606-nm InGaN Amber Micro-Light-Emitting Diodes With an On-Wafer External Quantum Efficiency of 0.56%

We demonstrated amber InGaN $47\times 47\,\,\mu \text{m}^{{2}}$ micro-light-emitting diodes ( $\mu $ LEDs) with the peak wavelength of 606 nm and full-width at maximum (FWHM) of 50 nm at 20 A/cm². The amber $\mu $ LEDs exhibited a 33-nm blue-shift of the peak wavelength and obtain broader FWHMs to approximately 56 nm at 5 to 100 A/cm². The peak on-wafer external quantum efficiency was 0.56% at 20 A/cm². The characteristic temperature was 50–80 K at 20 to 60 A/cm² but increased to 120–140 K at 80 to 100 A/cm². The strong increase in the characteristic temperature from 60 to 80 A/cm² could mainly be attributed to the saturation of the Shockley-Read-Hall non-radiative recombination at high current densities.

SECTION I. Introduction

Due to their high brightness levels, long lifetime, large modulation bandwidth, and small form factors, InGaN-based micro-light-emitting diodes ($\mu $ LEDs) have achieved expanding interests in many newly-emerging applications such as micro-displays in wearable and smart electronics, visible light communication, and biomedical sensors [1, 2]. Although InGaN blue and green LEDs have been commercialized owing to their mature technology and reliability, many studies have reported that the external quantum efficiency (EQE) of InGaN $\mu $ LEDs decreases with chip size [3]–​[5]. This size dependence of the EQE was caused by the non-radiative recombination at the edge of the device active region [6]–​[8], which could be eliminated or avoided by using a combination of chemical treatment and atomic-layer deposition sidewall passivation [9] or the bottom-up growth method to form $\mu $ LED mesas [10].

Another important issue to consider is red, green, and blue $\mu $ LEDs for full-color micro-displays. Generally, InGaN LEDs suffer from a significant reduction in EQE as the In content increases in InGaN quantum wells (QWs) [11]. EQE reduction can mainly be attributed to the degradation of the crystal quality for InGaN QWs because of the large lattice mismatch between high In-content InGaN and GaN templates [8]. To reduce lattice mismatch, a partially relaxed InGaN pseudo-substrate fabricated by Soitec was proposed for high In-content InGaN red LEDs [12]. Another lattice-matched InGaN template grown on ScAlMgO₄ (0001) substrates also proved to be capable of remarkably improving the internal quantum efficiency of InGaN QWs, showing the potential of these lattice-matched templates for long-wavelength InGaN emitters [13].

Recently, $6\times 6\,\,\mu \text{m}^{2}$ size 632-nm InGaN $\mu $ LEDs on porous GaN were demonstrated to have an on-wafer EQE of 0.2%, which was the first reported value for red $\mu $ LEDs with the dimension $ < 10~\mu \text{m}$ [14]. Besides, orange/red InGaN LEDs could realize high efficiency on silicon substrates [15] because the tensile strain of the GaN on silicon during growth was favorable for In incorporation [11]. Our group chose to adjust the thickness of the GaN on sapphire substrates to realize highly efficient InGaN red LEDs [16]. Besides, the proposed micro-flow growth method [17] and AlN/AlGaN strain-compensated barriers [18] were also useful for improving the crystal quality of high-In-content InGaN QWs.

For most micro-displays, $\mu $ LEDs need to be driven at low current densities. However, some applications such as head-mounted displays and image sources in projection systems, which need very high brightness levels, require $\mu $ LEDs to be operated at high current densities. Besides, the optogenetic stimulation of chrimson by $\mu $ LEDs also requires a minimal optical power density, which could not be realized at low current densities [19]. Because strong quantum-confined Stark effect (QCSE) in InGaN QWs makes a large blue-shift of the peak wavelength, InGaN long-wavelength-emitting $\mu $ LEDs operated at high current densities need higher In contents compared to $\mu $ LEDs operated at low current densities, which will be much more challenging.

In this work, we demonstrated 606-nm InGaN amber $\mu $ LEDs with dimensions of $47\times 47\,\,\mu \text{m}^{2}$ at 20 A/cm². We examined the current–voltage ($I$ –$V$ ) curve to investigate the leakage current and operating voltage of our $\mu $ LEDs. Optical properties, including the peak wavelength, full-width at half maximum (FWHM), light output power, and on-wafer EQE of the amber $\mu $ LEDs, were characterized by on-wafer testing. We finally measured temperature-dependent electroluminescence (EL) to evaluate the temperature stability of the $\mu $ LEDs.

