High performance InGaN/GaN nanorod light emitting diode arrays fabricated by nanosphere lithography and chemical mechanical polishing processes

1. Introduction

The synthesis of nanowires or nanocolumns has attracted much attention in recent years as it brings huge impact to the research and applications in electronics, biology and optoelectronics. The exploration on nanoscale device properties not only helps the research of the fundamental physics but also improves the device performance. For example, nanostructure LEDs have the advantage of increasing surface area from the sidewalls of nanorods, which may have the chance to improve optical output. Furthermore, the availability of nanostructure provides an excellent tool to study carrier transport and material strain that limit the performance of light sources, especially in the GaN based devices. The synthesis of nanowire or nanocolumn structure is generally divided into bottom-up and top-down approaches. In the bottom-up method, the vapor-liquid-solid (VLS) growth has been widely employed [1–3], which metal nanoparticles such as Fe, Au and Ni are used as the catalysts during the growth procedure. While in the top-down approach, the process simply uses nano-scale etching masks and followed by dry or wet etching to form nanorod structure. E-beam lithography, for example, is carried out to achieve high resolution and well defined nano-patterns. However, the disadvantages of low-throughput and high-cost make e-beam lithography unsuitable for commercial device applications. Alternatively, self-assembled metal nanomasks or self-organized nanoparticles provide high-throughput yet less organized nanopatterns [4–6].

Despite numerous research works on nanowire/nanorod synthesis, only a limited number of groups reported fabrication and characterizations on the nanorod LED arrays [5–10]. The critical issue of nanorod LED fabrication is to prevent shorting the p- and n-type semiconductors when depositing the top metal contact. In addition, for most nanorod arrays, the leakage current is inevitable and the device fails before light emission as the junction temperature becomes too high. There have been several solutions proposed, including the insertion of spin-on glass (SOG) [5] or polymer (such as SU-8) as the space layer [6], the usage of oblique indium tin oxide (ITO) deposition [7], and the photo-enhanced chemical (PEC) wet oxidation process of GaN nanorod sidewalls [8] to achieve nanorod devices. Some reported the reduction of the reverse bias current to μA range [5,8,10]. The optical power of GaN based nanorod arrays is as high as 3700mW/cm² at the injection current of 20mA (corresponding to the current density of 22.22A/cm²) [7]. Despite the report of light emission from nanorod arrays, a reliable manufacturing process with high production yield and excellent performance isn’t available. The current-voltage curves of such nanodevices suggest that they suffer from large leakage currents, large ideality factors, and low optical output power, as compared with conventional planar GaN based LEDs.

In this work, the technology of nanosphere lithography was applied to fabricate nanorod arrays. By spin-coating a monolayer of nanospheres on top of the GaN based LED epi-structure, and followed by semiconductor etching, the InGaN/GaN nanorod structure was realized. With PECVD grown SiO₂ sidewall passivation, the leakage current can be significantly reduced. Furthermore, we performed a chemical mechanical polishing (CMP) technique to achieve uniform exposure of p-type GaN nanorod tips on the sample. The electrical and optical properties were characterized. We demonstrated a naorod LED array with high optical output power and a low ideality factor.

2. Fabrication

The nanorod LED fabrication started from preparing a multi quantum well (MQW) InGaN/GaN LED epi-structure. The GaN based LED sample is grown by metal organic chemical vapor deposition (MOCVD) on a c-plane sapphire substrate. The material structure is composed of a 25nm GaN buffer layer, a 2μm Si doped n-type GaN layer, a five period of In_xGa_1-xN/GaN multiple quantum well (MQW) structure in which each period is 17nm, and a 160nm Mg doped p-type GaN layer [see Fig. 1(a) ]. The average composition of indium (In) x is around 0.2 with the quantum well thickness of 3nm, which results in a photoluminescent peak wavelength at 467nm. The nanorod fabrication started from the definition of a 250x250 μm² mesa pattern and followed by spin-coating a monolayer of silica nanoparticles with the diameter 100 ± 10nm. Using the silica nanoparticles as the etch mask, the nanorods were then exposed on the sample after the subsequent ICP-RIE (inductively coupled plasma reactive ion etching) etching [Fig. 1(b)]. The nanorod array is shown from the SEM (scanning electron microscope) image in Fig. 2(a) . The etching depth is around 290nm. In the next step, a 100nm-thick SiO₂ insulating layer was blanket coated on the sample using PECVD, which prevents shorting the p-type and n-type semiconductor and meanwhile passivates defects created during the nanorod etching [Fig. 1(c)]. Next, CMP process employing Al₂O₃ particle slurry with the diameter around 800nm was applied to remove the SiO₂ deposited right on the tips of nanorods [Fig. 1(d)]. Since the surface microhardness of GaN is around 1200-1700 kg/mm² [11], while that of SiO₂ is 790kg/mm² [12], the difference ensures that the SiO₂ layer right on top of GaN nanorods can be easily removed without damaging the nanorod semiconductor material. The SEM image in Fig. 2(b) suggests that the tips of nanorods are exposed after the CMP process. In the last step, the via holes were opened by RIE (reactive ion-etching) to enable the deposition of the n-type [Au/Ti (120nm/10nm)] contact electrodes. Finally a thin metal stack (Au/Ni 5nm/5nm) was coated for current spreading (p-thin layer) and the p-type contact electrode (Au/Ti 120nm/10nm) was also deposited [Fig. 1(e)].

 

Fig. 1 The process flow of nanorod LED array fabrication. (a) The device employs a typical MQW LED (b) The nanorod array is realized by nanosphere lithography (c) The PECVD grown SiO₂ layer was deposited to prevent p-n shorting and to passivate the sidewalls (d) The CMP process to expose the nanorod tips (e) Deposition of contact pads.

