Cubic zincblende gallium nitride for green-wavelength light-emitting diodes

Abstract

Gallium nitride (GaN)-based light-emitting diodes (LEDs) are highly energy efficient and their widespread usage in lighting can induce significant worldwide electricity savings. To achieve white and colour-tuneable lighting, the mixing of light from red-, blue- and green-wavelength LEDs is desired. At present, the efficiency of green-wavelength LEDs is only about half of that of red- and blue-wavelength LEDs, which is also known as the ‘green gap’ problem. Cubic zincblende GaN has the potential to bridge the ‘green gap’ due to the theoretical absence of internal electric fields that plague the commonly used c-plane hexagonal wurtzite structure. This review first looks at the various methods of achieving metastable zincblende GaN, and examines the crystal defects present. Then the different components towards a full zincblende GaN LED structure, including p- and n-type doping, zincblende InGaN and AlGaN on GaN heterostructures, together with the problems that need to be overcome to achieve green-wavelength emissions are reviewed.

This review was submitted as part of the 2017 Materials Literature Review Prize of the Institute of Materials, Minerals and Mining run by the Editorial Board of MST. Sponsorship of the prize by TWI Ltd is gratefully acknowledged

Keywords

Gallium nitride cubic zincblende green gap light-emitting diodes defects doping InGaN

Introduction

In the UK and around the world, about 20% of electricity consumption is typically used on lighting [1]. Therefore, a significant amount of electricity can be saved using highly energy-efficient light-emitting diodes (LEDs) as light sources instead of the traditional incandescent and fluorescent lighting. At present, virtually all commercial LEDs that emit white light consist of a blue-wavelength gallium nitride-based (GaN) LED capped with a phosphor. The phosphor absorbs parts of the blue spectrum and re-emits the other colours in the visible spectrum of longer wavelengths, with the superposition of the different wavelengths producing white light. Since there is inevitable energy loss during the wavelength conversion in the phosphor, a more energy-efficient system of producing white light is to mix light from red, blue and green LEDs, with the additional benefit that their intensities can be varied to produce colour-tuneable lighting. The ability to tune the wavelengths of indoor lighting throughout the day has potential health benefits, as studies have suggested that exposure to bluewavelengths in the evenings through LED-lit devices suppresses the body's production of melatonin, a hormone that promotes sleepiness [2].

LED structure and operation

A schematic of the structure of an LED is shown in Figure 1. Applying a forward-bias across the p- and n-contacts results in holes generated in the p-type layer and electrons generated in the n-type layer moving towards the InGaN/GaN multiple quantum wells (MQWs) region. Since the GaN/InGaN heterostructure is Type I as shown in Figure 3(c), the carriers are spatially confined within the InGaN quantum wells where radiative recombination occurs between electrons and holes to produce photons. It is also possible that non-radiative recombination takes place instead, for example due to impurity states within the band gap to produce phonons. The efficiency of an LED is often expressed in terms of the internal quantum efficiency (IQE) and external quantum efficiency (EQE), which are the ratio of radiative electron-hole recombinations to number of injected carriers, and the ratio of emitted photons to injected carriers, respectively.

Figure 1

Schematic of an LED structure with InGaN/GaN MQWs as active layers and light extraction via a cleaved side facet.

The emission wavelength of the InGaN quantum wells can be tuned by the indium content (by changing the band gap energy, as shown in Figure 2) and the thickness of the quantum wells. Higher indium contents and thicker quantum wells are required to achieve green emission wavelengths. From Figure 2, which presents Γ-valley energy gap values for relaxed III-nitride layers at 300 K, the indium contents of In_xGa_1−xN alloys that have green emission band gaps of 2.3 eV are x = 0.26 and 0.31 for the zincblende and wurtzite structures, respectively. Since the InGaN quantum wells are strained in LEDs, the band gap is calculated to be 0.173 eV larger for fully strained In_xGa_1−xN (x = 0.26) layers compared with relaxed layers, using the method in Morkoc et al. [5] and material constants from Vurgaftman et al. [3].

Figure 2

Γ-valley energy gaps at 300 K of wurtzite and zincblende III-nitrides. Energy gaps are plotted as a function of lattice parameter a_wz for wurtzite and a_zb/√2 for zincblende, both of which correspond to the distance between group-III ions in the (001) plane. Values from Vurgaftman et al. [3] and As et al. [4].

The ‘green gap’ problem

Currently, green indium gallium nitride (InGaN) LEDs have only less than half the EQE of blue InGaN and red aluminium gallium indium phosphide (AlGaInP) LEDs [6], and this is known as the ‘green gap’ problem. Phosphide-based LEDs for green-wavelength emission suffer from an indirect bandgap which lowers the radiative recombination efficiency. For nitride-based LEDs, the band gap is direct, but there are several possible factors that can contribute to the ‘green gap’ problem, with all of them related to the requirement that a higher indium content of around 11% is required in the InGaN quantum wells for emission of green wavelengths (540 nm) compared to blue wavelengths (470 nm) for the wurtzite crystal structure.

Most of the commercial LEDs are currently grown in the c-plane (0001)-oriented hexagonal wurtzite crystal structure, which results in piezoelectric and spontaneous polarisation fields in the [0001] growth direction across the LED device structure. Electric fields across the LED structure result in the quantum confined Stark effect (QCSE), which is the spatial separation of electron and hole wavefunctions within a quantum well, as shown in Figure 3(d). (A more detailed discussion of the QCSE and its effects is given by Ren [8]). Although all working diodes are subjected to the QCSE due to external bias, the internal piezoelectric and spontaneous electric fields present in the c-plane wurtzite crystal structure exacerbate the effect. With a smaller spatial overlap between the electron and hole wavefunctions, there is a lower rate of radiative recombination. In order to reduce such internal electric fields, many groups have focussed their studies on wurtzite with non-polar (m-plane and a-plane) and semi-polar orientations, but the efficiencies such LEDs have not been as high as expected and fall behind those of c-plane wurtzite LEDs [8]. Throughout this review, the c-plane orientation is implied where the wurtzite structure is mentioned, unless explicitly stated otherwise.

Figure 3

(a) and (b) Crystal structures of cubic zincblende and hexagonal wurtzite III-nitrides, respectively. Reprinted from Hanada [7], with the permission of Springer. Band diagrams of an InGaN quantum well with GaN barriers without external bias: (c) in the absence of internal electric fields for the (001)-oriented zincblende structure, (d) in the presence of internal electric fields for the c-plane (0001)-oriented hexagonal wurtzite structure resulting in bending of the band structure. The blue and orange curves represent the electron and hole wavefunctions, respectively.

Increased indium concentrations and thickness in the InGaN quantum wells result in larger lattice mismatch and strains with the GaN quantum barriers, thus larger piezoelectric polarisation fields. Again, this results in a lower rate of radiative recombination. Moreover, with higher indium content, the density of point and extended defects increases and inhomogeneous indium concentrations are observed in c-plane wurtzite GaN [9] to result in a higher rate of non-radiative recombination. The lower rates of radiative recombination and higher rates of non-radiative recombination from the above factors result in the decrease in IQE and EQE of c-plane wurtzite GaN LEDs of decreasing wavelengths from blue to green.

Properties of cubic zincblende GaN and InGaN

The cubic zincblende () structure and the hexagonal wurtzite () structure are two different crystal polytypes of the same material, differing in the arrangement of the atoms within the material. The most obvious advantage that zincblende InGaN possesses over wurtzite InGaN for green-wavelength emission is the absence of spontaneous and piezoelectric polarisation fields in the [001] growth direction, due to the nature of the crystal structures as depicted in Figure 3(a) and (b) (Hanada [7]). The lattice parameters a = b = c for zincblende, whereas a ≠ c for wurtzite. Therefore, in the zincblende structure, strains due to the lattice mismatch between GaN barriers and InGaN wells should not give rise to piezoelectric polarisation fields and the associated QCSE described previously. Additional strains from higher indium content and thicker quantum wells for green-wavelength emission also do not contribute to internal electric fields for zincblende InGaN, thus both indium content and thickness of wells can be tuned in a wider range without reducing recombination efficiency. Moreover, zincblende InGaN has a smaller band gap than wurtzite InGaN by around 0.14 eV at 300 K, as shown in Figure 2. Therefore, zincblende InGaN quantum wells require lower In incorporation compared with equivalent non-polar or semi-polar InGaN quantum wells. However, zincblende GaN is a metastable phase, while wurtzite GaN is the thermodynamically more stable phase, so careful selection and optimisation of growth conditions are required.

