PSS templates

of Sciences, Guangzhou 510650, China. ‡School of Physics & Electronic Engineering, Guangzhou University, Guangzhou. 510006, China. KEYWORDS: GaN ...
1 downloads 0 Views 6MB Size
Research Article Cite This: ACS Appl. Mater. Interfaces 2017, 9, 43386−43392

www.acsami.org

High-Quality GaN Epilayers Achieved by Facet-Controlled Epitaxial Lateral Overgrowth on Sputtered AlN/PSS Templates Chenguang He,*,† Wei Zhao,† Kang Zhang,† Longfei He,† Hualong Wu,† Ningyang Liu,† Shan Zhang,‡ Xiaoyan Liu,† and Zhitao Chen*,† †

Guangdong Institute of Semiconductor Industrial Technology, Guangdong Academy of Sciences, Guangzhou 510650, China School of Physics & Electronic Engineering, Guangzhou University, Guangzhou 510006, China



S Supporting Information *

ABSTRACT: It is widely believed that the lack of high-quality GaN wafers severely hinders the progress in GaN-based devices, especially for defect-sensitive devices. Here, low-cost AlN buffer layers were sputtered on cone-shaped patterned sapphire substrates (PSSs) to obtain high-quality GaN epilayers. Without any mask or regrowth, facet-controlled epitaxial lateral overgrowth was realized by metal−organic chemical vapor deposition. The uniform coating of the sputtered AlN buffer layer and the optimized multiple modulation guaranteed high growth selectivity and uniformity of the GaN epilayer. As a result, an extremely smooth surface was achieved with an average roughness of 0.17 nm over 3 × 3 μm2. It was found that the sputtered AlN buffer layer could significantly suppress dislocations on the cones. Moreover, the optimized three-dimensional growth process could effectively promote dislocation bending. Therefore, the threading dislocation density (TDD) of the GaN epilayer was reduced to 4.6 × 107 cm−2, which is about an order of magnitude lower than the case of two-step GaN on the PSS. In addition, contamination and crack in the light-emitting diode fabricated on the obtained GaN were also effectively suppressed by using the sputtered AlN buffer layer. All of these advantages led to a high output power of 116 mW at 500 mA with an emission wavelength of 375 nm. This simple, yet effective growth technique is believed to have great application prospects in highperformance TDD-sensitive optoelectronic and electronic devices. KEYWORDS: GaN, facet-controlled, epitaxial lateral overgrowth, sputtered AlN, patterned sapphire substrate, MOCVD



higher than 1 × 108 cm−2,13,14 and the degradation of electronic devices cannot be eliminated until the TDD is lower than 1 × 108 cm−2.15−17 Therefore, it is urgent and challenging to achieve high-quality GaN with a TDD below 108 cm−2, which is a prerequisite for high-performance devices. So far, various growth techniques have been utilized to reduce the TDD of GaN epilayers. Among these techniques, the epitaxial lateral overgrowth (ELOG) and its derivatives, such as facet-controlled epitaxial lateral overgrowth (FACELO) and pendeo-epitaxy, are extremely effective in suppressing dislocations.18,19 By using a micro/nanoscale SiNx/SiO2patterned mask in various shapes, the TDD of GaN epilayers can be reduced to 105−107 cm−2.20,21 However, the traditional ELOG technique involves the multiple-step masking process

INTRODUCTION

In the past decades, GaN has attracted considerable attention for its extensive application potentials in optoelectronic and electronic devices.1−6 However, because of the limited availability and high cost of native substrates, GaN-based devices are generally manufactured on foreign substrates such as sapphire (Al2O3), silicon (Si), and silicon carbide (SiC).7 The large mismatch between GaN epilayers and foreign substrates typically results in a high threading dislocation density (TDD) of 108 to 109 cm−2, despite the use of lowtemperature AlN and GaN buffers.8−10 Fortunately, indiumcontaining AlxInyGa1−x−yN materials exhibit a defect-insensitive emission probability, which promotes the prosperity of blue and green light-emitting diodes (LEDs) and laser diodes.11,12 However, such a high TDD fails to meet the demands of defect-sensitive devices with weak indium-related carrier localization. It has been reported that the efficiency of ultraviolet LEDs has a drastic deterioration when the TDD is © 2017 American Chemical Society

Received: September 28, 2017 Accepted: November 22, 2017 Published: November 22, 2017 43386

