Optical performance of top-down fabricated InGaN/GaN nanorod light emitting diode arrays

1. Introduction

Improving the performance of Group-III nitride (AlGaInN) based light emitting diodes (LEDs) has been the intense focus of research and development efforts worldwide. While current LEDs are based on planar thin-film architectures, vertically aligned nanorods (also called nanowires or nanocolumns) are currently being explored as an alternative architecture. Several advantages of nanorod-based LEDs have been recently reported. For example, nanorod-based LEDs enhance light extraction due to light scattering, optical mode elimination, and efficient light out-coupling [1]. Strain-relaxed, bottom-up nanorod growth also enables high crystalline quality with significantly reduced threading dislocation densities [2]. Higher indium compositions, which are desired for longer green-red wavelength emission, can be achieved with nanowires because of their compliance properties and strain relief mechanisms [3,4]. Variability in the emission wavelengths across nanowires within an ensemble can lead to phosphor-free “white” LEDs [5,6]. Additionally, suppressed quantum confined Stark effect (QCSE) [7] and reduced droop in InGaN/GaN nanowires [8] have also been reported.

Growth of InGaN/GaN-based nanorod LEDs by bottom-up methods, including hydride vapor phase epitaxy [9] and molecular beam epitaxy [5,8,10], have been demonstrated. However, relatively low growth temperatures and low V to III ratio are commonly used to promote anisotropic one dimensional crystal growth. Metal catalyzed-grown nanowires also require narrow growth conditions which involves lower than optimal growth temperatures [11]. These growth conditions may introduce higher impurities and point defect densities [12,13] than the conditions used for creating commercial-quality planar LEDs and provide less flexibility for adjusting growth parameters to optimize doping concentrations and other desired material properties.

In contrast, nanorods fabricated by top-down methods are etched from planar thin film LED structures grown under optimized growth conditions, obviating these disadvantages. Tapered, non-faceted InGaN-based nanorod LEDs have been previously demonstrated by plasma etching planar LED structures [7,14]. However, the top-down plasma etching leads to damaged, rough, and non-faceted sidewalls with defects, and leakage currents that limits performance [7]. Here we demonstrate a top-down strategy for creating nanorod LEDs from planar LED wafers using a two-step process that adds a selective anisotropic wet etch after the initial plasma etch to remove the dry etch damage while enabling nanorods with straight and smooth faceted sidewalls and controllable diameters independent of pitch. The nanorod LEDs created by this two-step process show potential for enhancing the internal quantum efficiency (IQE) of InGaN/GaN multi-quantum wells (MQWs).

2. Experimental procedures

For the nanorod LED fabrication, a prototypical InGaN/GaN MQW planar LED structures were first grown on c-plane sapphire in a Veeco D-125 metal organic chemical vapor deposition reactor. The MQW structure consists of a 2.5 µm thick Si-doped n-type GaN layer grown at 1050°C followed by a 5-period MQW comprised of 2.5 nm thick In_0.13Ga_0.87N wells and 7.2 nm thick GaN barriers grown at 770 °C. In addition, a 22 nm thick p-Al_0.2Ga_0.8N layer and a 200 nm thick p-type GaN contact layer were grown sequentially after the MQWs. After the growth, a close-packed monolayer of 1 µm diameter silica spheres was then self-assembled on the GaN surface in a Langmuir-Blodgett trough as previously reported [15]. The silica colloid monolayer functions as a mask in subsequent inductively coupled plasma etches. Previously plasma etching has been used to create GaN nanorods [16] and GaN-based LED nanorods [7,14] using various etch masks. In this work we follow the plasma etch step with a selective KOH-based wet etch (AZ400K photoresist developer, AZ Electronic Materials USA Corp) [17,18]. With this etchant, the top Ga-polar c-plane surface has a near-zero etch rate while the {10-10} m-plane sidewalls have a relatively fast etch rate compared to the other planes. As a result, the height of nanorod LEDs remains constant and the tapered sidewalls are eventually replaced with six straight and smooth m-plane sidewalls. The nanorod LED array and the original planar LED control sample were characterized by scanning electron microscopy (SEM), x-ray diffraction (XRD), transmission electron microscopy (TEM), and temperature (4K-298K) and pump power dependent photoluminescence (PL).