Zincblende GaN growth methods

GaN- and GaN-based heterostructures are grown in an epitaxial manner, where a crystalline thin film is deposited monolayer by monolayer on a crystalline substrate. The orientation and crystal structure of the thin film are determined by the substrate. Zincblende GaN has been mainly grown by molecular beam epitaxy (MBE) and metal-organic vapour phase epitaxy (MOVPE). MBE is an ultra-high vacuum technique in which the ultra-pure forms of the elements are made to sublime, by either heating or electron beam evaporation. The fine molecular beams formed in this way are directed to a heated substrate, where the atoms of the different elements become physically adsorbed and subsequently incorporated into the crystal lattice through formation of bonds with neighbouring lattice atoms. The low growth rates (50–400 nm/h for zincblende GaN) of this process notionally allow precise atomic layer control of the growth. As for MOVPE, precursors are carried into the reaction chamber by a carrier gas that becomes saturated with the precursors as it is bubbled through solutions of the precursors. In the reaction chamber, the precursors flow laminar above the heated substrate and decompose by pyrolysis to leave behind the pure elements (e.g. Ga, In and N). Like MBE growth, the pure elements are then physically adsorbed on the heated substrate and incorporated into the crystal lattice. MOVPE does not require an ultra-high vacuum and has a faster growth rate compared with MBE, hence it is widely employed in industry for the growth of commercial GaN-based LEDs.

The growth of GaN is usually heteroepitaxial, which means that the substrate is not GaN but another material, since bulk GaN substrates are not readily available and expensive. The choice of substrate is vital to ensure zincblende GaN growth. The main substrates that have been studied include cubic GaAs (001) (by MBE [10 –14] and MOVPE [15 –21]), 3C-SiC (by MBE [22 –26] and MOVPE [27]), 3C-SiC on Si by MBE [28 –33], patterned 3C-SiC on Si [34] and patterned Si [35 –37] substrates. Some of the substrate properties to consider are lattice matching and difference in thermal expansion coefficient with zincblende GaN and cost. Cubic GaAs substrates and zincblende GaN have a large lattice mismatch of 20% resulting in a high density of defects in the zincblende GaN layer [38]. 3C-SiC (3C simply denotes the cubic zincblende structure of SiC) substrates have a much lower lattice mismatch of 3% with zincblende GaN which thus produces zincblende GaN with improved crystallinity and lower content of hexagonal inclusions [22,23,25]. However, the limiting factor to the success and commercialisation of the growth of zincblende GaN on 3C-SiC is the prohibitively high cost of the substrate compared with silicon substrates. By depositing only a few micrometres of 3C–SiC onto silicon substrates, as shown in Figure 4, the best of both worlds can be achieved. The commercial viability of zincblende GaN LEDs grown on such substrates is also increased by the availability of large area (up to 8 inches in diameter) Si substrates. 3C-SiC layers have been grown on silicon by chemical vapour deposition [32], carbonising with a carbon-containing gas such as C₂H₂ [39,40], carbonisation in rapid thermal processing reactor [31] and high-dose carbon ion implantation [33]. More recently, Stark et al. [35] and Bayram et al. [36] have both used patterned Si substrates with {111} side walls, although in slightly different geometries. Referring to Figure 5, the hexagonal wurtzite GaN phase first nucleates on the two opposite {111} side walls, and the zincblende phase forms where the two wurtzite crystals meet [35]. Such methods result in stripes of zincblende GaN crystal being formed, with a width of ∼1 µm and a length limited by the size of the substrate.

Figure 4

Zincblende GaN epilayer grown by MBE on thin -SiC/Si (001) substrates. Reprinted from As et al. [31], with the permission of AIP Publishing.

Figure 5

Coalescence of the two wurtzite crystal nucleated at the Si {111} side walls results in the formation of zincblende GaN phase. Reprinted from Stark et al. [35], with the permission of AIP Publishing.

Thin film growth of zincblende GaN typically consists of several steps. After annealing and cleaning of the substrate, a porous nucleation layer of a few nanometres is often grown at around 600°C [24,31,41 –43] that compensates the lattice mismatch with the substrate. An annealing and crystallisation step is used to smoothen the nucleation layer, before a thicker buffer layer of several hundred nanometres is grown. The substrate temperature during subsequent GaN growth can affect the kinetics of the growth surface and is important in controlling the polytype of GaN that nucleates on the surface. Zincblende GaN is typically reported to be stabilised at lower temperatures of 600–950°C, compared with 1000–1100°C for wurtzite GaN. Moreover, the substrate temperatures reported for MOVPE (900–950°C) [16,42,44,45] are generally higher than those of MBE (600–835°C) [29 –31,33,43,46,47]. The ratio of N to Ga, or V/III ratio, can affect surface diffusion since N adatoms have a higher sticking coefficient than Ga adatoms [39]. A high V/III ratio can thus limit surface diffusion and promote surface roughening, which has been reported to subsequently result in the nucleation of hexagonal inclusions [31,47]. Typical optimised V/III ratios used for MOCVD zincblende GaN growth are 1000–1500 [19,42].

Wurtzite inclusions

The aim of optimisation of growth conditions is to minimise the amount of wurtzite inclusions. The stacking sequence of the close-packed planes in the wurtzite structure is ABABABAB, while that of the close-packed {111} planes in the zincblende structure is ABCABCABC. A stacking fault in the zincblende structure arises when the stacking sequence is disordered to become ABCABABC, where the stacking fault is between the letters highlighted in bold. Therefore, a stacking fault in zincblende GaN has the same stacking sequence as the wurtzite structure, and is thought to act as nucleation sites for wurtzite inclusions. The presence of wurtzite inclusions and stacking faults has been found to be correlated by Kemper et al. [48] and Martinez-Guerrero et al. [32], with the wurtzite inclusions nucleating on the {111} planes.

Crystal defects

Stacking faults in zincblende GaN have been consistently found to lie on the {111} planes when grown on both GaAs [49,50], 3C-SiC [26,32] and patterned 3C-SiC/Si [34] substrates. The prevalence of stacking faults in zincblende GaN is likely due to stacking faults having the same stacking sequence as the wurtzite structure, which is more thermodynamically stable than the zincblende structure. Wang et al. [39] found that many stacking faults originate from the rougher regions of the interface between the zincblende layer and a 3C-SiC/Si substrate. However, it is not clear if the stacking faults originate at the substrate surface steps, at existing stacking faults in the substrate, or at other locations. It is important to understand the origin of stacking faults to reduce their densities. The stacking fault density is found to decrease with the distance from the GaN/substrate interface by Martinez-Guerrero et al. [32], and it has been suggested that stacking faults lying on the {111} planes in the zincblende structure can intersect and annihilate to create a sessile dislocation in the <110> directions.

The light-emitting MQW active region (refer to Figure 1) is arguably the most important part of an LED structure. The effect of stacking faults on the confinement of electrons and holes, and radiative and non-radiative efficiencies of the MQW region are so far not well understood. Kemper et al. [51] reported that stacking faults on the {111} planes in zincblende GaN led to a decrease of the scanning transmission electron microscopy-cathodoluminescence emission intensity. This seems to indicate that stacking faults reduce radiative recombination efficiencies, although it is possible that there is also a corresponding decrease in dislocation density with increasing thickness of GaN films that can also decrease radiative recombination efficiencies. In wurtzite GaN, zincblende stacking faults are thought to act as type–II quantum wells [52,53]. Since zincblende GaN has hexagonal wurtzite stacking faults, it is possible that stacking faults in zincblende GaN have similar type-II quantum well emission properties. Moreover, Kemper et al. [48] reported that stacking faults resulted in MQWs with an undulating morphology; however, whether this affects light emission properties still needs to be investigated. An obvious and simple remedy to avoid the effects of stacking faults is to grow thicker GaN films to reduce the stacking fault density near the MQW light-emitting region of the LED.

Owing to the lattice mismatch between GaN and the substrates, structural defects, such as misfit dislocations, stacking faults and microtwins, have been attributed to relieving strain at the interface [48,54]. Pure edge-type misfit dislocations occurring at regular intervals (every fifth {111} lattice plane of zincblende GaN grown on GaAs substrate) by Trampert et al. [54]. were revealed with high-resolution TEM. With a low formation energy in fcc crystals [55], it is questionable whether stacking faults help to relieve strain. Twinning has been reported for zincblende GaN grown on GaAs [50,54,56]; however, the mechanism by which strain is relieved is not clear.