DOI: 10.1021/acsami.7b14801 ACS Appl. Mater. Interfaces 2017, 9, 43386−43392

Research Article

ACS Applied Materials & Interfaces and regrowth. There also exists a high probability of contamination. Hybrid substrate technique, such as epitaxy on sapphire coated with the carbon nanotube, graphene, or boron nitride nanosheet, is another alternative way to improve the crystal quality of GaN. Adopting this technique can avoid regrowth and achieve a TDD of 107−108 cm−2.22,23 Nevertheless, the uniformity and repeatability of hybrid substrates still need further validation for commercial applications. The patterned sapphire substrate (PSS) technique, which is free from the masking process and regrowth, is the most common approach in commercial applications because of its enormous advantages in maturity and cost. Generally, the epitaxy of GaN on PSSs employs in situ GaN buffer layers, obtaining a TDD of 2 × 108 to 6 × 108 cm−2.24 Recently, ex situ AlN buffer layers on PSSs were proposed to further improve the electrical properties of GaN-based LEDs;25,26 but the TDD of GaN epilayers is still far from satisfactory in comparison with the ELOG and hybrid substrate techniques. In this work, an approach to obtain high-quality GaN epilayers was proposed by realizing mask-free FACELO on sputtered AlN/PSS templates. Because of the high growth selectivity and uniformity in the whole epitaxy process, the GaN epilayer achieved an extremely smooth surface with an average roughness of 0.17 nm over 3 × 3 μm2. Meanwhile, by suppressing dislocation generation and promoting dislocation bending, the TDD was reduced to 4.6 × 107 cm−2, which is about an order of magnitude lower than the case of two-step GaN on the PSS. The detailed evolution mechanism of the surface morphology and crystal quality was revealed using a well-developed growth model. Finally, the contamination and crack in the LED fabricated on the high-quality GaN were proved to be effectively suppressed. All of these advantages led to a strong electroluminescence (EL) at 375 nm. This simple, yet effective growth technique is believed to have great application prospects in high-performance TDD-sensitive optoelectronic and electronic devices.



Table 1. Characterizations of Samples A−F number sample sample sample sample sample sample

A B C D E F

3D time (s)

3D temperature (°C)

recovery temperature (°C)

FWHMs of (002)/(102) reflections (arcsec)

1500 1500 1500 210 1037 2000

975 990 975 975 975 975

1035 1015 1015 1015 1015 1015

150/127 234/205 140/119 223/205 216/191 149/133

interrupted time were labeled as C0 (0 s, the beginning of growth), C1 (210 s), C2 (980 s), C3 (1812 s), C4 (2750 s), and C5 (7175 s, the end of growth). Finally, a near ultraviolet (NUV) LED was grown to manifest the quality of the as-obtained GaN. The structure consists of an undoped GaN layer (the same growth conditions with sample C), 40 periods of GaN (2 nm)/Al0.1Ga0.9N (2 nm) superlattices, a 2 μmthick n-Al0.07Ga0.93N layer, 5 periods of In0.01Ga0.99N (5 nm)/ Al0.12Ga0.88N (10 nm) multiple quantum wells (MQWs), a 20 nmthick p-Al0.25Ga0.75N electron blocking layer, a 100 nm-thick p-GaN, and a 20 nm-thick p++-GaN. A conductive indium tin oxide layer was used as a p-type Ohmic contact layer. Cr/Au metallization was deposited as p-type and n-type electrodes, respectively. The size of the NUV LED chip on the wafer was 1143 × 1143 μm2. The surface morphology was characterized by a FINIAL FJ-3A optical microscope (OM), a Bruker Dimension Edge atomic force microscope (AFM), and a FEI Quanta 650 scanning electron microscope (SEM) with an energy dispersive spectrometer (EDS). The crystal quality of the samples was evaluated by a Rigaku SmartLab 9 kW high-resolution X-ray diffraction (HRXRD) system ω-scan rocking curve (RC), cathodoluminescence (CL), photoluminescence (PL), a FEI Tecnai Osiris TF-20 scanning tunnel electron microscope (STEM), and a transmission electron microscope (TEM). All measurements of RCs were based on a typical double-crystal X-ray diffraction mode. For the incident optics unit, a Ge(220) 2-bounce monochromator was adopted. As for the receiving optics unit, we did not add any analysis crystal except for two inherent slits. The width of these two receiving slits for measurement were both 1.000 mm, which was large enough in comparison with the line broadening of the detected X-ray. The length of the receiving slits can be considered as 8 mm. A closed cycle helium cryosystem provided the low temperature (6 K) for the PL measurement. A 325 nm He−Cd laser was used for the PL excitation source. The TEM/STEM-ready samples were prepared using the in situ focused ion beam (FIB) lift out technique on an FEI DualBeam FIB/SEM. The elemental analysis was conducted by secondary ion mass spectroscopy (SIMS). The evolution of the stress and surface morphology were monitored by in situ measurements of the wafer curvature (405 nm) and optical reflectivity (633 nm).

