Perspectives about growth of bulk gallium nitride crystals, fabricating high structural quality gallium nitride wafers and the market demand for them are presented. Three basic crystal growth technologies, halide vapor phase epitaxy, sodium flux, and ammonothermal, are described. Their advantages and disadvantages, recent development, and possibilities are discussed. The main difficulty with crystallization of thick GaN is determined. Some new solutions for bulk growth are proposed. It is shown that only crystallization on high structural quality native seeds will ensure proper progress. New ideas for fabricating gallium nitride crystals and wafers with a better control of their structural properties and point defect concentration are proposed.
INTRODUCTION
The 50th anniversary of the first paper devoted to vapor deposition of single-crystalline gallium nitride (GaN) was celebrated last year.1 Halide vapor phase epitaxy (HVPE), the technology previously known from crystal growth of gallium arsenide (GaAs), gallium phosphide (GaP), and gallium antimonide (GaSb), was applied by Maruska and Tietjen to crystallize GaN. A direct bandgap energy of 3.39 eV was determined. A sapphire single crystal was used as a seed. A crystalline layer of a very high free electron concentration, exceeding 1019 cm−3, was grown. Such a high value was explained by the presence of nitrogen vacancies. Today, it is known that the free carrier concentration should have rather been correlated with oxygen, one of the main unintentional dopants in GaN. In the 1960s and 1970s, commercially available ammonia (NH3) (gas used for the synthesis of GaN by HVPE) had a water content as high as 1000 ppm.2 Therefore, Maruska and Tietjen were not able to crystallize high-purity GaN. The oxygen concentration had to be extremely high. No doubt, the development in the chemical industry, including increased purity of commercially available gases, allowed for the progress in the nitride semiconductor science and technology to occur.
The pioneering work on nitride semiconductors took place in the 1980s. It allowed us to crystallize hetero-epitaxial GaN-based quantum structures and fabricate efficient blue Light Emitting Diodes (LEDs). For this work, three distinguished Japanese professors, I. Akasaki, H. Amano, and S. Nakamura, were awarded with a Nobel Prize in 2014.3 Today, GaN and its alloys with indium (In) and aluminum (Al) are basic materials for fabricating optoelectronic devices as LEDs and Laser Diodes (LDs).4 GaN is also an attractive semiconductor for electronic high-frequency and high-power devices. This is due to its wide bandgap as well as high values of electron velocity, breakdown field strength, and thermal conductivity.5 Most of the mentioned GaN-based devices are, however, built on foreign substrates [sapphire (Al2O3), silicon (Si), or silicon carbide (SiC)]. For many years, native substrates were not available in large quantities on the market. The challenging GaN crystal growth process can be held responsible for this. The compound melts at extremely high temperature (>2200 °C) and the nitrogen pressure necessary for congruent melting of GaN is expected to be higher than 6 GPa.6,7 Thus, today it is impossible to crystallize GaN from the melt. This compound should be grown by other techniques that require lower pressure and temperature. Crystallization from the already mentioned gas phase, solution, or any combination thereof must be included. For more than the last three decades, growing bulk GaN has remained challenging for the nitride community. Progress in bulk GaN growth was mainly motivated by the LD market. GaN-based LD structures have to be built on native substrates.8 The development of the LD market was not rapid enough to force a breakthrough in technologies for fabricating GaN substrates. The demand for GaN wafers was not significant. Recently, however, the situation has slightly changed. GaN substrates of relatively high quality became available.9–16 The nitride-based LD market also starts to increase. The main applications are copper welding, general and automotive lighting, free space optics (FSO), visible light communication (VLC), as well as light fidelity (Li-Fi). Considerable progress in GaN-based LDs fabricated by Nichia and Osram is observed.17,18 The next driving force for crystallizing bulk GaN and fabricating substrates seems to be the electronic industry and the demand for vertical high-power transistors and diodes. For these applications, GaN wafers of high structural quality and with a high free carrier concentration are necessary. The availability of semi-insulating GaN (SI-GaN) substrates for making lateral devices also remains an open issue. High electron mobility transistors (HEMTs) are mainly prepared on SiC substrates.19 Recently, a few kinds of such devices based on GaN-on-Si have been commercialized.20,21 For the industry, the very low cost as well as large size (8-in.) of a Si substrate proved to be more important than the relatively high price and small size (2-in.) of a native GaN wafer. The high structural quality of the latter one seems to be of low significance. Obviously, the yield and cost of epitaxy with a complicated buffer layer and hundreds of GaN/AlGaN superlattices applied for releasing stress in the GaN HEMT structures based on Si remain a secret. It appears possible that the cost of preparing a simpler AlGaN/GaN-on-GaN structure could be lower. Nobody has compared the costs, yield (in terms of the number of devices from one wafer), or the size of nitride HEMT transistors fabricated on an industrial scale on Si, SiC, and SI-GaN. The reason for that is very simple: lack of mass production of 2-in. SI-GaN wafers. Additionally, everybody is waiting for the availability of 4- or 6-in. GaN substrates.
