1. Introduction

In full-color micro-LED displays fabricated by combining blue, green, and red micro-LED chips, InGaN with varying indium content is used for blue and green, while AlGaInP is widely adopted for red. Compared to the blue micro-LEDs, the green micro-LEDs have higher indium content in multiple quantum wells (MQWs), causing large amount of material strain, an increase of defect density, piezoelectric fields, and morphological breakdown in the form of v-pits, and thus resulting in a significant efficiency loss of the micro-LEDs in the green spectral range, which is the so-called “green gap” [1,2]. Therefore, the fabrication of highly efficient green micro-LEDs is challenging [3–8]. In particular, the peak EQE of green micro-LEDs did not exceed 10% until now.

Recently, green micro-LEDs with a size of 25–200 µm showed high brightness of 38 – 68K nits at 1A/cm², which were the record-high brightness for green micro-LEDs, having a 5 nm Ni/Au layer as the current spreading layer (CSL) [5]. However, such a thin Ni/Au layer has a major drawback of low transmittance (∼70%) in the green region due to optical absorption and the issue of contact reliability [9]. As a CSL, several studies showed that the ITO provides more uniform current spreading in an LED chip due to its higher transparency (> 80%) in the visible spectrum, and higher conductivity (resistivity of ∼10⁻⁴ Ω·cm) [9–12]. In addition, due to the high refractive index of ITO (1.8–2.0), the ITO layer can reduce the total internal reflection losses of the generated light in the active layer, resulting in higher light extraction efficiency.

In this work, we present the electrical characterizations and optical properties of InGaN-based green micro-LEDs with three different sizes of 80×80 µm², 150×150 µm², and 200×200 µm². The optimized fabrication processes and the ITO as a CSL provide a peak EQE of 13.7%, a peak WPE of ∼ 11%, and a high brightness of over 10M nits, ranking among the highest performing green InGaN micro-LEDs reported to date. This could originate from the more uniform current spreading in the active layer and higher transparency due to the ITO.

2. Materials and methods

Green micro-LEDs devices with a size of 80×80 µm², 150×150 µm², and 200×200 µm² are designed for systematic electro-optical investigation. These green micro-LEDs are fabricated with commercial epi-wafers, which have ten periods of MQWs on a sapphire substrate. The cross-section diagram of such micro-LED is shown in Fig. 1(a) illustration.

 

Fig. 1. Size-dependent current density–voltage (J-V) characteristics. Figure 1(a) is forward current for different size in a semi-log scale, the insert is schematic diagram of micro-LED structure, and Fig. 1(b) is reverse leakage current for different size in a linear scale.

Download Full Size | PDF

The 300 nm SiO₂ layer is deposited on the wafer by plasma enhanced chemical vapor deposition (PECVD) as a hard mask for patterning. Hexamethyldisilazane (HMDS) is deposited using spin coating before coating the photoresist and the wafer is subjected to a 1-minute bake at 100°C (soft bake) to increase photoresist adhesion on the SiO₂. After photolithography, inductively coupled plasma (ICP) etching is used to form mesa on wafer. Then the ITO layer (100nm) is deposited on the surface of p-GaN as a CSL. Rapid thermal annealing is performed to achieve Ohmic contact between p-GaN and CSL. Electron beam evaporation deposition is used to deposit the p and n electrodes of Ti/Al/Ti/Au with 25 nm/100 nm/25 nm/50 nm. Among them, the bottom Ti layer is used to enhance the adhesion between CSL and Al layer. The thick Al layer can improve the current horizontal uniformity and increase the reflectivity, ensuring that light can escape from the bottom. The Ti layer between the Al layer and the Au layer is used to prevent the diffusion of aluminum and gold. The Au layer is adopted to reduce resistance and to avoid the electrode being oxidized. It should be noted that micro-LEDs have no sidewall passivation layer effectively reducing sidewall defects. All devices are on the wafer, without packaging. The emission measurements are performed from the top sides of the LEDs (electrode side). For EQE measurement, Al wire bonding with electrode is used after shearing and cutting without packaging, and integrating sphere spectroradiometer system is used. The current-voltage characteristics were analyzed by using Keysight (formerly Agilent) B1500 Analyzer.

