1. Introduction

In general, the light output power of a light-emitting diode (LED) depends on the quantum efficiency of the active layer and the light extraction efficiency (LEE). One of the most significant problems with LEDs is the occurrence of trapped light within semiconductors with high refractive indices [1,2]. To improve LEE, efforts have been made to overcome the significant photon loss that results from the total internal reflection inside LEDs. For GaN/sapphire–based blue/green LEDs, the refractive indices of sapphire and GaN are approximately 1.7 and 2.4, respectively. Therefore, photons emitted from the active layer are virtually trapped within the GaN-based layers if the GaN/sapphire and the GaN/air interfaces are specular. The LEE is enhanced by roughening the semiconductor/air and semiconductor/substrate interfaces or by shaping the die into a non-cubic geometry to increase the light escape cone in the LED. Methods to control the growth conditions to obtain rough surfaces in situ are theoretically superior to etching methods [3]. The applications of naturally textured surfaces, including V-shaped pits or truncated micro-pyramids that originate from a change in the conditions during the growth of p-GaN top contact layer, are well-known approaches [4,5]. Recently, GaN/sapphire–based LEDs grown on patterned sapphire substrates (PSS) to create a rough GaN/sapphire interface, and consequently enhance LEE, has become a well-accepted technology for mass production [6–8]. Tsai et al. demonstrated another approach using in situ SiH₄-treatment to create a naturally textured GaN/sapphire interface in GaN/sapphire–based LEDs. This process led to the formation of a series of voids at the GaN/sapphire interface, resulting in effective light scattering to enhance LEE. In contrast to the self-assembled texture of the top GaN/air and GaN/sapphire interfaces, which are naturally formed during epitaxial growth, the rough GaN/PSS interface needs additional processing, e.g., sapphire is initially patterned by wet chemical etching or by dry plasma etching before epitaxial growth [6–8]. In addition, the growth conditions are strongly dependent on the depth and spacing of the PSS patterns and surface morphology thereon [6–8]. In this study, an approach aimed at achieving guided-light scattering layers beneath the active layer of GaN-based LEDs was demonstrated to improve LEE. The epitaxial layer structures of InGaN/GaN blue LEDs are grown on n-GaN template layers with selective-area Ar ion implantation (Ar-implanted GaN). The dosage of Ar ion used in this study was as high as 1 × 10¹⁶/cm² to selectively create a series of shallow damage layers on the n-GaN templates. During the growth of the InGaN/GaN LED structures on the Ar-implanted GaN templates, the epitaxial layer was initially grown on the implantation-free regions whereas the deposition of GaN on the Ar-implanted regions did not occur because the lattice constant of the latter regions was different from the former. As a result, selective growth occurred in the GaN epitaxial layer on the Ar-implanted GaN templates, and a series of hexagonal concaves were formed on the GaN layer. In contrast to the conventional PSS or SiO₂ nanorod-embedded GaN templates with concaves and/or convexes [9], the Ar-implanted GaN templates were substantially flat. In this study, AlGaN layers were embedded in the InGaN/GaN LED structures that were grown on the Ar-implanted GaN templates to create guided-light scattering layers for improving the LEE of LEDs. The output power of the bare chip LEDs were measured with an integral sphere. Detailed processing procedures and related results, including the electrical and optical properties of the fabricated LEDs, are discussed subsequently.

2. Experiments

The samples used in this study were grown on c-face (0001) sapphire substrates in a vertical metal-organic vapor-phase epitaxy (MOVPE) system. Before the growth of the LED structures, n-type GaN epitaxial layers, including a 1 μm-thick undoped GaN layer and a 2 μm-thick Si-doped n-GaN layer, were grown on the sapphire substrates as templates for the subsequent regrowth process. The carrier concentration of the n-GaN template layer was around 8 × 10¹⁸/cm³. Ar ion implantation, in which the dosage and energy were 1 × 10¹⁶/cm² and 100 keV, respectively, was then selectively performed on the GaN templates to create circular pattern regions. A 200 nm-thick Ni layer served as the mask layer. Circular openings on the Ni mask layer with a diameter of 3 μm were defined using standard photolithography and wet etching. In addition, a 90 nm-thick SiO₂ layer was then deposited in the opening so that the implanted Ar ions accumulate next to the surface of the GaN layer. After removing the mask layers, the diameter of the Ar-implanted regions and the spacing between the circular patterns were 3 μm, as shown in the Fig. 1(a) . Next, the Ar-implanted GaN templates were loaded into the MOVPE reactor to regrow the LED structures; these labeled as LED-I. LED structures grown on conventional GaN templates without the Ar-implanted regions were also prepared for comparison and were labeled as LED-II. The LED structure included a 1.7 μm-thick Si-doped n-GaN layer, a 0.05 μm-thick Mg-doped p-Al_0.15Ga_0.85N electron blocking layer, and a 0.2 μm-thick Mg-doped p-GaN top contact layer all grown at 1000 °C, as well as a ten-pair In_0.3Ga_0.7N/GaN MQW structure grown at 750 °C. Afterwards, a heavily Si-doped short-period superlattice (SPS) structure was grown on the p-GaN contact layer [10]. During the regrowth of the GaN epitaxial layer on the Ar-implanted GaN templates, the layer was initially grown on the implantation-free regions whereas the deposition of GaN on the Ar-implanted regions did not occur. As a result, GaN layer with holes corresponding to the implanted regions was formed, as shown in Fig. 1(b). Afterward, the GaN layers coalesce laterally between the Ar-implanted regions to form GaN layer with V-shaped concaves, as shown in the schematic diagram of Fig. 1(c).

