, 2006 11. T. Hausken, Are GaN & SiC ready for prime time? In IEEE Applied Power Electronics Conference and Exposition, Feb. 24–28, 2008, Austin, TX, USA 12. http://www.businessweek.com/innovate/content/oct2005/id20051021_469296.htm. (October 2008) 13. http://compoundsemiconductor.net/cws/article/news/35847 (October 2008) 14. http://compoundsemiconductor.net/cws/article/news/30529 (October 2008) 15. http://www.sony.net/Products/SC-HP/cx_news/vol49/pdf/sld3131vf_vfi.pdf. (October 2008) 16. http://www.sony.net/Products/SC-HP/cx_news/vol49/pdf/sld3234vf_vfi.pdf. (October 2008) 17. http://www.eetimes.com/showArticle.jhtml;jsessionid=2GCT0UEZ5S0BEQSNDLRSKH0CJ UNN2JV%20N?articleIDD198001549. (October 2008) 18. M. Ikeda, S. Uchida, Phys. Stat. Sol. (a) 194, 407 (2002) 19. http://compoundsemiconductor.net/cws/article/news/32952. (October 2008) 20. R. Stevenson, Compound Semicond. 12, 22 (2006) 21. http://compoundsemiconductor.net/cws/article/business/36163 (October 2008) 22. http://www.eetimes.com/showArticle.jhtml?articleIDD194400941. (October 2008) 23. S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa, T. Yamada, T. Matsushita, H. Kiyoku, Y. Sugimoto, Jpn. J. Appl. Phys. 35, L74 (1996) 24. S. Porowski et al., Blue lasers on high pressure grown GaN single crystal substrates, Europhysics News 35, (3) 69 (2004) 25. S. Nakamura, Science 281, 956 (1998) 26. H.Y. Ryu, K.H. Ha, J.H. Chae, K.S. Kim, J.K. Son, O.H. Nam, Y.J. Park, J.I. Shim, Appl. Phys. Lett. 89, 171106 (2006) 27. M. Ikeda, T. Mizuno, M. Takeya, S. Goto, S. Ikeda, T. Fujimoto, Y. Ohfuji, T. Hashizu, Phys. Stat. Sol. (c) 1, 1461 (2004) 28. O.H. Nam, et al. Phys. Stat. Sol. (a) 201, 2717 (2004) 29. S. Ito, Y. Yamasaki, S. Omi, K. Takatani, T. Kawakami, T. Ohno, M. Ishida, Y. Ueta, T. Yuasa, M. Taneya, Jpn. J. Appl. Phys. 43, 96 (2004) 30. M.-H. Kim, M.F. Schubert, Q. Dai, J.K. Kim, E.F. Schubert, Appl. Phys. Lett. 91, 183507 (2007) 31. H. Masui, H. Kroemer, M.C. Schmidt, K.-C. Kim, N.N Fellows, S. Nakamura, S.P. DenBaars, J. Phys. D: Appl. Phys. 41, 082001 (2008) 32. M. Kunzer, M. Baeumler, K. Köhler, C.-C. Leancu, U. Kaufmann, J. Wagner, Phys. Stat. Sol. A 204, 236 (2007)
26
A.D. Hanser and K.R. Evans
33. Y.C. Shen, G.O. Mueller, S. Watanabe, N.F. Gardner, A. Munkholm, M.R. Krames, Appl. Phys. Lett. 91, 141101 (2007) 34. Y. Yang, X.A. Cao, C. Yan, IEEE Trans. Electron. Dev. 55, 1771 (2008) 35. I. Ho, G.B. Stringfellow, Appl. Phys. Lett. 69, 3701 (1996) 36. Y.-H. Cho, S.K. Lee, H.S. Kwack, J.Y. Kim, K.S. Lim, H.M. Kim, T.W. Kang, S.N. Lee, M.S. Seon, O.H. Nam, Y.J. Park, Appl. Phys. Lett. 83, 2578 (2003) 37. Y. Chen, T. Takeuch, H. Amano, I. Akasaki, N. Yamada, Y. Kaneko, S.Y. Wang, Appl. Phys. Lett. 72, 710 (1998) 38. C. Wetzel, T. Salagaj, T. Detchprohm, P. Li, J.S. Nelson, Appl. Phys. Lett. 85, 866 (2004) 39. Unipress InGaN MBE 40. Z. Ren, Q. Sun, S.-Y. Kwon, J. Han, K. Davitt, Y.K. Song, A.V. Nurmikko, W. Liu, J. Smart, L. Schowalter, Phys. Stat. Solidi (c) 4, 2482 (2007) 41. M. Faqir, G. Verzellesi, G. Meneghesso, E. Zanoni, F. Fantini, IEEE Trans. Electron Dev. 55, 1592 (2008). 42. K.K. Chu, P.C. Chao, M.T. Pizzella, R. Actis, D.E. Meharry, K.B. Nichols, R.P. Vaudo, X. Xu, J.S. Flynn, J. Dion, G.R. Brandes, IEEE Electron Dev. Lett. 25, 596 (2004) 43. D.F. Storm, D.S. Katzer, J.A. Roussos, J.A. Mittereder, R. Bass, S.C. Binari, D. Hanser, E.A. Preble, K.R. Evans, J. Cryst. Growth, 301–302, 429 (2007) 44. J.W. Hsu, M.J. Manfra, R.J. Molnar, B. Heying, J.S. Speck, Appl. Phys. Lett. 81, 79 (2002) 45. B.B. Simpkins, E.T. Yu, P. Waltereit, J.S. Speck, J. Appl. Phys. 94, 1448 (2003) 46. K.H. Baik, Y. Irokawa, J. Kim, J.R. LaRoche, F. Ren, S.S. Park, Y.J. Park, S.J. Pearton, Appl. Phys. Lett. 83, 3192 (2003) 47. Y. Zhou, M. Li, D. Wang, C. Ahyi, C.-C. Tin, J. Williams, M. Park, N.M. Williams, A. Hanser, Appl. Phys. Lett. 88, 113509 (2006) 48. H. Lu, R. Zhang, X. Xiu, Z. Xie, Y. Zheng, Z. Li, Appl. Phys. Lett. 91, 172113 (2007) 49. T. Uesugi, R&D Review of Toyota CRDL 35 1 (2000.6) 50. H. Ueda, M. Sugimoto, T. Uesugi, and T. Kachi, Wide-bandgap semiconductors for automobile applications, In CS MANTECH Conference, April 24–27, 2006, Vancouver, British Columbia, Canada 51. C. Gmachl, H.M. Ng, S.N.G. Chu, A.Y. Cho, Appl. Phys. Lett. 77, 3722 (2000) 52. G. Sun, R.A. Soref, J.B. Khurgin, Superlattices Microstruct. 37, 107 (2005) 53. J.B. Limb et al. Appl. Phys. Lett. 89, 011112 (2006) 54. O. Jani, H. Yu, E. Trybus, B. Jampana, I. Ferguson, A. Doolittle, C. Honsburg, Effect of phase separation on performance of III–V nitride solar cells, In 22nd European Photovoltaic Solar Energy Conference, September 3–7, 2007, Milan, Italy 55. C. Mion, J.F. Muth, E.A. Preble, D. Hanser, Appl. Phys. Lett. 89, 092123 (2006). 56. S. Kubo, K. Fujito, H. Nagaoka, T. Mochizuki, H. Namita, S. Nagao, Bulk GaN crystals grown by HVPE, In Second International Symposium on Growth of III-Nitrides, July 6–9, 2008, Laforet Shuzenji, Izu, Japan 57. A. Sarua, H. Ji, M. Kuball, M.J. Uren, T. Martin, K.P. Hilton, R.S. Balmer, IEEE Trans. Electron Dev. 53, 2438 (2006) 58. A. Sarua, H. Ji, K.P. Hilton, D.J. Wallis, M.J. Uren, T. Martin, M. Kuball, IEEE Trans. Electron Dev. 54, 3152 (2007) 59. A.D. Hanser, E.A. Preble, L. Liu, K. R. Evans, B.K. Wagner, H.M. Harris, Thermal benefits of native GaN substrates, In Advanced Technology Workshop on Advanced Substrates and Next-Generation Semiconductors, April 30–May 1, 2008, Linthicum Heights, Maryland, USA 60. M. Harris, B. Wagner, S. Halpern, M. Dobbs, C. Pagel, B. Stuffle, J. Henderson, K. Johnson, Full two-dimensional electroluminescent (EL) analysis of GaAs/AlGaAs HBTs, In Proc. IEEEIRPS, vol. 121 (1999) 61. C. G. Michel, G. J. Vendura, H. S. Marek, J. Elect. Mat. 16, 295 (1987) 62. G. Mueller, P. Berwian, E. Buhrig, and B. Weinert, GaAs substrate for high power laser diodes, in High Power Diode Lasers, ed by R. Diehl, Topics Appl. Phys. 78, 121 (2000) 63. M.J. Rosker, in IEEE Compound Semiconductor Integrated Circuit Symposium 2007, pp 1–4 64. I.P. Smirnova, L.K. Markov, D.A. Zakheim, E.M. Arakcheeva, M.R. Rymalis, Semicond. 40, 1363 (2006)
1
Development of the Bulk GaN Substrate Market
27
65. http://compoundsemiconductor.net/cws/article/fab/36330. (October 2008) 66. K. Okamoto, H. Ohta, S.F. Chichibu, J. Ichihara, J. Takasu, Jpn. J. Appl. Phys. 46, L190 (2007) 67. M.C. Schmidt et al. Jpn. J. Appl. Phys. 46, L190 (2007) 68. J.P. Liu et al. Appl. Phys. Lett. 92, 011123 (2008) 69. H. Zhong, A. Tyagi, N.N. Fellows, F. Wu, R.B. Chung, M. Saito, K. Fujito, J.S. Speck, S.P. DenBaars, S. Nakamura, Appl. Phys. Lett. 90, 233504 (2007) 70. http://www.ledsmagazine.com/news/5/7/22. (October 2008) 71. http://www.ledsmagazine.com/news/4/9/9. (October 2008) 72. O.B. Shchekin, J.E. Epler, T.A. Trottier, T. Margalith, D.A. Steigerwald, M.O. Holcomb, P.S. Martin, M.R. Krames, Appl. Phys. Lett. 89, 071109 (2006) 73. http://www.ledsmagazine.com/news/4/3/7. (October 2008) 74. W. Huang, T.P. Chow, and T. Khan, Phys. Stat. Sol (a) 204, 2064 (2007) 75. J.S. Jur, V.D. Wheeler, M.T. Veety, D.J. Lichtenwalner, D.W. Barlage, and M.A.L. Johnson, Epitaxial rare earth oxide growth on GaN for enhancement-mode MOSFETs, in CS MANTECH Conference, April 14–17, 2008, Wheeling, Illinois, USA 76. Yole Développement, 400 epi-wafer needs for GaN-based RF devices market 2005–2012, 2008 77. R. Dwili´nski, R. Doradzi´nski, J. Garczy´nski, L.P. Sierzputowski, A. Puchalski, Y. Kanbara, K. Yagi, H. Minakuchi, H. Hayashi, J. Cryst. Growth 310, 3911 (2008) 78. I. Grzegory, M. Bo´ckowski, S. Porowski, in Bulk Crystal Growth of Electronic, Optical, and Optoelectronic Materials, ed. by P. Capper (Wiley, UK, 2005), Chap. 6, pp. 173–207 79. F. Kawamura, M. Morishita, M. Tanpo, M. Imade, M. Yoshimura, Y. Kitaoka, Y. Mori, T. Sasaki, J. Cryst. Growth 310, 3946 (2008)
Part II
Vapor Phase Growth Technology
Chapter 2
Hydride Vapor Phase Epitaxy of GaN Akinori Koukitu and Yoshinao Kumagai
Abstract Hydride vapor phase epitaxy (HVPE) is the growth method known for a long time, where a halide vapor precursor, such as GaCl, and a hydride, such as NH3 , are used as the group III and the group V precursors, respectively. HVPE growth can achieve a high growth rate and a high crystal quality due to thermal stability of a halide source, the use of high purity starting materials without containing carbon, and a high surface migration of halide molecules. Therefore, HVPE growth attracts attention as a thick film growth method.
2.1 Introduction The first hydride vapor phase epitaxy (HVPE) of GaN was reported by Maruska et al. in 1969, using a sapphire substrate [1]. The growth temperature of GaN was 825 ıC, which is extremely lower than that used for a high quality GaN HVPE growth recently. The GaN epitaxial layer was enough for measuring the band-edge absorption and lattice constant the first time, although epitaxial layer should not be a high quality GaN from the viewpoint of growth temperature. In the 1990s, HVPE growth of GaN attracted attention for the preparation of freestanding GaN wafers. The first thick GaN (approximately 400 m) with a smooth surface was reported by Detchprohm et al. in 1992, using a ZnO buffer layer on sapphire [2]. Although the epitaxial layer of GaN on the ZnO buffer layer was transparent and smooth, the ZnO buffer layer was so thin that the ZnO layer could not be etched off by a solution. From the late 1990s, most of the attention has been focused on the heteroepitaxy of GaN on foreign substrates such as sapphire, Si, NdGaO3 , and GaAs to realize freestanding GaN wafers [3–7]. Usui et al. [6] achieved preparation of a freestanding GaN wafer of 2-in. (approximately 5 cm) diameter size by the use of a new technique, epitaxial lateral overgrowth (ELO), to reduce significantly the density of dislocations, and by a laser irradiation from behind to remove the GaN thick layer from the sapphire substrate. Unfortunately, the laser irradiation method did not seem to satisfy the need for high reproducibility in mass production.
31
32
A. Koukitu and Y. Kumagai
An important characteristic of GaN is that depending on the atomic configuration of the substrate, it crystallizes in either a wurtzite structure having hexagonal symmetry or a zincblende structure with cubic symmetry. From the point of view of atomic configuration, the use of GaAs as the substrate has some advantages, i.e., (a) the atomic configuration on the GaAs (100) surface is equivalent to the cubic GaN structure, (b) the atomic configuration on the GaAs (111) surface is equivalent to the wurtzite GaN structure, (c) the difference of the thermal expansion coefficient between GaAs and GaN is about a quarter of that between sapphire and GaN, and (d) the GaAs substrate is easy to lap and etch off to separate the GaN layer from the GaAs substrate. In this chapter, we describe HVPE of GaN using the GaAs substrate. First, in order to get the information on suitable growth conditions and understand a growth mechanism, a thermodynamic analysis of HVPE is described for GaN. Then, we show the HVPE growth of cubic GaN on GaAs (100) substrates, and the comparison of GaN growth on (111)A and (111)B GaAs substrates. The orientation dependence of GaN buffer layer growth on GaAs is also investigated using density-function theory (DFT) to better understand GaN growth on GaAs. Next, it is shown that the growth of thick GaN layers on GaAs (111)A surface at 1;000 ı C or above is promising for the preparation of a freestanding GaN substrate. Finally, characteristics of undoped and Fe-doped freestanding GaN substrates obtained by removing GaAs (111)A substrates are presented.
2.2 Thermodynamic Analysis on HVPE Growth of GaN Knowledge of the mechanisms that determine the growth rate is important for the growth of compound semiconductors. It provides us with information on suitable growth conditions for the preparation of compound semiconductors, as well as insight into the crystal growth mechanism. We have shown that an equilibrium model is useful for predicting the growth rate and the element incorporation of III–V compound semiconductors grown by vapor phase epitaxy (VPE) [8–17], including metalorganic vapor phase epitaxy (MOVPE), HVPE and molecular beam epitaxy (MBE). In this chapter, thermodynamic analysis on the HVPE growth of GaN is described for understanding the equilibrium partial pressures of vapor species and the driving force for the GaN deposition by HVPE.
2.2.1 Calculation Procedure The following six species were chosen as the necessary vapor species in analyzing the HVPE growth of GaN: GaCl, GaCl3 , NH3 , HCl, H2 and inert gas (IG) such as nitrogen or helium. The chemical reactions which connect the species at
2
Hydride Vapor Phase Epitaxy of GaN
33
the deposition zone are: GaCl.g/ C NH3 .g/ D GaN.s/ C HCl.g/ C H2 .g/ GaCl.g/ C 2HCl.g/ D GaCl3 .g/ C H2 .g/
(2.1) (2.2)
In HVPE, GaCl as the major source material of Ga is generally formed by the reaction between Ga metal and HCl at the source zone heated up about 800–900 ıC. The growth equipment of HVPE becomes complex compared with those of MOVPE and MBE, because two zones called source zone and deposition zone are required in the HVPE growth reactor. On the other hand, metalorganic hydrogen chloride vapor phase epitaxy (MOHVPE) [5, 18] was developed with a view to simplifying the growth reactor. In this system, alkyl gallium (trimethyl or triethyl gallium, TMG or TEG) and ammonia .NH3 / are used as Ga and nitrogen sources, respectively. Alkyl Ga is supplied with H2 , and mixed with HCl in the upstream region called mixing zone of the reactor heated at 750 ıC. The reaction between alkyl Ga and HCl forms gaseous GaCl, which is transported to the downstream region called deposition zone. When GaCl and NH3 are mixed, the final reaction for the GaN growth is the same as that in conventional HVPE growth, as shown in reaction (2.1). The equilibrium equations for reactions (2.1) and (2.2) are K1 D
PHCl PH2 ; PGaCl PNH3
(2.3)
K2 D
PGaCl3 PH2 : 2 PGaCl PHCl
(2.4)
and
The total pressure of the system and the stoichiometric relationship for III-nitride deposition in the vapor phase can be written as X
Pi D PGaCl C PGaCl3 C PNH3 C PHCl C PH2 C PIG ;
(2.5)
and o o PGaCl PGaCl D PNH PNH3 ; 3
(2.6)
where P ı i ’s indicates the input partial pressure of the source materials. In addition, the parameter A and F are introduced. AD
1=2PGaCl C 3=2PGaCl3 C 1=2PHCl ; PH2 C 3=2PNH3 C 1=2PHCl C PIG
(2.7)
F D
PH2 C 3=2PNH3 C 1=2PHCl : PH2 C 3=2PNH3 C 1=2PHCl C PIG
(2.8)
34
A. Koukitu and Y. Kumagai
A is the ratio of the number of chloride atoms to the number of hydrogen and IG atoms in the system. F is the mole fraction of hydrogen relative to IG atoms. These parameters are kept invariant under a given growth condition. Thermodynamically, almost all the NH3 is decomposed into N2 and H2 at temperature higher than 300 ı C. However, it is well known that the decomposition rate of NH3 under typical growth conditions is very slow without a catalyst, and the extent of the decomposition strongly depends on the growth conditions and equipment. Therefore, a parameter ’ is introduced into the calculation as the mole fraction of decomposed NH3 . NH3 .g/ ! .1 ’/NH3 .g/ C ’=2N2 .g/ C 3’=2H2 .g/:
(2.9)
In the following calculation, the change of ’ affects the change of H2 ; N2 , and NH3 partial pressures in the growth system. A similar result is produced by changing the parameter F . Consequently, we performed the following calculation by changing the parameter F, and fixed ’ as 0.03, according to Ban’s experiments in hot-wall quartz reactor at 950 ı C [19]. The equilibrium partial pressures can be calculated using (2.3)–(2.8).
2.2.2 Equilibrium Partial Pressure and Driving Force for the GaN Deposition Variation of the equilibrium partial pressures for each vapor species is shown in Figs. 2.1 and 2.2 as functions of growth temperature and input partial pressure of GaCl. The general features of the equilibrium partial pressures are similar to those of the hydride systems of arsenides and phosphides of Ga and In [20,21]. It is seen that NH3 , GaCl, N2 , and HCl except the hydrogen carrier gas are major vapor species, and their pressures increase with an increase of the input partial pressure of GaCl. Since Ga is transported as GaCl from the mixing zone and GaN is deposited to form HCl, the reaction governing the growth at the substrate may be described as GaCl C NH3 D GaN C HCl C H2 :
(2.10)
Figure 2.3 shows the relationship between the partial pressures versus parameter F, including those of H2 and IG. The partial pressure of IG in the figure involves N2 produced by the decomposition of NH3 . In the present calculation, IG brings about the same effect irrespective of the IG species. The partial pressure of HCl increases at the expense of GaCl while decreasing the mole fraction of H2 in carrier gas. This means that GaN tends to deposit into the solid phase with an increase of the IG mole fraction. Consequently, hydrogen plays an important role in the deposition of GaN. In Fig. 2.4, the driving force for the GaN deposition, PGa ŒP ı GaCl .PGaCl C PGaCl3 / is shown as a function of temperature. The driving force for the deposition decreases with an increase of F . Therefore, it is expected that a higher growth rate is obtained in the inert carrier gas system than in the H2 system. This fact agrees
2
Hydride Vapor Phase Epitaxy of GaN
35
SPi:1.0 atm PGaCl :5.0×10–3 atm V/III:50 F:1.0 a:0.03
1
H2 NH3
10–2
GaCl
Partial Pressure (atm)
N2 HCl
10–4
10–6
10–8 GaCl3
10–10
10–12 500 600 700 800 900 1000 1100 1200
Temperature (°C) Fig. 2.1 Equilibrium partial pressures over GaN as a function of growth temperature. Total pressure: 1.0 atm, input partial pressure of GaCl: 5 103 atm, input V/III ratio: 50, F : 1.0 and ’: 0.03
1
SPi:1.0 atm
Temp.:1000°C F:1.0 a:0.03 H2
10–2
NH3
GaCl N2
Partial Pressure (atm)
V/III:50
HCl
10–4 10–6 10–8
GaCl3
10–10 10–12 10–14 0.001
0.005
0.01
0.015
Input Partial Pressure of GaCl (atm) Fig. 2.2 Equilibrium partial pressures over GaN as a function of input partial pressure of GaCl at 1;000 ı C. Total pressure: 1.0 atm, input V/III ratio: 50, F : 1.0 and ’: 0.03
36
A. Koukitu and Y. Kumagai
1
Temp.:1000°C SPi:1.0 atm PGaCl :5.0×10–3 atm V/III:50 a:0.03 NH3
H2
IG
Partial Pressure (atm)
10–2 GaCl HCl
10–4
10–6
10–8 GaCl3
10–10 0.0
0.2
0.4
0.6
0.8
1.0
F Fig. 2.3 Equilibrium partial pressures over GaN as a function of hydrogen partial pressure in the carrier gas, parameter F at 1;000 ı C. Total pressure: 1.0 atm, input partial pressure of GaCl: 5 103 atm, input V/III ratio: 50 and ’: 0.03. The partial pressure of IG involves N2 , which is produced by the decomposition of NH3
DPGa (atm)
5.0×10
–3
SPi:1.0 atm F:0.0 – 1.0
1.0×10–3
1.0×10–4 500
PGaCl :5.0×10–3 atm a:0.03
V/III:50
F:0.0 0.01 0.05 0.1 0.2 0.4 0.6 0.8 1.0
600
700
800
900 1000 1100 1200
Temperature (°C) Fig. 2.4 Driving force for the GaN deposition as a function of growth temperature with various parameters F . Total pressure: 1.0 atm, input partial pressure of GaCl: 5 103 atm, input V/III ratio: 50 and ’: 0.03
2
Hydride Vapor Phase Epitaxy of GaN
Growth Rate (mm/h)
100
37
experimental calculated
80 60 40
Temp.: 1000°C
20
F :1.0 a :0.03
PNH3 :0.26 atm
2.0×10–3
4.0×10–3
6.0×10–3
8.0×10–3
1.0×10–2
1.2×10–2
Input Partial Pressure of GaCl (atm) Fig. 2.5 Comparison between calculated growth rates and experimental data
well with Ban’s experiments [19]. In [19], it has been shown that the extent of GaCl consumption, i.e., the deposition of GaN, was significantly greater for He as the carrier gas than for H2 . The growth rate, r, under the mass transport type II or diffusion limit at constant pressure is expressed as (2.11) r D KgPGa ; where Kg is the mass transport coefficient [22]. Figure 2.5 shows a comparison between the calculated growth rate using (2.11) and the experimental data reported by Usui et al [6]. In the calculation of the growth rate, the value of Kg was determined to be 1:18 105 m=.h atm/. The agreement between the calculated and experimental growth rate is quite good, and shows that the growth of GaN using HVPE is thermodynamically controlled.
2.3 Cubic GaN Epitaxial Growth on (100) GaAs Substrate [5, 18] In the 1990s, epitaxial growth of cubic GaN was investigated on various substrates [23–28] such as GaAs, Si, SiC, and MgO. The growth of cubic GaN on GaAs (100) substrate had been reported using gas source MBE (GSMBE) and MOVPE [27,28]. Tsuchiya and Hasegawa et al. [29] demonstrated the homoepitaxial growth of cubic GaN by HVPE on cubic GaN/GaAs (100) substrates prepared by GSMBE. They obtained cubic GaN films with a growth rate as high as 1:6 m=h. At that time, it was generally supposed that a low-temperature buffer layer of GaN could not be grown on a foreign substrate using HVPE system since only nonequilibrium growth system realized a GaN buffer layer.
38 Fig. 2.6 Schematic diagram of the metalorganic hydrogen chloride vapor phase epitaxy (MOHVPE) system
A. Koukitu and Y. Kumagai TMG + H2 HCl + H2 NH3 + H2 Furnace Substrate
Exhaust
However, from the present knowledge, the cubic GaN layers grown, at the time of writing of those reports should be a mixture of some wurtzite type structure and zinc blende structure. In this section, the HVPE growth on GaAs (100) substrate is described to grow cubic GaN. It will be seen in the following chapters that the predeposited GaN buffer layer plays an essential role in the subsequent growth of high-quality cubic GaN films on GaAs (100) substrate by HVPE.
2.3.1 Experimental A schematic illustration of the reactor used in this work is shown in Fig. 2.6. The quartz reactor was vertically placed and the growth of buffer layers and epilayers was carried out at atmospheric pressure. The substrate was GaAs (100) which was etched in a solution of 3 W 1 W 1 D H2 SO4 W H2 O2 W H2 O for about 1 min. The MOHVPE system [5, 18] used is a simple-structured HVPE using TMG HCl NH3 H2 system as described above. TMG was coinjected with HCl into a hot-wall reactor to form gaseous GaCl and CH4 . GaN was then deposited on GaAs substrates through the reaction with NH3 . Prior to the growth of cubic GaN, a GaN buffer layer was grown at 500 ı C. Then the substrate temperature was raised to the growth temperature of 850 ı C in the flow of H2 and NH3 . Subsequently, TMG and HCl were supplied simultaneously to grow the GaN layer. Typical input partial pressures of TMG, HCl and NH3 over the substrate were 8 104 atm; 8 104 atm and 1:6 101 atm, respectively.
39
30
c-GaN (200)
GaAs (200)
Hydride Vapor Phase Epitaxy of GaN
Intensity (arb. units)
2
40
50
2q (deg) Fig. 2.7 XRD profile of the GaN film grown on (100) GaAs
2.3.2 Cubic GaN Growth The growth rate of the GaN films obtained under these conditions was about 3:0 m=h. Figure 2.7 shows a typical X-ray diffraction (XRD) profile for the obtained GaN films. The strong diffraction peak at about 40:0ı corresponding to (200) of cubic GaN can be seen for the grown layer. However, a quality of cubic GaN cannot be evaluated only by a typical XRD, ™ 2™ method. In addition, the full-width at half-maximum (FWHM) of XRD rocking curve of the (200) cubic GaN was about 37.7 min. When the growth was carried out without the GaN buffer layer at high temperature, the GaAs substrate broke into pieces due to the reactions with HCl and NH3 . These findings indicate that the buffer layer prevents the reaction between the GaAs substrate and HCl or NH3 , which leads to high-quality GaN films. Figure 2.8 shows the FWHM values of the XRD rocking curve of the (200) cubic GaN with the epitaxial layer thickness of about 3 m as a function of the thickness of the GaN buffer layer on GaAs (100). The FWHM value initially decreases with increasing thickness of the GaN buffer layer up to 30 nm and then gradually increases. These results indicate that the optimum thickness of the GaN buffer is around 40 nm. GaN epitaxial layers with about 30 nm thick GaN buffer layer were observed by high resolution TEM and electron diffraction patterns to evaluate the quality of the GaN epitaxial layers. Three samples, A, B and C were evaluated. Sample A was taken out from the growth system just after the growth of a 30 nm thick GaN buffer layer, sample B was taken out after annealing at 850 ı C for 10 min and sample C was taken out after growth of the GaN epitaxial layer. Figure 2.9a shows a high-resolution TEM image observed from the [110] direction for sample A. Grain structures consisting of many small polycrystals with various orientations are observed in the GaN buffer layer (arrow 1 in Fig. 2.9a). The interface between the GaN buffer layer and the GaAs substrate is almost flat, but dark regions caused
40
A. Koukitu and Y. Kumagai PTMG :8.0×10–4 atm PHCI :8.0×10–4 atm
140
V/III:200
120
FWHM (min)
100 80 60 40 20 0
0 20 40 60 80 100 120 140 Thickness of GaN buffer layer (nm)
Fig. 2.8 The FWHM values of the X-ray rocking curve of cubic GaN (200) as a function of the GaN buffer layer thickness
a
GaN buffer layer GaAs substrate
b
5 nm
Fig. 2.9 (a) High-resolution TEM image and (b) electron diffraction pattern of sample A
2
Hydride Vapor Phase Epitaxy of GaN
41
by stress are observed in part of the GaAs near the GaN buffer layer (arrow 2 in Fig. 2.9a). The stress may be due to the difference in the lattice parameters between GaAs and GaN. Figure 2.9b shows an electron diffraction pattern taken at the interface with the incident beam along the [110] zone axis of GaAs in sample A. The pattern consists of strong diffraction spots (arrow 1 in Fig. 2.9b) for the GaAs substrate and a ring pattern (arrow 2 in Fig. 2.9b) for the GaN buffer layer. These findings show that the GaN buffer layer consists mainly of polycrystals after the growth at 500 ıC. Figure 2.10a shows a TEM image of sample B. A TEM image and an electron diffraction pattern of the interface for sample C are very similar to those of sample B (not shown here). The dark regions observed in the interface in Fig. 2.9a have disappeared due to stress relaxation by thermal annealing. No (111) facet can be observed near the interface region of the GaAs substrate, and the GaAs interface is fairly flat. The period of the dark lines (arrow 1 in Fig. 2.10a) is 4 in terms of the dot spacing in the GaAs image, and 5 in terms of that of the GaN image [30]. The dark lines (arrow 1 in Fig. 2.10a) indicate misfit dislocations due to the difference in the lattice parameters. Therefore, the lattices of GaN are almost strain-free and form single cubic crystals with some dislocations in the direction (arrow 2 in Fig. 2.10a). Also, dark and bright regions are observed in the GaN buffer layer near the GaAs substrate (arrow 3 in Fig. 2.10a), implying that the GaN buffer layer near the GaAs substrate consists of a set of crystals having slightly different orientations. Figure 2.10b shows an electron diffraction pattern observed from the [110] direction for sample B. It is seen that the ring pattern observed in sample A changed to a spot pattern (arrow 1 in Fig. 2.10b). This finding indicates that the GaN buffer layer becomes a single crystal upon thermal annealing. Line patterns caused by stacking faults in the direction are also observed (arrow 2 in Fig. 2.10b). These
a
GaN buffer layer GaAs substrate
b
Fig. 2.10 (a) High-resolution TEM image and (b) electron diffraction pattern of sample B
5 nm
42
A. Koukitu and Y. Kumagai
results show that thermal annealing after the growth of a GaN buffer layer plays an important role in changing the polycrystalline GaN to a single crystal.
2.4 Comparison of GaN Growth on (111)A and (111)B GaAs Substrates [7, 31] There have been several attempts to prepare a bulk GaN crystal to avoid defect generation. Porowski et al. reported the synthesis of GaN bulk crystal under high pressure and high temperature [32]. Kurai et al. employed the sublimation method [33]. However, GaN crystals obtained by these methods are too small for use as substrate at this stage. Another potential method for a GaN substrate preparation is to use a high growth rate of HVPE [6, 34, 35]. Large freestanding GaN substrates have been prepared by separating thick HVPE GaN layers from sapphire substrates [35]. However, it is very difficult to separate the GaN layer from the sapphire substrate because sapphire is very hard and stable. Growth of GaN on GaAs (111) substrate is very attractive, since the GaAs starting substrate can be easily removed by aqua regia after the GaN growth. In addition, the polarity of wurtzite GaN (0001) grown on GaAs (111) is expected to be controlled by the polarity of GaAs used, i.e., GaN (0001) (Ga-polarity) on GaAs (111)A N (N-polarity) on GaAs (111)B. However, there is the problem of and GaN .0001/ how to prevent deterioration of GaAs due to the unstable nature of GaAs at high temperature around 1;000 ıC (widely used temperature for the epitaxial growth of GaN). Hasegawa et al. introduced a 3 m thick GaN intermediate layer grown at 850 ıC on a 50 nm thick GaN buffer layer to prevent deterioration of a GaAs substrate at high temperature, and obtained a hexagonal GaN layer with a fairly smooth surface at 1;000 ıC on GaAs (111)B [4]. However, there is no report of the GaN growth on GaAs (111)A surfaces at temperatures as high as 1;000 ı C. In this chapter, comparison of the GaN buffer layers and the GaN epitaxial layers grown on GaAs (111)A and (111)B surfaces is described.
2.4.1 Experimental The MOHVPE system used in this study employs a vertical quartz reactor as shown in Fig. 2.6. A GaAs (111) substrate with both the sides polished was used and placed parallel to the gas stream so that GaN could be grown on both GaAs (111)A and (111)B surfaces simultaneously. The GaAs substrate was first cleaned in a solution of HCl W H2 O D 1 W 19 for 1 min and then set into the reactor. TMG and NH3 were used as Ga and N sources. The mixing zone of TMG and HCl was maintained at 750 ıC throughout the entire growth procedure to form gaseous GaCl by the reaction between TMG and HCl. GaN was deposited on the both surfaces of GaAs substrate in the downstream region, where GaCl and NH3 were mixed.
2
Hydride Vapor Phase Epitaxy of GaN
43
First, a thin GaN buffer layer of about 50 nm thick was grown on both sides at 550 ıC. The growth was performed under atmospheric pressure and the input partial pressures of TMG, HCl and NH3 were 8 104 atm; 8 104 atm and 1:6 101 atm, respectively. Then the substrate temperature was raised to 1;000 ıC in an NH3 ambient. Subsequently, TMG and HCl were supplied again and the growth of GaN was performed for 1 h. The input partial pressures of TMG, HCl and NH3 at this stage were 5:8 103 atm; 5:8 103 atm and 2:9 101 atm, respectively, where the growth rate of GaN was about 25 m=h.
2.4.2 Comparison of GaN Growth on (111)A and (111)B GaAs Surfaces Figure 2.11 shows the thickness of the low temperature GaN buffer layers grown on GaAs (111)A and (111)B surfaces as a function of growth time. Although the growth rate of buffer layers on GaAs (111)A and (111)B surfaces is almost the same after about 3 min of growth, the thickness on GaAs (111)A surface is about 1.3 times larger than that on GaAs (111)B surface due to the difference in growth rate in the early stages of the growth. The reason for this result may be related to the nucleation manner of the buffer layer. From this point of view, GaN initial growth processes on GaAs (111)A and B surfaces will be described in the next chapter. Figure 2.12 shows AFM images of as-grown buffer layers on GaAs (111)A and (111)B surfaces after 15 min of growth. The difference in surface morphology is clearly seen. The buffer layer grown on GaAs (111)A surface shows a smooth sur-
180 160
Thickness (nm)
140 120
on GaAs(111)A
100 80
on GaAs(111)B
60 40 20 0
0
5
10
15
20
25
30
Growth time (min) Fig. 2.11 Thickness of the GaN buffer layers grown by MOHVPE as a function of growth time: on GaAs (111)A surface (circles) and on GaAs (111)B surface (triangles)
44
A. Koukitu and Y. Kumagai
a
b
nm 25
nm 25
0 0.5
0 0.5
0.5
0.25
0.25
μm
0.5
0.25
0.25
μm
μm
0 0
μm
0 0
Fig. 2.12 AFM images of as-grown GaN buffer layers after 15 min of growth: (a) on GaAs (111)A and (b) on GaAs (111)B surface
a
b
5m
m
Fig. 2.13 Photographs of the GaAs substrate covered with a 50-nm-thick GaN buffer layer grown at 550 ı C and heated to 1;000 ı C in NH3 ambient. (a) on GaAs (111)A and (b) on GaAs (111)B. The word “GaN” is projected onto the surface
face and covers the GaAs surface homogeneously. Peak-to-valley height of this surface is within 5% of the grown layer. On the other hand, the GaN buffer layer on GaAs (111)B surface shows a very rough surface with many pits. Peak-to-valley height of this surface is about 30% of the grown layer. Also, there are some pinholes reaching the GaAs substrate on the surface. Next, we investigated the surface feature of 50 nm thick GaN buffer layers grown on GaAs (111)A and (111)B surfaces after heating in NH3 ambient up to 1;000 ı C. Photographs of the substrate after heating are shown in Fig. 2.13. In this figure, the word “GaN” has been projected onto the surface in order to reveal the surface features clearly. The difference in surface morphology between the two surfaces is very clear. The surface on the GaAs (111)A still appears to be mirror-like even after 1;000 ıC heating (Fig. 2.13a). In addition, there is no change in the GaAs substrate beneath the GaN buffer layer. On the contrary, numerous pinholes are seen in the GaN buffer layer grown on the GaAs (111)B surface (Fig. 2.13b). Furthermore, serious deterioration of the GaAs substrate beneath the GaN buffer layer is observed. This may be caused by arsenic (As) desorption from the GaAs substrate through the pinholes. The above results suggest two important points. One is the difference between the surface polarities of the GaN buffer layers grown on GaAs (111)A and (111)B surfaces, and the other is the possibility of high-quality GaN layer growth on GaAs (111)A surface. It was found that the uniform GaN layer with a mirror-like surface can be grown at 1;000 ıC on the 50 nm thick GaN buffer layer on GaAs (111)A surface.
2
Hydride Vapor Phase Epitaxy of GaN
45
Fig. 2.14 SEM image of the top and cleaved surface of a GaN layer grown at 1;000 ı C for 1 h on a 50-nm-thick GaN buffer layer grown on a GaAs (111)A
Figure 2.14 shows a scanning electron microscopy (SEM) image of the top and cleaved surface of GaN layer grown at 1;000 ı C for 1 h on the 50 nm thick GaN buffer layer grown on the GaAs (111)A surface. It is clearly seen that the surface of the grown layer is mirror-like, although there are a few hexagonal pits. As not shown here, the surface behind the substrate (GaAs (111)B surface) is seriously damaged due to the deterioration of the 50 nm thick GaN buffer layer at 1;000 ı C. Another interesting point we can see from Fig. 2.14 is that the porous shape of the GaAs substrate proceeds during GaN growth at 1;000 ıC from the surface behind the substrate, i.e., from the GaAs (111)B surface.
2.5 Ab Initio Calculations of GaN Initial Growth Processes on (111)A and (111)B GaAs Surfaces [36] As described in the preceding section, the GaAs (111)A substrate covered with a 50 nm thick GaN buffer layer did not deteriorate even after subsequent heating up to 1;000 ı C, whereas numerous holes were formed in the GaN buffer layer that was grown on GaAs (111)B surface. However, little is known why the GaN buffer layer on the GaAs (111)A does not deteriorate at 1;000 ı C. Therefore, the growth mechanisms, especially between GaN buffer layer and GaAs (111) surfaces, should be investigated to better understand GaN growth on GaAs. In this chapter, we show the orientation dependence of GaN buffer layer growth on GaAs by employing density-function theory (DFT). At the initial stage, we
46
A. Koukitu and Y. Kumagai
assume that As atoms on a GaAs surface react with hydrogen in the carrier gas, and then As atoms desorb from GaAs surface through AsHn molecules. In the next stage, Ga atoms on the surface react with NHn (NH3 , NH2 , or NH) to form GaN on the GaAs surface.
2.5.1 Calculation Procedure The total energy calculations and geometry optimization were performed employing DFT [37–39] within the generalized gradient approximation (GGA) [40–42] and the first-principles pseudopotential approach [43]. The energy cutoff value, which determines the number of plane-waves in the basis set, was 15 Ry for GaAs. The calculations were carried out using the package fhi98md [44]. The models used for the ab initio calculations represented the As-terminated GaAs (111)A and (111)B surfaces by supercell geometries with four GaAs atomic layers containing eight Ga atoms and eight As atoms per layer, and 20 Å vacuum region. The Ga atoms in the bottom layer are terminated by fictitious H atoms with 1.25 elementary charges in order to form perfect covalent bonds at the bottom surface of GaAs [45]. The initial structures of the models were obtained from ab initio calculations, where only the positions of fictitious hydrogen atoms terminated on the surface were varied. The positions of Ga and As atoms were fixed to keep the crystal structure of bulk GaAs. Then, in order to perform energy minimization, the position of Ga and As atoms in the top three layers were varied, and the positions of As precursors over the GaAs surface were varied along the a and b axes. The positions of Ga atoms in the fourth layer and the optimized fictitious H atoms were fixed in their bulk GaAs positions.
2.5.2 GaN Initial Growth Processes on (111)A and (111)B GaAs Surfaces The surface on the GaAs (111)A still appeared to be mirror-like even after 1;000 ıC heating as shown in Fig. 2.13a, whereas numerous pinholes were seen in the GaN buffer layer grown on the GaAs (111)B surface as shown in Fig. 2.13b. Therefore, we assumed that the As atoms, which remain on the GaAs surface, prevent N precursors such as NH3 ; NH2 , and NH from reacting with Ga atoms on the surface, and we calculated the As desorption energies from the GaAs (111)A and (111)B surfaces by ab initio calculations. First, the GaAs (111)A and (111)B surfaces before As desorption were optimized. In VPE, the uppermost atoms on the GaAs surface might be adsorbed by hydrogen atoms because hydrogen is generally used as a carrier gas. Therefore, we selected four models as the initial surfaces; Model 1: the bare surface of GaAs (111) As plane, and the GaAs (111) As surfaces terminated with monohydride (Model 2), dihydride (Model 3), and trihydride (Model 4). Next, we simulated the As
2
Hydride Vapor Phase Epitaxy of GaN
47
Potential energy change(eV)
–2147
Non-hydride (-As)
–2148 –2149
461 kJ/mol
Monohydride (-AsH)
As
–2150
400 kJ/mol
Dihydride (-AsH2)
(111)A –2151
164 kJ/mol
Trihydride (-AsH3)
–2152
71.9 kJ/mol
–2153 –2154
0
1
2
3
4
5
6
Distance from surface (Å) Fig. 2.15 Potential energy changes for the desorption processes of AsHn molecules (n D 0, 1, 2, 3) from nonhydride .As/, monohydride .AsH/, dihydride .AsH2 / and trihydride .AsH3 / surfaces of GaAs (111)A
desorption process from these initial surfaces, where it was thought that As atoms on the four surfaces were removed as As, and AsHn with n D 1–3. In the calculations, As or AsHn were moved from the initial surface perpendicularly to the surface in steps of about 1.0 Å, and then the total energies and the optimized structures were obtained for each height of the molecule from the surface. Figure 2.15 shows the potential energy change in the As desorption process on the GaAs (111)A surface with respect to the distance between the GaAs (111)A surface and the desorbed AsHn (n D 0, 1, 2, 3) molecules. The total energy of the initial monohydride surface .AsH/ is the smallest, and then comes trihydride .AsH3 /, nonhydride .As/, and dihydride .AsH2 / in that order. However, the differences of the potential energies in the initial surface are small. As shown in the figure, the activation energies on the GaAs (111)A surface are 71.9 kJ/mol for desorption of AsH3 molecules from the AsH3 surface, 164 kJ/mol for that of AsH2 from the AsH2 surface, 400 kJ/mol for that of AsH from the AsH surface, and 461 kJ/mol for As from the As surface. These results directly suggest that As atoms on GaAs (111)A desorb from the GaAs surface in AsH3 molecules, and it is likely that As atoms completely desorb from the surface at about 550 ı C before the growth of GaN buffer layer because the activation energy of the reaction is extremely small. The potential energy change in the As desorption process from GaAs (111)B is shown in Fig. 2.16. The initial surfaces are AsH2 ; AsH; As and AsH3 surfaces in increasing order of the value of total energy. The activation energy for desorption of AsH3 from the GaAs (111)B surface is shown with a negative value. However, the AsH3 desorption directly from the AsH3 surface does not occur, because the formation of the AsH3 surface is impossible due to the large potential
48
A. Koukitu and Y. Kumagai –1954 Non-hydride (-As)
Potential energy change(eV)
–1955 –1956 As
–1957 –1958
397 kJ/mol
Monohydride (-AsH)
364 kJ/mol
Dihydride (-AsH2)
(111)B
235 kJ/mol
Trihydride (-AsH3)
–1959 –1960 –1961
0
1
2
3
4
5
6
Distance from surface (Å) Fig. 2.16 Potential energy changes for the desorption processes of AsHn molecules (n D 0, 1, 2, 3) from nonhydride .As/, monohydride .AsH/, dihydride .AsH2 / and trihydride .AsH3 / surfaces of GaAs (111)B
energy of the AsH3 surface as shown in the figure. From these calculations, we can speculate on the As desorption process from the GaAs (111)B surface by considering the stability of AsH3 molecules as follows. First, hydrogen molecules from the carrier gas attach to the surface, and then each of them react with As atoms on the surface to form an AsH2 surface. Next, the AsH2 molecules desorb from the surface with activation energy of 235 kJ/mol, and then the AsH2 molecules react with hydrogen in the vapor phase to form AsH3 . Figure 2.17 shows the arithmetical mean deviation of GaN buffer surface as a function of the activation energy for As desorption process from the surface. The activation energy from the GaAs (001) surface shown in this figure (144 kJ/mol) was obtained by the in situ gravimetric monitoring method [46, 47]. From the figure, the small activation energy on the GaAs (111)A suggests that all As atoms can desorb as AsH3 from the surface thus leaving a completely Ga-terminated surface for the subsequent GaN buffer growth. On the other hand, the large activation energy on GaAs (111)B prevents the complete desorption of all As atoms before the GaN buffer growth. Accordingly, the As atoms remaining on the surface probably prevent N precursors from reacting with the Ga atoms on the surface during GaN buffer growth. The above results can explain the experimental finding that the GaN buffer layers on GaAs (111)B surface had many holes. Therefore, the complete removal of As atoms from the surface is required for the growth of high quality GaN buffer layers.
Hydride Vapor Phase Epitaxy of GaN
Arithmetical mean deviation (nm)
2
49
calculated experimental
As
2.0 As
(111)B
(001)
As 1.0
(111)A
50
100
150
200
250
Activation energy (kJ/mol) Fig. 2.17 Arithmetical mean deviation of the surface as a function of the activation energy for As desorption from GaAs (111)A, (001) and (111)B surfaces
2.6 Thick GaN Growth on (111)A GaAs Substrate [48, 49] As described in the above section, it was found that the uniform GaN layer with a mirror-like surface could be grown at 1;000 ıC on (111)A GaAs surface. Growth of a thick GaN layer on a GaAs substrate is worth investigating because GaAs is easily removed by lapping or etching. In this chapter, the growth temperature dependence of the surface morphology, optical property and crystalline quality of the GaN epitaxial layers are described. In addition, it is shown that the growth of thick GaN layers on (111)A GaAs surface at 1;000 ı C or above is promising for the preparation of a freestanding GaN substrate.
2.6.1 Experimental The MOHVPE system used in this study is the same as that described above. In this system, a vertical quartz reactor is employed and the growth of GaN is carried out under atmospheric pressure using H2 carrier gas. TMG and NH3 were used as Ga and N sources, respectively. The mixing zone of TMG and HCl was kept at 750 ıC throughout the entire growth procedure to form gaseous GaCl by the reaction between TMG and HCl. The substrate used was (111)A GaAs. First, a thin GaN buffer layer of about 50 nm in thickness was grown on the GaAs (111)A surface at 550 ı C. The input partial pressures of TMG, HCl and NH3 were 8 104 atm; 8 104 atm and 1:6 101 atm, respectively. Then the substrate temperature was raised to the growth temperature in an NH3 ambient. Subsequently,
50
A. Koukitu and Y. Kumagai
TMG and HCl were supplied again and the growth of GaN was performed at various temperatures ranging from 920 ıC to 1;000 ıC. The input partial pressures of TMG, HCl and NH3 at this stage were 5:8 103 atm; 5:8 103 atm and 2:9 101 atm, respectively.
2.6.2 Thick GaN Growth on (111)A GaAs Surface First, we investigated the growth rate of GaN on the GaAs (111)A surface in the temperature range from 920 ıC to 1;000 ı C. It was found that the growth rate of GaN was almost constant at around 25 m=h in this temperature range and increased linearly with/while increasing the input partial pressure of TMG and HCl. Therefore, the growth process is limited by mass transportation in the temperature range from 920 ıC to 1;000 ıC. On the other hand, the surface morphology of the grown layer strongly depends on the growth temperature. Figure 2.18 shows surface morphologies of 25 m thick GaN layers grown at various temperatures ranging from 920 ıC to 1;000 ı C. Improvement of the surface morphology with the increasing growth temperature is
Fig. 2.18 Surface morphologies of the 25-m-thick GaN layers grown on the GaAs (111)A surfaces at various temperatures observed by SEM: (a) 920 ı C, (b) 940 ı C, (c) 960 ı C, (d) 980 ı C and (e) 1;000 ı C
2
Hydride Vapor Phase Epitaxy of GaN
51
quite clear. The GaN layer grown at 920 ı C shows a rugged surface, and no-mirrorlike regions can be seen in the surface. However, the area of the mirror-like regions significantly increases with the increase of the growth temperature. As shown in the figure, the GaN layer grown at 1;000 ıC shows a mirror-like surface without hexagonal pits. These results indicate that the growth temperature of 1;000 ı C or above is essential to obtain a GaN layer with a mirror-like surface. The photoluminescence (PL) spectra measured at room temperature using a He Cd laser excitation are shown in Fig. 2.19. One important feature in this figure is the increase of the near band edge emission intensity at 363.1 nm with the increase of growth temperature. Although the band edge emission cannot be seen in the layer grown at 920 ıC, an intense emission with a narrower FWHM of 96 meV can be seen in the layer grown at 1;000 ıC. Another important feature is the decrease of the broad emission peaking at 568.3 nm with the increase of growth temperature. The broad emission, or so-called yellow emission, is related to the deep-level traps due to lattice imperfections. PL analysis also suggests that the growth temperature at 1;000 ıC or above is essential for the growth of high quality GaN layers on GaAs (111)A surface. The crystalline structure and/or quality of the GaN layers grown on GaAs (111)A surfaces were examined by means of XRD and double crystal XRD (DCXRD). First, the 25-m-thick GaN layers grown at various temperatures shown in Fig. 2.18
568.3
363.1
920 °C
×0.5
940 °C
PL intensity (arb. units)
×0.5
×0.5
960 °C
×0.5
980 °C
×0.5
400
500 600 Wavelength (nm)
1000 °C 700
800
Fig. 2.19 Room temperature PL spectra of the GaN layers grown on GaAs (111)A surface at various temperatures shown in Fig. 2.18
52
A. Koukitu and Y. Kumagai 60
FWHM (min)
50
25-μm-thick MOHVPE GaN
40 30 20 10 0 900 920 940 960 980 1000 1020
Temperature(°C) Fig. 2.20 FWHM values of the ¨-mode DCXRD profiles of the (0002) GaN as a function of growth temperature. Note that the thickness of the GaN layer is 25 m at each temperature
were analyzed by XRD in ™–2™ mode. In the obtained XRD profiles, only the hexagonal GaN (0002) and (0004) peaks were seen, irrespective of the growth temperature, which indicates that the hexagonal GaN grows epitaxially on the GaAs (111)A surface to satisfy the relation of GaN (0001) jj GaAs(111) in the temperature range from 920 ı C to 1;000 ıC. Therefore, we examined the FWHM value of the ¨–scan for the (0002) GaN plane by DCXRD. The dependence of the FWHM value on the growth temperature is shown in Fig. 2.20. It is seen in this figure that the FWHM value decreases with increasing growth temperature up to 960 ı C, which is consistent with the results obtained in Figs. 2.18 and 2.19 which show, that both surface morphology and optical qualities are improved with an increase of the growth temperature. On the other hand, at growth temperatures above 960 ı C, the FWHM value increases with increasing growth temperature. This opposite tendency is thought to be due to the bending of the thin grown layer .25 m/ with improvement of crystallinity. Fortunately, however, an answer to the problem was also found. In Fig. 2.21, the FWHM values in the ¨ mode scan for the GaN (0002) plane by DCXRD are shown as a function of grown layer thickness. The growth was performed at 1;000 ı C in this case. While the FWHM value is large when the grown thickness is around 13 m (41.8 min), the FWHM value decreases with the increasing grown layer thickness. By growing a 100 m thick GaN layer with a mirror-like surface, the FWHM value of 4.6 min can be obtained. From these results, we believe that the growth of a thick GaN layer on a (111)A GaAs surface at 1;000 ı C or above is a promising method for the preparation of freestanding GaN substrates. In actual fact, the HVPE growth of a freestanding GaN substrate over 2 in. in size was reported according to the above mentioned findings [48]. In that case, a 0.1m-thick SiO2 layer with 2-m-diameter round openings is first formed directly on a (111)A GaAs substrate, and a 60-nm-thick GaN buffer layer was selectively grown in the openings at 500 ıC. Then, a thick GaN layer was grown at 1;030 ı C. After growing an about 500-m-thick GaN layer, the (111)A GaAs substrate was
2
Hydride Vapor Phase Epitaxy of GaN
53
60 MOHVPE GaN Temp.: 1000°C
FWHM (min)
50 40 30 20 10 0
0
20
40
60
80
100
Thickness (mm) Fig. 2.21 FWHM values of the ¨-mode DCXRD profiles of the GaN (0002) as a function of grown layer thickness. The GaN layer was grown at 1;000 ı C
dissolved by aqua regia. Surface SEM micrographs in the GaN growth process at 1;030 ıC is shown in Fig. 2.22. It is clearly seen that the growth of GaN begins selectively in the openings in the SiO2 mask (Fig. 2.22a), and hexagonal pyramid of GaN N surrounded by f1101g appear (Fig. 2.22b). As the growth continues, overgrowth of GaN on the SiO2 mask occurs, and coalescence of the GaN begins (Fig. 2.22c). After removing the (111)A GaAs substrate, a freestanding GaN substrate of about 500 m thick was prepared by lapping and polishing. The SiO2 layer on backside was also removed during this process. Figure 2.23 shows a photograph of the freestanding GaN substrate obtained. DCXRD profile of the (0002) GaN showed a narrow FWHM of 106 arcsec, and plan-view TEM observation at the top surface revealed the dislocation density of the GaN substrate was as low as 2 105 cm2 . Hall measurement showed that the GaN substrate had n-type conductivity with a carrier concentration and electron mobility of 5 1018 cm3 and 170 cm2 V1 s1 , respectively, indicating that the obtained freestanding GaN substrate is suitable for use in fabrication of violet laser diodes on it.