GaN-on-GaN electronic devices are still at the beginning of their road to commercialization. Their main competitors are devices based on SiC (already commercialized), gallium oxide (Ga2O3), and probably (in the near future) AlGaN-on-AlN and diamond (C), as well as boron nitride (BN). All these materials are very important for new electronic applications. It has to be, however, underlined that the mentioned materials, including GaN, will not replace but only supplement Si. In the case of SiC and Ga2O3, native substrates (highly conductive and SI) up to 6-in. in diameter are available. This facilitates to build high-quality device structures by epitaxy and/or implantation. SiC has been developed for almost 40 years; thus, its commercialization should not be a surprise. Research on Ga2O3 began less than 5 years ago, but first devices have already been demonstrated.22 This material develops extremely fast and cannot be left out from the group of significant competitors of GaN.
Two different technologies can be applied for preparing GaN-on-GaN vertical devices (e.g., MOSFET). The structure can be grown by epitaxial techniques with some procedures as regrowth, needed due to the device architecture, or by ion implantation.23,24 The latter approach seems to be much less demanding and more perspective. Regrowth is combined with an appropriate etching of part of a device structure and subsequent growth of nitride in given semi-polar and/or non-polar directions. On the other hand, it is well known that ion implantation is one of the basic tools for semiconductor device fabrication. The implantation process has been commonly applied for selective area n- and p-type doping. This allows us to reduce the device size and to control the electric field configuration in devices. In the case of GaN, high n-type carrier concentration and conductivity have already been demonstrated.25 Recently, very effective activation by ultra-high pressure annealing (UHPA) of magnesium (Mg)-implanted p-type GaN has been announced.26 Diffusion of Mg during the UHPA process has also been intensively studied.27 Without going further into details of the device construction, it should be mentioned again that the first step for building them is to have GaN substrates of the highest structural quality.
Today, GaN is crystallized by three methods: (i) HVPE28 and its derivatives as Oxide VPE (OVPE)29 or halide free VPE (HFVPE);30 (ii) sodium flux (Na-flux) crystallization;31 and (iii) basic or acidic ammonothermal growth.32,33 Perspectives of the above-mentioned technologies will be presented in this paper.
It will be shown that growth from the vapor phase requires a new philosophy. It seems that crystallization on foreign seeds (sapphire and GaAs) will not develop any further. Despite many tricks performed during the crystal growth process, the structural quality and yield of obtained GaN crystals and wafers are very low and cannot be improved. The only right approach is GaN-on-GaN crystallization. Since there is a lack of high-quality GaN seeds, the Na-flux and ammonothermal methods could be very helpful.
The Na-flux growth of GaN, performed from a solution of gallium and sodium at relatively low nitrogen pressure, is focused on fabricating high-quality crystals of large lateral size (6-in. and 8-in.), which can be used as seeds in other crystallization methods (mainly HVPE or OVPE). Advantages and disadvantages of the new approach in the Na-flux technology, promotion of lateral growth of GaN crystals on point seeds, will be discussed. Difficulties and barriers of this method will be pointed out and analyzed.
Recently, significant progress has also been observed in the ammonothermal method, both basic and acidic approaches. High structural quality crystals and of 4 in. in diameter were obtained.34 Next challenges and future perspectives of this technology will be described in detail. Basic ammonothermal technique allows us to crystallize highly conductive (n-type) and SI crystals. Wafering procedures for fabricating GaN substrates will be discussed. Obviously, the use of ammonothermally grown GaN as seed for HVPE will also be presented. All challenges and difficulties in the growth of bulk GaN from the vapor phase on a native seed will be demonstrated. Some new ideas, as overcoming the equilibrium crystal shape, will be shown and analyzed. Scenarios for further development of bulk GaN crystallization and wafering will be presented. Special attention will be paid to the two techniques developed in Poland: basic ammonothermal growth and HVPE, as well as their combination. Parameters which make GaN substrates suitable for epitaxy of electronic and optoelectronic device structures will be discussed. These specifications motivate the development of bulk GaN growth, and, therefore, they will be presented before showing the state-of-the-art results of different crystallization methods.