EXPERIMENTAL SECTION

The 2 in. c-plane (0001) cone-shaped PSSs were prepared by photolithography and inductively coupled plasma etching techniques. The bottom diameter, interval spacing, and height of the pattern were 2.55, 0.45, and 1.6 μm, respectively. After the patterning process, a NAURA iTops A230 AlN sputter system was used to deposit 30 nmthick AlN buffer layers at 650 °C by feeding 120 sccm N2, 30 sccm He, and 1 sccm O2. A 2 in. aluminum disk (99.999%) was used as the sputtering target. Then, the epitaxy of GaN was carried out in an Aixtron low pressure metal−organic chemical vapor deposition (MOCVD) system, with a vertical close-coupled showerhead reactor. Trimethylgallium (TMGa) and ammonia (NH3) were used as precursors for Ga and N, respectively. High-purity hydrogen (H2) was employed as the carrier gas. For samples A−F, the AlN/PSS templates were first heated to the three-dimensional (3D) temperature under an ammonia atmosphere. After that, 3D GaN layers were grown on the AlN/PSS templates. On top of the 3D GaN layers, GaN recovery layers were introduced to promote lateral coalescence. The recovery layers were terminated once the oscillation amplitude of the reflectance signals saturated or reached their maximum. Finally, the samples were finished with 3.2 μm-thick two-dimensional (2D) GaN layers deposited at 1045 °C. The growth pressures of the 3D layer, recovery layer, and 2D layer were 600, 300, and 150 mbar, respectively, and the V/III mole ratios of the 3D layer, recovery layer, and 2D layer were 1316, 1663, and 1221, respectively. The detailed characterization of samples A−F is listed in Table 1. In addition, for sample C, a series of growth interruptions were introduced to investigate the evolution of the surface morphology and crystal quality. The stages with different



RESULTS AND DISCUSSION Figure 1a−f shows the OM images of samples A−F, respectively. It was found that the surfaces of the samples B, C, and F were smooth and clean. In comparison, those of samples A, D, and E were full of hexagonal pyramidal hillocks. It has been reported that the formation of hillocks is related to the growth instability started on the coalescence boundaries.27 In this work, a relatively high recovery temperature (sample A) and a short 3D growth time (samples D and E) both accelerated the lateral coalescence. Consequently, surface undulation happened around the coalescence boundaries, leading to the formation of hillocks. This indicates that the 3D and recovery process both had influence on the surface morphology. Table 1 summarizes the full width at halfmaximums (FWHMs) of the (002)/(102) reflections for samples A−F. The FWHMs of the (002)/(102) reflections 43387

DOI: 10.1021/acsami.7b14801 ACS Appl. Mater. Interfaces 2017, 9, 43386−43392

Research Article

ACS Applied Materials & Interfaces

Figure 1. (a−f) OM images of samples with different 3D and recovery growth conditions with a magnification of 100. Many hexagonal pyramidal hillocks appeared on the surface of samples A, D, and E. In comparison, smooth and clean surfaces were achieved for samples B, C, and F.

Figure 2. (a−f) Plan-view SEM images from stage C0 to C5 with a magnification of 20 000. The inset in (f) shows the AFM image at stage C5. Its RMS value over 3 × 3 μm2 was 0.17 nm. (g) Evolution of the growth fronts from stage C2 to C3. The inclined facets evolved from (6 × {11̅01} + 6 × {1122̅ }) to 6 × {1122̅ }. (h) Evolutionary trend of the surface morphology from stage C3 to C4. The inclined facets gradually evolved from 6 × {112̅2} to 6 × {11̅01}. The outer contour of the aqua hexagon had a 30° rotation relative to the wine hexagon. (i−k) EDS mappings for Al (blue) + Ga (red) elements of stage C1−C3, respectively. All of these plan-view pictures (a−k) were taken in the same direction.

was interrupted at different stages. The entire morphology evolution process is presented in Figure 2a−f by SEM planview imaging. Before epitaxy (C0), the pattern of the AlN/PSS template was uniform in size and well-arranged. In the 3D process, the valley region was quickly covered up by GaN (C1). Then, the (0001) facet (c-plane) shrank drastically (even disappeared in certain areas), and the inclined facets formed on the sidewalls (C2).18 The inclined facets were composed of 6 × {11̅01} and 6 × {112̅2} planes, as illustrated in Figure 2g. The evolution of the facet structure is attributed to a faster vertical ([0001] direction) growth rate than the lateral ones, which is a typical 3D behavior for GaN.18,28 In the recovery process, the (0001) facet extended rapidly because of the faster lateral growth rates.18,28 The inclined facets first (C3) evolved from ({11̅01} + {112̅2}) to {112̅2} and then (C4) evolved from {112̅2} to {11̅01}. The evolution from ({11̅01} + {112̅2}) to

for samples A, B, and C were 150/127, 234/205, and 140/119 arcsec, respectively. Apparently, the best crystal quality was obtained by sample C. By contrast, the FWHMs of sample A (higher recovery temperature) demonstrated a slight recession, whereas those of sample B (higher 3D temperature) declined obviously. This implies that the crystal quality is more sensitive to the 3D temperature than the recovery temperature. Furthermore, from the comparison among samples C−F, it was found that narrower FWHMs also required an adequate 3D process longer than 1500 s. Therefore, the 3D process was thought to play a critical role in the control of crystal quality. Overall, through an optimized multiple modulation (3D → recovery → 2D) technique, sample C achieved the best surface morphology and crystal quality at the same time. To further investigate the evolution mechanism of the surface morphology and crystal quality, the growth of sample C 43388

DOI: 10.1021/acsami.7b14801 ACS Appl. Mater. Interfaces 2017, 9, 43386−43392

Research Article

ACS Applied Materials & Interfaces

Figure 3. (a) HRXRD FWHMs of symmetric (002) and asymmetric (102) reflections at different stages. The FWHMs of (002) and (102) reflections at stage C5 were 140 and 119 arcsec, respectively. (b) Plan-view CL mapping for GaN at stage C5. The dark spot density was 4.6 × 107 cm−2. (c) Low-temperature (6 K) PL spectrum of sample C. Peak positions were moved to the high-energy side by compressive stress in the GaN epilayer. (d−f) Cross-sectional BF STEM images, which exhibit predominantly strain contrast and thus visualize screw and edge dislocations at the same time.