SECTION II. Experimental Details

Our amber InGaN LED epitaxial wafers were grown on $c$ -plane patterned sapphire substrates via metalorganic vapor-phase epitaxy. The epitaxial structures have been reported in our previous study [16]. The thick GaN template [16] and hybrid InGaN QWs [20] were used to reduce the lattice mismatch of the InGaN amber QWs. The n-Al_0.03Ga_0.97N layer could realize an extremely low resistivity by high Si doping [21]. The schematic structure of InGaN $\mu $ LEDs is shown in Figure 1(a). Indium tin oxide (ITO) was deposited as the transparent conductive layer, and a two-step annealing with and without O₂ gas was done to achieve ohmic contacts with p-GaN [22]. The sheet resistivity of ITO layers could be further improved after the second annealing without O₂ gas. The $\mu $ LED mesa was formed by etching through the ITO layer, InGaN QWs, and InGaN/GaN superlattices (SLs) to the n-type Al_0.03Ga_0.97N layer using inductively-coupled plasma. Before fabricating the n- and p-electrodes (Cr/Pt/Au), a SiO₂ layer was deposited using plasma-enhanced chemical vapor deposition to passivate the $\mu $ LED sidewalls. This SiO₂ layer could also serve as an isolated layer between the p-electrode and n-AlGaN.

Fig. 1.

Schematic structure of InGaN $\mu $ LED devices.

The morphology of $\mu $ LEDs was examined via a scanning electron microscope (SEM). The $\mu $ LEDs were characterized at a probe station using a semiconductor parameter analyzer. The EL properties were measured under different currents at stage temperatures ranging from 295 K (room temperature) to 373 K.

SECTION III. Results and Discussion

Figure 2(a) shows the top-view SEM image of a $47\times 47\,\,\mu \text{m}^{2} \mu $ LED. The n-electrode was around the four sidewalls of the $\mu $ LED, and the p-electrode on ITO was designed as a cross shape. The designs of both the n- and p-electrodes were expected to guarantee uniform current injection into the $\mu $ LED. Because the p-electrode extended from the bottom to the top of the $\mu $ LED, the SiO₂ isolated layer covering the sidewall of the $\mu $ LED was critical to avoid the connection between the p-electrode and the n-AlGaN layer.

Fig. 2.

(a) Top-view SEM image of a $47\times 47\,\,\mu \text{m}^{{2}}$ InGaN $\mu $ LED. (b) Absolute current and current densities of a typical $\mu $ LED at the different bias voltages. (c) Measurement configuration picture of our $\mu $ LEDs. (d) EL spectra of a $\mu $ LED at 5-100 A/cm². The inset is the emission image at 5 A/cm².

The $\vert I\vert $ –$V$ curve was measured under the applied voltage ranging from −4 to 4 V. The absolute current was plotted on a logarithmic scale as shown in Figure 2(b). At a forward voltage below 1 V, the absolute current was at the detection floor. After increasing the forward voltage above 1 V, the current was increased linearly with two different slopes on the semi-logarithmic scale. The first linear part corresponded to the tunneling leakage current [23], while the second linear part was the typical operation region of a p-n junction diode. The transition point between the two linear parts, which was regarded as the turn-on voltage, was around 2.5 V. This $\vert I\vert $ –$V$ behavior was quite similar to other yellow/red InGaN $\mu $ LEDs [6], [14]. The operating voltage at 20 A/cm² was around 3.1 V.

At the reverse voltage, the reverse current also stayed at the detection floor. However, it started to increase after the reverse voltage increased above −3 V. The increment of the reversed currents illustrates that some leakage channels existed in our InGaN $\mu $ LEDs, which might be caused by the defects in the InGaN active region [24] and the Shockley-Read-Hall (SRH) non-radiative recombination at the $\mu $ LED sidewalls [9].

Figure 2(c) shows the EL measurement configuration of $\mu $ LEDs. $\mu $ LEDs were driven at the probe station, and the integrating sphere was located above the sample to collect the light output power of $\mu $ LEDs. Figure 2(d) shows the EL spectra of a $\mu $ LED at 5 to 100 A/cm². The typical single peaks can be observed for all EL spectra. The peak wavelength of the $\mu $ LED was 606 nm at 20 A/cm² (non-uniformity ~ 603–611 nm for 2- inch wafer). No additional blue peaks in this work demonstrated less In fluctuation and phase separation in the $\mu $ LED compared to previous works [16], [25]. The EL emission of the $\mu $ LED at 5 A/cm² in the inset of Fig. 2(d) exhibited non-uniform luminescence due to defects in the amber QWs.

