1. Introduction

AlGaN compound semiconductor has been widely explored for fabricating short-wavelength optoelectronic devices, such as ultraviolet (UV) light-emitting diodes (LEDs) [1] and laser diodes (LDs) [2] that are promising light sources for disinfection, bio-medical and UV curing, etc. However, strong piezoelectric field exists in AlGaN/(Al)GaN MQWs, leading to spatial separation of electrons and holes, or named the quantum confined stark effect (QCSE), which decreases the internal quantum efficiency (IQE) and brings a red-shift in emission spectrum [3, 4]. Another issue is the low light extraction efficiency (LEE) of the AlGaN-based UV LED, which is caused by the internal total reflection at the epi-layers’ interfaces [5, 6], as well as the UV absorption in the p-GaN layer commonly used on top of the p-AlGaN to enhance the hole injection [7–9]. Although nanopillars and nanowires have been reported as effective structures to address the QCSE and low LEE issues in InGaN/GaN-based blue LEDs [10–17], however, up to data the nanopillar structures have not been applied to AlGaN/GaN-based UV LEDs probably due to the difficulty to achieve high-quality AlGaN epitaxial materials, and nanopillars’ optical properties also have not been studied in depth yet.

In this paper, we fabricated nanopillar AlGaN/GaN LED arrays using nanosphere lithography (NSL) technique [17–19] followed by a dry-etching process. Compared to other methods for fabricating nanopillar structures such as self-organized nickel islands [10], nano-imprinting [11], electron-beam lithography [12] and interferometric lithography [20], NSL has advantages in cost, efficiency and large area uniformity. The optical properties of the nanopillar LEDs were investigated by cathodoluminescence (CL), temperature-dependent photoluminescence (PL) and time-resolved PL (TRPL) measurements, showing suppressed QCSE compared versus the as-grown LEDs. The strain relaxation within the nanopillar MQWs was confirmed by micro-Raman measurement. Besides, the finite difference time domain (FDTD) method was performed to simulate how the UV photons escape out of the LED device by comparing the LEE from top, bottom and side for both the as-grown and nanopillar samples.

2. Experiment

The AlGaN/GaN LED wafer was grown by low-pressure metal-organic chemical vapor deposition (LP-MOCVD) on a 2-inch c-plane sapphire substrate. The LED structure consists a 25-nm-thick AlN nucleation layer grown at 550 °C, a 200-nm-thick AlN template layer grown at 1200 °C, 20 pairs of AlN/AlGaN superlattices (SLs), a 3-μm-thick Si-doped n-Al_0.15Ga_0.85N layer, five pairs of Al_0.15Ga_0.85N (12 nm)/GaN (3nm) MQWs, an Al_0.2Ga_0.8N electron blocking layer and a 120-nm-thick p-GaN contact layer. The nanopillar structure was fabricated by NSL technique, according to the schematic process flow illustrated in Fig. 1. First, a 200-nm SiO₂ film was deposited on the AlGaN/GaN LED wafer by plasma enhanced chemical vapor deposition, and a highly ordered self-assembled monolayer of polystyrene (PS) nanospheres with a diameter of 630 nm was dip-coated on the wafer. Then, oxygen plasma was applied to the wafer using a reactive ion etching tool to shrink the nanospheres, that were the etching mask to define the SiO₂ nanopillar in subsequent CF₄-based inductively coupled plasma (ICP) etching. After the PS masks dissolving in toluene, the wafer was etched down to n-AlGaN layer using the SiO₂ nanopillar mask by ICP etching with mixture plasmas of BCl₃/Cl₂/Ar. Finally, the SiO₂ mask was removed by buffer oxide etchant and the nanopillar LEDs were treated in HCl to remove ICP-induced sidewall etching damages.

 

Fig. 1 Schematic process flow of fabricating AlGaN/GaN nanopillar LEDs using the nanosphere lithography and dry-etching.