 

Fig. 1 (a) Illustration of Ar-implanted on GaN template (b) GaN regrowth rate on implantation-free regions more than Ar-implanted regions (c) GaN layer to form V-shaped concaves typical tilted-angle-view SEM image of LED-I (d) interted AlGaN/GaN pyramidal sells on the concaves (e) epitaxial blue LED structure on AlGaN pyramidal shells.

Download Full Size | PDF

Figure 2(a) shows the typical top-view micrographs obtained through scanning electron microscopy (SEM). A series of hexagonal holes are observed on the GaN layer. This could be attributed to the relatively lower growth rate of GaN at the Ar-implanted regions than at the implantation-free regions. In general, the crystal structure of semiconductors undergoing high-dose and/or high-energy ion bombardment produces an amorphous layer [11]. In this study, the implanted Ar ions created a damage layer with a depth of approximately 50 nm from the GaN surface estimated by simulation. The difference in the growth rates of the two aforementioned regions resulted from the difference in their lattice constants. According to previous reports that critical doses for the amorphization of GaN films were shown to be around 6 × 10¹⁵/cm², and the lattice constant of Ar-implanted GaN increased with an increase of implantation dose [11]. Figure 2(b) shows a typical cross-sectional image under SEM. The V-shaped concaves observed on the GaN layer regrown on the Ar-implanted GaN templates indicate that selective and lateral overgrowth indeed occurred during the epitaxial regrowth. After the formation of the V-shaped concaves(i.e., in cross-section view) in the regrown GaN layer, inverted AlGaN pyramidal shells were formed by the deposition of two Al_0.27Ga_0.83N layers with a thickness of 150 mm separated by a 250 nm-thick n-GaN layer on the concaves, as shown in Fig. 1. Figure 2(c) shows a typical SEM micrograph taken near the Al_0.27Ga_0.83N/GaN interfaces, which corresponds to Fig. 1.

 

Fig. 2 (a) Top-view micrographs obtained through scanning electron microscopy (SEM) (b) cross-sectional view by SEM (c) typical SEM micrograph taken near the Al_0.27Ga_0.83N/GaN interfaces (d) shows schematic structure of the LED-I.

Download Full Size | PDF

Accordingly, GaN-based LEDs with embedded inverted AlGaN pyramidal shells beneath the MQW active layer could be formed using GaN templates with selective Ar implantation. Figure 2(d) shows schematic structure of the LED-I. In this study, the processing methods applied to the formation of mesa and electrodes were similar to our previous report [12].

3. Results and discussions

Figure 3(a) shows the relative output power-current (L-I) characteristics of LED I and LED II. With a 20 mA current injection, the output power of LED-I is enhanced by 10% compared with LED-II without embedded Al_0.27Ga_0.83N layers. The embedded inverted Al_0.27Ga_0.83N pyramidal shells created two interfaces with index steps that led to an increased escape probability for the photons emitted from the active layer of the LEDs. In other words, light emitted from the active layer of LED-I experiences a strong reflection and redirection at the GaN/Al_0.27Ga_0.83N interfaces, and thereby results in a shorter average path before the photons escape into the free space, as shown in Fig. 3(b) [13]. Figure 3(c) shows the micrograph taken from the LED-I under an operation current of 20 mA. Periodic feature resulted from the light scattering at the embedded AlGaN pyramidal shells could be observed on the LED, as indicated in the inset of Fig. 3(c). Previous researches have established that the emission efficiency of InGaN-based blue/green LEDs is imperceptibly sensitive to structural defects [14] even at densities as high as 10⁹–10¹⁰ cm⁻², six orders of magnitude higher than conventional AlGaAs or AlGaInP LEDs. A popular model has been proposed to account for these defect-insensitive emission characteristics of highly defective InGaN materials. The presence of InN-rich or pure InN clusters in the InGaN quantum assumes the role of quantum dot-like localized states to confine the injected electron-hole pairs that have radiatively recombined without diffusing toward the dislocations [15–17]. In this study, the Al content of the embedded AlGaN layers was approximately 27%, which was calibrated using the AlGaN bulk layer grown on a planar GaN template under the same growth conditions. The total thickness of the Al_0.27Ga_0.83N layers was kept below the critical thickness. Consequently, the material quality between the LED-I and LED-II was assumed as essentially identical. Thus, the enhanced light output was mainly attributed to light scattering at the textured Al_0.27Ga_0.83N/GaN interfaces, which led to a relatively lower internal absorption in the LED-I.