2.7 Preparation of Fe-Doped Semi-insulating GaN Substrates [50, 51] Recent progress in the preparation of 2-in. -diameter freestanding GaN substrates has enabled mass-production of high-power and long-lifetime violet laser diodes. However, nominally undoped GaN layers grown by any technique, including HVPE, generally show n-type conductivity due to unintentionally incorporated background donor impurities such as Si and O [52]. A next target of production of freestanding GaN substrates is preparation of high-quality semi-insulating (SI) GaN substrates
54
A. Koukitu and Y. Kumagai
Fig. 2.22 SEM micrographs in the process of GaN growth at 1;030 ı C on a (111)A GaAs substrate with a SiO2 mask: (a) 0.5 min, (b) 4 min and (c) 10 min of GaN growth
2
Hydride Vapor Phase Epitaxy of GaN
55
Fig. 2.23 Photograph of a freestanding GaN substrate prepared by HVPE using a (111)A GaAs as an initial substrate
for high-frequency applications. One approach for the growth of SI GaN is to introduce deep acceptor states to compensate for the background donors. It has been reported that the growth of SI GaAs and InP is possible by Fe doping [53, 54]. Fe has been also known to act as a deep acceptor in GaN [55]. Therefore, the method for efficient generation of Fe source and its transportation is essential for growing SI GaN substrate by HVPE. In this chapter, the growth of Fe-doped GaN layers on both (0001) sapphire and (111)A GaAs substrates is shown. Then, preparation of a SI GaN substrate using (111)A GaAs substrate and properties of the SI GaN substrate is described.
2.7.1 Experimental A MOHVPE system with a vertical hot-wall quartz reactor was used. Ga source was GaCl formed by the reaction between TMG and HCl, and Fe source was FeCl2 formed by the reaction between ferrocene ŒCp2 Fe W .C5 H5 /2 Fe and HCl. Both GaCl and FeCl2 were formed independently in the upstream region of the reactor maintained at 750 ıC. The substrates used were (0001) sapphire and (111)A GaAs. First, a 50-nm-thick Fe-doped GaN buffer layer was grown with input partial pressures of TMG, HCl to form GaCl, and NH3 of 8:0 104 atm; 8:0 104 atm and 1:6 101 atm, respectively. The growth was carried out at 500 ı C and 550 ıC for the sapphire and GaAs substrates, respectively. Then, Fe-doped thick GaN layers were grown at 1;000 ıC for 1 h with input partial pressures of TMG, HCl to form GaCl, and NH3 of 3:0 103 atm; 3:0 103 atm and 1:5 101 atm, respectively. During the growth, CP2 Fe and HCl to form FeCl2 were additionally introduced into the reactor. Input partial pressure of the HCl to form FeCl2 was twice as that of Cp2 Fe, and FeCl2 supply was varied by using the input partial pressure ratio of Cp2 Fe to TMG as an index.
56
A. Koukitu and Y. Kumagai
2.7.2 Fe-Doped GaN Layer Grown on Sapphire and GaAs First, GaN layers were grown on sapphire and GaAs substrates with input partial pressure ratios, Cp2 Fe=TMG, of 0 (undoped) and 0.34 (Fe-doped). The growth rate of GaN was 11–12 m=h, regardless of the substrate type and the supply of FeCl2 . After the growth, resistivity was determined by measuring the direct current through the GaN layer at a fixed bias. It was found that the undoped GaN layers grown on sapphire and GaAs substrates show low resistivities about 2:5 103 cm. On the other hand, the resistivities of Fe-doped GaN layers were 2:5 104 cm on sapphire substrate and 3:3 102 cm on GaAs substrate. These results indicate that resistivity of the Fe-doped GaN layer grown on GaAs substrate remained low. To clarify the mechanism of the Fe doping, secondary ion mass spectrometry (SIMS) analysis of GaN layers was performed. It was found that undoped GaN layers grown on sapphire and GaAs substrates contain large quantities .>1015 cm3 / of O (donor), C (acceptor) and Si (donor) impurities, among which O had the highest concentration of about 5 1019 cm3 . This is the reason why undoped GaN layers show n-type conductivity. Figure 2.24 shows the SIMS profiles of GaN layers grown at a CP2 Fe=TMG ratio of 0.34 on sapphire (left) and GaAs (right) substrates. It was found that the Fe-doped GaN layer on a sapphire substrate has impurity concentrations comparable to those of the undoped layer, while a high concentration of Fe in excess of O is incorporated. Thus, it is clear that a high doping level of Fe above 1019 cm3 is possible and that the increase in resistivity by Fe doping on sapphire substrate is attributed to the compensation of O donors by Fe. On the other hand, the Fe-doped GaN layer grown on GaAs substrate also contains O, C and Si impurities with concentrations at roughly the same level as those on (111)A GaAs
on (0001) sapphire GaN
Concentration (cm–3)
1021 1020
GaN
O
Fe
O
1019
C
C 1018
Fe As
1017
Si
Si
1016 0
4
8
Depth (mm)
12
16
0
4
8
12
Depth (mm)
Fig. 2.24 SIMS profiles of Fe-doped GaN layers grown on (0001) sapphire (a) and (111)A GaAs substrate (b). Growth was performed at 1;000 ı C with an input ratio of FeCl2 /TMG of 0.34
2
Hydride Vapor Phase Epitaxy of GaN
57
in Fe-doped GaN layer grown on sapphire substrate. However, it is found that the Fe concentration drops by two orders of magnitude with respect to that on sapphire. In addition, a high concentration of As (neutral) impurity is incorporated on the order of 1017 cm3 . Thus, the low-resistivity of the Fe-doped GaN layer grown on GaAs is due to incomplete compensation of O donors by the decrease of Fe incorporation, even though the same growth conditions were used as on the sapphire substrate. The trade-off between the incorporation of Fe and As is surprising since Fe atoms occupy Ga sites [56] while As atoms is thought to occupy N sites in GaN lattice. As shown later, it was found that the use of a GaAs substrate coated on its back by NiTi layer could increase the concentration of Fe. Thus, it is speculated that the increase in As vapor-species in the growth reactor due to the degradation of the back surface of the GaAs substrate hinders FeCl2 from reacting with NH3 . In order to grow a SI GaN substrate using GaAs substrate, protection of the back surface of the GaAs substrate is essential. Utilizing the above findings, preparation of a SI freestanding GaN substrate was successfully achieved using (111)A GaAs substrate. In this case, GaCl was formed by the reaction between Ga metal and HCl gas for a long period of growth. Before placing a (111)A GaAs substrate in the reactor, the back of the GaAs substrate was covered by a 0:1 m-thick NiTi protective layer. First, a 60-nm-thick GaN buffer layer was grown at 550 ıC. Then, a 400-m-thick Fe-doped GaN layer was grown. Input partial pressures of GaCl, FeCl2 and NH3 at this stage were
Fig. 2.25 Photograph of a 400-m-thick Fe-doped SI GaN substrate after removal of GaAs substrate
58
A. Koukitu and Y. Kumagai
Fe concentration (cm–3)
1021 1020 Fe-doped GaN 1019 1018 1017 undoped GaN 1016 1015 0
2
4
6
8
Depth from surface (μm) Fig. 2.26 SIMS depth profiles of Fe in undoped (dashed line) and Fe-doped (solid line) freestanding GaN substrates
GaN (1010)
GaN (0002) undoped GaN
undoped GaN FWHM = 353 arcsec
Fe-doped GaN
Fe-doped GaN
FWHM = 410 arcsec
FWHM = 360 arcsec
Intensity (normalized)
FWHM = 443 arcsec
–1000
0
ω (arcsec)
1000
–1000
0
1000
φ (arcsec)
Fig. 2.27 XRD rocking curves of (0002) and .101N 0/ GaN measured by ¨- and ¥-mode scan, respectively: undoped (dashed lines) and Fe-doped (solid lines) freestanding GaN substrates
8:0 103 atm; 8:0 104 atm and 2:4 101 atm, respectively, which yielded a GaN growth of about 75 m=h. A photograph of the Fe-doped freestanding GaN substrate taken after removal of the GaAs substrate is shown in Fig. 2.25. The shape of the GaN substrate is the
2
Hydride Vapor Phase Epitaxy of GaN
59
same as that of the GaAs substrate, and no polishing process was employed. It is seen that a transparent GaN substrate with a smooth surface is obtained although the GaN substrate is slight grayish. Figure 2.26 shows SIMS depth profiles of Fe for the undoped and Fe-doped GaN substrates. It is found that the Fe-doped GaN substrate has a Fe concentration of 1:5 1019 cm3 , while the undoped GaN substrate has a Fe concentration almost equal to the detection limit .5 1015 cm3 /. Thus, a high Fe doping level above 1019 cm3 is achieved even on a GaAs substrate. The resistivity of the Fe-doped freestanding GaN substrate measured at room temperature by three-electrode guard method was 8:8 1012 cm. Thus, a SI freestanding GaN substrate can be successfully prepared. Crystalline quality of the Fe-doped SI freestanding GaN substrate was examined by XRD. Figure 2.27 shows the ¨- and ¥-mode XRD rocking curves of the (0002) N of the undoped (dashed lines) and Fe-doped (solid lines) freestanding and .1010/ N of the FeGaN substrates. It is found that the FWHMs of the (0002) and .1010/ doped SI freestanding GaN substrate are fairly small and are 410 and 360 arcsec, respectively. It is also found that the FWHMs do not vary by the Fe doping, indicating that the crystalline quality of the Fe-doped GaN substrate is comparable to that of the undoped GaN substrate.
References 1. H.P. Maruska, J.J. Tietjen, Appl. Phys. Lett. 15, 327 (1969) 2. T. Detchprohm, K. Hiramatsu, H. Amano, I. Akasaki, Appl. Phys. Lett. 61, 2688 (1992) 3. A. Wakahara, T. Yamamoto, K. Ishio, A. Yoshida, Y. Seki, K. Kainosho, O. Oda, Jpn. J. Appl. Phys. 39, 2399 (2000) 4. F. Hasegawa, M. Minami, K. Sunaba, T. Suemasu, Jpn. J. Appl. Phys. 38, L700 (1999) 5. Y. Miura, N. Takahashi, A. Koukitu, H. Seki, Jpn. J. Appl. Phys. 34, L401 (1995) 6. A. Usui, H. Sunakawa, A. Sakai, A.A. Yamaguchi, Jpn. J. Appl. Phys. 36, L899 (1997) 7. Y. Kumagai, A. Koukitu, H. Seki, Jpn. J. Appl. Phys. 39, L149 (2000) 8. H. Seki, A. Koukitu, J. Cryst. Growth 98, 118 (1989) 9. A. Koukitu, N. Takahashi, T. Taki, H. Seki, Jpn. J. Appl. Phys. 35, L673 (1996) 10. A. Koukitu, H. Seki, Jpn. J. Appl. Phys. 36, L750 (1997) 11. A. Koukitu, N. Takahashi, T. Taki, H. Seki, J. Cryst. Growth 170, 306 (1997) 12. A. Koukitu, N. Takahashi, H. Seki, Jpn. J. Appl. Phys. 36, L1136 (1997) 13. A. Koukitu, S. Hama, T. Taki, H. Seki, Jpn. J. Appl. Phys. 37, 762 (1998) 14. A. Koukitu, Y. Kumagai, N. Kubota, H. Seki, Phys. Stat. Sol. B 216, 707 (1999) 15. A. Koukitu, Y. Kumagai, H. Seki, Phys. Stat. Sol. A 180, 115 (2000) 16. A. Koukitu, Y. Kumagai, H. Seki, J. Cryst. Growth 221, 743 (2000) 17. Y. Kumagai, K. Takemoto, T. Hasegawa, A. Koukitu, H. Seki, J. Cryst. Growth 231, 57 (2001) 18. Y. Miura, N. Takahashi, A. Koukitu, H. Seki, Jpn. J. Appl. Phys. 35, 546 (1996) 19. V.S. Ban, J. Electrochem. Soc. 119, 761 (1972) 20. H. Seki, S. Minagawa, Jpn. J. Appl. Phys. 11, 850 (1972) 21. A. Koukitu, H. Seki, Jpn. J. Appl. Phys. 16, 1967 (1977) 22. D.W. Shaw, in Crystal Growth, vol. 1, ed. by C.H.L. Goodman (Plenum, New York, 1978), p.1 23. M. Mizuta, S. Fujieda, Y. Matsumoto, T. Kawamura, Jpn. J. Appl. Phys. 25, L945 (1986) 24. T. Lei, M. Fanciulli, R.J. Molnar, T.D. Moustakas, R.J. Graham, J. Scanlon, Appl. Phys. Lett. 59, 944 (1991)
60
A. Koukitu and Y. Kumagai
25. M.J. Paisley, Z. Sitar, J.B. Posthill, R.F. Davis, J. Vac. Sci. Technol. A 7, 701 (1989) 26. R.C. Powell, G.A. Tomasoch, Y.W. Kim, J.A. Thornton, J.E. Greene, Mater. Res. Soc. Symp. Proc. 162, 525 (1990) 27. H. Okumura, S. Yoshida, S. Misawa, E. Sakuma, J. Cryst. Growth 120, 114 (1992) 28. S. Miyoshi, K. Onabe, N. Ohkouchi, H. Yaguchi, R. Ito, S. Fukatsu, Y. Shiraki, J. Cryst. Growth 124, 439 (1992) 29. H. Tsuchiya, T. Okahisa, F. Hasegawa, H. Okumura, S. Yoshida, Jpn. J. Appl. Phys. 33, 1747 (1994) 30. N. Kuwano, Y. Nagatomo, K. Kobayashi, K. Oki, S. Miyoshi, H. Yaguchi, K. Onabe, Y. Shiraki, Jpn. J. Appl. Phys. 33, 18 (1994) 31. Y. Kumagai, H. Murakami, H. Seki, A. Koukitu, Phys. Stat. Sol. A 188, 549 (2001) 32. S. Porowski, J. M. Baranowski, M. Leszczynski, J. Jun, M. Bockowski, I. Grzegory, S. Krukowski, M. Wroblewski, B. Lucznik, G. Nowak, K. Pakula, A. Wysmolek, K.P. Korona, R. Stepniewski, Proc. Int. Symp. Blue Laser and Light Emitting Diodes, Chiba, 1996 (Ohmsha, Tokyo, 1996) p.38 33. S. Kurai, Y. Naoi, T. Abe. S. Ohmi, S. Sakai, Jpn. J. Appl. Phys. 35, L77 (1996) 34. S.T. Kim, Y.J. Lee, D.C. Moon, C.H. Hong, T.K. Yoo, J. Cryst. Growth 194, 37 (1998) 35. M.K. Kelly, R.P. Vaudo, V.M. Phanse, L. Görgens, O. Ambacher, M. Stutzmann, Jpn. J. Appl. Phys. 38, L217 (1999) 36. Y. Matsuo, Y. Kumagai, T. Irisawa, A. Koukitu, Phys. Stat. Sol. A 188, 553 (2001) 37. P. Hohenberg, W. Kohn, Phys. Rev. B 136, 864 (1964) 38. R.M. Dreuzler, E.K.U. Gross, Density Functional Theory (Springer, Berlin 1990) 39. W. Kohn, A.D. Bedcke, R.G. Parr, J. Phys. 100, 12974 (1996) 40. A.D. Becke, Phys. Rev. A 38, 3098 (1988) 41. J.P. Perdew, Phys. Rev. B 33, 8822 (1986) 42. J.P. Perdew, J.A. Chevary, S.H. Vosko, K.A. Jackson, M.R. Pederson, D.J. Singh, C. Fiolhais, Phys. Rev. B 46, 6671 (1992) 43. D.R. Hamann, Phys. Rev. B 40, 2980 (1982) 44. M. Bockstedte, A. Kley, J. Neugebauer, M. Scheffler, Comput. Phys. Commun. 107, 187 (1997) 45. K. Shiraishi, J. Phys. Soc. Jpn. 59, 3455 (1990) 46. A. Koukitu, N. Takahashi, Y. Miura, H. Seki, J. Cryst. Growth 146, 239 (1995) 47. A. Koukitu, T. Taki, K. Narita, H. Seki, J. Cryst. Growth 198/199, 1111 (1999) 48. Y. Kumagai, H. Murakami, A. Koukitu, K. Takemoto, H. Seki, Jpn. J. Appl. Phys. 39, L703 (2000) 49. K. Motoki, T. Okahisa, N. Matsumoto, M. Matsushima, H. Kimura, H. Kasai, K. Takemoto, K. Uematsu, T. Hirano, M. Nakayama, S. Nakahata, M. Ueno, D. Hara, Y. Kumagai, A. Koukitu, H. Seki, Jpn. J. Appl. Phys. 40, L140 (2001) 50. Y. Kumagai, K. Takemoto, H. Murakami, A. Koukitu, Jpn. J. Appl. Phys. 44, L1072 (2005) 51. Y. Kumagai, F. Satoh, R. Togashi, H. Murakami, K. Takemoto, J. Iihara, K. Yamaguchi, A. Koukitu, J. Cryst. Growth 296, 11 (2006) 52. W.J. Moore, J.A. Freitas, Jr., G.C.B. Braga, R.J. Molnar, S.K. Lee, K.Y. Lee, I.J. Song, Appl. Phys. Lett. 79, 2570 (2001) 53. P.L. Hoyt, R.W. Haisty, J. Electrochem. Soc. 113, 296 (1966) 54. K. Tanaka, K. Nakai, O. Aoki, M. Sugawara, K. Wakao, S. Yamakoshi, J. Appl. Phys. 61, 4698 (1987) 55. B. Monemar, O. Lagerstedt, J. Appl. Phys. 50, 6480 (1979) 56. R. Togashi, F. Satoh, H. Murakami, J. Iihara, K. Yamaguchi, Y. Kumagai, A. Koukitu, Phys. Stat. Sol. B 244, 1862 (2007)
Chapter 3
Growth of Bulk GaN Crystals by HVPE on Single Crystalline GaN Seeds B. Łucznik, B. Pastuszka, G. Kamler, I. Grzegory, and S. Porowski
Abstract In this chapter, the problems related to bulk crystallization of GaN by HVPE are discussed. High quality small single crystalline GaN seeds grown under high pressure and large, flat, free standing GaN substrates grown by HVPE were used in long duration HVPE experiments. The characterization results of the obtained bulk crystals suggest that strong dependence of physical properties (mostly oxygen content) on the orientation of the crystallization front is the main reason for strain and structural defects generation in the new grown material.
3.1 Introduction The hydride vapor phase epitaxy (HVPE) method is, at present, the only one allowing mass production of free standing GaN substrates with quality sufficient for fabrication of laser diodes. However, the technology of making the GaN substrates by HVPE is again very specific (the GaN-based technologies are usually “unusual”) because it is heteroepitaxy (on GaAs [1] or sapphire [i.e., 2]) followed by subsequent substrate lift-off giving just one GaN substrate in one HVPE run (if the single wafer HVPE reactor is used). This is in contrast with the typical approach in semiconductor technology where the substrates (for example of Si or GaAs) are obtained by slicing and polishing of bulk high quality single crystals typically grown from stoichiometric liquids by Czochralski or Bridgman methods. For GaN, it is difficult because of its extreme conditions of melting (TM D 2;220 ıC and pM D 6:0 GPa) [3]. It is possible to grow GaN by HVPE with relatively high rates of the order of 100 m=h in various crystallographic directions, which suggests that bulk crystallization by this method should be attainable in principle. There are a few reports [i.e., 4, 5] on the long duration .>10 h/ growth of GaN by HVPE on a large area, flat sapphire or GaN seeds (substrates) giving 2 in. diameter crystals with thickness approaching 10 mm. However, the progress (in terms of crystal size and quality) in bulk crystallization of GaN by HVPE is not very fast, which is related mostly with technical limitations of the method like parasitic deposition of GaN out of the seed, 61
62
B. Łucznik et al.
deposition of solid reaction products (mostly NH4 Cl) or exhausting of the liquid gallium source. It seems that the sensitivity of physical properties of GaN on the orientation of the growth surface can also cause serious problems in the crystallization of large crystals by HVPE. It is well established [i.e., 6] that for GaN growth with intentional variations of growth direction (i.e., in Epitaxial Lateral Overgrowth – ELOG technique) by MOVPE or HVPE methods, significant differences in optical and electrical properties are observed in different sectors of the crystals. It can be particularly important for growth starting from small seed crystals because the extension of the crystal size proceeds in three dimensions and therefore in different crystallographic directions. In this chapter, the HVPE growth on both small and large (flat) seed crystals is discussed. The discussion is based mostly on the results obtained in the Institute of High Pressure Physics PAS in Warsaw, Poland.
3.2 Experimental 3.2.1 Seed Crystals The crystals used in this study for seeding of bulk crystallization of GaN by HVPE were grown from solutions in liquid gallium at nitrogen gas pressure of 1.0 GPa [7]. This method results in crystals of two dominating morphological forms (Fig. 3.1): thin hexagonal platelets and hexagonal needles elongated into c-direction of wurzite structure. The needles grown usually at high supersaturations are morphologically unstable N faces being partially hollow or skeletal but they often have well developed f1010g which makes these crystals good seeds for growth by HVPE. Crystals of both forms are limited in size to about 1 cm due to still insufficient control of supersaturation in gallium solutions in large volume high pressure reactors. The important advantage of the High Pressure Solution method is that the resulting crystals of GaN are almost free of structural defects. Defect selective etching (DSE) in molten bases [8, 9] of both (0001) polar and nonpolar faces of the crystals
Fig. 3.1 Two morphological forms of GaN crystals grown under high N2 pressure: (a) hexagonal platelets, (b) hexagonal needles. Grid: 1 mm
3
Growth of Bulk GaN Crystals by HVPE on Single Crystalline GaN Seeds
63
Fig. 3.2 Free standing 1–2 mm thick GaN crystals separated spontaneously from GaN/sapphire substrates after 10–12 h HVPE growth: (a) susceptor with large hexagonal crystal just after the HVPE process, (b) the same crystal, (c) and (d) – similar crystals. Grid: 1 mm
of both forms has shown that dislocation densities are negligible and usually do not exceed 100 per cm2 . On the other hand, the crystals contain a high number of point defects due to the high activity of oxygen always present as nonintentional impurity in the high pressure growth system. Due to this, the GaN crystals grown without intentional doping are highly n-type with free electron concentrations exceeding 1019 cm3 . This leads to the slight expansion of the crystal lattice, corresponding to a mismatch to pure GaN of the order of 0.01%. The free electrons can be removed by compensation of the oxygen donor by magnesium added to the growth solution. Then the lattice constants become the same as for undoped material. Both n-type GaN and highly resistive GaN:Mg platelets have been used as seeds for HVPE growth. The (0001) polar surfaces of the seed crystals were prepared for the growth by mechanical and chemical polishing. The n-type GaN needles were used for seeding in their “as grown” state. The large (0001) oriented free standing GaN substrates were grown on GaN– sapphire templates as thick (1–2 mm) GaN crystals which usually separated from sapphire in large pieces if their thickness was sufficiently uniform [10]. Figure 3.2 shows some examples of the thick GaN crystals with characteristic hexagonal shape often observed as a result of spontaneous separation of the nitride from sapphire in the HVPE system used in this study. Such crystals were polished mechanically and etched by reactive ions to be used as substrates for further crystallization. Dislocation density in these substrates was in the range 1–5 106 cm2 depending on the “as grown” crystal thickness.
3.2.2 HVPE Reactor and Growth Conditions The HVPE system used in this work was the horizontal home-built quartz reactor with the rotating quartz susceptor designed for 2 in. diameter substrates. GaCl was supplied vertically, just over the surface of the susceptor with the use of shower head type quartz line. A growth temperature of about 1;050 ı C, temperature of GaCl synthesis of 870 ıC, HCl flow in the range of 5–30 ml=min diluted in 500 ml/min of
64
B. Łucznik et al.
N2 ; NH3 flow of 1,200 ml/min and 3,000 ml/min of N2 or H2 as carrier gases were applied for runs of 4–15 h. The growth rates observed for this geometry and the set of conditions varied, from 20 to 200 m=h and were dependent mainly on the flow of HCl. The main factor limiting the duration of the experiments was deposition of GaN microcrystals at the outlet of GaCl line thus changing the effective size of the openings during growth. All the crystals discussed in this study were grown without an intentional doping therefore their electrical properties were due to either intrinsic point defects or impurities incorporated nonintentionally during growth.
3.2.3 Characterization The bulk GaN crystals grown by HVPE were characterized by various methods allowing determination of structural quality as well as specification and distribution of point defects in the samples. The structural properties were evaluated by X-ray diffraction (XRD) DSE in molten KOH–NaOH eutectics [8, 9] of different surfaces of the crystals and also by transmission electron microscopy (TEM) [11]. For mapping of electron concentration, photo-electro-chemical etching (PEC) in basic solutions under UV illumination was used [12, 13]. The PEC etching rate is dependent on the free carrier concentration in the crystal so the crystal sectors with different electrical properties can be visualized. The PEC method was calibrated with the use of micro-Raman scattering technique [12] allowing mapping of local electron concentration (in certain range) with resolution of about 1 m.
3.3 Experimental Results 3.3.1 Crystals Grown on Small Near Dislocation Free GaN Platelet-Like Seeds The growth of GaN by HVPE on (0001) Ga-polar surface of the pressure grown platelets (HP platelets)is usually stable that is the crystallization front remains flat and the material is continuous. Some typical crystals are shown in Figs. 3.3a–d. The nonpolar cross-sections of the HVPE-grown crystals are shown in Fig. 3.3e, f. It can be seen that the crystal is macroscopically continuous (no inclusions or voids) and that the growth proceeds in both normal and lateral directions. Beside {0001} N and semipolar (mostly polar faces, the low index side walls of both nonpolar {1010} N {1011}) orientations are present. On polar (0001) surfaces of the crystals, macroscopic defects in the form of inverted polyhedral pits are observed. They can be due to defects related to the nonperfect surface preparation or noncoherent inclusions of GaN microcrystals
3
Growth of Bulk GaN Crystals by HVPE on Single Crystalline GaN Seeds
65
Fig. 3.3 Morphology of GaN crystals grown by HVPE on HP platelets: (a–d) General view of the crystals, (e) (101N 0) nonpolar cross-section including HP substrate (dark stripe) and the HVPE GaN, (f) nonpolar cross-sections non including the HP substrate: upper edges correspond to semipolar {101N 1} and nonpolar {101N 0} side walls of the crystal, bottom edges are cut., Grid 1 mm on all images
especially at high supersaturations where nucleation and growth is relatively fast. The latter can be particularly relevant for shower head type geometry of GaCl outlet in the HVPE reactor. N nonpolar surfaces of such crystals revealed DSE of both (0001) polar and (1010) that despite a very high structural quality of the substrates (in terms of dislocation density) a significant number of lattice defects such as dislocations (at a density of 104 –106 cm2 ) and low angle grain boundaries is observed in the HVPE grown thick GaN [14, 15]. Typical DSE result is shown in Fig. 3.4. For growth on the n-type HP substrates, in (0001) direction with a rate of 100 m=h or higher and in N2 atmosphere, the defects are generated in HVPE crystals with thickness exceeding 30–50 m. Such critical thickness can be explained by a small lattice mismatch between strongly n-type substrate and HVPE grown GaN showing typical free electron concentration of less than 1017 cm3 for HVPE growth in Ga-polar (0001) direction. This problem was analyzed by Kry´sko et al [16] who determined curvature of HPGaN/HVPEGaN system in dependence on the HP substrate thickness. The substrates were gradually removed by reactive ion etching (RIE) in order to follow changes of deformation of the crystal. It was found that for “subcritical” HVPE layers, the system behaved in quite a good agreement with Stoney equation modified for the case of thick layers [17]. For layers thicker than 36 m, the crystal shape was constant, which indicated plastic deformation of the system. It seems however that the simple lattice mismatch between the substrate and the new grown crystal is not the only explanation of strain and defect generation in HVPE GaN crystals grown on small seeds because the phenomenon is observed also for HP substrates doped with magnesium for which the free electrons are fully compensated by the acceptor and no lattice mismatch between these crystals and
66
B. Łucznik et al.
dislocations
Low angle grain marker 50 µm
200 µm
Fig. 3.4 GaN crystal grown by HVPE on the n-type HP platelet after DSE: defects visualized as etch pits on both (0001) and (101N 0) surfaces of the crystal
Fig. 3.5 X-ray rocking curves for (0002) reflection for: (a) Mg-doped 110 m thick GaN substrate, (b) the same substrate with 100 m thick HVPE layer, (c) the HVPE layer after removing of the substrate. Both intensity and angle scales are the same for all diagrams
undoped GaN has been found. For the Mg-doped substrates and similar growth conditions (100 m=h; N2 atmosphere) the “critical” thickness was about 100 m and the behavior of the system was quite similar to the one with the n-type substrates. Such a behavior is illustrated in Fig. 3.5 where the X-ray rocking curves (XRC) are shown for the Mg-doped substrate, for the substrate with 100 m HVPE
3
Growth of Bulk GaN Crystals by HVPE on Single Crystalline GaN Seeds
67
layer deposited on (0001) surface and for the HVPE layer after removing of the substrate by mechanical polishing. Deposition of the HVPE layer induced a strong bowing of the system reflected by widening of the rocking curve with corresponding decrease of its intensity. The initial width of the XRC can be partially recovered by the removal of the substrate. Such a behavior shows presence of strain in this virtually lattice and thermally matched system. Difficulties in reproduction of excellent quality of the small pressure grown GaN crystals by bulk crystallization by HVPE can be better understood after analysis of physical properties of HVPE grown GaN in dependence of orientation of the crystallization front. The application of characterization methods sensitive to the concentration of free carriers to the bulk GaN crystals grown by HVPE has shown [12, 15, 18, 19] that physical properties of GaN grown by HVPE are very much dependent on the orientation of the growth surface. This is illustrated in Figs. 3.6a and b, where some typical results of PEC of HVPE grown GaN crystals, are shown. The results are schematically summarized in Fig. 3.6c. For the Ga-polar (0001) growth surface, typical free electron concentration is usually lower than 1017 cm3 N one as well as for semipolar (1011) N surfaces the conwhereas for the opposite (0001) centration is higher by more than two orders of magnitude and exceeds 1019 cm3 . N surface (see next section) the concentration is For growth on the non polar (1010) also quite high and usually achieves 5–8 1018 cm3 . In the considered experimental system, the main source of free electrons in GaN is the oxygen donor. We have shown by secondary ion mass spectroscopy (SIMS) that GaN grown by HVPE contains much more oxygen at places where the crystallization front is rough with predominant semipolar orientation in contrast to the one corresponding to the flat (0001) face of the crystal. Figure 3.7 shows one such result where oxygen concentration has been measured in cross-sectional sample showing two types of growth front shapes: rough with a large fraction of semipolar orientation and a smooth (0001) oriented one. The shape of the crystallization front was recovered again by PEC.
High n – no etching
100 µm
High n – no etching
> 1019
Low n – etching
< 1017
5 × 1018
Low n
> 1019 5 × 1018
> 1019
a
100 µm
b
C
Fig. 3.6 Anisotropy in physical properties of the GaN crystals revealed by photo-electro-chemical etching: (a) (101N 0) cross-section of GaN crystal with flat crystallization front grown by HVPE without an intentional doping, on n-type HP substrate: the substrate and “semipolar” part of the crystal shows no etching features due to very high electron concentration, the (0001) polar part of the crystal is etched, (b) similar crystal with inverted pyramid pits: both substrate and “semipolar” GaN inside the pits are not etched, (c) schematic illustration of electron concentration distribution for main low index faces of GaN during growth by HVPE without an intentional doping
68
B. Łucznik et al.
Fig. 3.7 Oxygen concentration in GaN crystal grown by HVPE in dependence on the shape of the crystallization front: (a) SIMS spectrum showing high oxygen content in part 1 of the crystal shown on the right, (b) cross-section of GaN crystal grown by HVPE after PEC showing changing shape of the crystallization front, (c) SIMS spectrum showing low oxygen content in part 2 (flat growth front) of the crystal
The observed dependences are very important and should be taken into account if bulk crystallization of GaN by HVPE is designed. It is particularly important for fast growth where the local changes of the growth surface orientation can take place. It can happen at the edges of three dimensional islands, macroscopic growth steps or at the walls of the pits. The effect of bowing of the substrate-HVPE layer system as well as structural properties of the final bulk crystals (never as good as the quality of the substrates) observed for both n-type and even the Mg-doped substrates support that growth mechanisms play a crucial role in the generation of strain and defects in GaN crystals grown by HVPE. One can easily imagine that the laterally grown part of the crystal (with high electron concentration) surrounding the internal low concentration sector of the crystal grown on (0001) face of the seed, can lead to deformation (bowing) of the substrate-layer sandwich. An additional (positive) support for this interpretation is the recent observation of a strong influence of the carrier gas on behavior of the investigated system. It seems that it is much easier to grow low defect density crystal on the HP plate-like seeds using hydrogen as a carrier gas in the HVPE experiment. It was shown that near dislocation free, over 1 mm thick crystals can be grown on (0001) surface of the n-type GaN platelet at a rate of about 100 m=h in H2 atmosphere. The only structural defects revealed by DSE of the (0001) surface of the resulting thick crystal were the ones corresponding to defects in the substrates (Fig. 3.8). This effect can be due to relative decrease of the rate of growth in lateral directions (observed also in [20]). The nonpolar GaN quasi wafers were made by slicing the GaN crystals grown by HVPE on the pressure grown GaN substrates Fig. 3.9a. Up to now this approach has been tried for GaN crystals grown on the high pressure platelets with relatively high rates, in N2 atmosphere. Therefore, the nonpolar wafers contained some disN oriented slices locations and grain boundaries (Fig. 3.4). The 300 m thick, (11 20) were prepared for epitaxy by mechanical and chemical polishing. Then, 10-fold
3
Growth of Bulk GaN Crystals by HVPE on Single Crystalline GaN Seeds
b
Defects in the substrate
a Defects in HVPE-GaN
Defects in the substrate
69
c
500 μm
Fig. 3.8 Ga-polar surface of 1 mm thick GaN crystal grown by HVPE on the n-type HP GaN substrate in H2 atmosphere: (a) (middle) – Nomarski contrast (DIC) image focused on (0001) surface of the HVPE grown GaN after DSE: dark etch pits indicate dislocations, large part of the surface is almost free of dislocations, (b,c) DIC images focused on (0001) surface of the substrate showing etch pits corresponding defected area in HVPE grown crystal
Fig. 3.9 GaN/AlGaN MQW deposited by PA MBE on GaN bulk substrates: (a) the substrates, (b) TEM image of the structure grown in nonpolar direction (courtesy of Julita Smalc), (c) excitonic spectra of the nonpolar GaN–AlGaN MQWs with different thicknesses
GaN/AlGaN sequences of 2, 3, and 4 nm thick GaN quantum wells separated by 7 nm Al0:11 Ga0:89 N barriers were deposited by plasma assisted MBE [15, 21] on both the nonpolar and the “usual” polar GaN substrates (for comparison of the optical properties). An advantage of the PA MBE method was that both thicknesses and compositions of the epitaxial layers composing the MQWs were the same for both polar and nonpolar growth directions. The structural quality of the “nonpolar” MQW was checked by XRD measurements and TEM. The TEM image of the nonpolar quantum structure is shown in Fig. 3.9b whereas excitonic photoluminescence (PL) spectra for the nonpolar MQWs of different thicknesses in Fig. 3.9c. The well resolved excitons in the PL spectra are good evidence of an excellent quality of the “nonpolar” quantum walls.
3.3.2 Crystals Grown on Small Near Dislocation Free GaN Needle-Like Seeds In case of crystallization on the needle-like HP seeds, the HVPE growth occurs N mostly in the nonpolar directions, since the seed crystals are elongated in the direction . For conditions corresponding to the fast growth on the
70
B. Łucznik et al. rd
3 HVPE growth
HP seed nd
2 HVPE growth
HVPE GaN HP GaN
st
1 HVPE growth
HP seed
a
b
c
Fig. 3.10 GaN bulk crystals grown by HVPE on the needle-shaped high pressure seeds: (a) schematic illustration of the process (b) example crystals, (c) polar cross-section slice of the crystal grown in threefold 10 h. HVPE processes, interfaces between the HVPE growth runs are visible in optical microscope. Grid: 1 mm
Fig. 3.11 Images of (0001) cross-section of four times HVPE grown GaN crystal on HP GaN seed after DSE: (a) photograph of the cut platelet, HP seed crystal (area of dark contrast) and interfaces between subsequent HVPE growths are indicated. (b) optical microscopy image of the same platelet as in image (a) after etching in molten bases. Feature 1 shows the seed crystal, features 2 and 4 are areas with etch pits spreading from morphological imperfections of the seed, features 3 and 5 show low defect density parts of the HVPE bulk GaN, features 6 and 7 are parasitic, strongly defected grains. (c) scanning electron microscopy image of the boxed area on the image (b)
platelets in N2 atmosphere, the growth on the needles results in prismatic crystals with surprisingly stable morphology. Some examples are shown in Fig. 3.10b. Stability of the growth front can be perturbed by parasitic nucleation due to strain induced by macroscopic imperfections in the seed crystals (see further explanations in this section). The crystals can be reused for further crystallization which allows increasing their size (Figs. 3.10c and 3.11a). The crystals grown by HVPE on the needleshaped seeds were of n-type electrical conductivity with free electron concentration of about 5 1018 cm3 . However, as it was shown by micro-Raman scattering measurements [22], at the interfaces between subsequent HVPE runs the concentration of electrons was more than one order of magnitude higher than in the bulk. Due to this difference in point defect concentration the interfaces between subsequent HVPE runs are visible in optical microscope as shown in Figs. 3.10c and 3.11a. Structural defects in the new grown crystals have been studied by DSE in molten KOH/NaOH eutectics of both polar and nonpolar cross-section samples sliced from
3
Growth of Bulk GaN Crystals by HVPE on Single Crystalline GaN Seeds
71
the bulk material as well as by TEM [11]. Current results of the DSE studies can be summarized as follows (also Fig 3.11): 1. DSE of [0001] surfaces indicate that its significant part (about 50%) is of very low defect density . 0/, and GaN is transported from the high-temperature zone to the low-temperature zone. First, the basic character of solution is nondestructive on autoclave material. Second, one has to take into account that even Ni Cr alloys have limited strength and that certain temperature-pressure conditions cannot be exceeded. Therefore, it is favorable to crystallize GaN in the hot zone and take advantage of maximum temperature in the system, as it leads to better growth quality and lower dislocation density. Third, as was reported in [26, 27], the acidic regime suffers from the appearance of cubic-GaN phase at certain conditions, whereas in ammonobasic regime one deals with the purely hexagonal phase, as shown in Sect. 7.3.1. And last but not least, the retrograde GaN transport direction enables to decrease a contamination level of growing GaN with other undesirable phases (e.g., microprecipitates), as they are typically transported to the cold zone of the system.
7.2.5 Doping In the ammonothermal method the bulk monocrystalline GaN can be easily doped by different donors and acceptors by introducing appropriate precursors of dopants into the autoclave and adding them to the feedstock. In general, different donor dopants, acceptor dopants, and magnetic dopants can be selected according to the desired properties of the substrates. Preferably, they are selected, from the group consisting of Si and O, while Zn and Mg are preferred as acceptor dopants. The concentration of dopants can be controlled in order to alter optical and electrical properties of crystals, ranging from highly conductive to semi-insulating [31] (Sect. 7.3). Typically, the concentrations of these dopants range from 1017 to 1021 cm3 and depend on desired final application of the crystal.
7.2.6 Crystal Machining Figure 7.3a shows one of our first 2-in. seed monocrystals, from which 2-in. substrates can be sliced. In Fig. 7.3b we present our typical 1-in. AMMONO-GaN substrate, which is a c-plane oriented and polished wafer, sliced and round-shaped from a larger crystal. Due to a perfect scalability of the ammonothermal method with autoclave diameter we were able to enlarge the size to 1.5-in. (Fig. 7.3b). Very recently, AMMONO Sp. z.o.o. began to increase the population of crystals (Fig. 7.3a) used for the manufacturing of 2-in. wafers. This may facilitate more effective production of optoelectronic devices grown on truly bulk GaN substrates.
7
Ammonothermal Growth of GaN Under Ammono-Basic Conditions
145
a
b
4
3
2
1
0
1
2
3
4
5
6
7
Fig. 7.3 (a) Photograph of 2-in. seed crystal obtained at AMMONO Company, (b) Photograph of 1-in. and 1.5-in. A-GaN wafer manufactured at AMMONO company
For photonic applications non-polar, m-plane GaN substrates are even more promising and interesting (e.g. green lasers) due to lack of internal electric fields. Since now such substrates have been obtained by hydride vapor phase epitaxy (HVPE) method [32, 33] by slicing the c-direction grown bulk boule along the proper direction, yielding about 1-cm square plates. Similar approach was applied to GaN crystals grown by ammonothermal method. In this case the larger size of semi-polar or non-polar GaN substrates can be obtained from much smaller diameter of boule, when compared to 2-in. HVPE GaN. Figure 7.4a shows the photograph of the thick 1-in. GaN crystals, from which a 10:5 19 mm nonpolar substrate can be made (Fig. 7.4b) [34]. Production of semipolar, for example (201) plane oriented wafers, is also achievable. The wafers must be properly polished by mechanical and chemico-mechanical polishing (CMP) in order to obtain an epi-ready surface. Such prepared A-GaN substrates were applied for homoepitaxy by metal-organic vapor phase epitaxy and epilayers of excellent properties were obtained this way. The results of structural and optical investigations of these epilayers are the subject of Sect. 7.4.
146
R. Doradzi´nski et al.
a
b
Fig. 7.4 (a) Photograph of a thick 1-inch A-GaN crystal, (b) typical m-plane A-GaN substrate fabricated from such a crystal.
7.3 Crystal Characterization 7.3.1 Structural Properties Since the AMMONO-Bulk Method enables growth of high-quality seeds, that can be further multiplied and regrown without a noticeable loss of quality, the AMMONO-GaN crystals are characterized by excellent structural properties. All structural data shown below was measured on as-grown AMMONO-GaN crystals of various shapes and dimensions which did not influence their properties. The pure hexagonal phase of ammonothermally grown GaN was confirmed by powder diffractometry of pulverized single crystal (Fig. 7.5) [21]. In addition, we performed X-ray rocking curve measurements on (002) plane of N-face GaN obtained by ammonothermal method. The measurements were conducted using Panalytical X’Pert Pro MRD high resolution with a Cu K’1 line source, U D 40 kV; I D 20 mA, angle resolution 0:0001ı, slit width 0:1 0:1 mm2 for the incident beam and an open-detector mode for the diffracted beam. Excellent crystallinity was manifested by full width at half maximum (FWHM) value as low as 16 arcsec (Fig. 7.6). We also stress that similar value was obtained on asymmetric (014) and (115) planes. Further narrowing of FWHM to about 10 arcsec was achieved by applying a highly monochromatic and perfectly collimated lowdiameter (below 1 m) X-ray microbeam of synchrotron radiation Spring-8 facility located in Hyogo, Japan [21]. This is still very low FWHM value as compared with that obtained for GaN obtained in ammonoacidic conditions directly on HVPE seeds (about 100 arcsec) [35]. The natural cause of such structural properties is the possibility of using native ammonothermal GaN as seeds for further proliferation in order to create more and more new generations of AMMONO-GaN crystals without any loss of quality. Structural properties of AMMONO-GaN were also illustrated by reciprocalspace maps, as shown in Fig. 7.7. The left image corresponds to .Qx ; Qy / crosssection of reciprocal space map for (002) symmetrical reflex of truly bulk A-GaN crystal and the right one, for highly stressed GaN layer grown on sapphire substrate
7
Ammonothermal Growth of GaN Under Ammono-Basic Conditions
147
25000
20000
15000
10000
5000
0 15
25
35
45 2
55
65
㧛 deg.
Fig. 7.5 X-ray diffraction pattern of pulverized A-GaN crystal 100 μm x 100 μm slit
intensity (a.u.)
150
100
FWHM = 16 arcsec
50
17.34
17.36
17.38
17.40
17.42
omega (degrees)
Fig. 7.6 (0002) X-ray rocking curve of A-GaN crystal
measured for comparison (Fig. 7.7). The general conclusion is that AMMONO-GaN crystals reveal excellent single-crystal structure, without mosaicity and low-angle grain boundaries, as broadening along the Qx direction .FWHM Qx D 0:0008 rlu/ is very small, especially in comparison to broadening of GaN epilayer grown on sapphire. Additionally, one can determine the lattice parameters a, c of AMMONO-GaN monocrystals by solving a set of two equations (7.3)
148
R. Doradzi´nski et al.
Qy·104
Qy·104
AMMONO-GaN
2973
2973
2971
2971
2969
GaN on sapphire
2969 –8.2
Qx·104 8.2
0.0
0.0 Qx·104
–8.2
FWHM Qx = 0.80
8.2
FWHM Qx = 4.67
Fig. 7.7 Reciprocal space map collected for (0002) point collected for A-GaN truly bulk crystal (left panel) and GaN epilayer grown on sapphire (right panel)
1 2 dhkl
D
4 3
h2 C hk C k 2 a2
C
l2 c2
(7.3)
(hkl – Miller indices) related with two interplane spacings d114 and d015 evaluated from reciprocal space maps collected in asymmetric (114) and (015) reflexes. The procedure was repeated for several measurement points of the same crystal, yielding mean lattice parameters a D .3:18919 ˙ 0:00030/ ; c D .5:18529 ˙ 0:00015/ for intentionally undoped n-type crystals. Semi-insulating crystals reveal lattice parameters a D .3:18896 ˙ 0:00023/ , c D .5:18539 ˙ 0:00014/ and p-type crystals give values a D .3:18871 ˙ 0:00028/ ; c D .5:18551 ˙ 0:00020/ . We stress high resolution of the measurement and very low gradients of lattice parameters across the crystal. The obtained values can be used as reference lattice parameters of unstrained GaN. For comparison, the lattice parameters of high nitrogen pressure grown crystals are a D .3:1881 3:1890 ˙0:0003/ ; c D .5:1846 5:1864 ˙0:0002/ [36]. Measurements of the radius of curvature of the AMMONO-GaN lattice were also performed. For the sake of comparative analysis similar measurements were first performed for a standard free-standing HVPE-GaN. The HVPE-GaN measurement consisted in collecting about a dozen X-ray rocking curves from a (002) plane, measured in colinear points spaced by 5 mm along the wafer. A typical result of such a measurement is presented in Fig. 7.8. Because of the lattice bending, the maxima of diffraction peaks shift systematically on the -axis. This effect reflects the fact that the inclination of (0002) plane changes systematically when moving along the measurement line. For each pair of maxima one can calculate the corresponding radius of curvature. As shown in the table on the right side of Fig. 7.8, the values for HVPE-GaN range between 2 and 12 m. In contrast to that, X-ray rocking curves measured at the AMMONO-GaN sample remained essentially in the same position, as illustrated in Fig. 7.9. When
7
Ammonothermal Growth of GaN Under Ammono-Basic Conditions
149
counts/s 1400 1200
HVPE GaN R curv = 2 – 12 m
Position Radius of on sample curvature [m]
1000
–25 : –20 –20 : –15 –15 : –10 –10 : –5 –5 : 0 0:5 5 : 10 10 : 15 15 : 20 20 : 25 25 : 30
800 600 400 200 0 17.0
17.2
17.4
17.6
17.8
18.0
18.2
18.4
12,26 4,79 3,48 2,64 2,27 2,17 3,36 3,38 2,28 2,10 1,88
18.6 Omega (°)
Fig. 7.8 Family of X-ray rocking curves for HVPE GaN and radius of curvature evaluation counts/s 2500
R curv > 1000 m 2000
1500
1000
AMMONO-GaN
500
0 16.80
16.82
16.84
16.86
16.88
16.90
16.92
16.94 Omega (°)
Fig. 7.9 Family of X-ray rocking curves for AMMONO-GaN and radius of curvature evaluation
calculating the radius of curvature, values higher than 1,000 m were obtained. Although such an excellent result cannot be treated as representative for the population of AMMONO-GaN crystals, it was confirmed in many cases, and the averaged value for the whole population is of the order of hundred meters. These data prove
150
R. Doradzi´nski et al. AMMONO-GaN M N
HVPE-GaN N 64 x 64 4 mm
CL (420 nm) ( 2 nm)
CL C (420 0 nm) DSD = 3 x106 /cm2
DSD unmeasurable
Fig. 7.10 Cathodoluminescence result for HVPE-GaN (left panel) and AMMONO-GaN (right panel) 1 x 1 mm m
EPD = 5 x 103 /cm2
Fig. 7.11 Etch pits in AMMONO-GaN
that crystal lattice of AMMONO-GaN crystals is extremely flat, leading to a very low level of the built-in stress. In order to estimate the surface dislocation density, cathodoluminescence (CL) imaging was performed on both free-standing HVPE-GaN and AMMONO-GaN samples. The representative pictures are shown in Fig. 7.10, where the detectedlight wavelength is 420 nm and the picture area is 64 64 m2. The dark spot density (DSD) of the HVPE material (left panel) can be estimated at the level of 3 106 cm2 . On the contrary, in the AMMONO-GaN sample (right panel) one can hardly notice any dark spot and therefore estimation of DSD was not possible. Finally, we performed etching experiments in molten KOH (400 ıC, 5 min) in order to determine the dislocation density in AMMONO-GaN. The results of 1 mm2 of surface area are presented on differential interference contrast (DIC) microscopy photograph in Fig. 7.11. In general, selective etching reveals hexagonal pits resulting from increased etching rate on dislocations. All defects were revealed in 400 ı C,
7
Ammonothermal Growth of GaN Under Ammono-Basic Conditions
151
since no new defects appeared by etching in higher temperature 520 ıC. The value of etch pit density (EPD) is estimated at the level of 5 103 cm2 , which is an outstanding result as compared with typical dislocation density of HVPE material (of the order of 106 cm2 ).
7.3.2 Optical Properties Photoluminescence (PL) is one of the best tools to identify the existence of specific defects and their participation in different radiative recombination processes. In this respect, microphotoluminescence (-PL) experiments were performed at temperature 4.2 K using a 3.813 eV line of He Cd laser as an excitation source. The laser spot was focused on surface using a microscopic reflection with the magnification of 25. Resulting excitation spot size was estimated to be 10 m in diameter. The -PL spectra were collected at an as-grown N-side of AMMONO-GaN crystal at temperature T D 4:2 K in wide spectral range of 2.0–3.5 eV. The results of -PL are shown in Fig. 7.12. The luminescence is dominated by strong band edge emission, consisting mainly of neutral donor bound exciton .D0 X/ recombination (3.4716 eV) [37] together with its single and double
60000
D0 X
T=4.2K
Intensity (Counts)
50000
40000
D0X-LO 30000
20000
D0X-2LO
10000
D-A
0 2.2
2.4
2.6
2.8
3.0
3.2
3.4
Energy (eV) Fig. 7.12 -photoluminsescence spectra for as grown AMMONO-GaN surface measured at T D 4:2 K. The peaks are marked in the following way: D0 X – exciton bound to neutral donor, D0 X LO: optical phonon replica of exciton bound to neutral donor line, D0 X–2LO: two optical phonons replica of exciton bound to neutral donor line, D–A: donor–acceptor transition
152
R. Doradzi´nski et al.
optical phonon replica .D0 X–LO; D0 X–2LO/ and donor-acceptor pair transition (D–A, 3.283 eV). Intense excitonic emission is accompanied by diminishing of typical of GaN, parasitic yellow luminescence .2:2 eV/, attributed mainly with Ga vacancy acceptors [38]. This observation proves high optical quality and unique purity of the studied material, enabling effective excitonic recombination.
7.3.3 Electrical Properties The carrier concentration of the wafers can be controlled by appropriate doping [39], as mentioned in Sect. 7.2.5. Both n-type .n 1018 1019 cm3 , D 103 102 cm/, p-type .p 1018 cm3 , D 101 –102 cm/ and semi-insulating . 106 cm–1012 cm/ substrates can be grown via the ammonothermal method [39], as measured by both Hall effect experiments and contactless methods. Tuning of the electrical properties suggest, that truly bulk AMMONO-GaN crystals may find application in both optoelectronics (highly conductive substrates) and high power electronics (semi-insulating substrates for high electron mobility transistors production).