Download Full Size | PDF

3. Results and discussions

Figures 2(a)-2(c) show typical top-view, tilted and cross-sectional SEM images of the AlGaN/GaN MQWs nanopillar structures, respectively. The nanopillars are in a highly uniform array determined by the highly ordered PS nanospheres. From Fig. 2(a), the nanopillar density is ~2.5 × 10⁸ cm⁻². As shown in Figs. 2(b) and 2(c), nanopillars consistently have smooth top and sidewall surfaces with narrow top and width bottom. The top diameter, bottom diameter and height average around 410 nm, 630 nm and 630 nm, respectively.

 

Fig. 2 The top-view (a), tilted (b) and cross-sectional (c) SEM images of the AlGaN/GaN nanopillar structure.

Download Full Size | PDF

Figures 3(a) and 3(b) show the plan-view and cross-sectional monochromatic CL images of the nanopillar sample at an electron acceleration voltage of 5 kV. In Fig. 3(a), the emitting regions of the nanopillars are uniform, showing a strong peak around 354 nm and a weak emission one around 332 nm, corresponded with the MQWs and n-Al_0.15Ga_0.85N layer, respectively, as revealed in the spatial distribution monochromatic CL in Fig. 3 (b).

 

Fig. 3 The CL results of the AlGaN/GaN MQWs nanopillar LED sample measured at room temperature. The plane (a) and cross-sectional (b) monochromatic CL images of the nanopillar LEDs. The inset shows the CL spectrum of the nanopillar LEDs.

Download Full Size | PDF

The PL spectra were measured using a 325-nm He-Cd laser and a spectrometer. Both the optical excitation and light collection were from samples’ top surfaces. Figures 4(a) and 4(b) show the room-temperature (RT, 300 K) and 10K PL spectra for both the as-grown and nanopillar samples, respectively, under identical condition. As shown in Fig. 4(a), the peak PL intensity of the nanopillar sample is increased by 2.3 times when compared with that of the as-grown sample at RT. Moreover, the small emission peak blue shift from 353.9 nm to 353.2 nm (7 meV) in the nanopillar sample can be attributed to the alleviated QCSE because of the strain relaxation in the MQWs region after fabrication of nanopillar structures [21–24]. The reduced QCSE suggests the relieved band bending, the increased wave function overlap of electrons and holes and therefore the increase of IQE. In addition, the large difference of PL intensity at 10 K, shown in Fig. 4(b), indicates the enhanced LEE in the nanopillar sample, assuming that IQE equals to 100% for both samples at 10 K. Thus, the enhancement of room temperature PL intensity can be attributed to the increase of both IQE and LEE (mainly the light extraction enhancement from the top side) in the nanopillar sample.

 

Fig. 4 PL spectra of as-grown sample and nanopillar sample at room temperature (a) and 10 K (b).

Download Full Size | PDF

The ratio of the PL intensities of RT and 10 K is widely used to numerically evaluate the IQE at RT, assuming that the non-radiative recombination at 10 K is negligible and the corresponding IQE is 100%. Figure 5 shows the Arrhenius plots of the temperature-dependent PL intensities of these two samples from 10 K to 300 K. The RT IQEs are calculated to be 33% and 47% for the as-grown sample and the nanopillar sample, respectively, demonstrating a great increase of 42% for the nanopillar sample. In addition, in Fig. 5 the PL intensities of the as-grown and nanopillar samples sharply decreased when the temperatures were above 70 K and 160 K, respectively. The higher temperature for the nanopillar sample means that its non-radiative recombination centers need higher energy to be activated.

 

Fig. 5 Arrhenius plots of the PL intensities of the as-grown and nanopillar samples as a function of temperature from 10 K to 300 K.

Download Full Size | PDF

Time-resolved PL (TRPL) measurements were performed on the as-grown and nanopillar samples at 7 K using a 267-nm mode-locked Ti:sapphire laser (frequency: 76 MHz, pulse width: 150 fs). TRPL signals were collected by a stand streak-camera acquisition system with a resolution of 15 ps. As shown in Fig. 6, the TRPL spectra of these two samples both can be fitted with single exponential curves described by $I\left(t\right)=I_0\exp (-\frac{t}{\tau })$, where $I\left(t\right)$is the PL intensity as a function of time (t), and τ is the photo-excited carrier lifetime. The best fitted τ for as-grown and nanopillar samples are 870 ps and 621 ps, respectively. It is commonly accepted that non-radiative recombination centers freeze up at low temperature [25, 26], so the shorter τ of the nanopillar sample indicates a faster radiative recombination rate, resulting from the improvement in the QCSE caused by the strain relaxation. This result is in accordance with the blue shift observed in RT PL spectra for the nanopillar sample.