 

Fig. 3 (a) Relative output power of the LED-I and LED-II as a function of injection current (b) illustration of photons escape into the free space (c) light scattering at the embedded AlGaN pyramidal shells.

Download Full Size | PDF

Figure 4(a) and (b) show the cross-sectional ray trajectory images of conventional LED-I and LED-II, respectively. Compared with LED-II, the extracted ray density through the top face of LED-I is markedly higher. Although the ray density in the sapphire substrate of the LED-I is smaller than that of the LED-II due to the emitted light is reflected at the GaN/Al_0.27Ga_0.83N interfaces and redirected to the top surface, this means a relative smaller incident angle of the light into the top GaN/air interface and thereby leads to a relatively lower internal absorption in the LED-I. In principle, the output power of LED-I can be further enhanced by increasing the Al content of the embedded AlGaN layers because the refractive index contrast between the AlGaN and GaN around the inverted AlGaN pyramidal shells increases with a corresponding increase of the Al content in the AlGaN layers. The higher index contrast results in a more significant light scattering around the AlGaN/GaN interfaces, which enhances the LEE of the LEDs. Based on the ray tracing simulation, if Al composition reaches to 50% in the embedded AlGaN layers, the output power of LED-I enhances over 20% compared with that of LED-II, as shown in the inset of Fig. 3(a).

 

Fig. 4 (a) and (b) Show the cross-sectional ray trajectory images of conventional LED-I and LED-II, respectively.

Download Full Size | PDF

The room-temperature current-voltage (I-V) characteristics of LED-I and LED-II were measured with an HP4156 semiconductor parameter analyzer. As shown in Fig. 5 , considering an injection current of 20 mA, the typical forward voltages (V_f) were 3.35 and 3.25 V for LED I and LED II, respectively. The slight difference in the V_f can be attributed to the relatively higher strain on LED-I compared with LED-II due to the additional embedded AlGaN layers in the LED-I. The excess strain caused an additional band tilt in the quantum wells and hence an effective band discontinuity that substantially hinders the carrier transport. Structural defects such as threading dislocations (TDs) strongly affect the GaN-based LEDs that raise the leakage current. As shown in Fig. 5, the reverse I-V characteristics of the LED-I and LED-II were almost the same, implying that the embedded inverted AlGaN pyramidal shells did not cause a marked difference in material quality such as the amount of TD density. Although the V-shaped concaves resulted from the selective growth that occurred around the Ar-implanted regions in the LED-I, the effect of the lateral growth mechanism on the reduction in the TD density was negligible because the spacing between the Ar-implanted regions was relatively small compared to those of the so-called epitaxial laterally overgrown processes, which are 3–10 μm apart between the SiO₂ mask layers [18]. Therefore, the effect of TD bending on the reduction in TD density is limited. On the other hand, previous research has reported the apparently insignificant effect of TD density on the efficiency of GaN-based LEDs emitting at visible wavelengths [15–17]. In this study, the enhancement of output power is mainly due to the light scattering at the AlGaN/GaN interfaces, which leads to a higher escape probability for the photons.

 

Fig. 5 Room-temperature current-voltage (I-V) characteristics of LED-I and LED-II.

Download Full Size | PDF

4. Conclusions

In summary, we have demonstrated that GaN-based LEDs grown on Ar-implanted GaN templates form inverted AlGaN pyramidal shells beneath an active layer. GaN-based epitaxial layers grown on the selective Ar-implanted regions had lower growth rates compared with those layers grown on the implantation-free regions, resulting in selective growth and the formation of V-shaped concaves in the epitaxial layers. Accordingly, the inverted AlGaN pyramidal shells were formed after the AlGaN and GaN layers were subsequently grown on the V-shaped concaves. Experimental results indicate that the light-output power of the LEDs was higher than that of conventional LEDs with inverted AlGaN pyramidal shells. The enhancement in output power was mainly due to the light scattering at the AlGaN/GaN interfaces, which leads to a higher escape probability for the photons, that is, to an increase of light-extraction efficiency, instead of the improvement in material quality, that is, internal quantum efficiency.