Because the decomposition pressure at the melting point (∼ 2200  ° C) is extremely high (∼ 6  GPa), ^[18] GaN decomposes under the pressure instead of growing.^{[15, 16]} And the standard techniques of crystal growth (Bridgman, Czochralski) cannot be employed for GaN growth.

Using the direct reaction between gallium and nitrogen, scientists developed the high-pressure nitrogen solution (HPNS) method and Na-flux method for GaN crystal growth.^{[15, 16]} To realize the GaN growth, it is necessary to enhance the N solubility in Ga melt for both of the two methods. The N solubility is relatively high in an HPNS system benefiting from high temperature and high pressure growth conditions. While in the Na flux system, high solubility of N is achieved by adding sodium into gallium melt.

The growth mechanism of HPNS is as follows: Nitrogen molecules dissociate on gallium surface and dissolve in the metal, and then nitrogen atoms transport from the hot region of the solution to the cooler region, and finally GaN crystallizes and grows. Due to the very low solubility of nitrogen in Ga less than 0.5%, the growth rate is very slow, and the crystal sizes have been limited to several millimeters. It is difficult to grow crystals for industrialization due to the slow growth rate and the critical growth conditions. Despite these disadvantages, the HNPS method grows very high-quality “ truly” bulk GaN with dislocation density less than 2 × 10²  cm^{− 2} by spontaneously nucleation.^[16]

In the Na-flux method, by adding sodium into Ga melt, the growth temperature and pressure can be greatly decreased to 750  ° C– 900  ° C and 3  MPa– 5  MPa, which are only 50% and 0.3% of the growth conditions used in HPNS, respectively. The basic growth process is as follows: First of all, the nitrogen molecules are ionized by Na at the gas– liquid interface, and the ionized nitrogen is easily dissolved in the Ga– Na melt system. Then the nitrogen atoms combine with gallium atoms or additives (carbon, lithium, etc.). When the N concentration exceeds the critical growth concentration in the system, spontaneous nucleation of GaN crystal takes place. On the other hand, forming complex compounds, the nitrogen atoms could be transported to the surface of the GaN seed; generally the GaN seed was grown by MOCVD or HVPE method. As time goes on, most N atoms will contribute to the GaN seed in the bottom of a crucible. When the N concentration exceeds the critical growth concentration, Liquid Phase Epitaxy (LPE) will happen.^[19] Normally, high super-saturation generated by increased nitrogen concentration would grow transparent GaN single crystal at a fast growth rate. In summary, in order to obtain high quality GaN crystal by the Na-flux method, one way is to grow GaN crystal by spontaneous nucleation, and the other way is to grow GaN crystal by LPE.

Inspired by mass production of quartz grown by the hydrothermal method, the ammonothermal method was developed for GaN crystal growth. The ammonothermal process is analogous to the hydrothermal method, which allows solubilization of polycrystalline GaN nutrient or feedstock in supercritical ammonia under high pressure. The process of ammonothermal growth is as follows: The autoclave is divided into growth zone (lower solubility of GaN) and feeding zone (higher solubility of GaN) by a baffle through the temperature gradient control. By convection, dissolved GaN nutrient with higher solubility is transported to the growth zone from the feeding zone, and GaN crystallizes and grows on the seeds due to the supersaturation in the growth zone. And the remaining dissolved nutrient is transmitted to the feeding zone in the role of convection, which is unsaturated relative to the solubility of feeding zone. Therefore, the GaN nutrient continues to dissolve. Thus, a gallium nitride growth cycle is realized.^[20]

However, the solubility of GaN in supercritical ammonia remains in insufficiency for ammonothermal growth. And mineralizers are used to increase the solubility of nutrients. The mineralizers commonly used at present can be divided into two typical types: basic mineralizers, such as MNH₂ (M = Na, Li, K), and acidic mineralizers, such as NH₄X (X = Cl, Br, I). As a consequence of different chemical nature, basic mineralizers and acidic mineralizers of the ammonothermal technique are different: the coefficient of the solubility is positive in the ammonoacidic solutions, while negative in the ammonobasic solutions.^[21]