3. Result and discussion

The current density (J) against the voltage (V) is compared in Fig. 1(a) in a semi-log scale while Fig. 1(b) is reverse leakage current for different sizes in a linear scale. The leakage currents revealed under low forward bias and under reverse bias are similar, as shown in Fig. 1(b). It is shown that the current density increases with decreasing chip size and increasing reverse bias. The leakage currents can be attributed to the defects at the sidewall due to the damage caused by the etching [13]. In addition, the reverse leakage current density of the 80µm device is about 300 nA/cm² at −2 V, indicating the leakage characteristics of the micro-LEDs with passivation layers [14]. In Fig. 1(a), the forward voltages (V_f) measured at 10 mA/cm² are equal to about 1.5 V, which is lower than V_f of InGaN LEDs with Ni/Au CSL (1.7 V) [5]. This can be attributed to the fact that the ITO layer leads to an Ohmic contact by forming a p-n junction at the interface of ITO/p-GaN and induces the carrier transport by tunneling from the ITO to p-GaN, resulting in the improvement of V_f [15]. Note, V_f is the voltage when the LED has forward-conducting current, before the LED emits light. When applying a higher forward bias voltage, the current no longer increases exponentially with the increase of voltage. The slope of the J-V curve is mainly affected by the series resistance (R_s), coming from the current crowding effect [16]. At the same voltage, the smaller the device size, the higher the current density, indicating better current uniformity for small size devices [17–21]. The small size contributes to reducing the current crowding effect, and low current crowding could reduce heat.

The optical power-voltage plot in Fig. 2(a) shows the turn-on voltage (V_on) of the 80µm device is 2.13 V. We define the V_on as the voltage when the light output power occurs. Interestingly, V_on decreases to 2.06 V with the increase in size. According to qV_on= E_g, V_on is related to the semiconductor band gap (E_g). The electroluminescence (EL) spectrum of LED is dominated by spontaneous emission. Thus, it is expected to observe the red shift of EL spectrum with the increase in size. To verify our assumption, the EL spectra of the three devices are measured at the current density of 20 A/cm², displayed in the Fig. 2(b). As we expected, it exhibits clearly that the peak of the EL spectrum shifts to lower energy with increasing size. Two mechanisms can be considered: 1) strain relaxation and 2) thermal effect. The relaxed strain reduces the piezoelectric field inside the quant well so that the spectrum shows a blue shift of the emission wavelength [22]. The smaller size micro-LEDs have a larger area affected by strain relaxation [23]. Therefore, the EL spectrum has a redshift with increasing size. In addition, thermal effects also affect the emission wavelength. Current crowding effect causes heat accumulation in LEDs. The size-reduction of LED leads to improved current spreading on the active layers of LED, alleviating the current crowding, and reducing heat generation [18,19,21]. Yang et al. reported the band gap shrinkage of the active layer in the green micro-LEDs with the increase of temperature, showing a red shift in wavelength [24]. Thus, it is speculated that the heat dissipation capacity of green micro-LEDs increases, and the operating temperature decrease with the decrease in size.

 

Fig. 2. (a) Light output power with voltage. (b) Normalized EL spectrum at 20 A/cm² for difference size device. The inset is an enlarged view at the peak energy.

Download Full Size | PDF

Figure 3(a) shows the EQE and the calculated internal quantum efficiency (IQE) of our devices. The peak EQEs are 13.7% (at the current density of 28.1 A/cm²) for the 80µm device, 13.3% (25.3 A/cm²) for the 150µm device, and 12.2% (23.8 A/cm²) for the 200µm device. The smaller devices have higher EQEs than the larger devices but also have less drop. This could be attributed to large size micro-LEDs having more obvious current crowding effect [17,19,20]. And an aggravated current crowding effect contributes to a decreased IQE, LEE and deteriorated EQE droop [25]. The peak EQE shift to lower current density with the increase in size. This could be attributed to the less non-radiative recombination centers in bigger-size devices, based on the ABC model [26]. The IQEs were extracted from the EQE values by using the room-temperature reference-point method (RTRM) [27], and the result is shown in the inset of Fig. 3(a). The calculated IQEs show a similar trend as the EQEs, as expected. However, surprisingly, the calculated peak IQE is between 70% and 80%, strongly indicating that our devices are well optimized, and the current crowding effect is significantly suppressed due to the ITO layer. The current densities at the peak EQE value in our devices are higher than Liu’s report (∼3 A/cm²) [5]. In the frame of ABC model [26], it can be interpretated as the existence of more non-radiative recombination centers due to the sidewall effect. However, the 13.7% peak EQE for the 80µm device provides the best EQE in green micro-LED in our knowledge [4,6,28,29]. Thus, the suppressed current crowding using ITO is the key factor for the good performance of our devices.