7.4 Homoepitaxy on Ammonothermal GaN The epilayers were deposited by metalorganic chemical vapor phase deposition (MOCVD) using a RF heated AIXTRON AIX-200 low pressure horizontal reactor on both c-plane (Ga face) and m-plane oriented substrates. Trimethylgallium and ammonia were used as Ga and N precursors, respectively. The growth procedure was started by annealing the substrate for 5 min at high temperature .1; 100 ıC/ in a H2 C NH3 atmosphere at a total reactor pressure of 350 mbar. For all samples grown on A-GaN substrates the ammonia .NH3 / was kept open from the moment the temperature in the reactor reached 500 ıC till the end of the process when the grown sample was cooled down to 500 ı C. This ammonia gas flow was kept constant using 2 standard liters per minute (slm) till the end of the growth process. This step was used to prevent first GaN substrate and then epitaxial layers from decomposition [40]. After the annealing, the growth time was set such as to obtain the 1–2 m thick GaN layer. The growth temperature was 1;170 ıC for the whole process and was carried out at a total reactor pressure of 50 mbar. The epilayers were also characterized by various techniques: X-ray diffraction, defect selective etching (DSE), -PL, PL, and reflectance. In order to check the overall structural quality of GaN epilayers grown on A-GaN polar substrates, the X-ray rocking curves mode measurements were performed. Representative rocking curves measured on (002) plane of epilayer grown on n-type and SI substrate are shown on Fig. 7.13. The resulting FWHM values (22–24 arcsec) are, to our knowledge, outstanding results, that strongly indicate the
7
Ammonothermal Growth of GaN Under Ammono-Basic Conditions
153
(a) 2 μm MOCVD GaN/ n-type A-GaN substrate Intensity (counts)
1000 100 μm x 100 μm slit
FWHM = 22 arcsec 500
0
(b) 2 μm MOCVD GaN/ SI A-GaN substrate Intensity (counts)
1000
100 μm x 100 μm slit
FWHM = 24 arcsec 500
0 –0.04
-0.02
0.00
0.02
0.04
angle (deg) Fig. 7.13 X-ray rocking curves of the GaN epilayers grown on n-type (a) and SI (b) A-GaN substrates for the symmetric (002) reflections
excellent crystalline properties of the studied material. This result is a direct consequence of the high crystalline quality of A-GaN substrates adapted to MOCVD growth, described in Sect. 7.3.1. All optical date are collected in the band-edge emission region. Presented in Fig. 7.14 is the -PL spectra for an epilayer grown on SI polar substrate which is dominated by very intensive, perfectly-resolved excitonic emission. This structure consists of both lines related to emission of free excitons .FXA ; FXB /, and very sharp and narrow lines (of the FWHM value of the order of 0.3 meV) related to excitons bound to neutral acceptor .A0 X/ and both excitons A and B bound to different neutral donors D1 0 XA ; D2 0 XA ; D3 0 XA ; D0 XB . Exact identification of individual donors involved with the latter four transitions requires further research (for example magnetooptic measurements in high magnetic fields). Two kinds of free excitons .FXA ; FXB / are formed with participation of holes originating from crystal field and spin-orbit split valence band. The energies of all the observed transitions are the following: 3.4779 eV, 3.4831 eV, 3.4717 eV, 3.4730 eV,
154
R. Doradzi´nski et al. 100000
0
AX
T=4.2 K
0
FWHM=0.3 meV
D1XA 0
Intensity (Counts)
10000
D3XA
SI substrate
1000
0
D2XA FXA 0 D XB
TES
FXB 100
10
1 3.44
3.46 Energy (eV)
3.48
Fig. 7.14 -PL spectra of GaN epilayer grown on SI A-GaN substrate measured at T D 4:2 K. The peaks are marked in the following way: FXA – free exciton A, FXB – free exciton B, D1 0 XA ; D2 0 XA ; D3 0 XA – exciton A bound to different neutral donors, D0 XB – exciton B bound to neutral donor, A0 X – exciton bound to neutral acceptor, TES – two electron satellite
3.4707 eV, 3.4754 eV, 3.4663 eV for FXA ; FXB ; D1 0 XA ; D2 0 XA ; D3 0 XA ; D0 XB ; A0 X peaks, respectively. They are very close to those reported for unstrained homoepitaxial GaN (3.4776 eV, 3.4827 eV for free excitons A and B recombination [41] and 3.4719 eV and 3.4668 eV for exciton bound to donor and acceptor emission [37]. In addition to excitonic structure we observed two electron transition (TES) – a recombination of exciton bound to donor that leaves the donor in the excited state [42]. We also stress that the optical spectra is reproducible in the entire sample surface range. In other words, the optical properties are highly homogeneous independently on excitation place (Fig. 7.14). High PL homogeneity corresponds well with structural and microscopic measurements performed on these layers. Moreover, the width of bound excitons peaks .FWHM 0:3 meV/ is one of the smallest ever reported for homoepitaxial GaN. It proves that ammonothermal GaN substrates with perfect crystalline properties enable the growth of excellent quality, strain-free homoepitaxial layers. Additional PL measurements were also performed at various impulse excitation conditions in a standard configuration using a Si multichannel detector, 0.55 m length monochromator and YAG laser emitting at 266 nm with an average output power of 2 mW (0:1 J in one 10 ns pulse). The excitation beam was defocused to a diameter of 2 mm and its intensity was changing from 0.01 to 2 mW. The same experimental set-up was used to measure reflectance spectra. The sample was illuminated by white light from a halogen lamp (100 W) at near normal incidence.
7
Ammonothermal Growth of GaN Under Ammono-Basic Conditions
155
For low temperature measurements the samples were mounted on a cold finger in a continues-flow liquid helium cryostat. Those additional PL measurements performed under pulsed laser excitation (Fig. 7.15) seem to corroborate the above results. In this case the bound exciton recombination is saturated and suppressed in the scale of the spectra due to the limited number of donors and acceptors. Instead, the biexciton emissions (marked by XX1 ; XX2 ) appear and dominate the PL spectra together with free exciton .FXA ; FXB / recombination under high excitation conditions (Fig. 7.15). Exact study of biexciton emission is beyond the scope of this chapter and will be published elsewhere. In the reflectance spectra (Fig. 7.16a) two dispersion lines .FXA ; FXB /, related to free excitons A and B described earlier, are clearly visible. In addition, there is a third one, a bit weaker, assigned as FXC , that consists of the electron and hole originating from the third remaining valence subband resulting from crystal field and spin-orbit splitting. The energy position of the dispersion lines FXA and FXB are in reasonable agreement with PL and -PL results and literature data [41]. More thorough analysis of reflectance and PL spectra will be a subject of an upcoming paper. In case of GaN epilayers deposited on nonpolar m-plane A-GaN substrates it should be expected that the FXA exciton will be completely polarized in the epilayer plane [34]. Indeed, the studied films exhibit intrinsic narrow exciton lines, which are very sensitive to the optical selection rules typical for hexagonal
GaN epilayer on n-type GaN substrate 1000
XX1 P max
PL intensity (arb. u.)
10
FXA FX B
XX2
100
T=10K
P min
1
0.1
0.01
1E-3 3.45
3.46
3.47
3.48
3.49
Energy (eV) Fig. 7.15 PL spectra measured for GaN epilayers grown on n-type A-GaN substrate collected at T D 10 K at various excitation conditions. The peaks are marked in the following way: FXA – free exciton A, FXB – free exciton B, XX1 ; XX2 – biexciton lines
T=10K
B
FXC
XX
c-plane
(b)unpolarized
FXA
FXB
FXC
E^ c E II c
I m-plane
3.42
3.44
3.46
3.48
Reflectance (arb. u.)
PL intensity (arb. u.)
FXA FX
(a)
Reflectance (arb. u.)
R. Doradzi´nski et al.
PL intensity (arb. u.)
156
3.50
Energy (eV) Fig. 7.16 (a) Reflectance and PL spectra of GaN epilayers grown on n-type polar A-GaN substrate collected at T D 10 K, (b) Polarized reflectance spectra measured at 10 K for GaN homoepitaxial layer grown on m-plane A-GaN substrate. Polarization configurations are ¢ .k?c; Ejjc/ and .k?c; E?c/. The lines are marked in the following way: FXA – free exciton A, FXB – free exciton B, FXC – free exciton C, XX – biexciton, I – bound exciton line
symmetry – the free exciton A line completely disappears in Ejjc configuration, proving truly nonpolar character of m-plane oriented A-GaN crystals (Fig. 7.16b). The crystals have an ideal hexagonal symmetry without any local structural imperfections which could destroy this symmetry and change the optical selection rules. Again, observation of both free excitons and biexciton transitions is a fingerprint of high quality studied material. High quality of homoepitaxial layers deposited on polar c-plane A-GaN substrates was also confirmed by modulation spectroscopy: photoreflectance (PR) and contactless electroreflectance (CER) measurements. We note that the PR technique has never been applied to study homoepitaxial films because of the limiting access to bulk GaN substrates. The PR spectra (Fig. 7.17) exhibits the following
7
Ammonothermal Growth of GaN Under Ammono-Basic Conditions
0.03
FXA
0.02
FXB
T = 10 K FXC
0.01
DR/R
157
0.00 –0.01 –0.02
GaN(2μm)/A-GaN
–0.03
Exp. Fit
–0.04 3.47
3.48
3.49
3.50
Energy (eV) Fig. 7.17 Photoreflectance spectra from a GaN epilayer grown on polar c-plane A-GaN substrate. Solid line represents Lorentzian line fit. The lines are marked in the following way: FXA – free exciton A, FXB – free exciton B, FXC – free exciton C
features: (1) Typical three resonances corresponding to intrinsic free exciton transitions (FXA, FXB and FXC) of the energies evidently lower (3.4760 eV, 3.4817 eV, 3.4991 eV) than those typical for GaN layers grown on sapphire substrates (3.483 eV, 3.491 eV, 3.512 eV, respectively) [43] were observed, (2) Exciton linewidth for the homoepitaxial layer are significantly smaller than those for the heteroepitaxial layer, the observed transitions are also about 20–50% narrower than the exciton transitions reported in [41] measured by reflectivity for homoepitaxial layers grown on HNP crystals. Such narrow resonances and homogeneous broadening of excitonic transitions mean that no residual strain exists in GaN epilayer. Very narrow CER resonances .15 meV/, typical of high quality material, have been observed for considered epilayers, while broad CER resonances typical of band-to-band absorption in bulk material were recorded for A-GaN substrates [44]. It confirms the excellent usefulness of ammonothermal GaN substrates for homoepitaxy.
7.5 Conclusions The results of measurements performed for bulk GaN single crystals obtained by the AMMONO-Bulk Method, such as: (a) FWHM value of (0002) X-ray rocking curve of about 16 arcsec, (b) the radius of curvature of the crystal lattice of the order of 102 –103 m, and (c) EPD of the order of 5 103 cm2 (d) wide spectrum of electrical properties, all unambiguously prove excellent crystalline quality of the grown material. The excellent structural properties of crystals, combined with advantages of the method itself (seed-multiplication, repeatability, scalability), allow us to claim that AMMONO-Bulk Method is an extremely promising technology for fabricating
158
R. Doradzi´nski et al.
high-quality, large diameter GaN substrates. Confirmed high scalability of the method will play a crucial role in further improving its cost-effectiveness when implemented on an industrial scale. Moreover, as no deterioration of the quality within crystal thickness was observed, the growth of sufficient sized non-polar, mplane substrates was succeeded. Such substrates eliminate the problem of internal piezoelectric field, which reduces efficiency of devices and are for example essential for achieving green lasers. Structural and optical properties of GaN MOCVD layers grown on ammonothermal GaN substrates have been studied by X-ray diffraction, defect selective etching, PL, -PL microphotoluminescence, reflectance, and modulation spectroscopy. We have succeeded in performing homoepitaxy by MOVPE resulting in high quality, strain free GaN epilayer, which was documented by the small width of bound exciton peaks, observation of free excitons and biexciton emission (and absorption), highly homogeneous luminescence properties, and small FWHM of X-ray rocking curve. Presented results indicate that both polar and nonpolar (m-plane) A-GaN substrates may be excellent for homoepitaxy. Moreover, controllable conductivity of A-GaN substrates (from n-type to SI) can be very useful in fabrication of various devices, like LD operating in blue light range, UV detectors, or high electron mobility transistors (HEMTs). Acknowledgements This project was partially supported by the European Union fund “Sektorowy Program Operacyjny Wzrost Konkurencyjno´sci Przedsie¸biorstw – Program 1.4.1” (Europejski Fundusz Rozwoju Regionalnego). The authors acknowledge the following colleagues for their contribution to this work: Dr. Robert Kucharski and his team for machining A-GaN crystals in order to prepare samples for epitaxy and measurements, Tomasz Tracz and his team for growing bulk A-GaN crystals, dr Mariusz Rudzi´nski for performing homoepitaxy, Arkadiusz Puchalski and Dr. Jarosław Serafi´nczuk for X-ray measurements, Dr. Andrzej Wysmołek for -PL measurements, and Dr. Robert Kudrawiec for PL, reflectivity, PR measurements.
References 1. See, for example, S. Nakamura, G. Fosol, The Blue Laser Diode (Springer, Berlin, 1998) 2. See, for example, H. Morkoc, Nitride Semiconductors and Devices (Springer, Heidelberg, 1999) 3. S. Porowski, Mater. Sci. Eng. B 44, 407 (1997) 4. I. Grzegory, M. Bo´ckowski, B. Łucznik, S. Krukowski, Z. Romanowski, M. Wróblewski, S. Porowski, J. Cryst. Growth 246, 177 (2002) 5. B. Łucznik, B. Pastuszka, I. Grzegory, M. Bo´ckowski, G. Kamler, E. Litwin-Staszewska, S. Porowski, J. Cryst. Growth 281, 38 (2005) 6. H. Yamane, M. Shimada, S.J. Clarke, F.J. DiSalvo, Chem. Mater. 9, 413 (1997) 7. A. Denis, G. Goglio, G. Demazeau, Mater. Sci. Eng. R 50, 167 (2006) 8. S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa, T. Yamada, T. Matsushita, H. Kiyoku, Y. Sugimoto, T. Kozaki, H. Umemoto, M. Sano, K. Chocho, Appl. Phys. Lett., 72, 211(1998) 9. K. Byrappa, M. Yoshimura, Handbook of Hydrothermal Technology (Noyes, NJ, USA, 2001) 10. R. Dwilinski, A. Wysmolek, J. Baranowski, M. Kaminska, R. Doradzinski, J. Garczynski, L. Sierzputowski, Acta Physica Polonica 88(5), 833 (1995)
7
Ammonothermal Growth of GaN Under Ammono-Basic Conditions
159
11. R. Dwilinski, J. Baranowski, M. Kaminska, R. Doradzinski, J. Garczynski, L. Sierzputowski, Acta Physica Polonica 90(4), 763 (1996) 12. R. Dwilinski, R. Doradzinski, J. Garczynski, L. Sierzputowski, J. Baranowski, M. Kaminska, Mater. Sci. Eng. B 50, 46 (1997) 13. M. Palczewska, B. Suchanek, R. Dwilinski, K. Pakula, A. Wagner, M. Kaminska, MRS Internet J. Nitride Semicond. Res. 3, Article 45 (1998) 14. J.W. Kolis, S. Wilcenski, R.A. Laudise, Mater. Res. Soc. Symp. Proc. 495, 367 (1998) 15. D.R. Ketchum, J.W. Kolis, J. Cryst. Growth, 222, 431 (2001) 16. A.P. Purdy, Chem. Mater. 11, 1648 (1999) 17. A. Yoshikawa, E. Ohshima, T. Fukuda, H. Tsuji, K. Oshima, J. Cryst. Growth 260, 67 (2004) 18. R.T. Dwilinski, R.M. Doradzinski, J.S. Garczynski, L.P. Sierzputowski, Y. Kanbara, Polish Patent Application no. P-347918 (06.06.2001) 19. R.T. Dwilinski, R.M. Doradzinski, J.S. Garczynski, L.P. Sierzputowski, Y. Kanbara, International Patent Application no. PCTnIB02n04185 (17.05.2002) 20. R.T. Dwilinski, R.M. Doradzinski, J.S. Garczynski, L.P. Sierzputowski, Y. Kanbara, United States Patent no. 6,656,615 B2 (02.12.2003) 21. R. Dwilinski, R. Doradzinski, J. Garczynski, L.P. Sierzputowski, A. Puchalski, Y. Kanbara, K. Yagi, H. Minakuchi, H. Hayashi, J. Cryst. Growth 310, 3911 (2008) 22. B. Wang, M.J. Callahan, J. Cryst. Growth 291, 455 (2006) 23. T. Hashimoto, K. Fujito, B.A. Haskell, P.T. Fini, J.S. Speck, S. Nakamura, J. Cryst. Growth 275, e525 (2005) 24. T. Hashimoto, K. Fujito, R. Sharma, E.R. Letts, P.T. Fini, J.S. Speck, S. Nakamura, J. Cryst. Growth 291, 100 (2006) 25. M.P. D’Evelyn, H.C. Hong, D.-S. Park, H. Lu, E. Kaminsky, R.R. Melkote, P. Perlin, M. Leszczynski, S. Porowski, R.J. Molnar, J. Cryst. Growth 300, 11 (2007) 26. A.P. Purdy, Cryst. Growth Des. 2(2), 141 (2002) 27. A.P. Purdy, A.C. Cas, N. Muratore, J. Cryst. Growth 252, 136 (2003) 28. Y. Kagamitani, D. Ehrentraut, A. Yoshikawa, N. Hoshino, T. Fukuda, H. Tsuji, K. Oshima, Jpn. J. Appl. Phys. 45, 4018 (2006) 29. T. Hashimoto, K. Fujito, M. Saito, J.S. Speck, S. Nakamura, Jpn. J. Appl. Phys. 44, L1570 (2005) 30. T. Fukuda, D. Ehrentraut, J. Cryst. Growth 305, 304 (2007) 31. R.T. Dwilinski, R.M. Doradzinski, J. Garczynski, L.P. Sierzputowski, Y. Kanbara, International Patent Application no. PCTnPL2005n000036 (10.06.2005) 32. T. Paskova, R. Kroeger, D. Hommel, P. P. Paskov, B. Monemar, E. Preble, A. Hanser, N.M. Wiliams, M. Tutor, Phys. Stat. Sol. (C) 4, 2536 (2007) 33. K. Fujito, K. Kiyomi, T. Mochizuki, H. Oota, H. Namita, S. Nagao, I. Fujimura, Phys. Stat. Sol. (A) 205, 1056 (2008) 34. R. Kucharski, M. Rudzi´nski, M. Zaja¸c, R. Doradzi´nski, J. Garczy´nski, L. Sierzputowski, R. Kudrawiec, J. Serafi´nczuk, W. Strupi´nski, R. Dwili´nski, Appl. Phys. Lett. 95, 131119 (2009) 35. D. Ehrentraut, Y. Kagamitani, C. Yokoyama, T. Fukuda, J. Cryst. Growth 310, 891 (2008) 36. M. Leszczy´nski, H. Teisseyre, T. Suski, I. Grzegory, M. Bo´ckowski, J. Jun, S. Porowski, K. Pakuła, J.M. Baranowski, C.T. Foxon, T.S. Cheng, Appl. Phys, Lett. 69, 73 (1996) 37. K. Pakuła, A. Wysmołek, K.P. Korona, J.M. Baranowski, R. Ste¸pniewski, I. Grzegory, M. Bo´ckowski, J. Jun, S. Krukowski, M. Wróblewski, S. Porowski, Sol. State Commun. 97, 919 (1996) 38. J. Neugebauer, C. van de Walle, Appl. Phys. Lett. 69, 503 (1996) 39. R.T. Dwilinski, R.M. Doradzinski, J. Garczynski, L.P. Sierzputowski, Y. Kanbara, International Patent Application no. PCTnPL2005n000036 (10.06.2005) 40. D.D. Koleske, A.E. Wickenden, R.L. Henry, M.E. Twigg, J.C. Culbertson, R.J. Gorman, Appl. Phys. Lett. 73, 2018 (1998); Erratum Appl. Phys. Lett. 75, 1646 (1999).
160
R. Doradzi´nski et al.
41. K.P. Korona, A. Wysmołek, K. Pakuła, R. Ste¸pniewski, J.M. Baranowski, I. Grzegory, B. Łucznik, M. Wróblewski, S. Porowski, Appl. Phys. Lett. 69, 788 (1996) 42. A. Fiorek, J.M. Baranowski, A. Wysmołek, K. Pakula, M. Wojdak, I. Grzegory, S. Porowski, Acta Phys. Pol. A 92, 742 (1997) 43. A.K. Viswanath, J.I. Lee, S. Yu, D. Kim, Y. Choi, C.H. Hong, J. Appl. Phys. 84, 3848 (1998) 44. R. Kudrawiec, J. Misiewicz, M. Rudzi´nski, M. Zaja¸c, Appl. Phys. Lett. 93, 061910 (2008)
Chapter 8
A Pathway Toward Bulk Growth of GaN by the Ammonothermal Method Tadao Hashimoto and Shuji Nakamura
Abstract In this chapter, research progress of bulk GaN growth by the ammonothermal method at the University of California, Santa Barbara is reviewed. Starting from a fundamental study on chemistry by powder synthesis, ammonothermal GaN growth process has been developed through solubility measurement, growth from metallic Ga nutrient and growth from polycrystalline GaN nutrient. Based on accumulated knowledge on the growth process, polyhedron-shaped GaN bulk crystals were achieved. Future challenges are presented through characterization of GaN wafers sliced from the bulk crystals.
8.1 Introduction Until the early 1990s, many researchers in the field of compound semiconductors aspired to become the “first” to realize blue light emitting diodes (LEDs) and blue laser diodes (LDs). Despite the earlier achievements of blue LEDs by silicon carbide [1] and blue LDs by II–VI materials [2], commercially reliable blue/green/white LEDs and blue LDs have been achieved with GaN-based solid solutions. After the successful demonstration of InGaN blue LEDs [3] and blue LDs [4], many crystal growers’ interest has shifted toward growth of bulk GaN crystals. Although not many material scientists are privileged to pursue this “second dream” due to limited access to special growth apparatus, the entire body of nitride research has longed for realization of GaN wafers and bulk GaN crystals. This is because GaN or AlN substrates provide a direct and ultimate solution for fundamental growth problems arising from heteroepitaxy (e.g., on sapphire). Growth of bulk GaN, however, is extremely challenging due to the high melting temperature and high equilibrium N2 pressure at high temperature. Several growth techniques based on a Ga melt have been investigated to grow bulk GaN crystals [5–8]. Nevertheless, these methods encountered difficulties to produce large enough bulk GaN crystals which can be sliced into 2 in. wafers. In the meanwhile, hydride vapor phase epitaxy (HVPE) was developed to take a quasi-bulk approach to fabricating GaN wafers. GaN wafers by HVPE made a significant contribution to mass 161
162
T. Hashimoto and S. Nakamura
production of blue LDs for blu-ray DVDs; however, cost reduction is one of the most challenging issues in the quasi-bulk approach. Considering a rapid decrease in the market prices of blue and white LEDs grown on sapphire or SiC, a bulk growth method must be expandable to a volume production. Ammonothermal growth, a solvothermal method using NH3 as a fluid, is a promising candidate owing to its potential scalability. Ammonothermal growth is regarded as a highly expandable method since it is an analogue of hydrothermal growth widely used for the quartz mass production. The recent demonstration of bulk GaN grown by the ammonothermal method [9] opened up a new substrate choice for nitride-based LEDs. For instance, LEDs fabricated on nonpolar or semipolar bulk GaN substrates have already realized performance comparable to state-of-the-art c plane LEDs [10, 11]. Nitride-based solid-state lighting which requires even brighter LEDs is expected to be realized with low-cost and high-quality nonpolar/semipolar GaN substrates sliced from ammonothermally grown bulk GaN. Since Peters reported the AlN synthesis in supercritical NH3 in 1990 [12] several groups have reported key achievements in the ammonothermal growth of GaN [9, 13–27]. Ammonothermal activity at the University of California, Santa Barbara (UCSB) started in July 2000. At that time, there were only a few publications on the ammonothermal growth of nitride: AlN synthesis by Peters [12], microcrystalline GaN growth in ammonobasic conditions [13–15], and microcrystalline GaN growth in ammonoacidic conditions [16]. As our group had no background in highpressure technology, we explored our own pathway toward bulk GaN growth with limited landmarks. In this chapter, we will trace our research findings until bulk GaN crystals were obtained by the ammonothermal method.
8.2 Impact of Mineralizer on Ammonothermal Synthesis of GaN A mineralizer is an ionic additive to increase solubility of solute in solvothermal growth. The choice of the mineralizer is very important since it determines the chemistry of crystal growth. The type of mineralizers can be categorized into three groups in terms of the resulting acidity of the fluid. Since NH3 dissociates into NHC 4 C and NH 2 , mineralizers that generate NH4 are called acidic mineralizers whereas those which generate NH 2 are called basic mineralizers. Mineralizers which do not dominantly generate either NHC 4 or NH2 are called neutral mineralizers. One of the major impacts of the mineralizer in the ammonothermal synthesis of GaN is its influence on phase selection. When a basic mineralizer is used, wurtzite GaN is preferentially formed at temperature as low as 600 ıC, whereas mixed phase (i.e., wurtzite and zinc-blende) GaN tends to be synthesized with acidic and neutral mineralizers. Another important factor is its chemical compatibility with autoclave materials. Acidic NH3 severely corrodes steel, stainless steel, and Ni Cr based superalloys whereas basic and neutral NH3 does not attack most of Ni Cr based superalloys. To avoid corrosion of the autoclave and related components (e.g., valves
8
A Pathway Toward Bulk Growth of GaN by the Ammonothermal Method
163
and pressure transducers) in acidic ammonothermal growth, all wet surface must be protected with Pt-based liner. Synthesis of microcrystalline GaN in autoclaves made of NiCr based alloy with various mineralizers is summarized in Table 8.1. Detailed experimental procedure and method for calculating the ratio of phases are explained elsewhere [22]. When a mineralizer was not added, Ga metal did not react with NH3 . The synthesized powder did not show any definite crystal facets under scanning electron microscope (SEM) observation. Among neutral mineralizers, only KI, NaI, and LiCl yielded an appreciable amount of GaN powder. The powder yield appeared to decrease at higher concentration of KI, although the mechanism of this is unclear. The powder yield with KI was about 1–2 wt% and the powder yield with NaI was slightly higher (3.3 wt%). Among neutral mineralizers in this study, LiCl resulted in the highest yield (16.7 wt%). Considering that KCl and NaCl did not yield GaN, Li seems to enhance GaN formation compared to K and Na. In acidic conditions, powder yield for NH4 Cl; NH4 I, and GaI3 was 6.3 wt%, 3.3 wt%, and 22 wt%, respectively. GaI3 mineralizer resulted in very high powder yield, because GaN was formed not only from the Ga nutrient but also from GaI3 . In the case of NaNH2 (a basic mineralizer), the powder yield was even higher .>20 wt%/. However, in some occasions GaN was not synthesized (Table 8.1). When GaN powder was not formed in basic conditions, the Ga charge became a yellowish powder which dissolved in water with an accompanying emission of NH3 . Judging from this emission, the yellowish powder is thought to be some form of amide, possibly NaGa.NH2 /4 [28, 29]. Therefore, under basic conditions, it is surmised that GaN was synthesized from further ammonolysis of gallium-containing amide. This implies that the reaction path in basic conditions is different from that in neutral and acidic conditions. Figure 8.1 shows X-ray 2 ! scans of the synthesized GaN powders. GaN powders synthesized with NH4 Cl (acidic), NaI (neutral), and KI (neutral) showed diffraction from zinc-blende GaN (zb-GaN) as well as wurtzite GaN (wz-GaN). In contrast, GaN powder synthesized with NaNH2 CNaI (basic) at 525 ı C did not show a clear peak from zb-002 diffraction. For GaN powder synthesized at higher temperature and pressure .600 ıC; 68:3 MPa/ with NaNH2 C NaI, the diffraction pattern showed phase-pure wz-GaN. The quantified zb/wz ratios are shown in Table 8.1. At 525 ı C the zb/wz ratios are about one order of magnitude lower for the basic condition than acidic and neutral conditions. As calculated in [22], the Gibbs free energy of formation for wz-GaN is smaller than that of zb-GaN by small energy difference: only 720 J/mol (8 meV) at 600 K and 1,310 J/mol (14 meV) at 800 K. In acidic and neutral conditions, higher zb/wz ratio at higher formation rate (Table 8.1) indicates that there is a kinetic mechanism which enhances zb-GaN formation at faster reaction. In the case of basic conditions, the ratio of zb-GaN to wz-GaN was about one order of magnitude lower at 525 ıC, and phase-pure wz-GaN was obtained at 600 ıC, at which temperature even vapor-phase epitaxy on c-plane sapphire experience inclusion of zinc-blende phase [30, 31]. These results indicate that the phase selection in the ammonothermal synthesis of GaN is primarily influenced by kinetics related to mineralizer type.
0:5
0:5
0:5
0:5
2:96
3:29
1:1
2:96
NaNH2
NaNH2
NaNH2
NaNH2
600
600
525
525
525
0:5
Basic mineralizers 1:77 NaNH2
525 525 525 525 525 525 525 525 525
525 525 525
3 3 3 3 3 3 3 3 3
525
Temp. (C)
Acidic mineralizers 2:83 NH4 C1 NH4 I 2:83 GaI3 2:83
3
Ga (g)
Neutral mineralizers KCl 2:87 KBr 2:87 KI 2:87 KI 2:87 KI 2:87 NaCl 2:87 NaBr 2:87 NaT 2:87 LiCl 2:87
NH3 (mol)
Without mineralizer 2:87
Mineralizer
68:3
29:3
56:9
53:1
40
34:5 32:8 32:4
32:8 31:7 32:4 32:4 32:8 32:4 32:6 32:4 33:1
33:8
Pressure (MPa)
18
48
24
24
37
7 7 7
7 7 7 7 7 7 7 7 7
19
Time (h)
Process Conditions
1
1
1
1
1
0:7 0:7 0:7
0:7 0:7 0:14 0:7 1:4 0:7 0:7 0:7 0:7
0
%mol of Mineralizer to NH3
Table 8.1 Summary of GaN synthesis with various mineralizers
140
no yield
no yield
no yield
100
190 100 660
No yield No yield 70 50 20 No yield No yield 110 480
No yield
GaN Powder Weight (mg)
28.00%
no yield
no yield
no yield
20.00%
6.33% 3.33% 22.00%
No yield No yield 2.33% 1.67% 0.67% No yield No yield 3.67% 16.00%
No yield
GaN Powder Yield (wt%)
Water soluble yellowish powder Water soluble yellowish powder Water soluble yellowish powder Pure wz-GaN
Mostly wz-GaN
wz, zb-GaN wz, zb-GaN wz. zb-GaN
N/A N/A wz. zb-GaN wz. zb-GaN wz, zb-GaN N/A N/A wz. zb-GaN wz, zb-GaN
N/A
Product
Result
0
0
N/A
N/A
0.66
0.41 0.21 1.14
N/A N/A 0.26 0.21 0.22 N/A N/A 0.33 0.76
N/A
zb/wz Ratio
With comineralizer of 0.1% N With comineralizer of 0.1% N With comineralizer of 0.1% N With comineralizer of 0.1% N With comineralizer of 0.1% N
Ga metal left unreacted
Comments
164 T. Hashimoto and S. Nakamura
8
A Pathway Toward Bulk Growth of GaN by the Ammonothermal Method
165
Fig. 8.1 X-ray 2 ! scans of GaN powders synthesized with various mineralizers
In ammonothermal growth of GaN, growth temperature is limited by the creep temperature of autoclave materials, which is approximately at 600 ı C. Therefore, the phase selection nature of basic mineralizer is beneficial to avoiding potential phase mixing in grown crystals. In addition, chemical compatibility of basic NH3 with Ni Cr based superalloys is advantageous over acidic NH3 because the autoclave price becomes enormous when the entire wet surface is covered with a Pt-based liner.
8.3 Solubility of GaN in Ammonobasic Solutions Dissolution of a nutrient into a supercritical fluid is the first step to attain successful crystal growth in solvothermal growth. It is reported that reasonable solubility for hydrothermal crystallization is 2–5 wt% [32]. To achieve high solubility and optimum supersaturation for crystal growth, detailed solubility data are indispensable. Although comprehensive data on solubility of GaN in supercritical NH3 are not available for either basic or acidic conditions, a few measurement data have been presented thus far [17, 20, 23, 33, 34]. As Peters implied in his paper on synthesis of AlN in supercritical ammonobasic solutions [12], GaN shows retrograde solubility in supercritical ammonobasic solutions. At UCSB, we also observed an indication of retrograde solubility in the early stage of growth experiment [20], and confirmed it with a well-controlled weight loss method [35] as shown below. In the weight loss measurement presented in this section, the following points were taken into consideration to minimize measurement errors: (1) sufficient amount
166
T. Hashimoto and S. Nakamura
.5 g/ of polycrystalline GaN was loaded in a small-volume autoclave .22 ml/; (2) the small-volume autoclave was heated in a one-zone furnace to minimize temperature gradient in the autoclave; (3) polycrystalline GaN granules with large grain size (2–3 mm) were used; (4) NH3 was released at elevated temperature followed by rapid air-cooling to avoid precipitation of dissolved GaN on the granules; (5) GaOx and metallic Ga were removed from the surface of polycrystalline GaN. HVPE-grown polycrystalline GaN granules were first baked in an H2 =N2 ambient at 1;000 ı C for 4 h to remove oxides on the crystal surface. After H2 =N2 baking, the GaN granules were dipped in a 50% HCl aq. solution for 3 h to remove residual metals on the crystal surface. Then, the GaN granules were rinsed with DI water and dried in an oven at 120 ı C for more than 8 h. The pretreated GaN granules placed in a Ni-basket and NaNH2 (1.5% mol to NH3 ) were loaded in the autoclave and NH3 was filled. Then, the autoclave was heated for 120 h followed by rapid release of high-pressure NH3 and immediate air-cooling. The GaN granules were rinsed with isopropanol and DI water to remove Na and Na compounds. Before weighing, the GaN granules were dipped in a 50% HCl aq. solution for 3 h, rinsed with DI water, and dried in an oven at 120 ı C for more than 8 h to remove residual Ga and other metallic compounds. The surface of the GaN granules was characterized with SEM and electron-dispersive X-ray analysis (EDX). The total weight of the GaN granules was carefully weighed before and after the process with an electric balance with a resolution of 0.1 mg. The weight loss of the GaN granules was determined by subtracting the final weight from the initial weight. The possible measurement errors are weighing errors of GaN granules due to the resolution of the balance .˙0:2 mg/, errors of the GaN weight due to surface contamination of the granules .C0:5 mg/, and weighing errors of NH3 due to resolution of the balance .˙0:1 mg/. The upper and lower limit of the measurement errors were calculated for all dissolution values or solubility values, and presented as error bars in the figures. Figure 8.2a shows a surface of a GaN granule after H2 =N2 baking observed by SEM and Fig. 8.2b is a corresponding EDX spectrum collected from the entire area of the SEM image. No O was detected by EDX analysis indicating that the surface was effectively reduced by H2 =N2 baking. White particles on the surface were also confirmed to be O-free by EDX. Weak N peak indicates that the surface of the GaN granule was Ga rich, although no evident Ga droplet was observed by SEM. Aluminum was also detected indicating possible physical transport of Al from Al2 O3 reactor tube during H2 =N2 baking. Figure 8.3a shows a surface of a GaN granule after HCl dip observed by SEM and Fig. 8.3b is a corresponding EDX spectrum collected from the entire area of the SEM image. The enhanced N peak and the absence of Al peak in the EDX spectrum indicate effective removal of metallic Ga and Al from the surface. The crater-like holes observed by SEM are probably due to etching of Ga islands on the surface. Baking in H2 =N2 followed by HCl dip is an important pretreatment to minimize measurement error in the weight loss method. A surface treatment of the GaN granules after run was also important to minimize measurement error. Figure 8.4a shows an SEM image of a GaN granule surface before weighing. The granules were rinsed with isopropanol and DI water, dipped in HCl, followed by a rinse with DI water. Although a wide area EDX spectrum
8
A Pathway Toward Bulk Growth of GaN by the Ammonothermal Method
167
Fig. 8.2 (a) A surface of a GaN granule after H2 =N2 baking observed with SEM. (b) An EDX spectrum collected from the entire area of the SEM image
(Fig. 8.4b) did not show any other peak than Ga and N, there were many white dots observed on the surface. The focused EDX measurement on the white dot revealed that the dot contained O and Cl (Fig. 8.4c). These white dots were probably Ga O Cl complex formed from gallium amide or gallium imide compounds during the surface treatment after run. No evidence of etching on the GaN surface was observed after HCl dip. To estimate solubility of GaN in supercritical ammonobasic solutions, weight loss was measured as a function of time at 450 ıC and 650 ı C. The NH3 fill factors at each temperature were determined so that the pressure at high temperature became 76 ˙ 12 MPa .11;000 ˙ 1;800 psi/. The wide pressure allowance was due to pressure increase or a minor NH3 leak during the run. As shown in Fig. 8.5, weight loss saturated after about 120 h at 450 ıC. This means that the supercritical ammonobasic solution was saturated with GaN after 120 h. From the amount of loss, the solubility of GaN in this solution was estimated to be 0.11 wt%. On the other hand, the saturation occurred in about 26 h at 650 ı C. This means that a kinetic process in dissolving was enhanced at higher temperature. Based on this experiment, the duration of 1 run was determined to be 120 h (5 days) for the rest of the experiments. Figure 8.6 shows solubility of GaN in supercritical ammonobasic solution as a function of temperature. Retrograde solubility was observed at temperature higher than 600 ıC in this experiment.
168
T. Hashimoto and S. Nakamura
Fig. 8.3 (a) A surface of a GaN granule after HCl dip observed with SEM. (b) An EDX spectrum collected from the entire area of the SEM image
There are three possible reasons for retrograde solubility of GaN in supercritical ammonobasic solutions. The first possible reason is dissociation of NH3 at high temperature. The solubility decreased at higher temperature range .600–650 ı C/ at which NH3 might start to dissociate into N2 and H2 [36]. Although the equilibrium temperature and pressure for the dissociation of supercritical NH3 cannot be calculated due to lack of thermodynamic data at high temperature and high pressure, pressure increase observed during the typical ammonothermal growth indicates possible dissociation of NH3 . If NH3 dissociates into N2 and H2 , the total amount of NH3 to dissolve GaN decreases, resulting in underestimation of solubility. The second possible reason is formation of gallium amide or gallium imide compounds. As shown in the previous section, formation of gallium amide or gallium imide compounds is more favored than formation of GaN at temperature below 600 ıC. This means that more of GaN might be transformed into gallium amide or gallium imide at lower temperature, resulting in higher solubility of GaN at lower temperature. The third possible reason is the effect of entropy in dissolution process. The standard free energy of solution, G is expressed as follows: G D H T S; where H and S are changes in enthalpy and entropy associated with dissolution, and T is temperature. In dissolution of ionic salts, an entropy contribution to the free
8
A Pathway Toward Bulk Growth of GaN by the Ammonothermal Method
169
Fig. 8.4 (a) An SEM image of a GaN surface after run. (b) An EDX spectrum collected from the entire area of the SEM image. (c) An EDX spectrum collected from the white dot observed in the SEM image
energy cannot be neglected since the enthalpy changes associated with dissolution process are usually comparatively small [37]. It is reported that the entropy of solution is exceedingly negative when the anion and cation that would be formed on dissociation are both triply charged [37]. This might be the case in GaN dissolution, i.e., the entropy of GaN dissolution into NH3 is negative, resulting in higher free energy at higher temperature. The situation where entropy becomes negative on dissolution can be explained as follows: If GaN generates Ga3C on dissolution, three 3C NH 3NH 2 molecules are attracted, forming an ordered structure of Ga 2 complexes. Therefore, the randomness of the NH3 GaN system without Ga3C 3NH 2 complexes becomes higher than that with Ga3C 3NH 2 complexes.
170
T. Hashimoto and S. Nakamura
DISSOLUTION (wt%)
0.300%
NaNH2 1.5%, 450°C NaNH2 1.5%, 650°C
0.200%
0.100%
0.000%
0
100 200 TIME (hour)
300
Fig. 8.5 Weight loss of GaN granules as a function of time. The pressure was 76 ˙ 12 MPa .11;000 ˙ 1;800 psi/ and the mineralizer was 1.5%mol NaNH2
Fig. 8.6 Solubility of GaN in supercritical ammonobasic solutions as a function of temperature. The pressure was 76 ˙ 12 MPa .11;000 ˙ 1;800 psi/ and the mineralizer was 1.5%mol NaNH2
Currently, there is no evidence determining which of the three factors has a major influence on the retrograde solubility of GaN. Further detailed study must be carried out to answer this question.
8.4 Seeded Growth of GaN with Metallic Ga Nutrient To apply the ammonothermal method to bulk GaN growth, it is necessary to grow GaN on seed crystals via fluid transport. As shown in the previous section, the solubility of GaN in supercritical NH3 with 1.5 mol% NaNH2 was at most 1.2 wt%. In addition, it took a few days to saturate GaN in supercritical NH3 . Given this low solubility and slow kinetic process, the growth rate of GaN using GaN nutrient was expected to be extremely low for low mineralizer concentration. In the early stage of experiment at UCSB, there was a limitation in the mineralizer concentration due to clogging of high-pressure tubing. Therefore, growth of GaN using metallic
8
A Pathway Toward Bulk Growth of GaN by the Ammonothermal Method
171
Ga nutrient was attempted first. In this section, successful growth of GaN on HVPE-grown GaN seed via fluid transport of metallic Ga nutrient is reviewed. Seeded growth was carried out in an autoclave made of Ni Cr-based superalloy. The inner room of the autoclave was divided into two regions with a baffle. Considering the retrograde solubility, the Ga nutrient was placed at a lower-temperature zone (an upper zone) and single-crystalline GaN seed crystals were placed at a higher-temperature zone (a lower zone). 1.5 mol% NaNH2 was added to NH3 as a mineralizer. The NH3 pressure was about 150–200 MPa. The surface morphology and the thickness of the GaN films grown on the seeds were characterized by SEM. Selected single-crystal samples were characterized by Philips high-resolution fouraxis X-ray diffractometer using the Cu K’ line monochromated with a four-crystal Ge (002). The X-ray tube was operated at 45 keV and 40 mA. The sample was evaluated by measuring both on-axis and off-axis rocking curves. The line-width of the 0002 reflection was used for evaluating dislocations with a screw component N reflection was used to evaluate dislocations with an and the line-width of the 2021 edge component. The microstructure of the films was characterized by transmission electron microscopy (TEM: FEI Tecnai 20) operated at 200 kV. TEM samples were prepared for observation nearly along the axis by mechanical polishing followed by ArC ion milling (Gatan PIPS). For determination of polarity, convergent beam electron diffraction (CBED) was observed when the samples were characterized by TEM. The experimental CBED patterns were compared with patterns simulated by the electron microscopy simulation (EMS) software package with 33 zero-order reflections [38]. All TEM and CBED characterizations were carried out by Dr. Feng Wu. Figure 8.7a, b are cross-section of GaN films grown with 100 g of Ga nutrient on Ga-face and N-face, respectively. The temperatures measured inside the autoclave in the seed zone and the nutrient zone were 550–565 ı C and 520–530 ı C, respectively. After 19-day growth, 17 m-thick GaN and 45 m-thick GaN were grown on the Ga-face and the N-face, respectively. The resulting growth rates were 0:9 m/day for the Ga-face and 2:4 m/day for the N-face. As explained below, the actual growth happened within approximately 1 day. The film on the Ga-face consisted of hexagonal columns as shown in Fig. 8.7c, whereas the N polar surface was featureless as shown in Fig. 8.7d. A cross-sectional TEM observation of the film grown on the N-face revealed numerous voids and defects at the interface between the seed and the film. However, at the surface of the film no void was observed and defects were greatly reduced. The TEM images taken with g D 000 2N and g D 1 1N 00 showed that the majority of the dislocations were mixed-character dislocations. It was also confirmed that there was no evidence of basal-plane stacking faults in the surface region. The dislocation density estimated from the cross-sectional TEM images was in the low-109 cm2 level. The TEM images of this sample can be found in [39]. Due to the voids at the interface, the grown film was easily separated from the seed crystal. The free-standing GaN platelet separated from the N-polar surface of the seed crystal was characterized by X-ray rocking curve (XRC). The FWHM of XRC from 0002 reflection and 20 2N 1 reflection were 3,400 arcsec and
172
T. Hashimoto and S. Nakamura
Fig. 8.7 (a, b) Cross-sectional SEM images, and (c, d) plan-view SEM images of the GaN films grown on a GaN seed crystal with 100 g of Ga. (a, c) the Ga-side, (b, d) the N-side
6,300 arcsec, respectively. The high value of FWHM is considered to originate from poor structural quality near the interface between the film and the seed crystal. Hall measurement was carried out on the free-standing GaN platelet. The electron con1 centration and mobility were 5:4 1019 cm3 and 73 cm2 Vs , respectively. The high carrier concentration is probably due to high impurity concentrations and high defect density. One of the greatest advantages of solvothermal growth is its excellent uniformity. Since the convective flow of the solvent during the growth is typically high enough to attain low distribution of temperature within an independent region (i.e., a nutrient region or a seed region), growth uniformity is excellent compared to other growth methods. Growth shown in Figs. 8.8 and 8.9 demonstrate the excellent uniformity of GaN film grown on 3 4 cm2 oval shaped seed. The growth temperature was 675 ıC and the nutrient temperature was 625 ı C. Note that these temperatures are values measured on the external wall of the autoclave. The temperature inside the autoclave was estimated to be about 50–100 ıC lower than these values. Growth was carried out for 1 day with 40 g of Ga nutrient. The surfaces of the grown films on the Ga-face and N-face are shown at the top left corner of the Figs. 8.8 and 8.9, respectively. The grown films showed yellowish color and the Ga-face was relatively rough whereas the N-face showed a specular surface. The surface morphology at various points
8
A Pathway Toward Bulk Growth of GaN by the Ammonothermal Method
173
Fig. 8.8 A photograph of GaN film grown on a large-sized GaN seed crystal (Ga-face) and surface morphology at various points observed with SEM
Fig. 8.9 A photograph of GaN film grown on a large-sized GaN seed crystal (N-face) and surface morphology at various points observed with SEM
on the Ga-polar surface was observed with SEM as shown in Fig. 8.8. The Ga-polar surface was filled with numerous pits whose sizes ranged from submicron to 10 m. However, the surface morphology was uniform over the entire growth area, indicating the excellent scalability of the ammonothermal method. Similarly, the surface morphology at various points on the N-polar surface was observed with SEM as shown in Fig. 8.9. The morphology of N-polar surface was featureless except for scattered dots of GaN. The N-face was also very uniform over the entire growth area. This growth demonstrated excellent uniformity of ammonothermal growth. Despite these achievements, there was a critical problem in the growth with a metallic Ga nutrient. Figure 8.10 shows the relationship between the thickness of the film and the growth time when metallic Ga was used as a nutrient. As clearly shown in the figure, the thickness did not increase after 12 h. We have achieved maximum growth rate of 88 m per day for optimized 12-h growth; however, the growth did not continue for extended time. This was because the metallic Ga nutrient was
174
T. Hashimoto and S. Nakamura
Fig. 8.10 Thickness of grown films as a function of growth time in the case of a metallic Ga nutrient
transformed into GaN in the nutrient crucible during growth. Once the metallic Ga is converted to GaN, the dissolution of Ga-containing species into NH3 decreased significantly and GaN ceased to grow. Another problem associated with metallic Ga nutrient is that GaN film started to grow at very low temperature .200–300 ıC/ due to high reactivity of metallic Ga in the supercritical ammonobasic solution resulting in poor growth interface. It is challenging to control the growth initiation in the closed system of ammonothermal growth. Due to these critical issues, we decided to switch to polycrystalline GaN nutrient.
8.5 Seeded Growth of GaN with Polycrystalline GaN Nutrient In this section, seeded growth of GaN by the ammonothermal method with polycrystalline GaN nutrient is reviewed. As explained in the previous section, GaN growth from Ga nutrient continued only for about 12 h. By utilizing polycrystalline GaN nutrient with increased mineralizer concentration (3.9–4.2 mol% of NaNH2 ), continuous growth up to 90 days was achieved as shown in Fig. 8.11. Improved microstructure in the crystal grown for 50 days was confirmed with TEM as shown in Fig. 8.12a, c. The cross-section samples were fabricated with focused ion beam (FIB). It was confirmed with an optical microscope that the location of the FIB hole (i.e., the place where the cross-sectional sample was obtained) was on the growth interface as shown in Fig. 8.12b. Cross-sectional TEM images were observed under bright field condition with g D 11 2N 0 and g D 0002. Figure 8.12a shows cross-sectional TEM micrographs at the interface on the Ga-polar
8
A Pathway Toward Bulk Growth of GaN by the Ammonothermal Method
175
Fig. 8.11 Thickness of grown crystal along Cc, c and a direction as a function of growth duration
side. The interface could not be identified and no dislocation was observed. Figure 8.12c shows cross-sectional TEM micrographs on the N-polar side. Unlike the Ga-polar side, a clear interface was observed on the N-polar side. One void (indicated as “V” in the image) and one mixed-character dislocation propagating horizontally (indicated as “D” in the image) were identified. To estimate the threading dislocation density (TDD), plan-view TEM images were observed under bright field condition with g D 11 2N 0. There were no dislocations observed on the Ga-face and only several dislocations were observed on the N-face. The TDD on the Ga-face was estimated to be less than 1 106 cm2 and TDD on the N-face was estimated to be 1 107 cm2 . Despite the excellent microstructure, the grown crystals contained multiple grains, which deteriorated FWHM of XRC as summarized in Table 8.2. The FWHMs of XRC from the Ga-polar surface of the 40-m crystal (50-day growth) were approximately 290 arcsec and 110 arcsec for 0002 and 20 2N 1 reflections, respectively. Although these numbers are comparable to the FWHMs for the seed crystals, the additional 90-day growth (total thickness of 160 m) resulted in higher FWHM values of 1,400 arcsec and 650 arcsec for 0002 and 20 2N 1 reflections, respectively. Similarly, the FWHMs of XRC from the N-polar surface of the 180-m crystal (50-day growth) were approximately 840 arcsec and 490 arcsec for 0002 and 20 2N 1 reflections, respectively and the additional 90-day growth (total thickness of 430 m) resulted in FWHMs of 1,400 arcsec and 570 arcsec for 0002 and 20 2N 1 reflections, respectively. It is surmised that GaN crystal grew under strain on HVPEgrown GaN seed, of which the crystal lattice is bowed and after reaching certain thickness the stress is relaxed resulting in increase in FWHM. Figure 8.13 is a direct observation of the grain boundary on Ga-polar surface with Nomarski micrograph after selective etching in a hot .160 ı C/ H3 PO4 solution. The detail of the etching is explained elsewhere [26, 27]. As indicated with an arrow, the grain boundary was clearly observed on the surface. The dimension of each grain was in the range of 500 m.
176
T. Hashimoto and S. Nakamura
Fig. 8.12 (a) Cross-sectional TEM images of the interface on the Ga-face with g D 0002 and g D 112N 0. No dislocation was observed. The interface was not identified with TEM. Note that the horizontal line is not an interface but damage during sample preparation. (b) Comparison of cross-sectional surface before and after the FIB process observed with a Nomarski microscope. It was confirmed that the “FIB holes” lie on the interface for both N-polar side and Ga-polar side. (c) Cross-sectional TEM images of the interface on the N-face with g D 0002N and g D 112N 0. One void (indicated as “V”) and one dislocation propagating horizontally (indicated as “D”) were observed. The interface was clearly identified as indicated with lines in the picture
Preliminary SIMS measurement was carried out for the crystal grown for 50 days. The crystal on the Ga-polar side contained 2 1019 cm3 of oxygen and 1 1018 cm3 of carbon. The crystal on the N-polar side contained 6 1019 cm3 of oxygen and 4 1018 cm3 of carbon. In addition, other impurities such as Li, Na,
8
A Pathway Toward Bulk Growth of GaN by the Ammonothermal Method
177
Table 8.2 FWHMs of XRC obtained from GaN crystals grown by the ammonothermal method with a polycrystalline GaN nutrient Sample 0002 reflection (arcsec) 202N 1 reflection (arcsec) Ga-face HVPE-grown seed crystal 280 90 40 m (50 days) 290 110 160 m (90-day regrowth 1;400 650 on 50-day crystal) N-face HVPE grown seed crystal 360 100 180 m (50 days) 840 490 430 m (90-day regrowth 1;400 570 on 50-day crystal)
Fig. 8.13 A Nomarski micrograph of the Ga-face after selective etching. The arrow indicates a grain boundary
Mg, Al, Ni, Cr, and Mn were detected. Reduction of impurities will be one of the future challenges in the ammonothermal method.
8.6 Growth of Bulk GaN Crystals and Sliced Wafers GaN in bulk shape has been obtained in 82-day growth with increased growth rate. Figure 8.14 shows bulk GaN crystals. The crystals showed dark brownish color. We surmise that the origin of the color center is a combination of impurities and point defects. The crystals were in polyhedron shape with a flat hexagonal bottom and sidewalls. The flat hexagonal surface was c plane and the sidewalls were m planes. The Ga-polar surface was decorated with semipolar facets. The height (i.e., thickness) of the crystal was 4.5 mm and 5.1 mm for the crystals in Fig. 8.14a (Sample A), b (Sample B), respectively. The grown thickness along c directions, which represents the addition of grown thicknesses along Cc and c directions was
178
T. Hashimoto and S. Nakamura
Fig. 8.14 Bulk GaN crystals grown by the ammonothermal method. (a) Sample A, (b) Sample B
approximately 4.1 mm and 4.6 mm for Sample A and Sample B, respectively. The lateral dimension of Sample A and Sample B was approximately 5.7 mm (pointto-point) and 7 mm (the longest side), respectively. Since the polycrystalline GaN nutrient had been completely consumed when the autoclave was open, we suspect the actual growth duration was less than 82 days. Therefore, the average growth rate along the c direction is estimated to be more than 50 m per day for Sample A and 55 m per day for Sample B. Since Sample B was placed at a lower position than Sample A in the autoclave, where the temperature was higher, the growth rate was higher due to larger temperature difference. One crystal was sliced to obtain 400-m-thick c plane wafers as shown in Fig. 8.15. The picture on the left in Fig. 8.15 shows the side view of the crystal before slicing. The parallel lines and the rectangle represent the approximate wire positions and the seed position, respectively. As clearly seen in Fig. 8.15, the growth nature was strongly dependent on growth directions. The wafer #1 has a relatively transparent rectangular region of the seed and laterally grown regions. The growth rate along a directions was much faster than that along m directions. The dimensions of the grown crystal along a directions and m directions were 1.5 and 0.3–0.5 mm, respectively. As seen in the wafer #2, the crystal grown along c direction was
8
A Pathway Toward Bulk Growth of GaN by the Ammonothermal Method
179
Fig. 8.15 Photographs of c plane wafers. The left picture is a crystal before slicing and right pictures are sliced wafers. The parallel lines represent the approximate wire positions and the rectangle represents the approximate seed position. The sliced wafers were labeled from top to bottom as #1 to #4
Fig. 8.16 Photographs of m plane wafers. The left picture is a crystal before slicing and right pictures are sliced wafers. The parallel lines represent the approximate wire positions and the trapezoid represents the approximate seed position. The arrows on the wafers represent the lateral interface between the seed and the grown crystal. The sliced wafers were labeled from top to bottom as #1 to #7
significantly colored. Also, one can see that the dark region grew laterally as crystal grew along c direction, indicating the growth along c direction was dominant over domains growing along a or m directions. Another crystal was sliced to obtain m plane wafers as shown in Fig. 8.16. The picture on the left in Fig. 8.16 shows the top view of the crystal before slicing. The
180
T. Hashimoto and S. Nakamura
Fig. 8.17 XRD rocking curves from 10 1N 0 reflection of the m plane wafers shown in Fig. 8.16. The wafer numbers correspond with those in Fig. 8.16
parallel lines and the trapezoid represent the approximate wire positions and the seed position, respectively. The arrows in the pictures on the right represent the lateral boundary between the seed platelet and the grown crystal. Again, one can clearly see growth anisotropy: the growth along Cc direction shows more grain structure with f10 1N 1g facets, the growth along c direction shows the darkest color and the growth along a direction shows the lightest color merging with domains growing along Cc and c directions. Since the wafers #1 and #7 consist of laterally grown domains, they showed better structural quality as shown later. The grown thickness along Cc and c directions was 2.5 and 2.1 mm, respectively. Figure 8.17 shows XRC from 10 1N 0 reflection of the m plane wafers shown in Fig. 8.16. The XRC indicates the multiple grain structure. As discussed above, the wafer #1 which predominantly consists of a laterally grown domain showed the narrowest FWHM of 220 arcsec. The wafer #7 also showed relatively narrow FWHM of 420 arcsec, although the rocking curve is not shown in Fig. 8.17. This implies that the degradation of the crystal quality arises from insufficient structural quality of the HVPE-grown seed crystals.