Generally, an HVPE reactor for GaN mainly includes two reacting zones. One is the source zone for providing chloride gas of gallium and the other is the deposition zone, in which the GaN is formed from the reaction of gallium source and nitride source (NH₃). Therefore, in HVPE growth, the main chemical reactions can be divided into two parts. One is in the source zone where the gallium is kept at a certain temperature and reacts with HCl gas, which is commonly used as reactive gas and introduced into the source zone over metal to form chloride gas of gallium. Then the gaseous species formed at the source zone are transported by carrier gas to the deposition zone for further react with nitride source to form GaN. The carrier gases are generally H₂ and inert gas (N₂, He, Ar, etc.).

In the source zone, the gaseous species and their equilibrium partial pressures are very important for the formation of GaN in the deposition zone. Generally, the chemical reactions simultaneously happen in the source zone:

The equilibrium partial pressures in the source zone can be calculated by thermodynamic analysis.^[22] In HVPE growth, the temperature of the source zone is usually around 850  ° C, at which the major gaseous species of gallium is formed by Eq.  (1). It should be noticed that the input partial pressure of HCl should not be too high in order to make almost all the HCl introduced into the source zone react with gallium. Therefore, the equilibrium partial pressure of GaCl should be almost equal to that of HCl at the source zone. From the thermodynamic analysis, if the temperature of the source zone is too low or the partial pressure of input HCl is too high, or the reaction area of gallium with HCl is too small, the reactions might be insufficient in the source zone, which means the reactions will be kinetically limited.

The GaCl formed at source zone and NH₃ are transported to the deposition zone separately by a carrier gas mixture of H₂ and inert gas. At the deposition zone, the source species are mixed and the following chemical reactions occur simultaneously

Then, gaseous species at the growth zone include GaCl, GaCl₂, GaCl₃, (GaCl₃)₂, NH₃, HCl, H₂, and inert gas. The partial pressures of these gaseous species at the growth zone as a function of temperature at deposition zone have been calculated by Koukitu et al.^[23] The partial pressures of (GaCl₃)₂ and GaCl₂ are very small under typical growth conditions. The equilibrium constants for these reactions can be calculated from the equilibrium equations.^[24] GaCl₃ also reacts with NH₃ to form GaN, whose equilibrium constant is close to zero at the usual growth temperature. Though the equilibrium constants are close for both reactions using GaCl and GaCl₃, the partial pressure of GaCl₃ is far lower than that of GaCl. Therefore, equation  (5) is the dominant reaction.

The driving force for the deposition can be obtained from the difference between the number of Ga atoms put in and the amount of Ga atoms remaining in the vapor phase, which can be written as^[23]

where , P_GaCl, and P_GaCl3 are the input partial pressure of GaCl, the partial pressure of GaCl and GaCl₃ in the vapor phase in deposition zone, respectively. From the calculation, ^[23] the growth temperature has an important influence on the driving force. Note that the influence of H₂ on the decrease of the driving force is more significant at high growth temperatures, and the driving force decreases with the increase of H₂ in the carrier gas.^[25]

In contrast to classical semiconductors like Si and gallium arsenide (GaAs), for which device structure is based on high quality native substrate with low cost, the GaN-based device technology is far more advanced than that of native GaN substrate crystal growth. Many scientists and engineers are working hard to change this situation and have made great progress with different methods of GaN crystals in recent years. The following table lists the contrast of growth conditions and progress for different bulk GaN growth methods.

Table 1.

Table 1.

