1. Introduction

AlGaN-based deep ultraviolet light-emitting diodes (DUV-LEDs) have attracted considerable attention for day-to-day applications [1,2], such as sterilization, water purification, and curing skin conditions, owing to their compact size and adjustable wavelength. The wavelength of DUV-LEDs can be tuned by varying the Al composition of Al_xGa_1-xN materials. DUV-LEDs have a high potential for replacing traditional mercury lamps [3]. Nevertheless, the external quantum efficiency (EQE) of DUV-LEDs was lower than 5% [4], which is significantly lower than that of InGaN-based blue LEDs (>70%) [5]. Moreover, the maximum wall-plug efficiency (WPE) of commercially available DUV-LEDs was usually lower than 2% for the regular size of DUV-LED device such as, 10 mil ×10 mil, to 50 mil ×50 mil. The chip size of DUV-LEDs needs to reduce in order to increase the maximum WPE due to the alleviation of the current crowding effect and the maximum WPE would be higher than 3% using the chip size of 10 µm × 10 µm [6–7]. In addition, the p-AlGaN with high Al content (50–70%) was used for the transparent p-contact layer of DUV-LEDs to reduce UV-light absorption and increase the light extraction efficiency (LEE) and EQE [8]. However, the low hole concentration of the p-AlGaN epilayer resulted in a high contact resistance of the p-electrode. This increased the operating voltage of DUV-LEDs and further reduced the WPE of DUV-LEDs. In general, the p-GaN has been used as the p-contact layer in most studies to reduce the contact resistance of the p-electrode and operation voltage of DUV-LEDs. In addition, indium tin oxide (ITO) was used as the current spreading layer. However, the p-GaN contact layer and ITO current spreading layer strongly absorb light at DUV wavelengths owing to the material limitation. To overcome the abovementioned problems, a thin p-GaN contact layer and ITO have been designed in the structures of DUV-LEDs [9].

In general, the LEE of DUV-LEDs is enhanced by the optimized epitaxial structure and improved LED structure design, such as reflective p-contact electrodes for AlGaN-based DUV-LEDs. Because the DUV light absorption in the p-GaN contact layer causes strong absorption loss, the reflective p-contact electrodes with high reflectance in the DUV wavelength region were limited. Only aluminum (Al) has a reflectance higher than 80% in the DUV wavelength range [10]. The Al cannot form an ohmic contact with p-GaN owing to the lower work function of Al. A thin Ni layer was deposited before the thick Al mirror to overcome this issue [11]. To increase the LEE of DUV-LEDs, the p-GaN should be as thin as possible to minimize light absorption loss and maximize the reflectivity of the reflective p-contact electrodes in the DUV region. Therefore, omnidirectional reflectors (ODRs), which comprise SiO₂/Al [12] and MgF₂/Al [13], were deposited on the inclined sidewall to increase the output power of DUV-LEDs. Al_xGa_1-xN is the primary material for near-UV and DUV-LEDs and other optoelectronic devices owing to the tunability of the bandgap in the large energy range of 3.4–6.2 eV [14]. DUV-LEDs comprise Al_xGa_1-xN material with different Al content for the AlN template, n-Al_xGa_1-xN for the n-contact, multiple quantum well (MQW), p-Al_xGa_1-xN for the electron blocking layer (EBL), and p-GaN for the p-contact. Therefore, the optoelectronic properties of DUV-LEDs depend on the material properties of Al_xGa_1-xN and epitaxial structure design of the LED [15]. The basic properties and bandgap of Al_xGa_1-xN alloys are modified by the different Al content and varied ambient temperatures, which further influence the optoelectronic properties of DUV-LEDs, such as the wavelength (noted as WLP), forward voltage (noted as VF), and output power.