 

Fig. 6 Time-resolved PL (TR-PL) spectra of the as-grown and nanopillar sample measured at low temperature (7 K).

Download Full Size | PDF

Micro-Raman measurement was carried out to confirm the strain state in as-grown and nanopillar samples. The Raman spectra are shown in Fig. 7. The spectra are mainly dominated by strong E₂(high), E₂(low), A₁(LO) phonon modes. Usually, the E₂(high) phonon mode is used to represent the stress state in the epitaxial layers, and the in-plane compressive stress σ of epitaxial layers can be expressed by the following equation [27]: ${\omega }_{E_2\left(high\right)}-{\omega }_0=C\sigma$, where C is the biaxial strain coefficient (−2.25 cm⁻¹/GPa for GaN), ${\omega }_{E_2\left(high\right)}$and ${\omega }_0$ are the Raman shifts for ${\omega }_{E_2\left(high\right)}$ mode of the GaN epitaxial layers in our study and the stress-free GaN, respectively. The ${\omega }_0$ value is reported to be 568.0 cm⁻¹ [28]. However, the values of ${\omega }_{E_2\left(high\right)}$ for the as-grown and nanopillar samples are located at 574.3 cm⁻¹ and 570.0 cm⁻¹, respectively. The in-plane compressive stress σ in GaN layers of the as-grown and nanopillar samples can be estimated to be −2.8 GPa and −0.9 GPa, respectively. Therefore, the less in-plane compressive stress for nanopillar sample demonstrates that the strain relaxation has occurred in the epitaxial layers of the nanopillar sample.

 

Fig. 7 Raman spectra of the E₂ (high) phonon mode of the as-grown and nanopillar samples, the inset shows a typical Raman spectra of the as-grown and nanopillar samples.

Download Full Size | PDF

FDTD simulation was performed to investigate the effect of nanopillar structure on the LEE with a simplified 2D simulation model because of the symmetry of the nanopillars. Figures 8 (a) and 8(b) present the 2D distribution of the simulated near-field intensity of light propagation in as-grown and nanopillar LEDs, As shown in Table 1, the top-, bottom- and side-LEE for the nanopillar LED respectively is 2.2, 1.8 and 0.73 times of those for the as-grown LED. Although the side-LEE is decreased, the total-LEE is increased from 22.5% to 33.7% thanks to the greatly enhanced LEE in the vertical direction. The reasons are as follows. First, each nanopillar has an effect of waveguide to vertically confine photons. Second, photons extracted from the nanopillars’ sidewall can also propagate vertically due to Bragg scattering of nanopillars. Third, the etched p-GaN layer decreases the absorbing of photons.

 

Fig. 8 Comparison of FDTD simulations of light propagation in (a) as-grown LED sample and (b) nanopillar LED sample.

Download Full Size | PDF

Table 1. the LEE of different directions by FDTD simulation.

View Table

4. Conclusion

In conclusion, we have fabricated highly uniform AlGaN/GaN nanopillar LED arrays using nanosphere lithography technique. The AlGaN/GaN MQWs in nanopillars emit at a wavelength of 353.2 nm. The strain relaxation due to the formation of nanopillar MQWs results in three main effects compared with the as-grown sample. First, RT PL peak wavelength displays an approximate blue shift of 7 meV. Second, the IQE of nanopillar sample is improved by 42% suggested by temperature-dependent PL measurements. Third, the carrier lifetime at 7 K decreases, and the corresponding radiative recombination rate increases. Additionally, FDTD simulation has demonstrated that more photons are able to propagate vertically in the nanopillar MQW LED, thereby increasing the total LEE from 22.5% to 33.7%.