{1122̅ } facets is due to an in-plane growth rate anisotropy.29 It is well-established that the closely packed {112̅2} facets have a higher density of dangling bonds at the step edge, which means a higher incorporation probability of Ga. In other words, the fronts of {112̅2} facets had a faster growth rate toward the centers of the cones. As a result, the growth fronts of {11̅01} facets first diminished (green dashed line) and then disappeared (wine dashed line), with only {1122̅ } facets left, as shown in Figure 2g. Meanwhile, the outline of the inclined facets varied from dodecagon (blue full line) to hexagon (wine dashed line). The evolution from {112̅2} to {11̅01} facets was determined by the little GaN islands on the cones, whose outer contour (aqua hexagon) had a 30° rotation relative to the wine hexagon, as shown in Figure 2h. In the 2D process, the inverted pyramids were healed, and an extremely smooth surface was obtained with a root mean square (RMS) of 0.17 nm over 3 × 3 μm2 (C5), as shown in Figure 2f. It is noted that GaN epilayers showed high growth selectivity and uniformity in the whole epitaxy process, which was in favor of the smooth surface morphology. The high growth selectivity and uniformity can be more clearly reflected by EDS mappings in Figure 2i−k. GaN was first grown on the valleys and then advanced toward the centers of the cones uniformly from the outside, producing a close growth rate and crystal orientation for GaN from the adjacent regions. This phenomenon is quite different from the epitaxy of two-step GaN grown on micro/ nanoscale PSS, in which case GaN will be grown everywhere at random.30,31 It is speculated that the sputtered AlN buffer layer could modify the surface energy. Thus, the growth on the valleys was enhanced, and the growth on the cones was suppressed. To figure out the evolution process of crystal quality, HRXRD, CL, PL, and STEM measurements were carried out. Figure 3a shows the HRXRD FWHMs of (002) and (102) reflections at different stages. For the sputtered AlN buffer layer, the FWHMs of (002) and (102) reflections were 2066 and unmeasurable, respectively. According to our experience,

the unmeasurable FWHM was at least larger than 4000 arcsec. The poor crystal quality of the AlN buffer layer resulted in large FWHMs of GaN epilayers in the initial stage of growth. The FWHMs of (002) and (102) reflections then underwent a drastic decrease to ∼500 arcsec in the 3D process, whereas they demonstrated a relatively slow decrease in the recovery and 2D process. It supports our opinion that the 3D process played a critical role in the control of crystal quality. The final FWHMs for (002) and (102) reflections were 140 and 119 arcsec, respectively. Accordingly, the TDD in the GaN epilayers could be roughly estimated to be 8.5 × 107 cm−2 by using the following equation

N random =

β2 4.35 × b2

(1)

where β is the FWHM of tilt/twist and b is the length of the dislocation Burgers vector for screw/threading dislocation.32 However, this empirical model will significantly overestimate the TDD by a factor of 2−5, when the dislocations are not randomly distributed but localized at grain boundaries.32 So, plan-view CL mapping was conducted for a more precise estimation, as shown in Figure 3b. It was found the dislocations indeed exhibited a localized distribution. The TDD was estimated to be 4.6 × 107 cm−2 based on the dark spot density. This value is about an order of magnitude lower than the average level of two-step GaN grown on the PSS. The excellent crystal quality was also confirmed by the lowtemperature (6 K) PL measurement, as shown in Figure 3c. The PL spectrum was dominated by the donor bound exciton (DX) and exciton A (XA) line at 3.500 and 3.506 eV, respectively. The linewidth of DX and XA peaks were 2.6 and 3.8 meV, indicating a very good crystal quality. In addition, acceptor bound exciton (AX), exciton B (XB), and two longitudinal-optical phonon replicas could be clearly observed.33,34 So many fine structures also demonstrated the excellent crystal quality. It is noted that the peak positions of DX and XA in this study moved to the high-energy side 43389

DOI: 10.1021/acsami.7b14801 ACS Appl. Mater. Interfaces 2017, 9, 43386−43392

Research Article

ACS Applied Materials & Interfaces

Figure 4. (a−f) Schematic of the entire growth process viewed in a cross-section along the ⟨112̅ 0⟩ direction.

