1. Introduction

Gallium nitride (GaN) has attracted much attention due to its excellent optoelectronic property with a direct band gap of 3.4 eV, high mobility and excellent thermal stability^[1]. It is widely accepted that GaN nanostructures are highly promising materials for producing devices with excellent performance, such as LDs^[2], LEDs^[3], solar cells^[4], sensors^[5], and piezo-electric nanogenerators^[6]. GaN nanostructures have many superior properties due to the possibility of quantum confinement, coaxial heterostructures, correlated photon emission, and photonic crystal effects, which make it a potential contender for nanolasers and nanoLEDs^{[7, 8]}. In particular, these GaN nanostructures could provide additional advantages for light emission, quality improvement of regrowth, etc^{[9, 10]}. GaN nanostructures such as nanorods (NRs) with high aspect-ratio and large surface-to-volume ratio can dramatically reduce the dislocation density in the upper part of the NRs^[11]. Besides, the nanopillars with small footprints help to relieve the strain induced by thermal expansion mismatch and avoid crack generation. The light-extraction efficiency is expected to increase owing to the non-planar geometry of nanowires. Compared to thin films, heterostructures on GaN nanostructure arrays with low defect density are much easier to fabricate and lead to high-performance devices^[12]. Also, GaN nanostructures can provide a unique opportunity for understanding the electronic, optical, and mechanical properties of the material.

Nano-GaN has been successfully synthesized via different methods, such as molecular beam epitaxy^[13], selective growth on patterned substrates by metalorganic chemical vapor deposition (MOCVD)^[14], and hydride vapor-phase epitaxy (HVPE)^[15]. These above-mentioned methods are so-called "bottom-up" techniques, where nanostructures are grown from atoms or molecules to clusters. However, they often show a broad distribution of size, length, and orientation.

Contrary to bottom-up synthesis, a top-down etching process is an alternate way to achieve nanostructures^{[16, 17]}. Top-down methods, such as inductively coupled plasma (ICP) etching, is more economic and extensively controlled and could be applied in the fabrication of a large-area nano-GaN in the future. In this work, we fabricate uniform GaN nanopillar arrays by inductively coupled plasma etching using Ni nano-islands as the masks. A high density of $c$-axis oriented nanopillars ($\sim$10$^{8}$ cm$^{-2})$ with a tip diameter of about 350 nm and a height of 2 $\mu$m has been reported.

2. Experiment

Figure 1 shows a schematic diagram of the formation of GaN nanopillars on sapphire. A 2 $\mu$m GaN layer on (0001) sapphire by MOCVD was used as the original substrate^[18]. A Ni metal film was first deposited on GaN/sapphire by electron-beam evaporation. Then, Ni film/GaN/sapphire is annealed under the ammonia for the Ni nano-islands mask on GaN/sapphire. Finally, GaN not covered with Ni nano-islands was removed by ICP etching. During the ICP etching, the flow rates of chlorine and BCl$_{3}$ were maintained at 48 and 6 sccm, respectively. For the whole process, the chamber pressure, ICP power, and RF power were 7.50 $\times$ 10$^{-9}$ Pa, 300 W, and 100 W, respectively. Photoluminescence (PL) spectra were recorded by using a continuous wave He-Cd laser source at room temperature (RT). Scanning electron microcopy (SEM), energy dispersive X-ray (EDS), and Raman scattering measurements have been used to characterize the morphologies and structural properties.

3. Results and discussion

Figure 2 shows the SEM images of 35 nm thick Ni film annealed using ammonia at 850 ℃ for different times. Ni film was transformed to nano-islands after the NH$_{3}$ annealing, which was attributed to the heating effect and partially etched off by NH$_{3}$ plasma at the same time^[19]. The average size of nano-islands decreases from 600 to 400 nm as the annealing time increased from 4 to 12 min (Figs. 2(a)-2(c)) and then the bigger Ni nano-islands begin to absorb any smaller islands nearby and the average size of the Ni nano-islands increase until all the Ni islands are fully discrete (Figs. 2(d) and 3). Figure 3 shows the change of the average size of the Ni nano-islands from Ni film with different thicknesses at the same NH$_{3}$ conditions. They have the similar trend for the size of Ni islands as the annealing time changes.