Antiphase domains present in 3C-SiC/Si substrates can template on to subsequent zincblende GaN growth, as observed by Kemper et al. [57], as shown in Figure 6. A silicon substrate is not atomically flat and exhibits surface steps. Figure 7(a) and (b) schematically illustrate two neighbouring nuclei that have nucleated on different steps of step height . If n is odd, the two nuclei have an antiphase relationship between them and an antiphase boundary results. If n is even, the two nuclei do not have an antiphase relationship and only a vertical displacement boundary results. Antiphase domains in 3C-SiC layers template the subsequent growth of zincblende GaN [57] and can result in lattice defects such as stacking faults in subsequent growth of GaN [58]. It is well known that introducing a miscut to the substrate by cutting the Si wafer at a slight angle (< ∼ 5°) reduces the presence of antiphase domains in GaAs [59] and Si [60]. Huang et al. [60] proposed that the double step height, h₁ was critical for achieving antiphase domain suppression on the miscut Si substrate.

Figure 6

SEM (a) and AFM (b) images of antiphase domains in zincblende GaN. Reprinted from Kemper et al. [57], with the permission of AIP Publishing.

Figure 7

(a) and (b) Schematic of two neighbouring 3C-SiC nuclei that have nucleated on different steps of a stepped (001) Si substrate. The height between the two steps is , where in (a) n = 1 and in (b) n = 2. It is assumed that SiC is more thermodynamically stable than Si and that the first layer that is deposited on the silicon (white circles) surface is a layer of carbon atoms (black circles). ‘DB’ and ‘APB’ are ‘displacement boundary’ and ‘antiphase boundary’, respectively. Reprinted from Pirouz et al. [58], with the permission of AIP Publishing.

Doping of zincblende GaN

Doping basics

GaN is doped to create mobile electrons and holes that can be injected into the MQW active region for radiative recombination. Doping generates donor and acceptor energy levels within the band gap of GaN, as shown in Figure 8, and donors and acceptors need to be ionised (usually thermally) to generate electrons and holes. Donors (acceptors) can be identified as shallow or deep depending on how close or far their energy levels are from the conduction (valence) band. P-type doping of GaN to generate holes can be achieved by replacing the group III gallium site with a group II element (e.g. Mg) or the group V nitrogen site with a group IV (e.g. C). On the other hand, n-type doping to generate electrons can be achieved by replacing the gallium site with a group IV element (e.g. Si) or the nitrogen site with a group VI element.

Figure 8

Energy level diagram showing acceptor, donor and deep donor states. The arrows going across the energy levels represent various radiative recombination mechanisms: (1) free excitons or near band edge emission, (2) from conduction band to shallow acceptor state, (3) from conduction band to deep acceptor state.

The degree of p-type and n-type dopant incorporation can be determined from secondary ion mass spectroscopy (SIMS) measurements. Carrier concentration and mobility are usually measured through Hall effect measurements with the Van der Pauw method. Finally, the ionisation energy of a dopant is often determined from photoluminescence (PL) spectroscopy, where the sample is optically excited by a laser and the emission energy and intensity of subsequent radiative recombinations are detected. Peaks can be assigned to radiative recombination mechanisms by comparing with spectra from known samples or with theoretical predicted values.

P-type doping

Magnesium doping

Since magnesium has been the most successful p-type dopant in wurtzite GaN, many studies have applied magnesium p-type doping to zincblende GaN, despite the high ionisation energy and resulting low hole concentration. Table 1 shows a summary of the Mg concentrations and corresponding hole concentrations and mobilities reported by studies of Mg-doped zincblende GaN epilayers. The low Mg concentration (5 × 10¹⁸ cm⁻³) reported by As et al. [63] stands out. The Mg-doped zincblende GaN epilayers from As et al. [63] were grown by MBE at Mg cell temperature of 300–400°C, which is higher compared with 260°C, used by Martinez-Guerrero et al. [62]. As et al. [63] suggested that the saturation of Mg concentration despite increasing Mg flux from higher Mg cell temperatures was due to the desorption of Mg atoms from the substrate surface at high vapour pressures of Mg. On a possibly related note, Martinez-Guerrero et al. [62] reported that Mg cell temperatures above 250°C resulted in increased roughness of epilayers, which suggests that a high Mg flux affects the surface atomic arrangements of the epilayer during growth.

Table 1

Mg concentrations and corresponding hole concentrations and mobilities reported in studies of Mg-doped zincblende GaN epilayers.

Paper	Growth method	Substrate temperature /°C	Mg concentration /cm⁻³	Hole concentration /cm⁻³	Mobility /cm² V-s⁻¹
Vilchis et al. [61] (2015)	MOVPE	830–870	1 × 10¹⁸–1 × 10¹⁹	5 × 10¹⁷–8 × 10¹⁸	1–30
Martinez-Guerrero et al. [32] (2002)	MBE	Not reported	1 × 10¹⁷–2.5 × 10²⁰	Not reported	Not reported
Martinez-Guerrero et al. [62] (2001)	MBE	680	1 × 10¹⁹	3–5 × 10¹⁸	35
As et al. [63] (1998)	MBE	720	5 × 10¹⁸	1 × 10¹⁶	215
Lin et al. [64] (1993)	MBE	580		8 × 10¹⁶–8 × 10¹⁸	39

Data taken from Vilchis et al. [61], Martinez-Guerrero et al. [32,62], As et al. [63] and Lin et al. [64].

The Mg-doped zincblende GaN epilayers grown by MOVPE from Vilchis et al. [61] have a much higher ratio of hole to Mg concentrations compared with those grown by MBE from Martinez-Guerrero et al. [62] and As et al. [63]. To explain the low hole to Mg ratios for MBE-grown epilayers, it was suggested that background oxygen concentrations from the Mg source resulted in Mg having to compensate the n-type doping, and the formation of Mg complexes and precipitates [32]. It is possible that MOVPE produced higher quality epilayers with fewer structural defects, which trap impurities, [62] compared with MBE to result in more ionised Mg. The composition of wurtzite inclusions of optimised epilayers by Vilchis et al. [61] is less than 5%; however, there are no further structural studies from Vilchis et al. [61], Martinez-Guerrero et al. [62] and As et al. [63] to confirm the connection between defects and ionisation of Mg. Future studies correlating Mg concentration to structural defects, presence of Mg complexes or precipitates and background oxygen levels of Mg-doped zincblende GaN layers will be useful.

The low hole concentrations reported by Martinez-Guerrero et al. [62] and As et al. [63] also suggest that not all the Mg act as shallow acceptors, with some acting as deep Mg-related acceptors, and a broad blue emission band observed in PL spectra [61,65,66] attributed to such deep Mg-related acceptors supports this. Various authors have assigned peaks observed in the PL spectra of Mg-doped zincblende GaN films to the different recombination mechanisms, with Table 2 providing a summary and Figure 8 depicting the mechanisms. From the energies of the PL peaks assigned to shallow Mg acceptors and deep Mg-related acceptors, Xu et al. [66] calculated the respective ionisation energies to be 213 meV and between 215 and 570 meV from the valence band. Regarding the assignment of PL peaks to recombination mechanisms, Xu et al. [66] compared how peaks shifted in energy at different excitation intensities, and verified them using the excitation-power dependence of the intensity of PL peak of different recombination mechanisms [67]. On the other hand, As et al. [63] compared the intensity of PL peaks with Mg concentrations obtained from SIMS Mg concentration data to assign peaks, noting that the intensity of the broad blue peak showed a clear increase with increasing Mg flux. However, it is possible that defects such as C impurities and stacking faults also give rise to peaks in the PL spectra. Moreover, since stacking faults are essentially wurtzite material within zincblende GaN, they can result in quantum wells to give rise to PL peaks. Future studies should correlate PL and SIMS measurements of Mg-doped p-type zincblende GaN layers with structural studies of defects.

Table 2

PL spectra peaks for Mg-doped zincblende GaN.

Recombination mechanism	Assigned peak in PL spectra/eV
Free exciton	3.25 (Powell et al. [65], Vilchis et al. [61])
Near band edge	3.15 (Powell et al. [65])
Shallow Mg acceptor (and conduction band)	3.05 eV (Powell et al. [65])
	3.00 (Vilchis et al. [61])
	3.037 (Xu et al. [66])
Deep Mg-related acceptors (and conduction band)	2.95 (Powell et al. [65])
	2.82 (Vilchis et al. [61])
	2.895, 2.716, 2.639 (Xu et al. [66])
	Broad peaks observed in all

Several authors [32,63,64] have reported the tendency for Mg to segregate towards the surface of the Mg-doped zincblende GaN films at such Mg concentrations, with Lin et al. [64] suggesting that a possible cause is the difference in covalent radius between Mg and Ga. As et al. [63] also observed Mg segregation towards the GaN/GaAs substrate interface that could be due to increased Mg diffusion into GaAs or the high density of defects near the interface.