In our previous work, the HfO₂/SiO₂ stacks functioned as a reflective passivation layer (RPL) and effectively improved the performance of DUV-LEDs [9]. Although the output power of DUV-LEDs increased, the maximum external quantum efficiency (EQE) was only 3.09%. However, the effect of heating in the DUV-LEDs during the operation on their performance has been studied. DUV-LEDs are always operated in harsh environments, such as high temperature, and the performance of DUV-LEDs operated at high temperatures has not been studied previously.

In this study, the thermal behavior of DUV-LEDs with different p-GaN contact layer thickness and ITO current spreading layer with and without RPL were studied. The wavelength variation and junction temperature of the above DUV-LEDs were measured. Furthermore, the characteristics and performances of DUV-LEDs operated at different ambient temperatures were also studied.

2. Experiments

The DUV-LED structure was grown on a (0001)-oriented c-plane sapphire substrate by metalorganic chemical vapor deposition (MOCVD). It consisted of a 25 nm AlN template fabricated by physical vapor deposition (PVD), 1.5 µm-thick AlN layer, 2 µm-thick Si-doped Al_0.45Ga_0.55N layer, Si-doped Al_0.4Ga_0.6N/Al_0.6Ga_0.4N multiple quantum well (MQW), 30 nm-thick p-Al_0.75Ga_0.25N as EBL p-GaN:Mg layer, and Mg-doped p⁺-GaN of 5 nm-thick contact layer. The total thickness of p-GaN and p⁺-GaN were 100 nm and 40 nm, respectively, in this study. Disilane (Si₂H₆) and bis (cyclopentadienyl) magnesium (Cp₂Mg) were used for n- and p-type doping, respectively. Trimethylgallium (TMGa) and trimethylaluminum (TMAl), and ammonia (NH3) were used for the group-III and group-V elements, respectively. Details of the fabrication process of DUV-LEDs have been described elsewhere [9]. The ITO fhin film of 100 nm and 12 nm was used as p-contact with p⁺-GaN. A KLA-Tencor profilometer P-10 was used to control the thicknesses of HfO₂, SiO₂, and metal electrode. The reflectance of the RPL was measured by an n&k 1280 analyzer (n&k Technology, Inc.). The dimensions the of DUV-LED were 40 mil × 40 mil. The junction temperatures of the packaged DUV-LEDs were measured by a TERALED and T3Ster combined system with calibrated integrating sphere, respectively. The junction temperatures of the packaged DUV-LEDs were measured at room temperature(25°C) for different injection currents. The thermal behaviors of the packaged DUV-LEDs with different ambient temperature under injection current of 500 mA were also measured by a TERALED and T3Ster combined system. The thermal behaviors study of the packaged DUV-LEDs with different ambient temperature was heated by hot-plate, which was controlled by thermocouple. The three types of the LED samples were fabricated using the same epitaxial structure with thick/thin p-GaN and ITO (110 and 12 nm thick) and used to evaluate the effects of the RPL on the performance of the DUV-LEDs. The following configurations were defined: (1) Ref-LED with 100 nm-thick p-GaN and 110 nm-thick ITO thin film, (2) Thin GaN-LED with 40 nm-thick p-GaN and 110 nm-thick ITO thin film, and (3) Thin GaN-ITO-RPL-LED with 40 nm-thick p-GaN, 12 nm-thick ITO thin film and RPL structure. All the DUV-LEDs were packaged using the flip chip on the circuit boards, without epoxy.