REQUIREMENTS FOR GaN SUBSTRATES
The most important feature of a GaN substrate is its structural quality. The threading dislocation density (TDD) should be as low as possible and uniform across a wafer. Today, the lowest values of TDD, of the order of 104 cm−2, are noted for ammonothermally grown GaN on its (0001) surface.33–35 In commercially available HVPE-GaN, this value is two orders of magnitude higher. The value of TDD will be presented for each discussed crystallization method. Since TDD is correlated with etch pit density (EPD),36 the latter data will mainly be shown. In our opinion, low TDD and/or EPD (we will use these parameters interchangeably) is not the most important feature of GaN substrates. Obviously, we remember that screw and mixed dislocations are the source of leakage current in a device. It should be remarked that there are not a lot of results presenting the density of dislocations distinguished between screw, mixed, and edge ones in GaN substrates or crystals. The data for ammonothermal GaN and HVPE-GaN grown on it show that the density of screw dislocations are in the range 100–101 cm−2 and the density of mixed and edge ones is at the level of 5 × 104 cm−2.35 In our opinion, the flatness of crystallographic planes (see Fig. 1) seems to be a more important parameter. It guarantees a uniform off-cut of the substrate. This, in turn, allows for epitaxial growth with atomic step flow. Generally, the variation of off-cut across the surface cannot be higher than 0.1°. This is the main and basic requirement for promoting the already mentioned bilayer step flow, controlling the composition of ternary alloys in the device layers as well as the incorporation of dopants and unwanted impurities. It was shown that both the quality of the epilayer and the incorporation of dopants depend on the off-cut degree of the substrate.37,38 It was also found that a higher misorientation corresponds to a higher concentration of the hole carriers at the same density of incorporated magnesium acceptor.39 An optimal and uniform substrate misorientation is important for AlGaN grown on GaN substrates. The evidence of strain relaxation of the AlGaN in samples grown on substrates with a larger off-cut was demonstrated.40
For a 2-in. GaN wafer, the required radius of curvature should exceed 15 m. In the commercial HVPE-GaN substrates, it is at the level of 5–10 m. For Na flux and ammonothermal GaN, the value of the radius of curvature31,33–35 fulfills the presented above requirement. But again, as in the case of HEMTs based on Si, SiC, and GaN, nobody has compared the yield in terms of the number of devices fabricated on 2-in. GaN substrates with bent and flat crystallographic planes. Nobody has even demonstrated the difference in the uniformity of epitaxial layers. Again, such results are missing due to the lack of 2-in. GaN wafers of high structural quality (flat crystallographic planes) on the market.
The uniform off-cut allows us to prepare an epi-ready surface of a substrate. An epi-ready surface is the next requirement for a wafer. The surface, mainly (0001) in the case of GaN, should be epi-ready without any subsurface damage. The value of root means square (RMS) of epi-ready GaN should be lower than 0.1 nm with atomic steps visible and the surface should be clean. Figure 2 shows an epi-ready (0001) surface (c-plane) with a RMS data of a typical ammonothermal 2-in. GaN wafer fabricated at the Institute of High Pressure Physics of the Polish Academy of Sciences (IHPP PAS).
Last but not least, there is a requirement associated with the free carrier concentration in a GaN wafer. For vertical devices, as already mentioned in the Introduction, substrates of high and uniform carrier concentration are needed. P-type GaN wafers have not been demonstrated, so these will not be discussed. In the case of n-type GaN, a uniform electron concentration of the minimum order of 1018 cm−3 is required. This allows us to fabricate in a relatively easy way a stable and low-resistance ohmic contact to the bottom part of the substrate. Obviously, lateral devices require SI substrates with resistivity higher than 108 Ω cm at room temperature. This value cannot decrease rapidly with increasing the temperature. It would be nice to have substrates highly resistive at 300 °C (the operating temperature of high power–high frequency electronic devices). After an epitaxial growth (thus, annealing at high temperature, of the order of 1000 °C), the resistivity should not change. In order to obtain SI-GaN or n-type GaN, uniform doping with acceptors or donors, respectively, is needed.