The current density dependence of the peak wavelength and FWHM for the $\mu $ LEDs is shown in Figure 3(a). The peak wavelength of the $\mu $ LED exhibited a total 33-nm blue-shift from 624 to 591 nm at 5 to 100 A/cm². However, the blue-shift behavior was different at low and high current densities. At the current densities < 40 A/cm², the blue-shift of the peak wavelength was as large as 26 nm, which was caused by the strong QCSE and band filling effect in high-In-content InGaN QWs [8], [11], [26]. However, at the current densities above 40 to 100 A/cm², the band filling effect should have been negligible, and the QCSE was also partially compensated. Therefore, the peak wavelength of the $\mu $ LED had a slight blue-shift of 7 nm, as shown in Figure 3(a). Because the QCSE and the band filling effect were respectively originated from the in-plane strain and In fluctuation in InGaN QWs, the strain relaxation and the improvement of the InGaN crystal quality were vital for suppressing this large blue-shift.

Fig. 3.

(a) Peak wavelength and FWHM and (b) EQE and output power of a typical $\mu $ LED at different current densities based on on-wafer testing.

The FWHM at the current density below 20 A/cm² remained 50–51 nm, which was comparable to the best values of other InGaN orange and red LEDs [15], [27]. However, the FWHM increased to around 56 nm with a current density up to 100 A/cm². Our previous work found that the reason was the heat generation in devices under high direct current injection [25].

The output power of the $\mu $ LED increased almost linearly with the current density. At 20 A/cm², we obtained an output power of $5~\mu \text{W}$ at the wavelength of 606 nm. The output power density was calculated as 2.26 mW/mm², which corresponded to the peak on-wafer EQE of 0.56% (Fig. 3(b)).

To estimate the absolute EQE in the integrating sphere, a green $\mu $ LED in the same size was used to obtain the calibration factor for the on-wafer testing and measurement in the integrating sphere. This calibration method was also used in other works [6], [14]. The output power of the green $\mu $ LED (bare chip without resin) measured in the integrating sphere was enhanced by ~2.17 compared to that measured by the on-wafer testing. Therefore, we could expect that the absolute peak EQE of our amber $\mu $ LEDs was estimated to exceed 1.2%. At the current density above 20 A/cm², the on-wafer EQE in Fig. 3(b) exhibited a typical behavior of the efficiency droop, which was calculated around 10% from 20 to 100 A/cm².

The characteristic temperature indicates the temperature stability of $\mu $ LEDs [27]. A larger characteristic temperature implies a weaker temperature dependence. The EL intensity of the $\mu $ LEDs at 20 to 100 A/cm² at different stage temperatures from 295 (RT) to 373 K was normalized by EL intensity at RT and plotted on a logarithmic scale in Figure 4(a). The characteristic temperature could be obtained using the following formula shown below: \begin{equation*} I=I_{T=295 K} \exp \left ({-\frac {T-295 K}{T_{0}}}\right)\tag{1}\end{equation*} where $I$ is the normalized EL intensity, $I_{T=295K}$ is the EL intensity at 295 K, $T$ [K] is the stage temperature, and $T_{0}$ [K] is the characteristic temperature.

Fig. 4.

(a) EL integral intensity of a typical $\mu $ LED at different current densities. (b) Fitted characteristic temperatures at different current densities.

By fitting the experimental data in Figure 4(a), we obtained the characteristic temperatures and their fitting errors at 20 to 100 A/cm² in Figure 4(b). The characteristic temperature was 50–80 K at the current density < 60 A/cm², but increased strongly to 120–140 K at 80 to 100 A/cm². We found that the characteristic temperature of InGaN $\mu $ LEDs was much lower than our standard red LEDs (higher than 300 K) [25], which was caused by more SRH non-radiative recombination at the sidewalls of $\mu $ LEDs compared to standard LEDs.

Furthermore, the SRH non-radiative recombination was also the main reason for the current density dependence of the characteristic temperature for InGaN $\mu $ LEDs [28]. At this point, we believe that the SRH non-radiative recombination included not only the surface recombination at the sidewalls but also the defect-related non-radiative recombination in high-In-content QWs. The SRH non-radiative recombination could be saturated at high current densities. Therefore, the EL intensity of the $\mu $ LEDs at high current densities would be less influenced by the SRH non-radiative recombination and exhibited less thermal droop and higher characteristic temperatures.

SECTION IV. Conclusion

In summary, we demonstrated amber InGaN $47\times47\,\,\mu \text{m}^{2} \mu $ LEDs with a wavelength of 606 nm and an FWHM of 50 nm at 20 A/cm². A large blue-shift of 33 nm for the amber InGaN $\mu $ LEDs was observed at 5 to 100 A/cm². The peak on-wafer EQE was 0.56% (estimated to exceed 1.2% if measured in the integrating sphere) at 20 A/cm², corresponding to a high output power density of 2.26 mW/mm². The characteristic temperature was 50–80 K at 20 to 60 A/cm² but increased to 120–140 K at 80 to 100 A/cm². These higher characteristic temperatures under higher current densities were caused by the saturation of the SRH non-radiative recombination at high current densities.