Table 1. The contrast of growth conditions and progress for different bulk GaN growth methods. The unit: 1  atm = 1.01325 × 105  Pa, 1  inch = 2.54  cm. Growth method Growth pressure Growth temperature/° C Growth rate/(μ m/h) Crystal thickness/lateral size/dislocation density Dislocation density/cm− 2 HPNS 1  GPa– 2  GPa ∼ 1700 1– 3 platetlets or prisms, mm grade 102, Ref.  [16] Na flux 3  GPa– 5  MPa 800 10– 40 mm grade/2– 4 inch 102– 104, Ref.  [26] HVPE 1  atm 1020– 1050 100– 200 mm grade/2-6 inch 104– 106 Ammonothermal 100  GPa– 600 MPa 500– 750 1– 30 mm grade/2 inch 103, Ref.  [27]

Table 1. The contrast of growth conditions and progress for different bulk GaN growth methods. The unit: 1  atm = 1.01325 × 105  Pa, 1  inch = 2.54  cm.

To obtain high quality and crack-free bulk GaN substrate, dislocation reduction and strain control are the most important issues during HVPE growth. Epitaxial lateral overgrowth (ELOG) is an effective and traditional method to bend the dislocation line and form void-structure in GaN films simultaneously, which helps to enhance the dislocation  annihilation for dislocation reduction and helps to release the growth and thermal stress.^{[28, 29]}

A standard ELOG process is as follows: (i) Depositing mask film such as SiN, SiO₂, etc. on GaN film; (ii) Etching dielectric film to expose GaN as window area; (iii) GaN layer grows from the window and then laterally overgrows to cover the mask. By carefully adjusting the mask/window shape and the growth parameters for the overgrowth, reasonable lateral overgrowth rate and smooth surface can be achieved.^{[28, 30– 32]} Other technologies based on ELO have been proposed, like double-layer ELO^{[33, 34]} and three-step ELO, ^[35] in order to eliminate the remaining threading dislocations (TDs) in the coalescence region.

By improvement of ELOG technology, some special technologies have been developed in HVPE system for the growth of high quality and crack-free GaN layers, such as dislocation elimination by epitaxial growth with inverse-pyramidal pits (DEEP), TiN-based nano-mask and photoelectron-chemical etched nanowires.

DEEP was reported by Kensaku Motoki from Sumitomo Electric.^[36] During the growth of the GaN layer, there are numerous large hexagonal inverse-pyramidal pits constructed mainly by {11-22} facets appearing on the surface. While GaN grows, dislocations are collected to the center of the hexagonal pits parallel to (0001) in the ⟨ 11-20⟩ or ⟨ 1-100⟩ direction, and therefore dislocations are eliminated within the hexagonal pits except for its center. The DEEP method can produce a high quality GaN layer, but the dislocation distribution on the surface of GaN is not uniform. Far from the center of a pit, dislocations are few; in contrast, the center of a pit gathering dislocations possesses very high dislocation density (DD).^{[37, 38]}

Besides ELOG and DEEP, nano-mask is recognized as a good approach that not only reduces the DD but also easily separating the freestanding GaN from sapphire. Yuichi Oshima et al. developed a novel technique for preparing large-scale freestanding GaN wafers called VAS (void-assisted separation) by thin TiN film.^[39]

ELOG, DEEP, and nano-mask all use dielectric film as the mask for the overgrowth of GaN. Another unique method without dielectric film as mask was developed in our Institute.^[40] By electrode-less photoelectron-chemical (PEC) etching, long and straight GaN nanowire arrays are obtained, which have a density about 10⁷  cm^{− 2}, with diameters ranging from 150  nm to 500  nm, and corresponding lengths ranging from 10  μ m to 20  μ m. It is found that the GaN nanowires (NWs) are almost dislocation- and strain-free (Figs.  1(a) and 1(b)), because PEC etching begins from the dislocation core. Based on these GaN NWs, we can get high-quality and crack-free GaN layer (Figs.  1(c) and 1(d)), with thickness about 400  μ m and dislocation density about 10⁴  cm^{− 2}– 10⁶  cm^{− 2}.

Fig.  1.

	Figure Option ViewDownloadNew Window
	Fig.  1. (a) SEM image of GaN NW array, [40] (b) weak beam dark-field TEM image with g = 11-22, revealing that the GaN NW does not possess dislocations, [40] (c) photo of a 400-μ m GaN based on the PEC GaN NW arrays template, (d) panchromatic CL image shows that the DD of panel (a) is 3 × 106  cm− 2.