 

Fig. 3. Optical properties of L = 80 µm, 150 µm, and 200 µm device. (a) WPE and EQE. IQEs are in inset, (b) Luminance.

Download Full Size | PDF

It was reported that for current (I) ≤ 1 mA, the variation in emission intensity of different pixels in micro-LED displays is negligible [30]. Thus, the working current density of micro-LEDs for display applications is in the range of 0.02 to 2 A/cm² [31]. The EQEs of our devices are still over 5% at 1 A/cm², which are the typical EQEs in other reports about green micro-LEDs [6,28]. It is worthwhile to note that the devices in other reports have a passivation layer and their size is much smaller than our devices [5,7,32,33]. However, the devices with a passivation layer show the experimental trend of higher EQE than devices without a passivation layer and no size-dependency [34,35]. These factors strongly support the superior performance of our devices.

Energy efficiency reducing electricity demand in electrical products is one of the key factors to fight against climate change. Besides, the challenge of micro-LEDs is to maintain high WPE at a much lower current density than the current density for the peak EQE. The WPE of conventional LEDs can be as high as 60%, and the current density at peak WPE value is generally 2 A/cm² [36–38]. However, as device size decreases, the peak WPE is achieved at higher current densities, producing a higher operating voltage. Interestingly, there is no experimental report for the WPE of micro-LEDs until now, although it is very important as an indicator of energy-efficient devices. A recent simulation study expected that the WPE is less than 30% for the micro-LED with a size of less than 100×100 µm² [39], which could be the theoretical maximum value after optimizing the device shape and structure. The obtained WPEs of our devices are exhibited in Fig. 3(a). The peak WPE are 11.3% (at the current density of 20 A/cm²) for the 80µm device, 11.6% (18 A/cm²) for the 150µm device, and 11.0% (14 A/cm²) for the 200µm device. The current densities at peak WPEs are clearly less than those at the peak EQEs, although they are about 10 times larger than the conventional LEDs. Importantly, our devices provide the relatively high WPE though the chip shape and the structure are not optimized for enhanced the WPE.

The EQEs of our devices are record-breaking high, resulting in very high brightness green micro-LEDs, i.e., very high level of light output. The relationship with brightness and current density is exhibited in Fig. 3(b). Smaller-sized devices have higher brightness, which shares the same trend as EQE. It can be attributable to the improvement of current spreading [40]. The luminance of the 80 µm, 150 µm, and 200 µm devices are 6.0M nits, 5.5M nits, and 4.9M nits, at 10 A/cm², respectively. They are about 10 times higher than Liu’s report (∼ 0.5M nit at 10 A/cm²) [5]. Particularly, the brightness is over 10M nits from 20 A/cm² for all devices. Even at 1 A/cm², our devices provide over 200K nits. In the display industry, the required brightness for indoor and outdoor displays are ∼200 nits and over 5K nits, respectively. Our green micro-LEDs without passivation layer provide more than 40 times higher brightness than which the display industry requires.

4. Conclusion

In our study, it is found that the ITO provides excellent uniform current spreading, leading to a peak EQE of ∼13%, a peak WPE of ∼11%, and record-high luminance for InGaN-based green micro-LEDs. The better current spreading in smaller device results in higher EQE and brightness. However, the current densities at the peak EQE and the peak WPE of our devices are still much higher compared to conventional LEDs or other green micro-LEDs. For energy-saving displays, the current density of the peak EQE/WPE is desired to be in the range of 0.01 - 1 A/cm². The superior performance could be attributed to the combination of the optimized process in each fabrication step, the use of ITO, and the good quality of the commercial epi-wafer though more systematic studies are required to understand. Our devices would exhibit the practical maximum performance among non-passivated devices. The passivation of the active layer and better current spreading using a mixture layer of ITO/metal help to improve the EQE and the WPE at the lower current densities [38,41]. Furthermore, optimized chip shape and design could help to enhance the WPE.