Another problem, which applies mainly to MOVPE growth, is the formation of neutral H-Mg complexes. Since hydrogen is generally present in a MOVPE reactor, either as a carrier gas or an impurity in precursors, it is likely to incorporate into the thin film to form H-Mg complexes. Thermal annealing at 700–900°C is usually used for Mg-doped wurtzite GaN to activate neutral Mg-H complexes to form Mg acceptors, by providing the energy and time for the complex to dissociate and H to diffuse out of the structure [6]. However, Vilchis et al. [61] reported that no thermal annealing was required for MOVPE-grown Mg-doped zincblende GaN, based on a study by Guarneros et al. [68] on similar Mg-doped wurtzite GaN films that showed no increase in hole concentrations with annealing. A high Mg to hole concentration ratio was obtained by Vilchis et al. [61], and it is possible that the growth temperature of 830–870°C (around the annealing temperatures used for wurtzite Mg-doped GaN), long growth time of 60 mins and thin film of 1 µm allowed most of the H to dissociate and diffuse out. As et al. [69] also reported that no annealing step after the growth of Mg-doped zincblende GaN was necessary without providing reasoning, but the Mg-doped zincblende GaN films were grown by MBE in ultra-high vacuum with an absence of a hydrogen source.

Carbon doping

Owing to the difficulties with magnesium p-type doping, a few studies have investigated p-type doping of zincblende GaN by carbon [70 –72]. Carbon is a group IV element, which can either replace the group V nitrogen site as an acceptor or the group III gallium site as a donor, resulting in potential amphoteric behaviour. Since carbon is more similar to nitrogen in atomic radius and electronegativity, it is expected to replace the nitrogen site instead of the gallium site.

Carbon doping has been achieved through electron beam evaporation of a graphite rod [70] and sublimation of solid CBr₄ powder [71,72]. The maximum carbon incorporation concentrations were 2 × 10²⁰ cm⁻³ from As et al. [70] and 1 × 10²⁰ cm⁻³ from Zado et al. [72] before the zincblende GaN film degraded in quality. The two authors who studied carbon doping of zincblende GaN each identified factors to explain the observed variation in hole concentrations with flux of carbon. With increasing electron beam power on the carbon source, As et al. [70] observed an initial rise of hole concentration to a maximum of 6 × 10¹⁷ cm⁻³ before a dramatic decrease, which suggested the introduction of compensating centres at high carbon flux. It is also possible that the amphoteric nature of carbon causes carbon to act as a p-type dopant at low carbon flux and n-type dopant at higher carbon flux with increasing electron beam power on the graphite rod; however, Zado et al. [72] did not observe tendency towards n-type behaviour with increasing carbon incorporation. Instead, the net donor concentration decreased before the net acceptor concentration increased to a maximum of 1 × 10¹⁹ cm⁻³, as shown in Figure 9. The result was explained by the carbon atoms first compensating unintentional n-type doping of the undoped layer, possibly due to oxygen from the nitrogen source for GaN, before acting as acceptors. It is difficult to compare the results from As et al. [70] and Zado et al. [72] directly since the two sources of carbon are different, which might affect the mechanism of incorporation of carbon into the zincblende GaN layers.

Figure 9

As carbon concentration in an undoped zincblende GaN film increased (diamonds), carbon first compensated the unintentional n-type doping (squares show net donor concentration), before acting as an acceptor to result in p-type doping (circles show net acceptor concentration). Reprinted from Zado et al. [72], with the permission of Elsevier.

Overall, the results from the three initial studies of carbon p-type doping seem encouraging for further investigation, with questions that include whether hole mobilities are improved compared with magnesium doping, and whether the amphoteric nature of carbon dominates above a certain degree of incorporation.

N-type doping

Traditionally, the group III element Si is used for n-type doping in wurtzite GaN, due to the similar size of Ga and Si ions resulting in a high incorporation efficiency of Si on Ga-sites. Moreover, the shallow donor ionisation energy of Si in wurtzite GaN leads to high concentrations of free carriers with high mobilities at room temperature. Owing to the similarities between wurtzite and zincblende GaN, similar doping properties can be expected when using Si as n-type dopant.

The Si and electron concentrations reported by studies of Si-doped zincblende GaN layers are plotted in Figure 10. The Si concentrations (1 × 10¹⁸ cm⁻³ to 1 × 10²⁰ cm⁻³) and electron concentrations (2 × 10¹⁸ cm⁻³ to 3.9 × 10²⁰ cm⁻³) are both higher compared with Mg and hole concentrations, and in the case of Martinez-Guerrero et al. [62], the electron concentration is actually higher than the Si donor concentration. The two following factors described can help explain these observations. First, the electron concentration of Si-doped zincblende GaN layers was found to follow the Si concentration curve, as shown in Figure 11, which indicates that most of the Si atoms are located on the Ga sites and ionised [62,75]. Second, unintentional n-type doping can increase the overall electron concentration, either by Si that diffuses from an underlying SiC substrate into GaN layer (typical unintentional Si doping is 2–7 × 10¹⁷ cm⁻³) [62], or by oxygen (e.g. from NH₃ nitrogen source) since Xu et al. [73] reported that undoped GaN grown on GaAs substrate by MOVPE has an electron concentration of 1 × 10¹⁷ cm⁻³.

Figure 10

Si (diamonds) and electron (circles) concentrations in cm⁻³ of Si-doped n-type zincblende GaN from the literature, with information on growth method and substrate. Data from [31,39,61,63,73,74].

Figure 11

Electron concentrations (squares) from Hall measurements and Si concentrations (circles) from SIMS of Si-doped zincblende GaN films grown on SiC/Si (001) substrate. Reprinted from Martinez-Guerrero et al. [62], with the permission of Elsevier.

There has been concern whether high concentrations of Si incorporation lead to defects and wurtzite inclusions in the zincblende GaN layers. {111} stacking fault density was reported to increase with Si incorporation when Wang et al. [40] compared reflection high-energy electron diffraction (RHEED) patterns and cross-sectional TEM micrographs of undoped and Si–doped layers with Si concentration of 1 × 10¹⁹–1 × 10²¹ cm⁻³. By comparing samples of varying Si concentration, Xu et al. [73] attributed the UV 364 nm emission peak of the room temperature PL spectrum of a film with Si incorporation of 4 × 10¹⁹ cm⁻³ to the wurtzite phase. Since stacking faults are essentially wurtzite inclusions in zincblende GaN and often thought to act as nucleation sites for the wurtzite phase, the interpretations of the presence of stacking faults by Wang et al. [40], and wurtzite phase by Xu et al. [73] of Si-doped zincblende GaN layers support each other.

The dislocation density of Si-doped sample with a high Si concentration (1 × 10²¹ cm⁻²) was found to be ∼1 × 10⁹ cm⁻², mostly at small-angle grain boundaries within the grains, while the dislocation density of the undoped sample was ∼5 × 10⁸ cm⁻² found mostly at grain boundaries from plan-view TEM images by Wang et al. [40]. However, at lower Si concentrations between 10¹⁹ and 10²⁰ cm⁻³, Martinez-Guerrero et al. [32] reported that the FWHM of an X-ray diffraction ω-scan showed similar values of mosaicity (misorientation of different crystallites within the film), which means the Si-doped layers should have similar dislocation densities.

Li et al. [76] claimed that the broadening of FWHM of PL spectra of c-GaN as a function of Si doping concentration was due to the microscopic fluctuations of doping concentration. However, it is unclear what sort of length scales the author means for the fluctuations, nor is there more direct evidence of such doping fluctuations.

With the variety of defects that seem to accompany Si incorporation into zincblende GaN films, it will be useful to understand the extent of their effects on the mobility of electrons.

Zincblende InGaN on GaN heterostructures

The ability to form InGaN on GaN heterostructures is vital for the light-emitting InGaN/GaN MQW region of an LED. In terms of the interfaces within the zincblende InGaN/GaN MQWs, Li et al. [77] and Kitamura et al. [43] concluded that the interfaces have high structural qualities, with X-ray diffraction superlattice satellite peaks up to fifth order. Within zincblende InGaN films, several authors reported the presence of In-rich phases [78,79] or phase-separated quantum dots [80], and suggested the cause to be the spinodal decomposition of InGaN. Such films studied were thick (300 nm) and therefore relaxed, which is not particularly relevant to LEDs where InGaN layers are thin (typically <10 nm) and biaxially strained.