Fig.  2.

	Figure Option ViewDownloadNew Window
	Fig.  2. Thickness from 5  μ m to 5  mm and panchromatic CL images: (a) 5  μ m, 2 × 108  cm− 2; (b) 20  μ m, 5 × 107  cm− 2; (c) 200  μ m, 2 × 107  cm− 2; (d) 0.7  mm, 5 × 106  cm− 2; (e) 1.2  mm, 3 × 106  cm− 2; (f) 1.8  mm, 8 × 105  cm− 2; (g) 3  mm, 3.4 × 105  cm− 2; (h) 5  mm, 8.0 × 104  cm− 2; (i) the relation between thickness and DD.

In addition to the above mentioned methods, increasing the thickness of GaN is another effective approach to decrease the defect density of GaN substrate. The dislocation density reduced to 10⁶  cm^{− 2} as the thickness increased to 5  mm.^[41] According the same research from our group, it is found that the dislocation density declines sharply from 10⁸  cm^{− 2} to 10⁴  cm^{− 2} when the thickness increases from several micrometers to several millimeters (Figs.  2(a)– 2(h)). The related panchromatic CL images of different thickness GaN are shown in Fig.  2.

In 1997, Yamane et al.^[15] reported that GaN crystal can be grown in a Ga– Na mixed solution in relatively low pressure nitrogen atmosphere (< 50  atm) and at a relatively low temperature range of 600  ° C– 900  ° C called the Na-flux method. The Na-flux method has a significant advantages in synthesizing high quality GaN crystal through spontaneous nucleation process with very simple equipment. Imade et al.^{[78– 80]} reported that the GaN bulk crystal could be grown on the small spontaneously nucleated GaN seed by a long growth period, as shown in Figs.  5(a) and 5(b). In our group, we also obtain the GaN single crystals by spontaneous method, with the help of high-temperature and high-pressure home-made autoclave. Figures  5  [(a1)– (a3) and (b1)– (b3)] show the morphology of the spontaneously nucleated GaN crystals grown by Na flux. However, it is difficult to grow large GaN crystals with a moderate growth rate, because of the difficulty in controlling the spontaneous nucleation process.

Fig.  5.

	Figure Option ViewDownloadNew Window
	Fig.  5. (a) and (b) Photographs of GaN bulk crystal grown on the small spontaneously nucleated GaN seed by a long growth period. (from Refs.  [78] and [79]); (a1)– (a3) and (b1)– (b3): Morphology of the spontaneously nucleated GaN crystals grown by Na-flux method. (Temperature: 973  K∼ 1073  K, pressure: 2.0  MPa∼ 3.5  MPa).

In 2003, a group at Osaka University employed the liquid phase epitaxy (LPE) method using Na flux and made major progress in terms of the GaN single crystal size and quality, as well as scalability of the crystal growth system.^{[81– 83]} In 2008, a 3-mm thick 2-inch GaN crystal was obtained for the first time by adding carbon additive in the Na-flux method.^[84] In recent years, Osaka University has succeed in growth of 4-inch GaN bulk crystal^[80] and high quality GaN bulk crystal with dislocation densities less than 10³/cm^2[85] (Fig.  6). These important results were encouraged by the collaboration between Osaka University and Nichia/Ricoh Co., Ltd.

Aiming to improve the yield of the liquid phase epitaxy GaN, Kawamura et al.^[84] found that the addition of carbon could effectively suppress the generation of polycrystals, which could also improve the growth rate. Mori et al.^[19] investigated the dependence of growth rate on the concentration of carbon, and reported that an increase of carbon in the solution enhanced the growth of GaN at low concentrations and tended to suppress the GaN growth at higher concentrations. In our study, we found that the type of carbon used has an effect on the generation and morphology of GaN polycrystals, while the growth rate of the liquid phase epitaxy single crystals decreased in the following order: OMC, G4N, G3N, G2N, GRA, AC, and CNT, ^[86] as shown in Fig.  7. Though only 18 years have passed since the Na-flux method was discovered, significant results have been reported. The Na-flux method has a great potential to realize the GaN bulk single crystal and industrialized production.