8.7 Summary Starting from powder synthesis, we have developed our own technology of ammonothermal growth at UCSB. Although there still remain a lot of uninvestigated matters in this method, the bulk crystals with three-dimensional polyhedron shape in appreciable size demonstrate the high feasibility of ammonothermal growth. The obtained growth rate was in the practical range for commercialization of ammonothermal growth. At this moment, multiple grains and coloration are the major challenges in our research; however, these problems will be solved in the near future by
8
A Pathway Toward Bulk Growth of GaN by the Ammonothermal Method
181
engineering. We believe that the ammonothermal growth is the enabling method to realize our “dream” bulk crystals of GaN. Acknowledgements This research was conducted as a part of Nakamura Inhomogeneous Crystal Project of Exploratory Research for Advanced Technology (ERATO) program funded by Japan Science and Technology Agency (JST). The authors appreciate the long-term and flexible funding of JST. The authors thank Professor James S. Speck and Dr. Feng Wu at UCSB for TEM observation and fruitful discussions. Also, the authors thank Mr. Kenji Fujito and Mr. Makoto Saito at Mitsubishi Chemical Corporation for participating in the project as visiting researchers. In addition, the authors acknowledge Mitsubishi Chemical Corporation for supplying HVPE-grown GaN platelets. The authors are also thankful to Takatori Corporation and Yasunaga Wire Saw Systems Corporation for slicing the crystals with wire saws.
References 1. R.W. Brander, R.P. Sutton, J. Phys. D (Appl. Phys.) 2, 309 (1969) 2. M.A. Haase, J. Qiu, J.M. DePuydt, H.C. Cheng, Appl. Phys. Lett. 59, 1272 (1991) 3. S. Nakamura, T. Mukai, M. Senoh, Jpn. J. Appl. Phys. Part I 32, L16 (1993) 4. S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa, T. Yamada, T. Matsushita, H. Kiyoku, Y. Sugimoto, Jpn. J. Appl. Phys. Part II 35, L74 (1996) 5. S. Porowski, MRS Internet J. Nitride Semicond. Res. 4S1, G1.3 (1999) 6. T. Inoue, Y. Seki, O. Oda, S. Kurai, Y. Yamada, T. Taguchi, Phys. Stat. Sol. (B) 223, 15 (2001) 7. H. Yamane, M. Shimada, T. Sekiguchi, F.J. DiSalvo, J. Cryst. Growth 186, 8 (1998). 8. F. Kawamura, M. Morishita, K. Omae, M. Yoshimura, Y. Mori, T. Sasaki, Jpn. J. Appl. Phys. 42, L879 (2003) 9. T. Hashimoto, F. Wu, J.S. Speck, S. Nakamura, Jpn. J. Appl. Phys. 46, L889 (2007) 10. M.C. Schmidt, K.C. Kim, H. Sato, N. Fellows, H. Masui, S. Nakamura, S. DenBaars, J.S. Speck, Jpn. J. Appl. Phys. 46, L126 (2007) 11. A. Tyagi, H. Zhong, N.N. Fellows, M. Iza, J.S. Speck, S.P. DenBaars, S. Nakamura, Jpn. J. Appl. Phys. 46, L129 (2007) 12. D. Peters, J. Cryst. Growth 104, 411 (1990) 13. R. Dwilinski, A. Wysmolek, J. Baranowski, M. Kaminska, R. Doradzinski, J. Garczynski, L. Sierzputowski, H. Jacobs, Acta Phys. Polonica A 88, 833 (1995) 14. R. Dwilinski, R. Doradzinski, J. Garczynski, L. Sierzputowski, M. Palczewska, A. Wysmolek, M. Kaminska, MRS Internet J. Nitride Semicond. Res. 3, 25 (1998) 15. D.R. Ketchum, J.W. Kolis, J. Cryst. Growth 222, 431 (2001) 16. A.P. Purdy, R.J. Jouet, C.F. George, Cryst. Growth Des. 2, 141 (2002) 17. R.T. Dwilinski, R.M. Doradzinski, J. Garczynski, L.P. Sierzputowski, Y. Kanbara, U. S. Patent 6656615 B2 (2003) 18. M.J. Callahan, B. Wang, L.O. Bouthillette, S.Q. Wang, J.W. Kolis, D.F. Bliss, Mater. Res. Soc. Symp. Proc. 798, Y2.10 (2004) 19. A. Yoshikawa, E. Ohshima, T. Fukuda, H. Tsuji, K. Oshima, J. Cryst. Growth 260, 67 (2004) 20. T. Hashimoto, K. Fujito, B.A. Haskell, P.T. Fini, J.S. Speck, S. Nakamura, J. Cryst. Growth 275, e525 (2005) 21. T. Hashimoto, K. Fujito, M. Saito, J.S. Speck, S. Nakamura, Jpn. J. Appl. Phys. 44, L1570 (2005) 22. T. Hashimoto, K. Fujito, R. Sharma, E.R. Letts, P.T. Fini, J.S. Speck, S. Nakamura, J. Cryst. Growth 291, 100 (2006) 23. M. Callahan, B.G. Wang, K. Rakes, D. Bliss, L. Bouthillette, M. Suscavage, S.Q. Wang, J. Mater. Sci. 41, 1399 (2006)
182
T. Hashimoto and S. Nakamura
24. B. Wang, M.J. Callahan, K.D. Takes, L.O. Bouthillette, S.-Q. Wang, D.F. Bliss, J.W. Kolis, J. Cryst. Growth 287, 376 (2006) 25. Y. Kagamitani, D. Ehrentraut, A. Yoshikawa, N. Hoshino, T. Fukuda, S. Kawabata, K. Inaba, Jpn. J. Appl. Phys. 45, 4018 (2006) 26. T. Hashimoto, F. Wu, J.S. Speck, S. Nakamura, Jpn. J. Appl. Phys. 46, L525 (2007) 27. T. Hashimoto, F. Wu, J.S. Speck, S. Nakamura, Nat. Mater. 6, 568 (2007) 28. P.P. Molinie, R. Brec, J. Rouxel, P. Herpin, Acta Crystallogr. B29, 925 (1973) 29. H. Jacobs, B. Nocker, Zeitschrift für Anorganische und Allgemeine Chemie 619, 381 (1993) 30. X.H. Wu, D. Kapolnek, E.J. Tarsa, B. Heying, S. Keller, B.P. Keller, U.K. Mishra, S.P. DenBaars, J.S. Speck, Appl. Phys. Lett. 68, 1371 (1996) 31. A. Munkholm, C. Thompson, C.M. Foster, J.A. Eastmen, O. Auciello, G.B. Stephenson, P. Fini, S.P. DenBaars, J.S. Speck, Appl. Phys. Lett. 72, 2972 (1998) 32. A.A. Ballman, R.A. Laudise, Hydrothermal Growth, The Art and Science of Growing Crystals, ed. by J.J. Gilman (Wiley, New York 1963), p. 231 33. T. Hashimoto, M. Saito, K. Fujito, F. Wu, J.S. Speck, S. Nakamura, J. Cryst. Growth 310, 311 (2008), note that the non-polar plane is mislabeled as m-plane which should be corrected as a-plane 34. D. Ehrentraut, Y. Kagamitani, C. Yokoyama, T. Fukuda, J. Cryst. Growth 310, 891 (2008) 35. K. Byrappa, M. Yoshimura, Handbook of Hydrothermal Technology, Chap. 4 (Noyes, London, 2001) 36. H. Jacob, S. Schmidt, High-pressure ammonolysis in solid-state chemistry, Current Topics in Materials Science, Vol. 8, Chap.5, ed. by E. Kaldis (North-Holland, Amsterdam, 1982) 37. D.A. Johnson, Some thermodynamic aspects of inorganic chemistry, 2nd edn., Chap. 5 (Cambridge University Press, Cambridge, 1968, 1982) 38. P.A. Stadelmann, Ultramicroscopy 21, 131 (1987) 39. T. Hashimoto, K. Fujito, F. Wu, B. A. Haskell, P. T. Fini, J.S. Speck, S. Nakamura, Jpn. J. Appl. Phys. 44, L797 (2005)
Chapter 9
Acidic Ammonothermal Growth Technology for GaN Dirk Ehrentraut and Yuji Kagamitani
Abstract The ammonothermal technique for the crystal growth of bulk GaN is presented. General aspects are discussed and the use of acidic mineralizer is focused on. This method is a recent development to challenge the problem of latticematched substrate crystal for future group-III nitride device technology. Chemical and physical properties are discussed and an outlook to the future prospective and technological requirements is given.
9.1 Introduction The accessibility of lattice and thermally matched substrates is typically a great benefit in the design and fabrication of optimized device structures providing high efficiency and lifetime [1]. The hexagonal phase of the wide bandgap material gallium nitride (GaN; space group P63 mc, with lattice parameters a D 3:189 Å and c D 5:186 Å), to which we refer as GaN unless otherwise mentioned, is of significant importance for numerous applications in the big market of electronics and optoelectronics because of its excellent properties (Eg D 3:4 eV, breakdown 1 voltage 3 MV cm1 , electron mobility 1; 000 cm2 V s1 , and high saturation velocity); see, for example, Chap. 1 in this book. It is widely recognized that providing high-quality GaN substrates at a reasonable price is the key toward further development of next-generation electronic and optoelectronic devices. Different approaches have been chosen in the extensively employed vapor phase technologies like hydride vapor phase epitaxy (HVPE) or molecular beam epitaxy (MBE) to lessen the problem of lacking lattice and thermally matched substrates by some means. Templated single crystalline ’-Al2 O3 (sapphire) substrate of different crystallographic orientations (polar, nonpolar) is the most prominent one realized to grow GaN crystals of about 10 mm thickness along the h0001i axis [2]. Other mateN 6H-SiC [4, 5], (111) and rials of choice involve (111)GaAs [3], (0001) and .0001/ N (001)Si [6], .100/”-LiAlO2 [7], (0001) and .0001/ ZnO [8, 9] to mention but a few. The GaN crystals grown on these substrates yet contain a large amount of crystal defects like grain boundaries, threading dislocations, stacking fault, etc. Whatever 183
184
D. Ehrentraut and Y. Kagamitani
be the foreign substrate material, it is the thermal and lattice mismatch between substrate and GaN that causes the formation of crystal defects in high densities, strain and subsequently bowing of the final product epi-ready GaN wafer. What is more, the substrate must often be removed if realization of a transparent and electrically conductive back contact is considered. Free-standing GaN wafers of two inches in size (minimum requirement) and low defect concentration around 104 cm2 are in vital demand in the industry. The answer to the problem could lie in the growth of GaN by a process from liquid solution. Growth from the solution is generally attractive in that grown crystals are of high structural perfection and high throughput can be achieved. Quartz .’-SiO2 / and more recently ZnO have paved the way for the solvothermal growth technique [10]. More specifically, the solvothermal growth technique basically employs a polar solvent to dissolve and successively form metastable products with the solute. Later, after transport to the growth zone, under slightly different temperature conditions, crystallizes the desired phase of GaN on a seed of similar crystal structure. Mineralizers are essential to amplify the solubility of the solute in the solvent. Since a closed system is utilized, i.e., exchange of matter with the ambient is impossible, the solvent takes over a supercritical (SC) state, which also improves the solubility of the nutrient. Basically, acidic and basic mineralizers are employed in the ammonothermal growth of GaN. Whereas the latter are known for its retrograde solubility (also compare Chaps. 7 and 8 of this book), acidic mineralizers can be utilized under classic conditions, i.e., crystal growth occurs in the cooler zone of an autoclave or upon lowering system temperature. The solvothermal crystal growth technology, if compared to the vapor phase growth technologies, is characterized by the following merits: (1) the operation near the thermodynamic equilibrium for a given temperature–pressure range. This leads to generate an extremely low supersaturation in the solution and consequently a high crystallinity can be expected. (2) The temperature gradient at the growing interface is practically zero which results in stable growth conditions. (3) Scalability: large crystal quantities can be controlled over long process time, thus enabling a high throughput. Growth processes employing multiple seeds (around 2,000 quantity) are applied for the low temperature phase of quartz, ’-SiO2 . (4) Environmentally benign conditions for production and capability for complete recycling of the used solution. (5) This technology is used for a long time, i.e., >60 years to mass-produce crystals like ’-SiO2 . Like with most of the regularly used technologies, there are certain limitations in that the incorporation of constituents or reaction products from the solution on a mesoscopic scale may occur if the temperature inside the growth vessel is not controlled precisely. The homogeneous incorporation of dopants over long growth cycles may be hard to manage since the reaction rate of the host crystal to be grown and the dopant under similar conditions usually differ. This effect will result in graded crystals, i.e., changes in the concentration of the dopant along the growth direction of a crystal facet. It should be known here that the process performed at the vicinity to the thermal equilibrium does not allow the growth of thermodynamically unstable solid solutions. This in turn limits the fabrication of lattice-matched
9
Acidic Ammonothermal Growth Technology for GaN
185
alloyed substrates, but is a general feature for growth techniques from the liquid solution. However, the main purpose for solvothermal growth technologies lies in the fabrication of crystals of superior crystallinity to slice substrates for use in the following step of device fabrication, i.e., the main purpose is to provide lattice-matched, low-strain substrates of large size at reasonable production cost. Almost naturally, a high-throughput technique like the hydrothermal (ammonothermal) growth employing a solution indeed holds great promise to satisfy the market demand. As side note, an isostructural compound to GaN is the hexagonal zinc oxide (ZnO; space group P63 mc, with lattice parameters a D 3:253 Å and c D 5:213 Å), which is the first semiconductor crystal being fabricated by the hydrothermal technique: 3 in. size wafers have already been demonstrated [10]. However, the situation in the ZnO device technology is opposed to that in the GaN device technology such that ZnO wafers of high crystallinity are available but the device technology seemingly needs more progress before ZnO-based light emitters (LEDs) make their way to the market. Contrary, in case of GaN the device technology is already a multibillion US dollar business but high-quality and cheap GaN substrates are not yet available. This chapter comprehensively reports on the growth of GaN single crystals by the acidic ammonothermal path, including detailed analysis of important growthrelated parameters like solubility and chemistry of the solution, and will start with the historical developments from the early beginning.
9.2 Brief History of the Ammonothermal Growth Technique of GaN The ammonothermal technique is the most recent among the methods to produce bulk GaN crystals for future wafer fabrication as laid out in the book. The HVPE technique is currently the most widely used one (Chaps. 2–5). Other techniques with potential comprise the high-pressure solution (HPS, Chap. 10) growth and the Na-flux method (Chap. 11). Early experiments using SC ammonia have been focused on the materials synthesis, specifically of complex nitrides [11]. In the 1990s, nano- and microcrystals of AlN and GaN have been synthesized by the ammonothermal approach [12–15] and the potential of the method as a prospective candidate for bulk GaN crystal growth had been realized. It was, however, not until around the year 2000 that free-standing millimetersized GaN crystals have been grown by the ammonothermal technique: Ketchum and Kolis have grown 0:5 0:2 0:1 mm3 crystals in a basic ammonothermal environment over 10-day cycles at 400 ı C and 240 MPa [16]; Callahan et al. have used polycrystalline and HVPE-GaN as seed crystals to grow thick films using basic mineralizer and system pressure around 100–300 MPa [17]; Wang et al. have reported in 2006 the growth of GaN crystals up to 10 10 1 mm3 at growth rates 50 m
186
D. Ehrentraut and Y. Kagamitani
Fig. 9.1 First 2-in. size GaN produced by the acidic ammonothermal route grown from the autoclave shown in Fig. 9.4. HVPE GaN crystals served as seeds
per day [18]. Dwilinski et al. and Hashimoto et al. have reported their huge efforts over the past years in the basic ammonothermal technology; for details see Chaps. 7 and 8. Acidic mineralizers have been employed recently by Purdy et al., Yoshikawa et al., and Kagamitani and Ehrentraut et al. to synthesize free-standing hexagonal GaN crystals and homoepitaxial films [19–21]. More recently, the first successful trial to use a 2-in. size HVPE-GaN substrate to produce ammonothermally grown GaN was published [22]. Figure 9.1 shows a roughly 2-in. size GaN crystal, the first ever grown under acidic ammonothermal conditions. The acidic ammonothermal technique is by far the most juvenile, though swiftly developing, technology with substantial prospect to fabricate bulk size GaN crystals.
9.3 Growth Technology A closed system, so-called autoclave, is employed rugged enough to bear high pressure under elevated temperature. Important is the choice of a proper alloy for the autoclave to withstand the chemically harsh conditions over long term. In the case of the hydrothermal method, SC water .H2 O/ is used as solvent and Fe-based alloys are mostly chosen as for the autoclave material. Fe-based alloys have excellent mechanical properties; however, owing to corrosion they cannot be used under the conditions of SC NH3 C acidic mineralizers. Therefore,
9
Acidic Ammonothermal Growth Technology for GaN
187
corrosion-resistant Ni-based alloys (alloy 625, Rene 41, Nimonic 90, etc.) may be used as material of choice for the main body of the autoclave. Even for the aforementioned Ni-based alloys, though, little corrosion may occur depending on the acidity of the solution. Autoclaves for acidic ammonothermal growth processes are therefore equipped with an inner liner made of chemically inert platinum (Pt) to fully prevent the corrosion of the autoclave and successively incorporation of impurities into the growing crystal. The baffle plate, wire to fix seed crystals, and the basket to contain the precursor are all made of Pt. Figure 9.2 reveals a typical autoclave setup. A baffle serves to separate the volume into dissolution zone with higher temperature and growth zone with slightly lower temperature. Ga metal and/or polycrystalline GaN or other Ga-contained materials can be used as precursor. Gallium metal (Ga; 99.9999% purity) had been used initially as the precursor and was gradually replaced by polycrystalline GaN. The latter is a by-product obtained from the synthesis of hexagonal GaN by HVPE. The ratio of the molar concentration of mineralizer and precursor, calculated to .1 x/ GaN C x Ga, was up to 0.15. NH4 X (X D Cl, Br, I) is added as mineralizer. The mineralizers are of 99.999% phase purity and have been used in the molar ratio of mineralizer and ammonia in the range 102 . GaN single crystals are prepared as seed. Precursors and mineralizes are placed in the high-temperature zone and the seed crystals are mounted in the low-temperature zone. The autoclave is then sealed and thereafter vacuum pumped and successively flooded with nitrogen to remove oxygen. Generally, several methods maybe used to fill the autoclave with ammonia. In the high-pressure injection system, ammonia is injected into the vessel by a
Fig. 9.2 Photograph and sketch of an autoclave for the acidic ammonothermal growth of GaN at the research scale
188
D. Ehrentraut and Y. Kagamitani
high-pressure injection syringe. However, dirt-like oil in the cylinder can mix in the container, thus polluting the system. Another, yet better method is to process under low temperatures, so-called cooling introduction. This is suitable for small autoclaves. The autoclave is vacuum-pumped and cooled below the boiling point Tb of NH3 (under atmospheric pressure Tb D 33:35 ı C, melting point Tm D 77:7 ı C). Then, ammonia of >99:999% purity is flushed into the autoclave in a controlled way using a mass flow controller. The ammonia in the autoclave takes over the liquid state. Typically, we worked with the fill ranging 40–70% of NH3 . Already well working for small-size autoclaves, this filling technique is currently under development for large-scale autoclaves to hold 2-in. crystals. The final step prior growth is heating up the autoclave in a controlled way. We have performed the GaN growth in the temperature range T D 360–550 ı C under pressures p D 130–170 MPa. The temperature is kept constant during the growth process.
9.4 Chemistry of the Solution and Growth Mechanism 9.4.1 Solubility When compared to H2 O; NH3 is a weaker solvent for ionic substances since its dielectric constant ", a measure of the energy of solvation compared to the lattice energy, is smaller, i.e., " D 16:5 for NH3 compared to 80 for H2 O. This will result in the formation of low-crystallinity phases; however, working at much higher pressures may lead to partly overcome this problem as " increases at higher density of the solvent [11]. Anyway, it has become obvious that mineralizers are needed to enhance the solubility of GaN while processing under reasonable conditions. It has been shown in the case of GaN that the solubility strongly depends on the chemical environment in the ammonothermal solution. Retrograde solubility was reported for basic mineralizers like KNH2 [23], consequently, the seed crystals have to be assembled in the hot zone to achieve crystal growth while the precursor is placed in the colder zone of the autoclave where the solubility is highest. Contrary, the acidic mineralizers NH4 X in SC NH3 provides a positive solubility coefficient [24]. The precursor is placed in the hotter zone and the GaN seed crystals in the cooler zone inside the growth vessel. An advantage of the acidic mineralizer NH4 Cl over basic mineralizers such as KNH2 or NaNH2 is the high solubility in NH3 around room temperature where 124 g NH4 Cl .24:8 ı C/, 3.6 g KNH2 .25 ı C/, and 0.163 g NaNH2 .20 ı C/ per 100 g NH3 were reported [25]. Thus, a larger amount of GaN would be dissolvable in ammonia containing an acidic mineralizer in comparison to a basic mineralizer. We have worked on the measurement of the solubility of GaN under acidic ammonothermal conditions, in particular with NH4 Cl, and different molar concentrations c of the mineralizer. A high-pressure cell of 10 mL volume has been
9
Acidic Ammonothermal Growth Technology for GaN
a Solubility Gan [mol]
0.25
189
GaN: Molar ratio NH4Cl/NH3 = 0.0127; 0.032; 0.127 AIN: Molar ratio NH4Cl/NH3 = 0.0318
0.20
0.127
0.15 0.10 contains 8 mol% NH4I
0.05
0.032 AIN 0.0127
0.00 250
b
350 400 450 500 Temperature [°C]
550
Molar ratio mineralizer/NH3 = 0.31– 0.34
0.008 Solubility GaN [mol]
300
T = 431– 437 °C, p = 95 –100 MPa
0.006
NH4 Cl1 x Xx with: X=l X = Br
0.004 0.002 0.000 0.0
0.6 0.8 1.0 0.2 0.4 Molar content x in mineralizer NH4Cl1-x Xx
Fig. 9.3 Solubility of GaN (a) and AlN for different concentrations of the mineralizer NH4 Cl with respect to 1 mol NH3 and (b) for mixed mineralizer NH4 Cl1x Xx with X D Br, I
employed at pressures around 100 MPa [24, 26]. Experiments were conducted for 120 h under isothermal conditions so as to avoid partial recrystallization of the dissolved GaN. Very similar to the ammonothermal growth process for GaN, polycrystalline HVPE GaN has been used as precursor. In Fig. 9.3a is shown the solubility of GaN in NH4 Cl-containing NH3 . It gets clear that an increase of CNH4Cl as well as temperature results in an increase of CGaN in the solution. Also shown in Fig. 9.3a is a data set of the solubility of the isostructural AlN under comparable conditions, i.e., similar T and molar ratio NH4 Cl=NH3 about 0.032. AlN shows a somewhat higher solubility than the GaN. Increasing the acidity of the solution by adding NH4 I lowers the solubility of AlN. Derived from the single exponential growth of CGaN over T D 250–600 ı C, we have calculated the energy of formation HF D 15:9 kcal mol1 [24]. It was found that this value is independent of the concentration of mineralizer in NH3 and will provide a tool to calculate the overall energy needed to control the solubility of GaN,
190
D. Ehrentraut and Y. Kagamitani
such way controlling the growth of GaN under acidic ammonothermal conditions for a given system. In Fig. 9.3b is shown the solubility of GaN when the mixed mineralizer NH4 Cl.1x/ Xx with X D Br, I is employed. Increasing x results in a lowered solubility of GaN. Utilizing a fraction of NH4 Br gives a bit higher solubility than making use of NH4 I, but does not differ remarkably when the pure mineralizer NH4 Br or NH4 I, i.e., x D 1 is used.
9.4.2 Growth Rate and Chemistry of the Solution In general, the growth rate is firstly depending on the polarity, i.e., electrical charge and density of the ion attached to the crystal surface. The second contribution comes from the abundance of growth species drawn from the solution in front of the growing face of a crystal. Unlike the case of say NaCl dissolved in H2 O, no clear features exist from the chemical constitution of the acidic ammonothermal system at high pressure and temperature. Wang and Callahan [23] have recently summarized from the literature the results on the synthesis of GaN. They reported that the most favorable conditions for the growth of GaN under basic ammonothermal conditions with either purely KNH2 or mixed KNH2 and KN3 as mineralizers are at temperatures around 550 ıC in the hot zone and 172–310 MPa pressure. The ammonia fill must be >60%. In the case of acidic mineralizers such as NH4 Cl, the formation of ammonium chlorogallates being similar to .NH4 /3 GaCln is likely to happen, quite in analogy to ammonium hexafluorogallate, .NH4 /3 GaF6 [27]. The ammonolysis of the latter compound tends to produce GaN at about 400 ı C. As for acidic ammonothermal conditions, pentaaminechlorogallium (III) dichloride ŒGa.NH3 /5 ClCl2 has recently been suspected to be an effective precursor for the growth of GaN. This specie has been synthesized at 840 K [28], i.e., it is stable under the temperature conditions of acidic ammonothermal growth of GaN. The structure is described as cationic [Ga(NH3 )5 Cl]2C octahedra which are surrounded by distorted cubes of Cl anions. The later are weakly bonded and therefore relatively easily disconnect from the cationic octahedra in a given environment. Besides, the free Cl anions might easily form NH4 Cl, a mineralizer molecule again with the NH4C derived from the autoprotolysis of NH3 , accordingly: 2NH3 $ NHC 4 C NH2
(9.1)
If we now assume that the double positively charged ŒGa.NH3 /5 Cl2C does exist in the solution of SC NH3 and NH4 Cl, it might be the reason for the higher growth N face in comparison to the (0001) face. Quite in analogy to the rate of the .0001/ hydrothermal ZnO system with basic mineralizers like LiOH and KOH [10], positively charged growth species like ŒGa.NH3 /5 Cl2C are likely to be attracted by the N face rather than the positively charged (0001) face. We negatively charged .0001/
9
Acidic Ammonothermal Growth Technology for GaN
191
have earlier noted that only one process is controlling the chemical equilibrium since the Arrhenius plot from the solubility chart clearly shows a linear slope. More investigations are required to really understand whether ŒGa.NH3 /5 Cl2C or another, perhaps very similar, complex molecule is the key to the chemical equilibrium in the reactions of dissolving and depositing GaN under acidic ammonothermal conditions. The estimation of the maximum stable growth rate Vmax , at which high crystal quality may still obtain, gives an idea of whether a solution growth method is of principal interest in terms of future production process. Above Vmax entrapment of inclusions, columnar growth mechanism and formation of related crystal defects are most likely to happen. Thus far, nothing is known for GaN grown under ammonothermal condition, even ZnO is not well treated this way. The concept of the maximum stable growth rate basically derived from Nernst’s principle [29] and later modified by Carlson [30]: Vmax D
0:214Du 2 Ce2 Sc1=3 2 L
1=2 :
(9.2)
Here, the parameters are: D – diffusion coefficient, u – solution flow rate, – thickness of the diffusion boundary layer in front of the crystal face, Ce – equilibrium solute concentration obtained from the solubility, – density of the fluid, L – length of the crystal surface, Sc – Schmidt number. The Sc gives the ratio of friction and mass transport by diffusion, and can be calculated: Sc D = D
(9.3)
Playing the parameters by means of D 4:6 105 Pa s; .673 K/ D 0:39 g cm3 ; 1 D D 2104 cm2 s , calculated after [31], we would obtain Sc D 5:77. Generally, Sc D 0:2–3 for gas mixtures and 100–1,000 for mixtures of liquids. Our calculated value of Sc D 5:77 suggests that we are dealing with a gas phase rather than purely liquid phase. The calculation of Vmax is based on the bulk transport of the solute and therefore only realistic for macroscopic crystal length. As can be seen from (9.2), higher supersaturation would yield a higher Vmax . This fact has been proven experimentally [32]. As shown above, increasing CNH4Cl in NH3 from almost 0 to >1 mol% gives rise to an improved growth rate up to >50 m per day. Back to reality, very high flow rates are not realistic in closed systems without forced convection like stirring. Consequently, the determining parameter in the calculation is the maximum solution flow rate which should be established in a sustainable way. The effect of the other parameters of the equation is basically through their temperature dependences, and is therefore relatively small. Typical cases of flux growth does show a Vmax D 72–180 m h1 (1;700–4; 320 m per day) [33]. In the production of large size hydrothermal ’-SiO2 and ZnO crystals, a growth rate of 500–1,000 and 200–300 m per day, respectively, is
192
D. Ehrentraut and Y. Kagamitani
typically observed for the (0001) face to produce the large-size, high-quality crystals [10]. Now, further playing the parameters for the case of ammonothermal growth of 1 GaN, thus making following assumptions that D D 2104 cm2 s ; u D 10 cm s1 as obtained from simulation [34], D0:1 cm; Ce .773 K/D1 (i.e., 1 mol GaN/1 mol NH4 Cl), Sc D 5:77; .673 K/ D 0:39 g cm3 , and L D 1 cm; Vmax D 15:48 m h1 (or 371:5 m per day) for a crystal of 1 cm in length. Accordingly, a GaN seed of two inch diameter could be grown at Vmax D 164:6 m per day. 1 Uncertainties are due to D (ranging 104 6 104 cm2 s ); was approximated, and Sc was calculated on considered D. D and however roughly compensate so that Sc and also Ce play the central role in the calculation of Vmax . Main increase of Vmax will come from an improved flow rate u, which in turn reduces , which then allows to increase Ce . This can be achieved by a suitable temperature gradient T between the dissolution and growth zone as shown for the case of SC H2 O [35]. Increasing the T from 15 to 25 K almost doubled the main flow rate. We have found experimentally that a higher flow rate can easily be established in larger autoclave like the one shown in Fig. 9.4. A higher T can be established which causes the flow rate going up. The growth rate does have a high impact on the overall economy of a crystal growth process and strongly determines the price of wafers cut from a crystal. It is known that the growth of a single crystal from a solution is a much slower process
Fig. 9.4 A large size autoclave, mantled by the heater assembly, capable of growing 2-in. size GaN crystals
9
Acidic Ammonothermal Growth Technology for GaN
193
than say growth from the melt and the reasons for that have been discussed extensively in the various literatures, e.g., [36]. If, however, multiple crystals are grown during a growth cycle then the growth from a solution gets much more efficient in terms of total crystal yield. The growth rate as determined from experiments is a function of the concentration of the mineralizer and the absolute temperature in the autoclave. A growth rate exceeding 50 m per day for long-term growth cycles requires 1 mol% NH4 Cl per 1 M NH3 and higher growth rates close to 100 m per day were already achieved in very recent experiments. A stable growth rate around 80–100 m per day would already be satisfying for a future mass production of GaN by the ammonothermal route by making use of large size autoclaves for hundreds of GaN seed crystals per growth cycle [23]. It is interesting to note that the growth rates of GaN obtained for the acidic and basic ammonothermal system are very similar, i.e., the differences in the chemistry behind does not prefer one method over the other in terms of the growth rate.
9.4.3 Effect of Acidity on Formation of GaN The mineralizer NH4 X (X D Cl, Br, I) as employed show an increase in the acidity from X D Cl toward X D I, i.e., the ligands X are determining the concentration of free NH 2 in the solution. A very welcomed effect of increasing the acidity by switching to NH4 Br or even better NH4 I is the increase of the yield of formed GaN, indicated by the quantity of used up GaN precursor, as was recently demonstrated by Ehrentraut et al. [37], see Fig. 9.5. Whereas employing NH4 Cl does not substantially change the yield over increasing the reaction temperature from 360 to 550 ıC, the mineralizers NH4 Br and NH4 I do, leading to raise the yield at T 450 ıC. This might be favorable for an increased growth rate; however, much work has still to be performed to apply this phenomenon. What else can be seen in Fig. 9.5 is that phase stability to the favor of the hexagonal phase is better achievable at T 500 ı C when NH4 Br and NH4 I are used, quite in agreement to the work reported by Purdy et al. [19]. The increasing acidity on the other hand goes along with an increased quantity of self-nucleated cubic GaN, space group F 43N m, crystals as evidenced by powder X-ray scanning [37]. This is strictly true only for the growth of GaN without providing a seed crystal of hexagonal GaN. The admixture of up to 20 mol% NH4 Br or NH4 I to NH4 Cl is still gaining yields around 95% while the hexagonal phase of GaN is still preserved. Figure 9.6 shows that the (200) reflection around 2 D 40:416ı due to cubic GaN, according to the JCPDS International Centre for Diffraction Data card no. 88–2,364, is hardly visible from the powder X-ray scan on self-nucleated GaN. First experiments have evidenced that a higher growth rate of hexagonal GaN on HVPE GaN seed crystals was obtained [37].
194
D. Ehrentraut and Y. Kagamitani
Fig. 9.5 Yield of self-nucleated GaN in terms of quantity of dissolved GaN precursor and phases as evidenced by XRD. (After [37]). The h and c indicates a purely hexagonal and hexagonal with cubic phase, respectively, as revealed by XRD measurements
Fig. 9.6 Phase purity of self-nucleated GaN crystals grown from mixed mineralizers .1x/ mol% NH4 Cl=x mol%NH4 Br and .1 x/ mol% NH4 Cl=x mol%NH4 I. x D 20, 30, and 50 is denoted by blue, green, and red graphs, respectively
9
Acidic Ammonothermal Growth Technology for GaN
195
The company of oxygen in the system at high levels at >1018 cm3 causes the formation of crystalline gallium oxide (Ga2 O3 , space group R 3N c) as precipitates if unseeded growth is performed.
9.5 Properties of Ammonothermal GaN Characterization of produced GaN crystals was performed using a Nomarski differential interference microscopy (NDIM), scanning electron microscope (SEM; JEOL JSM-7000F), energy-dispersive spectroscopy (EDS; JEOL JSM-7000F, 15 kV, 1 nA, detection range 0–20 keV, accumulation time 60 s) and secondary electron mass spectroscopy (SIMS; primary beam species CsC and O2C at 5 kV and around 200– 400 nA) to identify elements and their concentrations in comparison to our standard reference material, HVPE-GaN from Mitsubishi Chemical Corporation. Powder X-ray diffractometry (XRD; Rigaku RINT-2000, 40 kV and 40 mA) was employed using CuK’1 radiation in combination with a Ge (220) channel monochromator (12 arcsec beam divergence, scan speed 0:01ı min1 , step width 104ı ) to evaluate the crystallinity. The internal standard reference material for XRD was silicon (JCPDS-International Centre for Diffraction Data card no. 27–1,402). Steadystate photoluminescence (PL) was performed using a He–Cd laser (Omnichrome 3056-M-A01: D 325 nm; Pout D 10 mW) for excitation and a cryostat (Daikin UV202CL) for temperature control. Spectra were obtained by a photomultiplier using a monochromator (Jobin Yvon Spex HR320: 1,200 lines/mm gratings, 320 mm focal length). The first crack-free ammonothermal GaN crystal has been grown on a 1 cm2 large (0001) HVPE-grown GaN seed crystal with a mechanically untouched surface around the year 2005 as shown in Fig. 9.7. The grown GaN crystal in Fig. 9.7 after the experiment was completely covered with a film mainly containing the mineralizer. This film could have been removed easily upon treatment with distilled water and 2-propanol. The quality of the surface of a seed crystal has a tremendous impact on the nucleation and successive growth of the GaN [21,37]. A mechanically untouched surface would certainly provide the best surface condition with no damage. However, proper CMP or purely chemical etching if one gets very smooth surfaces without macrosteps has shown to deliver a satisfying quality [21]. A similar observation has been made in the growth on m-plane seed crystals as we will discuss later. Figure 9.8 provides some images of a c-plane ammonothermal crystal grown on a HVPE GaN seed crystal. Here, the result was obtained from remodeled heater system which allowed a better control of the overall thermal conditions of the autoclave. Although the surface of the crystal has not been polished, just as-grown, a highly transparent ammonothermal crystal was obtained. The thickness was around 15 m for both faces, Ga and N-polar. The XRC FWHM from the (0002) reflection in Fig. 9.9 is very homogeneous with values around 108 and 339 arcsec for the Ga-polar and N-polar face, respectively. Such substrates can certainly be
196
D. Ehrentraut and Y. Kagamitani
Fig. 9.7 Image of the first thick GaN crystal grown under acidic ammonothermal conditions on a mechanically untreated HVPE GaN crystal
Fig. 9.8 Optical micrographs of high-quality ammonothermal GaN grown on HVPE GaN seed shown in .a/ reflection, .b/ transmission, .c/ NDIM from Ga-polar face, and .d / NDIM from N-polar face. Reprinted with permission. Copyright (2009), Springer-Verlag
used for successive deposition of simple device structures. This work is currently progressing. It has recently been taken into consideration [21] that cobble patterning of an as-grown (0001) surface might be related to formed dislocations, quite in analogy to quartz. It can be seen from Fig. 9.8c, d that the density of this hillocks is N than for the (0001) face. Hashimoto et al. about three times larger for the .0001/
9
Acidic Ammonothermal Growth Technology for GaN
197
Ga polar face
Intensity [a.u.]
1000
100
10
1 17.15
17.20 17.25 17.30 17.35
17.40 17.45
N polar face
Intensity [a.u.]
10000 1000 100 10 1 16.0
16.5
17.0
17.5
18.0
18.5
Fig. 9.9 XRC from the (a) Ga and (b) N-polar face of an ammonothermally grown GaN crystal
[38] have confirmed by transmission electron microscopy (TEM) that the threading dislocation density is higher for the N-polar face, i.e., about 106 and 107 cm2 for the Ga and N-polar face, respectively. The cross-sectional SEM in Fig. 9.10a shows about 100 m GaN had been grown on both polar faces, and pyramidal facets already had commenced to form. Zooming into the area of such formed interface between the HVPE seed and ammonothermal GaN crystal (Fig. 9.10b) from a typical short-term growth reveals a high degree of flatness and the cracking pattern just continue from the HVPE into the ammonothermal material. Voids have not been found. Figure 9.10c features the interface region N HVPE seed. The of a grown ammonothermal GaN crystal on the nonpolar .1010/ growth of GaN structures on nonpolar substrates draw a lot of attention recently since it has been demonstrated that improved internal quantum efficiency can be achieved because of the lack of an electrostatic field [39].
198
D. Ehrentraut and Y. Kagamitani
Fig. 9.10 Cross-sectional SEM (a) showing the Ga-polar and N-polar face, both about 100 m thick, including commencing formation of pyramidal facets and featuring the flat interface region between (b) the (0001) HVPE seed and the ammonothermal crystal and (c) the .101N 0/ HVPE seed and the ammonothermal crystal
Fig. 9.11 Optical micrographs of high-quality ammonothermal GaN grown on m-plane HVPE GaN seed shown in (a) reflection, (b) transmission, (c) NDIM at higher magnification displaying growth features because of mechanical damaging and (d) NDIM revealing a very flat surface. Reprinted with permission. Copyright (2009), Springer-Verlag
N We have used .1010/, or m-plane, HVPE seed crystals which are available in 2 sizes of up to 1 cm and excellent quality with the (0002) XRC FWHM around 100 arcsec and a dislocation density of 2 105 cm2 [40]. Figure 9.11 demonstrates the result from the ammonothermal growth. A very smooth surface morphology is obtained. It must however be noted here that damage from slicing is still sometimes visible under NDIM observation (Fig. 9.11c). We have recently demonstrated the acidic ammonothermal growth on a 1in. seed crystal [26] and have shown in Fig. 9.1 the first result of the acidic ammonothermal growth of GaN on a 2-in. large HVPE (0001) GaN seed under similar conditions as above, i.e., temperature and pressure around 500 ıC and 100 MPa, respectively. The curvature radius of the seed crystal was around 9 m and the miscut orientation ˙0:15ı . The autoclave used for the process measures about 100 1; 000 mm2 in inner diameter inner length, presently the largest one in use for theammonothermal growth of GaN from acidic mineralizer, see Fig. 9.4a. Although the crystal quality is still not satisfactory, in spite of the few attempts
9
Acidic Ammonothermal Growth Technology for GaN
199
however this result is very promising in so far that we have shown the practical feasibility of growing large size GaN crystals under moderate process temperature and pressure conditions. This is an important issue for future industrial production. It should be noted here that strain-free (no bending) and large size, i.e., 2 in. GaN seed crystals are yet not available. Only Dwilinski et al. have recently shown the growth of large size, extremely flat and low-strain GaN crystals by using their own GaN crystals as seeding material, see Chap. 7. Impurities in a crystal are a major concern and the sources for them are quite manifold when autoclave technology is employed: the autoclaves, gas pipes, baffle and seed crystal holder in addition to precursor, mineralizer and the ammonia itself are contributing sources. In summary of the many experiments we have carried out thus far, we note levels of impurities such as Cr, Fe, and Ni of the order around 1015 –1018 cm3 ; 1017 –1020 cm3 , and 1017 –1020 cm3 , respectively. Platinum from the inner liner contributed with maximum 1016 cm3 . Silicon (Si) and oxygen (O) ar detected at levels around 1019 cm3 and 1018 –1020 cm3 , respectively. The values for Si and O are somewhat comparable to ammonothermal GaN fabricated from basic mineralizers where around 1019 cm3 for both elements [41] and around 1018 cm3 Si and 1020 cm3 O were reported [23, 41]. In all cases, it was found that the N-polar face contains higher amounts of impurities, quite in agreement with ZnO where the O-polar face easily traps impurities [10]. Also, the lower levels are the result of advanced growth technology. PL measurements were made to evaluate the optical quality of some of the ammonothermal GaN samples [24, 42, 43]. A general observation made by PL is that the GaN nucleated on the N-polar face seems to be of higher quality than the GaN nucleated on the Ga-polar face. Moreover, we have recently reported on the strong improvement of the optical quality of acidic ammonothermal GaN and shown that an ammonothermal GaN samples is similar to standard HVPE-grown GaN wafer [24]. In Fig. 9.12, the PL spectrum obtained at 10 K from ammonothermal GaN is compared with that from HVPE-grown GaN. Emission from the exciton bound to neutral donor .D0 X/ line of the hexagonal structure at 3.472 eV was observed from the HVPE GaN and at lower intensity also from the ammonothermal GaN. However, an additional peak at 3.357 eV was sometimes found which was assigned to the Y4 line [44]. Temperature dependence of this line however shows characteristics similar to the D0 X emission from 2H -GaN and to the reported exciton bound to point defects trapped at threading dislocations. It was therefore suggested that the Y4 line might be related to another crystal structure when comparing the intensity distribution of the (0006) ! 2 XRD satellite peak and the Y4 emission. From bandgap energy consideration, the Y4 line could be linked to the D0 X emission from the 6H structure of GaN which is a metastable phase [43]. A deep level luminescence peaking at 1.93 eV, red luminescence, not shown in the figure, was observed from the ammonothermal GaN crystal at 10 K [42]. The characteristics of the temperature dependence of the PL intensity and TR-PL suggest that the high oxygen concentration of the ammonothermal GaN may be the origin of
200
D. Ehrentraut and Y. Kagamitani
Fig. 9.12 Photoluminescence at 10 K from the Ga and N-polar face from a high-quality HVPE and an ammonothermal crystal
this observed red luminescence, quite in good agreement to the high oxygen levels in other GaN samples.
9.6 Prospects and Future developments for Ammonothermal GaN For the first time, a high-quality, free-standing 1 in. GaN wafer had been shown in 2007 by the Ammono Company [41]. Since then, the ammonothermal growth of GaN is being considered a technique competitive to the other, a more established vapor phase technique to grow GaN crystals. Fukuda & Ehrentraut have recently compared the developments in the ammonothermal technology to precedent developments in ZnO and SiO2 [26]. Supposing a comparable technological time route for scaling-up of the autoclave, GaN seed crystal size, and optimization of the process regime including a better understanding of the rate-determining process, one could expect that commercially available wafers of ammonothermal GaN of 2 in. in size may soon be available. The technological developments in the hydrothermal growth of ZnO were purely based on the achievements from hydrothermal growth of SiO2 . In contrast, the ammonothermal growth of GaN is very much profiting from both SiO2 and ZnO
9
Acidic Ammonothermal Growth Technology for GaN
201
and problems such as the design of a large size autoclave including choice of the alloy, establishing of the thermal gradient or the process control are much better understood. Furthermore, numerical simulation of mass and heat transport under ammonothermal ambient to optimize growth conditions has seen quite some progress recently [45]. We therefore may expect the growth of large quantities of large size .2 in:/ GaN by the ammonothermal technique within the near future [26]. By comparison of the growth rates of some semiconductor crystals with those of SiO2 and ZnO, it gets clear that the melt growth is clearly dominant in terms of growth rates, e.g., up to 60 mm h1 for Si. Hydrothermal ZnO, however, grows with about 10 m h1 only. The target number for GaN would be about half that of ZnO; however, only 2 m h1 is currently achieved. The presently low growth rate, roughly 10% of that of quartz, seems to impede from the application of ammonothermal growth technique to produce GaN in an economic fashion. However, several factors need to be taken into account: (1) Firstly, scalability is a main advantage strongly speaking for the ammonothermal rather than for any of the other above mentioned technologies to produce GaN crystal. Analogous to SiO2 , i.e., using 1,000 seeds to grow 1,000 crystals from one autoclave at the same time would drastically bring down the price per one GaN wafer. Also, the size of the crystal to be grown is basically limited mostly by the size of the autoclave inner diameter. Taking all the mentioned points into account, we have recently [26] estimated that the price for one ammonothermal GaN wafer is comparable to that of InP, despite the low growth rate and high-pressure equipment. (2) Secondly, the growth rate of SiO2 in the early work of Spezia [46] was estimated to be 1 m h1 [47]. However, the growth rate of SiO2 has been improved by a factor of about 20 through better understanding of the rate determining parameters along with better knowledge of the chemistry in the hydrothermal solution. Right now we are facing a similar situation in the ammonothermal growth of GaN. (3) Thirdly, the period that ammonothermal growth technique is seriously under development is pretty short, about 15 years [13]. Moreover, only a very few groups have been working on the subject. This is in contrast to the melt growth of now dominating semiconductors like Si, Ge and to a lesser amount GaAs, which is the object for scientific investigation for up to half a century already by an incomparably larger community of researchers as well as monetary funding. The major tasks for the successful ammonothermal growth of GaN have recently been taken into account by Fukuda & Ehrentraut [26]: Profound knowledge of the solubility of the GaN feedstock and the mineralizers; availability of large-size, highquality GaN seeds; stable feeding with the GaN feedstock to achieve long-term growth; scaling-up of the size of present autoclaves for 2 in: crystals. N N N and Next, semi- and nonpolar oriented GaN wafers, i.e., f1011g; f1122g; f1010g, N f1120g will be fabricated from large-size GaN crystals, which will allow the fabrication of device structures free of electrostatic fields, therefore yielding an improved quantum efficiency [39, 48]. The dislocation density of GaN wafers must be decreased to 103 cm2 which is attainable as demonstrated for the case of LPE-grown GaN [49] and more recently also for the basic ammonothermal growth technique of GaN [41]. Large GaN seed
202
D. Ehrentraut and Y. Kagamitani
crystals of highly crystalline perfection, i.e., unstrained with no bending would then become available. If we look about 5–10 years ahead from now, we believe that a countable fraction of the GaN wafers for devices is likely to be produced from GaN crystals grown by the ammonothermal growth technique [26].
References 1. T. Koyama, S.F. Chichibu, J. Appl. Phys. 95, 7856 (2004) 2. Mitsubishi Chemical Corporation, Tokyo, Japan http://www.m-kuyaku.co.jp 3. K. Motoki, T. Okahisa, N. Matsumoto, M. Matsushima, H. Kimura, H. Kasai, K. Takemoto, K. Uematsu, T. Hirano, M. Nakayama, S. Nakahata, M. Ueno, D. Hara, Y. Kumagai, A. Koukitu, H. Seki, Jpn. J. Appl. Phys. 40, L140 (2001) 4. T. Sasaki, T. Matsuoka, J. Appl. Phys. 64, 4531 (1988) 5. V. Härle, B. Hahn, H.-J. Lugauer, S. Bader, G. Brüderl, J. Baur, D. Eisert, U. Strauss, U. Zehnder, A. Lell, N. Hiller, Phys. Stat. Sol. A 180, 5 (2000) 6. Y. Honda, Y. Kawaguchi, Y. Ohtake, S. Tanaka, M. Yamaguchi, N. Sawaki, J. Cryst. Growth 230, 346 (2001) 7. P. Waltereit, O. Brandt, M. Ramsteiner, R. Uecker, P. Reiche, K.H. Ploog, J. Cryst. Growth 218, 143 (2000) 8. Z. Sitar, M.J. Paisley, B. Yan, R.F. Davis, Mater. Res. Soc. Symp. Proc. 162, 537 (1990) 9. E.S. Hellman, D.N.E. Buchanan, D. Wiesmann, I. Brener, MRS Internet J. Nitride Semicond. Res. 1, 16 (1996) 10. D. Ehrentraut, H. Sato, Y. Kagamitani, H. Sato, A. Yoshikawa, T. Fukuda, Prog. Cryst. Growth Charact. Mater. 52, 280 (2006) 11. H. Jacobs, D. Schmidt, High-pressure ammonolysis in solid-state chemistry, In Current Topics in Materials Science, vol. 8, ed. by E. Kaldis (North-Holland, Amsterdam, 1982), p. 379 12. D. Peters, J. Cryst. Growth 104, 411 (1990) 13. R. Dwili´nski, A. Wysmolek, J. Baranowski, M. Kami´nska, R. Doradzi´nski, J. Garczy´nski, L.P. Sierzputowski, Acta Phys. Pol. 88, 833 (1995) 14. J.W. Kolis, S. Wilcenski, R.A. Laudise, Mater. Res. Soc. Symp. Proc. 495, 367 (1998) 15. A.P. Purdy, Chem. Mater. 11, 1648 (1999) 16. D.R. Ketchum, J.W. Kolis, J. Cryst. Growth 222, 431 (2001) 17. M.J. Callahan, B. Wang, L.O. Bouthillette, S.-Q. Wang, J.W. Kolis, D.F. Bliss, Mater. Res. Soc. Symp. Proc. 798, Y2.10.1 (2004) 18. B. Wang, M.J. Callahan, K.D. Rakes, L.O. Bouthillette, S.-Q. Wang, D.F. Bliss, J.W. Kolis, J. Cryst. Growth 287, 376 (2006) 19. A.P. Purdy, R.J. Jouet, C.F. George, Cryst. Growth Des. 2, 141 (2002) 20. A. Yoshikawa, E. Ohshima, T. Fukuda, H. Tsuji, K. Oshima, J. Cryst. Growth 260, 67 (2004) 21. Y. Kagamitani, D. Ehrentraut, A. Yoshikawa, N. Hoshino, T. Fukuda, S. Kawabata, K. Inaba, Jpn. J. Appl. Phys. 45, 4018 (2006) 22. D. Ehrentraut, K. Kagamitani, T. Fukuda, F. Orito, S. Kawabata, K. Katano, S. Terada, J. Cryst. Growth 310, 3902 (2008) 23. B. Wang, M.J. Callahan, Cryst. Growth Des. 6, 1227 (2006) 24. D. Ehrentraut, K. Kagamitani, C. Yokoyama, T. Fukuda, J. Cryst. Growth 310, 891 (2008) 25. D. Nicholls, Inorganic Chemistry in Liquid Ammonia, in Topics in Inorganic and General Chemistry, Monograph 17, ed. by R.J.H. Clark (Elsevier, Amsterdam, 1979) 26. T. Fukuda, D. Ehrentraut, J. Cryst. Growth 305, 304 (2007) 27. M. Roos, J. Wittrock, G. Meyer, S. Fritz, J. Strähle, Z. Anorg. Allg. Chem. 626, 1179 (2000) 28. H. Yamane, Y. Mikawa, C. Yokoyama, Acta Crystallogr. E 63, i59 (2007) 29. W. Nernst, Z. Physik. Chem. 47, 52 (1904)
9
Acidic Ammonothermal Growth Technology for GaN
203
30. A. Carlson, in Growth and Perfection of Crystals, ed. by R.H. Doremus, B.W. Roberst, E. Turnbull, (Wiley, New York, 1958), p. 421 31. C.-H. He, Y.-S. Yu, Ind. Eng. Chem. Res. 37, 3793 (1998) 32. D. Ehrentraut, Y. Kagamitani, Ammonothermal growth of GaN, IWBNS-V, Itaparica, Salvador, Brazil, September 24–28, 2007, Workshop Program and Abstracts 33. H.J. Scheel, D. Elwell, J. Cryst. Growth 12, 153 (1972) 34. Y. Masuda, Private communication 35. K. Nagai, J. Asahara, J. Jpn. Assoc. Cryst. Growth 27, 68 (2000) 36. A.A. Chernov, in Modern Crystallography III: Crystal Growth, Springer Series in Solid-State Sciences, vol. 36 (Springer, Berlin, Heidelberg, New York, Tokyo, 1984) 37. D. Ehrentraut, N. Hoshino, Y. Kagamitani, A. Yoshikawa, T. Fukuda, H. Itoh, S. Kawabata, J. Mater. Chem. 17, 886 (2007) 38. T. Hashimoto, F. Wu, J.S. Speck, S. Nakamura, Jpn. J. Appl. Phys. 46, L525 (2007) 39. P. Waltereit, O. Brandt, A. Trampert, H.T. Grahn, J. Menniger, M. Ramsteiner, M. Reiche, K.H. Ploog, Nature 406, 865 (2000) 40. F. Orito, K. Katano, S. Kawabata, Y. Kagamitani, D. Ehrentraut, C. Yokoyama, H. Yamane, T. Fukuda, Growth strategy for gallium nitride bulk crystals, 5th International Workshop on Bulk Nitride Semiconductors (IWBNS-V), Sept. 24–28, (2007), Itaparica, Bahia, Brazil, Workshop Program and Abstracts 41. R. Dwili´nski, R. Doradzi´nski, J. Garczy´nski, L.P. Sierzputowski, A. Puchalski, Y. Kanbara, K. Yagi, H. Minakuchi, H. Hayashi, J. Cryst. Growth 310, 3911 (2008) 42. K. Fujii, G. Fujimoto, T. Goto, T. Yao, Y. Kagamitani, N. Hoshino, D. Ehrentraut, T. Fukuda, Phys. Stat. Sol. (A) 204, 3509 (2007) 43. K. Fujii, G. Fujimoto, T. Goto, T. Yao, Y. Kagamitani, N. Hoshino, D. Ehrentraut, and T. Fukuda, Phys. Stat. Sol. (A) 204, 4266 (2007) 44. M.A. Reshchikov, H. Morkoc, J. Appl. Phys. 97, 061301 (2005) 45. Q.-S. Chen, V. Prasad, W.R. Hy, J. Cryst. Growth 258, 181 (2003) 46. G. Spezia, Atti. Accad. Sci. Torino 35, 95 (1900) 47. F. Iwasaki, H. Iwasaki, J. Cryst. Growth 237–239, 820 (2002) 48. A. Chakraborty, T.J. Baker, B.A. Haskell, F. Wu, J.S. Speck, S.P. DenBaars, S. Nakamura, U.K. Mishra, Jpn. J. Appl. Phys. 30, L945 (2005) 49. Y. Mori, T. Sasaki, Oyo Buturi 75, 529 (2006), in Japanese
Part IV
Flux Growth Technology
Chapter 10
High Pressure Solution Growth of Gallium Nitride Michal Bo´ckowski, Pawel Strak, ˛ Izabella Grzegory, and Sylwester Porowski
Abstract State-of-the-art high pressure solution (HPS) growth of gallium nitride is presented in this chapter. The spontaneous crystallization and seeded growth are described in detail from the technological point of view. Two classes of pressuregrown crystals and their use as substrates for blue laser diodes (LDs) are demonstrated.