3. Results and discussion

It is important to evaluate the effect of the thickness of p-GaN and ITO on the performance of DUV-LEDs with and without the RPL structure. Figure 1(a) shows the output powers and forward voltages of the Ref-LED, Thin GaN-LED, and Thin-GaN-ITO-RPL-LED as functions of the injection current. The output power was found to increase with the injection current and did not saturate even when the injection current was 1600 mA. The output powers of Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED at an injected current of 500 mA were 40.2, 48.4, and 67.5 mW, respectively, while the corresponding voltages were 6.007, 6.741 and 7.277 V, respectively. The reduction of the thickness of p-GaN from 110 nm to 40 nm was observed to contribute only 8.2 mW, which corresponded to an improvement of approximately 21% in the output power. However, the thin ITO and RPL structures contributed an output power of 19.1 mW compared to the output powers of the Thin GaN-LED and Thin GaN-ITO-RPL-LED, which was increase 40% in the output power. Therefore, reducing the thickness of p-GaN from 100 nm to 40 nm, and ITO from 110 nm to 12 nm and adding the RPL structure contributed 27.3 mW compared to the output power of the Ref-LED, which increases the 68%.Thus, the improved output power of the DUV-LEDs was strongly related to the p-GaN, ITO thickness, and RPL structure. In particular, the thin ITO and RPL contributed more output power than the thin p-GaN. These results are summarized in Table 1. Although the DUV-LEDs with thin p-GaN, thin ITO, and RPL structure improve the output power, it is necessary to evaluate the wall plug efficiency (WPE) of DUV-LEDs owing to the operated voltage, which could be different for these structures. Figure 1(b) shows the WPE characteristics of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED as functions of the injection current. The WPE decreased with increasing injection currentfor all LEDs. Furthermore, the WPE of the Thin GaN-ITO-RPL-LED was higher than those of the Ref-LED and Thin GaN-LED. The Ref-LED exhibited the lowest WPE because the thick p-GaN and ITO film absorbed most of the DUV light. The WPE of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED were 1.34%, 1.44%, and 1.86% at 500 mA injection current, respectively. Although thin p-GaN reduced the absorption of DUV light, it resulted in a high forward voltage of the Thin GaN-LED even if it was deposited with a thick ITO (110 nm) current spread layer.

 

Fig. 1. (a) Output power characteristics, (b) WPE, (c) LEE characteristics, and (d) junction temperature as a function of the injection current measured at room temperature.

Download Full Size | PDF

Table 1. Opto-electronic Properties of DUV-LEDs with Different Structures

View Table

Although the DUV-LEDs with thin p-GaN, thin ITO, and RPL structure improve the output power, it is necessary to evaluate the wall plug efficiency (WPE) of DUV-LEDs owing to the operated voltage, which could be different for these structures. Figure 1(b) shows the WPE characteristics of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED as functions of the injection current. The WPE decreased with increasing injection currentfor all LEDs. Furthermore, the WPE of the Thin GaN-ITO-RPL-LED was higher than those of the Ref-LED and Thin GaN-LED. The Ref-LED exhibited the lowest WPE because the thick p-GaN and ITO film absorbed most of the DUV light. The WPE of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED were 1.34%, 1.44%, and 1.86% at 500 mA injection current, respectively. Although thin p-GaN reduced the absorption of DUV light, it resulted in a high forward voltage of the Thin GaN-LED even if it was deposited with a thick ITO (110 nm) current spread layer.

The maximum WPE of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED were 1.77%, 1.80%, and 2.25%, respectively. Since the WPE was defined [16] as WPE = P/IV, where P is the light output power of the LED (unit: mW), IV was the electrical power provided to it. The forward voltages of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED were 6.007, 6.741 and 7.277 V, respectively, under an injection current of 500 mA. The difference between the forward voltages of the Ref-LED and Thin GaN-LED was 0.734 V because the thin p-GaN resulted in higher forward voltage. Moreover, the difference between the forward voltages of Thin GaN-LED and Thin GaN-ITO-RPL was 0.536 V owing to thin ITO which caused higher forward voltage. Because of the reduced absorption for the Thin-GaN LED, absorption by the 110 nm ITO and without RPL occurred. It increased the WPE slightly (ΔWPE = 0.03%) for the Thin GaN-LED compared to that of the Ref-LED. Moreover, the Thin GaN-ITO-RPL-LED exhibited the highest forward voltage and output power because of the thin p-GaN, thin ITO and with RPL structure. Based on the above results, reducing the light absorption by p-GaN and ITO and un-absorption light reflected by the RPL were crucial for achieving the highest WPE.