Fig.  6.

	Figure Option ViewDownloadNew Window
	Fig.  6. (a) 2-inch and (b) 4-inch GaN crystals grown on HVPE-GaN seed crystals by Na-flux method (from Ref.  [80]), (c) morphology of 2-inch coalesced GaN crystal (from Ref.  [85]).

Fig.  7.

	Figure Option ViewDownloadNew Window
	Fig.  7. The LPE growth rate of GaN single crystal grown on HVPE substrate with different carbon sources (from Ref.  [86]).

A full understanding of dislocation nucleation, multiplication, and movement behavior is essential to decrease the dislocation density in GaN substrate and develop advanced technologies for improving GaN-based device performance.

Nanoindentation is an ideal method for studying the dislocation behavior in a crystal.^[108] There has been a considerable effort to determine the properties of dislocation in GaN during plastic deformation using indentation.^{[109, 110]} However, most of these earlier studies mainly focused on the microstructure of the plastic deformation in GaN thin film, so the fundamental dislocation multiplication and movement mechanism of GaN substrate are not understood fully.

A combination of nanoindentation and CL techniques has been performed to investigate the movement mechanism of dislocations in a GaN substrate under nanoindentation by Huang and Xu et al.^[111] The experimental results show that the plastic deformation of c-plane GaN is primarily due to slip occurring on both {001} and {101-1} planes. Dislocation loops can multiply and move from plane to plane by cross-slip after nucleation, enabling the plastic deformation to proceed further. This mechanism is further supported by the remarkable movement of the indentation-induced dislocations during annealing (Figs.  9(a)– 9(d)).

Fig.  9.

Figure Option ViewDownloadNew Window

Fig.  9. (a) and (b) Dark-field XTEM images of indentation in c-plane GaN; Room-temperature panchromatic CL images of conical indent in GaN (c) Dislocation luminescence in c-plane GaN under nanoindentation; (d) Dislocation pattern in m-plane GaN under nanoindentation; (e) The topographic image around a nanoindentation with a scan area of 10  μ m × 10  μ m. A V-pit of thread dislocation is marked with A, and a nearby plane position is marked with B. Panels  (f) and (g) are the CPD images of the same area acquired under dark conditions and under UV illumination with wavelength of 360  nm, respectively. The curves in the images are profiles along the white line.[114]

Further Huang and Xu et al. also demonstrated the nanoscale anisotropic dislocation behavior in GaN subs-trate.^[112] They presented nanoindentation experiments performed on the three principal surfaces, (0001), (11-20), and (10-10), of GaN substrate. CL and cross-sectional transmission electron microscopy (XTEM) measurements show that there are two primary dislocation slip planes ((0001) and (10-11) planes) for c-plane GaN, while there is only one most favorable dislocation slip plane ((0001) plane) for nonplanar GaN during plastic deformation. They suggest that the anisotropic elasto-plastic mechanical properties of GaN are relative to its anisotropic plastic deformation behavior.

More interesting, of course, are the intrinsic optical properties of dislocations in GaN substrate. A VL band peaking at about 3.12  eV from the region near the dislocations is characterized and identified by Huang and Xu et al.^[113] A comprehensive study encompassing CL, Raman, and annealing experiments allowed the assignment of VL band to e– A transitions involving V_Ga. They proposed the V_Ga is formed by the motion of jogged dislocations under shear stress.

