1.Introduction

Indium gallium nitride (InGaN) light-emitting diodes (LEDs) can cover the entire range of visible wavelengths, thereby attracting significant attention for different applications, such as color displays and general lighting devices. Commercial white LEDs (WLEDs) are mainly produced using two methods. The most well-known method for producing WLEDs is by combining gallium nitride (GaN)-based blue LEDs and yellow phosphors. Another method is the selective and proper combination of red, green, and blue LEDs to achieve white light emission. Monolithic WLEDs have been validated using two or three vertically integrated InGaN-based quantum wells (QWs) in the LED structure on the same substrate. However, the limited diffusion length of holes causes difficulties in controlling the emission spectra [1, 2]. A number of numerical simulations have shown that white light or multiple wavelength emission from a single GaN-based chip could be achieved by cascading two or three LEDs with different band gaps using tunneling junctions as interconnections [3–6]. However, such a structure cannot be achieved because of the lack of feasible tunneling junctions [3–6]. In principle, monolithic integration could be conducted by laterally connecting LEDs with different emission wavelengths in a single chip. This setup may require numerous runs of epitaxial growth on the same substrate. However, the in-plane modulation of the epitaxial structure using the selective-area growth (SAG) technique is also challenging. InGaN/GaN multiple quantum wells (MQWs) were grown on a GaN template with microfacets using the SAG technique to overcome inherent material problems and to achieve multiple wavelength LEDs [7]. This technique is interesting because semipolar/nonpolar GaN planes can be formed on a c-plane polar substrate, thereby providing GaN films with different facets and growth rates that depend on the direction or shape of the mask layer [8]. GaN bumps featuring different shapes with semipolar/nonpolar facets can be attributed to the different growth rates in the directions of the a-, m-, and c-axes, which can be tuned by growth conditions of the metalorganic vapor phase epitaxy (MOVPE) [9–12]. Consequently, the SAG of InGaN/GaN QWs on a patterned c-plane GaN template causes in-plane wavelength modulation ranging from 371 nm to 438 nm [10]. At this emission range, the emission wavelength is mainly governed by the thickness modulation of the InGaN QWs. Furthermore, SAG techniques can be used to produce microstructures, such as micropyramids, quantum dots, and quantum wires. These microstructures can produce InGaN layers with spatially inhomogeneous distribution of indium (In) content, thereby resulting in multiple wavelength emission [13–17]. Mask layers, such as a patterned SiO₂ film, were formed on the GaN templates before epitaxial growth to achieve SAG during epitaxial growth. InGaN-based LEDs grown on the SiO₂-masked template may have poor electrical properties, such as high turn-on voltage if the SiO₂ mask layer is embedded under the regrown layers. In this study, InGaN/GaN MQW LEDs, which emit multiple wavelengths, are grown on n-GaN templates with truncated hexagonal pyramids (THPs) featuring c-plane and r-plane facets. The THP array was formed by the regrowth of the GaN layer on a selective-area Si-implanted GaN (SIG) template. The n-GaN epitaxial layer is initially grown on the implantation-free regions to form a series of THPs during the growth of InGaN/GaN LED structures on the SIG templates. By contrast, the epitaxial growth of n-GaN on the Si-implanted regions did not occur because the implanted regions with lattice constant differed from the implantation-free regions [18,19]. In particular, the subsequent MQW and p-type layers were grown on the nonplanar surface featuring the LED structure with semipolar and polar facets. The dosage of Si ions used in this study was 1 × 10¹⁶ cm⁻² to create shallow lattice-distortion areas selectively on the n-GaN templates. The SIG layer has low resistivity, thereby not impeding current conduction, unlike the aforementioned SiO₂ mask layer in GaN-based LEDs [17].