In order to investigate the effect of biaxial strain, Tabata et al. [81] performed calculations of mixing free energy for zincblende In_xGa_1−xN alloy as a function of composition at a temperature range between 700 and 1100 K, for both fully relaxed and biaxial-strained films. Results predicted for films with x ≅ 0.4 that spinodal decomposition would take place on relaxed films but be suppressed on biaxial-strained films. Subsequently, Tabata et al. [81] compared the micro-Raman spectra of thin (3 nm) layers to thicker (30 nm) layers, which correspond to strained and relaxed layers notionally, between a 400 nm GaN buffer layer and 30 nm GaN cap. The thinner layers did not show the peak corresponding to the In-rich phase (x ≅ 0.5–0.6) present in relaxed layers, even though they were annealed so that total time at growth temperatures was similar as for thicker layers, giving rise to the conclusion that biaxial strain suppressed spinodal decomposition in InGaN quantum wells. However, it might be the case that the strained layers used are so thin (3 nm) in the study that the Raman measurements do not pick up the presence of any In-rich phase. Pacheco-Salazar et al. [82] also claimed that the minute volume taken up by the In-rich phase means that only PL measurements can detect their presence due to high recombination efficiency of the In-rich phase, but not by other optical methods such as Raman spectroscopy. Moreover, Husberg et al. [83] found the 30 nm In_0.56Ga_0.44N layer between a 300 nm GaN buffer and 30 nm GaN layer to be strained instead of relaxed through high-resolution X-ray diffraction measurements, in contrast to Tabata et al. [81]. The InGaN layer also showed evidence of an In-rich phase (x = 0.56) through PL measurements and reciprocal space map from X-ray diffraction.

Using first principles total energy calculations, Teles et al. [84] showed that biaxial strain suppresses spinodal decomposition of InGaN alloys by encouraging the formation of ordered [n,m] superlattices consisting of n planes of In and m planes of Ga stable up to 1000 K, with the [3,3] superlattice at x ≅ 0.5 having the lowest excess energy. Therefore, Teles et al. [84] believe that the In-rich phases (x = 0.56) reported by Husberg et al. [83] could be such an ordered structure of InGaN; however, the reciprocal space map from X-ray diffraction does not show oscillation fringes that would be expected from the periodic structure. No stable ordered structure was found for thick unstrained InGaN alloys, hence spinodal decomposition is predicted to occur as mentioned in Silveira et al. [78], Tabata et al. [79] and Lemos [80]. The direct observation of such ordered InGaN structures and their effects on localisation of carriers and light emission properties have yet to be investigated.

Zincblende AlGaN on GaN heterostructure

AlN has the largest band gap in the III-nitrides and, if alloyed with GaN, results in AlGaN that has a larger band gap than InGaN. Therefore, AlGaN with low Al content would be useful as confinement and electron blocking layers (refer to Figure 1) for zincblende GaN LEDs, similar to wurtzite GaN LEDs. In wurtzite GaN LEDs, electron blocking layers are usually made from higher Al content AlGaN layers to create an even higher band gap layer, since the diffusion constant and hence leakage current of electrons are usually greater than holes in III-V semiconductors [85]. An alternative is to use Al_xIn_1−xN, which has a larger band gap compared with InGaN and GaN at all x compositions with the possibility of providing very close lattice matching, as shown in Figure 2.

So far, there is no known literature on AlGaN on GaN as an electron blocking layer in zincblende GaN LEDs. Instead, studies on zincblende AlGaN have been performed for AlGaN epilayers [86,87], AlN/GaN MQWs [26], AlGaN/GaN tunnel diodes [88], distributed Bragg reflectors [89] and laser diodes [90]. No major issues with the growth of such AlGaN/GaN heterostructures, such as phase separation, wurtzite inclusions and lattice matching, were reported. Zainal et al. [88] used in situ RHEED, and ex situ X-ray diffraction and PL to confirm the presence of high-quality zincblende GaN and Al_0.3Ga_0.7N layers with negligible wurtzite inclusions, while Li et al. [89] found no cracks for the 15.5 layer-stacks of c-Al_0.3Ga_0.7N/GaN, where the thicknesses of the AlGaN and GaN layer were 55 nm and 52.7 nm, respectively, from scanning electron microscopy.

Optical properties of zincblende InGaN/GaN quantum wells

The optical properties of zincblende GaN-based structures ranging from simple p-n junctions [18,30,69], to MQWs [77], to full LED structures [35] have been reported. Since one of the major motivations for zincblende GaN is to produce efficient LEDs that emit green wavelengths, it is encouraging that Li et al. [77] showed strong green (520 nm) PL emission from In_xGa_1−xN/GaN six-period MQWs [77]. In theory, zincblende GaN MQWs should have an absence of internal polarisation fields along the (001) growth direction. Li et al. [77] and Stark et al. [35] provided results that supported this. For wurtzite InGaN/GaN MQWs, as the quantum well thickness increased, PL intensity was observed to decrease due to larger internal electric fields separating the electrons and holes [91]. On the other hand, Li et al. [77] observed that PL intensity increased with quantum well thickness for zincblende InGaN/GaN MQWs. Stark et al. [35] showed that the peak wavelength of the electroluminescence spectra of zincblende InGaN/GaN MQWs remained constant as current density was increased, while the peak wavelength for wurtzite LEDs usually shifts. The internal piezoelectric and spontaneous electric fields result in larger band bending of wurtzite InGaN quantum wells compared with those with zincblende structure. If the current density across the wurtzite LED increases, the electron and hole wavefunctions across a quantum well shift and become more spread out, resulting in overlap across different band gap energies and shift in peak wavelength of the electroluminescence spectra. For the zincblende LED, the band bending of an InGaN quantum well is not as severe, hence an increase in current density does not result in pronounced shifts in carrier wavefunction peaks and their overlap.

Other issues related to zincblende GaN-based LEDs

Diode leakage

Several studies [18,30,35,69] have reported leakage currents when the structures are under reverse bias, at voltages as low as − 3 V [30]. There are different opinions regarding the cause of such leakage currents. For p-n junctions formed from Mg-doped p-type GaN and Si-doped n-type GaN, Gamez-Cuatzin et al. [30] and As et al. [69] suggested deep levels in the band gap resulted in the leakage currents. From room temperature electroluminescence spectra, Gamez-Cuatzin et al. [30] and Yang et al. [18] observed a blueshift in the luminescence peak with increasing injection current and attributed this to the saturation of deep states. PL at 10 K showed a broad 2.8 eV band associated with deep Mg states, and the electroluminescence spectra showed an increasing Mg-related blue band (between 2.5 and 2.8 eV) at higher temperatures as Mg deep levels are ionised [30]. Yang et al. [18] also reported the blue (∼2.8 eV) peak that dominated in a room temperature electroluminescence spectrum. However, As et al. [69] did not observe the blue Mg-related band in a room temperature electroluminescence spectrum. Instead, the leakage current and deep levels were attributed to the large dislocation density (∼10¹⁰ cm⁻²) in the GaN layer.

Contacts

Low resistance Ohmic contacts are vital to connect the n-type and p-type layers of an LED structure to an external power source for forward biasing. An Ohmic contact has a negligible potential barrier to carriers flowing in either direction between the metal and semiconductor. This is as opposed to a Schottky contact which has a significant potential barrier and is undesirable for LEDs. The formation of an Ohmic contact is influenced by the following factors: the work functions of the metal and zincblende GaN and resultant band bending of the semiconductor, Fermi level pinning by surface states, and tunnelling effects from a thin contact or heavy doping of the semiconductor.

At the time of writing, there is no known literature that specifically investigates the contact formation for zincblende GaN. However, contacts have been formed on zincblende GaN structures to study electrical properties, and the metals chosen are summarised in Table 3. Different metals have been used as contacts for p-type and n-type zincblende GaN because if only the resultant band bending of the contact is considered, a metal with a larger (smaller) work function compared with p-type (n-type) zincblende GaN is required to form an Ohmic contact. Since zincblende GaN has a relatively wide band gap of 3.302 eV and electron affinity of 4.1 eV, it has a work function of ∼7.4 eV. To form an Ohmic contact to p-type zincblende GaN, metals with higher work functions, such as gold and nickel, are used.