The dislocations strongly affect the carrier properties in GaN, including minority diffusion lengths and surface recombination velocities, both of which are important for device performance. For example, in photovoltaic detectors, due to a large absorption coefficient of GaN, carriers are generated close to the surface and recombine. A sufficiently long minority diffusion length and the suppression of surface recombination velocity are helpful in the realization of high sensitivity.^[115] However, for most characterization methods, such as photoluminescence, ^[116] surface photovoltage, ^[117] and photocurrent^[118] measurements, the spatial resolution of as-measured carrier properties is low, which makes it difficult to reveal the relationship between the experimental results and the local dislocation structures. Electron beam-induced current (EBIC) method is capable of achieving the inhomogeneity of minority diffusion length along a depth gradient, but a p– n junction or a Schottky barrier has to be made at a cross section.^[119] Recently, a combination of surface photovoltage spectroscopy (SPS) method and Kelvin probe force microscopy (KPFM) was reported for simultaneous measurement of the topography, the local minority diffusion length and the surface recombination velocity at a single thread dislocation.^[114] The contact potential difference (CPD) at nanometer scale, varying with the incident photon energy, was measured with a corresponding topography image. SPS responses at single thread dislocations near a nanoindentation on an HVPE-grown GaN surface can be distinguished. As shown in Figs.  9(e)– 9(g), the thread dislocations introduced by a nano-indentation were observed as V-pits, where the photovoltage was lower than that on the plane surface under ultra-violate illumination. Compared with those on the plane surface, the calculated hole diffusion length is 90  nm shorter and the surface electron recombination velocity is 1.6 times higher at an individual thread dislocation.

GaN substrates can offer three different benefits for fabrication of InGaN-based laser diodes (LDs). First, smooth facets can be obtained by cleaving along the m-plane by using GaN substrates, which helps reduce the threshold current density and improve fabrication yield. Second, vertical LD structure can be fabricated by depositing n-electrode on the N-face of GaN substrates, which eliminates the current crowding effect that exists in lateral LD structures, and thus reduces operation voltage. Third, the lifetime of InGaN-based LDs can be improved by using GaN substrates with low dislocation density since dislocation defects can seriously reduce the lifetime of InGaN-based LDs operated at current density of more than 5  kA/cm²

By employing high quality GaN substrates, the performance of GaN-based laser diodes has made great advances in the past several years. Blue laser diodes with threshold current density lower than 2  kA/cm² and output power higher than 1  W have been commercialized by Nichia^[120] and Osram.^[121] However, epitaxial growth on bulk GaN substrates is not straightforward and the details about home-epitaxy have rarely been reported. At Suzhou Institute of Nano-tech and Nano-bionics, we have carried out extensive studies on the homo-epitaxy of GaN-based laser diodes on bulk GaN substrates. We have found that one key factor that affects the quality of epilayer grown on bulk GaN substrates is the miscut angle. A miscut angle larger than 0.2° is needed to suppress hillocks and wavy morphology.^[122] We have also obtained blue LDs with low threshold current, low voltage, and high output power. Power– current– voltage curves of a blue InGaN LD packaged with TO-56 under room-temperature (RT) and continuous-wave (CW) conditions are shown in Fig.  10(a). At the current of 650  mA, light output power is 500  mW and voltage is 5.8  V, which means a wall-plug efficiency of 13%.

In contrast, it took more than a decade to realize GaN-based green laser diodes due to several challenges, which include incompatibility of growth conditions of In-rich InGaN quantum wells (and GaN and AlGaN as well), large lattice mismatch between In-rich InGaN quantum wells and GaN, and reduced optical confinement due to smaller difference of refractive index between InGaN and AlGaN. The former two issues result in a high density of defects^{[123– 126]} and broadened gain spectra.^{[127– 130]} After great efforts to develop green laser diodes by many groups during the past decade, lasing in the green spectrum range has been realized by several groups such as Nichia, ^[131] Sony and Sumitomo, ^{[132, 133]} Osram, ^{[134, 135]} UCSB, ^{[136, 137]} Corning, ^[138] and Soraa.^{[139, 140]} We have also been developing green laser diodes on c-plane bulk GaN substrates. Green LD structures with excellent luminescence homogeneity and small linewidth have been fabricated on Nanowin’ s ultra-high quality GaN substrates.^{[141, 142]} Lasing has been achieved, ^[142] as shown in Fig.  10(b). The threshold current at room temperature is 400  mA under pulse operation with 300-ns pulse width and 10-kHz repetition frequency.