Download Full Size | PPT Slide | PDF

 

Fig. 2 SEM images of (a) the nanorod structure after ICP etching and (b) the surface profile after SiO₂ passivation and CMP process.

Download Full Size | PPT Slide | PDF

3. Results & discussion

The current-voltage curves were extracted from the nanorod device using an Agilent 4155c semiconductor parameter analyzer. As demonstrated in Fig. 3 , the device shows a well-rectified diode behavior with a current ratio 3983 at ± 5V. And the 4.77nA reverse current at −5V is significantly lower than that of previous reported approaches [5–10]. Furthermore, the ideality factor is calculated to be 7.35. Sah-Noyce-Shockley theory [13] suggests ideality factors between 1 and 2 for typical diodes due to the competition between the drift-diffusion and generation-recombination processes. However, for GaN based diodes, the ideality factors usually fall in the range between 5 and 7 [14,15], which results from the large p-type contact resistance and the polarization-induced triangular band profiles of the quantum barriers [16]. As for nanorod diodes, the formation process using the top-down etching may damage the rod sidewalls and creates a current leakage path; thus larger ideality factors such as 11.2 ± 0.56 in [17] and 18 in [18] were reported. The sidewall passivation and the subsequent CMP process in our work can significantly reduce the sidewall leakage as well as prevent the shorting paths along the nanorods. Furthermore, it was suggested from [19–21] that the formation of nanorods can reduce the strain in the InGaN/GaN quantum well structure and thus mitigate the influence of polarization induced band profile adjustment on the ideality factor.

 

Fig. 3 The I-V curve of a nanorod LED. The dash line indicates an ideality factor of 7.35

Download Full Size | PPT Slide | PDF

The luminescence-current (L-I) behavior of the nanorod LED array was measured at room temperature. Both the optical power (on the right axis) and injection current are normalized to the p-type mesa area (250x250μm²). As demonstrated in Fig. 4 , at an injection current density of 32A/cm², the corresponding optical intensity is 6807mW/cm². The image of blue light emission from the nanorod LED array at 1mA (1.6A/cm²) is shown in the inset of Fig. 4. The output power of our device is the highest among similar devices reported so far. The external quantum efficiency (EQE), defined as the ratio between the photons extracted to free space and carriers injected to the device (assuming the same area size as the p-mesa) [22], is demonstrated in the right axis of Fig. 4. The EQE is above 8% under the injection current range and reaches its maximum of 16.61% at 1.6A/cm². For such an EQE curve, a so-called efficiency droop is observed. The EQE suffers from a rapid decrease at the current density between 1.6A/cm² and 8 A/cm² and then a smaller decreasing slope. Such an interesting phenomenon leads us to study the droop effect in the nanorod LED array. Typically, for planar GaN based LED structure, the root causes of efficiency droop may be attributed to various device phenomena such as the Auger recombination, junction heating, and polarization induced carrier overflow [22–25]. For our device, to understand the effect of junction temperature, EQE of nanorod device was extracted at both DC (100% duty cycle) and pulsed (1% duty cycle with the cycle period of 50ms) currents. The normalized EQE vs. injection current for both bias conditions are plotted in Fig. 5 . Basically, the heating effect can be mitigated at the 1% duty cycle. The EQE curve at 1% duty cycle shows a nearly constant decreasing trend (with the slope −6.45x10⁻³ (A/cm²)⁻¹), which is close to that of the 100% duty cycle case at the high current level (slope = −7.69 x10⁻³ (A/cm²)⁻¹)). As we compared both cases, the rapid EQE decrease between 1.6 and 8A/cm² (slope = −8.3 x10⁻² (A/cm²)⁻¹)) in Fig. 4 is mainly associated with the thermal effect, while the decrease beyond 8A/cm² is related to other factors.

 

Fig. 4 Optical output power density (left axis) and the corresponding EQE (right axis) of the nanorod LED array at different injection current densities. (Inset) Light emission image from the nanorod LED array at 1.6 A/cm².

Download Full Size | PPT Slide | PDF

 

Fig. 5 Normalized EQE at (a) 100% and (b)1% duty cycles

Download Full Size | PPT Slide | PDF

Moreover, the far field radiation profile of nanorod LED array is demonstrated in Fig. 6 . The angle at half maximum intensity, θ_1/2, occurs at 30°, suggesting that the far field pattern is a Lambertian profile. Since the Lambertian emission refers to a source emitting the same quality of light in all directions, it indicates that the nanorod LED array is mainly consisted of point light sources with excellent uniformity.

 

Fig. 6 The radiation profile of a nanorod LED array. The red line indicates that θ_1/2 of the nanorod LED is 30°.

Download Full Size | PPT Slide | PDF

4. Conclusion

Nanorod LED array devices were fabricated using the technology of nanosphere lithography. By spin-coating a monolayer of silica nanospheres on top of the GaN based LED epi-structure, and followed by semiconductor etching, the nanorod structure was realized. The nanorod sidewalls were passivated by PECVD grown SiO₂ and the p-type GaN tips were exposed by chemical mechanical polishing. We achieve a reverse current of 4.77nA at −5V, an ideality factor of 7.35, and an optical output intensity of 6807mW/cm² at the injection current density of 32A/cm² with very high nanorod light emitting uniformity. Moreover, the study of droop effect for such a nanorod LED array reveals that junction heating is responsible for the sharp decrease at the low current between 1.6A/cm² and 8 A/cm². Finally, the Lambertian radiation profile indicates that the nanorod LED array is mainly consisted of point light sources.