10.1 Introduction The melting temperature of gallium nitride (GaN) is extremely high 2; 493 K- and was determined with very good precision by Utsumi et al. [1] in 2003. The nitrogen pressure needed for congruent melting of GaN is 6 GPa. Before 2003, people had known that the melting parameters were high from theoretical approximation [2]. Since it was impossible to crystallize GaN from the melt, other crystal growth methods were developed. One of them was crystallization from the high temperature liquid solution of gallium under high nitrogen pressure. According to the definition given by Elwell and Scheel [3], in the high temperature solution growth the constituents of the material to be crystallized are dissolved in a suitable solvent and crystallization occurs as the solution becomes critically supersaturated. The supersaturation can mainly be promoted by cooling the solution or by a transport process in which the solute flows from a hotter to a cooler region [3]. Obviously, the main advantage of the solution growth is that the growth temperature is lower than that required for growth from the melt. This advantage often results in a better crystal quality with respect to point defects, dislocation density, and low angle grain boundaries, compared to crystals grown directly from their melts [3]. The main disadvantages of the high temperature solution growth method are associated with microscopic and macroscopic inclusions of solvent or impurities in the crystals, nonuniform doping, and a slow growth rate [3]. The high temperature solution growth of gallium nitride proceeds at temperature of the order of 1,800 K and at nitrogen pressure up to 2 GPa. The molecular nitrogen being under pressure is dissociated on the hot liquid gallium surface and the atomic nitrogen is dissolved in 207
208
M. Bo´ckowski et al.
the liquid metal. Then, the convective flow of gallium, due to applied temperature gradients, is exploited in order to provide atomic nitrogen to the supersaturated zone of the solution, where the GaN crystallization takes place. Since the high temperature solution growth of GaN requires high nitrogen pressure, gallium nitride has always been in the field of interest of the Institute of High Pressure Physics Polish Academy of Sciences (formerly High Pressure Research Center PAS). In 1984 Karpinski et al. [4] determined the equilibrium nitrogen pressure–temperature curve for gallium nitride in the temperature range up to 2,000 K. At the beginning of the nineties, on the wave of nitrides popularity, the thermodynamic background of the system Ga–GaN–N2 and thermodynamic properties of GaN have been determined and formulated [5]. By the end of 1995 small GaN platelets (up to 2 mm) had been obtained. In 1996, 1 cm2 and 100 m thick GaN crystal (platelet) of perfect structural quality and low dislocation density .102 cm2 / was obtained for the first time. Based on this success, relatively fast development of the Crystal Growth Laboratory was observed. The quantity of high pressure (HP) reactors (Institute’s home-made apparatus) increased several times during a few years. The interior diameter, and thus the working volume of the HP chamber increased too. A small pilot production of GaN crystalline platelets was started thanks to the Polish Government Project [6]. In parallel, the epitaxial techniques (MOCVD and MBE) and laser processing technology were created and implanted to the Institute. In 2002, the first blue laser diode (LD) was built on the pressure grown GaN crystals by the MOCVD method [7]. Three years later, GaN-based blue LDs were obtained by plasma-assisted MBE [8]. The strategic Polish Government Program “Blue Optoelectronics” [9] helped develop the nitrides technology in the Institute of High Pressure Physics and created a spin-off laser company called TopGaN Ltd. The technology of the blue lasers diodes made by TopGaN has just been based on homoepitaxial growth on pressure-grown gallium nitride crystals. The idea of the homoepitaxy has always been the base for the nitrides technology in the Institute of High Pressure Physics. Unfortunately, the spontaneous crystallization by the HP technique has not allowed growth of bigger than 3 cm2 GaN crystalline platelets. Therefore, hydride vapor phase epitaxy (HVPE) was introduced in the Institute. The new idea has been to combine the HVPE with the HP technique in order to create 2 in., low dislocation density crystals with good structural quality and appropriate physical properties to be the substrates. In this chapter, this idea is partially described. Partially, since the use of HVPE-GaN crystals as seeds for the HP crystallization is just shown. The use of HP-GaN crystals as seeds for the HVPE technique is presented in detail in Chap. 3 of this book. The chapter starts with the brief description of the high pressure solution (HPS) growth method and its thermodynamic and kinetic aspects. Then, the HP experimental setup and the spontaneous crystallization are described. Physical properties of the pressure grown crystals (platelets and needles) are presented. Next part of the chapter is focused on a seeded crystallization under pressure. The results of the liquid phase epitaxy (LPE) technique are shown. The relatively new approach of seeded growth, with convective flow of gallium under control, is described in details. At the
10
High Pressure Solution Growth of Gallium Nitride
209
end of this chapter the application of the pressure grown GaN crystals as substrates for the LDs is briefly reported.
10.2 Growth Method The HPS method is temperature gradient growth method based on direct reaction between gallium and nitrogen at high temperature and high nitrogen pressure. Figure 10.1 presents schematic illustration of the crystallization by the HPS method. Nitrogen molecules dissociate on gallium surface and dissolve in the metal. Therefore, the crystals are grown from the solution of atomic nitrogen in liquid gallium. Supersaturation, the driving force for the crystallization process, is created by the application of temperature gradient along the liquid gallium. Thus the HPS method consists of three stages: (1) dissociative adsorption of nitrogen on liquid metal surface, (2) dissolution and transport of nitrogen atoms from the hot end of the solution to the cooler one, and (3) crystallization process. Due to the temperature gradient in the system, the atomic nitrogen is transported from the hot end of the solution to the cooler part. It has been shown that the convection mechanism plays the dominant role in the transport phenomenon [10]. The nitrogen in the solution is uniformly distributed due to the strong convection flow. Thus, the solution at the cold zone must be supersaturated. It means that the excess nitrogen concentration related to the equilibrium value for a given temperature exists at the cold zone of the liquid metal. Hence, GaN crystal growth takes place at this zone.
crucible N2 molecules under pressure Dissociative adsorption of N2 on liquid metal surface Dissolution and transport to the crystallization zone
ΔT
Ga GaN
Spontaneous crystallization
T
Fig. 10.1 Schematic illustration of the crystallization process in a temperature gradient by the high pressure solution growth method; three basic stages of the method are marked
210
M. Bo´ckowski et al.
10.2.1 Thermodynamic and Kinetic Aspects of HPS Growth As mentioned in Sect. 10.1, the equilibrium pressures of N2 for the Ga–GaN system have been determined by Karpinski et al. [4]. The direct synthesis and decomposition experiments performed by gas pressure technique for nitrogen pressures up to 2 GPa were practiced. The curve following from these data is shown in Fig. 10.2a. Figure 10.2b shows, in turn, the N solubility data resulting from the annealing of Ga in N2 atmosphere at the three phase Ga–GaN–N2 equilibrium conditions for the temperature range from 1,273 to 1,973 K. The solid line has been calculated in the ideal-solution approximation. The nitrogen concentrations are not too high; they are below 1 at% [5]. This curve allows us to determine the maximum of the relative supersaturation reached in the solution. If the convective transport in the solution is faster than the crystallization, the relative supersaturation can be expressed as: x.T2 / x.T1 / ; x.T1 /
D
(10.1)
where represents the relative supersaturation and x.T2 / x.T1 / is the difference between equilibrium nitrogen concentration at hot and cold zone of the solution. The determination of is also schematically presented in Fig. 10.2b. The analysis of thermodynamic properties of GaN and the system of its constituents explains the role of the use of HP [5]. It is a factor increasing thermodynamical potential of GaN constituents. It makes the GaN crystal stable at higher temperatures necessary for crystallization. The pressure, however, is also important for kinetics of the GaN synthesis. As mentioned, one of the stages of GaN synthesis from its constituents is dissolution of nitrogen in hot liquid gallium. This process was analyzed by quantum mechanical calculations [11]. The nitrogen molecule approaching the metal surface is repelled by the metal which results in a potential
a
temperature, K 2500 2250 2000 1750
1500
1250
b 1900 1 45 GPa
10000 pressure limit of 2 GPa
1000
GaN Ga+1/2N2
100 0.0004
temperature, K
pressure, bar
100000
1800 T2 1700 T1 1600 1500 1400
0.0005
0.0006 1/T, K–1
0.0007
0.0008
1 21 GPa
1 01 GPa 0 81 GPa 0 73 GPa 0 6 GPa
0 3 GPa
x(T2)-x(T1) x(T1)
x(T2)
0.001 0.002 0.003 0.004 0.005 XN, at.fr.
Fig. 10.2 (a) p T equilibrium curve for GaN determined by Karpinski et al. [4]; (b) liquidus curve for Ga–GaN system [5]; determination of the relative supersaturation in the solution
10
High Pressure Solution Growth of Gallium Nitride
211
barrier. If the molecule has enough energy to overcome the potential barrier, it comes closer to the gallium surface and dissociates into two atoms, forming new bonds with the metal. The potential barrier is lower than the bonding energy in the nitrogen molecule. However, its value of 3.5 eV seems to be quite high. Therefore, the density of the interacting gas (thus its pressure) is of crucial importance. Here, it should be pointed out that for oxygen interacting with gallium, there is no potential barrier [12]. One can always expect the oxygen atoms in liquid gallium. This fact is very important for the properties of GaN crystals grown by HPS method, what will be shown later. The high potential barrier for nitrogen dissociation on the liquid gallium surface also suggests that this dissociation process is kinetically controlled even for relatively high temperatures. Determining the rate of dissociation at 2 GPa, Krukowski et al. [11] showed that for gallium nitride the rate of dissociation is not fast. For effective synthesis (10 mg cm2 during 100 h), the temperature should be higher than 1,500 K. Comparing this last result with the equilibrium p T curve for GaN (see Fig. 10.2a), it may be concluded that only application of the high nitrogen pressure allows the synthesis of gallium nitride from its constituents.
10.2.2 Experimental One of the typical experimental systems for crystal growth of GaN at nitrogen pressure is shown in Fig. 10.3. This system, consisting of two HP chambers shown in the photograph, is connected to the central pressure line where 1 GPa of nitrogen can be obtained by a set of membrane compressors and intensifiers. Vertically positioned, technological gas pressure chambers (reactors) of internal diameters of 4, 6, or 10 cm, allowing crucibles with the working volume of 10, 25, and 150 cm3 respectively, are connected by capillaries to the special gas chambers with compressors, which serve as reservoirs of nitrogen and allow stabilizing pressure in the chambers during experiments. The multizone cylindrical graphite furnaces, capable of reaching temperatures up to 2,000 K, are placed inside the gas pressure reactors. The HP chambers are equipped with additional systems necessary for annealing in vacuum, cooling of the reactor, electronic stabilization and programming of pressure and temperature. To register the temperature during the crystal growth experiments, PtRh6%–PtRh30% thermocouples are used. They are arranged along the furnace and coupled with the input power control electronic systems. The pressure is measured by manganine gauges positioned in the low temperature zone of the HP systems. The pressure and temperature are stabilized with an accuracy of ˙1 MPa and ˙0:1 K, respectively. The typical crystal growth experiment is that the metal (Ga) placed in the graphite crucible is heated in the furnace inside the HP chamber with a constant rate to given axial and/or radial temperature profiles. Then, the system is annealed at these conditions under high nitrogen pressure for 100–500 h. After that, the furnace is cooled down at a constant rate, the system is decompressed, and the metal with
212
M. Bo´ckowski et al. central pressure line 1 GPa to the next high pressure system
the set of membrane compressors and intensifiers
to the next high pressure system
2m
Fig. 10.3 One of the experimental system for crystal growth of GaN at N2 pressure; two visible high-pressure chambers are connected to the central pressure line. The photograph at the bottom left corner of this figure shows a furnace being introduced into the reactor
GaN crystals inside is removed. The charge from the crucible is etched with boiling HNO3 =HCl acid solution to extract the crystals. During the crystallization process, the pressure and the temperature in the whole gallium sample correspond to the nitride stability range. Thus, the highest measured temperature is lower than the equilibrium temperature for a given nitrogen pressure at three phase Ga–N2 –GaN equilibrium conditions. The gallium nitride crystals grow spontaneously from the wall of the crucible. They are perpendicular to the wall surfaces and distributed randomly in the supersaturated cold zone of the solution. The size of this zone depends on the applied axial temperature profile and temperature distribution in the liquid gallium. Two types of the axial temperature gradient configurations are used: positive (direct) and negative (reversed). They are schematically shown in Fig. 10.4a, b. The temperature distribution is evaluated by three, four, or five thermocouples positioned in the crucible wall. Three types of the axial temperature profiles – linear, convex, and concave are observed. Figure 10.4c represents the radial temperature distribution on the gallium top and crucible bottom. The radial gradient seems to be very important since it drives the convective flow in the liquid gallium. A few experiments with thermocouples positioned close to the center of the gallium top surface were carried out. These experiments showed that the radial temperature distribution on the gallium top was flat, as is shown in Fig. 10.4c. At the bottom of the crucible the temperature has always been measured at the sides and in the center. Three radial temperature profiles represented in Fig. 10.4c by curves 1, 2, and 3 have been detected.
10
High Pressure Solution Growth of Gallium Nitride
a
linear
positive (direct)
213
convex
z
concave
z
z
Ga
T
b
T
linear
T
convex
negative (reversed) z
concave
z
z
Ga
T
T
T
c radial temperature distributions on the crucible′s top and bottom
T
Ga
1 2 3 r Fig. 10.4 (a) Positive (direct) temperature gradient configuration with three possible temperature profiles; (b) negative (reversed) temperature gradient configuration with three possible temperature profiles; (c) radial temperature distributions on the crucible’s top and bottom; at the bottom of the crucible three temperature profiles can be obtained; filled circle denotes the positions of thermocouples
For seeded crystallization, a seed crystal is immersed in the gallium and back melted by the metal [13]. Then, the procedures are similar to that described above. The details of seeded and spontaneous growth configurations and procedures are presented separately in the corresponding subsections of this chapter.
214
M. Bo´ckowski et al.
10.3 Spontaneous Crystallization by HPS Growth Method The crystallization without seeds in the solution is called here the spontaneous crystallization. The dominating morphological form of GaN crystals grown by the spontaneous method is a hexagonal platelet. In order to crystallize big (1 cm2 or more) platelets the positive temperature gradient configuration with the convex temperature profile is usually applied. The maximum temperature of the solution is about 1,750 K at nitrogen pressure of 1 GPa. Typical duration of a crystal growth run is 100 h. A difference between the temperatures measured at the hottest and the coldest places in the crucible (a supercooling) is always lower than 50 K. The average temperature gradient along the z-axis of the crucible attains 10 K/cm. The typical relative supersaturation determined from (10.1) varies from 5% to 20%. The radial temperature distribution is represented by the curve 3 from Fig. 10.4c, i.e., the temperature at the center of the crucible is lower (several degrees) than those measured at the sides. In this configuration the crystals are grown at the bottom of the crucible and on the crucible wall close to the bottom. Up to five 1 cm2 platelets can be obtained during one crystallization run in the reactor of internal diameter of 6 cm. Other crystals are smaller. An increase in the crucible’s volume - thus crystallization in bigger reactors - does not increase the size of the platelets. Only number of the crystals grows up. The linear or concave temperature distributions allow increasing the supersaturated zone in the solution and then the crystals can grow on the crucible wall, relatively far from the bottom. It does not mean though that the crystals are bigger. They are just distributed in the crucible in a different way. A significant increase in the temperature gradient (to 50 K/cm) does not lead to a significant acceleration of the growth rate but to a change of the growth mode. Then the needles become the dominating morphological form of GaN crystals grown by the spontaneous method.
10.3.1 Habit and Morphology of the Crystals Figure 10.5a presents a scheme of an ideal GaN platelet. Its large hexagonal surfaces correspond to {0001} polar crystallographic planes of the wurzite structure. ˚ ˚ The N planes and semipolar 1011 N and side faces of the crystals are nonpolar 1010 ˚ N planes. For the real GaN platelet (see Fig. 10.5b) these last planes are not 1012 well developed. The crystals in the form of hexagonal platelets grow slowly at a rate N lower than 0.1 mm/h into directions (perpendicular to the c-axis). They are single crystals, slightly gray or transparent, very often with flat mirror like faces. The maximum lateral size of the platelets is 3 cm2 whereas the thickness is about 150 m. The tendency for unstable growth is stronger for one of the polar {0001} faces of the platelets. The morphological features like macrosteps, periodic inclusions of solvent, or cellular growth structures are observed on this side. The opposite surface is mirror-like and often atomically flat. For crystals grown without an intentional doping, the unstable surface always corresponds to the Ga-polar
10
High Pressure Solution Growth of Gallium Nitride
215
Fig. 10.5 Two habits of GaN crystals grown by the HPS method (a) ideal GaN wurzite platelet; (b) real GaN pressure grown platelet, N-side as grown; grid 1 mm; (c) ideal GaN wurzite needle; (d) real GaN pressure grown needle
(0001) face of GaN. The polarity of the crystal surfaces has been identified by etching in hot alkali aqua solutions since the Ga-polar surface is inert to etching whereas the N-polar one etches very well. The method was calibrated by convergent beam electron diffraction (CBED) and X-ray photoelectron spectroscopy (XPS) measurements [14–16]. Figure 10.5c represents the second possible form of the pressure grown crystals – a needle. The real crystal is presented in Fig. 10.5d. The needles are elongated into c-direction, with diameter of about 1 mm and up to 10 mm in length. It is clearly seen in Fig. 10.5d that only some facets of the needle are well formed and grown. The others are not well developed or they do not simply exist. Some of them collapse and then disappear. Thus, the GaN needles are strongly unstable crystals. They are dark due to the presence of microdefects like dislocation loops and gallium inclusions being the result of the fast and unstable growth in the c-direction [17]. These crystals are often hollow from inside. However, they have very well developed faces of orientations different from the ones, which give an opportunity to check the growth behavior in these directions using the needles as seeds (see Chap. 3).
10.3.2 Physical Properties of the Crystals As was already mentioned, for oxygen interacting with Ga, there is no potential barrier for dissociation. Therefore, even traces of this impurity in the growth system are sources of the unintentional oxygen doping of GaN. Consequently, the HP-GaN crystals (both platelets and needles) are strongly n-type with uniform free electron
216
M. Bo´ckowski et al.
concentration of about 5 1019 cm3 (metallic conductivity) and mobility of about 60 cm2 =Vs [18]. These free carriers can be fully eliminated by Mg acceptor added into the growth solution [19]. The presence of the native point defects in the crystals (platelets) was checked by positron annihilation measurements [20]. High concentration of Ga vacancies .VGa / was found in the conductive crystals in contrast to the Mg-doped samples where VGa were not observed. This agrees with the theoretical prediction that the formation energy of VGa decreases with an increase of the Fermi level energy [21]. It suggests that the creation of these defects is thermodynamically controlled. The difference in the PL spectra of the conductive (strong yellow emission) and Mgdoped crystals (no yellow emission, but blue Mg-related signal) revealed that VGa are involved in strong yellow luminescence in GaN. The structure of the pressure grown GaN crystals has been studied by X-ray diffraction [22]. The crystallinity of the conductive platelets depends on the size of the crystal. The full widths at half maximum (FWHM) for the reflection 002 are of the order of 0.02 degree for 1–3 mm crystals. For larger platelets the rocking curves often split into a few peaks showing the presence of low angle boundaries separating grains of a few millimeters in size. Misorientation between grains increases monotonically from end to end of the crystal [21]. The crystallinity of the needles is similar to that of the platelets [17]. It has been shown by TEM examination [14, 23] that the N-polar 0001N surface of the n-type pressure grown GaN platelets is often atomically flat and that the crystals with this surface are practically free of extended defects. On the opposite, rough surface, a number of extended defects like stacking faults, dislocation loops, and Ga microprecipitates are observed. The relative thickness of this part usually consists of 10% of the entire thickness of the platelet. In order to determine dislocation densities in GaN crystals, the defect selective etching (DSE) method has been developed [24]. It has been shown that etching in molten KOH–NaOH eutectic (at 723 K) reveals dislocations in GaN pressure-grown single crystals. The etch pit density in HP-GaN (estimated by counting the etch pits in a given area) is very low, of the order of 102 cm2 . More detailed description of the properties of the pressure grown GaN crystals can be found in [25].
10.4 Seeded Growth by HPS Method The main disadvantage of spontaneous crystallization by the HPS method is poor reproducibility of the size and distribution of GaN crystals in the crucible. Better control of these two parameters may be achieved by growth with seed crystals introduced intentionally into the solution. Two techniques have been developed in the Crystal Growth Laboratory of the Institute of High Pressure Physics: LPE and seeded growth with convective flow of gallium under control.
10
High Pressure Solution Growth of Gallium Nitride
217
10.4.1 Liquid Phase Epitaxy in the c-Direction on Various Substrates The LPE technique, i.e., deposition of GaN on the substrate to force the growth in particular direction has been examined at two types of gradient configurations: positive and negative and with five kinds of substrates: HP-GaN, HVPE-GaN, SiC, sapphire, and sapphire/GaN MOCVD templates. The experiments have been carried out in three brands of the HP reactors, what allowed us to use three sizes (in diameter) of substrates: 1 cm, 1 in., and 2 in., respectively. The substrate has been placed at the bottom or at the top of the crucible. A thin tube has always been located in the crucible in such a way that the edges of the seed have not been exposed to the liquid gallium. In this way, the edge nucleation has been eliminated. The crucible with a baffle plate positioned close to the seed has been used. This baffle plate has allowed obtaining a flat crystallization front on the substrate and consequently maintaining a flat GaN surface during a long crystallization run [26]. The crystallization without the baffle has always led to nonuniform crystallization front,and thereby formation of the growth hillock at the center of the substrate [13]. This last phenomenon has perturbed the crystallization process. The crystallization process has also been perturbed if the nitrogen polar surface of the substrate has been exposed to the liquid gallium (case of HP-GaN and HVPE-GaN). Then, the presence of several growth centers has been observed. The simultaneous nucleation and growth of randomly oriented crystals have been noticed, too [13]. In turn, on the gallium polar surface (case of HP-GaN, HVPE-GaN and sapphire/GaN MOCVD template) the growth mode has been stable and proceeded by the steps propagation from the center of the sample to its sides [27]. The LPE crystallization on pure sapphire has been dominated by the polycrystalline GaN growth [28]. On the 4H-SiC substrates the main mechanism of crystallization has been the formation of many growth hillocks, coalesced by small steps propagating from each of the hillock [28]. The permanent feature of GaN grown on SiC has been that the GaN has totally been cracked. These cracks have been formed during cooling procedure, just after the main growth process, due to the difference in thermal expansion coefficients between GaN and SiC at low temperature [28]. It is worth noting that cracking has never been observed for the growth on HP-GaN, HVPE-GaN, or sapphire/GaN MOCVD template. After HP LPE on the sapphire/GaN MOCVD template the bending of the sample has just been observed. It has been found that for the same temperature gradient and supercooling the growth rate in negative (reversed temperature gradient) configuration is twice as fast [27]. It could indicate that the growth rate is governed by a nitrogen transport mechanism. In the configuration with reverse temperature gradient, the convection flow is always stronger. Therefore, for such configuration the crystal can be grown in relatively higher supersaturation compared to the direct gradient system. It has been shown however that there are two factors responsible for GaN growth in the c-direction and these factors are not depended on the configurations or the kind of the substrates used. For a short time .30 h/, the growth rate is governed mainly
218
M. Bo´ckowski et al.
by nitrogen transport to the crystallization zone, and the observed average growth rate can attain 10 m=h. After a longer time .100 h/, even at high temperature gradients, the surface kinetic factor becomes more important and the average rate decreases to 1 m=h. In our opinion, the macrosteps and terraces have been created on the growing crystal surface which finally hindered or even stopped the growth [26]. However, Hussy et al. [29] have proposed a different explanation. During LPE at low pressure solution growth process, it has been observed that at relatively high supersaturation the parasitically growing GaN consumed nearly all nitrogen which has been supplied in the solution. The amount of nitrogen that has been transported through the whole solution volume to the seed has been decreasing with increasing process time. Therefore, the growth rate of the seed has decreased. The analysis of the GaN mass crystallized at configuration with and without seed has confirmed that the growth rate on the seed is governed by surface kinetics. The GaN mass crystallized at the bottom of the crucible without seed was bigger than the mass crystallized on the seed (with no parasitic growth observed) under the same experimental conditions [28]. Thus, not all nitrogen atoms from the solution reaching the seed surface were adsorbed there. Figure 10.6a, b present a general view of the samples deposited on HP-GaN and sapphire/GaN MOCVD templates, respectively. The average dislocation density in
Fig. 10.6 (a) General view of 100 m thick GaN layer deposited on HP-GaN; (b) general view of 100 m thick GaN layer deposited on 2 in. MOCVD GaN/sapphire template; (c) general view of 100 m thick GaN layer deposited on HVPE-GaN – grid 1 mm; (d) Nomarski microphotograph of the surface after DSE. Homogenous distribution of etch pits is visible
10
High Pressure Solution Growth of Gallium Nitride
219
GaN crystallized on the HP-GaN is always as low as 102 cm2 . The dislocation density in the pressure-grown material on sapphire/GaN templates is of the order of 3–5 107 cm2 , for layers thicker than 40 m. This value is more than one order of magnitude lower than in the outgoing templates where the dislocation density varies from 8 108 to 109 cm2 . Lower defect density can be obtained if the HVPE-GaN is used as a substrate. Then, the dislocation density in the starting material is of the order of 5 106 cm2 . For 40 m thick GaN layer the defect density is 2 106 cm2 . For the layer 250 m thick the etch pit density attains even 8 105 cm2 . Figure 10.6c,d present newly deposited material on HVPE free-standing crystal and its surface with etch pits, respectively. Two kinds of etch pits are visible: smaller are associated with edge dislocations, larger associated with screw or mixed-type dislocations. The number of edge dislocations is bigger by a factor of two. The new deposited material on HP-GaN, HVPE-GaN, SiC, or sapphire/GaN template is always n-type with free electron concentration of about 5 1019 cm3 and mobility of 60 cm2 =V s. It also shows strong yellow luminescence. The FWHM of GaN grown on the sapphire/GaN MOCVD template consists of a few narrow peaks, suggesting that the new deposited GaN is composed of small grains of about 0.1– 0.2 mm, misoriented by a few angular minutes. The curves have relatively large total width due to the significant bending of the sapphire/GaN system [26]. The FWHM of GaN grown on the HVPE-GaN is of the order of 0:1–0:15ı for the reflection 002 and the beam dimension of 3 10 mm2 . The 60 m thick pressure grown material, cut from the HVPE-GaN seed for the reflection 002 and beam 0:310 mm2 , showed the FWHM of 0:04ı and the bending radius of 30 cm. In turn, during the LPE runs on pressure grown GaN platelets the bending phenomenon is never observed and the rocking curve is always very narrow .0:01–0:02ı /, which suggests the best structural quality of the deposited material.
10.4.2 Modeling of the Convective Transport in Gallium for LPE Growth (Time Independent Solutions) The finite element calculation has been used for modeling the convective transport in the solution, during LPE process in positive temperature gradient configuration, especially for analyzing the effects of the crucible wall, seeds, and baffle. The LPE configuration with seed of diameter of 1 in. and thickness of 400 m has been selected for the modeling. The relatively small volume of Ga used in this configuration .5 cm3 / has resulted in its small Reynolds number which has assured convergences of the solutions of the calculations. The shortness of the crucible has permitted the obtaining of symmetrical temperature distribution in it. Thus, the radial temperature distribution at the bottom of the crucible could be approximated as parabolic and the vertical temperature distribution as linear (as is shown in Fig. 10.7a). The radial temperature distribution on the crucible’s top has been assumed as constant. The analysis of the convective transport in the gallium at twodimensional approximation has been carried out with FIDAP 8.5 supplied by Fluent Inc. The convectional flow velocity in the gallium and the temperature distribution
220
M. Bo´ckowski et al.
a
N2
z
T(r) 1723 1673
1623
1623 1593
T substrate T(r)
b
Fig. 10.7 (a) Scheme of the experimental LPE configuration with the baffle plate; the temperature values and assumed radial and axial temperature distributions are marked; (b) stream lines in the gallium for the presented configuration
in the liquid, in the seed, and in the crucible wall have been determined. One has considered Ga as the incompressible, viscous and Newtonian fluid, where density changes have been produced only by temperature changes. The boundary conditions have been fixed by temperatures measured during the LPE process. They are presented in Fig. 10.7a. The mesh composed of 41,000 elements has been introduced. Assuming the laminar flow in the gallium the following set of equations has been solved in order to find stationary and time independent solutions: 1. continuity equation for Ga r .Eu/ D 0;
(10.2)
where uE is a velocity vector for the given mesh’s element. 2. momentum balance equation for Ga
0 uE.r uE/ D rp C r 2 uE 0 ˇT0 .T T0 /g; E
(10.3)
10
High Pressure Solution Growth of Gallium Nitride
221
where T0 is the arbitrary chosen reference temperature .T0 D 1; 673 K/; 0 is the gallium density at T0 ; the viscosity, ˇT0 the gallium thermal expansion coefficient at T0 ; T the temperature, gE the gravity vector and p pressure in the liquid metal given by the following formula: 1 p D r uE "
(10.4)
where " is a penalty parameter equal to 107 [30]. 3. energy balance equation for Ga,
0 cp .Eu rT / D r .krT /
(10.5)
where cp is the specific heat capacity of gallium, and k is its thermal conductivity coefficient. Since uE D 0 in the crucible and seed, for graphite and seed the following equation has been taken: r .ki rT / D 0; (10.6) where ki represents the thermal conductivity coefficient for graphite or the seed respectively. The solution of the equations presented above has allowed us to determine: uE and T and present them graphically as the stream lines, velocity vectors, and isotherm lines for the examined system (graphite crucible, seed, and first of all the liquid gallium). Some of the material properties of gallium, graphite, and seeds used in the simulations are presented in Table 10.1. Since the temperature dependence of the specific heat of liquid gallium is not strong, one has considered cp as constant for the temperature range used [31]. However the thermal conductivity of gallium has been determined for the given temperature from the following equation [31]: k0 D 22:3 C 1:8 .T 273:15/0:5 :
(10.7)
The gallium viscosity has been approximated by the relation [31]: .T / D 1:729
T 273 350
n ;
(10.8)
Table 10.1 Properties of gallium, graphite GaN, and sapphire used in the simulations Density Specific Thermal Thermal expansion heat conductivity coefficient Gallium Graphite GaN
5;360 kg=m3 [31] 2;300 kg=m3 [32] 6;100 kg=m3 [33]
374 J/kgK [31] 202 J/kgK [33] 500 J/kgK [34]
See the text 30 W/mK [32] 24 W/mK [35]
0:11 103 1=K [31] – –
222
M. Bo´ckowski et al.
where n.T / D 1:135 0:41 103 .T 623/ C 0:80 107 .T 623/2:
(10.9)
In general, the modeling of the convective transport in the solution has allowed us to demonstrate that the measurements of the temperature distribution in the wall and bottom of the crucible have not reflected the real temperature distribution in the liquid gallium [27]. For the given temperature distribution the convective rolls’ behavior has been totally different for the gallium filling the crucible and the gallium being free (crucible neglected). It has also been shown that the substrate in the crucible can change the order of the convectional flow in the solution [36]. It depends on its thermal conductivity coefficient and a relation of this coefficient to the thermal conductivity of the crucible material. Since the temperature gradients realize in the material of the lowest coefficient of thermal conductivity, the seed of the coefficient lower than for graphite (case of sapphire) can cut off the influence of the radial gradient (controlled in the bottom) for the mass of liquid gallium. For the same reason an introduction of the graphite material to the gallium (case of the baffle) also perturbs and changes the roll’s and isotherm lines’ distributions in the liquid. Figure 10.7b shows the streamlines for the configuration with the baffle plate and the GaN crystal as seed. One can see two pairs of the convectional rolls above the baffle. They are separated by the graphite stick. Below the baffle two rolls are observed. There, the gallium flows along the baffle to the center, then comes downwards and backs along the substrate to the sides of the crucible. The maximum velocity of the convectional flow above the baffle is 1 mm/s. Under the baffle the convection is very weak; the maximum flow velocities in two rolls achieve 0.2 mm/s. The nitrogen is transported to the baffle by relatively strong convection, then the N atoms are dispersed on the baffle but by the sides (opening areas) they are supplied below the baffle to the seed region, where their velocity is low and thus by very weak convection process they are transported to the seed. It seems that due to very low convectional flow velocity the growth is macroscopically stable but the rate as slow as 1 m=h.
10.4.3 Seeded Growth with Convective Flow of Gallium Under Control Bulk GaN crystallization may take place when a properly prepared GaN seed is wholly immersed in the liquid Ga:N solution. Then, the growth can proceed in any of the available crystallographic directions. Additionally, the control of the convective flow of gallium can give a proper nitrogen transport to a supersaturated zone of the solution and therefore stable and relatively fast GaN growth. Figure 10.8a presents one of typical experimental configurations used for the bulk growth by HPS method: a graphite crucible with special graphite baffles system, tube, seeds, thermocouples positions, and temperature values measured during crystallization runs.
10
High Pressure Solution Growth of Gallium Nitride
223
a 1623 GaN seed 1623 1630
1643 1673
Fig. 10.8 (a) Scheme of typical experimental configurations used for the bulk growth by HPS method: a graphite crucible with special graphite baffles system, tube, seeds, thermocouples positions, and temperature values expected during typical crystallization runs; (b) typical HVPE seed; in its upper part the special slot allowing the hanging of this crystal in the crucible is visible; (c) photograph of the real experimental configuration; four hung seeds, the baffle, and the system of tubes are visible
This is a two-zone reversed temperature gradient configuration; the hot part of the solution is at the lower part of the crucible and the cooler part is above. The convective flow in the liquid gallium is provided and controlled by the radial temperature gradient. The seeds have been placed in the Ostwald–Miers zone of the solution, supersaturated metastable zone in which crystallization is possible, but spontaneous
224
M. Bo´ckowski et al.
nucleation does not take place. The search for this zone for presented configuration is described in details in [37]. The HVPE crystals have been used as the seeds. Their surfaces have been prepared by mechanical polishing and reactive ion etching. The crystals have had a special slot (see Fig. 10.8b), which allowed their hanging and holding by the graphite sticks, at fixed places in the liquid solutions. The use of the HP reactor of the internal diameter of 10 cm has allowed placing in the crucible several, bigger than 1 in., seeds in one crystallization run. The photograph of the configuration described above is shown in Fig. 10.8c. The seeds have been positioned perpendicular to the crucible radius, although they can also be positioned in parallel or in free chosen directions. The growth rate has not been dependent on the crystallographic directions or planes and varied from 1 to 3 m=h. Again, like for LPE configuration, a decreasing of growth rate in time has been observed. Therefore, for obtaining more than 300 m thick GaN layer on one of the seed planes, the minimum 200 h crystallization process has had to be carried out.
10.4.4 Growth on HVPE Seeds in the c-Directions During the long period of crystallization under pressure .400 h/, the HVPE seeds have been grown up to 700 m per side. The growth in nonpolar directions has not been observed. The parasitic crystals have been grown on the seed’s edges. The growth of GaN “wings” from the edges of the seed, on the gallium side, has also been detected. These phenomena have been associated with the local (close to the seed edges) increasing of the supersaturation. The cross section analysis of the growing samples (see Fig. 10.9) has shown that the new grown material has been deposited in a uniform way, from the thickness point of view. The nitrogen side has been grown showing the presence of several hillocks (growth centers). On the contrary, the gallium side has been grown by macrosteps propagation. Cellular instabilities have been observed on this side. They have increased in the time of the crystallization run. Figure 10.10a represents the photograph of one of the seed used for the bulk growth. Figure 10.10b shows the same seed after 300 h of the crystallization process. The part of this crystal has been cut perpendicular to the c-direction and the pressure grown free-standing material has been obtained (see Fig. 10.10c). The Hall effect measurements on Ga side of the HP free standing crystal have shown that the carrier concentration in the material has been uniform- about 5 1019 cm3 . This result has been confirmed by the X-ray measurements of the c parameter in the crystal. The variation of this parameter has been 0.0003 Å (see Fig. 10.11a). This variation corresponds to the difference in the free carrier concentration of 1:4 1019 cm3 [38]. In turn, the FWHM has been of the order of 0:015ı for the reflection 004 at the beam of 0:2 1 mm2 (see Fig. 10.11b). The bowing of the crystal has been of the order of 10 m. An evident change in the FWHM of the HP-GaN coupled to the HVPE seed and the HP free standing GaN crystal should
10
High Pressure Solution Growth of Gallium Nitride
N side
HVPE seed
225
{1120} cleaved
{1010} polished
Ga side
Fig. 10.9 Cross section of GaN crystal grown by HPS method on HVPE seed; grid 1 mm
Fig. 10.10 (a) Free standing HVPE crystal (700 m thick) used as a seed; (b) free standing HVPE crystal after HP crystallization process (1.5 mm thick); (c) 300 m thick hp-GaN crystal cut from the HVPE seed; grid 1 mm
be noted. The FWHM for the HP-GaN coupled to the seed has been of the order of 0:12ı for the reflection 002 and beam dimension of 3 10 mm2 . The FWHM of the free standing HP-GaN has been of the order of 0:025ı for the same reflection and beam dimension. It seems that for the free standing material, some stresses due to the difference in lattice constant between HVPE seed and pressure grown GaN crystal have been released. The etching process in KOH/NaOH eutectic at 450 ıC of the HP free standing GaN crystal has shown that the etch pits density on the gallium side has been 1 106 cm2 . The main disadvantage of this material has been the presence of graphite precipitates. Their density has locally attained 102 cm2 .
226
M. Bo´ckowski et al.
a 11
5.18625 5.18628 5.18630 5.18633
Y [mm]
10 9
5.18636 5.18639
8
5.18641 5.18644
7
5.18647 5.18649 5.18652
6
5.18655 5.18658 5.18660 5.18663 5.18666
5 4
5.18668 5.18671
3
5.18674 5.18676
2
5.18679 5.18682 5.18685
1
5.18687 5.18690
0
b
1
12000
2
3
5 4 X [mm]
6
7
8
FWHM=0.015 deg. Beam: 0.2 mm x 1 mm
counts
8000
4000
0 36.40
36.47 omega, deg
36.54
Fig. 10.11 (a) Variation of the c parameter in the pressure grown free-standing crystal; the difference between green and red areas corresponds to 0.0003 Å; (b) X-ray rocking curves for this crystal
10
High Pressure Solution Growth of Gallium Nitride
227
10.4.5 Growth on HVPE Seeds in Nonpolar Directions The seeded growth in nonpolar directions under high nitrogen pressure was carried N out only in direction. The seed for this growth was prepared by HVPE crystallization of thick GaN layer (few millimeters) on the sapphirenGaN MOCVD template. The˚HVPE free standing material was cut in such a way that small platelets N planes were obtained. The photograph of the seed is presented of GaN with 1010 in Fig. 10.12a. The crystallization was carried out at typical parameters already presented in this chapter (see Fig. 10.8a). After 100 h the seed grew about 150 m per side. The growth mode was unstable and it was very difficult to detect. Some regions with rectangle steps were observed. The parasitic deposition of small GaN crystals on the seed surfaces was watched. Thus, the crystal was polished on both sides. Since we were interested in free-standing material the sample was polished in such a manner that one of its side and the seed were removed. The X-ray measurement showed that the FWHM of the free standing crystal was 0:05ı for the reflection 100 and beam diameter of 3 10 mm2 . In the luminescence spectrum measured at 10 K the strong peak of the yellow luminescence was found. This suggested that the material was n-type. The Raman spectroscopy confirmed the good structural quality of the crystal and showed that the free electron concentration was higher than 5 1018 cm3 (the Raman spectrum is shown in the Fig. 10.12b). The Hall Effect measurements confirmed it. The free carrier concentration was of the order of 1:6 1019 cm3 (at mobility of 95 cm2 =Vs) thus lower by a factor three than the free carrier concentration in HP-GaN grown in polar direction. It ˚ suggests that more N planes. oxygen can be built on the polar surfaces than on the nonpolar 1010 The crystal was etched in KOH/NaOH eutectic at 450 ı C in order to reveal the etch pits. Only one kind of pit (in sense of their size) was detected. The defect density on the 70% of the crystal surface was 3 106 cm2 , on 20% was 5 105 cm2 , on 10% was 5 107 cm2 . Figure 10.13 presents the surface of the crystal after the DSE procedure. The regions with various defect densities are marked.
10.4.6 Modeling of the Convective Transport in Gallium (Time Dependent Solutions) In comparison to the previous calculations presented in the Sect. 10.4.2, the stationary and time independent solutions have been resigned. The time-dependent set of equations has been used. The change of the equations has been associated with an increase of the size of the experimental setup and changing of the dimension of the calculated system from 2D to 3D. With an increase in the volume of gallium, the Reynolds number has increased too and the solutions have not been convergent in the stationary model. In a case of time-dependent solutions the convergence has been obtained. Thus, the flow in any moment of the simulation and the change in the mass transfer in time could be analyzed.
228
M. Bo´ckowski et al.
˚ Fig. 10.12 (a) Seed; 101N 0 plane of HVPE-GaN grid – 1 mm; (b) Raman spectrum of the pressure grown free standing material; the crystal is also presented; the c-direction is marked; grid – 1 mm
Fig. 10.13 Pressure grown nonpolar GaN crystal after the DSE procedure with three regions of various defect densities
10
High Pressure Solution Growth of Gallium Nitride
229
The analysis of the convective transport in the gallium at three-dimensional approximation for the configuration presented in Fig. 10.8a has been carried out with the code Fluent supplied by Fluent Inc. All assumptions have been the same as for calculations presented in Sect. 10.4.2. Only the mesh has been changed and composed of 2 106 elements. In order to find time dependent solutions the modified set of the equations (10.2)–(10.6) have been solved: r uE D 0; @Eu E
0 C uE.rEu/ D rp C r 2 uE 0 ˇT0 .T T0 /g; @t @T C uE rT D r .krT /:
0 cp @t
(10.9) (10.10) (10.11)
Since uE D 0 in the crucible and seeds, the following equation was taken for graphite and GaN crystals: @T D r .ki rT /;
i cpi (10.12) @t where ki represents the thermal conductivity coefficient, cpi the specific heat capacity, ¡i the density, for the graphite and gallium nitride, respectively. The solution of equations presented above has allowed us to determine uE and T in any time of the simulation as the initial conditions uE D 0 has been taken. The simulations have been carried out to the moment of time when the solution for isotherm lines has stopped its evolution and has not been changed in time. The material properties of gallium, graphite, and GaN have been the same as those presented in Table 10.1. The heat capacity .cp / of gallium, graphite, and GaN has been taken as constant for the temperature range used. The thermal conductivity and viscosity of gallium for the given temperature have been determined according to (10.7)–(10.9) presented earlier. Figure 10.14a represents the convectional velocity vectors in the gallium at 2,270th second of the simulations for the temperature distribution presented in Fig. 10.8a. As seeds, GaN platelets with the dimensions 30 20 1 mm3 have been assumed. The hot gallium flows up through the graphite tube with the maximum velocity of 2.20 cm/s at the upper part of the crucible (red area in Fig. 10.14a). The main flow jets from the tube and hits the graphite closure of the crucible. Then, it flows to the sides and down. The flow close to the seeds is slower, up to 3 mm/s. Figure 10.14b shows the region just close to the seed. Between the wall of the crucible and the seed the gallium flows down but in the region between the tube and the seed it flows up. The correlation between direction of the gallium flow and morphology has not been found. The experimental configuration presented in Fig. 10.8a appears to be very interesting and important for further crystallization processes. Varying the tube diameter, thickness of the tube’s wall, or its length, the flow velocities and rolls positions can be changed for the given temperature distribution. Varying the radial temperature gradient (the temperature at the bottom of the crucible) the flow velocities can be changed for the given crucible configuration. For example at radial temperature
230
M. Bo´ckowski et al.
Fig. 10.14 (a) Convectional velocity vectors in the gallium at the end of the simulations (in 2,270th second); simulation for the temperature distribution presented in Fig. 10.8a, (b) convectional velocity vectors in the gallium close to the seed
gradient of 5 K, the maximum convective velocity has only been 2 mm/s and the flow velocity close to the seed has been lower than 1 mm/s [37]. Finally, it should be possible to find an appropriate seed place in the solution, an appropriate place for relatively fast but stable growth of GaN. For that, however, an equation for the nitrogen concentration in liquid gallium should be added to the set of equations presented in this subchapter. Then, the supersaturation in any place of the solution might be determined.
10.5 Applications of Pressure Grown GaN Substrates: Blue Laser Diodes in TopGaN Ltd It is well known that GaN substrates for nitride based LDs technology should have high structural quality and low threading dislocation density. The substrates have to be relatively thick, for a possibility to miscut them appropriately for epitaxial growth and so that they are highly conducting which allows for preparation of stable and low-resistance ohmic contact. The pressure grown GaN crystals (HPGaN) almost fill these conditions. Spontaneous HP-GaN platelets have, as shown in Sect. 10.3.2, very good structural quality and very low defect density. They are also highly conductive. Unfortunately, they are relatively thin and miscuting them is rather impossible. For the use of the platelets as substrates for epitaxy by both MOCVD [39] and MBE [40] methods, the mechano-chemical polishing procedures have been used.
10
High Pressure Solution Growth of Gallium Nitride
231
Fig. 10.15 Structure of a laser diode grown on bulk GaN crystals in TopGaN Ltd
Figure 10.15 shows schematically a typical structure of a LD fabricated in TopGaN Ltd. The active layer of the device has been composed of one to five In0:1 Ga0:9 N quantum wells. The diode has been mounted p-side down on metalized diamond heat-spreaders. TopGaN LDs during CW operations have shown a threshold voltage of around 6 V@5 kA=cm2 at emission wavelength from 395 nm to 420 nm. The lifetime of the devices has achieved a few thousands hours. The LDs structures have been characterized by a density of dislocations at the level of around 105 cm2 . The origin of the dislocations has been the large mechanical strain existing in cladding and blocking AlGaN layers, high magnesium content in p-type part of the epi-structure, and segregation tendency of indium in InGaN quantum wells [41]. In spite of the relatively high defect density and graphite precipitates, pressure grown free standing HP-GaN crystals, described in Sect. 10.4.4, have also rendered good substrates for fabrication of the blue LDs. The defect density in these laser structures (made by MOCVD) has never exceeded 5 106 cm2 . During CW operation the lasers of the emission wavelength of 395 nm showed a threshold voltage of around 5:5 V@5 kA=cm2 . The lasers also show very satisfying spatial distribution of the emission wavelength with the variation 395 ˙ 2 nm. No doubt, it is due
232
M. Bo´ckowski et al.
to good substrate properties - first of all applicable thickness allowing miscut of the substrate for the epitaxy. Today, at TopGaN and the Institute of High Pressure Physics laser technology level, it appears that the more important features of good substrates are their thickness and the possibility of miscutting them rather than their exclusively low defect density.
10.6 Summary and Perspectives of HPS Growth Method In this chapter the state of the art HPS growth method has been described. The spontaneous growth of platelets and their use as substrates for blue LDs by TopGaN Ltd. has been presented. The progress in seeded growth by HPS method has also been shown. The pathway from LPE to the crystallization on the seeds with convective flow of gallium under control has been demonstrated. The modeling of the convective transport in gallium has been used to better control the gallium flow during the crystallization process. The new class of HP-GaN crystals has been obtained. These crystals, grown by HPS method on the HVPE seeds, satisfy well the conditions required for the substrates for LDs. Today, their main disadvantage is the presence of graphite precipitates and defect density of the order of 106 cm2 . Despite these, they have been successfully used as substrates for the LDs. In the future, the growth on the HVPE-GaN crystals as seeds would be in obtaining HP-GaN crystals bigger than spontaneously grown platelets, uniform in shape, and in physical properties. Obviously, the spontaneous growth of the platelets is still being carried out, since these crystals seem to be very good seeds for crystallization with convective flow of gallium under control and/or for the HVPE growth (see Chap. 3 of this book) due to their extremely low dislocation density. Generally in the HPS method, the growth rate should be increased. Most probably, addition of appropriate impurities to the solution may be very helpful. As has been shown by Kawamura et al. [42] the LPE at sodium flux method with Li additive allows the obtaining of growth rate of the order of 20 m=h. Another approach is to crystallize ˚ in nonpolar and semipolar directions. The first tests of the growth on the N planes have shown that the growth rate has not been changed or increased. 1010 It seems that rather highly indexed planes of the seeds should be used. However for that, thicker HVPE free standing crystals are needed. Thus, the HVPE technique and crystallization on the sapphire/GaN MOCVD templates and the pressure grown crystals have to be developed fast. The growth with appropriate impurity in the solution and on appropriate seed (with highly indexed planes exposed to the solution) becomes the goal of our future work. Acknowledgements This chapter could be written thanks to the results obtained inter alia during realization of two scientific projects: Research Grant No R0203601 supported by the Polish Ministry of Science and Higher Education and Structural Founds Grant No WKP_1/1.4.3/2/2005/ 13/132/322/2007/U supported partially by the Polish Ministry of Science and Higher Education (34%) and the European Union (66%). The authors would like to thank Drs E. Litwin-Staszewska, J. Domagala, M. Krysko, H. Teisseyre, A. Khachapuridze, and G. Kamler for the Hall, X-ray, Raman, luminescence, and DSE measurements.