Fig.  10.

	Figure Option ViewDownloadNew Window
	Fig.  10. (a) Power– current– voltage curves of a blue InGaN LD packaged with TO-56 under room-temperature and continuous-wave conditions; (b) EL spectra of green LD below and above threshold current. Inset shows micro-PL image of the green LD structure, indicating homogeneous luminescence and no dark spot.

The LED market grew rapidly in recent years, reaching 50  million US dollars last year. However, most LEDs in the market are based on foreign substrates because the volume production GaN bulk substrates are still under improvement, and the cost is relatively high. The main challenge inherent in today’ s LEDs is the efficiency droop effect; the related influencing aspects are still under discussion.^[143] Growing LEDs by homo-epitaxy based on GaN bulk is a possible route. Cao et al. were first to demonstrate UV and blue LEDs grown on GaN bulk substrate. The internal quantum efficiency of the UV LED on GaN was higher compared to the UV LED on sapphire, whereas the performance of the blue LEDs was found to be comparable.^[144] Since then, more and more research groups and companies are interested in research on LEDs on GaN bulk substrates.^{[145, 146]} The performance of LEDs grown on GaN bulk substrate has achieved quite remarkable progress in recent years. In 2012, Soraa reported an external quantum efficiency of 68% at 180  A· cm^{− 2} and no current crowding is observed at high current density.^[147] Then Soraa upgrade the  extraction efficiency to 89%^[148] in 2014, and total power conversion efficiency to 84%^[149] in 2015. Besides the above-mentioned LEDs grown on the Ga-face GaN, LEDs grown on non-polar GaN substrates are another hot research field because of the reduction of quantum-confined Stark effect in this structure. Sato et al. demonstrated a green LED with a peak emission wavelength of 516  nm grown on semipolar (11-22) bulk GaN substrate in 2007.^[150] The performance has improved significantly in recent years.^[151] However, suffering from the size limitation of non-polar GaN bulk substrates, the market application of non-polar LEDs is quite confined.

Power semiconductor devices are the key building block for power electronic systems like solar inverters, electric vehicle drives, the three-phase motor driver in hybrid vehicles, and so on. With the wide band gap and thus high critical field, GaN is a promising material for high power devices and has attracted much attention since silicon power devices are approaching physical limits of the materials. While high power devices based on SiC are more mature, GaN-based devices have the potential to be better because the fundamental (material based) figure of merit (FOM) parameters for GaN are larger than those of SiC.^[152] However, the performance and reliability of most lateral GaN-based power devices have fallen short of their potential because the quality of GaN layers grown on foreign substrates (e.g. sapphire or Si) is poor due to the high dislocation density (> 10⁸  cm^{− 2}) and because lateral device architectures are less well suited for high voltage and high current power applications. Low defect density is important for power devices because it can affect the performance characteristics (e.g. breakdown voltage and off-state leakage current), yield, and reliability. By fabricating the power devices on high-quality bulk GaN substrates, realizing the material limit potential of GaN is expected to be possible including true avalanche breakdown capability, to create vertical architectures that do not suffer from the thermal management issues, and to permit an increased number of die on a wafer and improve device reliability. Recently, a variety of GaN-based power devices have been successfully obtained on the bulk GaN substrates. For example, Nie et al.^[153] fabricated vertical GaN transistors using 2-inch bulk GaN substrates with a low dislocation density (10⁴  cm^{− 2}), and the transistors exhibit saturation current exceeding 2.3 A, breakdown voltages of 1.5  kV, area differential specific on-resistance of 2.2  mΩ · cm², and an FOM of 1.0 × 10⁹  V² · Ω ^{− 1}· cm^{− 2} with a junction termination extension scheme. Freestanding GaN substrates were also used to fabricate submicrometer gate-length (L_G) high electron mobility transistors (HEMTs) with excellent dc and RF performance, ^[154] and L_G = 100  nm devices exhibited a drain current density of 1.5  A/mm, current gain cutoff frequency f_T of 165  GHz, a maximum frequency of oscillation f_max of 171  GHz, and intrinsic average electron velocity of 1.5 × 10⁷  cm/s. The vertical breakdown voltage of the heterostructure field effect transistor (HFET) on the Na-flux grown GaN substrate is reported to be over 3000  V for a GaN buffer layer thickness of 10  μ m, giving a breakdown field of 3.2  MV/cm, almost equal to the physical limit of GaN (3.3  MV/cm).^[78] Besides  that, vertical p– n diodes are also successfully fabricated on pseudo-bulk low defect density (10⁴  cm^{− 2}– 10⁶  cm^{− 2}) GaN substrates, ^[155] and the measured devices demonstrate near power device figure of merit, that is, differential specific on-resistance of 2  mΩ · cm² for a breakdown voltage of 2.6  kV and 2.95  mΩ · cm² for a 3.7  kV device, respectively. In 2010, Saitoh et al.^[156] fabricated the vertical GaN Schottky barrier diodes (SBDs) with field plate structure on freestanding GaN substrates, and the specific on-resistance (R_onA) and the breakdown voltage (V_B) of the SBDs were 0.71  mΩ · cm² and over 1100  V, respectively. The figure of merit was 1.7  GW/cm², which surpassed the theoretical limit of 4H– SiC for the first time. Ozbek et al.^[157] also obtained high-voltage GaN Schottky barrier diodes on a bulk GaN wafer with a finite termination by high-dose argon ion implantation to create a high resistivity layer at the periphery of the device, and the breakdown voltage reaches 1700  V.