COMMERCIALLY AVAILABLE HVPE-GaN WAFERS—CRYSTALLIZATION ON FOREIGN SEEDS
HVPE is a method of crystallization from gas phase. In the low-temperature zone of the quartz HVPE reactor (800–900 °C), hydrochloride (HCl) reacts with gallium to form gallium chloride (GaCl). GaCl is then transported by the carrier gas (mainly N2, H2, Ar, He, or their mixtures) to the high-temperature zone (1000–1100 °C). Herein, GaN is crystallized due to a reaction of GaCl with ammonia. NH3 is flown separately to the crystal growth zone. The details of this technology and the reactor configurations were presented in many papers.28,35,41 GaN is mainly crystallized on a foreign seed. It is usually a GaAs substrate with a low-temperature buffer layer of GaN or a metal-organic vapor phase epitaxy (MOVPE)-GaN/sapphire template.42,43 The main crystallographic growth direction in HVPE-GaN technology is the [0001] (c-direction). Strengths of HVPE include a relatively high growth rate, which can exceed 100 μm/h, and a possibility to crystallize high-purity material. Concentrations of main dopants (silicon and/or oxygen) are lower than 1017 cm−3. Fujikura et al. reported HVPE-GaN with main impurities below the secondary-ion mass spectrometry (SIMS) detection limits.44 Doping processes in HVPE, with silicon or germanium for obtaining highly conductive crystals as well as with iron or carbon for SI ones, are well developed and described in detail in the literature.45–49 Thanks to these attributes, HVPE is a well-established method for fabricating GaN substrates. This growth technique is employed by Sumitomo Electric Industries (SEI),9 SCIOCS by Sumitomo Chemical,10 Mitsubishi Chemical Corp. (MCC),11 Furukawa,12 Nanowin,13 Lumilog by Saint Gobain,14 and Eta Research.15 Two main growth technologies are applied. SEI used GaAs as seeds and developed the Dislocation Elimination by the Epitaxial growth with inverse pyramidal Pits (DEEP) and Advance-DEEP (A-DEEP) technologies for fabricating GaN substrates. They consist of areas (e.g., arranged in stripes) of high (108 cm−2) and low (104 cm−2) dislocation density.42,50,51 Thus, the (0001) surfaces of the substrates are neither macroscopically uniform nor flat, which make them difficult or simply impossible to prepare to a proper epi-ready state. Fortunately, the bending of crystallographic planes is not observed thanks to the similar thermal expansion coefficients of GaN and GaAs as well as the existence of inversion domains that reduce stress in GaN.42
The second technology was introduced by Hitachi Cable (today SCIOCS). The Void Assisted Separation (VAS) method uses MOVPE-GaN/sapphire templates as seeds.43,52,53 It seems that all other companies, except SEI, use some derivatives of the VAS technology. VAS involves the formation of voids between the template and the new-grown HVPE-GaN. A thin titanium (or other metallic) layer is deposited on the surface of an MOVPE-GaN/sapphire template. The layer is annealed in an ammonia atmosphere and a TiN nanonet is formed. Due to the decomposition of GaN, voids are formed under the nanonet. Next, during the HVPE growth, the GaN layer nucleates through the openings in TiN and the voids remain at the GaN-sapphire interface. Stress induced by the difference in thermal expansion coefficients of a new-grown thick HVPE-GaN and sapphire results in a well-controlled self-separation of GaN during the cooling process. Free-standing (FS) crystals and then, after proper wafering procedures, grinding, lapping, mechanical and chemo-mechanical polishing, GaN substrates can be obtained. They have macroscopically flat c-plane surfaces, uniform TDD of the order of 106 cm−2, but bent crystallographic planes. The last one results from the difference between the lattice constants and thermal expansion coefficients of sapphire and GaN. The trick used to avoid the bending of crystallographic planes is to start the crystallization process in a three-dimensional (3D) growth mode and, by changing the supersaturation, switch it in time into a two-dimensional (2D) one.54 In spite of this, the value of bowing radius of crystallographic planes of a typical GaN substrate is not higher than 10 m. The substrates are plastically deformed and have dislocation bundles that create a cellular network.55 Table I summarizes the advantages and disadvantages of two technologies described above.
A-DEEP . | VAS . |
---|---|
No bending of crystallographic planes | Bending of crystallographic planes |
Non-uniform threading dislocation density (104–108 cm−2) | Uniform threading dislocation density (106 cm−2) |
Non-uniform and non-flat surface (c-plane) | Uniform and flat surface (c-plane) |
A-DEEP . | VAS . |
---|---|
No bending of crystallographic planes | Bending of crystallographic planes |
Non-uniform threading dislocation density (104–108 cm−2) | Uniform threading dislocation density (106 cm−2) |
Non-uniform and non-flat surface (c-plane) | Uniform and flat surface (c-plane) |
As mentioned in the Introduction, the HVPE technology requires a fresh approach. The crystallization on foreign seeds has already achieved the best possible results. Using MOVPE-GaN/sapphire template, one can crystallize 4-in., 6-in., and 8-in. GaN. However, bending of crystallographic planes increases with the diameter of the crystal. It is clearly shown in Fig. 3. Large bending makes it impossible to prepare a good substrate from such a crystal. A uniform off-cut and an epi-ready surface are not available. A proper misorientation can only be fully maintained for small substrates. For 2-in. wafers, this is no longer guaranteed. On the other hand, the TDD of the order of 106 cm−2 seems to be low enough for devices like LDs. Time will show if this value is also suitable for electronic devices.
GaN grown on GaAs can also give substrates of lateral size larger than 2 in. Additionally, it should be marked that the LDs applied in BluRay are built on the A-DEEP SEI GaN substrates.42 On the other hand, the future of BluRay does not look bright. This technology has been defeated by memory sticks (Si-based technology), fast Internet, and techniques like iClouds. It is not certain if the 400-μm-width stripes of low TDD are not too narrow for preparing high-power optoelectronic or electronic devices. Their size can reach several square millimeters. Thus, in the authors' opinion, the only way for the future development of GaN substrates is GaN-on-GaN crystallization. Since there is a lack of native HVPE seeds, the Na-flux and ammonothermal methods can be very useful.