2.Experiment

Si-doped GaN (n-GaN) epitaxial layers with a thickness of 3 µm used in this study were sequentially grown on c-faced (0001) sapphire substrates in a vertical MOVPE system (EMCORE D180). SiO₂ and aluminum (Al) layers with thicknesses of 90 and 200 nm, respectively, were deposited in sequence on the GaN epitaxial wafers. The Al layer was patterned as circular dots with diameters of 3 µm to cover the Si ion implantation. The SiO₂ film served as ion stopping layer to alleviate the channeling effect and cause the implanted Si ions that accumulate next to the surface of the n-GaN layer. Figures 1(a) and 1(b) show the schematic layer structures of n-GaN epitaxial and mask layers. Before growing the LED epitaxial structures, Si ion implantation at 70 keV with a dosage of 1 × 10¹⁶ cm⁻² was conducted on the wafers to prepare the SIG templates. The mask layers produced selective implantation to form a circular implantation-free area on the n-GaN layer. The circular implantation-free area had a diameter of 3 µm, and the spacing between the circular areas was 3 µm, as shown in Fig. 1(c). After Si ion implantation, the SiO₂ stopping layer and Al mask layer were removed using the buffer oxide etchant and H₃PO₄-based solution, respectively. Then, the SIG templates were loaded into the MOVPE chamber to regrow the epitaxial layers of the LED structure. Figure 1(d) shows the schematic layer structure of the SIG templates. The layer structure of LED consisted of a 1.7 µm thick Si-doped n-GaN layer grown at 1,035 °C, a 20-pair GaN (1.3 nm)/In_0.05Ga_0.95N (1.8 nm) superlattice as strain-relaxed layer, a 10-pair In_0.2Ga_0.8N (3.5 nm)/GaN (13.5 nm) MQW structure grown at 750 °C, a 0.03 µm thick magnesium (Mg)-doped p-Al_0.12Ga_0.85N electron blocking layer, and a 0.15 µm thick Mg-doped p-GaN top contact layer grown at 950 °C. The schematic structure of the GaN-based THP array is shown in Fig. 1(e). The carrier concentrations of the n-GaN and p-GaN layers were approximately 8 × 10¹⁸ and 5 × 10¹⁷ cm⁻³, respectively. Then, a heavily Si-doped n⁺-InGaN top layer was grown on the p-GaN contact layer [20,21]. Notably, reference wafers using implantation-free GaN templates with planar surface were also grown at the same time for comparison. The structural properties were characterized by transmission electron microscopy (TEM) and scanning electron microscopy (SEM) on the LED wafers, the optical and electrical properties of which were also discussed.

 

Fig. 1 Schematic illustrations of (a) n-GaN grown on sapphire substrate (b)Al mask and SiO₂ ion-stopping layers n-GaN/sapphire template(c)selective-area Si implantation into the n-GaN layer (d) Si-implanted GaN/sapphire template (e)LED epitaxial structure with truncated hexagonal pyramid array.