Figure 5. (a) Wafer structure of the NUV-LED. (b−d) Cross-sectional TEM images of NUV-LED. (e) SIMS profiles for H, C, and O elements. The concentrations of C, H, and O elements in the n-Al0.07Ga0.93N epilayer are 3 × 1016, 1.5 × 1017, 1 × 1016 cm−3, respectively. (f) The in situ reflectance (at 633 nm) and curvature (at 405 nm) transients vs time during the growth of GaN. The GaN epilayer bore a compressive stress because of the smaller in-plane lattice constant of the sputtered AlN buffer layer. The tensile stress in the subsequent n-AlGaN was significantly decreased. (g) EL spectrum of the NUV-LED. The inset shows the output power as a function of the forward current. A high output power of 116 mW @ 500 mA was achieved with an emission wavelength of 375 nm.

first propagated vertically, but then bended by 90° to adopt a horizontal direction (circled in green).28 When the thickness reached 700 nm (in the range of 3D layer), the dislocation bending phenomenon was nearly complete. The preexisting dislocations were reduced to a very low level. That is the reason why the 3D process played a critical role in the control of crystal quality. By suppressing dislocation generation and promoting dislocation bending, the final crystal quality was significantly improved in comparison with two-step GaN on the PSS. Up to this point, we have realized controlling of the facet structure, “blocking” of dislocations, and promotion of dislocation bending. These behaviors are the typical characteristics of FACELO, indicating we had realized mask-free FACELO through an optimized multiple modulation technique. To explain the evolution process of the morphology and dislocations more vividly, a simplified schematic of growth behavior was constructed, as shown in Figure 4a−f. Before epitaxy (C0), the ex situ AlN buffer layer provided a uniform coating on the PSS. In the initial stage of 3D growth (C1), influenced by AlN-modified surface energy, GaN was almost entirely grown on the valleys with dislocations generated on the

compared with other reports.35,36 Moreover, the linewidth in this study was a little larger.35 They were attributed to the compressive stress in GaN epilayers, which will be discussed later. Figure 3d−f shows the cross-sectional bright-field (BF) STEM images, which exhibit predominantly strain contrast and thus visualize screw and edge dislocations at the same time. It can be seen that the GaN epilayer grown on the cones was nearly free from dislocations, whereas plenty of dislocations were generated on the valleys in the initial stage of growth. This phenomenon is attributed to the high growth selectivity. As mentioned above, direct growth of GaN on the cones was very difficult. Before coalescence was completed, the growth process was dominated by the 3D growth on the valleys and lateral overgrowth toward the cones. In this case, dislocations would not be generated on the cones until the end of the recovery process.18,21 Therefore, the cone region played a role of “mask” here, like SiNx/SiO2 in the traditional ELOG, and GaN epilayers on the cone region corresponded to the “wing.” Considering that the valley region only accounted for 34.5% of the total area in the plan-view, the generation of dislocation was significantly suppressed. The existing dislocations on the valleys 43390

DOI: 10.1021/acsami.7b14801 ACS Appl. Mater. Interfaces 2017, 9, 43386−43392

Research Article

ACS Applied Materials & Interfaces

morphology. The generation of dislocations on the cone region was greatly suppressed by the sputtered AlN buffer layer. Moreover, the existing dislocations on the valley region could be effectively eliminated by a 90° bending through optimizing the 3D process. Under their combined action, the TDD could be reduced to 4.6 × 107 cm−2, which is about an order of magnitude lower than the case of two-step GaN on PSS. Besides, this technique also exhibited great advantages in contamination and stress. All of these contributed to a high output power of 116 mW at 500 mA with an emission wavelength of 375 nm. Combined with the superiority in maturity and cost, this technique would have great application potential for high-performance TDD-sensitive optoelectronic and electronic devices.

interface. With the development of the 3D process (C2), a faster vertical growth rate led to the gradual disappearance of the c-plane and the expansion of inclined facets. As a result, the cross-sectional shape of the GaN epilayer transmitted from a trapezoid to a triangle. Accompanied by the evolution of morphology, the vertical dislocations had a 90° bending when they got close to the edge of the lateral facets and the top (0001) facet. This process would not stop until the crosssectional shape became a triangle. Because dislocations were mainly eliminated by this means, an inadequate 3D time or a high 3D temperature would severely discount the effect and aggravate the crystal quality. When the epitaxy came into the recovery process (C3 and C4), a faster lateral growth rate resulted in the expansion of the c-plane, and the cross-sectional shapes of the GaN epilayer transmitted from a triangle to a trapezoid again. During this process, reasonable growth conditions would produce a close growth rate and crystal orientation for GaN from the adjacent regions, suppressing the generation of hillocks and dislocations on the boundary of coalescence. In the 2D process, the surface morphology and TDD were further improved. Through the multiple modulation of growth modes, FACELO was successfully realized without any mask or regrowth. Finally, a NUV-LED was manufactured to manifest the quality of the as-obtained GaN. Figure 5a−d shows the schematic diagram and cross-sectional TEM images of the NUV-LED wafer. Few threading dislocations can be observed in the MQW region. Figure 5e shows the SIMS profiles of carbon (C), hydrogen (H), and oxygen (O) elements. The concentrations of C, H, and O elements in the n-Al0.07Ga0.93N epilayer were only 3 × 1016, 1.5 × 1017, 1 × 1016 cm−3, respectively.37 Such low impurity concentrations indicate that the contamination can be well-controlled by this growth technique. In addition, as shown in Figure 5f, GaN on AlN/PSS demonstrated obvious compressive stress because of the smaller in-plane lattice constant of the sputtered AlN buffer layer. This phenomenon is quite different from the case of the in situ GaN buffer layer because the GaN epilayer grown on the MOCVD GaN buffer layer usually bears tensile stress originated from the coalescence process of 3D GaN islands.38 Obviously, the compressive stress in the GaN epilayer was beneficial for the elimination of cracks in the subsequent AlGaN.39 Finally, the chip on the wafer achieved an output power of 116 mW @ 500 mA with an emission wavelength of 375 nm, as shown in Figure 5g. In comparison, Feng et al. reported an excellent output power of 110 mW (nearly saturated) @ 350 mA.22 However, the emission wavelength of their work was 410 nm. It is known that the output power usually experiences a steep reduction when the emission wavelength decreases.40 Therefore, for a 375 nm LED with a weak indium-related carrier localization, our EL performance is good, considering it is only a “chip-on-wafer” result. Further improvement can be expected by introducing laser lift-off, surface roughening, reflecting electrode, and encapsulation.