10
High Pressure Solution Growth of Gallium Nitride
233
References 1. W. Utsumi, H. Saitoh, H. Kaneko, T. Watanuki, K. Aoki, O. Shimomura, Nat. Mater. 2, 735 (2003) 2. J.A. Van Vechten Phys. Rev. B7, 1479 (1973) 3. D. Elwell, H.J. Scheel, Crystal Growth from High-Temperature Solutions (Academic, London, 1975) 4. J. Karpinski, J. Jun, S. Porowski, J. Cryst. Growth 66, 1 (1984) 5. S. Porowski, I. Grzegory, J. Cryst. Growth 178, 174 (1997) 6. Grant No 1270/C T08–7/95 of the Polish Committee for Scientific Research (1995) 7. I. Grzegory, M. Bockowski, S. Krukowski, B. Lucznik, M. Wroblewski, J.L. Weyher, M. Leszczynski, P. Prystawko, R. Czernecki, J. Lehnert, G. Nowak, P. Perlin, H. Teisseyre, W. Purgal, W. Krupczynski, T. Suski, L.H. Dmowski, E. Litwin-Staszewska, C. Skierbiszewski, S. Lepkowski, S. Porowski, Acta Phys. Pol. A 100, 229 (2001) 8. C. Skierbiszewski, Z.R. Wasilewski, M. Siekacz, A. Feduniewicz, P. Perlin, P. Wisniewski, J. Borysiuk, I. Grzegory, M. Leszczynski, T. Suski, S. Porowski, Appl. Phys. Lett. 86(1), 11114 (2005) 9. Grant No 2700/C.T11–8/2000 of the Polish Committee for Scientific Research (2000) 10. I. Grzegory, M. Bockowsk, B. Lucznik, S. Krukowski, Z. Romanowski, M. Wroblewski, S. Porowski, J. Cryst. Growth 246, 177 (2002) 11. S. Krukowski, Z. Romanowski, I. Grzegory, S. Porowski, J. Cryst. Growth 189/190, 159 (1999) 12. S. Krukowski, M. Bockowski, B. Łucznik, I. Grzegory, S. Porowski, T. Suski, Z. Romanowski, J. Phys. Condens. Matter 13, 8881 (2001) 13. M. Bo´ckowski, I. Grzegory, S. Krukowski, B. Łucznik, Z. Romanowski, M. Wróblewski, J. Borysiuk, J. Weyher, P. Hageman, S. Porowski, J. Cryst. Growth 246, 194 (2002) 14. Z. Liliental-Weber, EMIS Datareview Series No23, vol. 230 (INSPEC, The Institution of Electrical Engineers, London, 1999) 15. J.L. Rouviere, J.L. Weyher, M. Seelmann-Eggebert, S. Porowski, Appl. Phys. Lett. 73, 668 (1998) 16. M. Seelmann-Eggebert, J.L. Weyher, H. Obloh, H. Zimmermann, A. Rar, S. Porowski, Appl. Phys. Lett. 71, 18 (1997) 17. M. Bockowski, I. Grzegory, G. Kamler, B. Łucznik, S. Krukowski, M. Wróblewski, P. Kwiatkowski, K. Jasik, S. Porowski, J. Cryst. Growth 305, 414 (2007) 18. P. Perlin, J. Camassel, W. Knap, T. Talercio, J.C. Chervin, T. Suski, I. Grzegory, S. Porowski, Appl. Phys. Lett. 67, 2524 (1995) 19. E. Litwin-Staszewska, T. Suski, R. Piotrzkowski, I. Grzegory, M. Bockowski, J.L. Robert, L. Konczewicz, D. Wasik, E. Kaminska, D. Cote, B. Clerjaud, J. Appl. Phys. 89, 7960 (2001) 20. K. Saarinen, T. Laine, S. Kuisma, P. Hautojarvi, L. Dobrzy´nski, J.M. Baranowski, K. Pakuła, R. St˛epniewski, M. Wojdak, A. Wysmołek, T. Suski, M. Leszczy´nski, I. Grzegory, S. Porowski, Phys. Rev. Lett. 79, 3030 (1997) 21. J. Neugebauer, C.G. Van de Walle, Phys. Rev. B 50, 8067 (1994) 22. M. Leszczynski, I. Grzegory, H. Teisseyre, T. Suski, M. Bockowski, J. Jun, J.M. Baranowski, S. Porowski, J. Domagala, J. Cryst. Growth 169, 235 (1996) 23. S.H. Christiansen, M. Albrecht, H.P. Strunk, C.T. Foxon, D. Korakakis, I. Grzegory, S. Porowski, Phys. Stat. Sol. (A) 176(1), 285 (1999) 24. J.L. Weyher, P.D. Brown, J-L. Rouviere, T. Wosinski, A.R. Zauner, I. Grzegory, J. Cryst. Growth 210, 151 (2000) 25. I. Grzegory, M. Bockowski, S. Porowski, in Bulk Crystal Growth of Electronic, Optical and Optoelectronic Materials, ed. by P. Capper (Wiley, Chichester, 2005), p. 173 26. M. Bockowski, I. Grzegory, S. Krukowski, B. Lucznik, M. Wróblewski, G. Kamler, J. Borysiuk, P. Kwiatkowski, K. Jasik, S. Porowski, J. Cryst. Growth 274, 55 (2005) 27. M. Bockowski, P. Stak, P. Kempisty, I. Grzegory, S. Krukowski, B. Lucznik, S. Porowski, J. Cryst. Growth 307, 259 (2007)
234
M. Bo´ckowski et al.
28. M. Bo´ckowski, I. Grzegory, S. Krukowski, B. Łucznik, M. Wróblewski, G. Kamler, J. Borysiuk, P. Kwiatkowski, K. Jasik, S. Porowski, J. Cryst. Growth 270, 409 (2004) 29. S. Hussy, E. Meissner, P. Berwian, J. Friedrich, G. Müller, J. Crystal Growth 310, 738 (2008) 30. Fidap 7.0 User Manual, 1984–1993 Fluid Dynamic International, Inc. 31. S. Krukowski, I. Grzegory, M. Bockowski, B. Lucznik, T. Suski, G. Nowak, J. Borysiuk, M. Wróblewski, M. Leszczynski, P. Perlin, S. Porowski, J.L. Weyher, Int. J. Mater. Prod. Technol. 22, 226 (2005) 32. Material Property Data http://www.matweb.com 33. International Nuclear Safety Center http://www.insc.anl.gov/matprop/graphite/ 34. J.C. Nipko, C-K. Loong, C.M. Balkas, R.F. Davis, Appl. Phys. Lett. 73, 34 (1998) 35. A. Jezowski, P. Stachowiak, T. Plackowski, T. Suski, S. Krukowski, M. Bockowski, I. Grzegory, B. Danilchenko, T. Paszkiewicz, Phys. Stat. Sol. (B)240(2), 447 (2003) 36. M. Bockowski, P. Stak, P. Kempisty, I. Grzegory, B. Lucznik, S. Krukowski, S. Porowski, Phys. Stat. Sol. (C) 6, 1539 (2008) 37. M. Bockowski, P. Strak, I. Grzegory, B. Lucznik, S. Porowski, J. Cryst. Growth 310, 3924 (2008) 38. M. Krysko, M. Sarzynski, J. Domagala, I. Grzegory, B. Lucznik, G. Kamler, S. Porowski, M. Leszczy´nski, J. Alloys Comp. 401, 261 (2005) 39. L. Marona, P. Wisniewski, P. Prystawko, I. Grzegory, T. Suski, S. Porowski, P. Perlin, R. Czernecki, M. Leszczy´nski, Appl. Phys. Lett. 88(20), 201111 (2006) 40. C. Skierbiszewski, P. Wisniewski, M. Siekacz, P. Perlin, A. Feduniewicz-Zmuda, G. Nowak, I. Grzegory, M. Leszczynski, S. Porowski, Appl. Phys. Lett. 88(22), 221108 (2006) 41. P. Perlin, M. Leszczynski, P. Prystawko, M. Bockowski, I. Grzegory, C. Skierbiszewski, T. Suski, in III-Nitride Devices and Nanoengineering, ed. by Zhe Chuan Feng (World Scientific, China, 2008), p. 240 42. M. Morishita, F. Kawamura, M. Kawahara, M. Yoshimura, Y. Mori, T. Sasaki, J. Cryst. Growth 284, 91 (2005)
Chapter 11
A Brief Review on the Na-Flux Method Toward Growth of Large-Size GaN Crystal Dirk Ehrentraut and Elke Meissner
Abstract The growth of bulk GaN by the Na-flux method is reviewed. The largest GaN crystal thus far produced is two inches in size with thickness of a few millimeters along (0001) axis. The historical development of the method, experimental conditions, and quality of the grown GaN crystals are summarized.
11.1 Introduction The sodium (Na)-flux method in its original version is one of the older approaches to grow GaN; however, a relatively new development among all the techniques toward the crystal growth of bulk GaN is the Na-flux method as a liquid phase epitaxy (LPE) variant. Firstly, Na was named as the flux component, the other part being gallium (Ga), and later a larger number of other group-I and -II elements like lithium (Li), potassium (K), calcium (Ca), strontium (Sr), barium (Ba) have been worked with and their effects on the crystal formation was studied by various research groups in the past. Major progress in terms of crystal size, quality, and scalability of the crystal growth system has been reported by the group from the Osaka University who employs the LPE technique [1]. The first growth of a 2-inch size GaN crystal has been reported in 2006 [2]. This chapter will therefore have its main focus on the achievements reported by the Osaka group, which by now is considered the leader in this particular technology, and will begin with a brief overview over the historical developments in the Na-flux technique.
11.2 Historical Development in Brief Yamane et al. have observed in the mid-1990s that the synthesis of ternary nitrides like Ba2 ZnN2 ; Sr2 ZnN2 , etc. from a Na flux sometimes yielded hexagonal GaN crystals as stable phase [3]. They consequently focused on the preparation of GaN 235
236
D. Ehrentraut and E. Meissner
crystals from the Na flux using sodium azide .Na3 N/ as Na source and were first to report the growth of hexagonal GaN crystals up to 2 mm in size in 1997 [4]. Sealed steel tubes were used at temperatures .T / ranging between 600 and 800 ı C, pressure .p/ 11 MPa, and under nitrogen (N) excess. A high Na content is required to form GaN and it was reported that the formation of GaN requires less amount of Na when the temperatures rises. Moreover, the GaN size increased with decreasing Na content in the Ga–Na solution. This technology had been improved and soon Aoki and Yamane reported the growth of up to 3 mm large GaN crystals with an X-ray full-width half-maximum (FWHM) from the (0004) diffraction peak as small as 25 arcsec, synthesized by heating a Na–Ga melt placed in a boron nitride (BN) crucible at T D 750 ı C and p D 5 MPa of N2 for 200 h growth time [5]. Yields and morphology of GaN single crystals varied with the molar ratio Na=.Na C Ga/ in the starting composition. Electrical properties were measured: resistivity of 0:04 cm, carrier concentration of 1–2 1018 cm3 (n-type), and carrier mobility of 100 cm2 V1 s1 . Cathodoluminescence (CL) spectra at room temperature revealed a single peak around D 362–365 nm. The crystal size was further increased to 3 5 mm2 platelets and bulky prismatic crystals of 1 mm along the c-axis [6]. The average growth rates were 14 and 2 m h1 along c direction for the platelet and prismatic crystals, respectively. Sodium was later speculated to act as catalyst for the dissociation of N2 molecule into N ions by donating electrons [7]. Consequently, a high content of Na would accelerate the introduction of N ions into the Na–Ga melt where Nax –.GaNy / complexes would be formed. This phenomenon has been further investigated by Iwahashi et al. and Kawamura et al. [8,12]. They concluded that N2 would be ionized at the gas–liquid interface of the Na flux at T > 900 K and successively entered into the Ga–Na flux. Pressure is also promoting the solubility of N radicals. The additives Ca and Li in the Ga–Na flux improve the quantity of dissolved N, obviously by the formation of a metastable phase in the flux, thus decreasing the N activity. Some of the larger crystals are shown in Fig. 11.1. A maximum size obtained for self-nucleated GaN crystals was around 6 mm along the c-axis. Using a selfnucleated GaN crystal as seed, a second generation of GaN was deposited at T D 760 ıC at low average growth rates around 0:2 m h1 for the nonpolar faces [9]. Li3 N was added to the flux to improve the solubility of N. Figure 11.2 compares the size of a seed and a GaN crystal grown on such seed. The growth rate for seeded growth was increased to about 2 m h1 though they are assumed to be higher because of initial dissolution of the seed crystal [10]. Growth temperature and pressure were T D 900 ıC and 0:8 p 7 MPa of N2 for 72 h and the Ga melt was heated in Na vapor, which forms the Na–Ga melt. The presently largest GaN crystal grown by the Na-flux method was produced by a group from the Osaka University [2, 12]. Figure 11.3 depicts the 2-inch size GaN crystal fabricated on a (0001) HVPE-GaN substrate. The first report by the Osaka group was published in 1998 [13]. In order to increase the crystal size as well as obtaining oriented crystals, the LPE technique employing a substrate was used successfully [14].
11
A Brief Review on the Na-Flux Method
237
Fig. 11.1 GaN prismatic and needle crystals prepared at 900 ı C and 9.5 MPa for 400 h. (Image courtesy of H. Yamane, Tohoku University, Japan.)
Fig. 11.2 Self-nucleated GaN seed and GaN crystal grown on the seed. (Image courtesy of H. Yamane, Tohoku University, Japan.)
It was recognized that a forced flow of the solution in the vicinity of the growing GaN crystal would lead to improve growth conditions, hence, leading to increase the growth rate. The flux-film-coated LPE method (FFC-LPE) was introduced in 2003 [15]. The heated growth chamber with the flux-containing crucible was mounted such that swinging could be applied. At a swing frequency of 1:5 min1 , the growth rate of 4 m h1 at T D 800 ıC was achieved although the nitrogen pressure was as low as p D 0:95 MPa. Nonpolar substrates are also in the focus and the fabrication of an a-plane GaN substrate has recently been reported [16]. Strontium (Sr) was added as the additive to
238
D. Ehrentraut and E. Meissner
Fig. 11.3 A 2-inch GaN crystal grown by the Na-flux LPE method. About 0.5 at% C had been used as additive to reduce the parasitic nucleation of GaN [12] (Reprinted with permission. c 2008, Elsevier Publishers.)
gain the desired morphology. X-ray diffraction revealed the improved crystallinity, N reflection plane decreased from 1,152 to 236 arcsec i.e., the FWHM from the .1120/ for the GaN template (MOCVD GaN film on Al2 O3 ) and the LPE-grown crystal, respectively. Aside from the Na flux, there are some more attempts using alkaline and earth alkaline salts to obtain a reasonable solubility of GaN under moderate conditions. Feigelson and Henry have recently employed the system LiF BaF2 Li3 N to grow GaN single crystals about 0.5 mm long [17]. The growth temperature was 800 ıC and pressure about 2.5 MPa. Song et al. used the mixture Li3 N C Ga [18] and Wang et al. made use of the mixture Li3 N C Ca C Ga [19] at temperature around 800 ı C and around 200 kPa N2 pressure. Free-standing GaN platelets of up to 4 mm in size were produced. As appealing as all the above mentioned low temperature and pressure approaches are, it must be however noted here that apart from the technology currently under development at the Osaka University, bulk GaN crystals of sufficient size by the Na-flux or related flux methods are yet to be demonstrated.
11.3 Experimental Conditions for the LPE Growth of GaN by the Na-Flux Method Experimental conditions are typically subject to modification in order to target a specific goal like doping, control of morphology, and the likes. Basically, the following procedure is employed according to [20]: Starting materials are Ga and Na,
11
A Brief Review on the Na-Flux Method
239
both placed in an alumina crucible. A GaN=Al2 O3 template would serve as seeding substrate and is placed at the bottom of the crucible prior to filling with the starting materials. A stainless steel tube resistant to high temperature and pressure is housing the crucible. Before mentioned steps are carried out in a glove box under argon (Ar) atmosphere to prevent the Na from the reaction with moisture from the air. Next, the steel tube is pressurized with N2 gas and heated in an electric furnace. GaN crystals are typically grown in the Ga–Na solution by maintaining the temperature (around 850 ıC) and pressure (around 5 MPa) over the entire growth time. After the growth has been completed, the tube is taken out from the furnace to remove the crucible, and the Na is dissolved in cold ethanol and water. Careful handling is requisite to prevent from strong reaction of the Na. Remaining Ga can be removed with ethanol and hot concentrated chloride acid.
11.4 Growth Mechanism and Dislocations 11.4.1 Effect of Flux Composition on Growth Stability and Crystal Morphology The effect of additives on solubility, growth stability, crystal morphology as well as alloying with GaN was studied to improve the GaN crystal growth and eventually to prepare new phases. Lithium increases the solubility of GaN probably through increased solubility of N [21]. Lithium-based fluxes like the Li3 N C Ga [18] and also Li3 N C Ca C Ga [19] were applied; however, it has been noted that excessive Ca in the flux hinders the formation of GaN. Bao et al. have recently reported the use of a flux composed of Ba3 N2 C Ga to grow small crystals of GaN [22]. As for the flux with Ca, higher concentration of Ba strongly reduces the formation of GaN and also affects the growth rates of the different crystal facets; consequently, the morphology of the crystal can be modified. Sekiguchi et al. have reported that using K instead of Na would yield cubic GaN crystals [23]. According to CL measurements, the cubic GaN crystals prepared and peaking around 3.2 eV were highly defective. Iwahashi et al. illustrated the change of the crystal morphology that is caused by Sr as an additive to the melt [16]: As the portion of Sr ˚ was increased from 0 to N 1.5 mol% with respect to Na, the formation of m-faces 1100 gained strength at ˚ N the cost of the pyramidal facets 1011 , Fig. 11.4. At 1.5 mol% Sr in the Na flux, self-nucleated GaN would only be bound by both basal faces (0001) and ˚ crystals N and the 1100 N .0001/, faces. Further increase of the Sr content to 3.5% did not yield GaN crystals at all. Kawamura et al. stated on the effect carbon (C) would have as an additive [24]. The growth rate was significantly increased above 20 m h1 (even 30 m h1 was reported [24]) because of suppressing theparasitic nucleation and stabilizing the
240
D. Ehrentraut and E. Meissner
Fig. 11.4 Photograph of GaN produced on a-plane GaN templates in (a) the pure Na flux and c 2007, The Japan Society of (b) the Sr-added Na flux. (After [16]. Reprinted with permission. Applied Physics)
formation of the nonpolar m-face. Already 1 at% C at T D 800 ıC would cause complete suppression of the parasitic nucleation. Moreover, seemingly C does not lead to strongly affect the C-level in the grown GaN. SIMS analysis revealed concentrations about 2 1017 cm3 which is the same amount as for those crystals grown under intentionally C-free conditions. The transition metal additives chromium (Cr), manganese (Mn), iron (Fe), cobalt (Co), and nickel (Ni) have been employed [25]. Chromium resulted in coprecipitation of CrN single crystals whereas Mn, Fe, Co, and Ni rather had an effect on the GaN crystal morphology in that growth in the c- direction was enhanced with Ni illustrating the largest effect. Only Mn could be incorporated into GaN at concentrations up to 0.35 at% as detected by ICPMS. Small crystals of the —-phase of Mn2 N could be prepared from an indium In–Na flux at 700 ıC [26].
11.4.2 Growth Mechanism and Effect on Dislocation Population Several factors do influence the defect formation and propagation during seeding and crystal growth, which was studied for the growth of GaN on (0001) MOCVDGaN substrates [27]. Namely, this is the concentration of dislocations at the initial growth stage (nucleation of GaN islands on the substrate), which can be as high as this of the substrate typically for MOCVD layers 109 cm2 . This is followed by the dislocation lines bending within a few micrometers above the interface of the substrate and the LPE grown GaN, which results in the low dislocation density of 103 cm2 [28] (Fig. 11.5).
11
A Brief Review on the Na-Flux Method
241 104~5 cm–2 LPE – GaN
106 cm–2
Interface MOCVD – GaN
108~9 cm–2
Fig. 11.5 Reduction of dislocations during the LPE growth of GaN on a MOCVD-GaN substrate c 2008, Elsevier Publishers.) by the Na-flux method. [12] (Reprinted with permission.
11.4.3 Solubility and Growth Rate The Solubility of GaN is an essential parameter that would govern the crystal growth process. Kawamura et al. have investigated the solubility and reported that only 0.036 mol% of GaN can be dissolved in a flux of 27 mol% Ga and 73 mol% Na, while keeping the process running for 200 h at T D 973 K [12]. Adding Na to a Ga melt obviously improves the solubility of GaN and also allows applying lower pressure conditions. By comparison, Madar et al. reported a solubility of GaN in the pure Ga melt of only 0.008 mol% at T D 1; 473 K and p D 800 MPa [29] and Karpinski et al. have observed a rise in the solubility to about 0.16 mol% at T D 1; 773 K and p D 1; 600 MPa [30]. The solubility of GaN can be expressed in terms of system pressure and temperature. A minimum pressure is required above which crystal growth of GaN is commencing. Above T D 900 K, Na is clearly promoting the dissolution of N in the Ga–Na melt. As a result, about 0.12 at% N dissolves at T D 1; 073 K and p D 0:75 MPa in the flux of 27 mol% Ga and 73 mol% Na [9]. What is more, it was reported that the minimum pressure to grow GaN in the Ga–Na melt decreases until T D 1;023 K and after that increases. This phenomenon is explained in terms of the stronger rise of the solubility of GaN which exceeds that of N [9]. The growth rate depends on the thermal gradient between the substrate and the Na flux. Thermal convection inside the flux container is formed such way. Kawamura et al. recently reported a growth rate as large as 20 m h1 in the direction of h0001i [20]. They furthermore found that a dwell time of about 20 h from starting the experiment until first growth appears is needed, which is mainly caused by the slow saturation of the Na flux with nitrogen until reaching a level of supersaturation that initiates the growth process on the substrate. The control over parasitic nucleation is also a big challenge in the Na-flux method. It w as observed that they preferentially crystallize at the interface of the
242
D. Ehrentraut and E. Meissner
Table 11.1 Properties of GaN crystals grown by the Na-flux method Self-seeded GaN
GaN grown on substrate /
Impurities
a D 3:1903.3/ a D 3:1896.1/ a D‹ Å [18] Å [4] c D 5:1877 Å [5] c D 5:1864.6/ c D 5:1854.2/ Å [18] Å [4] / / 25 arcsec (0004) 40–60 arcsec (0002) [9] / / /
Resistivity
/
/
Carrier / concentration Mobility /
/
Lattice parameter
XRC FWHM
/
/
Na D 4:2 1014 cm3 [2] C D 2 1017 cm3 [24] /
0:04 cm (T D 110 350 K) [5] 1–2 1018 cm3 / (RT), n-type [5] 100 cm2 V1 s1 / (RT) [5]
liquid Na flux to the gas atmosphere, thus consuming a large fraction of the free N radicals needed to form the GaN on the substrate crystal. However, adding carbon (C) as additive would completely suppress the formation of parasitic GaN crystals and therefore support the growth of GaN on the substrate crystal [20].
11.5 Properties of GaN There is only very limited data available on the structural and physical properties of GaN produced by the Na-flux method. Table 11.1 has listed the lattice parameter a and c, XRC FWHM, concentration of the impurities Na and C, electrical resistivity, carrier concentration, and carrier mobility as far as available. Skromme at al. have investigated some optical properties of colorless prismatic GaN crystals grown by Yamane [31] by photoluminescence (PL), reflectance, and micro-Raman scattering. Raman scattering revealed an A1 .LO/ phonon mode at 739 cm1 indicating a free electron concentration around 2–3 1017 cm3 for 3–4 mm large platelets. Smaller platelets grown in a pyrolytic BN (pBN, purity > 99:9999%) crucible showed the A1 (LO) phonon mode at 733 cm1 , implying a free electron concentration in the mid 1016 cm3 . PL measurements disclosed a residual donor species with the binding energy of 33.6 meV, probably because of ON . Residual zinc (Zn) acceptors have also been evidenced peaking around 2.9 eV with a broader peak in the low-temperature .T D 1:7 eV/ spectrum. Other
11
A Brief Review on the Na-Flux Method
243
impurities evidenced were Si and Mg, both at low levels. The yellow band around 2.2 eV, and usually associated with gallium vacancies, was rarely observed. In the low-temperature .T D 1:7 eV/ PL measurement, the neutral donor-bound exciton .Do X/ emission was high in intensity with a small FWHM of only 2.2 meV. The authors concluded that the material was of excellent optical quality.
11.6 Industrialization Potential for the Na-Flux Method The Na-flux method has proven the possibility to grow GaN crystals under relatively mild conditions with respect to pressure and temperature, i.e., p 10 MPa and 600 T 900 ı C, respectively. The growth rate of up to 30 m h1 seems reasonable for industrial production purposes, particularly in the light of the intrinsically lower growth rates for growth techniques from the solution; refer also to Chaps. 7–10 and 12 in this book. The structural quality certainly can further be improved, especially in the view of single crystallinity and related grain boundaries for the wafer-sized area. A critical issue in terms of industrialization of the Na-flux technology, essentially for all technologies considered to produce GaN crystals for GaN wafer fabrication, is the feasibility for upscaling in order to achieve a high crystal throughput, which would positively affect the cost per epi-ready GaN substrate. The ammonothermal technique seems to be most promising in this regard. In fact, the cost for an epiready GaN substrate is an increasingly important issue considered by device makers as it covers the largest fraction in the cost structure for a mass-produced GaN-based device. For a vast sector of GaN devices, the cost for an epi-ready GaN substrate has to compete with that for the widely used sapphire-based templates that can achieve high-quality GaN devices through sophisticated template technology in combination with epitaxial lateral overgrowth (ELO).
References 1. Department of Electrical Engineering, Osaka University, Japan 2. F. Kawamura, H. Umeda, M. Morishita, M. Kawahara, M. Yoshimura, Y. Mori, T. Sasaki, Y. Kitaoka, Jpn. J. Appl. Phys. 45, L1136 (2006) 3. H. Yamane, F.J. DiSalvo, J. Solid State Chem. 119, 375 (1995) 4. H. Yamane, M. Shimada, S.J. Clarke, F.J. DiSalvo, Chem. Mater. 9, 413 (1997) 5. M. Aoki, H. Yamane, M. Shimada, T. Sekiguchi, T. Hanada, T. Yao, S. Sarayama, F.J. DiSalvo, J. Cryst. Growth 218, 7 (2000) 6. M. Aoki, H. Yamane, M. Shimada, S. Sarayama, F.J. DiSalvo, Cryst. Growth Des. 2, 119 (2001) 7. H. Yamane, D. Kinno, M. Shimada, T. Sekiguchi, F.J. DiSalvo, J. Mater. Sci. 35, 801 (2000) 8. T. Iwahashi, F. Kawamura, M. Morishita, Y. Kai, M. Yoshimura, T. Sasaki, J. Cryst. Growth 253, 1 (2003) 9. F. Kawamura, M. Morishita, K. Omae, M. Yoshimura, Y. Mori, T. Sasaki, J. Mater. Sci. Mater. Electron. 16, 29 (2005)
244
D. Ehrentraut and E. Meissner
10. M. Aoki, H. Yamane, M. Shimada, S. Sarayama, H. Iwata, F.J. DiSalvo, J. Cryst. Growth 266, 461 (2004) 11. T. Yamada, H. Yamane, Y. Yao, M. Yokoyama, T. Sekiguchi, Mater. Res. Bull. 44, 594 (2009) 12. T. Sasaki, Y. Mori, F. Kawamura, M. Yoshimura, Y. Kitaoka, J. Cryst. Growth 310, 1288 (2008) 13. Y. Mori, M. Yano, M. Okamoto, T. Sasaki, H. Yamane, Bull. Solid State Phys. Appl. 4, 198 (1998) (In Japanese) 14. M. Yano, M. Okamoto, Y.K. Yap, M. Yoshimura, Y. Mori, T. Sasaki, Jpn. J. Appl. Phys. 38, L1121 (1999) 15. F. Kawamura, M. Morishita, K. Omae, M. Yoshimura, Y. Mori, T. Sasaki, Jpn. J. Appl. Phys. Part II 42, L879 (2003) 16. T. Iwahashi, Y. Kitaoka, M. Kawahara, F. Kawamura, M. Yoshimura, Y. Mori, T. Sasaki, R. Armitage, H. Hirayama, Jpn. J. Appl. Phys. 46, L103 (2007) 17. B.N. Feigelson, R.L. Henry, J. Cryst. Growth 281, 5 (2005) 18. Y. Song, W. Wang, W. Yuan, X. Wu, X. Chen, J. Cryst. Growth 247, 275 (2003) 19. G. Wang, J. Jian, W. Yuan, X. Chen, Cryst. Growth Des. 6, 1157 (2006) 20. F. Kawamura, M. Morishita, M. Tanpo, M. Imade, M. Yoshimura, Y. Kitaoka, Y. Mori, T. Sasaki, J. Cryst. Growth 310, 3946 (2008) 21. M. Aoki, H. Yamane, M. Shimada, S. Sarayama, H. Iwata, F.J. DiSalvo, Jpn. J. Appl. Phys. 42, 7272 (2003) 22. H.Q. Bao, H. Li, G. Wang, B. Song, W.J. Wang, X.L. Chen, J. Cryst. Growth 310, 2955 (2008) 23. T. Sekiguchi, H. Yamane, M. Aoki, T. Araki, M. Shimada, Sci. Technol. Adv. Mater. 3, 91 (2002) 24. F. Kawamura, M. Tanpo, Y. Kitano, M. Imade, M. Yoshimura, Y. Kitaoka, Y. Mori, T. Sasaki, 5th International Workshop on Bulk Nitride Semiconductors (IWBNS-V), Itaparica, Salvador, Bahia, Brazil (2007) 25. M. Aoki, H. Yamane, M. Shimada, S. Sarayama, H. Iwata, F.J. DiSalvo, Jpn. J. Appl. Phys. 42, 5445 (2003) 26. M. Aoki, H. Yamane, M. Shimada, T. Kajiwara, Mater. Res. Bull. 39, 827 (2004) 27. F. Kawamura, H. Umeda, M. Kawahara, M. Yoshimura, Y. Mori, T. Sasaki, H. Okado, K. Arakawa, H. Mori, Jpn. J. Appl. Phys. 45, 2528 (2006) 28. F. Kawamura, M. Imade, M. Yoshimura, Y. Mori, Y. Kitaoka, T. Sasaki, Proc. SPIE, 7216, 72160B-1 (2009) 29. R. Madar, G. Jacob, J. Hallais, R. Fruchart, J. Cryst. Growth 31, 197 (1975) 30. J. Karpinski, J. Jun, S. Porowski, J. Cryst. Growth 66, 1 (1984) 31. B.J. Skromme, K. Palle, C.D. Poweleit, H. Yamane, M. Aoki, F.J. DiSalvo, J. Cryst. Growth 246, 299 (2002)
Chapter 12
Low Pressure Solution Growth of Gallium Nitride E. Meissner, S. Hussy, and J. Friedrich
12.1 Introduction Because of the extremely high decomposition pressure of GaN at its melting point [1], it is technically unfeasible to grow GaN from the melt by means of classical crystal growth techniques, like Czochralski, Bridgman, or Vertical Gradient Freeze, commonly used for the production of industrially important materials like silicon, GaAs, and many more. At present there are several promising techniques under development to produce thick GaN layers or bulk crystals, which are described in detail in other chapters of this book. Among the different crystal growth processes, the solution growth methods do have theoretical, as well as the already demonstrated practical potential for the production of GaN with the lowest possible dislocation density. The theoretical expectations on the solution growth are based on the fact that such a process would operate close to the thermodynamic equilibrium compared to any of the gas-phase methods and at much lower absolute temperatures. This expectation was successfully proven by Porowski et al. [2] with their high pressure solution growth method (HPSG) resulting in GaN material with an extraordinary low dislocation density of 10–100 cm2 . The HPSG method is performed under very high nitrogen pressure ranging up to 17 kbar and temperatures between 1,300 and 1;600 ıC [2], which indicates the obvious drawback of the method – the high pressure which needs to be applied. High temperatures are necessary in order to achieve a reasonable solubility of nitrogen in the gallium melt. The solubility of nitrogen in liquid gallium metal is very low [3] compared to that in classical solution growth processes. GaN can also be grown under much lower pressures and temperatures if it is not pure Ga-melt but gallium with the addition of a solvent. The solvent is required to enhance the reactivity and solubility of nitrogen in the gallium solution. Then, one could either still use pure nitrogen gas with a certain over pressure of up to, e.g., 50 bar to deliver nitrogen to the system or one can apply other alternative chemical species containing atomic nitrogen in order to get the process pressure down and maintain nitrogen solubility at a reasonable level. A certain variety of alkali or earth alkali metals, their respective nitrides, or mixtures thereof were already investigated by several researchers. The corresponding method originating from that approach 245
246
E. Meissner et al.
is the so-called Na-flux technique, originally developed by Yamane et al. mainly between 1997 and 2000 [4, 5] and recently further investigated for the purpose of the design of a liquid phase epitaxy process (LPE) by Kawamura et al. [6]. The sodium flux method is based on the usage of a gallium–sodium melt made out of a mixture of the pure metals and NaN3 or through Na vapor generated in the system [7]. Other group one or group two elements from the periodic system were also tried as flux compositions, like lithium or lithium nitride [8, 9], lithium – potassium mixtures [10], or additions of certain amounts of calcium or carbon [11, 12]. The choice of the solvent is not easy, as it has to fulfill several criteria. In the ideal case, the solution additive should have a high solubility for nitrogen, but should not itself form a stable compound with nitrogen, which could compete with the formation of GaN under the given conditions, nor should it produce stable binary or ternary alloys, especially in the case of more than one element added to the solution. Finally, a handful of metallic elements from the periodic table remain, which could be promising candidates as additives for a GaN solution growth process, e.g., under reduced pressure. Systems with more than two elements other than gallium were not reported for the GaN crystal growth, since multicomponent systems are even more difficult to control in terms of solubility, thermodynamics, and reaction kinetics of possible phases formed. Logan and Thurmond [13] started already in 1972 with a very detailed investigation of the solution growth of GaN under a flowing gas atmosphere of ammonia and hydrogen, using a solution of Ga and Bi. They succeeded in growing epitaxial GaN layers on sapphire substrates. Elwell et al. [14], who had similar experimental conditions like Logan and Thurmond, used Bi and Sn as solvents and demonstrated again in 1984 that the growth of GaN from a solution is possible even under ambient pressure. Meissner et al. [15] and Hussy et al. [16] used other metal additives to the Ga melt, not applied so far by other research groups and managed to grow GaN under ammonia atmosphere from those solutions. The metals applied in their study do not have the same difficulties as sodium or potassium that have high vapor pressure. Their growth process was performed without any overpressure applied to the system. Table 12.1 shows a compilation of approaches, with the respective pressures, temperatures, and solvents, made for the solution growth of GaN (ordered after pressure). The sodium flux method still needs the application of certain overpressure and getting the pressure threshold down is unlikely, unless the chemical species delivering the nitrogen to the system is one other than the stable nitrogen molecule. Ammonia is the simplest choice to add the nitrogen into the growth process from the gas phase and adjust a condition close to the equilibrium curve of GaN, within the GaN stability field, thus preventing the GaN from decomposition during the process. Recently, over the last few years, a solution growth process for the growth of GaN under room pressure was developed and carefully investigated in detail at the Crystal Growth Department of the Fraunhofer Institute of Integrated Systems and Device Technology. The idea was to pick up the basic aims for the design of a crystal growth process for GaN, which could be of technical relevance because of simplicity, materials quality, scalability, and cost effectiveness. The result was the so-called low pressure solution growth method (LPSG), which is the subject of this chapter.
12
Low Pressure Solution Growth of Gallium Nitride
247
Table 12.1 Data available in the literature for experimental conditions and solvents used for the crystal growth of Gallium nitride from a solution Solvents
Atmosphere
Temperature Pressure Reference, Year [bar] Œı C
NaN3 Na C NaN3 Na K Li Na C Ca Na C Li Na Li Li3 N Bi Bi, Sn Bi Ge Ge, Sn, Au, Ag Ge, Sn, Bi, Au, Ag and mixtures thereof
N2 N2 N2 N2 N2 N2 N2 N2 C NH3 NH3 N2 H2 C NH3 H2 C NH3 N2 C NH3 H2 C NH3 H2 C NH3 H2 C N2 C NH3
600–800 650–840 750 750 800 800 800 800 300–750 740–800 850–1,050 870–990 900 930–990 960 900–1,050
110 120 50 70 50 30 15 10–20 1 1–2 1 1 1 1 1 1
Yamane [17], 1997 Yamane [4], 2000 Aoki [5], 2000 Yamane [18], 2000 Morishita [19], 2003 Kawamura [10], 2002 Morishita [19], 2003 Iwahashi [20], 2003 Barry [8], 1998 Song [9], 2003 Logan [13], 1972 Elwell [14], 1984 Klemenz [21], 2000 Hussy [16], 2008; this chapter Hussy [22], 2008; this chapter Hussy [23], 2008
Most of the studies mentioned above lack a thorough investigation of the solubility of nitrogen in the chosen system and detailed considerations on the formation reaction of GaN from the particular system. Therefore, important information for the design of a technically relevant production process based on solution growth technique is missing for the crystal growers. The mechanism of action of the solvents in the solution is not sufficiently clarified. The solvents influence not only the solubility of nitrogen, but also the kinetics of nucleation and crystallization of GaN in the Ga-solution. In turn, very relevant questions necessary to be addressed like control over the nucleation density, growth rates, crystallization mechanism and others, which are related to the basic understanding of the influence of the solvents on the formation reaction of GaN as a function of the process parameters could not be answered so far with data available from the literature. The thermodynamic aspects of the “pure system,” namely the GaN formation from a Ga melt under nitrogen atmosphere and the solubility of nitrogen in pure liquid Ga, were, however, quite well investigated for the HPSG process [3, 24].
12.2 Technology of Solution Growth Under Ambient Pressure, the LPSG Method The LPSG technique is a relatively simple method for growing GaN, as it can be performed in easy horizontal or vertical furnace setups. Basically, the LPSG technique is very flexible. It can be modified in order to aim the growth of GaN bulk crystals, as
248
E. Meissner et al.
well as it can be considered as a variant of multiwafer liquid phase epitaxy methods. Like with any other method, some important features must be addressed in order to gain a stable growth process and good quality material. A thorough study of this technique was performed during the last few years [15, 16, 25–27] and a detailed review of our work will be given in this chapter. The study answers at least some of the basic questions mentioned above. This chapter will also emphasize and discuss most recent achievements and the remaining challenges in LPSG GaN growth.
12.2.1 The Formation Reaction of GaN Under Ammonia Atmosphere The Fig. 12.1 shows the equilibrium pressures of N2 (upper region) and NH3 respectively over GaN as a function of temperature. The data were compiled from the literature and can be found in [28] and [29]. The stability curve was determined by thermal decomposition of GaN, the open and solid symbols represent the experimental data. The HPSG growth of GaN happens in the stability region of the stable nitride close to the equilibrium curve. The sodium flux technique works slightly away from the GaNGa C N2 phase boundary, but well situated within the GaN stability field, for the benefit of lower process pressure. Logan and Thurmond [13] and Elwell et al. [14] observed the decomposition or formation of GaN under ammonia atmosphere instead of pure nitrogen, in order to determine the boundary curve for the coexistence between GaN with NH3 . The LPSG growth of GaN and the LPE approaches from [13] and [21] were located in this respective lower region of the diagram in Fig. 12.1. Many efforts have been made in the last few years to control nucleation density, avoiding homogeneous nucleation in the solution and to establish a stable growth process by a careful choice of the solution composition, gas fluxes, process pressure, and temperature. Sun et al. [25] studied the formation reaction of GaN from a solution under ammonia atmosphere in order to retrieve the necessary data for the crystallization of GaN during the LPSG process. The work of Sun contains a systematic approach by means of high resolution thermogravimetry (TG); the details of the investigation can be found in the related publications [28, 30]. These considerations have the unquestionable advantage that an in situ technique was applied and that the measurements not only reflect a comparable situation to the LPSG growth process, but also to other solution growth techniques. The overall chemical reaction for the nitridation of gallium in an ammonia atmosphere can be simplified in the following expression: 2Ga C 2NH3 ! 2GaN C 3H2 During the TG experiments, the conversion of Ga from the solution to GaN crystals or epilayers was monitored as a weight signal as a function of reaction time, which compares to the ongoing crystal growth process. The formation of GaN from the
12
Low Pressure Solution Growth of Gallium Nitride
249
Fig. 12.1 Equilibrium pressure of N2 (upper region) and NH3 (in H2 ) (lower region) over GaN as a function of temperature. The solid symbols denote experimentally found stable GaN and the open ones denote decomposed GaN. For the high and low pressure solution growth methods the calculated equilibrium (partial) pressures are indicated by dashed and solid lines, respectively
solution is expressed as conversion ratio ˛ which is plotted vs. the reaction time. ˛ is defined as mt ˛ D Ga ; m0Ga where m0Ga is the initial mass of Ga and mtGa the mass of gallium which has reacted to form GaN until the time t. As described in [29], the conversion ratio ˛ can be related to the total weight change measured during the TG experiments as a function of process time. Generally it is assumed that the formation of GaN from the Ga-solution occurs by a nucleation step followed by subsequent growth of a nucleus into a crystal as discussed, e.g., by [31]. For the solution growth process like LPSG, there are in fact several stages for the formation of GaN, first characterized by an initial reaction upon the introduction of NH3 into the system, followed by the so-called induction period for the formation of nuclei. The reaction accelerates when the nuclei start to grow and pass into a period of crystal growth with a constant formation rate .d˛=dt D const:/, which compares to the finding of a constant growth rate of GaN with time during the LPSG process if a stable growth condition is maintained (as discussed later in this chapter). The induction time tind is an important parameter, and although it is not a fundamental property for the system, it can be used to describe the nucleation rate J , which can be written as, 1 J DK tind with K being a constant.
250
E. Meissner et al.
However, some assumptions are needed for this formulation, which can be found in a thorough study and interpretation of the induction period of crystallization processes in the paper from Söhnel and Mullin [32] and for the LPSG process in [29]. An empirical formula was expressed by He [33], which can be applied to describe the influence of temperature on the induction time and hence can give an idea about nucleation. Generally it is found that a high supersaturation leads to a high reaction rate and a short induction period, and in turn a high nucleation rate in the solution is expected. Thus, from the observation of the induction time, one can draw conclusions as to the supersaturation present in the growth system, since the nucleation rate J is proportional to the supersaturation S by the following formulation, where KN and n are constants: J D KN S n : Sun et al. [29] demonstrated how this relation can be used to retrieve the equilibrium partial pressure of ammonia for the LPSG growth process, by the observation of the length of the induction period as a function of ammonia partial pressure, p.NH3 /. It is essential to know the equilibrium ammonia partial pressure, since this parameter must be carefully controlled in order to achieve a stable growth condition, as will be explained later in this chapter in detail. The study of the induction period as a function of growth temperature generated valuable knowledge about the activation energy for homogeneous nucleation in the system. It was found that the activation energy for the reaction of Ga under ammonia remained the same if the ammonia partial pressure was changed [29]. If the activation energy for the reaction is not a function of p.NH3 /, the underlying reaction mechanism is the same for all p.NH3 / applied. However, the activation energy of the reaction was found to be different if the GaN was crystallized from a solution rather than from the pure Ga melt, which gives interesting implications as to the action of the solvents in the growth solution. The solvents seem to act kinetically as well as in terms of enhancement of solubility for nitrogen. In view of the growth rates, it would be important to define the rate limiting steps in the formation reaction of GaN. However, the determination of rate limiting steps is not easy. The soluble species is unknown and is expected to differ for any particular composition of a solution. Because the nitrogen as a component is supplied from the gas phase, the first interface that could define the reaction rates is the gas–liquid interface. Here the nitrogen compound must be dissociated and the atomic nitrogen must be dissolved in the melt or a soluble species must be formed. Krukowski et al. [34] examined the dissociation of molecular nitrogen on the surface of a gallium melt in the HPSG process by theoretical means and also Yamane et al. [35] described the dissociation of molecular nitrogen and the formation of a complex species which solves and contains the nitrogen. In the later case, sodium solvent plays an important role. The knowledge of the kind of soluble species would be a key issue in order to describe and understand the formation of GaN from a solution and to actively handle this parameter in view of the growth rates. But the soluble species remains subject to speculations or assumptions in theoretical considerations, as it is not easily accessible by any experimental technique. The approach made by Sun et al. [30] resulted in the cognition that the kinetics at the gas liquid interface plays
12
Low Pressure Solution Growth of Gallium Nitride
251
an important role and represents a first trial of an experimental proof. The scenario of processes happening at the second rate limiting interface, the liquid–solid interface is not easily amenable by experimental means either; it consequently depends on the situation manifested in the solution “behind” the kinetics at the melt surface and may be even more difficult to investigate. Thus, the role of the solvents added to the solution growth is not fully clarified up to now, but it is definitely a combination of the enhancement of solubility for nitrogen in the Ga-melt and a kinetic influence on the key reactions at the respective interfaces.
12.2.2 Solubility of Nitrogen in Gallium-Metal Solutions The knowledge of the total concentration of nitrogen present in the Ga-melt is of great interest, in order to be able to quantitatively express a supersaturation state for the system and define the expected Ostwald–Miers region for stable growth. Moreover, if known, it would be possible to choose the most effective solvent with regard to nitrogen solubility. However, the access to the solubility by direct measurement is not easy, due to the very low values expected for the solubility of nitrogen in gallium or other metals. Thus, most of the solubility data were produced by ex situ analysis after quenching the sample [3,11,13,36] and the accuracy could suffer from desorption of N upon cooling the specimens. But, even at the highest available temperatures of 1;687 ıC, the solubility was not higher than 1 at%. In order to gain data for the solubility which would represent the condition in the low pressure solution growth under ammonia atmosphere and at room pressure, we made the approach to measure the solubility in-situ by means of thermogravimetry [28]. We used the same experimental setup as for the study of the reaction kinetics, as described above and the experimental procedure was very similar, but much more delicate in terms of artifacts and the ammono partial pressures are adapted to the equilibrium condition. Any kind of disturbing reaction, despite the dissolution of nitrogen to the melt, has to be avoided, like the formation of GaN in the solution, oxidation, and formation of GaN on the surface. But also the weight change from evaporation of Ga from the sample will interfere with the weight signal arising from the dissolution of nitrogen in the solution. All those parasitic reactions must be carefully avoided or at least minimized or quantitatively controlled such that a correction is possible. As the actual soluble species is subject to speculations a simple model for the reactions taking place in that system is formulated, assuming the dissolution of atomic nitrogen into the melt. This can be justified by the consideration of the weight signal resulting from the addition of nitrogen atoms only, which would be independent of the soluble species itself. According to Henry’s law for diluted systems and if the ammonia partial pressure is kept below the equilibrium partial pressure, one can treat the NH3 H2 equilibrium thermodynamically and the amount of dissolved nitrogen or solubility of nitrogen is expected to be a function of p.NH3 /. Figure 12.2 depicts the schematic illustration of the process of adsorption of nitrogen from an ammonia containing atmosphere to the solution in the LPSG process.
252
E. Meissner et al.
NH3
process gas atmosphere
(1)
NH3,ad Nad + 3Had (2) (3) [N] (4) +Ga=GaN Fig. 12.2 Schematic illustration of the processes of absorption and desorption of nitrogen and the formation of GaN in a sequence of steps (numbers 1–4) (after [28])
The interaction of NH3 and the Ga solution includes the following steps according to the figure above: (1) NH3 adsorption to the free solution surface (2) Dissociation of the adsorbed NH3 molecules and dissolution of the atomic nitrogen into the melt (3) Transportation of the nitrogen atoms through the bulk solution (4) Formation of GaN from the solution if the solubility limit is reached A detailed discussion of the model may be found elsewhere [28]. Figure 12.3 shows as an example the graph of the corrected weight change as a function of time for a Ga–Sn solution under an ammonia–hydrogen atmosphere. The region ’ represents the deoxidation of the sample under ArH2 flow, followed by the dissolution of nitrogen of the ammonia containing atmosphere until equilibrium is attained (sector “) and the nitrogen desorption process in the region ” after the NH3 - flow was stopped. The curve must be corrected for the gallium evaporation taking place at all stages of the reaction. The solubility was calculated from the total weight gain in the equilibrium state. Table 12.2 combines exemplary solubility data from this study and data existing in the literature. It can be seen that a solubility that can be achieved with the LPSG method is comparable to those values reported for high pressures and high temperatures. This means that the solvent indeed enhances the solubility of nitrogen. In comparison to the modeled solubility of nitrogen in pure gallium under the same conditions, the solvents enhance the solubility by roughly three orders of magnitude. However, care has to be taken when comparing solubility data at different conditions as the gas solubility is a function of temperature and pressure. For many dilute solutions of gases in metal melts, the relation between lgXN (with XN D solubility mole fraction) and the inverse temperature is linear at a given pressure. This indicates that the changes in enthalpy and entropy of the dissolution of nitrogen in those systems is virtually independent of temperature. For and the assumption that the solubility of N is very XN D 0:11 exp 12 28;745 T
12
Low Pressure Solution Growth of Gallium Nitride
253
0.8 0.6 0.4
b
0.2
g
a
Dm (mg)
0.0 Ar+H2
–0.2 –0.4 –0.6
Ar+H2+NH3
Ar+H2
(1) Ga evaporation (1) Ga evaporation (2) Nitrogen incorporation (2) Deoxidization (3) Nitrogen transportation
(1) Ga evaporation (2) Desorption
–0.8 –1.0
900°C, Ga (28 at%)+Sn (72 at%) Ar=60ml/min; H2=40ml/min; NH3=0.075ml/min PNH =0.75 mbar
–1.2
3
–1.4 200
800
1400
2000
2600
3200
t (minutes) Fig. 12.3 Plot of weight change vs. time for a Ga–Sn solution. The curve is corrected by the subtraction of the Ga evaporation as well as by an offset (redrawn after [28]) Table 12.2 Comparison of nitrogen solubility data for different conditions present in GaN solution growth Study
T Œı C
P [bar]
Solution composition
Atmosphere
Solubility mole fraction N2 ; XN
Sun 2006 [28] Sun 2006 [28] Sun 2006 [28] Logan 1972 [13] Morishita 2005 [11] Madar 1975 [36] Grzegory 2002 [3]
900 930 950 1;150 900 1;200 1;520
1 1 1 1 42 8;000 11;000
Ga C Sn Ga C Ge Ga C Ge Ga Ga C Na Ga Ga
Ar C H2 C NH3 Ar C H2 C NH3 Ar C H2 C NH3 H2 C NH3 N2 N2 N2
6 104 8 104 8 104 3 105 104 1 103 3 103
small and therefore, Henry’s law is obeyed all gases can be treated as ideal. Using existing data in the literature and the solubility measured by the TG experiments, one can express the Gibbs free energy change as a function of temperature as shown in Fig. 12.4a and subsequently calculate the temperature dependence of the nitrogen solubility. The run of the curve in Fig. 12.4a shows that G for the dissolution of N into Ga is positive in a wide temperature range and implies that the nitrogen solubility will increase with temperature under NH3 atmosphere. Figure 12.4b depicts the respective comparison of the TG data gained by Sun, with the solubility data for pure gallium systems under high N2 pressure determined by Grzegory [3] and Logan [13] for pure gallium under ammonia atmosphere. For details of the calculation see Sun [28]. Overall, it could be experimentally shown by direct in situ measurements that the nitrogen solubility is indeed enhanced by the addition of suitable solvents. The
254
E. Meissner et al. T (°C)
b
1650 1500
200
0.01
150
1E-3
XN(mole fraction)
ΔG (kJ/mol)
a
100
ΔG1
50
0
–50 800
1400
T(K)
1600
1800
2000
1050
900
1E-5 Grzegory (Pure Ga+N2 at high pressure)
Nitrogen dissolution in Ga melt under NH3 atmosphere
1200
1200
1E-4
1E-6
1000
1350
1E-7 5.0
Logan (Pure Ga, NH3+H2 at 1 bar) Calculation for pure Ga + NH3 at 1bar This study (Ga+Solvent, NH3+H2+Ar, at 1 bar) Detection limit for TG measurements
5.5
6.0
6.5
7.0
7.5
8.0
8.5
1/T×10000(1/K)
Fig. 12.4 (left) Calculated Gibbs free energy change of nitrogen dissolution in a Ga melt under ammonia atmosphere as a function of temperature (right) Temperature dependent nitrogen solubility for different systems
nucleation as well as the growth rates will depend on temperature and ammonia partial pressure. Moreover, the solvent acts kinetically as was discussed in the previous chapter. The situation in the case of LPSG growth is therefore very much comparable to the high pressure solution growth but with much favorable process conditions like flowing gases and low temperatures. The TG method was also found to be a suitable way for studying the kinetics of the reactions involved and defining rate limiting steps as well as selecting the best solvents and best composition for the growth solution.
12.2.3 Growth Setup, Process, and Basic Challenges Different growth setups, vertical [21] as well as horizontal, [13, 14] were used by the respective working groups. Only Logan and Thurmond used a thermal gradient in the melt for most of experiments, claiming that it enhances the growth rate [13], while Elwell et al. found that “the temperature gradient transport is not a significant effect in crystal growth of GaN from the solution” [14]. Most of the growth runs reported in the literature have the substrates fixed or floating in the melt during the whole process [13, 14, 16], which has some pros (simple setup), but also some negative effects (like thermal etching or homogeneous nucleation, see later in this chapter). The furnaces discussed in the literature were resistively heated up to temperatures of 1;150 ıC and quartz glass was used as main furnace tubing material. Only small samples were processed in the furnaces of Logan et al. [13], Elwell et al. [14], or Klemenz et al. [21], whereas recent studies in our laboratory developed the method even further leading to growth furnaces capable of processing several 2 in. substrates in one run and demonstrating the first 3 in. GaN template grown from the solution [16]. However, one has to cope with some basic challenges with this method in order to attain stable growth conditions and a high quality material, which will be discussed in the following section.
12
Low Pressure Solution Growth of Gallium Nitride
255
Formation of a Crust on the Melt Surface The formation of a crust of nitrides or oxynitrides on top of the melt can principally impede the reaction of ammonia with the solution and hence reduces or stops crystal growth. Logan and Thurmond [13] reported that the crust will start to form when process temperatures were set below 950 ı C. As a measure to prevent this, Logan [13] suggested to agitate the melt by mechanical vibration or a temperature gradient, whereas our recent studies indicated that this effect is more likely related to the purity of the gas phase. If oxygen and moisture is kept out of the system, it is possible to perform LPSG growth at temperatures as low as 800 ıC without the formation of a surface blocking layer. This also points out that the gas purity in general must be a major concern and the application of point-of-use purifiers for the process gases is advisable.
The Creeping Phenomenon The upwards flow of the Ga containing melts along the inner wall of the crucible is a more “striking feature.” This effect was already observed by Logan [13] who mentioned that the flow of the solution occurs if the ammonia partial pressure is set higher than the respective equilibrium partial pressure. Under such conditions, a thin GaN layer is probably formed on the inner crucible wall right above the melt surface, which is then wetted by the gallium solution. The thin layer in the vicinity of the new melt surface forms again, is subsequently wetted, and thus kicks off a fast progressing up flow of the melt along the crucible wall. The higher the ammonia partial pressure (supersaturation), the stronger is this effect. According to our studies, there are three ways to overcome or at least reduce the creeping problem. First, the supersaturation must be kept well within the Ostwald–Miers region, which will help to avoid parasitic formation of GaN and also the melt creeping phenomenon. Secondly, a short stop of the growth process after some hours for a certain time by reducing the supersaturation and/or the temperature can help to pin the up flowing melt at the position reached until that stage. Lastly the choice of the solvent has a strong influence on this inconvenient behavior of the melts. A solvent with a high vapor pressure compared to gallium, like sodium, potassium or others that are equally evaporating, will have a strong disadvantageous effect as the gallium concentration increases locally and therewith the supersaturation accelerating the creeping, whereas solvents with low vapor pressures will help to circumvent this phenomenon.
Materials in the Hot Zone and Dissociation of Ammonia Attention has to be turned also on the materials present in the interior of the furnace in order to avoid uncontrolled or unwanted dissociation of ammonia on other surfaces than the solution and corrosion of the furnace interior. But, this is not a too serious issue as long as no graphite containing parts are used inside the hot zone.