Download Full Size | PDF

3.Results and discussions

Figure 2(a) shows a typical top-view SEM image taken from the n-GaN layer regrown on the SIG templates. The THP array formed by SAG in the <0001> direction exhibited six equivalent semipolar facets and a (0001) top surface in each pyramid because of the hexagonal wurtzite crystal symmetry of GaN. Cross-sectional TEM images taken from the samples were evaluated to determine the crystal planes of the semipolar facets. The angle between the semipolar and c-plane facets observed in the images was approximately 62°, indicating that the semipolar facets were $\left\{1\overline{1}01\right\}$ planes [22,23]. The average heights of the THPs were approximately twice as large as those of the reference samples, that is, the epitaxiallayers grown on implantation-free n-GaN templates. This difference was due to the loading effect during the SAG process. The SAG on the SIG templates could be attributed to the n-GaN preferentially grown on the implantation-free regions rather than on the Si-implanted regions. The absence of epitaxial growth on the Si-implanted regions was attributed to the fact that the sticking coefficient of Ga and/or N species on the surface of implanted regions, which have amorphous-like property caused by the high-dose implantation, might be significantly lower than that of implantation-free regions. The crystal structure of the GaN layer subjected to high doses and/or energy levels of ion bombardment produces an amorphous layer [18,19]. Based on previous studies, the critical dose for the amorphization of GaN was approximately ~10¹⁶ cm⁻². Moreover, the lattice constant of the Si-implanted GaN increases with the implantation dose and/or energy [18,19]. In this study, the implanted Si ions induced a damaged layer with a depth of approximately 60 nm from the surface, as indicated by the simulation results. In particular, the implanted regions with distorted lattices behaving like a mask layer, such as the SiO₂ layer, used in the conventional SAG process. During the growth of n-GaN on the SIG templates, reaction species (i.e., Ga atoms) were preferentially absorbed on the implantation-free regions rather than on the Si-implanted regions. On the other hand, the formation of the top (0001) facet on each THP could be attributed to the fact that the diffusion length of Ga was sufficiently high to reach a balance between desorption and absorption rates of the Ga species on the (0001) facet because the growth temperature was as high as 1,035 °C. In general, Ga atoms on the $\left\{1\overline{1}01\right\}$ facets are readily desorbed and might evaporate to the gas phase to be absorbed on the top (0001) facet [22–24]. Furthermore, Ga atoms absorbed on the (0001) facet migrate outward to the $\left\{1\overline{1}01\right\}$ facets when the diffusion length of Ga atoms is larger than the width of the (0001) facet. As a result, the THP with a stable and self-limited facet structure is achieved as the incoming and outgoing Ga species to the (0001) facet are balanced [24,25]. A typical cross-sectional TEM image taken from the GaN layer regrown on the SIG template is shown in Fig. 2(b). As shown in the inset of Fig. 2(b), a dark layer with a thickness of approximately 60 nm can be observed near the surface of the GaN template layer. The clear contrast between the implanted and implantation-free regions indicates that the lattice distortion of Si-implanted GaN was significant when compared with the implantation-free GaN. In addition, no additional layer could be observed on the surface of the implanted region (i.e., Si-implanted GaN layer). On one hand, this finding indicated that the regrowth of the GaN layer did not occur on the surface of the Si-implanted region. On the other hand, a lateral overgrowth phenomenon can be observed near the intersection between the semipolar and c-face planes, as shown in the inset of Fig. 2(b). A typical cross-sectional SEM image of an LED structure grown on the SIG template is shown in Fig. 2(c). The THP array can be clearly observed on the wafer after the growth of the LED structure because the total thickness of the LED epitaxial layers, including the InGaN/GaN MQW and p-GaN layers, was insufficient to even out the depressions between n-GaN THPs. However, as shown in Fig. 2(c), the spacing between the neighboring THPs was decreased as compared with that shown in Fig. 2(a). The LED structure grown on the n-GaN THP array composed of the semipolar and c-plane (0001) facets. As shown in Figs. 2(a) and 2(c), the average spacing between the THPs decreased from 2 µm to 0.2 µm after the regrowth of the LED structure on the n-GaN THP array. In Fig. 3(a), the typical TEM image shows a V-shaped depression, indicating that the TEM image was taken from a local region between two THPs. A lateral bright line with a width of approximately 60 nm was observed under the V-shaped depression. This bright line was attributed to the Si ion implantation-induced damaged layer. The line width was consistent with the estimates of simulation using the Transport of Ions in Matter program. Considering the selected area electron diffraction patterns determined around the implanted region (data not shown here), the lattice of Si-implanted GaN exhibited severe distortion. Moreover, the implantation-induced damage was not recovered to a single crystal phase after regrowth. In contrast to the n-GaN THP array, as shown in Fig. 2(a), a considerable thickness of epitaxial layers was observed between the THPs, indicating that crystal growth occurred during the regrowth of the LED structure. Notably, AlGaN- or InGaN-based films were directly deposited on the Si-implanted regions after the growth of the n-GaN layer on the SIG template. For the n-GaN layer on the SIG template, the growth temperature and pressure were 1,035 °C and 100 Torr, respectively. In fact, the selectivity of GaN-based layers regrown on SiO₂-masked GaN templates was strongly dependent on growth temperature and pressure [24,25]. Selective area epitaxy of GaN on the SIG templates exhibited similar results, wherein the selectivity of GaN regrown on the SIG templates increased with the increase in growth temperature and/or the decrease in growth pressure. However, this contention is unsuitable for the case of In_xGa_{1 − x}N and Al_xGa_{1 − x}N epitaxial layers regrown on the SIG templates because the In_xGa_{1 − x}N is generally grown at a temperature of approximately 750 °C and a pressure of 200 Torr to 300 Torr. The Al_xGa_{1 − x}N epitaxial layer contains Al atoms with higher sticking coefficient (i.e., low diffusion coefficient) on the growth surface, even when grown at high temperature. In this study, the LED epitaxial layer structure including a 20-pair GaN/In_0.05Ga_0.95N superlattice was grown at a temperature of approximately 750 °C and a pressure of 300 Torr, followed the n-GaN layer grown at 1,035 °C. Therefore, the deposition of films on the implanted regions started from the growth of GaN/In_0.05Ga_0.95N superlattices. As a result, the subsequent layers, including GaN/InGaN MQW, Mg-doped AlGaN, and GaN layers, were stacked on the GaN/In_0.05Ga_0.95N superlattices.