3. Results and discussion

Figure 1(a) shows a SEM image of a planar LED structure covered with a hexagonal close-packed monolayer of silica spheres. After plasma etching, truncated cone-shaped nanorods are formed (Fig. 1(b)). After, the anisotropic wet etch using AZ400K, nanorods with straight and smooth sidewalls are produced, as seen in Fig. 1(c). The n-type GaN etches more quickly than the p-type GaN, leading to “flashlight” or “golf-tee” shaped nanorod LEDs, as shown in Figs. 1(c) and 1(d). The mechanism responsible for the faster n-type etch rate is not currently known. Previously, p-type GaN has been reported to be resistant to photoelectrochemical etching, as photogenerated holes needed for oxidation of the surface are swept away from the depletion region near the surface into the bulk [19]. While our etch process is neither photo- nor electrically assisted, it is possible that the presence of holes at the surface or the different surface potential between n-type and p-type GaN [20] is responsible for the difference in wet etch rates observed here.

 

Fig. 1 SEM images of (a) a planar LED wafer covered with a monolayer of self-assembled silica spheres, (b) tapered nanorod LEDs created by plasma etch, and (c) “flashlight” shaped nanorod LEDs array following wet etch. (d) A STEM image of nanorod “flashlight” LEDs showing the position of the InGaN MQWs (bright stripes).

Download Full Size | PPT Slide | PDF

In addition to allowing for control of the nanowire geometry, the wet etch step also serves to remove damage in the nanorod sidewalls caused by high energy ion bombardment during the plasma etch. This ion damage was reported to cause a Ga-rich surface in GaN films [21] with efficient nonradiative recombination centers [22] and a deterioration in PL intensity. Moreover, vacancies may form near the surface region and trap or scatter carriers leading to high static resistance and reduced carrier mobility [23]. Room-temperature PL (266 nm pump) showed the appearance of strong defect-related yellow luminescence from GaN nanorods etched from a GaN epilayer after the plasma etch, as shown in Fig. 2 . However, after subsequent wet etching the nanorod samples in AZ400K, the yellow luminescence intensity was significantly reduced to a level similar to that of the planar GaN thin film prior to the plasma etch. This result unambiguously shows the wet etch step successfully removes the damage caused by the plasma etch. The slight redshift of the band edge luminescence in the nanorod samples compared to the planar sample may be a result of laser-induced heating effects caused by poorer thermal transport properties [24]. We note that drawing conclusions based on comparing the absolute band edge and yellow intensities of the samples is difficult, as various factors can impact the PL intensities. For example, the nanorod samples likely experience enhanced pump laser absorption and higher light extraction efficiencies, but have only a small fraction of the original volume of the planar sample. However, the relative contribution of each of these factors cannot be deconvolved.

 

Fig. 2 PL spectra of GaN thin film (solid black) and GaN nanorods after plasma etch (dashed-dotted red) and subsequent wet etch (dotted blue).

Download Full Size | PPT Slide | PDF

Bright-field TEM imaging was performed to study the dislocation morphology of the nanorods under multi-beam conditions to reveal dislocations with different Burger’s vectors. Typical TEM images are shown in Fig. 3 . TEM imaging of 100 randomly sampled nanorods reveals that 94 nanorods are free of threading dislocations. The threading dislocation density in the initial planar LEDs epilayers before etching was measured to be ~5 dislocations/µm² (5 × 10⁸ cm⁻²), which is typical for commercial quality LEDs. Thus, for nanorod diameters ~150 nm (with cross-sectional area ~0.02 µm⁻²), the average number of dislocations per nanorod is ~0.1. In reality, each nanorod either has one or more dislocations or is dislocation free. Therefore, unlike bulk LEDs, which have a distribution of dislocations throughout, a nanorod LED array is comprised of individual LED elements which are largely dislocation-free. This nearly dislocation-free nanorod architecture could, compared to planar LEDs, have lower leakage currents, less non-radiative recombination, higher IQE, and higher lifespans. This result also shows that nanorods formed by top-down etching can be nearly dislocation free even though they are not grown strain-relaxed like bottom-up nanowires.

 

Fig. 3 TEM images of etched nanorod LED structures. A dislocation is indicated by the arrow in (a).