SODIUM FLUX METHOD
The sodium flux method is similar to the high nitrogen pressure solution (HNPS) crystallization, the first technology that allowed IHPP PAS to obtain FS monocrystalline GaN.56 In both methods, GaN grows from a liquid solution. In Na-flux, under a relatively low pressure of nitrogen (5–50 atm.) and at constant temperature of the order of 900 °C, gallium and sodium are mixed in the crucible. Sodium increases the solubility of atomic nitrogen in the flux to a few percent.31 This is the basic difference from the HNPS method where the solubility of nitrogen in pure gallium did not even reach 1%. In the Na-flux technology, there is much more sodium than gallium (73%– 27%, respectively). Nitrogen molecules dissociate on the surface of the flux and dissolve into it. The mass transport is governed by convection caused by mechanical stirring of the flux.31 The crystallization process proceeds on a foreign seed (mainly MOCVD-GaN/sapphire template) placed at the bottom of a crucible. The Na-flux GaN crystals are of high structural quality and purity.57 In the undoped material, the free carrier concentration is not higher than 1016 cm−3. The main donor is oxygen. Germanium is intentionally incorporated in order to obtain highly conductive n-type crystals. There are no reports about SI material.
During the last 20 years, different approaches to the Na-flux method were developed. Today, the point seed technique is mainly used.58 An MOCVD-GaN/sapphire template is patterned into GaN point seeds (up to 1-mm in diameter). At the beginning of a crystallization process, all small seeds are overgrown in the c-direction. A pyramidal crystallization takes place (see Fig. 4; step 1). In order to switch from 3D to 2D growth, the seed is pulled out from the solution when GaN is pyramidal. Then, the Na–Ga solution remains only between the pyramids and the lateral growth is enhanced (the seed is still under nitrogen pressure). Since there is a small amount of the solution between the pyramids, the template should be again dipped into the solution. The pulling up and dipping procedures are repeated until the GaN surface becomes flat (see Fig. 4; step 6). Then, the crystallization run is continued in the 2D growth mode.
The presented above technique corresponds to the already mentioned switching of growth mode from 3D to 2D in the HVPE technology. However, this procedure in the Na-flux seems much more difficult. In case of HVPE the supersaturation can be easily changed by (for example) switching the carrier gas. It is well known that in H2 vertical growth is favored and in N2 the lateral one. In the Na-flux method the change in supersaturation is forced by pulling up the growing crystal from the solution. Outside the solution, but still under nitrogen pressure, the growth in the lateral directions is favored. The pulling and dipping procedures have to be performed a few times. The 3D part of the grown crystal should always be removed. Only the 2D GaN can be analyzed and applied (see Fig 4; step 6; yellow part). In our opinion, this technology seems very sophisticated and expensive. However, today it gives crystallographically flat 6-in. crystals.59 Eight-inch crystals are expected soon. Additionally, it was demonstrated that Na-flux GaN can be applied as a seed for the HVPE growth. As already mentioned, it seems that this is the future for crystals obtained by the Na-flux. First results presented by SCIOCS and Osaka University were very promising. HVPE-GaN of high structural quality, totally crystallographically flat (no bending) and with TDD/EPD lower than 105 cm−2, was obtained.60
AMMONOTHERMAL GaN
The ammonothermal technology is an analog of the hydrothermal one, used in industrial production of quartz.61 The difference is the presence of supercritical ammonia instead of water. The idea of ammonothermal growth is the following: GaN feedstock is dissolved in supercritical ammonia in the first of two zones of a high-pressure autoclave. The dissolved feedstock is transported to the second zone, where the solution is supersaturated and crystallization of GaN on native seeds occurs. An appropriate temperature gradient between the dissolution and crystallization zones enables the convective mass transport. Some mineralizers are added to ammonia in order to accelerate its dissociation and enhance the solubility of GaN. Thus, the growth can be proceeded in a different environment: basic or acidic. The type of environment is, obviously, determined by the choice of mineralizers. Ammonobasic growth makes use of alkali metals or their amides as mineralizers, while in ammonoacidic growth, halide compounds are present (for details, see Ref. 62). A negative temperature coefficient of solubility is observed in the ammonobasic approach.63,64 As a consequence, the chemical transport of GaN is directed from the low-temperature solubility zone (with the feedstock) to the high-temperature crystallization zone (with the seeds). The pressure of ammonia in the reactor is usually between 100 and 600 MPa. The typical growth temperature is in the range 400–750 °C. Schemes of the time-temperature relation of the feedstock and seed zones for the two described above approaches, basic and acidic, are presented in Fig. 5. Green and red curves represent the temperatures of the feedstock and growth zones, respectively. At the beginning of a basic ammonothermal process, the feedstock zone is heated up and the seeds placed in the crystal growth zone (with a lower temperature) start to dissolve in ammonia. At this stage (dissolution stage), a back-etching process occurs; the seeds are coupling with the solution. Then, the crystal growth zone is heated up and the temperature of the feedstock zone decreases. The crystal growth run starts. For acidic ammonothermal run, the temperature vs time profiles look opposite. The back-etching process takes part at a higher temperature and the crystallization proceeds at a lower one.