 

Fig. 2 Typical SEM images of (a) an n-GaN layer regrown on the SIG template (b) a cross-section-view TEM image taken from the boundary between the regrown GaN layer and the Si-implanted GaN surface(c) LED structure including n-GaN, InGaN/GaN MQW and p-GaN layers grown on the SIG template.

Download Full Size | PDF

 

Fig. 3 Typical cross-sectional TEM images taken (a) between two THPs near the valley in the $\left[11\overline{2}0\right]$ direction (b) inspection at the bottom of V-shaped depression (c) inspection at top surface of THP (d) inspection at the intersection between the semipolar and top facets.

Download Full Size | PDF

Figure 3(b) shows an enlarged TEM image of the bottom of the V-shaped depression above the Si-implanted layer. Given the significant lattice mismatch between the GaN/In_0.05Ga_0.95N superlattices and the Si-implanted GaN, the crystal quality of the GaN/In_0.05Ga_0.95N superlattices on the Si-implanted regions was poor. This contention is supported by the finding that a layer-by-layer contrast could not be observed at the bottom of the V-shaped depression, as shown in Fig. 3(b). In addition, the MQW active layer grown on the superlattice exhibited a disk-like structure, as shown in Fig. 3(b). This result could be attributed to the fact that the V-shaped defects originating from the pits on the InGaN layers in the GaN/InGaN superlattice resulted in the in-plane spatial discontinuity in the subsequent GaN/In_0.2Ga_0.8N QWs. The V-shaped defects were eventually filled by the subsequent Mg-doped AlGaN and GaN layers. A similar result was also observed at each top (1000) surface of the THP array, as shown in Fig. 3(c). Figure 3(d) shows an enlarged TEM image taken from a THP at the intersection between the $\left\{1\overline{1}01\right\}$ semipolar and top (0001) facets. Notably, different thicknesses and potentially indium compositions in QWs were observed on the different facet types. As shown in Fig. 3(d), the QW thickness of the top (0001) and $\left\{1\overline{1}01\right\}$ facets were determined to be 7 and 2 nm, respectively. This result was consistent with the observations reported by Wang et al. [26]. The difference in QW thickness between different facets could be attributed to the fact that different growth rates on the facets were a result of the different diffusion lengths of the reaction species adsorbed on those planes [25–27]. In addition, the indium incorporation rate in In_xGa_{1 − x}N QWs grown on the semipolar facets should be different from those on the top facets [27,28]. In addition to the issue of crystal orientation, the aforementioned discrepancy in growth rate and/or indium incorporation efficiency for the growth of In_xGa_{1 − x}N on the GaN template with THP array is also dependent on the growth temperature. In general, a higher temperature results in a higher diffusion length of the reaction species during growth. Considering Figs. 3(c) and 3(d), the barrier thickness for the top (0001) and $\left\{1\overline{1}01\right\}$ facets were determined to be 46 and 3.5 nm, respectively. However, the growth rate of the p-layer (p-GaN and p-AlGaN) for the top (0001) and $\left\{1\overline{1}01\right\}$ facets exhibited an opposite trend to that of the InGaN/GaN MQW. The p-layer (p-GaN and p-AlGaN) thickness for the top (0001) and $\left\{1\overline{1}01\right\}$ facets were determined to be approximately 35 and 80 nm, respectively. This result could be attributed to the fact that the higher growth temperature of the p-layer (950 °C) resulted in a higher diffusion length of the reaction species, especially for Ga atoms, compared with the InGaN/GaN MQW grown at a temperature of approximately 750 °C. Therefore, Ga atoms absorbed on the (0001) facet readily migrate outward to the semipolar facets during the growth of p-GaN and p-AlGaN layers. In addition, the growth rates of Mg-doped (Al)GaN layer grown on the semipolar and top (0001) facets may be drastically affected by Mg incorporation [21,29]. Consequently, the total thickness of the p-layer on the semipolar facet was significantly larger than that on the top (0001) facet on each THP. By contrast, the InGaN/GaN MQW was grown at a relatively lower temperature compared with the growth of the Mg-doped (Al)GaN layer. Therefore, the diffusion length of the reaction species, especially for gallium and indium atoms, might be lower than the width of the top (0001) facet in each THP. As a result, the growth rates of InGaN/GaN layers during the growth of the MQW structure on the top (0001) facet was higher than that on the semipolar facets.Photographs of the LED wafer with THP arrays under different injection currents are shown in Fig. 4(a). The In balls were directly attached to the wafer to serve as the n-type and p-type contact electrodes. Light emission changed from red to near white when the current was increased from 5 mA to 100 mA. Figure 4(b) shows the typical electroluminescence (EL) spectra taken from a reference LED wafer. The epitaxial growth procedure was the same as that of the aforementioned LED structure grown on the SIG template, except that the epitaxial layers of reference wafer were grown on the planar and implantation-free n-GaN templates. In contrast to the LEDs with THP array, the EL spectra exhibited a single-band property that peaked at approximately 470 nm. The inset of Fig. 4(b) shows the blue emission under a driving current of 100 mA. A standard device process was conducted on the LED wafers grown on the SIG templates to evaluate in detail the electrical and optical properties on the LEDs with THP array. In summary, indium-tin-oxide(ITO)-based and Cr/Au-based films served as the anode and cathode electrodes, respectively, after Cl₂-based dry etching was conducted to expose the underlying n-GaN layer [21,30].