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsami.7b14801. X-ray rocking curve scan for sample C, illustration of different facets, and in situ curvature monitoring of GaN epilayers grown on different buffers (PDF)



AUTHOR INFORMATION

Corresponding Authors

*E-mail: [email protected] (C.H.). *E-mail: [email protected] (Z.C.). ORCID

Chenguang He: 0000-0002-3683-9401 Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work is supported by the National Natural Science Foundation of China (grant nos. 11304048 and 61604045), Innovation-driven Development Capacity Construction Project of the Guangdong Academy of Sciences (grant no. 2017GDASCX-0845), the Science and Technology Program of Guangdong (grant nos. 2014TQ01C707, 2015B010132004, 2015B010134001, and 2016B070701023), and Pearl River S&T Nova Program of Guangzhou (grant no. 201610010142). The authors are also grateful to Dr. Dong Boyu and Guo Bingliang from Beijing NAURA Microelectronics Equipment Co., Ltd. for providing sputtered AlN/sapphire templates.



REFERENCES

(1) Ponce, F. A.; Bour, D. P. Nitride-based semiconductors for blue and green light-emitting devices. Nature 1997, 386, 351−359. (2) Fasol, G. Room-temperature blue gallium nitride laser diode. Science 1996, 272, 1751−1752. (3) Mishra, U. K.; Parikh, P.; Wu, Y.-F. AlGaN/GaN HEMTs-an overview of device operation and applications. Proc. IEEE 2002, 90, 1022−1031. (4) Lupan, O.; Pauporté, T.; Viana, B.; Tiginyanu, I. M.; Ursaki, V. V.; Cortès, R. Epitaxial electrodeposition of ZnO nanowire arrays on p-GaN for efficient UV-light-emitting diode fabrication. ACS Appl. Mater. Interfaces 2010, 2, 2083−2090. (5) Pauporté, T.; Lupan, O.; Zhang, J.; Tugsuz, T.; Ciofini, I.; Labat, F.; Viana, B. Low-temperature preparation of Ag-doped ZnO nanowire arrays, DFT study, and application to light-emitting diode. ACS Appl. Mater. Interfaces 2015, 7, 11871−11880. (6) Zhang, S. G.; Zhang, X. W.; Yin, Z. G.; Wang, J. X.; Si, F. T.; Gao, H. L.; Dong, J. J.; Liu, X. Optimization of electroluminescence from n-



CONCLUSIONS In summary, we demonstrated that the quality of GaN epilayers could be significantly improved by using sputtered AlN/PSS templates. Through optimizing the growth modes, FACELO was realized without any mask or regrowth. The uniform coating of the sputtered AlN buffer layer and the optimized multiple modulation guaranteed high growth selectivity and uniformity of the GaN epilayer, leading to a smooth surface 43391