The LEE was used to evaluate the effect of RPL on the performance of DUV-LEDs to obviate the forward voltage factor, which is shown in Fig. 1(c). It was defined as LEE = EQE/IQE, where EQE and IQE denote the external and internal quantum efficiencies, respectively. These have been studied in our previous work [9]. The maximum LEE of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED are 4.68%, 5.75%, and 7.57% at injection currents of 20, 20, and 300 mA, respectively. The LEE increases significantly at first, reaching a maximum at a certain current and then decreases slightly. To further investigate the effect of different thickness of p-GaN, with/without the RPL structure, the droop efficiencies (defined as [(LEE_max-LEE_1600mA)/LEE_max] × 100%) of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED were calculated as 24.20%, 22.10%, and 20.93%, respectively. The droop efficiency in the LEE is clearly improved by reducing the UV light absorption of p-GaN layer and introducing the thin ITO and RPL structure.

Although the output power and droop can be improved by the thin p-GaN, ITO, and RPL structure, the WPEs of these LEDs are still low. The input power heated the DUV-LEDs. The junction temperatures (T_j) of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED are shown in Fig. 1(d). The junction temperatures of all the samples increased with the injection current owing to self-heating caused by the generated phonons. At high injection currents (above 500 mA), the junction temperatures of the Ref-LED with thick p-GaN increased linearly. The junction temperatures rate of Thin GaN-LED and Thin GaN-ITO-RPL-LED with thin p-GaN layers were lower than that of the Ref-LED. As mentioned in a previous study [9], the transmittance of thin p-GaN (T% = 45.7%, @280 nm) improved significantly and exceeded that of thick p-GaN (T% = 15.7%, @280 nm), which enabled more photons to escape from active layer of the LED. Meanwhile, the escaped photons emitted near the edge of the mesa could be reflected back to the active layer owing to total internal reflection between Al_xGa_1-xN and SiO₂ passivation layer. The light absorption in thick p-GaN and total internal reflection could be primarily responsible for self-heating in the junction [17] and lead to a high junction temperature in the Ref-LED. The light absorption in the thin p-GaN and thin ITO thin film samples improved significantly. Therefore, a considerable amount of light escaped from the LED with thin p-GaN, thin ITO, and RPL structure and reduced self-heating which alleviated the junction temperature rate in the Thin GaN-LED and Thin GaN-ITO-RPL-LED samples.