 

Fig. 4 (a)Typical photographs taken from the LED wafer with THP array under different current injection (b) typical EL spectra taken from the reference LED wafer under different current injection. The inset shows the photograph of reference LED wafer driven at current of 100 mA

Download Full Size | PDF

Figure 5(a) shows the schematic layer structure of the LED chip with the THP arrays. All of the experimental LEDs used in the present study had an area of 320 × 320 µm². Figure 5(b) shows a typical current–voltage (I–V) characteristic measured from the LED chips with the THP array. At an injection current of 20 mA, the typical forward voltage (V_f) was approximately 3.8 V. This V_f is slightly higher than the reference LEDs with V_f of approximately 3.5 V. The relatively higher V_f could be attributed to poor current spreading because of the nonplanar p-GaN layer. In other words, the effective lateral length of ITO current spreading layer deposited on the nonplanar p-GaN layer was longer than that of ITO layer on the reference LEDs with planar p-GaN layer. Photographs taken from the LED chips with the THP array under different injection currents are shown in Fig. 6(a). Light emission changed from red to near white when the current was increased from 3 mA to 100 mA. At a relatively low current of 3 mA, the LEDs exhibited red emission. Figure 6(b) shows the typical EL spectra taken from the LED chips with the THP array under different injection currents. Single-band spectra peaked at approximately 600 nm (red band) were observed when the injection currents did not exceed 5 mA. The red band showed a significant blueshift from 600 nm to approximately 560 nm as the injection currents were increased from 5 mA to 20 mA. As the injection currents exceeded 20 mA, the blueshift of the yellow-green(YG) band continued with the increase in the injection current, as shown in Fig. 6(c). The strain-induced polarization field resulting in the quantum-confined Stark effect (QCSE) caused the blueshift of the emitted spectra with the increase in the injection current [31].

 

Fig. 5 (a) schematic layer structure of LED chip with the THP arrays (b) typical current-voltage (I-V) characteristic measured from the LED chip with the THP arrays.

