The recent development of dielectric microsphere lithography has been able to open up new means of performing simple and easy patterning on the semiconductor surfaces. Here, we report uniform and high-density arrays of microspheres using a solution-based spin-coating method. The arrays of microspheres were used for etching mask to form the arrays of III-nitride microrods. By regrowing GaN layer on the microrod structures, high-quality GaN layer was achieved in terms of surface morphology as well as XRD characterization. To apply the advantages such as improved crystal quality and light extraction enhancement, light-emitting diodes (LEDs) were grown and then fabricated. The regrown LEDs with microspheres showed much improved optical output power and forward voltage characteristics in the same current injection. Therefore, we believe that this approach is quite useful for the development of high efficiency LEDs for future lighting.

1. Introduction

Colloidal particles have been widely used owing to their material/chemical/biological functionality [1]. Dimension of the colloidal particles can mainly be determined by the type of materials; it can thus be available for diverse applications such as slurries, clays, minerals, and aerosols [1]. The controlled size of the colloidal particles gives rise to special interests in the nanoscale patterning technology since they can provide a simple approach to form monodisperse masks [2]. In particular, the use of silica colloids is quite useful in inorganic solid-state lighting semiconductors since it not only gives realistic dimension of 10–1000 nm, but also enables change of light management [1]. Moreover, the submicrometer-scale patterns can open up new opportunities for the semiconductor to be used for biomimetic applications [3].

One practical application of silica colloids can be to improve light extraction efficiency of light-emitting diodes (LEDs). To date, internal quantum efficiency (IQE) of LEDs is almost saturated (>80%) owing to high-quality GaN epitaxial growth techniques [4, 5]. Besides a design of multiple quantum wells (MQWs), many efforts in LEDs have been made to improve external quantum efficiency such as a reduction of polarization field effects and surface roughening [6, 7]. In particular, the enhancement of light extraction efficiency (LEE) is of critical importance since the angle of light emission is limited by the large difference of refractive index between a semiconductor layer and air (or layer to layer) [8]. Hence, surface patterning and near field enhancement such as surface plasmon polariton (SPP) can be mostly highlighted for improving LEE [7, 911]. However, the fabrication cost to form a designed pattern on a surface or metal nanoparticles is relatively expensive, making it undesirable regarding the cost reduction issue in the LED industry.

In this study, we report high-density and monodisperse arrays of dielectric microsphere using solution-based silica colloids. It is used as a two-dimensional (2D) etching mask to fabricate arrays of III-nitride microrods. By using a regrowth process in metalorganic chemical vapor deposition (MOCVD), we achieved much improved crystal quality of the III-nitride layer. Furthermore, LED chips with embedded microspheres and microrods were fabricated, showing dramatically improved light output power compared to the controlled LEDs. This method can not only provide cost-effective fabrication compared with metal-used patterning, but also block the light absorption on Si substrates owing to the light reflection from the bottom side by inserting novel dielectric particles.

2. Experimental Method

Synthesis of silica particles was carried out by Stöber method. Arrays of dielectric microspheres (MS) were obtained by mixing silica particles in solution with deionized (DI) water, methanol, ammonia (NH3), and diluted tetraethyl orthosilicate (TEOS). The uniformity of silica particles was controlled by sonication, and the diameter of silica particles became ~1 μm. To generate a highly ordered and monodisperse layer, spin coating was performed at 600 rpm for 1 min. To prepare the GaN template on Si substrates, epilayers were grown on 2-inch Si (111) substrates by MOCVD (AIXTRON AIX 2000). Precursors of III and V sources were trimethylgallium (TMGa), trimethylaluminum (TMAl), trimethylindium (TMIn), and ammonia (NH3). As dopants, disilane (Si2H6) and cyclopentadienyl magnesium (Cp2Mg) were used to grow n-type and p-type GaN, respectively. Firstly, Si (111) substrates were heated at 1080°C in a H2 atmosphere for 10 min to remove native oxide on the surface. To avoid surface damage due to melt-back etching, TMAl source was first opened for 10 s before the NH3 injection. A 60 nm thick high-temperature AlN (HT-AlN) buffer layer was grown at 1020°C with Al nucleation process. Subsequently, a three-step AlGaN buffer layer with a thickness of ~600 nm was grown, where Al compositions were composed of 76%, 54%, and 23% from bottom to top layer, respectively. After growing a 500 nm thick u-GaN layer at 1045°C, a 15 nm thick low-temperature AlN (LT-AlN) layer was grown at 820°C to reduce the threading dislocation originated from initial growth procedures. A 400 nm thick u-GaN was additionally grown at 1045°C, after which we could obtain a GaN template to disperse MS particles.