Figure 2(a) shows the measurement/simulation results of the reflectance spectra of the RPL and single passivation layer. The measured and simulated reflectance data display similar trends in ranges of wavelength. The reflection bandwidth is defined as a wavelength range where the reflectivity is higher than 90%. The reflection bandwidth of RPL is 53 nm and 56 nm for measurement and simulation, respectively. At 280 nm, the reflectance of RPL for measurement and simulation data are 92.11% and 99.09%, respectively. The difference between the measured and simulated reflectances caused deviations in the extinction coefficient and thickness. Comparing the LED performance of DUV-LEDs with thick and thin p-GaN layers and those with and without RPL, the LEE of DUV-LEDs was improved by the RPL. Further, the junction temperatures of the DUV-LEDs with thin p-GaN thickness for Thin GaN-LED and Thin GaN-ITO-RPL-LED were reduced, especially under a large injection current of 500 mA. The single passivation layer exhibited a reflectance of 9.45% at a wavelength of 280 nm, which was significantly lower than that of the RPL. Meanwhile, the reflectance of different ITO thickness of 110 nm and 12 nm were 17.82% and 23.19%, which were lower than that of the RPL. the different Therefore, the LEE of Thin GaN-LED was higher than that of the Ref-LED because of reduced light absorption with thin p-GaN. The comparison shows that the RPL could efficiently improve the reflectance in the DUV and yield improved optical properties. Figure 2(b) shows the emission wavelengths of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED as a function of the injection current. At an injection current of 60 mA, the electroluminescence peaks of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED occur at 281.39, 281.03, and 280.32 nm, respectively. thin GaN-LED, and Thin GaN-ITO-RPL-LED showed small wavelength shift towards short wavelengths (blue-shift). The wavelength of the blue-shift resulted from the quantum confined Stark effect (QCSE) and typically Al_xGa_1-xN MQW designs [18]. A slight blue-shift of wavelength for these three samples was resulted from the uniformity of epilayer. In general, there exist serious stress in the epilayer of the DUV-LEDs due to the high Al composition. As the injection current increases to 400 mA, the electroluminescence peaks of Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED showed small wavelength shift towards short wavelengths (blue-shift). The wavelength of the blue-shift resulted from the quantum confined Stark effect (QCSE) and typically Al_xGa_1-xN MQW designs [17]. For the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED, the wavelengths red-shift were 0.67, 0.76, and 0.93 nm, respectively, for injection currents ranging from 400 to 1600 mA. These red-shifts, which were clearly insignificant for all the LEDs as the injection current increased from 400 mA to 1600 mA, may have originated from the thermal effect [19–20]. Owing to the lowering of the bandgap of semiconductors with increasing temperatures, the temperature dependence of the bandgap can be expressed as follows:

 

Fig. 2. (a) Measured and calculated reflectance spectra of the RPL and single passivation layer and (b) emission wavelengths of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LEDs as functions of the injection current.

Download Full Size | PDF

The thermal management of DUV-LEDs is an important issue, especially at high injection currents and ambient temperatures. In this study, the DUV-LEDs consist of p-GaN and ITO thin films of different thickness, which could alter the thermal stability of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED. The thermal stability of DUV-LEDs was evaluated by measuring their electrical and optical properties at the same high injection current (500 mA) under different ambient temperatures (from 20 °C to 120 °C).

Figure 3 shows the forward voltage, output power, and wavelength of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED measured at different ambient temperatures. From Fig. 3(a), the forward voltage (V_F) of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED are 6.007 V, 6.741 V, and 7.277 V, respectively, at an injection current of 500 mA and ambient temperature of 20 °C. Since the electrical power supplied to an LED can be transferred to the optical output power and heat, the temperature sensitive coefficients were used to evaluate the thermal stability performance of the LED devices. The temperature sensitive coefficients (dV/dT) of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED are 5, 9, and 11 mV/K, respectively. The thicknesses of the p-GaN and ITO were the main factors affecting the dV/dT of the DUV-LED. The p-GaN and ITO thickness were the main factors affecting the V_F of DUV-LEDs because the contact resistance and thin film material characteristics of p-GaN/ITO were changed. Although the thick p-GaN caused the UV-light absorption, it contributed to low temperature sensitive coefficients. In contrast, the dV/dT of the Thin GaN-LED and Thin GaN-ITO-RPL-LED were significantly higher than that of the Ref-LED, especially the Thin GaN-ITO-RPL-LED. Because the carrier concentration and mobility of the ITO thin film decreased as the ambient temperature increased, the resistivity of the ITO thin film increased further [21]. Therefore, the contact resistance of p-GaN/ITO increased with the ambient temperature. In addition, the thin p-GaN thickness increased the contact resistance of p-GaN/ITO, which resulted in an abrupt increase in the temperature sensitive coefficients from 5 to 9 mV/K for theThin GaN-LED compared to that of the Ref-LED. It is noteworthy that the temperature sensitive coefficient of DUV- LEDs changed by only 2 mW/K (from 9 mW/K to 11 mW/K) as the ITO thickness was reduced from 110 nm to 12 nm. Obviously, thin p-GaN dominated the temperature sensitive coefficient (dV/dT).