Download Full Size | PDF

 

Fig. 6 (a)Typical photographs taken from the LED chip with different driving currents (b) EL spectra taken from the LED chip with driving currents from 3 to 20 mA(c) EL spectra taken from the LED chip with driving currents from 25 to 100 mA.

Download Full Size | PDF

However, the magnitude of the shift became smaller when the LEDs were driven at an injection current of larger than 20 mA as shown in Fig. 6(b). This alleviation of the blueshift could be attributed to the fact that the increase in injection current increases the junction temperature to cause a redshift of the emission band, thereby compensating for the blueshift caused by the screening of QCSE. Another emission band that peaked at approximately 500 nm (green band) was observed as the injection currents exceeded 10 mA. In particular, the origin of the green band that peaked at approximately 500 nm is different from the YG band as the injection currents exceeded 10 mA. In this study, the well thickness on the (0001) and $\left\{1\overline{1}01\right\}$ facets in the THP LEDs were 7 and 2 nm, respectively, which were determined by TEM, as shown in Fig. 3. This result indicated that the growth rate of In_xGa_1-xN layers on the (0001) plane was higher than on the$\left\{1\overline{1}01\right\}$ planes. Funato et al. revealed that the In content and growth rate of In_xGa_1-xN layers in the semi-polar QWs was lower than in the c-axis plane [16,17]. In light of the aforesaid contention, one tentatively proposed that the green and red band emissions might most likely result from the c-plane MQWs, which were grown at the V-shaped depression between two THPs and the top facet in each THP, because of the relatively thicker wells and higher indium content. However, this contention should be further validated using microscopic techniques, such as the measurements of spatially resolved cathodoluminescence and energy-dispersive X-ray spectroscopy under TEM inspection because the indium contents in the MQWs grown on different facets might be different.

Figure 7 depicts the variation in the Commission Internationale de l’Éclairage(CIE) color coordinates taken from typical LEDs with THP array driven at different currents which correspond to the EL spectra shown in Fig. 6(a). In addition, the EL spectra corresponded to the changes in correlated color temperature (CCT) from 1,800 to 16,000 as the injection current increased from 5 mA to 100 mA, respectively. The significant increase in CCT with the increase in the injection current was attributed to the blueshift of the emission spectrum, especially when the injection current exceeded 20 mA because the thin wells grown on semipolar facets dominated the emission. Considering the THP LEDs, carriers experience relative higher barrier in energy when they were injected into the QWs in the $\left\{1\overline{1}01\right\}$ facets especially at low driving current(i.e., low forward bias). In other words, at low driving current, the light emission mainly resulted from the thick QWs on (0001) facets because the injected carries experiences relative lower energy barrier in the (0001) facets. Therefore, at lower driving current levels (3mA and 5mA), the emission is dominated by red color range lights. On the other hand, at higher injection current levels(i.e., high forward voltage), the emission peak intensity at the 470~490 nm increased steeply and surpassed the emission peak at around 560 nm because the carriers have energy enough to inject into the QWs in the $\left\{1\overline{1}01\right\}$ facets, as shown in Fig. 6(c). As a result, at high driving currents the THP LEDs exhibited white light emission with high CCT, as shown in Fig. 7.

 

Fig. 7 Typical CIE color coordinates taken from the LEDs driven with different currents.

Download Full Size | PDF

4. Conclusions

In summary, we have validated the selective-area regrowth of Si-implanted GaN template layers to form the THP arrays for achieving nonplanar InGaN/GaN MQWs in GaN-based LEDs. Unlike conventional designs that use a dielectric layer as the mask layer, the Si-implanted GaN template featuring a planar surface and the low-resistivity mask layer protected the devices from high operation voltage. The broad emission spectra were attributable to the spatial variation of the well thicknesses and the In contents in MQWs on the semipolar facets or c-planes. Facet control can be easily achieved by tuning the growth conditions during the regrowth of the epitaxial layers. Therefore, the Si-implanted GaN template is a potential choice for fabricating GaN-based LEDs that emit multiple peaks with a single chip to achieve white light emission without the use of phosphors.