At present, there are several companies and research institutes working on ammonothermal growth of GaN, such as IHPP PAS (formerly Ammono S.A., Poland),32 SixPoint Materials Inc. (USA),65 University of California Santa Barbara (USA),66 University of Stuttgart and University Erlangen-Nürnberg (Germany),67 MCC (Japan),33 Tohoku University (Japan),68 and Kyocera (formerly Soraa, Inc., USA/Japan).69 The leader in the acidic ammonothermal growth is MCC with their own patented SuperCritical Acidic Ammonia Technology (SCAAT). Recently, 4-in. (in lateral size) GaN crystals of the highest structural quality (EPD lower than 104 cm−2 and bowing radius of crystallographic planes higher than 60 m) have been presented.34 The crystals are n-type with a free carrier concentration of the order of 1018 cm−3. They are grown at a temperature ranging from 500 °C to 650 °C and under ammonia pressure of 0.15–0.3 GPa. In turn, the crystallization method used at IHPP PAS is basic ammonothermal growth, which proceeds in the temperature range 400–600 °C and pressure 0.3–0.4 GPa. The main idea of this growth is to increase the lateral size of the initial seed (e.g., high structural quality ammonothermal GaN), which is in the shape of a long stick (called a slender crystal; see the scheme presented in Fig. 6). At the same time, the seed is naturally overgrown in the vertical –c-direction, but the vertical growth rate is much lower (dozens of times) than the lateral one (for details, see Ref. 32). It is clearly seen in Fig. 6 that the growth in the lateral directions is performed in order to get rid of all (11-2) facets and form (10-10) ones. When the appropriate lateral size is reached (after a few ammonothermal processes), the crystallization takes place only in the –c-direction; the growth in the +c-direction is always mechanically blocked. After that, the crystal is sliced perpendicularly to the growth direction and it can increase the population of seeds used for subsequent growth runs or wafering process (GaN substrate fabrication).
Commercial HVPE-GaN, after an appropriate preparation, can also be used as a seed in an ammonothermal process. Figure 7(a) shows a typical ultra-slender seed, with its basic structural characteristics, used at IHPP PAS. The main trick, not marked in Fig. 6, is to remove this ultra-slender seed from the first new-grown ammonothermal crystal and apply ammonothermal GaN of high structural quality for further growth processes. As shown in Figs. 7(b) and 7(c), the ultra-slender crystal is not of high structural quality, with the FWHM value for the (00.2) reflection of 333 arc sec, bowing radius of crystallographic planes equal to 3 m and EPD at the level of 107 cm–2. Thus, this slender seed must be removed from the grown crystal in order to achieve the highest structural quality. Elementary calculations result in the desired length of the primary slender seed assuming a hexagonal equilibrium habit of a growing crystal. In order to grow a 2-in. wafer, the ultra-slender seed must be longer than 100 mm. It means that HVPE-GaN of a diameter equal to or even higher than 4 in. is needed. Since such crystals were not commercially available when the basic ammonothermal method was developed (the beginning of the 21st century), the growers chose a different way: tiling. Figure 8 presents a scheme of combining two crystals to get one larger. Obviously, one can join many crystals, which was shown in detail by Fujikura et al. for HVPE-GaN.70 It should be, however, kept in mind that the tiling technology always leads to lowering the structural quality of the obtained substrate. The area of tiling will always be visible in polarized light [see Fig. 8(b)]. The TDD in the tiling place will also be slightly higher.
Today, 4-in. or even bigger HVPE-GaN wafers can be fabricated. Therefore, going back to growing uniform 2-in. crystals without the tiling approach seems reasonable and indicates a new idea and solution for ammonothermal technology. Preparing a proper slender seed and overgrowing it in lateral directions to obtain a uniform GaN crystal from which one can prepare a 2-in. circle seems possible. Such an approach, with the low growth rates appearing in the basic ammonothermal method, could result in the first high structural quality 2-in. GaN substrates after 3–5 years. Fabricating 4-in. GaN wafers looks more complicated. A possible way is to bond two long slender seeds into one and then overgrow it in the lateral directions. Time (a few years) and large high-pressure autoclaves are required. For preparing the highest quality 4-in. basic ammonothermal GaN wafers, the market prospects for GaN-on-GaN devices would have to be unimaginably great. Then, money will be further invested in crystal growth. Since the market for GaN-on-GaN devices is not so large, in the authors' opinion, all efforts should be focused today on improving the structural quality of ammonothermal GaN crystals and obtaining 2-in. GaN substrates on a mass scale.