256
E. Meissner et al.
The presence of graphite in the hot part of the furnace can be very deleterious to the growth process [23]. However, the situation can be different if the reaction gases are changed. The same debate holds concerning the choice of the crucible material. The most stable material in direct contact to the melt is pyrolytic boron nitride. Other materials like quartz glass, graphite, glassy carbon, silicon nitride, or aluminum nitride can be corroded by the melt and thus cause a pollution of the solution with unwanted impurities, beside the effect of uncontrolled dissociation of the ammonia as discussed above. Also the concentration of hydrogen in the carrier gas can have a noticeable influence on the thermal decomposition reaction of ammonia on the inner surfaces of the furnace [37] due to a passivation effect. This effect develops stronger with increasing hydrogen concentration in the atmosphere [23, 37]. A very detailed discussion on how to overcome or circumvent all this technological obstacles is given in Hussy [23]. In contrast to standard LPE techniques, the supersaturation in the LPSG process is not attained by cooling of the solution, but by applying a higher ammonia partial pressure than the equilibrium one [14, 16, 22]. In turn, the careful control of the partial pressure of the gases used in the process is a very important parameter. Besides that, the temperature and composition of the solution play the major role as described before. Regarding a complicated gas phase process, the handling of process parameters in LPSG is comparably simple and the most important points will be discussed in the following section of this chapter.
12.2.4 The Influence of Process Parameters on the Epitaxial and Parasitic Growth of GaN Ammonia Partial Pressure As described before in this chapter, the nitrogen concentration in the growth solution is a function of the ammonia partial pressure. Therefore, the equilibrium nitrogen concentration in the melt (no growth of GaN, but also no dissociation of GaN) corresponds to a certain ammonia partial pressure peq .T /. The value ppara .T / is the maximum ammonia partial pressure which corresponds to a maximum supersaturation of the solution before homogenous nucleation starts under the given conditions. The homogeneous nucleation is considered parasitical as long as the stable growth on a seed is wanted. The difference between this supersaturation concentration and the equilibrium concentration (D solubility) corresponds to the so-called Ostwald–Miers region. The effect of the ammonia partial pressures on the formation or dissociation of GaN in hydrogen atmosphere was investigated in a temperature range from 870 to 1;150 ıC using different approaches for either pure gallium [14, 38] or gallium–germanium solutions [16]. Limiting values of p.NH3 / for the dissociation of GaN .pdiss / and the formation of epitaxial, respectively parasitic GaN were found. Therewith it could be considered that any GaN that was grown on the seed
12
Low Pressure Solution Growth of Gallium Nitride
257
Fig. 12.5 Limits of ammonia partial pressure in hydrogen over GaN(s)–Ga(l) for the formation of epitaxial and parasitic .pepi ; ppara / GaN respectively; the dissociation of GaN .pdiss / in the temperature range from 870 to 1;150 ı C. The calculated equilibrium ammonia partial pressure [38] is plotted as a solid line
was epitaxial GaN, .pepi / and any GaN that was nucleated in the solution elsewhere was parasitic GaN, .ppara /. In Fig. 12.5 the corresponding ammonia partial pressures .pdiss ; pepi ; ppara / are plotted as a function of temperature for the investigated temperature range. Thurmond et al. calculated the equilibrium ammonia partial pressure [38] which is increasing with temperature (see Fig. 12.5, solid line). The corresponding experimental results .pdiss ; pepi / show the same tendency [16, 38]; however, they can scatter widely especially above 1;000 ı C. For the formation of (parasitic) GaN crystallites in the solution, a higher ammonia partial pressure is necessary than for the growth of an epitaxial layer or for preventing the GaN from decomposition. This supersaturation is necessary to overcome a kinetic barrier. As the height of the kinetic barrier is temperature dependent itself (most likely decreasing with temperature) different temperature dependence for ppara compared to pepi or pdiss is anticipated and indeed the experimental results show this correlation. In Fig. 12.5 a region exists between pdiss and ppara and respective pepi and ppara , which corresponds to the classical Ostwald–Miers region, bordered by the solubility line (i.e., peq ) and the maximum supersaturation without homogeneous nucleation (super solubility line). A stable supersaturation exists within the Ostwald–Miers space. The corresponding sets of parameters (T and p.NH3 /) are favorable for the growth of seeded layers or crystals while the formation of parasitic GaN is suppressed.
258
E. Meissner et al.
It was demonstrated that the supersaturation within the Ostwald–Miers region .pepi .T / < p.NH3 / < ppara .T // leads to a quite low epitaxial growth rate, nearly independent of the growth time [16]. This means that thicker GaN epitaxial layers can be principially grown just by increasing the process time, which will however result in quite long growth runs. In contrast, the application of a supersaturation above the upper limit of the Ostwald–Miers space .p.NH3 / > ppara .T // results in a strongly time dependent epitaxial growth rate. The growth rate is high at the beginning of the process and decreases drastically with process time [16]. The formation and growth of parasitical GaN crystallites which consume reasonable amounts of the nitrogen present in the solution is the obvious reason for this. A similar observation was made for LPE of GaN by the Na-flux method under high supersaturation [39].
Influence of the Carrier Gas The influence of the carrier gas is manifold. Most experimental studies published up to now used hydrogen as a carrier gas [13–16] whereas [21] used nitrogen. First of all hydrogen itself has a high solubility in the gallium solution [28]. It is several orders of magnitude higher than the solubility of nitrogen introduced by ammonia under the same conditions. The influence of the composition of the carrier gas on the equilibrium ammonia partial pressure was experimentally investigated in [23]. At a given temperature of 960 ıC pepi is 1.35 mbar for 75–100% H2 in the carrier gas, dropping to 0.45 mbar for 30% H2 . As no stable epitaxial growth could be obtained at 960 ı C for lower hydrogen concentrations than 30%, pepi could not be investigated experimentally since the creeping phenomenon becomes severe under this growth condition. We found that technological difficulties like the above mentioned up flow of the solution on the crucible wall show an increasing tendency to occur with lower fractions of hydrogen in the carrier gas. In parallel, it was found that it is more difficult to attain stable epitaxial growth when lower amounts of hydrogen are present in the process gas. For much lower temperatures .850 ıC/ Klemenz et al. [21] achieved stable epitaxial growth with pure nitrogen as carrier gas and an ammonia partial pressure as high as 250 mbar. However, there was no description of up flow of the solution under such process conditions and using the same conditions in our furnaces did not result in epitaxial growth of GaN on the seed. But, according to our investigations this could also display the finding that the decomposition of the ammonia is not independent of the flow rates of the gases, especially in the case of low total flow rates and high nitrogen content. On the other hand, in the case of pure hydrogen atmosphere, the epitaxial as well as the parasitic growth rates were found to be independent of gas flow rates [23]. This might also be a result of the surface passivation effect as described before. With low flow rates, the time the gases remain in the hot zone mainly determines the dissociation of ammonia. If pure hydrogen is applied as a carrier gas, the passivation effect is dominant and the time for passing the gases through the hot zone does not play a major role. The thermal dissociation is then independent of flow rates, especially in the case of a high total gas flow.
12
Low Pressure Solution Growth of Gallium Nitride
259
Composition of the Solution The most crucial process parameter is the type and amount of the solvent added to gallium. As described before in this chapter, nitrogen solubility, as well as the kinetics of rate limiting steps, is strongly determined by the solvent and its concentration. Logan et al. [13] investigated the influence of the additive bismuth on the epitaxial GaN growth. They reported that as the ammonia partial pressure, necessary to initiate epitaxial growth, becomes higher, more bismuth is added to the solution. This is very pronounced for very high bismuth concentrations (>75 at%). But for solution additives with a high vapor pressure such as bismuth and lead, the exact values for pepi and ppara cannot be extracted easily from long-term experiments since the solution composition will change with time due to the evaporation loss of the components, resulting in nonstable growth conditions. Therefore, additives with a low vapor pressure should be favored, such as tin and germanium. Thus, instead of pepi and ppara , the ammonia partial pressure pgrowth , which leads to relatively stable epitaxial growth (up to 36 h growth time) was plotted against the initial solution composition, in the case of lead and bismuth as solvents. The results for the different solution additives, bismuth [13], germanium [23], tin, and lead are compiled in Fig. 12.6. The relationship between the concentration of the additives and the equilibrium ammonia partial pressure is qualitatively the same for all solvents shown in Fig. 12.6, whereas the effect of additives on the formation of parasitic GaN
ammonia partial pressure [bar]
0·1 pgrowth(NH3)
GaxBi1 x
1000 °C (Logan, 1972)
pgrowth(NH3)
GaxBi1 x
1000 °C
pgrowth(NH3)
GaxPb1 x 1000 °C
ppara(NH3)
GaxSn1 x
960 °C
ppara(NH3)
GaxGe1 x
960 °C (Hussy, 2008)
pepi(NH3)
GaxGe1 x
960 °C (Hussy, 2008)
0·01
1E-3 0·0
0·2
0·4 0·6 gallium atom fraction xGa
0·8
1·0 (Ga)
Fig. 12.6 Dependence of ammonia partial pressures .pepi ; ppara ; pgrowth / on the additive concen tration. For the solvents Bi and Pb the corresponding ammonia partial pressures .pgrowth / are given, which lead to a stable epitaxial growth in experiments with small process time (e.g., 20 h)
260
E. Meissner et al.
crystallites is controversially discussed in the literature. Logan found that the addition of bismuth suppresses the formation of parasitic GaN, while in Elwells experiments additions of both bismuth and tin increased the nucleation rate of GaN [13, 14]. The thermogravimetric measurements performed by Sun can be taken for clarification [28]. The additives germanium and tin help to suppress a high nucleation rate for GaN as they extend the so-called induction period for low supersaturations (time interval until first detectable weight gain related to GaN formation). This was proven by our recent studies suggesting that the concentration of the germanium has an effect on the width of the Ostwald–Miers region [23]. It was found, that small additions of germanium (10–40 at% Ge) lead to the suppression of the formation of parasitic crystallites and thereby enhance the epitaxial growth, but no stable epitaxial growth could be established for high additive concentrations (>75 at% Ge). This can also be explained by the study of Sun [25] who found that the nucleation rate is a function of ammonia partial pressure and can be different for any individual solution composition. Changing the composition of the solution does not only mean different equilibrium conditions and different widths of the Ostwald–Miers region but also affects the total density of the solution and the wetting angle of GaN. It could be shown that the solution density strongly influences the morphology of the as-grown GaN layers [22] and the formation of macroscopic defects resulting from parasitically grown GaN crystallites. If the formation of parasitic GaN can not be completely eliminated or is accepted for the benefit of higher growth rates, the solution density is a key issue to control as otherwise crystallites formed in the solution will drop on the growth interface and cause macroscopic defects. This needs special consideration since the LPSG process is a bottom seeded technique because of the fact that the nitrogen is supplied from the gas phase.
Temperature–Time Program Another important factor is the temperature – time program of the process, especially the cooling procedure. If the seed is kept in the solution for the whole time until the furnace is cold, it must be taken care that the solution does not get undersaturated with respect to the equilibrium partial pressure. This can happen if the ammonia flow is reduced too strongly at an early stage. Then, the grown GaN surface will be etched non-uniformly and the as-grown surface morphology will be affected negatively. On the other hand, if the ammonia partial pressure is higher than the equilibrium one during cooling, the melt may excessively supersaturate leading to parasitical GaN formation in the volume of the melt and subsequent development of macroscopic defects. To avoid both of the two scenarios, the ammonia partial pressure must be carefully adjusted according to the equilibrium partial pressure during the cooling procedure.
12
Low Pressure Solution Growth of Gallium Nitride
261
12.3 Evolution of Structure and Morphology of the GaN Layers 12.3.1 Importance of the Initial Growth Stage The growth of GaN on a GaN seed surface is considered homoepitxial. Therefore, it is expected, because supersaturation is low, that the growth should proceed by a layer by layer growth mechanism or by step flow along an off-cut direction, because the sapphire substrate with a MOCVD GaN layer applied as a seed, generally has an off orientation toward the c-direction of 0:25ı . However, it was observed [16] that a pronounced as-grown surface morphology is present after LPSG growth, characterized by hexagonal shaped features like terraces, macrosteps, or hillocks of various heights and densities. The roughness of the surface of the LPSG layers can vary strongly depending on the additives used for the solution growth process, the conditions of the growth process itself, and in turn, on the nucleation regime occurring in the very beginning of growth. Initial Growth of the LPSG Layer, Growth Stages The growth starts with the back etching of the seed, then 3D facetted islands nucleate. They are visible in Fig. 12.7 by faint contrast in secondary electron imaging. This contrast might develop from different doping levels of the material along the growth direction. It is not per se clear why the islands form. Respective experiments showed that their nucleation is suppressed if the seed applied in the growth process has a large off cut equal or larger than 1ı .
Fig. 12.7 SEM cross-section image of the interfaces between the seed layer and the LPSG GaN. The back etching region (grey line) is clearly visible. The nucleation islands developed on the seed surface are obvious from the darker contrast
262
E. Meissner et al.
The islands proceed with lateral growth until coalescence of the GaN layer is attained. The distribution of the islands over the seed surface is non-uniform. Their density and size depend on the growth temperature and the solvents applied. In order to clarify why the islands can form in a homoepitaxial growth case, it is important to investigate the nature of the islands in chemical and structural respect. Nature of the Growth Islands Investigations by spectroscopic and microscopic means were performed in the region where the growth islands occur. These investigations should help to clarify why the growth, unintentionally, always starts with the formation of nucleation islands. For the luminescence of GaN, it is generally expected that variations in dopant concentration scale reasonably with the luminescence intensity of the material if the luminescence from the donor acceptor pair (DAP) recombination is considered. Cathodoluminescence images taken at the wavelength of the near band edge emission, around 364 nm at room temperature, clearly revealed the presence of the growth islands at the seed–layer interface with different contrast. A complementary secondary ion mass spectroscopy measurement was carried out, monitoring the impurities in the GaN layers from the growth surface through the seed-layer interface and into the sapphire substrate. TEM microscopy mirrored the microstructure situation in the same region. Figure 12.8 shows a compilation of the corresponding observations. The impurity content of 1019 cm3 for silicon or oxygen characterizes a highly n-type doped GaN in this region; thus it’s counterintuitive that the luminescence intensity of the highly doped, island containing, layer of the LPSG GaN is as low as the seed layer which is a MOCVD GaN with a carrier density in the order of 1016 cm3 , since the luminescence intensity of the DAP usually increases with increasing n-doping level. However, it is visible from the corresponding TEM image that within the same region after the seed-layer interface there is a high recombination activity of the dislocation lines of the threading dislocations penetrating from the seed which are most likely the reason for the low luminescence intensity of this region because of nonradiative recombination. Figure 12.8 clearly shows that most of the reactions between the dislocations take place in the first micrometer of the LPSG layer after its interface with the seed. The related diagram of the impurity contents measured by SIMS does correspond to this situation. The highest measured concentrations of impurity elements are located in the island region. Figure 12.9 confirms that the dislocations act as non-radiative recombination centers destroying the luminescence in those particular parts of the layer, which explains the finding that the luminescence intensity is low although the island region is highly n-doped GaN. Though, in the first micrometer after the growth interface, the coincidence of effects is numerous and does not allow concluding too rigidly why the nucleation of islands happens although the growth is carried out homoepitaxially. The high impurity level in the island region can have more than one reason, e.g., gettering of impurities by dislocations and incorporation of different elements on the particular crystallographic facets. The accumulation of impurities on the growth interface can
12
Low Pressure Solution Growth of Gallium Nitride
263
Fig. 12.8 Correlation between the nucleation islands (SEM image, middle) and the abundance of dislocation lines and recombination action of dislocation in the first micrometer behind the seed–layer interface (TEM image, left) together with the depth distribution of impurity elements measured by SIMS analysis (right). Images are aligned towards each other along the interface Nucleation islands
Fig. 12.9 Panchromatic CL image of the interface region between the seed layer and the LPSG GaN. The nucleation islands are clearly distinguishable by darker contrast respectively lower CL intensity. The dislocations act as nonradiative recombination centers too and are visible as black lines (arrows)
however also result in a locally altered supersaturation with respect to the growing material and could cause a different nucleation regime although the general situation is homoepitaxial for GaN. On the other hand, the presence of the nucleation island is extremely beneficial in terms of reduction of the dislocation density. It was found in [41] that the dislocation density after the LPSG process was reduced by one to two orders of magnitude compared to the initial dislocation density of the seed as it is also obvious from the images above. The corresponding mechanism acting during the LPSG process is related to the nucleation of islands on the seed surface in the very beginning of the growth process. Therefore, it is very important to have sufficient control over the initial growth of the GaN in situ in the LPSG process, even though the nucleation situation is not fully understood up to now.
12.3.2 Macroscopic Defects in LPSG GaN Up to now most as-grown LPSG layers were reported not to be directly suitable for device production. This can be due to incomplete coalescence, existence of cracks,
264
E. Meissner et al.
Fig. 12.10 SEM images of the most prominent surface defects on as-grown GaN grown from the solution. Macrosteps are characteristic for most of the surface area (a), with hexagonal shaped flat pyramids as exemplarily shown in (b). The circular shaped surface defects are illustrated in (c) whereas (d) depicts particle induced morphological disturbances
blistering of the layer, or off-oriented GaN particles incorporated into the layers during growth (e.g., [13, 21]). The layers grown on c-plane sapphire or c-plane gas-phase seeding layers by Logan et al. exhibit strong surface structures characterized by hexagonal features. The surface morphology of recent LPSG layers show mainly terraces with macrosteps (Fig. 12.10a), interrupted by a variety of possible larger defects, like surface depressions and circular shaped defects [16] depending on seeding-layer-orientation, misfit, and process parameters (Fig. 12.10b–d). Most critical are surface depressions which have diameters up to 100 m and sometimes penetrate the whole epitaxial layer down to the seed surface. Therefore, they cannot be removed by polishing. These depressions are similar to the ones described for LPE studies of GaAs [42]. Several possible triggers for the formation of theses defects were described, with one reason being particles present on the seed or existing in the solution. Indeed, it was shown by Hussy et al. that GaN crystallites (particles) present in solution can induce the surface depressions in the LPSG GaN [22]. Avoiding the formation of particles in the solution is therefore a prerequisite for the production of GaN material free of such surface defects. One way to prevent the formation of such surface depressions is to adjust the density of the growth solution [22]. In those solutions with a density higher than that of the detrimental particles, the particles will float on the surface of the solution and hence cannot get into contact with the seed or the growth surface of the crystal at the bottom of the crucible. Another typical macro defect found in GaN-LPSG layers are circular shaped defects (Fig. 12.10c). They can be correlated to gas bubbles sticking on the growth
12
Low Pressure Solution Growth of Gallium Nitride
265
interface during the solution growth process [23]. However, the origin of those bubbles, which is among other things, correlated to the hydrogen content in the atmosphere is still under investigation. By adequate polishing technique most surface defects can be removed. Those layers as well as most recent as-grown layers were suitable for realization of LED test devices on the LPSG samples exhibiting an electroluminescence signal.
12.4 Properties of the LPSG GaN Material Many working groups made huge efforts to obtain GaN crystals with improved quality and structural perfection. Among all the methods tried, during the last few years, the solution growth is quite interesting as the crystallization of the material occurs under conditions which are close to thermodynamic equilibrium. Results of the high pressure solution growth demonstrated that very low dislocation density of some hundred dislocations per square centimeter can be attained [3]. Moreover, the growth situation under excess gallium can help to hinder the formation of point defects related to gallium vacancies as was discussed in [26]. The growth from a solution is however quite slow which is a great disadvantage for a technical process, but from many studies dedicated to the structural, optical, or electrical characterization of GaN, it is obvious that fast processes with simply high growth rates cannot deliver the desired materials quality [43], with the dislocation density of the materials being the most crucial issue. All fabrication methods rely on a full diameter seed in order to initiate growth of thick layers or quasi bulk GaN with usable diameters. In the majority of the cases, a seed with a GaN layer deposited by MOCVD is applied, which means that the initial dislocation density of the seed is in principle unacceptably high, compared to growth and seeding of classical compound semiconductor crystals. Measures have to be taken in order to reduce the dislocation density during the following growth process. The HVPE method reduces the dislocation density in the GaN boule as a function of thickness [44], whereas the common gas phase techniques, like MOCVD, take advantage from advanced measures like epitaxial lateral overgrowth (ELO) or related techniques [45]. The following section illustrates the situation in the LPSG material and describes how the dislocation density is reduced during the growth.
12.4.1 Structural Properties and Dislocation Density in LPSG GaN Material Dislocation Density of LPSG GaN Material Defect selective etching (DSE) was performed to evaluate the total dislocation density and the relation between the initial dislocation density (DD) and the one found
266
E. Meissner et al.
in the LPSG material. Many investigations about DD in GaN were published over the last few years. However, as can be seen from the literature, GaN originating from different production methods can behave quite differently in the wet etching procedures, such that it was suspected that many values given for the dislocation density might not represent the real DD present in the crystal [46]. The DD for the LPSG GaN was determined with TEM as well as with wet etching in NaOH/KOH eutectic melt. The etching temperature and time were carefully investigated and calibrated by comparison to TEM results. The optimum conditions for revealing the true DD density in LPSG material were found to be 420 ı C and 4 min etching time. Concerning the etching temperature, this is significantly higher than for the etching of common MOCVD layers as e.g., reported by [46]. The DSE of LPSG GaN reveals the same picture as often reported in the literature. The dislocations cropping out on the surface of the crystal were decorated with three different sizes of etch pits which could be counted separately. However, care has to be taken for sufficient statistics, especially for the large etch pits which are the fewest. Figure 12.11 also gives an example of the etched surface of an LPSG GaN material. By TEM it was validated that the large etch pits do decorate the c-type screw dislocations and the medium pits correspond to the mixed type dislocations, whereas the pure edge dislocations are decorated by small pits only. This agrees with the expectations from the literature [46]. Those small etch pits are the most difficult ones to etch and visualize because of the fact that the activation energy for the formation of such a pit is highest among the three types of defects [46]. The average DD in the LPSG GaN grown on a MOCVD GaN seed layer is some 108 cm2 , but can be as low as 1 107 cm2 .
Reduction of Dislocation Density by Bending of Dislocation Lines It was observed that during LPSG growth as well as epitaxial growth by the sodium flux LPE growth, the dislocation density was reduced by one to two orders of magnitude compared to the dislocation density offered by the MOCVD-GaN seed [41]. This mechanism functions because of a dislocation recombination action in a narrow range behind the seed-LPSG GaN interface. The annihilation can happen as a
s
m
l
Fig. 12.11 As grown c-plane GaN, Ga-polar surfaces of the MOCVD seed layer used for subsequent solution growth and the LPSG GaN in direct comparison. The samples were etched at 420 ı C for 4 min. The errors indicate the small (s), medium (m), and large (l) etch pits
12
Low Pressure Solution Growth of Gallium Nitride Dislocation lines LPSG GaN
267
VIIc V^c
Seedlayer
Substrate
Fig. 12.12 Schematic illustration of the dislocation reduction in a narrow region behind the seed surface. Nucleation of facetted islands (indicated by dashed lines) is the key for the induction of frequent interactions between bended dislocations. During further growth and lateral coalescence (indicated by grey colour) the dislocations can proceed towards each other and recombine or annihilate
result of the bending of the threading dislocations out of the c-growth direction. The bending enlarges the probability for a dislocation to meet another one and react or directly annihilate, in case of opposite Burgers vectors. The key issue is the presence of isolated, facetted nucleation islands in the very early growth state as described above and the tendency of the GaN material to form facets very easily, at least in the LPSG case as it was evaluated by [26]. This situation is illustrated by the schematic drawing in Fig. 12.12. The presence of crystal facets offers an energetic attraction to a dislocation to minimize its total line energy by changing its direction and proceed towards the facet surface as indicated in the figure. As the lateral growth velocity, v?c , is larger than the growth rate along the c-direction vIIc , the inclined dislocations can move towards each other upon the coalescence of the islands and interact if they get close enough [6, 41]. If the dislocation lines bend by large angles, a short distance after the seed surface is enough for a reduction of the dislocation density in the LPSG material. Bending by ninety degrees into the basal plane is most favorable. It will hinder the dislocations to propagate along the c-direction into the final active area of the layer, where the device is manufactured. But, a bending of dislocation lines by 90ı is not easily attained, because it requires the growth of steep side facets of the nucleation islands with particular crystallographic orientation. Only in that case the dislocation line can find its energy minimum and the driving force for the bending is high enough. Then the overall reduction of the number of dislocations is a function of the process time until island coalescence is attained, as the bending will only happen as long as the facets exist. Smaller bending angles of the dislocation lines would allow the coalescence time to be maintained for a longer time. Thus, in the LPSG growth also, the most beneficial combination are steep side facets of the nucleation islands provoking large bending angles. A detailed investigation and corresponding TEM investigations of the LPSG material by [41] showed, that dislocations behave differently, depending on their character with respect to their line and burgers vectors. The bending angle in relation to the c-direction is not only determined by the character of the dislocation, but also by the crystallographic index of the interacting
268
E. Meissner et al.
facet. A corresponding theoretical model for the observed behavior of dislocations in LPSG GaN was developed, which is on the basis of considerations as to the total energy budget of a dislocation and the minimization of line energy as the main driving force for various types of dislocations in GaN and their interaction with crystal surfaces. Further details can be found in [41], where also a thorough validation of the theoretical model by experimental observations give a good proof for the assumptions made. As can be determined from DSE images exemplarily shown in Fig. 12.11 it is obvious that the main difference between the DD in the seed and the LPSG GaN, lies in the number of small etch pits. This means that the dislocation reduction mechanism as described above does mainly act on the edge or the edge component of mixed type dislocations, if no special action – with respect to the type of facets of the nucleation islands – is taken as described in [41]. The pure c-direction screw dislocations are hard to influence as the screw configuration is energetically already very favorable. In general the DD can be reduced by one to two orders of magnitude by that mechanism compared to the initial DD present in the seed.
Structural Quality of the Lattice The structural quality of the synthesized materials was not only characterized by means of TEM and defect selective etching, but additionally investigated by high resolution x-ray diffraction. Despite the fact that no reliable model exists which directly correlates the dislocation density to the width of a x-ray rocking curve, one can qualitatively state that the smaller the width of the rocking curve, the smaller is the number of structural disturbances in the material measured. This includes all factors contributing to the width of a x-ray reflex, like twist and tilt of crystal blocks, strain, dislocations, lateral coherence length, and so on. Thus, rocking curve measurement is generally used as a relatively fast monitor for the crystal quality and for comparison purposes between different materials. In order to get the best idea about the real materials quality besides the effects of other structural contributions to the line width, one has to perform a different rocking curve measurement, which helps to deconvolute the effect. Hence, two types of rocking curve were measured for the characterization of the structural quality of the LPSG-GaN. First we investigated the typical omega scan rocking curves with open detector geometry. The spot size is large in order to integrate over a larger area of the sample. In this geometry, structural imperfections like block tilts towards the growth direction, residual strain, and threading screw dislocations contribute to the line width and shape of the rocking curve. Secondly we measured omega 2 theta scans of the material with the same large beam spot but with a 3 bounce 220 Ge analyzer crystal on the diffracted beam side, which monitors the real materials properties itself despite the other factors like the mosaicity. Figure 12.13 contains two exemplary curves measured for the [002] direction. Typically the samples from the LPSG process exhibit a rocking curve width
12
Low Pressure Solution Growth of Gallium Nitride
269
Fig. 12.13 Measurement of [002] rocking curves in two different diffraction geometries. (a) shows an example of a regular rocking curve measurement with an open detector, large beam spot and maximum acceptance angle for detection. Curve (b) exemplarily depicts a measurement with an analyzer on the diffracted beam side, monitoring the entire materials quality
around 200 ˙ 20 arcsec, measured with the open detector configuration as shown in Fig. 12.13a. This value is comparable to a high quality MOCVD GaN. No block tilts are visible from this curve, although the acceptance angle for detection of orientation differences is large. The entire materials quality of the LPSG GaN itself, blanking the other influences from the mosaicity, is expressed in the very narrow rocking curves as depicted in Fig. 12.13b. The width of those curves was found to be between 18 and 27 arcsec for all measured samples from standard LPSG processes, which was quite small and generally only found on material with high structural quality like GaN from high pressure solution growth or very recently from the ammonothermal method, although, principally the expectation holds that a process operated close to thermodynamic equilibrium should result in excellent structural properties of the material. Although the overall materials quality of the LPSG GaN is quite good and the dislocation reduction method does act usefully if the right conditions are chosen, the material shows a reduced DD compared to the seed layer, but it will not be possible to remove all the dislocations from the LPSG material that way. As soon as the dislocation density approaches a value on the order of 106 cm2 , the dislocation elimination by a recombination process is not effective any more as the interaction distances necessary for dislocation reactions are on the order of tens of nanometers, whereas the average distance of the dislocations at this DD is on the micrometer scale. However, it is a crucial fact for all existing growth processes for nitrides that the total number of dislocations in the material must be further reduced. But, by the first glance this can only be reached if either homogeneous nucleation is used or high quality seeds are applied. Here one could see a basic advantage of a solution growth process as free nucleation in the volume of the solution is theoretically possible, but it is a big challenge. So, one of the most urgent questions left to be solved now is how to generate a high quality seed for the GaN growth processes.
270
E. Meissner et al.
12.4.2 Electrical Properties Conductive n-type or p-type GaN substrates are attractive for optoelectronic devices, as the conductivity of the substrate offers the possibility for back side contacts and a free path to the front side for out coupling of light, whereas semi insulating material is required for RF applications. In this sense, the solution growth has an advantage and a disadvantage. Typically solution grown material can suffer from incorporation of the solvent into the material produced. In those cases where the solvent can be used as a dopant this can be taken as an advantage. In the case of the LPSG growth of GaN, it is easy to use a solvent like germanium which will result in a highly n-type doped gallium nitride. The electrical properties of the LPSG GaN were determined by classical Hall measurements as well as by Fourier Transformed Infrared (FTIR) reflectivity measurements. In comparison to the Hall measurements the FTIR measurements can be performed without the necessity of contacts and give reliable results down to n D 5 1017 cm3 [27]. The carrier density of the LPSG GaN is typically around 4 1019 cm3 if an n-type solvent like germanium is applied which readily incorporates into the GaN structure. Variations among the processes or within the wafers produced by multi wafer processing are marginal only. Other solvents like bismuth lead to GaN with around one order of magnitude lower carrier density compared to the germanium doped material. So far, p-type doping has not been tried.
12.4.3 Impurities An average analysis of impurities in LPSG GaN, in the case of germanium containing solutions, is given in the Table 12.3. The values were determined by secondary ion mass spectroscopy and averaged over the entire layer thickness. Note, that this is not entirely true if one compares Fig. 12.8, where an accumulation of impurities close to the seed surface is obvious. It is clear from those measurements that the system can be kept quite clean and the major impurity is the solvent resp. the dopant. Despite that, Si and O do have a Table 12.3 Concentration of impurities in LPSG GaN as measured by secondary ion mass spectroscopy Impurity Germanium Oxygen Silicon Hydrogen Aluminum Boron Carbon
Average concentration in LPSG GaN, per cm3 9 1019 3 1018 1 1018 1 1017 3 1016 2 1016 1 1016
12
Low Pressure Solution Growth of Gallium Nitride
271
noticeable abundance, whereas the other elements are quite low in concentration. No other impurities like alkali or earth alkali metals are present in this type of material because they do not find any use in the LPSG process. As discussed above, the impurities in the gallium nitride from the LPSG process mainly stem from the solvents added and therefore the gallium as well as the solvents should be of highest possible purity. However, an evaluation of the process concerning sources of impurities was necessary too. The oxygen contamination is most likely caused by the effect of hydrogen on the quartz glass ampoule. Boron can be mobilized from the boron nitride crucible. The process atmosphere containing ammonia and hydrogen is quite aggressive and it is important to choose appropriate materials in the interior of the furnace, especially in the hot zone. Graphite or oxide materials in the hot zone are no good choice as discussed in the Sect. 12.2.3 and are therefore not present in the process and the light carbon content obviously comes from another source.
12.5 Summary and Prospect The consequence of all the detailed studies discussed in this chapter is a deep understanding of the overall LPSG growth process as well as the, before more or less unclear, role of the solvents in terms of kinetics and solubility of nitrogen. Many of those findings are transferable to other solution growth methods which are probably very comparable among all the variants of solution growth. The intrinsic big capacity of a solution growth process is definitely the expectable materials quality from thermodynamic considerations, which were best demonstrated by the high pressure solution growth material. On the other hand, there are unwanted aspects like low supersaturations and in turn the very low growth rates leading to long process times. However, it was demonstrated by other groups like the sodium flux researchers that the growth rates can be higher by a factor of 10 or 20 compared to the actual state of the low pressure or high pressure solution growth. The materials quality which resulted from those Na-flux processes would need additional research efforts to be acceptable. The investigation and understanding of the action of the solvents as shown in this chapter can now help to perform a consequent screening of possible solution additives. With view to the enhancement of growth rates the real potential of this method can be tapped as much as possible by identifying the best solvents. But, all joint efforts did not clarify the experimentally maximal attainable growth rate for the solution growth methods up to now and more research efforts are necessary. It is anyway a mater of discussion whether bulk crystals are always needed for all application cases. The LPSG process can reduce the dislocation density remarkably and a model was built describing the underlying physical mechanism in a way that it can be used to give a forecast on the behavior of the dislocations. Because of the easiness of the whole LPSG setup, it is not difficult and very straight forward to up scale the process to larger diameters, which was successfully demonstrated. Seeing this, there seem to be no real obstacles for the growth on larger than 3 in. diameters,
272
E. Meissner et al.
despite the availability of seeds. The method can also be considered as a variant of a liquid phase epitaxy process, which found wide application as a multi wafer route and can produce large numbers of wafers in parallel. For the multi wafer production of GaN from the LPSG process we could demonstrate that it is possible to process e.g., three wafers in parallel without any loss in quality and comparable electrical properties within experimental errors. However, evaluating and exploiting the full potential of the low pressure solution growth method do need further research work and the question how to produce high quality seeds remains to be solved for all the nitride crystal growth methods. Acknowledgements The authors would like to acknowledge the support of all people who have contributed to this work by various amounts over the years. Especially, the authors are grateful to Rainer Apelt, Carmen Maier, and Michael Lang for their unremitting and perfect technical support. G. Li Sun for his contribution in the frame of thermogravimetry. B. Birkmann for his work in the development of the LPSG process during the first period of the project. I. Knoke for the TEM image and L. Kirste, Fraunhofer IAF, for discussion and help with the rocking curve measurements. Thanks to B. Sieber, University of Lille for the panchromatic CL image and J. Weyher for fruitful discussion about the wet etching results. The authors would also like to thank Prof. G. Müller for continuous discussions and valuable suggestions over the years. This work was supported by the German ministry of education and science (BMBF) under contracts number 01-BM-158 and 01BM-580. The TEM work by I. Knoke, contributing to this work was supported by a Fraunhofer PhD scholarship.
References 1. W. Utsumi, H. Saitoh, H. Kaneko, T. Watanuki, K. Aoki, O. Shimomura, Nat. Mater. 2, 735–738 (2003) 2. J. Karpinski, J. Jun, S. Porowski, J. Cryst. Growth 66, 1–10 (1984) 3. I. Grzegory, M. Bockowski, B. Lucznik, S. Krukowski, Z. Romanowski, M. Wroblewski, S. Porowski, J. Cryst. Growth 246, 177–186 (2002) 4. H. Yamane, D. Kinno, M. Shimada, T. Sekiguchi, F. DiSalvo, J. Mater. Sci. 35, 801–808 (2000) 5. M. Aoki, H. Yamane, M. Shimada, T. Sekiguchi, T. Hanada, T. Yao, S. Sarayama, F. DiSalvo, J. Cryst. Growth 218, 7–12 (2000) 6. F. Kawamura, H. Umeda, M. Kawahara, M. Yoshimura, T. Sasaki, Y. Mori, Y. Kitaoka, Jpn. J. Appl. Phys. 45, L1136–L1138 (2006) 7. T. Yamada, H. Yamane, Y. Yao, M. Yokoyama, T. Sekiguchi, Mater. Res. Bull. 44, 594–599 (2009) 8. S. Barry, S. Ruoff, A. Ruoff, Chem. Mater. 10, 2571–2574 (1998) 9. Y. Song, W. Wang, W. Yuan, X. Wu, X. Chen, J. Cryst. Growth 247, 275–278 (2003) 10. F. Kawamura, M. Morishita, T. Iwahashi, M. Yoshimura, Y. Mori, T. Sasaki, Jpn J. Appl. Phys. 41, L1440 (2002) 11. M. Morishita, F. Kawamura, M. Kawahara, M. Yoshimura, Y. Mori, T. Sasaki, J. Cryst. Growth 284, 91–99 (2005) 12. F. Kawamura, M. Morishita, M. Tanpo, M. Imade, M. Yoshimura, Y. Kitaoka, Y. Mori, T. Sasaki, J. Cryst. Growth 310(17), 3946–3949 (2008) 13. R. Logan, C. Thurmond, J. Electrochem. Soc. Solid-State Sci. Technol. 119, 1727–1734 (1972) 14. D. Elwell, R. Feigelson, M. Simkins, W. Tiller, J. Cryst. Growth 66, 45–54 (1983) 15. E. Meissner, G. Sun, S. Hussy, B. Birkmann, J. Friedrich, G. Müller, 21th Century COE Joint International Workshop on Bulk Nitrides (Institute of Pure and Applied Physics, Tokyo, Japan 2003)
12
Low Pressure Solution Growth of Gallium Nitride
273
16. S. Hussy, E. Meissner, P. Berwian, J. Friedrich, G. Müller, J. Cryst. Growth 310(4), 738–747 (2008) 17. H. Yamane, M. Shimada, S.J. Clarke, F.J. DiSalvo, Chem. Mater. 9, 413–416 (1997) 18. H. Yamane, T. Kajiwara, T. Sekiguchi, M. Shimada, Jpn. J. Appl. Phys. 39, L146–L148 (2000) 19. M. Morishita, F. Kawamura, T. Iwahashi, M. Yoshimura, Y. Mori, T. Sasaki, Jpn. J. Appl. Phys. 42, L565–L567 (2003) 20. T. Iwahashi, F. Kawamura, M. Morishita, Y. Kai, M. Yoshimura, Y. Mori, T. Sasaki, J. Cryst. Growth 253 (2003) 1–5 21. C. Klemenz, H.J. Scheel, J. Cryst. Growth 211, 62–67 (2000) 22. S. Hussy, P. Berwian, E. Meissner, J. Friedrich, G. Müller, J. Cryst. Growth 311(1), 62–65 (2008) 23. S. Hussy, PhD-thesis, Flüssigphasenepitaxie von Galliumnitrid unter ammoniakhaltiger Atmosphäre Universität Erlangen-Nürnberg (2008) 24. J. Karpinski, S. Porowski, J. Cryst. Growth 66, 11–20 (1984) 25. G. Sun, E. Meissner, P. Berwian, G. Müller, J. Friedrich, J. Cryst. Growth 305, 326–334 (2007) 26. B. Birkmann, S. Hussy, G. Sun, P. Berwian, E. Meissner, J. Friedrich, G. Müller, J. Cryst. Growth 297, 133–137 (2006) 27. B. Birkmann, C. Salcianu, E. Meissner, S. Hussy, J. Friedrich, G. Müller, Phys. Status Solidi. (C) 3, 575–578 (2006) 28. G. Sun, PhD-thesis, Thermogravimetric Studies of Nitrogen Solubility in Gallium Solutions and Reaction Kinetics of Gallium with Ammonia, Universität Erlangen-Nürnberg (2006) 29. G. Sun, E. Meissner, P. Berwian, S. Hussy, G. Müller, J. Cryst. Growth 305(2), 326–334 (2007) 30. G. Sun, E. Meissner, P. Berwian, G. Müller, Thermochim. Acta 474, 36–40 (2008) 31. D. Elwell, H.J. Scheel, Crystal Growth from High-Temperature Solutions (Academic, London, 1975) 32. O. Söhnel, J.W. Mullin, J. Coll. Interf. Sci. 123, 43–50 (1988) 33. S. He, J.E. Oddo, M.B. Tomson, J. Coll. Interface Sci. 174, 319–326 (1995) 34. S. Krukowski, Cryst. Res. Technol. 34, 785–795 (1999) 35. H. Yamane, D. Kinno, M. Shimada, T. Sekiguchi, F.J. Disalvo, J. Mater. Science 35, 801–808, (2000) 36. R. Madar, G. Jacob, J. Hallais, R. Fruchart, J. Cryst. Growth 31, 197–203 (1975) 37. C. Werkhoven, R. Peters, J. Cryst. Growth 31, 210–214 (1975) 38. C. Thurmond, R. Logan, J. Electrochem. Soc. Solid-State Sci. Technol. 119, 622–626 (1972) 39. M. Morishita, F. Kawamura, M. Kawahara, M. Yoshimura, Y. Mori, T. Sasaki, J. Cryst. Growth 270, 402–408 (2004) 40. H. Klapper, Phys. Status Solidi. A 14, 99–106 (1972) 41. I.Y. Knoke, E. Meissner, J. Friedrich, H.P. Strunk, G. Müller, J. Cryst. Growth 310(14), 3351– 3357 (2008) 42. E. Bauser, Appl. Phys. 15, 243–252 (1978) 43. T. Paskova, E. M. Goldys, R. Yakimova, E. B. Svedberg, A. Henry, B. Monemar, J. Cryst. Growth 208, 18–26 (2000) 44. R. Vaudo, X. Xu, C. Loria, A. Salant, J. Flynn, G. Brandes, Phys. Status Solidi (A) 194, 494–497 (2002) 45. F. Habel, P. Brückner, F. Scholz, J. Cryst. Growth 272, 515–519 (2004) 46. J. Weyher, Personal communication with E. Meissner (2006) 47. J.L. Weyher, S. Lazar, L. Macht, Z. Liliental-Weber, R.J. Molnar, S. Müller, V.G.M. Sivel, G. Nowak, I. Grzegory, J. Cryst. Growth 305, 384–392 (2007) 48. M. Aoki, H. Yamane, M. Shimada, S. Sarayama, F. DiSalvo, Mater. Lett. 56, 660–664 (2002)
Part V
Characterization of GaN Crystals
Chapter 13
Optical Properties of GaN Substrates Shigefusa F. Chichibu
Abstract Optical properties of GaN substrates are surveyed to discuss the superiority and critical issues as a substrate for device fabrication. At first, optical properties of a few hundred-micrometer-thick free-standing GaN substrates grown heteroepitaxially on (0001) Al2 O3 substrates by metalorganic vapor phase epitaxy and halide vapor phase epitaxy using lateral overgrowth techniques are introduced, as they can also be used as a seed substrate for the ammonothermal overgrowth. Then, spatiallyresolved cathodoluminescence (CL) spectra of GaN films grown on free-standing GaN seeds via fluid transport by the ammonothermal method are correlated with the microstructure and growth polarity. The spectral lineshape of local CL was nearly position-independent for a 4-m-thick N-polar film exhibiting featureless morphology: the spectra exclusively exhibited a broad near-band-edge (NBE) free-carrier recombination emission with Burstein–Moss shift. Conversely, CL spectra at 100 K N and (1012) N facets with ridges origof a 5-m-thick Ga-polar film having (1011) inating from central pits exhibited an NBE peak at 3.444 eV and emission bands at 3.27, 2.92, and 2.22 eV, all of which showed rich intensity contrasts in the CL mapping images. The NBE peak intensity was remarkably enhanced at crests of the ridges, where the density of threading dislocations (TDs) having edge components was greatly reduced by the dislocation bending. The results encourage one to grow low TD density GaN wafers by slicing thick crystals grown by the ammonothermal method.
13.1 Introduction Optoelectronic devices such as bright blue, green, and white light-emitting diodes, laser diodes operated at 405 nm [1–4], high electron mobility transistors, and bipolar transistors have been realized using GaN and related (Al,In,Ga)N alloys grown heteroepitaxially on (0001) Al2 O3 substrates. However, demands are thrown for obtaining low threading dislocation (TD) density, large-area, and low-cost homoepitaxial substrates to realize high-efficiency, highly reliable, and low-cost devices.
277
278
S.F. Chichibu
To comply with the requirements, the use of GaN wafer sliced from bulk GaN ingots is an ultimate solution. Several growth techniques using a Ga melt have been investigated to grow bulk GaN crystals [5–7]. However, as nitrogen has a very limited solubility in molten Ga, thin platelets of GaN were exclusively obtained. In addition, a bulk growth method must be expandable to an industrial mass-production scale to compete with existing GaN substrates prepared by metalorganic vapor phase epitaxy (MOVPE) [8] and halide vapor phase epitaxy (HVPE) on a patterned (111)A GaAs substrate [9] or on a GaN/(0001) Al2 O3 template whose surface is modified with TiN [10]. One of the promising techniques is an ammonothermal (AT) method, which is a solvothermal growth method that uses supercritical ammonia as a fluid medium, and is considered to have excellent scalability similar to the case for ’-quartz and ZnO hydrothermal [11] ones. Growth of AT-GaN has been investigated by several groups [12–18], and Hashimoto et al. [19–22] have reported on seeded growth of GaN on a free-standing (FS) GaN substrate prepared by HVPE [23] demonstrating the single crystalline growth [19] of large-area [20] .34 cm2 / and low TD density [21] (lower than 1106 cm2 for Ga-polar face) GaN films using basic mineralizers. Also, over 5-mm-long, 5mm-diameter hexagonal bulk GaN ingot has been grown recently [22]. Large-area (46-mm-diameter) seeded growth of GaN by AT method has also been reported by Ehrentraut et al. [24] using acidic mineralizers. Very recently, Dwilinski et al. have demonstrated 1-in.- [25] and 2-in.-diameter [26] GaN wafers grown by the AT method, both of which had nearly zero wafer bowing (curvature radius greater than 1 km). Structural properties of AT-GaN have been investigated by several groups. However, direct correlation between the microstructure and luminescence spectrum had not been reported before our work [27], although precise information obtainable from those analyses such as exciton energies versus strain, impurities, and carrier lifetime is needed to put AT-GaN in practical use as a homoepitaxial substrate. In this chapter, optical properties of GaN substrates are surveyed to discuss the superiority and critical issues as a substrate for device fabrication. At first, optical properties [28, 29] of a few hundred-micrometer-thick FS-GaN substrates grown heteroepitaxially on (0001) Al2 O3 substrates by MOVPE [8] and HVPE [30] using lateral epitaxial overgrowth (LEO) techniques are introduced, as they can also be used as a seed crystal for AT growth. Then, results of spatially-resolved cathodoluminescence (CL) measurements on Ga- and N-polar faces of GaN crystals grown on HVPE FS-GaN seeds via fluid transport by AT method are shown to correlate the local CL spectrum with the morphological feature, microstructure, and growth polarity [27]. We demonstrate the possibility to prepare low TD density GaN wafers by slicing thick crystals grown toward the Ga-polar direction by AT method, whose TD density is reduced by dislocation bending.
13
Optical Properties of GaN Substrates
279
13.2 Optical Properties of GaN Substrates Grown by Metalorganic Vapor Phase Epitaxy and Halide Vapor Phase Epitaxy 13.2.1 Photoreflectance Spectra of Excitonic Polaritons in MOVPE-LEO GaN Substrate As purely bulk, single crystals of GaN with sufficient area have not been available for a long time, the use of thick GaN films grown on heterogeneous substrates has been one of the alternative solutions to grow thick bulk GaN films. Nevertheless, one can pursue intrinsic material physics using such high quality GaN substrate with reduced TD and residual impurity density. Yamaguchi et al. [31, 32] have studied optical reflectance (OR) and photoluminescence (PL) spectra of thick GaN and GaN substrates prepared by HVPE-LEO to determine some valence band parameters. Concerning exciton–polariton [33] in GaN, Gil et al. [34] and Stepniewski et al. [35] have discussed exciton–polaritons in GaN=Al2 O3 [34] and homoepitaxial layers on bulk GaN crystals grown under very high-pressure N2 ambient [35]. Also, the authors have investigated low temperature OR and PL spectra of the GaN substrate grown by MOVPE-LEO [8] to show that the spectra were well described by a model exciton–polariton picture in which A, B, and C free excitons couple simultaneously to an electromagnetic wave [36]. Electromodulation spectroscopy such as electroreflectance (ER) yields a very sharp transition structure related to the third derivative of dielectric functions [37–39], and is a useful technique to obtain information on the bandgap. ER can be used to investigate exciton resonances in wide bandgap materials under the limit of low electric field [40] especially at low temperature. However, very little is known of the ER structure of FS-GaN. Here the spectra of photo-modulated electroreflectance, a contactless form of ER and is referred to as photoreflectance (PR), and PL of the high quality GaN substrate [8] are shown [29]. The GaN substrate [8] was grown by MOVPE-LEO with subsequent Al2 O3 removal. Details of the preparation procedures and structural properties are given in [8, 29, 36]. The lattice constants are a D 0:31898 ˙ 0:00002 nm and c D 0:51855 ˙ 0:00002 nm, and the crystal was close to strain-free. The residual electron density, n, at 300 K was too low to measure by the Hall effect measurement, i.e., n is estimated to be lower than 1014 cm3 , which is the limit of our equipment. Reduced nonradiative defect density was also characterized by time-resolved photoluminescence (TRPL) [28] and positron annihilation measurements [41–43]. The GaN substrate exhibited biexponential-type decay with decay constants 1 D 130 ps and 2 D 860 ps at 300 K, which will be shown later, while a standard 1-m-thick GaN=Al2 O3 generally exhibits a short decay of 2 shorter than 300 ps. OR spectra were measured at 10 K with incident angle smaller than 10ı . PR and PL spectra were measured as a function of temperature using the 325.0 nm line of a cw He–Cd laser as an excitor. The reflected white light from a Xe lamp and emitted light were dispersed by a f D 67-cm-focal-length grating monochromator, and phase
280
S.F. Chichibu
0.4 0.3
3.47
3.48
3.49
3.50
3.51
a
0.2 0.1 0.0
Theory Experiment
b
A
PR SIGNAL ΔR / R (arb. units)
REFLECTANCE
sensitive detection was carried out using a GaAs:Cs photomultiplier. The spectral resolution was ˙0:3 meV at a wavelength of 350 nm. The OR, PR, and PL spectra of the surface and the back of the substrate, namely Ga-polar and N-polar faces, were confirmed to be identical. Here we introduce the results on the Ga-polar face. Figure 13.1 shows (a) OR, (b) PR, and (c) PL spectra of MOVPE FS-GaN substrate measured at 10 K. In Fig. 13.1a, b, sharp reflection anomalies corresponding to the ground states of A, B, and C free excitons and the first excited states of A exciton .AnD2 / are found around 3.480, 3.485, 3.503, and 3.498 eV, respectively. The sample exhibited correspondingly sharp A, B, and AnD2 free excitonic PL peaks and a neutral donor bound exciton peak I2 . The full width at half maximum (FWHM) of I2 was as small as 0.7 meV. However, PL peaks corresponding to A and B free excitons were rather broad (2.6 meV for the A peak) and they cannot be fitted by single Lorentzian line-shape functions, suggesting the formation of excitonic polaritons. Indeed, emissions from lower and upper polariton branches of A exciton (LPBA and UPBA , respectively) and those of B exciton (LPBB and UPBB ) have been resolved using near-resonant excitation sources [36]. To interpret the OR spectrum, theoretical calculations have been carried out on the basis of a model exciton–
B
t (ps)
PL INTENSITY (arb. units)
C An=2
I2
A,Bn=2,3..
c 8 6 4 2 0
d
LPBA UPBB+LPBC
UPBA+LPBB
3.47
3.48
3.49
3.50
3.51
PHOTON ENERGY (eV)
Fig. 13.1 (a) Reflectance OR, (b) PR, and (c) PL spectra of FS-GaN measured at 10 K. The transition structures labelled A, B, C, and AnD2 in the PR spectrum (b) are related to A, B, and C free excitons and the first excited states of A exciton, respectively. The PL peak I2 is due to recombination of excitons bound to a neutral donor. Theoretically fitted reflectance spectrum is shown by the solid line in (a). The calculated lifetimes of LBPA , a branch combined UPBA and LPBB , and a branch combined UPBB and LPBC are shown in (d) where LPB’ .UPB’ / means the lower (upper) polariton branch associated with the ’ exciton. Dotted vertical lines through (a) to (d) show the transition energies obtained from the analysis of the PR spectrum (b). (After [29])
13
Optical Properties of GaN Substrates
281
Table 13.1 Values of free exciton energies at 10 K of the MOVPE-LEO FS-GaN substrate obtained from the fitting procedure of the OR spectrum. Transition energies and broadening parameters obtained from the analysis of the PR spectrum assuming the transitions as pure free excitons are also listed. (After [29]) Method
Parameter
Exciton/Transition A
B
C
Optical Exciton resonance 3:4791 ˙ 0:0002 3:4844 ˙ 0:0002 3:5027 ˙ 0:0002 reflectance energy (eV) (OR) Photoreflectance Transition 3:4803 ˙ 0:0003 3:4852 ˙ 0:0003 3:5033 ˙ 0:0003 (PR) energy (eV) Broadening parameter 1.0 1.5 1.1 (meV)
polariton picture in which A, B, and C free excitons couple simultaneously to an electromagnetic wave, where the effective mass anisotropy, the optical anisotropy, free exciton damping, and exciton “dead layer” are fully taken into account. Details of the calculation procedures are given in [36], in which the polariton lifetime (Fig. 13.1d) has also been estimated. The best-fit curve of R D j.1 n/=.1 N C n/j N 2, where nN is the effective refractive index, is shown by a solid line in Fig. 13.1a. Satisfactory agreement was achieved, and the exciton parameters are summarized in Table 13.1. The PR spectrum shown in Fig. 13.1b was analyzed by the low-field ER lineshape function [37–39], 9 8 p = direction [28]. The edge-character TD denSiO2 stripes oriented in the < 1100 sity in the wing region was lower than 106 cm2 [50, 51]. We note that the wing
S.F. Chichibu
PL INTENSITY (arb. units)
284
300K
103 τ1= 130 ps
FS-GaN substrate τ2= 860 ps
102
101
τ1= 87 ps 1 μm GaN/Al2O3
0.0
0.5
LEO GaN (wing) τ2= 400 psec
τ2= 260 ps
1.0
1.5
2.0
TIME (ns) Fig. 13.4 TRPL signals measured at 300 K of the FS-GaN substrate, the wing region of an LEO GaN, and a GaN epilayer on Al2 O3 , which were grown by MOVPE. (After [28])
region of the LEO GaN suffered from a compressive biaxial strain up to 0.18% [52–54] after the coalescence. The TRPL of the NBE excitonic emission peak was excited by an approximately 150 fs pulse of a frequency-tripled, tuneable mode-locked Al2 O3 W Ti laser operating at 80 MHz repetition rate. The excitation energy and power at the sample were typically 4.64 eV and approximately 200 nJ cm2 per pulse, respectively. The signal was collected using a standard streak-camera acquisition technique. As the TRPL decay shapes could be fitted by a double-exponential function, fast and slow time constants (1 and 2 , respectively) were deduced. The FWHM value of the NBE PL peak decreased from 32 meV to 29 meV at 300 K by using LEO technique. From their peak energies and FWHM values, the PL peaks are assigned as being due to the recombinations of A- and B-free exciton (FE) emissions [40]. In general, PL lifetime at room temperature is limited by nonradiative processes because of thermal activation of nonradiative recombination centers (NRCs). Therefore, the recombination lifetimes of FE emissions shown in Fig. 13.4 may reflect the free (minority) carrier lifetime in the GaN crystals. From Fig. 13.4, it is noted that 1 was constantly 130 ps for the two low TD density samples, and was 87 ps for the most defective GaN epilayer. These results indicate that TDs may be related to NRCs [28, 42, 43]. However, nonradiative lifetime in GaN crystals has been shown to be mainly limited by the concentrations of point defect complexes involving Ga vacancies .VGa / [42, 43, 55]. Indeed, both 1 and 2 changed remarkably by changing the growth parameters or background carrier concentration, even if a standard GaN on Al2 O3 with intermediate TD density .5 108 cm2 / was analyzed. The pronounced long 2 of FS-GaN (860 ps) may indicate improved purity of the crystal.