DOI: 10.1021/acsami.7b14801 ACS Appl. Mater. Interfaces 2017, 9, 43386−43392

Research Article

ACS Applied Materials & Interfaces

on cone-shaped patterned sapphire substrates. J. Appl. Phys. 2010, 107, 103506. (25) Yen, C.-H.; Lai, W.-C.; Yang, Y.-Y.; Wang, C.-K.; Ko, T.-K.; Hon, S.-J.; Chang, S.-J. GaN-based light-emitting diode with sputtered AlN nucleation layer. IEEE Photonics Technol. Lett. 2012, 24, 294−296. (26) Chang, L.-C.; Chen, Y.-A.; Kuo, C.-H. Spatial correlation between efficiency and crystal structure in GaN-based light-emitting diodes prepared on high-aspect ratio patterned sapphire substrate with sputtered AlN nucleation layer. IEEE Trans. Electron Devices 2014, 61, 2443−2447. (27) Liu, C.; Shields, P. A.; Denchitcharoen, S.; Stepanov, S.; Gott, A.; Wang, W. N. Pulsed epitaxial lateral overgrowth of GaN by metalorganic vapour phase epitaxy. J. Cryst. Growth 2007, 300, 104− 109. (28) Vennéguès, P.; Beaumont, B.; Bousquet, V.; Vaille, M.; Gibart, P. Reduction mechanisms for defect densities in GaN using one- or two-step epitaxial lateral overgrowth methods. J. Appl. Phys. 2000, 87, 4175−4181. (29) Kapolnek, D.; Keller, S.; Vetury, R.; Underwood, R. D.; Kozodoy, P.; Den Baars, S. P.; Mishra, U. K. Anisotropic epitaxial lateral growth in GaN selective area epitaxy. Appl. Phys. Lett. 1997, 71, 1204−1206. (30) Kim, Y. H.; Ruh, H.; Noh, Y. K.; Kim, M. D.; Oh, J. E. Microstructural properties and dislocation evolution on a GaN grown on patterned sapphire substrate: A transmission electron microscopy study. J. Appl. Phys. 2010, 107, 063501. (31) Li, J. Z.; Chen, Z. Z.; Jiao, Q. Q.; Feng, Y. L.; Jiang, S.; Chen, Y. F.; Yu, T. J.; Li, S. F.; Zhang, G. Y. Silane controlled three dimensional GaN growth and recovery stages on a cone-shape nanoscale patterned sapphire substrate by MOCVD. CrystEngComm 2015, 17, 4469−4474. (32) Chierchia, R.; Böttcher, T.; Heinke, H.; Einfeldt, S.; Figge, S.; Hommel, D. Microstructure of heteroepitaxial GaN revealed by x-ray diffraction. J. Appl. Phys. 2003, 93, 8918−8925. (33) He, C.; Qin, Z.; Xu, F.; Hou, M.; Zhang, S.; Zhang, L.; Wang, X.; Ge, W.; Shen, B. Free and bound excitonic effects in Al0.5Ga0.5N/ Al0.35Ga0.65N MQWs with different Si-doping levels in the well layers. Sci. Rep. 2015, 5, 13046. (34) Monemar, B.; Paskov, P. P.; Bergman, J. P.; Toropov, A. A.; Shubina, T. V.; Malinauskas, T.; Usui, A. Recombination of free and bound excitons in GaN. Phys. Status Solidi B 2008, 245, 1723−1740. (35) Hertkorn, J.; Brückner, P.; Thapa, S. B.; Wunderer, T.; Scholz, F.; Feneberg, M.; Thonke, K.; Sauer, R.; Beer, M.; Zweck, J. Optimization of nucleation and buffer layer growth for improved GaN quality. J. Cryst. Growth 2007, 308, 30−36. (36) Kornitzer, K.; Ebner, T.; Grehl, M.; Thonke, K.; Sauer, R.; Kirchner, C.; Schwegler, V.; Kamp, M.; Leszczynski, M.; Grzegory, I.; Porowski, S. High-Resolution Photoluminescence and Reflectance Spectra of Homoepitaxial GaN Layers. Phys. Status Solidi B 1999, 216, 5−9. (37) Li, X.-H.; Wang, S.; Xie, H.; Wei, Y. O.; Kao, T.-T.; Satter, M. M.; Shen, S.-C.; Douglas Yoder, P.; Detchprohm, T.; Dupuis, R. D. Growth of high-quality AlN layers on sapphire substrates at relatively low temperatures by metalorganic chemical vapor deposition. Phys. Status Solidi B 2015, 252, 1089−1095. (38) Wang, M.-T.; Brunner, F.; Liao, K.-Y.; Li, Y.-L.; Tseng, S. H.; Weyers, M. Optimization of GaN wafer bow grown on cone shaped patterned sapphire substrates. J. Cryst. Growth 2013, 363, 109−112. (39) Einfeldt, S.; Kirchner, V.; Heinke, H.; Dießelberg, M.; Figge, S.; Vogeler, K.; Hommel, D. Strain relaxation in AlGaN under tensile plane stress. J. Appl. Phys. 2000, 88, 7029−7036. (40) Kneissl, M.; Rass, J. III-Nitride Ultraviolet Emitters; Springer Series in Materials Science: Berlin, 2016.