Some defects were observed in polarized light in many basic ammonothermal crystals. This is clearly seen in Fig. 9. We called the appearance of such features the stress induced polarization effect (SIPE). It is interesting that SIPE does not exist in the material grown in the lateral direction (see Fig. 9). Additionally, the EPD in this material (laterally overgrown) is two orders of magnitude lower than that in the seed crystal. Most probably, SIPE is only created at the beginning of a crystallization process in the –c-direction due to non-uniform supersaturation at the relatively large surface of a crystal. There are a lot of speculations and models that should be experimentally verified. It should be stressed that SIPE is not always present in the grown crystals. This shows that there is still room for improvement in the structural quality of basic ammonothermal GaN.
The main dopants unintentionally incorporated into the ammonothermal crystals are oxygen and hydrogen (∼1019 cm−3). In basic ammonothermal GaN, high concentrations of carbon (>1017 cm−3) and metals as Al, Mg, Mn, Fe, or Zn are observed. These impurities come from ceramic and metal elements (i.e., holders) of the autoclave in the crystal growth zone. The free carrier concentration of ammonothermal GaN is usually at the level of 1019 cm−3. It can be decreased to 1 × 1018 cm−3 or even lower when some getters are added to ammonia. Then, also the oxygen concentration decreases. In order to grow a SI material, the donors are compensated by Mn acceptors. Thus, three types of substrates are prepared: n-type with free carrier concentration of 1 × 1019 cm−3 or 1 × 1018 cm−3 and SI with resistivity at room temperature higher than 1 × 108 Ω cm.
It should be noted that in a basic ammonothermal growth process without any getters or dopants (e.g., intentionally introduced acceptors), the growth rate in all crystallographic directions is the highest. Also, from the technological point of view, such a process is the simplest one. The crystals contain, however, a significant concentration of gallium vacancies (∼1019 cm−3) or their complexes with oxygen and/or hydrogen.71–73 One of the newest ideas (not investigated yet) is to introduce gallium atoms to the crystals by ion implantation. Then, due to the already mentioned UHPA process, Ga atoms might diffuse into the entire volume of the crystal. The gallium vacancies would be replaced by Ga atoms and the crystal would have a higher but well determined free carrier concentration (gallium vacancies act as acceptors in GaN). For decreasing the free carrier concentration, Mg or Mn ions would be implanted. Depending on their concentration, after the UHPA, GaN crystals might be conductive (n-type) or SI. The diffusion coefficients of the mentioned dopants should be higher than 10−10 cm2 s−1 in GaN. Such high values ensure a relatively fast diffusion. The duration of the UHPA processes would not be longer than 400 h. It is known that these types of processes are performed at temperatures of the order of 1450 °C and under nitrogen pressure of 1 GPa. The UHPA technology comes from HNPS one.56
P-type bulk crystals remain a separate problem. GaN with a low donor concentration is needed for Mg ion implantation and the UHPA process. Therefore, high structural quality and purity HVPE-GaN grown on a native seed must be considered.
HVPE-GaN-ON-NATIVE SEEDS
One of the main disadvantages of the ammonothermal crystal growth process is the low growth rate. The average value in the –c-direction varies from 1 to 2 μm/h. The growth rate in the sodium flux technology is slightly higher but still much lower than in HVPE. Thus, ammonothermal and sodium flux methods can be the source of seeds for HVPE growth. “They have the seed and we have the speed”— said Keith Evans, the former CEO of Kyma, negotiating using ammonothermal GaN crystals as seeds for HVPE.74 In fact, the first published results of HVPE growth on ammonothermal GaN were presented by the IHPP PAS team in 2013.75 Then, similar works (HVPE-GaN-on-SCAAT) were shown by MCC.34,76 Many results of such an approach were published. They are summarized in Refs. 35 and 77. As mentioned, doping processes, with silicon or germanium for obtaining highly conductive crystals or with iron, carbon, or manganese for SI ones, are well developed in the HVPE technology. The main problem for bulk growth has been determined. Finding a solution remains only a matter of time.
It was concluded that during homoepitaxial crystallization of HVPE-GaN in the c-direction the non-polar and semi-polar growth of “wings” (laterally grown material) leads to the formation of large stress in the growing crystal, close to its edges. This happens due to a different incorporation of dopants (mainly oxygen) into HVPE-GaN grown in the main c- and lateral directions. A different incorporation of dopants leads, in turn, to different lattice parameters of the crystallized GaN.35,78,79 The stress from the edges is much more significant than that generated by the lattice mismatch between the seed and the deposited layer [see Fig. 10(a)].78 Raman spectroscopy was applied to investigate the strain in bulk GaN crystallized by HVPE on ammonothermally synthesized seeds.79 Areas close to the edges of the growing crystals were analyzed. Shifts of phonon lines were examined thoroughly in order to determine the strain tensor. Three areas were checked: the seed, HVPE-GaN grown in the c-direction, and the laterally grown material. Differences in the spectra measured in these parts of GaN are shown in Fig. 10(b). The determined values of the stress tensor were confirmed by the stress values obtained by computer simulations based on experimental XRD data.79 Avoiding the lateral growth during crystallization in the c-direction seems crucial for developing the GaN bulk growth technology.