13
Optical Properties of GaN Substrates
285
13.2.3 Low Temperature PL Spectra of HVPE-LEO GaN Substrate As described in Sect. 13.1, HVPE method is one of the most promising ways to grow thick GaN substrates, according to its high growth rate. Consequently, FS-GaN grown by HVPE can be used as a seed crystal for the boule growth by AT method. However, undoped GaN grown by HVPE in general shows n-type conductivity because of the residual impurities such as O and Si, which may give rise to the degradation of device performances as a result of the parasitic capacitance, current leakage, and the resultant cross-talk. Also, the wafer bowing problem originating from in-plain compressive strain in FS-GaN is unavoidable, as long as lattice-mismatched Al2 O3 is used as an original substrate. Therefore, growths of purely strain-free FS-GaN crystal must be carried out to improve the device performances. In this subsection, low temperature PL spectra of a FS-GaN substrate [30] prepared by HVPE using the facet-initiated epitaxial lateral overgrowth (FIELO) method [56] are shown to survey its optical quality. The sample measured was a 180-m-thick undoped FS-GaN grown by HVPE using the FIELO method. The FIELO-GaN was grown on a 2-m-thick GaN template prepared by MOVPE on 2-in. diam. (0001) Al2 O3 . Ammonia .NH3 / was used for an N source and GaCl formed from HCl flown on metallic Ga was used for a Ga precursor. The TD density was lower than 107 cm2 . Steady-state PL was excited by the 325.0 nm line of a cw He–Cd laser .6 W cm2 /, and was dispersed by a 25cm-focal-length grating monochromator. Phase-sensitive detection was carried out using a GaAs:Cs photomultiplier. PL spectra of the FIELO FS-GaN measured at 8 K and 300 K are shown in Fig. 13.5a, b, respectively. The NBE PL spectrum at 8 K is magnified in Fig. 13.5c. As shown in Fig. 13.5a, c, the spectra exhibited sharp PL peaks originating from
WAVELENGTH (nm) 380
1LO 2LO
(a) 8K FEA,B
(b) 300K YL
1.5
BL
2.0 2.5 3.0 PHOTON ENERGY (eV)
3.5
(c) 8K
350
TES I23.473 A 3.478 B 3.484 A,Bn=2,3… 3.502
PL INTENSITY (arb. units)
PL INTENSITY (arb. units)
FEA
WAVELENGTH (nm) 370 360
3.453
350
400
3.457 3.461
500
1LO
600
2LO
800
3.3 3.4 3.5 PHOTON ENERGY (eV)
Fig. 13.5 PL spectra of undoped very thick FIELO-GaN measured at (a) 8 K and (b) 300 K. (c) The near-band-edge PL spectrum of undoped FIELO-GaN at 8 K
286
S.F. Chichibu
free A- and B-excitons at 3.478 and 3.484 eV, respectively [29]. A peak originating from the excited states of free A- and B-excitons .A; BnD2;3:: / was also detected at around 3.502 eV. In addition, a sharp PL peak due to the recombination of excitons bound to a neutral donor .I2 / at 3.473 eV as well as two-electron satellites (TES) at around 3.453 eV [57] were observed. Remarkable emission peaks labeled 1LO and 2LO were observed in the lower energy side of the NBE peaks. They are assigned to LO phonon replicas of the main peaks, since their energy separation is close to the longitudinal optical (LO) phonon energy of 91 meV [49, 58]. In addition to the NBE peaks, broad luminescence bands in blue and yellow color region (BL and YL, respectively) at around 2.8 and 2.2 eV were recognized, especially at 300 K. Comparing the results of PL and positron annihilation measurements, the origin of YL band has been assigned to acceptor-type defect complexes composed of VGa and/or VGa O [42,59–61]. The origin of BL band has been assigned to a donor-to-valence band transition, where the donor level has been proposed to be located at 0.58 eV below the conduction band minimum in undoped HVPE-GaN [62,63]. Another possible origin is a conduction band-to-acceptor transition like carbon. Anyhow, the intensities of BL and YL were more than two orders of magnitude lower than the NBE peak, indicating low concentrations of those point defects.
13.3 Effects of Growth Polar Direction on the Optical Properties of Seeded GaN Substrates Grown by Ammonothermal Method The samples investigated were a 5-m-thick Ga-polar and a 4-m-thick N-polar GaN films grown simultaneously on the front and back sides of a 400-m-thick (0001) FS-GaN seed, which was prepared by HVPE on a (0001) Al2 O3 followed by laser lift-off [23]. The TD density of the seed was approximately 107 –108 cm2 . The films were grown by AT method using a high-pressure autoclave [19–21] made of an Ni-based superalloy. Metallic Ga nutrient was placed in the low-temperature zone .520–530 ıC/ and the seed crystal was placed in the high-temperature zone .550–565 ıC/, because GaN has retrograde solubility at around 500 ıC in supercritical ammonia with basic NaNH2 NaI mineralizers [19] used in this study. The pressure was maintained at 150–200 MPa, and the growth was carried out for 72 h. Details of the growth have been described in [19, 20]. The surface morphology was observed with scanning electron microscopy (SEM) and atomic force microscopy (AFM). The microstructure was characterized with transmission electron microscopy (TEM) operated at 200 keV. TEM samples were N > axis by mechanical polishing prepared for observation nearly along the < 1010 C followed by Ar milling. For determination of polarity, convergent beam electron diffraction was simultaneously observed using the TEM specimen. Trace impurities were identified by secondary-ion mass spectrometry (SIMS). Wide-area .200 m¥/ PL spectra were recorded at 293 K using the 325.0 nm line of a cw He Cd laser .38 W cm2 / as an excitor.
13
Optical Properties of GaN Substrates
Fig. 13.6 Plan-view and cross-sectional SEM images of Ga-polar (a, c) and N-polar (b, d) ammonothermal GaN films grown on a 400-m-thick HVPE FS-GaN seed wafer. (After [27])
287 Ga-polar
N-polar
a
b
RMS: 2.1 nm
RMS: 0.6 nm
10 µm (surface)
c
d
5 µm (cross section)
WAVELENGTH (nm) PL or CL INT. (arb. units)
Fig. 13.7 Wide-area (a) CL spectra at 100 K and (b) PL spectra at 293 K of AT-GaN films. For comparison, PL spectrum of the HVPE FS-GaN seed is shown in (b). (After [27])
600
500
(a) 100K CL 15kV, 1nA
400
C
D
Ga-polar (b) 293K PL He-Cd 325.0nm 12mW
340
A B
H
N-polar
HVPE GaN seed (x 1/10)
2.0
2.5 3.0 3.5 PHOTON ENERGY (eV)
Plan-view and cross-sectional SEM images of the 5-m-thick Ga-polar (4-mthick N-polar) AT-GaN films are shown in Fig. 13.6a, c (Fig. 13.6b, d), respectively. Under the particular growth conditions used in this study, average film thickness was comparable with each other. The Ga-polar surface was filled with grooves originating from linearly aligned inverted hexagonal pyramids having preferentially N and (1012) N facets, which have a pit at the center, whereas N-polar surface (1011) was nearly featureless. However, the local root-mean-square (RMS) roughness for the flat region of the Ga-polar surface (0.6 nm) was smaller than that for the Npolar surface (2.1 nm). Suspicious source of the grooves and their influence on the behavior of TDs and on the CL spectra will be discussed later. As shown in Fig. 13.7a, b, the N-polar film exclusively exhibited a broad NBE emission band both at 100 and 293 K. As the peak energy (3.440 eV) was higher by 30 meV than the A- and B-free exciton peak of strain-free GaN . 3:41 eV/ [29] and the FWHM value (231 meV) was so large at 293 K, the emission is assigned as being due to the free-carrier recombination with pronounced Burstein–Moss (BM) shift [64]. From the FWHM value at 100 K being 200 meV, the residual electron concentration is estimated to be approximately 1020 cm3 [65]. Culprit impurities were inspected by SIMS measurement to be Si and O. On the other hand, PL spectrum at 293 K of the Ga-polar film resembled that of the HVPE FS-GaN seed, though the broadening of the NBE peak was obvious and a weak band at 3 eV was observed, as shown in Fig. 13.7b. Remarkable difference
288
S.F. Chichibu
against the N-polar film was found at 100 K, as shown in Fig. 13.7a: the wide-area CL spectrum exhibited a predominant peak at 3.444 eV and a higher energy shoulder around 3.57 eV, and bands at 3.27 eV, 2.92 eV, and 2.22 eV. They are labelled A, H, B, C, and D, respectively. Among these, D peak is the well-known YL band, of which origin has been assigned to VGa and/or VGa O complexes [42,59–61]. The A peak energy was lower by 30 meV than the free A-exciton [29] energy .EA / at 100 K but it approached EA at 293 K being 3.407 eV. The residual electron concentration of the sample is estimated from the FWHM value of A peak at 100 K being 94 meV to be of the order of 1019 cm3 [65], which is higher than the Mott density [40]. Therefore, the A peak is assigned to a free-to-bound (FB) transition at 100 K and a free carrier recombination at 293 K. Note that the excitonic contribution cannot be ruled out. The energy of H peak was approximately 3.58 eV and 3.51 eV at 100 and 293 K, respectively, and is assigned to a free carrier recombination with a BM shift. Apparently, the two NBE emissions A and H may not coexist in a homogeneous bulk film. As the film had odd-shaped morphology, the spectra in Fig. 13.7a, b are considered to be convolutions of position-specific local spectra.
13.4 Effects of Dislocation Bending on the Optical Properties of Seeded GaN Substrates Grown by Ammonothermal Method Spatially-integrated or resolved CL was excited by a cw electron beam with or without the beam scanning, and dispersed by a grating monochromator .f D 20 cm/ equipped with SEM. The acceleration voltage and probe current were 15 kV and 1 nA, respectively. Monochromatic CL intensity mapping images were taken under a fixed wavelength with the beam scanning at 100 K. Spot-excitation CL spectra at 100 K, the corresponding SEM image, and monochromatic CL intensity mapping images for A D peaks of the Ga-polar film are summarized in Fig. 13.8. The local spectra numbered 1–6 were taken at respective spots in Fig. 13.8b. As shown, local spectra varied spot to spot. The overall CL intensity was weak at the spots 2, 3, and 6, as shown in Fig. 13.8c–f, in which white areas correspond to those emitting bright lights. Considering our experimental configuration, low light collection efficiency might be the cause for the spots 2 and 3. The reason for the low CL intensity of spot 6 is due to the presence of the central pit that corresponds to the apex of a TD, which initiates the facet growth. At the planar region 5, BL band (C band) intensity was strong and therefore the incorporation efficiency of a specific impurity responsible for C band is higher for the (0001) face. The most suspicious impurity is Mg or C, both of which were detected by SIMS measurement. The peak B at 3.27 eV was found at all spots 1–6. Although the peak energy was close to that of the NBE emission of zincblende GaN [or, basal plane stacking fault (SF) in wurtzite GaN], noticeable SF was not found, as shown in the cross-sectional TEM images in Fig. 13.9a, c. Therefore, the origin of B peak is assigned to an impurity such as Mg. The most interesting
13
Optical Properties of GaN Substrates
289
WAVELENGTH (nm) CL INTENSITY (arb units)
600 500
Ga-polar GaN
400 340
15kV, 1nA, 100K wide-area C
D
(a) CL spectra B A H
1 2 3 4 5 6
6 4
2.0 2.5 3.0 3.5
1 mm
2
1
5
3
(b) SEM top view
PHOTON ENERGY (eV)
6 4
2 3
1
6 5
4
2 3
1
5
(c) CL map A (3.44 eV) (d) CL map B (3.27 eV)
6 4
1
2 3
6 5
4
1
2 3
5
(e) CL map C (2.92 eV) (f) CL map D (2.23 eV)
Fig. 13.8 (a) Wide-area and spot-excitation CL spectra at 100 K, (b) representative local SEM image, and (c)–(f) corresponding monochromatic CL intensity mapping images recorded for various photon energies at 100 K of the 5 m-thick Ga-polar AT GaN film. The spectra numbered 1–6 in (a) were taken at the numbered spots in (b)–(f). In (c)–(f), white areas correspond to those emitting bright lights. (After [27])
feature is that A peak intensity was remarkably enhanced at the spots 1, 2, and 3, where edge TD density was reduced by the dislocation bending owing to the facet growth [56], as shown in Fig. 13.9c. The average TD density was approximately 109 cm2 at the planar region and lower than 106 cm2 at the crests of the ridges (spot 1 for example). Therefore, wide-area low TD density region may be prepared by this mechanism when thicker film is continuously grown. Similar to the case for the FS-GaN substrate prepared by HVPE [9], a flat GaN wafer can then be prepared by slicing and polishing thick layers or boules. In contrast to the Ga-polar film, lineshape of the local CL spectra of N-polar film was nearly position-independent, as shown in Fig. 13.10a. The result may reflect the featureless morphology of the film. However, the NBE peak exhibited a noticeable intensity contrast in the CL mapping image, as shown in Fig. 13.10c. As the diameter of circular shaped bright (white in the figure) areas ranged between smaller than a few hundred nm and 1 m, they are considered to correspond to those between TDs.
290
S.F. Chichibu N-polar GaN
Ga-polar GaN
a
b
g=0002
g=0002
c
g=1120
1.3 mm
d
g=1120
1.1 mm
Fig. 13.9 Cross-sectional TEM images of (a,c) Ga-polar and (b,d) N-polar AT GaN films. The g vectors are < 0002 > for (a) and (b), < 1N 1N 20 > for (c), and < 112N 0 > for (d). (After [19, 27])
N-polar GaN CL INTENSITY (arb. units)
WAVELENGTH (nm) 600 500
1 mm
1
2
400 340
3
15kV, 1nA, 100K wide-area 1 2 3 4
2.0 2.5 3.0 3.5 PHOTON ENERGY (eV)
(a) CL spectra
4
(b) SEM top view 1
2 3
4
(c) CL map (3.50 eV)
Fig. 13.10 (a) Wide-area and spot-excitation CL spectra at 100 K, (b) representative local SEM image, and (c) monochromatic CL intensity mapping images recorded at 3.50 eV at 100 K of the 4 m-thick N-polar AT GaN film. The spectra numbered 1–4 in (a) were taken at the numbered spots in (b). In (c), white areas correspond to those emitting bright lights. (After [27])
Indeed, the TD density estimated from Fig. 13.9b, d is in the order of 109 cm2 , which correspond to the TD spacing of approximately 316 nm.
13.5 Summary In summary, optical properties of GaN substrates prepared by three methods were surveyed to discuss the superiority and critical issues of each method. Both the FSGaN substrates prepared by MOVPE using LEO technique and by HVPE using FIELO technique exhibited excellent PL properties, in terms of the observation of free exciton fine structures in their low temperature PL spectra. The intensities of
13
Optical Properties of GaN Substrates
291
characteristic YL and BL bands, which originate from point defect complexes and impurities, were more than two orders of magnitude lower than that of the NBE emission at room temperature for both cases. In addition, exciton–polariton features were found in the low temperature PR spectra of the MOVPE FS-GaN substrate, which exhibited a long PL lifetime of 860 ps at 300 K. The results indicate that they can also be used as a seed crystal for AT growth, except for the wafer bowing problem originating from in-plain compressive strain. Then, spot-excitation local CL spectra of Ga- and N-polar AT GaN films were correlated with submicrometer scale morphological features. The N-polar film contained high density residual electrons due presumably to the incorporation of O and Si, which gave rise to pronounced BM shift in the free carrier recombination emission band. On the other hand, the Ga-polar film exhibited rich spectral and intensity variations originating from facet growth. As a result of dislocation bending during the facet growth, TD density was greatly reduced at the crests of ridges. The results encourage one to grow low TD density GaN substrates by growing thick layers and/or boules by AT method and subsequent slicing. Acknowledgements The author would like to thank S. Nakamura, S.P. DenBaars, J.S. Speck, U.K. Mishra, A. Uedono, and T. Sota for having long-lived collaborations. He is also thankful to T. Hashimoto, S. Keller, K. Fujito, H. Marchand, M.S. Minsky, P.T. Fini, J.P. Ibbetson, and S.B. Fleischer of UCSB and Y. Ishihara and A. Usui of Furukawa Co. Ltd. for providing the samples. He is also grateful to T. Onuma and M. Kubota for optical measurements. This work was supported in part by Grant-in-Aids of CANTech, IMRAM, Tohoku Univ., NEDO project by METI and Scientific Research in Priority Areas No. 18069001 under MEXT, Japan.
References 1. S. Strite, H. Morkoc, J. Vac. Sci. Technol. B 10, 1237 (1992) 2. S. Nakamura, G. Fasol, The Blue Laser Diode (Springer, Berlin, 1997) 3. I. Akasaki, H. Amano, Jpn. J. Appl. Phys. 36, 5393 (1997) 4. S. Nakamura, S.F. Chichibu (Eds.), Introduction to Nitride Semiconductor Blue Lasers and Light Emitting Diodes (Taylor and Francis, London and New York, 2000) 5. H. Yamane, M. Shimada, T. Sekiguchui, F.J. DiSalvo, J. Cryst. Growth 186, 8 (1998) 6. S. Porowski, MRS Internet J. Nitride Semicond. Res. 4S1, G1.3 (1999) 7. F. Kawamura, M. Morishita, K. Omae, M. Yoshimura, Y. Mori, T. Sasaki, Jpn. J. Appl. Phys. 42, L879 (2003) 8. S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa, T. Yamada, T. Matsushita, H. Kiyoku, Y. Sugimoto, T. Kozaki, H. Umemoto, M. Sano, K. Chocho, Jpn. J. Appl. Phys. 37, L309 (1998) 9. K. Motoki, T. Okahisa, S. Nakahata, N. Matsumoto, H. Kimura, H. Kasai, K. Takemoto, K. Uematsu, M. Ueno, Y. Kumagai, A. Koukitu, H. Seki, J. Cryst. Growth 237–239, 912 (2002) 10. Y. Oshima, T. Eri, M. Shibata, H. Sunakawa, K. Kobayashi, T. Ichihashi, A. Usui, Jpn. J. Appl. Phys. 42, L1 (2003) 11. K. Maeda, M. Sato, I. Niikura, T. Fukuda, Semicond. Sci. Technol. 20, S49 (2005) 12. R. Dwilinski, R. Doradzinski, J. Garczynski, L. Sierzputowski, M. Palczewska, A. Wysmolek, M. Kaminska, MRS Internet J. Nitride Semicond. Res. 3, 25 0072 13. Y.C. Lan, X.L. Chen, Y.P. Xu, Y.G. Cao, F. Huang, Mater. Res. Bull. 25, 2325 (2000) 14. D.R. Ketchum, J.W. Kolis, J. Cryst. Growth 222, 431 (2001) 15. A.P. Purdy, R.J. Jouet, C.F. George, Cryst. Growth Des. 2, 141 (2002)
292
S.F. Chichibu
16. M.J. Callahan, B. Wang, L.O. Bouthillette, S.Q. Wang, J.W. Kolis, D.F. Bliss, Mater. Res. Soc. Symp. Proc. 798, Y2.10 (2004) 17. A. Yoshikawa, E. Ohshima, T. Fukuda, H. Tsuji, K. Ohshima, J. Cryst. Growth 260, 67 (2004) 18. T. Hashimoto, K. Fujito, B.A. Haskell, P.T. Fini, J.S. Speck, S. Nakamura, J. Cryst. Growth 275, e525 (2005) 19. T. Hashimoto, K. Fujito, F. Wu, B.A. Haskell, P.T. Fini, J.S. Speck, S. Nakamura, Mater. Res. Soc. Symp. Proc. 831, E2.8.1 (2005) 20. T. Hashimoto, K. Fujito, M. Saito, J.S. Speck, S. Nakamura, Jpn. J. Appl. Phys. 44, L1570 (2005) 21. T. Hashimoto, F. Wu, J.S. Speck, S. Nakamura, Jpn. J. Appl. Phys. 46, L525 (2007) 22. T. Hashimoto, F. Wu, J.S. Speck, S. Nakamura, Jpn. J. Appl. Phys. 46, L889 (2007) 23. B.A. Haskell, F. Wu, M.D. Craven, S. Matsuda, P.T. Fini, T. Fujii, K. Fujito, S.P. DenBaars, J.S. Speck, S. Nakamura, Appl. Phys. Lett. 83, 644 (2003) 24. D. Ehrentraut, Y. Kagamitani, T. Fukuda, F. Orito, S. Kawabata, K. Katano, S. Terada, J. Cryst. Growth 310, 3902 (2008) 25. R. Dwilinski, R. Doradzinski, J. Garczynski, L. P. Sierzputowski, A. Puchalski, Y. Kanbara, K. Yagi, H. Minakuchi, H. Hayashi, J. Cryst. Growth 310, 3911 (2008) 26. R. Dwilinski, R. Doradzinski, J. Garczynski, L. Sierzputowski, A. Puchalski, M. Rudzinski, M. Zajac, presented at International Workshop on Nitride Semiconductors 2008, Oct. 6–10, Montreux, Switzerland, No. We1-C1 (2008) 27. S.F. Chichibu, T. Onuma, T. Hashimoto, K. Fujito, F. Wu, J.S. Speck, S. Nakamura, Appl. Phys. Lett. 91, 251911 (2007) 28. S.F. Chichibu, H. Marchand, M.S. Minsky, S. Keller, P.T. Fini, J.P. Ibbetson, S.B. Fleischer, J.S. Speck, J.E. Bowers, E. Hu, U.K. Mishra, S.P. DenBaars, T. Deguchi, T. Sota, S. Nakamura, Appl. Phys. Lett. 74, 1460 (1999) 29. S.F. Chichibu, K. Torii, T. Deguchi, T. Sota, A. Setoguchi, H. Nakanishi, T. Azuhata, S. Nakamura, Appl. Phys. Lett. 76, 1576 (2000) 30. T. Kashiwagi, S. Sonoda, H. Yashiro, Y. Ishihara, A. Usui, Y. Akasaka, M. Hagiwara, Jpn. J. Appl. Phys. 46, 581 (2007) 31. A.A. Yamaguchi, Y. Mochizuki, H. Sunakawa, A. Usui, J. Appl. Phys. 83, 4542 (1998) 32. A.A. Yamaguchi, Y. Mochizuki, C. Sasaoka, A. Kimura, M. Nido, A. Usui, Mater. Res. Soc. Symp. Proc. 482, 893 (1998) 33. J.J. Hopfield, Phys. Rev. 112, 1555 (1958) 34. B. Gil, S. Clur, O. Briot, Solid State Commun. 104, 267 (1997) 35. P. Stepniewski, K.P. Korona, A. Wysmolek, J.M. Baranowski, K. Pakula, M. Potemski, I. Gregory, S. Porowski, Phys. Rev. B 56, 15151 (1997) 36. K. Torii, T. Deguchi, T. Sota, K. Suzuki, S. Chichibu, S. Nakamura, Phys. Rev. B 60, 4723 (1999) 37. M. Cardona, Modulation Spectroscopy, Solid State Physics. Suppl. 11, ed. by S. Seitz, D. Turnbull, H. Ehrenreich (Academic, New York, 1969) 38. D.E. Aspnes, Handbook on Semiconductors, Vol. 2, Chap. 4A, ed. by T.S. Moss (North-Holland, Amsterdam, 1980), p. 109 39. D.E. Aspnes, Surf. Sci. 37, 418 (1973) 40. S. Chichibu, T. Azuhata, T. Sota, S. Nakamura, J. Appl. Phys. 79, 2784 (1996) 41. A. Uedono, S.F. Chichibu, Z.Q. Chen, M. Sumiya, R. Suzuki, T. Ohdaira, T. Mikado, T. Mukai, S. Nakamura, J. Appl. Phys. 90, 181 (2001) 42. S.F. Chichibu, A. Uedono, T. Onuma, T. Sota, B.A. Haskell, S.P. DenBaars, J.S. Speck, S. Nakamura, Appl. Phys. Lett. 86, 021914 (2005) 43. S.F. Chichibu, A. Uedono, T. Onuma, B.A. Haskell, A. Chakraborty, T. Koyama, P.T. Fini, S. Keller, S.P. DenBaars, J.S. Speck, U.K. Mishra, S. Nakamura, S. Yamaguchi, S. Kamiyama, H. Amano, I. Akasaki, J. Han, T. Sota, Nat. Mater. 5, 810 (2006) 44. W. Shan, T.J. Shmidt, X.H. Yang, S.J. Hwang, J.J. Song, B. Goldenberg, Appl. Phys. Lett. 66, 985 (1995) 45. S. Chichibu, T. Mizutani, T. Shioda, H. Nakanishi, T. Deguchi, T. Azuhata, T. Sota, S. Nakamura, Appl. Phys. Lett. 70, 3440 (1997)
13
Optical Properties of GaN Substrates
293
46. G.D. Cody, T. Tiedje, B. Abeles, Y. Goldstein, Phys. Rev. Lett. 47, 1480 (1981) 47. L. Vina, S. Logothetidis, M. Cardona, Phys. Rev. B 30, 1979 (1984) 48. Y.P. Varshni, Physica 34, 149 (1967) 49. T. Azuhata, T. Matsunaga, K. Shimada, K. Yoshida, T. Sota, K. Suzuki, S. Nakamura, Physica B 219–220, 493 (1996) 50. H. Marchand, J. Ibbetson, P.T. Fini, P. Kozodoy, S. Keller, S.P. DenBaars, J.S. Speck, U.K. Mishra, Mater. Res. Soc. Internet J. Nitride Semicond. Res. 3, 3 (1998) 51. H. Marchand, X.H. Wu, J. Ibbetson, P.T. Fini, P. Kozodoy, S. Keller, J.S. Speck, S.P. DenBaars, U.K. Mishra, Appl. Phys. Lett. 73, 747 (1998) 52. S. Chichibu, A. Shikanai, T. Azuhata, T. Sota, A. Kuramata, K. Horino, S. Nakamura, Appl. Phys. Lett. 68, 3766 (1996) 53. A. Shikanai, T. Azuhata, T. Sota, S. Chichibu, A. Kuramata, K. Horino, S. Nakamura, J. Appl. Phys. 81, 417 (1997) 54. S. Chichibu, T. Azuhata, T. Sota, H. Amano, I. Akasaki, Appl. Phys. Lett. 70, 2085 (1997) 55. S.F. Chichibu, A. Uedono, T. Onuma, B.A. Haskell, A. Chakraborty, T. Koyama, P.T. Fini, S. Keller, S.P. DenBaars, J.S. Speck, U.K. Mishra, S. Nakamura, S. Yamaguchi, S. Kamiyama, H. Amano, I. Akasaki, J. Han, T. Sota, Philos. Mag. 87, 2019 (2007) 56. A. Usui, H. Sunakawa, A. Sakai, A.A. Yamaguchi, Jpn. J. Appl. Phys. 36, L899 (1997) 57. A. Wysmolek, M. Potemski, R. Stepniewski, J.M. Baranowski, D.C. Look, S.K. Lee, J.Y. Han, Phys. Status Solidi B 235, 36 (2003) 58. T. Azuhata, T. Sota, K. Suzuki, S. Nakamura, J. Phys. Condens. Matter 7, 129 (1995) 59. J. Neugebauer, C.G. Van de Walle, Phys. Rev. B 50, 8067 (1994) 60. J. Neugebauer, C.G. Van de Walle, Appl. Phys. Lett. 69, 503 (1996) 61. K. Saarinen, T. Laine, S. Kuisma, J. Nissilä, P. Hautojärvi, L. Dobrzynski, J. M. Baranowski, K. Pakula, R. Stepniewski, M. Wojdak, A. Wysmolek, T. Suski, M. Leszczynski, I. Grzegory, S. Porowski, Phys. Rev. Lett. 79, 3030 (1997) 62. P. Hacke, T. Detchprohm, K. Hiramatsu, N. Sawaki, K. Tadatomo, K. Miyake, J. Appl. Phys. 76, 304 (1994) 63. Z.-Q. Fang, D.C. Look, J. Jasinski, M. Benamara, Z. Liliental-Weber, R.J. Molnar, Appl. Phys. Lett. 78, 332 (2001) 64. E. Burstein, Phys. Rev. 93, 632 (1954) 65. E. Iliopoulos, D. Doppalapudi, H.M. Ng, T.D. Moustakas, Appl. Phys. Lett. 73, 375 (1998)
Chapter 14
Point Defects and Impurities in Bulk GaN Studied by Positron Annihilation Spectroscopy Filip Tuomisto
Abstract Positron annihilation spectroscopy is an experimental technique that allows for the selective detection of vacancy defects in semiconductors, providing a means to both identify and quantify them. This chapter first gives a short introduction to positron annihilation techniques, and then proceeds to present an overview of the positron studies of vacancy defects and impurities in bulk and quasi-bulk GaN. Both the in-grown and processing induced defects are addressed.
14.1 Introduction A variety of techniques can be applied to identify defects in semiconductors on the atomic scale. The main advantage of positron annihilation spectroscopy lies in its ability to selectively detect vacancy-type defects. The positron has two special properties that give this asset: it has a positive charge, and it annihilates with electrons. An energetic positron that has penetrated into a solid rapidly loses its energy and then lives a few hundred picoseconds in thermal equilibrium with the environment. During its thermal motion, the positron interacts with defects, which may lead to its getting trapped in a localized state. Hence, the final positron annihilation with an electron can happen from various states. The sensitivity of positrons to vacancy-type defects is rather easy to understand. A free positron in a crystal lattice feels strong repulsion from the positive ion cores, and an open-volume defect like a vacant lattice site is hence an attractive center where the positron gets trapped. The reduced electron density at the vacant site increases the positron lifetime, while the missing valence and core electrons cause substantial changes in the momentum distribution of the annihilating electrons. Two positron techniques have been efficiently used in defect studies in semiconductors, namely the positron lifetime and the Doppler broadening of the 511 keV line. These methods provide a straightforward tool for the identification of vacancy-type defects. In addition, the positron techniques are strongly supported by theory, as the annihilation characteristics can be calculated from first principles. A significant advantage of positron annihilation spectroscopy is that it can be applied to both bulk crystals and thin layers of any electrical conduction type. 295
296
F. Tuomisto
Vacancy defects in GaN can exist in both sublattices, those in the Ga sublattice (called Ga vacancies) being more likely to be observed by positrons because of their larger size and nonpositive charge states. Defect studies in GaN using positron annihilation spectroscopy have been carried out mostly by one research group, namely the one at TKK, since the late 1990s. The aim of this chapter is to give an overview of the results on vacancies and other point defects obtained with positron annihilation spectroscopy in bulk GaN, grown by either the high nitrogen pressure (HNP) method or by hydride vapor phase epitaxy (HVPE). The focus will be on the results obtained on the vacancy defects in bulk GaN; the principles of positron annihilation spectroscopy are described very shortly – for a full picture of positron annihilation in semiconductors, the reader is referred to earlier review articles [1–4], books, and book chapters [5–10], to the conference proceedings of the International Conference on Positron Annihilation (ICPA), and to other references therein. This chapter is organized as follows: The principles of positron annihilation spectroscopy are very briefly reviewed in Sect. 14.2. In Sect. 14.3, the results obtained on in-grown defects in HNP [11–17] and HVPE [16–29] GaN are reviewed. The positron results on artificial creation (e.g., by irradiation or implantation) and manipulation (e.g., by thermal annealing) of point defects in bulk GaN [30–35] are reviewed in Sect. 14.4. Finally, some concluding remarks are made in Sect. 14.5, summarizing the chapter.
14.2 Positron Annihilation Spectroscopy This section gives a brief overview of the principles of positron annihilation spectroscopy and of the experimental techniques. A thermalized positron in a crystal lattice behaves like a free electron or a hole. Analogously, positrons have shallow hydrogenic states at negative ions such as acceptor impurities. Furthermore, vacancies and other centers with open volume act as deep traps for positrons. These defects can be experimentally detected by measuring either the positron lifetime or the momentum density of the annihilating positron–electron pairs (Doppler broadening of the annihilation radiation). For the sake of clarity, we will concentrate on the measurements of those two quantities, as these methods are the most used in defect studies in semiconductors. Descriptions of other techniques can be found in, e.g., [9].
14.2.1 Positrons in Solids Positrons are easily obtained from radioactive .ˇ C / isotopes such as 22 Na, where the positron emission is accompanied by a 1.27 MeV photon. This photon is used in positron lifetime experiments as the time signal of the positron emission from the source. The stopping profile of positrons from the ˇ C emission is exponential
14
Point Defects and Impurities in Bulk GaN
297
[6, 36]: for the 22 Na source where the positron energy distribution extends to Emax D 0:54 MeV, the positron mean stopping depth is about 40 m in GaN. The positrons emitted directly from a radioactive source thus probe the bulk of a solid. Low-energy positrons are needed for studying thin overlayers and near-surface regions. Positrons from the ˇ C emission are first slowed down and thermalized in a moderator, consisting usually of a thin film placed in front of the positron source and made of a material (e.g., W) that has a negative affinity for positrons. Thermalized positrons close to the moderator surface are emitted into the vacuum with energy of the order of 1 eV, and a beam is formed using electric and magnetic fields. The positron beam can be accelerated to a tunable energy ranging from a few tens of eV to 100 keV (corresponding to mean implantation depths from a few nanometers to more than 10 m), giving the possibility to control the positron stopping depth in the sample. After implantation and rapid (taking a few picoseconds) thermalization, positrons in semiconductors behave like free carriers, i.e., the positron state is a Bloch state in a defect-free lattice (living for a few hundred picoseconds). Various positron states yield specific annihilation characteristics that can be experimentally observed in the positron lifetime and Doppler broadening experiments. The positron wave function can be calculated from a one-particle Schrödinger equation [3], where the positron potential consists of two parts: the electrostatic Coulomb potential and a term that takes into account the electron–positron correlation effects. Many practical schemes exist for solving the positron state ‰C from the Schrödinger equation [3, 37, 38]. A positron state can be experimentally characterized by measuring the positron lifetime and the momentum distribution of the annihilation radiation. These quantities can also be calculated once the corresponding electronic structure of the solid system is known. The positron annihilation rate , the inverse of the positron lifetime , is proportional to the overlap of the electron and positron densities: 1= D D r02 c
Z
d r j‰C .r/j2 n.r/ Œn.r/ ;
(14.1)
where r0 is the classical electron radius, c the velocity of light, n.r/ the electron density, and Œn the enhancement factor that accounts for the pileup of the electron density at the positron beyond the (average) density n.r/ [3]. The momentum distribution .p/ of the annihilation radiation is a nonlocal quantity and requires knowledge of all the electron wave functions ‰i overlapping with the positron. It can be written in the form ˇ2 ˇZ p ˇ r0 c X ˇˇ i pr ‰C .r/‰i .r/ i .r/ˇˇ ;
.p/ D dre ˇ V
(14.2)
i
where V is the normalization volume and i .r/ may depend only on i or on r. The Doppler broadening experiment measures the longitudinal momentum distribution along the direction of the emitted 511 keV photons, defined here as the z-axis:
298
F. Tuomisto
Z1 Z1
.pL / D
dpx dpy .p/ :
(14.3)
1 1
It should be noted that the momentum distribution .p/ of the annihilation radiation is mainly that of the annihilating electrons “seen by the positron,” because the momentum of the thermalized positron is negligible.
14.2.2 Positrons at Defects Positron States at Defects In analogy to free carriers, the positron also has localized states at lattice imperfections. At vacancy-type defects where one or several ions are missing, the repulsion sensed by the positron is lowered and the positron feels these kinds of defects as potential wells. As a result, localized positron states at open-volume defects are formed. The positron ground state at a vacancy-type defect is generally deep, the binding energy is about 1 eV or more [3]. In a vacancy defect, the electron density is locally reduced. This is reflected in the positron lifetimes, which are longer than in the defect-free lattice. For example, the calculated lifetimes in the unrelaxed Ga and N vacancies are 55 and 5 ps longer than in the perfect lattice [13]. The longer positron lifetime at VGa is due to the larger open volume compared with that of VN . The positron lifetime measurement is thus a probe of vacancy defects in materials. Positron annihilation at a vacancytype defect leads also to changes in the momentum distribution .p/ probed by the Doppler broadening experiment. The momentum distribution arising from valence electron annihilation becomes narrower due to a lower electron density. In addition, the localized positron at a vacancy has a reduced overlap with ion cores leading to a considerable decrease in annihilation with high momentum core electrons. In the case of GaN, where the dominant contribution to the high momentum part of the distribution comes from the Ga 3d electrons, the changes in the momentum distribution are more pronounced when positrons are trapped at Ga vacancies. The comparison of the measured positron lifetimes and Doppler broadening spectra with the theoretically calculated data for specific defects provides a very efficient tool for identification of the observed vacancy defects. Several ab initio approaches have been studied in recent years (see [37] and the references therein). The agreement between theory and experiment is excellent in terms of differences or ratios between the data for defects and the perfect lattice. Also quantitative agreement has been obtained in many materials, but all the theoretical schemes applied so far seem to have problems with the treatment of the partially bound 3d electrons, either under- or overestimating their contribution to the positron annihilation data. For example, in the case of GaN, the calculated lifetimes in the perfect lattice range from about 130 to 160 ps [13, 26], depending on the scheme, while the
14
Point Defects and Impurities in Bulk GaN
299
experimentally determined bulk lifetime is 160 ps [11,17,26,34]. On the other hand, the calculated differences in the lifetimes and the ratios of the momentum distributions between the defect and bulk states agree very well with the experiments [26, 37, 39]. A negatively charged impurity atom or an intrinsic point defect can bind positrons at shallow states even if these defects do not contain open volume [40, 41]. Being a positive particle, the positron can be localized at the hydrogenic (Rydberg) state of the Coulomb field around a negatively charged center. The situation is analogous to the binding of an electron to a shallow donor atom. The positron binding energy is typically Eion D 30–120 meV, indicating that positrons are thermally excited from the Rydberg states at 100–300 K. The hydrogenic positron state around a negative ion has a typical extension of 10–100 Å and thus positrons probe the same electron density as in the defect-free lattice. As a consequence, the annihilation characteristics (positron lifetime, positron–electron momentum distribution) are not different from those in the lattice. Although the negative ions cannot be identified with the experimental parameters, information on their concentration can be obtained in the positron lifetime and Doppler broadening experiments when they compete with vacancies in positron trapping.
Positron Trapping at Defects The transition of a positron from a free Bloch state to a localized state at a defect is called positron trapping. The trapping is analogous to carrier capture. However, in order to be observed, it must be fast enough to compete with annihilation. The positron trapping rate D onto a defect D is proportional to the defect concentration cD through D D D cD . The trapping coefficient D depends on the defect and the host lattice. As the positron binding energy at vacancies is typically > 1 eV, the thermal escape (detrapping) of positrons from the vacancies can be usually neglected. Because of the Coulomb repulsion, the trapping coefficient at positively charged vacancies is so small that the trapping does not occur during the short positron lifetime of a few hundred picoseconds [42]. Hence, the detection of vacancies or other defects in a positive charge state is impossible with positron annihilation techniques. The trapping coefficient at neutral vacancies is typically D 1014 –1015 s1 independently of temperature [42–44]. This value means that neutral vacancies are observed when their concentration is 1016 cm3 (i.e., about 100–200 ppb). The positron trapping coefficient at negative vacancies is typically D 1015 –1016 s1 at room temperature [42–44]. The sensitivity to detect negative vacancies is thus 1015 cm3 (i.e., about 10–20 ppb). The experimental fingerprint of a negative vacancy is the increase of D with decreasing temperature .D T 1=2 / [42– 44], which allows to experimentally distinguish negative vacancy defects from neutral ones. The positron trapping coefficient ion at the hydrogenic states around negative ions is of the same order of magnitude as that at negative vacancies [41, 45].
300
F. Tuomisto
Furthermore, the trapping coefficient exhibits a similar T 1=2 temperature dependence. Unlike in the case of vacancy defects, the thermal escape of positrons from the negative ions plays a crucial role at usual experimental temperatures. The principle of detailed balance yields the following equation for the detrapping rate ıion from the hydrogenic state [3]: ıion D ion
2 m kB T h2
3=2
Eion exp : kB T
(14.4)
Typically, ion concentrations above 1016 cm3 influence positron annihilation at low temperatures .T < 100 K/, but the ions are not observed at high temperatures .T > 300 K/, where the detrapping rate (14.4) is large. In practice, the positron annihilation data is analyzed in terms of kinetic rate equations describing the positron transitions between the free Bloch states and localized states at defects [8–10]. Very often, the experimental data show the presence of two defects, one of which is a vacancy and the other is a negative ion. In this case, the average lifetime ave (the center of mass of the lifetime spectrum), positron–electron momentum distribution .pL /, and shape parameters S and W of the Doppler broadened annihilation line (representing annihilations with low momentum valence electrons and high momentum core electrons, respectively) are given by P D B PB C ion Pion C V PV ;
(14.5)
Here P is the experimental parameter and the subscripts “B, ion, and V” denote the characteristic parameters of the bulk, negative ions, and vacancies, respectively. Equation (14.5) allows the experimental determination of the trapping rates V and ion related to the experimental trapping fractions through B D 1 ion V ; V V D ; B C V C 1Cıionion=ion ion ion D .1 C ıion =ion / B C V C
(14.6) (14.7)
ion 1Cıion =ion
:
(14.8)
and consequently the defect concentrations can be obtained. Furthermore, these equations enable the combination of positron lifetime and Doppler broadening results and various correlations between ave ; .pL /; S , and W can be studied. At high temperatures, all positrons escape from the hydrogenic state of the negative ions and no annihilations take place at them. Then ion D 0 and ıion =ion >> 1, and hence the determination of the positron trapping rate and vacancy concentration is straightforward using (14.5–14.8) V D V cV D B
ave B S SB W WB D B D B : V ave SV S WV W
(14.9)
14
Point Defects and Impurities in Bulk GaN
301
It should be noted that in this case, ave ; S , and W depend linearly on one another. The linearity of experimental points in the .ave ; S /; .ave ; W /, and (S, W) plots provides thus evidence that positrons annihilate from two distinguishable states, indicating that they are trapped at only a single type of vacancy defect in the samples.
14.2.3 Experimental Techniques Positron lifetime spectroscopy is a powerful technique in defect studies, because the various positron states appear as different exponential decay components. The number of positron states, their annihilation rates, and relative intensities can be determined. In a positron lifetime measurement, one needs to detect the start and stop signals corresponding to the positron entrance and annihilation times in the sample, respectively. The 1.27 MeV photon that follows the positron emission from the 22 Na isotope is a suitable start signal. The 511 keV annihilation photon serves as the stop signal. The positron source is prepared by sealing about 10 Ci (about 105 –106 Bq) of radioactive isotope between two thin foils. The source is then sandwiched between two identical pieces (e.g., 5 5 0:5 mm3 ) of the sample material. This technique is standard for bulk crystal studies. Pulsed positron beams have been constructed for lifetime spectroscopy in thin layers [46, 47], but so far they have not been used much in defect studies. Typically, about 106 lifetime events are recorded in 1 h. The experimental spectrum represents the probability of positron annihilation at time t and it consists of exponential decay components:
X dn.t/ Ii i ei t ; D dt
(14.10)
i
where n.t/ is the probability of the positron to be alive at time t. The decay constants i D 1=i are called annihilation rates and they are the inverses on the positron lifetimes i . Each positron lifetime has the intensity of Ii . In practice, the ideal spectrum of (14.10) is convoluted by a Gaussian resolution function which has a width of 200– 250 ps (full width at half maximum, FWHM). About 5–10% of positrons annihilate in the source material and proper “source corrections” must be made. Because of the finite time resolution, annihilations in the source materials, and random background, typically only 1–3 lifetime components can be resolved in the analysis of the experimental spectra. The separation of two lifetimes is successful only if the ratio œ1 =œ2 is 1.3–1.5. Figure 14.1 shows positron lifetime spectra recorded in two freestanding GaN samples grown by HVPE and the HNP method [17, 34]. Positrons enter the sample and thermalize at the time t D 0. The vertical axis of Fig. 14.1 gives the number of annihilations at a time interval of 25 ps. In the HVPE-GaN sample where the O concentration is low, the positron annihilation spectrum has a single lifetime
302
F. Tuomisto
105 n-type HNP GaN [O] ~ 1020 cm–3 τave = 190 ps
Number of counts
104
103 n-type HVPE GaN [O] ~ 1016 cm–3 τave = 160 ps
102
101 0.0
0.5
1.0 Time (ns)
1.5
2.0
Fig. 14.1 Examples of positron lifetime spectra in free-standing GaN crystals grown by two different methods with different concentrations of unintentional oxygen: hydride vapor phase epitaxy (HVPE) and the high nitrogen pressure (HNP) method. A constant background and annihilations in the source materials have been subtracted from the spectra, which consist of about 2106 recorded annihilation events. The solid lines are fits of the sum of exponential components convoluted with the resolution function of the spectrometer. The data in samples were recorded at 300 K. In the case of the HVPE sample, the annihilation spectrum has only a single component of 160 ˙ 1 ps. The spectrum in the HNP sample can be decomposed into two components of which the higher is 2 D 235 ˙ 5 ps
component of 160 ˙ 1 ps at 300 K corresponding to positron annihilations in the defect-free lattice. The spectrum measured in the HNP-GaN sample with high O content has two lifetime components, the longer of which .2 D 235 ps/ is due to positrons annihilating at Ga vacancies. The experimental results are often presented in terms of the average positron lifetime ave defined as Z1 ave D 0
Z1 X dn D dt n.t/ D dt t Ii i : dt 0
(14.11)
i
The average lifetime is a statistically accurate parameter, because it is equal to the center-of-mass of the experimental lifetime spectrum. Hence, it can be correctly calculated from the intensity and lifetime values even if the decomposition represented only a good fit to the experimental data without any physical meaning. For example, the positron average lifetimes in the two spectra of Fig. 14.1 are 160 ps
14
Point Defects and Impurities in Bulk GaN
303
(HVPE-GaN) and 190 ps (HNP-GaN). The difference is very significant as changes in the average lifetime less than 1 ps can be reliably observed in the experiments. Doppler broadening spectroscopy is often applied especially in the low-energy positron beam experiments, where lifetime spectroscopy is usually very difficult because of the missing start signal. The motion of the annihilating electron–positron pair causes a Doppler shift in the annihilation radiation with E D cpL =2, where pL is the longitudinal momentum component of the pair in the direction of the annihilation photon emission. This causes the broadening of the 511 keV annihilation line. The shape of the 511 keV peak gives thus the one-dimensional momentum distribution .pL / of the annihilating electron–positron pairs. A Doppler shift of 1 keV corresponds to a momentum value of pL D 0:54 a.u. .3:91 103 m0 c/. The Doppler broadening can be experimentally measured using a Ge gamma ray detector with a good energy resolution. For measurements of bulk samples, the same source–sample sandwich is used as in the lifetime experiments. For layer studies, the positron beam hits the sample and the Doppler broadening is often monitored as a function of the beam energy. The typical resolution of a detector is around 1–1.5 keV at 500 keV. This is considerable compared to the total width of 2–3 keV of the annihilation peak meaning that the experimental lineshape is strongly influenced by the detector resolution. Therefore, various shape parameters are used to characterize the 511 keV line. The low electron-momentum parameter S is defined as the ratio of the counts in the central region (typically pL < 0:4 a.u.) of the annihilation line to the total number of the counts in the line. In the same way, the high electron-momentum parameter W is the fraction of the counts in the wing regions of the line (typically pL > 1:5 a.u.). A good way of choosing the integration windows for the S and W parameters is best seen in Fig. 14.2. The S parameter is integrated from the Gaussian part (given by the momentum distribution of the unbound or only weakly bound valence electrons) so that it includes roughly 50% of the total counts in the peak. The lower limit of the W parameter window is chosen so that the dominant contribution to that part of the spectrum comes from the exponential tails (linear in the semilog plot) of the core electron distributions. In order to have as good statistics as possible, the upper limit is set as high as reasonable from the data scatter point of view. The proper choice of the lower limit of the W parameter window depends on the studied material. Because of the above considerations S and W are sometimes called the valence and core annihilation parameters, respectively. The high-momentum part of the Doppler broadening spectrum arises from annihilations with core electrons that contain information on the chemical identity of the atoms. Thus, the detailed investigation of core electron annihilation can reveal the nature of the atoms in the regions where positrons annihilate. In order to study the high-momentum part in detail, the experimental background needs to be reduced. A second gamma ray detector can be placed opposite to the Ge detector and the only events that are accepted are those for which both 511 keV photons are detected [48, 49]. Depending on the type of the second detector, electron momenta even up to p 8 a.u. .60 103 m0 c/ can be measured with the coincidence detection of the Doppler broadening (see Fig. 14.2).
304
F. Tuomisto Electron momentum (10–3 m0c) 10 20
0
30
Momentum distribution (a.u.)
1 Gallium nitride GaN lattice Ga vacancy 0.1
0.01
0.001
S 0
W 1 2 3 Electron momentum (a.u.)
4
Fig. 14.2 The positron–electron momentum distribution measured in the GaN lattice and in an electron irradiation-induced Ga vacancy. Possible definitions of the integration areas for the S and W parameters are shown. It should be noted that the Doppler broadened spectrum is symmetrical: it has been folded around zero momentum in order to increase statistics
14.3 In-Grown Defects GaN crystals grown by HVPE and the HNP method, both doped and nominally undoped, have been the subject of numerous positron defect studies since the first identification of the native Ga vacancy in GaN [11]. These samples are characterized by various impurity profiles and growth temperatures, giving rise to different vacancy concentrations. Furthermore, the growth polarity has an important role in determining both the incorporation of impurities and the formation of extended and point defects in the material. In the following, an attempt is made to gather the pieces of understanding presented in [11–13, 15–20, 23–29], providing a consistent picture of the formation of vacancy defects in bulk (or bulk-like) GaN growth.
14.3.1 Defect Formation: Growth Methods and Doping Figure 14.3 presents the average positron lifetime ave measured as a function of temperature in six differently doped samples grown by HVPE and HNP. In all samples where ave is above the bulk lifetime B D 160 ps indicating that positrons annihilate as trapped at vacancy defects, the annihilation spectrum is two-componential with the higher component being 2 D 235 ˙ 10 ps (not shown here). This component is characteristic of the Ga vacancy [11, 17, 24], and hence Ga vacancies are
Point Defects and Impurities in Bulk GaN
Fig. 14.3 The average positron lifetime measured as function of temperature in selected HVPE-GaN and HNP-GaN samples
305 undoped HNP GaN, [O] = 4 × 1019 cm–3 Mg-doped HNP GaN, [Mg] = 6 × 1019 cm–3
Average positron lifetime τave (ps)
14
Mg-doped HNP GaN, [Mg] = 1 × 1020 cm–3 210 200 O-doped HVPE GaN, [O] = 5 × 1019 cm–3 O-doped HVPE GaN, [O] = 4 × 1018 cm–3
190
undoped HVPE GaN, [O] = 1 × 1016 cm–3 180 170 160 0
100 200 300 400 Measurement temperature (K)
500
present at concentrations above 1015 cm3 (the sensitivity limit of the temperaturedependent positron measurements) in these samples. The slight increase of the average positron lifetime with increasing temperature in the heavily Mg-doped HNP-GaN sample, where positrons annihilate from the delocalized state in the defect-free lattice, is due to the thermal expansion of the GaN lattice. In the HVPE-GaN samples, the average positron lifetime increases with decreasing measurement temperature. This is a fingerprint of negatively charged vacancies, the trapping coefficient of which increases at low temperatures (see Sect. 14.2). Hence, the Ga vacancies are negatively charged in n-type HVPE-GaN, irrespective of the O content. In the nominally undoped and less heavily Mg-doped HNP-GaN samples the average positron lifetime decreases with decreasing measurement temperature. This indicates that negative ions are abundant enough to compete with vacancies in positron trapping at low temperatures, where the escape rate from the shallow Rydberg states is negligible. The data also show that the vacancies are negatively charged, as in the case of neutral vacancies the decrease would be faster in the Mg-doped sample, and the average positron lifetime could not stabilize at a value higher than the bulk lifetime in the nominally undoped sample. It should be noted that the negative ions cannot be identified on the basis of the positron data alone. However, their concentrations can be determined and compared to acceptor impurity concentrations measured with secondary ion mass spectrometry (SIMS). In the above HNP-GaN samples, this comparison has allowed the identification of the negative ions as Mg impurities [12]. The temperature dependence of the average lifetime can be modeled with kinetic trapping equations introduced in Sect. 14.2. The positron trapping coefficients at negative Ga vacancies V and negative ions ion vary as T 1=2 as a function of
306
F. Tuomisto
Table 14.1 Defect concentrations (given in 1017 cm3 ) in selected bulk or quasi-bulk GaN samples grown with different methods, doping and polarities [12, 17, 24, 28]. The impurity data were obtained by SIMS ŒVGa
[Mg]
cion
2 0.7