Download Full Size | PPT Slide | PDF

The room temperature PL spectra of the nanorod LED InGaN MQWs excited by a 413.1 nm Krypton ion laser are compared to the planar LED in Figs. 4(a) and 4(b). The PL spectra peak positions were determined using Gaussian fit and plotted as a function of pumping power in Fig. 4(c). For both structures there is a blue shift in the emission wavelength with pump power (carrier injection) caused by free carrier screening of the QCSE. The QCSE results from spontaneous polarization field caused by low symmetry of c-plane III-nitride crystal structure and piezoelectric polarization field caused by lattice mismatch strain at InGaN/GaN interfaces. These internal fields may lower the IQE by reducing carrier wavefunction overlap. By comparing (0002) ω/2θ XRD scans taken before and after nanorod etching and by applying an appropriate elasticity theory analysis, we estimate that the coherency strain of the individual InGaN QWs was reduced by 16 ± 4% because of the elastic response enabled by nanorod formation. A further comparison of our planar sample’s emission at low power (445 nm) to the bulk bandgap [25] of In_0.13Ga_0.87N (426 nm) indicates a QW Stark shift in the planar sample of ~19 nm. Assuming that the change in strain linearly alters the piezoelectrically dominated Stark shift, we estimate a ~3.0 nm blue shift (0.16 x 19 nm) for the nanorod QW emission relative to that of the planar sample. In Fig. 4(c), we see a measured blue shift of ~4 nm (from about 445 to 441 nm for the planar vs. nanorod sample) at low power, in approximate agreement with our estimates and in support of partial QW strain relaxation due to wire formation. A smaller maximum blue shift of ~2 nm in the nanorod InGaN peak position as the pumping power is increased from 0 to 100 mW compared to the 6 nm shift observed in the planar LED is also evidence for reduced QCSE in the nanorod LEDs, consistent with a previous report [7].

 

Fig. 4 MQW PL of (a) nanorod LEDs and (b) planar LED. (c) MQW PL peak position plotted as a function of pump power for planar LEDs and nanorod LEDs.

Download Full Size | PPT Slide | PDF

Due to the reduction in the piezoelectric polarization fields and relative lack of dislocations, the internal quantum efficiency of the nanorod LEDs might be expected to improve compared to the planar LED. Thus, the IQE (Fig. 5 ) of the nanorod and planar LEDs were measured by comparing their integrated PL intensity at room temperature with the maximum intensity measured at 4K and assuming unity IQE at 4K. Due to increased light scattering and light absorption, the nanorod LEDs likely experience a much higher equivalent pumping intensity (carrier concentration) at a given laser power compared to the planar sample, leading to the early peaking of the IQE. The IQE of the nanorod LEDs peaks at ~3 mW, (~24%), while the planar LED reaches its maximum IQE of ~27%, at a much higher pump power of ~55 mW. The peak IQEs are comparable so the formation of the nanowires did not compromise the IQE in an overly detrimental way. Additionally, it is possible that the decrease in IQE in the nanorod LEDs at relatively low pump powers may be caused by heating from the laser, due to the poor thermal transport properties of nanowires. This is supported by the red shift at higher pump powers observed in Fig. 4, consistent with a previous report where laser-induced produced redshifts in GaN nanowires during PL experiments [24].

 

Fig. 5 The IQE of the planar and nanorod LEDs plotted as a function of the optical pumping power.

Download Full Size | PPT Slide | PDF

Despite the reduced piezoelectric polarization and their mostly dislocation-free nature, the measured InGaN MQW IQEs of the nanorod and planar LEDs are not significantly different. It is possible that the benefit of fewer dislocations and reduced piezoelectric polarization field is counterbalanced by heating effects or the higher surface to volume ratio, which could promote nonradiative carrier recombination [26] and space-charge-limited carrier transport [27]. Future surface passivation experiments, for example with an AlGaN shell [28,29], may shed further light into the role of surface states in these wet-etched structures.

4. Conclusion

In conclusion, we have developed a two-step top-down process for creating nanorod LED arrays from planar LED epitaxial structures. In contrast to previous work, this top-down technique combines the plasma etch with a subsequent anisotropic wet etch, which allows for removal of the plasma etch damage, while enabling straight nanorod sidewalls and control over the nanorod geometry. Despite the fact that the planar LEDs have a dislocation density of 5 µm⁻², 94% of the nanorod LED structures are free of dislocations. The nanorod LEDs also show reduced QCSE compared to the planar LEDs due to reduced piezoelectric strain. The nanorod and planar LEDs were measured to have similar IQEs, perhaps due to heating effects and enhanced nonradiative recombination at the nanorod surfaces. These results suggest that these top-down nanorod LEDs have potential for higher-performance LEDs, although further work to investigate the role of thermal and surface effects is needed.