 

Fig. 3. (a) Forward voltage, (b) output power, and (c) wavelength of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LEDs at the injection current of 500 mA.

Download Full Size | PDF

Figure 3(b) shows that the output power of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED are 40.2, 48.2, and 67.5 mW, respectively, at an injection current of 500 mA and ambient temperature of 25 °C. The output power of the LED was varied by introducing the RPL and p-GaN and ITO layers of different thicknesses. At 500 mA, the output powers of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED decrease linearly with increasing ambient temperature. It is well known that the bandgap of semiconductor decreases as the temperatire increasing. It means the confinement of the barrier in the MQWs become less as increasing ambient temperature, which resulted in the LOP reducing [22]. The decrement in slope variation (dLOP/dT) of the output power for the DUV-LED varied in magnitude. The dLOP/dT of the Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED were 111, 97, and 157 mW/K, respectively. The dLOP/dT of the Thin GaN-LED and Thin GaN-ITO-RPL-LED changed significantly owing to the thin p-GaN and thin ITO of 12 nm. In addition, the modified dLOP/dT of these three LEDs were related to the junction temperatures, which were 97.08°C, 101.62°C, and 101.46°C, respectively, under an injection current of 500 mA and ambient temperature of 25 °C. The junction temperature analysis indicated that the junction temperatures of the Thin GaN-LED and Thin GaN-ITO-RPL-LED were very similar. However, the dLOP/dT of the Thin GaN-ITO-RPL-LED showed the worst value, owing to the electrical properties of thin p-GaN and thickness of the ITO. However, the performance of Thin GaN-ITO-RPL-LED remained the best when these three structures operated as LEDs at 120 °C.

To clarify the effect of ambient temperature on the wavelength shifts of the three types of LEDs, the wavelengths at 500 mA were measured as a function of the ambient temperature, as shown in Fig. 3(c). For injection currents higher than 500 mA, the red-shift wavelength of the DUV-LEDs depends strongly on the junction temperature, (shown in Fig. 2(b)). The junction temperatures of DUV-LEDs at a fixed injection current increased with the ambient temperature. Therefore, the wavelength shift of DUV-LEDs increased because of the high junction temperature. The increment in slope variation (dWLP/dT) of the wavelength Ref-LED, Thin GaN-LED, and Thin GaN-ITO-RPL-LED varied in magnitude, and were 0.0186, 0.0178, and 0.0168 nm/K, respectively. The dWLP/dT of the Ref-LED was clearly higher than those of the Thin GaN-LED and Thin GaN-ITO-RPL-LED. Since the bandgap of p-GaN narrowed after the ambient temperature increased to cause more light absorption, which results in self-heating in the junction was also increased. Meanwhile, the light absorption of thick p-GaN was also higher than that of thin p-GaN to cause more self-heating then further influence the wavelength [22]. Therefore, the lowest red-shift wavelength of Thin GaN-ITO-RPL-LED under different ambient temperature was related to p-GaN and ITO thickness because of the lowest self-heating level in the junction. The results are consistent with Fig. 1(d) and Fig. 2(b).

4. Conclusion

This study reported the effect of the RPL on the LEE of DUV-LEDs with p-GaN and ITO thin films of different thicknesses. The DUV-LEDs with thin p-GaN and RPL provided high LEE of 7.57% for DUV wavelength applications. A comparison of thick and thin p-GaN layers for DUV-LEDs with and without RPL demonstrated an improved LEE and a lower junction temperature for DUV-LEDs with thin p-GaN and ITO thickness. The DUV-LEDs comprising thick p-GaN and RPL exhibited higher junction temperatures because the light absorbed by thick p-GaN caused further self-heating of DUV-LEDs. Most of the DUV light trapped inside the DUV-LED caused self-heating. The DUV-LEDs with RPL, thin p-GaN, and thin ITO are promising candidates for high power applications of DUV-LEDs.