Growth in the lateral directions on the edges of the seed can be eliminated by placing a metal ring (e.g., molybdenum) on a seed [see Fig. 11(a)]. Then, the diameter of the new-grown crystal is smaller than that of the seed. Additionally, some facets exist on the HVPE-GaN crystal. Growth on them is obviously minimized. The metal ring increases the dissociation of ammonia. The supersaturation close to the edges of the crystal decreases. No growth takes place in the lateral direction. The same phenomenon can be observed when a moly ring is placed around the native seed [see Fig. 11(b)]. The size of the seed is not reduced. The facets are formed. The main issue is to find such growth conditions that will allow forming only the side facets that grow faster than the material in the c-direction. Then, the side facets will disappear. The crystal will increase its lateral size when growing only in the c-direction. According to a hypothesis by Sitar,80 this condition may be achieved by controlling the thermal field around the crystal. It has to reach its final shape by adapting to the thermal field rather than taking the equilibrium hexagonal habit. The equilibrium shape can be overpowered by a proper thermal field design. In this case the crystal will follow the thermal field and grow in a direction perpendicular to the isotherms. It is shown schematically in Fig. 11(c). This was demonstrated for aluminum nitride (AlN) growth by physical transport deposition (PVT) and presented by HexaTech.81 Obviously, there is a big difference in the formation of supersaturation in PVT and HVPE methods. The supersaturation is the difference of thermodynamic potentials at the interface between a crystal and its environment. In the case of PVT, it is almost unambiguous with the temperature distribution on the growing surface. In the case of HVPE, reactions of all vapor species should be considered. It should, however, be stated that if the equilibrium crystal shape of GaN can be overcome, it will be a transformative achievement for the HVPE technology. It has never been demonstrated before and will allow the growth of true bulk GaN crystals of high purity, eventually yielding several tens of wafers per boule.
To this day, the formation of side facets has not been defeated and they are still created during HVPE-GaN growth in the c-direction. In spite of this, MCC has already demonstrated 4-mm-thick HVPE-GaN-on-SCAAT.34 IHPP PAS showed 3-mm-thick HVPE-GaN-on-Ammono [see Fig. 11(d)].82
SUMMARY
Growth of high structural quality bulk GaN crystals can only be performed with GaN-on-GaN technology. However, since GaN crystals do not occur in the natural environment, the first growth of the material has to be carried out on a foreign wafer (seed). Then, GaN-on-GaN crystallization should be introduced. This paper clearly shows that sequential growth has to be carried out by various methods. HVPE-GaN growth on foreign seeds was presented. Then, the use of FS HVPE-GaN crystals as slender seeds for ammonothermal growth was described. After that, high structural quality ammonothermal GaN crystals were used as seeds for the HVPE process. A similar situation happened with the Na flux method and HVPE. High structural quality Na-flux GaN crystals, grown before on foreign wafers, were applied as seeds in the HVPE technology. All the presented combinations lead to structurally perfect GaN crystals. It seems, however, that today the main method for mass fabrication of GaN crystals is HVPE. This is due to its high growth rate and purity. Process repeatability is also the highest. Ammonothermal as well as Na-flux methods can provide seeds for HVPE growth. However, if the ammonothermal technology is improved, it will be a great competitor for HVPE. It should be remembered that during one ammonothermal process, hundreds of crystals can be grown. Table II summarizes advantages and disadvantages of all three crystal growth methods presented in this paper. Most probably, the economy will decide which technology will be applied in the future. In this paper, the costs of crystal growth processes and substrates fabrication have not been analyzed and compared.
. | HVPE . | Na-flux . | Ammonothermal . |
---|---|---|---|
Advantages |
|
|
|
Disadvantages |
|
|
|
. | HVPE . | Na-flux . | Ammonothermal . |
---|---|---|---|
Advantages |
|
|
|
Disadvantages |
|
|
|
In our opinion, in the near future, the following new ideas for growth of bulk GaN crystals should be realized:
fabricating 2-in high-quality ammonothermal GaN wafers without the tailing technology;
applying implantation and UHPA technologies to modify and control the free carrier and point defect concentrations in ammonothermal and HVPE-GaN substrates; and
developing a process for overcoming the equilibrium crystal shape in growth from the vapor phase.
ACKNOWLEDGMENTS
This research was supported by the TEAM TECH program of the Foundation for Polish Science co-financed by the European Union under the European Regional Development Fund (No. POIR.04.04.00-00-5CEB/17-00) as well as by the Polish National Science Center through Project No. 2018/29/B/ST5/00338. The authors are also grateful to Dr. Malgorzata Iwinska for helpful scientific discussions.
DATA AVAILABILITY
The data that support the findings of this study are available from the corresponding author upon reasonable request.