After forming spin-coated MS mask arrays, the samples were etched by inductively coupled plasma (ICP) etcher. The dry etching was performed at 5 mTorr with Cl2 (30 sccm)/BCl3 (28 sccm) gas mixtures. To proceed with a regrowth process, a 1.5 μm n-GaN (Si doping concentration is 5 × 1018 cm−3) was grown to merge the empty gaps of the microrod arrays. Then, 30 nm In0.05Ga0.95N/GaN superlattices (SLs) for a lattice-matched layer (LMI) and seven pairs of multiple quantum wells (MQWs) were grown. Here, a 20 nm p-Al0.21Ga0.79N layer was also grown, which plays a role in an electron blocking layer (EBL). Finally, a 100 nm p-GaN layer and a 20 nm p+GaN capping layer were grown at the same growth temperature of 960°C, where p-type doping concentration was maintained at ~1 × 1020 cm−3. Figure 1 shows the schematic of the LEDs with MS in detail. The surface morphology of grown epilayers was observed using scanning electron microscope (SEM, Hitachi S-4700) and optical microscope (OM, Olympus). Crystal quality of the grown GaN epilayer was characterized using X-ray diffraction (XRD, Philips X’Pert) equipped with a hybrid monochromator (2-bounce) in the incident beam and a channel-cut Ge (220) diffraction beam. Electrical and optical properties of fabricated devices were characterized by an LED chip measurement system. The size of a single chip was 500 × 500 μm2.

3. Results and Discussion

Prior to proceeding with the regrowth in MOCVD, a GaN template was prepared for dispersing MS particles, as shown in the schematic in Figure 2(a). It consists of a 60 nm HT-AlN layer, a 600 nm graded AlGaN layer, a 500 nm undoped GaN layer, a 15 nm LT-AlN layer, and a 400 nm undoped GaN layer. Here, we used 3-step AlGaN layers aiming to release the biaxial strain from GaN/Si interface, reducing cracks during cooling process of epitaxy [12]. The HT-AlN layer was grown to avoid melt-back etching from the chemical reaction between Si and Ga atoms, and an LT-AlN layer was inserted to reduce threading dislocation initiated from Si substrates [13]. Then, a monolayer with a high density of MS arrays was dispersed on the planar GaN template (Figure 2(a)). By performing the ICP etching process with Cl2-based gas mixtures, arrays of III-nitride microrods could be obtained, as shown in the schematic in Figure 2(b). To avoid the melt-back etching in high growth temperatures, dry etching was stopped at the HT-AlN layer. Figure 3 shows the SEM images for the fabricated microstructures. Figures 3(a) and 3(b) display the top-view SEM images of dispersed microspheres at center and edge regions, respectively. For further feasible applications of the spin-coated microstructures, the MS uniformity is of importance. It requires not only to be formed as a monolayer but also to maintain high density. If these two requirements are not satisfied, the regrown structures could not keep the smooth surface morphology due to large difference of coalescence distance [14]. The surface coverages of coated MS were ~80% and ~83% for center and edge regions of a 2-inch wafer, respectively. Moreover, the surface was almost covered by arrays of a monolayer, except for a few regions overlapped by the MS. Figure 3(c) shows a tilt-view SEM image of arrays of III-nitride microrods after the dry etching process. Each region for single microrods was clearly distinguished since the AlGaN alloys showed a distinct brightness compared with that of the GaN layer. The MS remained on the top of the microrods, indicating a chemical stability for dry etching process.

As the next step, an -GaN regrowth was carried out in MOCVD. For the full coalescence, we grew -GaN with thickness of ~1.5 μm. To fairly compare the regrown GaN layers, a planar GaN layer was also grown as a controlled sample. By performing an XRD scan for two samples, we could investigate crystal quality of epitaxial layers. In (002) reflection, the full width of half maximum (FWHM) of regrown GaN layers with MS was 394 arcsec, while that of the controlled sample without MS was 640 arcsec, as shown in Figure 4(a). Furthermore, the FWHM was decreased from 925 to 478 arcsec by applying the regrowth over the microrod structures in (102) reflection. The pits of the GaN surface were also dramatically reduced and surface smoothness with terrace-like atomic features was observed (not shown). This indicates the enhanced crystal quality, as observed from XRD characterization.

Figure 5 shows the SEM images of regrown GaN as growth time increases from 5 to 20 min. At the beginning, the three-dimensional (3D) island-like morphology was observed where the 3D layers were randomly distributed over the whole surface (Figure 5(a)). Then, 2D dominant morphology was shown except for partially empty regions (Figure 5(b)). Finally, smooth 2D layers were grown in 20 min, and after that these 2D layers were thickened by increasing the growth time further (Figure 5(c)). During the coalescence process, several factors can contribute to the improvement of crystal quality from MS structures: dislocation bending, defect blocking, strain suppression, and so forth [1518]. As evident from the improved FWHM in XRD measurement, we believe that not only dislocations from the templates were reduced but also strain above the MS was relaxed during the microrod etching and coalescence process [18, 19].