ZnO/AlN/p-GaN light-emitting diodes by tailoring Ag localized surface plasmon. J. Appl. Phys. 2012, 112, 013112. (7) Paskova, T.; Evans, K. R. GaN substratesProgress, status, and prospects. IEEE J. Sel. Top. Quantum Electron. 2009, 15, 1041−1052. (8) Gibart, P. Metal organic vapour phase epitaxy of GaN and lateral overgrowth. Rep. Prog. Phys. 2004, 67, 667−715. (9) Jain, S. C.; Willander, M.; Narayan, J.; Van Overstraeten, R. III− nitrides: Growth, characterization, and properties. J. Appl. Phys. 2000, 87, 965−1006. (10) Ambacher, O. Growth and applications of group III-nitrides. J. Phys. D: Appl. Phys. 1998, 31, 2653−2710. (11) Chichibu, S. F.; Uedono, A.; Onuma, T.; Haskell, B. A.; Chakraborty, A.; Koyama, T.; Fini, P. T.; Keller, S.; DenBaars, S. P.; Speck, J. S. Origin of defect-insensitive emission probability in Incontaining (Al,In,Ga)N alloy semiconductors. Nat. Mater. 2006, 5, 810−816. (12) Nakamura, S. The roles of structural imperfections in InGaNbased blue light-emitting diodes and laser diodes. Science 1998, 281, 956−961. (13) Kneissl, M.; Kolbe, T.; Chua, C.; Kueller, V.; Lobo, N.; Stellmach, J.; Knauer, A.; Rodriguez, H.; Einfeldt, S.; Yang, Z. Advances in group III-nitride-based deep UV light-emitting diode technology. Semicond. Sci. Technol. 2010, 26, 014036. (14) Ban, K.; Yamamoto, J.-i.; Takeda, K.; Ide, K.; Iwaya, M.; Takeuchi, T.; Kamiyama, S.; Akasaki, I.; Amano, H. Internal quantum efficiency of whole-composition-range AlGaN multiquantum wells. Appl. Phys. Express 2011, 4, 052101. (15) Ť apajna, M.; Kaun, S. W.; Wong, M. H.; Gao, F.; Palacios, T.; Mishra, U. K.; Speck, J. S.; Kuball, M. Influence of threading dislocation density on early degradation in AlGaN/GaN high electron mobility transistors. Appl. Phys. Lett. 2011, 99, 223501. (16) Marino, F. A.; Faralli, N.; Palacios, T.; Ferry, D. K.; Goodnick, S. M.; Saraniti, M. Effects of threading dislocations on AlGaN/GaN highelectron mobility transistors. IEEE Trans. Electron Devices 2010, 57, 353−360. (17) Cheng, J.; Yang, X.; Zhang, J.; Hu, A.; Ji, P.; Feng, Y.; Guo, L.; He, C.; Zhang, L.; Xu, F. Edge Dislocations Triggered Surface Instability in Tensile Epitaxial Hexagonal Nitride Semiconductor. ACS Appl. Mater. Interfaces 2016, 8, 34108−34114. (18) Hiramatsu, K.; Nishiyama, K.; Onishi, M.; Mizutani, H.; Narukawa, M.; Motogaito, A.; Miyake, H.; Iyechika, Y.; Maeda, T. Fabrication and characterization of low defect density GaN using facetcontrolled epitaxial lateral overgrowth (FACELO). J. Cryst. Growth 2000, 221, 316−326. (19) Linthicum, K.; Gehrke, T.; Thomson, D.; Carlson, E.; Rajagopal, P.; Smith, T.; Batchelor, D.; Davis, R. Pendeoepitaxy of gallium nitride thin films. Appl. Phys. Lett. 1999, 75, 196−198. (20) Wuu, D. S.; Wang, W. K.; Wen, K. S.; Huang, S. C.; Lin, S. H.; Huang, S. Y.; Lin, C. F.; Horng, R. H. Defect reduction and efficiency improvement of near-ultraviolet emitters via laterally overgrown GaN on a GaN/patterned sapphire template. Appl. Phys. Lett. 2006, 89, 161105. (21) Ji, Q.; Li, L.; Zhang, W.; Wang, J.; Liu, P.; Xie, Y.; Yan, T.; Yang, W.; Chen, W.; Hu, X. Dislocation Reduction and Stress Relaxation of GaN and InGaN Multiple Quantum Wells with Improved Performance via Serpentine Channel Patterned Mask. ACS Appl. Mater. Interfaces 2016, 8, 21480−21489. (22) Feng, X.; Yu, T.; Wei, Y.; Ji, C.; Cheng, Y.; Zong, H.; Wang, K.; Yang, Z.; Kang, X.; Zhang, G. Grouped and Multistep Nanoheteroepitaxy: Toward High-Quality GaN on Quasi-Periodic NanoMask. ACS Appl. Mater. Interfaces 2016, 8, 18208−18214. (23) Zhang, L.; Li, X.; Shao, Y.; Yu, J.; Wu, Y.; Hao, X.; Yin, Z.; Dai, Y.; Tian, Y.; Huo, Q. Improving the quality of GaN crystals by using graphene or hexagonal boron nitride nanosheets substrate. ACS Appl. Mater. Interfaces 2015, 7, 4504−4510. (24) Lee, K.-S.; Kwack, H.-S.; Hwang, J.-S.; Roh, T.-M.; Cho, Y.-H.; Lee, J.-H.; Kim, Y.-C.; Kim, C. S. Spatial correlation between optical properties and defect formation in GaN thin films laterally overgrown 43392

DOI: 10.1021/acsami.7b14801 ACS Appl. Mater. Interfaces 2017, 9, 43386−43392