Table 3

Contact metals for p-type and n-type zincblende GaN.

	p-type	n-type
Vilchis et al. [61] (2015)	Ni/Au	Ti/Au
As et al. [69] (2000)	Semi-transparent Au	Sn
Taniyasu et al. [92] (2000)	Semi-transparent Au	In (to backside of n-type GaAs substrate)
Yang et al. [18] (1999)	AuZn	AuGeNi
Gamez-Cuatzin et al. [30] (1999)	200 nm Au/ 30 nm Ni	Al/Ti

Internal and external quantum efficiencies (IQE and EQE)

With the lack of full LED structures based on zincblende GaN, the IQE and EQE of such devices have yet to be investigated, although Stark et al. [93] calculated the IQE of an LED grown on v-grooves of Si (001) substrate to be a maximum of 36% from ratio of intensity of CL peaks at 78 K and 300 K. The calculation assumes IQE to be close to 100% at the low temperature, but the assumption is not entirely valid as non-radiative recombinations still exist at such low temperatures [94]. Since much of the more recent research is focused on zincblende GaN grown on Si-based substrates, it is worth bearing in mind that the Si substrate needs to be removed to allow light extraction from the LED. This is evident from the low EQE of 3 × 10⁻⁵ reported by Stark et al. [35] of an LED grown on v-grooves of Si (001) substrate.

Concluding remarks

Zincblende GaN offers an alternative pathway to solving the ‘green gap’ problem, distinct from wurtzite GaN in semi-polar and non-polar orientations. The growth of zincblende GaN on 3C-SiC/Si and wurtzite GaN on Si patterned substrates appear promising, with good lattice matching with zincblende GaN and high commercial viability of Si-based substrates. It is now important to gain a deeper understanding of defects present in zincblende GaN, such as stacking faults and dislocations, including their origin, effects on light emission and elimination methods if required. P-type and n-type doping levels of zincblende GaN are comparable to levels in wurtzite GaN. For p-type doping, Mg incorporation and the ratio of Mg/hole concentrations can still be improved, although there is also the possibility of p-type doping by carbon to be explored further. As for n-type doping, defects resulting from high Si incorporation need to be controlled if they are found to hinder the mobility of electrons. The non-uniformity of In content within InGaN layers reported, with the possibility that they are ordered structures, need to be better understood. With evidence supporting the absence of internal electric fields in zincblende GaN structures, which is a major motivation for zincblende GaN research, the path ahead towards a successful zincblende GaN green-wavelength LED seems worthy of efforts, despite the above challenges that need to be addressed.

Footnotes

Acknowledgements

The author would like to thank her academic supervisor, Dr Rachel Oliver, for guidance and valuable feedback, and to Dr Martin Frentrup for helpful discussions.

Disclosure statement

No potential conflict of interest was reported by the authors.

ORCID

Lok Yi Lee

References

Gunn

, Griffiths

. ‘Lighting Technology’, Postnote 351, The Parliamentary Office of Science and Technology, UK, 2010.

Ford

. ‘Seeing Blue: the Impact of Excessive Blue Light Exposure’, Review of Optometry, 2016, available at https://www.reviewofoptometry.com/article/seeing-blue-the-impact-of-excessive-blue-light-exposure, accessed 19-Nov-2016.

Vurgaftman

, Meyer JR . Band parameters for nitrogen-containing semiconductors. J Appl Phys. 2003;94(6):3675–3696. doi: 10.1063/1.1600519

. Recent developments on non-polar cubic group III-nitrides for optoelectronic applications. Proc SPIE. 2010;7608:1–15.

Morkoc

. Electronic band structure and polarization effects, Chap. 2. In: Handbook of nitride semiconductors and devices, Volume 1, materials properties, physics and growth, Weinheim: Wiley-VCH; 2008. p. 176.

Krames

, Shchekin

, Mueller-Mach

, Status and future of high-power light-emitting diodes for solid-state lighting. IEEE/OSA J Disp Technol. 2007;3(2):160–175. doi: 10.1109/JDT.2007.895339

Hanada

. Basic properties of ZnO, GaN, and related materials. In: Yao

, Hong

S-K

, editors. Oxide and nitride semiconductors. Vol. 12. Berlin Heidelberg: Springer-Verlag; 2009. p. 1–19.

Ren

. Polarisation fields in III-nitrides: effects and control. Mater Sci Technol. 2016;32(5):418–433.

Oliver

. Gallium nitride-based visible light-emitting diodes. Mater Sci Technol. 2016;32(8):737–745. doi: 10.1080/02670836.2015.1116225

10.

Strite

, Ruan

, Li

, An investigation of the properties of cubic GaN grown on GaAs by plasma-assisted molecular-beam epitaxy. J Vac Sci Technol B Microelectron Nanom Struct. 1991;9(4):1924–1929. doi: 10.1116/1.585381

11.

Brandt

, Yang

, Trampert

, Growth of cubic GaN on (001) GaAs. Mater Res Soc Symp Proc. 1996;395:27–37. doi: 10.1557/PROC-395-27

12.

Okumura

, Ohta

, Nagatomo

, Observation of MBE-grown cubic-GaN/GaAs and cubic-GaN/3C-SiC interfaces by high resolution transmission electron microscope. J Cryst Growth. 1996;164:149–153. doi: 10.1016/0022-0248(95)01073-4

13.

Schikora

, Hankeln

, As

, Epitaxial growth and optical transitions of cubic GaN films. Phys Rev B. 1996;54(12):R8381–R8384. doi: 10.1103/PhysRevB.54.R8381

14.

Novikov

, Stanton

, Campion

, Growth and characterization of free-standing zinc-blende (cubic) GaN layers and substrates. Semicond Sci Technol. 2008;23(1):5018. doi: 10.1088/0268-1242/23/1/015018

15.

Kuznia

, Yang

, Chen

, Low pressure metalorganic chemical-vapor deposition of cubic GaN over (100) GaAs substrates. Appl Phys Lett. 1994;65(19):2407–2409. doi: 10.1063/1.112690

16.

Nakadaira

, Tanaka

. Growth of zinc- blende GaN on GaAs (100) substrates at high temperature using low-pressure MOVPE with a low V/lll molar ratio. J Electron Mater. 1997;26(3):320–324. doi: 10.1007/s11664-997-0171-z

17.

, Yaguchi

, Onabe

, Metalorganic vapor phase epitaxy growth of high quality cubic GaN on GaAs (100) substrates. Japanese J Appl Phys. 1998;37:1440–1442. doi: 10.1143/JJAP.37.1440

18.

Yang

, Zheng

, Li

, Cubic-phase GaN light-emitting diodes. Appl Phys Lett. 1999;74(17):2498. doi: 10.1063/1.123019

19.

Moret

, Ruffenach-Clur

, Moreaud

, MOCVD growth of cubic gallium nitride: effect of V/III ratio. Phys Status Solidi Appl Res. 1999;176(1):493–496. doi: 10.1002/(SICI)1521-396X(199911)176:1<493::AID-PSSA493>3.0.CO;2-6

20.

Feng

, Yang

, Zhagn

, Effect on the optical properties and surface morphology of cubic GaN grown by metalorganic chemical vapor deposition using isoelectronic indium surfactant doping. J Cryst Growth. 2002;235:207–211. doi: 10.1016/S0022-0248(01)01914-5

21.

Vilchis

, Sanchez

VM-R.

, Escobosa

. Cubic GaN layers grown by metalorganic chemical vapor deposition on GaN templates obtained by nitridation of GaAs. Thin Solid Films. 2012;520(16):5191–5194. doi: 10.1016/j.tsf.2012.03.123

22.

Paisley

, Sitar

, Posthill

, Growth of cubic phase gallium nitride by modified molecular-beam epitaxy. J Vac Sci Technol A Vacuum, Surfaces, Film. 1989;7(3):701. doi: 10.1116/1.575869

23.

Daudin

, Feuillet

, Hubner

, How to grow cubic GaN with low hexagonal phase content on (001) SiC by molecular beam epitaxy. J Appl Phys. 1998;84(4):2295–2300. doi: 10.1063/1.368296

24.

Gerthsen

, Neubauer

, Dieker

, Molecular beam epitaxy (MBE) growth and structural properties of GaN and AlN on 3C-SiC(001) substrates. J Cryst Growth. 1999;200(3–4):353–361. doi: 10.1016/S0022-0248(99)00060-3

25.

Pacheco-Salazar

, Li

, Cerdeira

, Growth and characterization of cubic InxGa1−xN epilayers on two different types of substrate. J Cryst Growth. 2005;284(3–4):379–387. doi: 10.1016/j.jcrysgro.2005.07.049

26.