Figure 4 shows the SEM images of Ni film with different initial thickness annealed using ammonia at 850 ℃ for 12 min. It can be seen that the size of the Ni islands increased with increasing the Ni film thickness, but the uniformity became worse. As can be seen above, we think that uniformly distributed Ni nano-islands can be obtained by controlling the NH$_{3}$ annealing time and initial thickness of Ni films. In our research, the optimal parameters for the NH$_{3}$ annealing are as follows: 12 min, 850 ℃ and 35 nm Ni film.

Figure 5 shows the SEM image of GaN with 35 nm Ni film by NH$_{3}$ annealing at 850 ℃ for 12 min after the ICP process, where the uniform array of GaN nanopillars is observed on the surface. GaN nano-pillars are quite homogeneous in their shape and size and almost of the same height, which indicates that GaN film under the Ni nano-island mask is not affected by the ICP process while the exposed GaN is etched away. From the top-view SEM images, the average tip diameter and the density of the pillars at different parts of the sample are calculated to be about 350 nm and 2.6 $\times$ 10$^{8}$ cm$^{-2}$, respectively. Comparing the final diameter of the pillars ($\sim$350 nm) with that of Ni islands ($\sim$400 nm), it is evident that there is a lateral etching in Cl$_{2}$ plasma, which produces tapered structures. Two possible sources of lateral etching are wider ion angular distribution function (IADF) and the mask erosion due to the formation of Ni chlorides under highly reactive Cl ions^{[16, 20]}.

Because GaN nano-pillars have a homogeneous vertical shape, we think that some Ni particles remain on the top of the GaN nanopillars. SEM with energy dispersive X-ray (EDX) spectrometry has been carried out for the presence of Ni and the residual Ni can be seen only on the top part of the pillars (Fig. 6).

The results of the Raman scattering measurements are shown in Fig. 7, where the E$_{2}$ (high) and A$_{1}$ (LO) modes can be seen. The strain within the GaN samples was estimated from the frequency shifts of the E$_{2}$ (high) phonon line in the Raman spectra^[21]. The E$_{2}$ (high) peak position for GaN film on sapphire was 568.11 cm$^{-1}$, and that for the GaN nanopillars was 565.73 cm$^{-1}$. Compared with GaN bulk materials, the strain was transformed from the press stress in GaN film to the tensile stress in GaN nanopillars due to the ICP etching.

PL spectra of the GaN film and GaN nanopillars array are measured at room temperature (RT). The PL spectrum of GaN nanopillars has a sharper and stronger near-band-edge emission as compared with GaN film, where a 61 meV redshift can be observed. ICP damage, a Stokes shift, or defect impurity states^{[22, 23]} may induce the redshift of the PL band. A biaxial stress relaxation of about 2.8 GPa is estimated by using the proportionality factor $K=$ 21.1 meV/GPa for the stress-induced PL peak shift^[24]. In addition, an enhancement of the near-band-edge PL intensity value indicates good optical quality of the GaN nanopillars. We attributed the enhancement of PL intensity to the reduction of the internal reflection^[25].

4. Conclusions

High-density GaN nanopillar arrays have been successfully fabricated using Ni nano-islands as the inert masks. The thickness and the NH$_{3}$ annealing time of Ni film have an important effect on the size and distribution of GaN nanopillars. The PL peak redshift indicates that the significant stress relaxation and the enhancement of PL intensity imply an improvement of optical properties. Further work will focus on the growth of GaN LEDs on the GaN nanopillar arrays for the achieving high luminescence efficiency.