Nitride materials are suitable for high-efficiency tandem solar cells.^{[158, 159]} The band gaps of InGaN ternary alloys have a broad range from 0.65  eV to 3.43  eV, which offers a high degree of freedom in the design of tandem solar cells. And a maximum conversion efficiency of close to 65% is expected by utilizing multiphoton absorption.^[160] However, in practice it is difficult to fabricate high-efficiency photovoltaic cells from InGaN. First of all, indium incorporation and segregation easily occur because of the large lattice mismatch between InN and GaN.^[161] Secondly In_xGa_{1− x}N films suffer from lattice mismatch with the most common substrates such as sapphire and silicon.^[162] To overcome these challenges and improve the crystal quality in the high-effeciency InGaN photovoltaic (PV) devices, it is currently considered that growing InGaN films on the freestanding GaN substrate is an important approach.^[163] The conversion efficiency of the InGaN-based solar cell on the GaN substrate reached 1.4%, approximately 1.5 times that of solar cells on sapphire substrate.^[164]

An investigation of correlations between PV performance and crystal defects demonstrated that the reduction of pit density using high-quality GaN substrate is essential for the realization of high-performance InGaN-based solar cells.^[165] Using a low-TDD GaN substrate prevented the generation of V-shaped defects and increased open-circuit voltage from 1.6  V to 2.2  V and conversion efficiency by 40%.^[166] By the growth of the high-quality InGaN/InGaN superlattice active layer on a GaN substrate, Kuwahara et al. obtained a high conversion efficiency of 2.5% for the InGN-based solar cell.^[164] Young et al. also confirmed that GaN substrate outperformed sapphire substrate in their investigation of high performance InGaN/GaN MQW solar cells.^[167] Most recently, Young et al. further obtained a conversion efficiency of 3.33% under the AM0 spectrum for InGaN solar cell fabricated on the GaN substrates by integrating well-designed broadband optical coatings.^[168]

In summary, the performance of InGaN-based solar cell on GaN substrate is still in progress, and it is necessary to improve the quality of GaN substrate further for the fulfillment of higher-efficient InGaN-based solar cells. Especially GaN substrates with TDD lower than 10³  cm^{− 2} ∼ 10⁴  cm^{− 2}, no V-shaped defects, and high doping will be the key for developing PV devices.