We further grew the LED structures on both the planar and regrown layers to investigate the effect of electroluminescence by using microrod structures. Figures 6(a) and 6(b) show the schematic of grown structures. The LED structures consist of 30 nm In0.05Ga0.95N/GaN SLs, seven multiple quantum wells (InGaN/GaN = 2.5 nm/12 nm), 20 nm p-AlGaN layer, 100 nm p-GaN layer, and 20 nm p+GaN layer. Here, the LMI was inserted for releasing the biaxial strain at InGaN/GaN interfaces of MQWs; it is thereby able to increase indium incorporation as well as internal quantum efficiency [20]. For fabricating LED chips, conventional photolithography and electrode formation were formed on both grown epitaxial structures. Figure 7 shows the typical spectra of both LED structures at the current injection of 20 mA. The EL intensity of LEDs with MS was significantly improved compared to that of LEDs without MS. The details of device performance are described in Table 1. By measuring the optical output power in the integrating sphere, it was improved from 3.08 to 4.11 mW, a 33% increase. Several factors could affect this optical power improvement, such as enhanced light extraction, crystal quality, and indium homogeneity [7, 9, 21]. In particular, the regrown GaN layer includes embedded MS structures for enhanced light extraction. We can expect more increased light scattering from this structure owing to the largely different refractive index () between GaN ( = ~2.4 at a wavelength of 555 nm) and MS ( = ~1.5) [2224]. This can be highly advantageous for GaN on Si epitaxy since the strong light absorption in Si can be reduced from this light management. In addition, the electrical properties of LEDs can be critically affected by a crystal quality of GaN layer on Si substrates [25]. Hence, the reduced forward voltage from 3.25 to 3.06 V can be attributed to the improved crystal quality, as evident from the FWHM of XRD in Figure 4. Even though the same growth condition was used for both LED samples, slight difference of peak wavelength might be due to the run-to-run deviation as well as spatial deviation.

In terms of prospective view for this method, we think that the density of MS (i.e., surface coverage of MS) is a very crucial parameter for LED performance. In brief, too-high a density of MS (e.g., >95%) is not useful for this regrowth approach since the fully covered MS layer does not provide enough room for a proper regrowth. In contrast, too-low a density of MS (e.g., <30%) cannot contribute significantly to the improvement of crystal quality as well as light extraction of LEDs because the regrown layers seemed to be 2D without additional scattering structures. Figures 8(a) and 8(b) show the representative SEM images for low density ( = ~300 nm) and high density ( = ~800 nm), respectively, where indicates a diameter of MS. In our case, the density of MS was somewhat increased by increasing the diameter of MS (Figure 8(c)). The highest density of MS was acquired at a diameter of 1000 nm. We believe that the slight decrease of the density on the diameter of 1500 nm compared to the MS with the diameter of ~800 nm is attributed to the reduced packing ratio due to the larger size of MS. Figure 8(c) also shows the optical output power of LEDs by applying different diameters of MS. Considerable improvement of the optical output power of MS was observed until the diameter became ~1000 nm. The exceptional decrease for the MS with a diameter of 1500 nm, despite the high density (~80%), is due to poor coalescence during the regrowth. Note that the diameter of ~1500 nm is quite similar to the total regrowth thickness of -GaN. To merge the whole surface, at least, the longer growth time is required under the same growth condition. Therefore, we could achieve the highest optical output power of LEDs with the moderate density of MS (~83%) and full coalescence of GaN over the whole surface.

4. Conclusion

In this paper, we reported improved quality of III-nitride epilayers on Si substrates using regrowth on microrod arrays. To demonstrate the microrod structures, MS arrays were uniformly dispersed on the surface by controlling the synthesis condition, thereby being used for masks. Depending on the density of MS arrays, the density of microrod arrays was considerably affected since we used a top-down approach to form the rod structure. From the XRD analysis, the FWHM of regrown GaN was dramatically reduced, indicating high crystal quality. The fabricated LEDs not only enhanced optical output power by 33%, but also reduced forward voltage characteristics at the same current injection. Therefore, we believe that this method can contribute to the improvement of the electrical and optical properties of LEDs for future lighting.

Conflict of Interests

The authors declare that there is no conflict of interests regarding the publication of this paper.


This research was supported by the Core Technology Development Program for Next-Generation Energy of Research Institute for Solar and Sustainable Energies (RISE).