, Kemper

, Mietze

, Spatially resolved optical emission of cubic GaN/AlN multi-quantum well structures. Mater Res Soc Symp Proc. 2014;1736. doi: 10.1557/opl.2014.944

27.

, Yaguchi

, Nagasawa

, Crystal structure of GaN grown on 3C-SiC substrates by metalorganic vapor phase epitaxy. Japanese J Appl Phys. 1997;36:4241–4245. doi: 10.1143/JJAP.36.4241

28.

Liu

, Frenkel

, Kim

, Growth of zincblende-GaN on beta -SiC coated (001) Si by molecular beam epitaxy using a radio frequency plasma discharge, nitrogen free-radical source. J Appl Phys. 1993;74(10):6124–6127. doi: 10.1063/1.355176

29.

Hiroyama

, Tamura

. Effect of very thin SiC layer on heteroepitaxial growth of cubic GaN on Si (001). Japanese J Appl Phys. 1998;37:L630–L632. doi: 10.1143/JJAP.37.L630

30.

Gamez-Cuatzin

, Tardy

, Rojo-Romeo

, Electroluminescence characterization of cubic gallium nitride p-n junctions grown on SiC/Si substrates by MBE. Phys Status Solidi Appl Res. 1999;176(1):131–135. doi: 10.1002/(SICI)1521-396X(199911)176:1<131::AID-PSSA131>3.0.CO;2-D

31.

, Frey

, Schikora

, Cubic GaN epilayers grown by molecular beam epitaxy on thin

-SiC/Si (001) substrates. Appl Phys Lett. 2000;76(13):1686–1688. doi: 10.1063/1.126136

32.

Martinez-Guerrero

, Bellet-Amalric

, Martinet

, Structural properties of undoped and doped cubic GaN grown on SiC(001). J Appl Phys. 2002, 91, (8), 4983–4987. doi: 10.1063/1.1456243

33.

Gerlach

, Murphy

, Lindner

JKN

, Structural characterization of cubic and hexagonal GaN thin films grown by IBA – MBE on SiC/Si. J Cryst Growth. 2010;312:762–769. doi: 10.1016/j.jcrysgro.2009.12.048

34.

Kemper

, Weinl

, Mietze

, Growth of cubic GaN on nano-patterned 3C-SiC/Si (0 0 1) substrates. J Cryst Growth. 2011;323:84–87. doi: 10.1016/j.jcrysgro.2010.12.042

35.

Stark

CJM

, Detchprohm

, Lee

, Green cubic GaInN/GaN light-emitting diode on microstructured silicon (100). Appl Phys Lett. 2013;103(23):232107. doi: 10.1063/1.4841555

36.

Bayram

, Ott

, Shiu

K-T

, Cubic Phase GaN on Nano-grooved Si (100) via Maskless Selective Area Epitaxy. Adv Funct Mater. 2014;24(28):4492–4496. doi: 10.1002/adfm.201304062

37.

Liu

, Bayram

. Maximizing cubic phase gallium nitride surface coverage on nano-patterned silicon (100). Appl Phys Lett. 2016;109(4):42103. doi: 10.1063/1.4960005

38.

Powell

, Lee

, Kim

, Heteroepitaxial wurtzite and zinc-blende structure GaN grown by reactive-ion molecular-beam epitaxy: growth kinetics, microstructure, and properties. J Appl Phys. 1993;73(1):189–204. doi: 10.1063/1.353882

39.

Wang

, Hiroyama

, Tamura

, Heteroepitaxial growth of cubic GaN on Si(001) coated with thin flat SiC by plasma-assisted molecular-beam epitaxy. Appl Phys Lett. 2000;76(13):1683. doi: 10.1063/1.126135

40.

Wang

, Yoshida

, Ichikawa

. Si-doped cubic GaN grown on a Si(001) substrate with a thin flat SiC buffer layer. Appl Phys Lett. 2002;80(14):2472. doi: 10.1063/1.1467971

41.

Yang

, Brandt

, Ploog

. MBE growth of cubic GaN on GaAs substrates. Phys Status Solidi B. 1996;194:109–120. doi: 10.1002/pssb.2221940112

42.

Wei

, Xie

, Li

, MOCVD growth of cubic GaN on 3C-SiC deposited on Si (100) substrates. J Electron Mater. 2000;29(3):317–321. doi: 10.1007/s11664-000-0070-z

43.

Kitamura

, Cho

, Ishida

, Growth and characterization of cubic InGaN epilayers on 3C-SiC by RF MBE. J Cryst Growth. 2001;227–228:471–475. doi: 10.1016/S0022-0248(01)00745-X

44.

, Yaguchi

, Nagasawa

, Crystal structure of GaN grown on 3C-SiC substrates by metalorganic vapor phase epitaxy. Japanese J Appl Phys. 1997;36:4241–4245. doi: 10.1143/JJAP.36.4241

45.

, Yaguchi

, Onabe

, Optical properties of cubic GaN grown on 3C-SiC (100) substrates by metalorganic vapor phase epitaxy. Phys Status Solidi A. 2000;180(403):403–407. doi: 10.1002/1521-396X(200007)180:1<403::AID-PSSA403>3.0.CO;2-A

46.

Kim

, Frenkel

, Liu

, Growth by molecular beam epitaxy and electrical characterization of Si-doped zinc blende GaN films deposited on beta-SiC coated (001) Si substrates. Appl Phys Lett. 1994;65(1):91–93. doi: 10.1063/1.113085

47.

Tsuchiya

, Sunaba

, Suemasu

, Growth condition dependence of GaN crystal structure on (001)GaAs by hydride vapor-phase epitaxy. J Cryst Growth. 1998;189–190:395–400. doi: 10.1016/S0022-0248(98)00322-4

48.

Kemper

, As

, Lindner

. Cubic GaN on nanopatterned 3C-SiC/Si (001) substrates. In: Li

, Wu

, Wang

Z. M.

, editor. Silicon-based nanomaterials. New York: Springer; 2013. p. 381–406.

49.

Trampert

, Brandt

, Ploog

. Phase transformations and phase stability in epitaxial

-GaN films. Angew Chem Int Ed Engl. 1997;36(19):2111–2112. doi: 10.1002/anie.199721111

50.

Chandrasekhar

, Smith

, Strite

, Characterization of group III-nitride semiconductors by high-resolution electron microscopy. J Cryst Growth. 1995;152(3):135–142. doi: 10.1016/0022-0248(95)00041-0

51.

Kemper

, Veit

, Mietze

, STEM-CL investigations on the influence of stacking faults on the optical emission of cubic GaN epilayers and cubic GaN/AlN multi-quantum wells. Phys Status Solidi C. 2015;12(4–5):469–472. doi: 10.1002/pssc.201400154

52.

Albrecht

, Christiansen

, Salviati

, Luminescence related to stacking faults in heteroepitaxially grown wurtzite GaN. Mater Res Soc Symp Proc V. 1997;468:293–298. doi: 10.1557/PROC-468-293

53.

Stampfl

, Van De Walle

. Energetics and electronic structure of stacking faults in AlN, GaN, and InN. Phys. Rev. B. 1998;57(24):52–55. doi: 10.1103/PhysRevB.57.15251

54.

Trampert

, Brandt

, Yang

, Direct observation of the initial nucleation and epitaxial growth of metastable cubic GaN on (001) GaAs. Appl Phys Lett. 1997;70(5):583–585. doi: 10.1063/1.118281

55.

Hull

, Bacon

. Introduction to dislocations. 5th ed. Oxford: Butterworth-Heinemann; 2011. Chapter 5, Dislocations in face-centered cubic metals; p. 85–107.

56.

Parinyataramas

, Sanorpim

, Thanachayanont

, TEM investigation of anisotropic defect structure in cubic GaN/AlGaAs/GaAs(001) grown by MOVPE. Phys Status Solidi C. 2011;8(7–8):2255–2257. doi: 10.1002/pssc.201001170

57.

Kemper

, Schupp

, Häberlen

, Anti-phase domains in cubic GaN. J Appl Phys. 2011;110:123512. doi: 10.1063/1.3666050

58.

Pirouz

, Chorey

, Powell

. Antiphase boundaries in epitaxially grown

-SiC. Appl Phys Lett. 1987;50(4):221–223. doi: 10.1063/1.97667

59.

Song

, Chang

, Oh

, Optimization of ZnSe growth on miscut GaAs substrates by molecular beam epitaxy. J Cryst Growth. 2003;249(1–2):128–143. doi: 10.1016/S0022-0248(02)02129-2

60.

Huang

, Balakrishnan

, Khoshakhlagh

, Simultaneous interfacial misfit array formation and antiphase domain suppression on miscut silicon substrate. Appl Phys Lett. 2008;93(7):10–13.

61.

Vilchis

, Sánchez-R

. Electrical behavior of Mg doped cubic GaN on c-GaN structure. Mater Sci Semicond Process. 2015;37:68–72. doi: 10.1016/j.mssp.2015.01.052

62.

Martinez-Guerrero

, Daudin

, Feuillet

, p and n type doping of cubic GaN on SiC (001). Mater Sci Eng B. 2001;82:59–61. doi: 10.1016/S0921-5107(00)00719-4

63.

. Electrical and optical properties of Mg doped MBE grown GaN epilayers. Phys Status Solidi B. 1998;210(2):445–449. doi: 10.1002/(SICI)1521-3951(199812)210:2<445::AID-PSSB445>3.0.CO;2-P

64.

Lin

, Xue

, Zhou

, P-type zinc-blende GaN on GaAs substrates. Appl Phys Lett. 1993;63(7):932–933. doi: 10.1063/1.109848

65.

Powell

REL

, Novikov

, Foxon

, Photoluminescence of magnesium and silicon doped cubic GaN. Phys Status Solidi C. 2014;11(3–4):385–388. doi: 10.1002/pssc.201300468

66.

, Yang

, Zhao

, Room-temperature optical transitions in Mg-doped cubic GaN/GaAs(100) grown by metalorganic chemical vapor deposition. J Appl Phys. 2000;87(4):2064–2066. doi: 10.1063/1.372139

67.

Schmidt

, Lischka

. Excitation-power dependence of the near-band-edge photoluminescence of semiconductors. Phys Rev B. 1992;45(16):8989–8994. doi: 10.1103/PhysRevB.45.8989

68.

Guarneros

, Sánchez

. Magnesium doped GaN grown by MOCVD. Mater Sci Eng B. 2010;174(1–3):263–265. doi: 10.1016/j.mseb.2010.03.022

69.

, Richter

, Busch

, Electroluminescence of a cubic GaN/GaAs (001) p–n junction. Appl Phys Lett. 2000;76(1):13. doi: 10.1063/1.125640

70.

, Köhler

, Lübbers

, p-type doping of cubic GaN by carbon. Phys Status Solidi Appl Res. 2001;188(2):699–703. doi: 10.1002/1521-396X(200112)188:2<699::AID-PSSA699>3.0.CO;2-8

71.

Zado

, Tschumak

, Gerlach

, ‘Doping of MBE grown cubic GaN on 3C-SiC (001) by CBr4’, 2010 Wide Bandgap Cubic Semiconductors: From Growth to Devices, Strasbourg, France, October 2010, 1292, (1), 181–184.

72.

Zado

, Tschumak

, Gerlach

, Carbon as an acceptor in cubic GaN/3C–SiC. J Cryst Growth. 2011;323(1):88–90. doi: 10.1016/j.jcrysgro.2010.12.044

73.

, Yang

, Li

, Effect of Si doping on cubic GaN films grown on GaAs(100). J Cryst Growth. 1999;206(1–2):150–154. doi: 10.1016/S0022-0248(99)00365-6

74.

, Chen

, Liu

, Influence of Si doping on optical characteristics of cubic GaN grown on (001) GaAs substrates. Appl Phys Lett. 2000;76(25):3765. doi: 10.1063/1.126774

75.

, Lischka

. Heteroepitaxy of doped and undoped cubic group III-nitrides. Phys Status Solidi Appl Res. 1999;176(1):475–485. doi: 10.1002/(SICI)1521-396X(199911)176:1<475::AID-PSSA475>3.0.CO;2-6

76.

, Chen

, Liu

, Photoluminescence study of Si doping cubic GaN grown on (001) GaAs substrates by molecular beam epitaxy. J Cryst Growth. 2001;227–228:420–424. doi: 10.1016/S0022-0248(01)00736-9

77.

, Schörmann

, As

, Room temperature green light emission from nonpolar cubic InGaN/GaN multi-quantum-wells. Appl Phys Lett. 2007;90(7):71903. doi: 10.1063/1.2475564

78.

Silveira

, Tabata

, Leite JR , Evidence of phase separation in cubic In x Ga 1 − x N epitaxial layers by resonant Raman scattering evidence of phase separation in cubic InxGa1-xN epitaxial layers by resonant Raman scattering. 1999;75(23):3602–3604.

79.

Tabata

, Leite JR , Lima

, Raman phonon modes of zinc blende In x Ga 1 − x N alloy epitaxial layers Raman phonon modes of zinc blende InxGa1-xN alloy epitaxial layers. Appl Phys Lett. 1999;75(8):1095–1097. doi: 10.1063/1.124608

80.

Lemos

. Evidence for phase-separated quantum dots in cubic InGaN layers from resonant Raman scattering. Phys Rev Lett. 2000;84(16):3666–3669. doi: 10.1103/PhysRevLett.84.3666

81.

Tabata

, Teles

, Scolfaro

LMR

, Phase separation suppression in InGaN epitaxial layers due to biaxial strain. Appl Phys Lett. 2002;80(5):769–771. doi: 10.1063/1.1436270

82.

Pacheco-Salazar

, Leite JR , Cerdeira

, Photoluminescence measurements on cubic InGaN layers deposited on a SiC substrate. Semicond Sci Technol. 2006;21(7):846–851. doi: 10.1088/0268-1242/21/7/003

83.

Husberg

, Khartchenko

, Vogelsang

, Photoluminescence associated with quantum dots in cubic GaN/InGaN/GaN double heterostructures. Phys E. 2002;13:1090–1093. doi: 10.1016/S1386-9477(02)00310-7

84.

Teles

, Ferreira

, Leite JR , Strain-induced ordering in InxGa1-xN alloys. Appl Phys Lett. 2003;82(24):4274–4276. doi: 10.1063/1.1583854

85.

Schubert

. Light-emitting diodes. 2nd ed. New York: Cambridge University Press; 2006.

86.

Okumura

, Hamaguchi

, Koizumi

, Growth of cubic III-nitrides by gas source MBE using atomic nitrogen plasma: GaN, AlGaN and AlN. J Cryst Growth. 1998;189:390–394. doi: 10.1016/S0022-0248(98)00321-2

87.

Nakadaira

, Tanaka

. Metalorganic vapor-phase epitaxial growth and characterization of cubic AlxGa1-xN alloy. Jpn J Appl Phys. 1998;37:1449–1452. doi: 10.1143/JJAP.37.1449

88.

Zainal

, Novikov

, Mellor

, Current-voltage characteristics of zinc-blende (cubic) Al 0.3 Ga 0.7 N/GaN double barrier resonant tunneling diodes. Appl Phys Lett. 2010;97(11):112102. doi: 10.1063/1.3488819

89.

, Schörmann

, Pawlis

, Cubic InGaN/GaN multi-quantum wells and AlGaN/GaN distributed Bragg reflectors for application in resonant cavity LEDs. Microelectronics J. 2005;36(11):963–968. doi: 10.1016/j.mejo.2005.04.003

90.

Nakadaira

, Tanaka

. Optically pumped stimulated emission from cubic GaN/AlGaN double heterostructure grown on GaAs(100) using metalorganic vapor-phase epitaxy. J Cryst Growth. 1998;189:411–414. doi: 10.1016/S0022-0248(98)00326-1

91.

Bai

, Wang

, Sakai

. Influence of the quantum-well thickness on the radiative recombination of InGaN/GaN quantum well. J Appl Phys. 2011;88(8):4729–4733. doi: 10.1063/1.1311831

92.

Taniyasu

, Suzuki

, Lim

, Cubic InGaN/GaN double-heterostructure light emitting diodes grown on GaAs (001) substrates by MOVPE. Phys Status Solidi Appl Res. 2000;180:241–246. doi: 10.1002/1521-396X(200007)180:1<241::AID-PSSA241>3.0.CO;2-A

93.

Stark

CJM

, Detchprohm

, Wetzel

. Cubic GaInN/GaN multi-quantum wells for increased smart lighting system efficiency. Conference on Lasers and Electro-Optics 2012, OSA Technical Digest (online), Paper JTh4J.3, Optical Society of America; 2012.

94.

Matioli

, Weisbuch

. Direct measurement of internal quantum efficiency in light emitting diodes under electrical injection. J Appl Phys. 2011;109:73114. doi: 10.1063/1.3549730