Study on aging performance of 1W silicon-based blue LED on different substrates

Since the 1990s, GaN materials have been widely used in display, indication, backlight and solid-state lighting, and have formed a huge market. So far, gallium nitride (GaN)-based light-emitting diodes ( LEDs ) prepared on three substrates (sapphire, silicon carbide and silicon) have been commercialized. In recent years, silicon substrate GaN-based LED technology has attracted much attention. Because silicon (Si) substrates have the advantages of low cost, large crystal size, easy processing and easy transfer of epitaxial films, they have excellent performance-price ratio in the application of power LED devices.

　　Many research groups have grown GaN epitaxial films on Si substrates and some have obtained devices or studied the related properties of Si-based GaN. In the process of LED preparation, the GaN film is transferred to a new supporting substrate to prepare a vertical structure device, which obtains better optoelectronic performance than the same-side structure device.

　　In this paper, the GaN epitaxial film grown on the Si substrate was transferred to a copper support substrate, a copper-chromium support substrate by electroplating and to a Si support substrate by pressure welding to obtain a vertical structure light-emitting device, and a comparative aging study was conducted on the three samples.

　　experiment

　　The epitaxial wafer used in the experiment is a 2in (50.8mm) blue light InGaN/GaN multi-quantum well epitaxial wafer grown on a silicon (111) substrate by MOCVD method. The chip size is 1000Lm@1000Lm , and the growth method has been reported. The experiment prepared three epitaxial wafers grown in the same furnace. One of them uses the pressure welding technology and chemical etching method to transfer the GaN epitaxial film to the Si substrate and obtain a light-emitting device, which is called sample A. The other two use electroplating and chemical etching methods to transfer the GaN epitaxial film to the electroplated copper substrate and the electroplated copper-chromium substrate respectively to obtain light-emitting devices, which are called sample B and sample C respectively. Except for the different epitaxial film transfer methods and supporting substrates, the other device manufacturing processes of the three samples are consistent.

　　Since there are slight differences between individuals of the same type of samples, samples A, B, and C were initially tested, and representative chips were selected for experiments and tests. Each chip is a bare die package. The working current of a chip with a size of 1000Lm@1000Lm is usually 350mA. In order to accelerate aging, a DC current of 900mA was passed through samples A, B, and C at room temperature. The current-voltage (IV) characteristic curves, electroluminescence (EL) spectra, and relative light intensity of each sample at each current before and after aging were tested using a power supply KEITHLEY2635 and a spectrometer Compact Array Spectrometer (CAS) 140CT.

　　Results and Discussion

　　IV Characteristic Analysis

　　Table 1 shows the Vf and Ir values ​​of the three samples before aging, after aging for 80, 150 and 200 hours. The aging condition is 900 mA at room temperature, where Vf is the voltage value at 350 mA and Ir is the leakage current value at 10 V in the reverse direction. Usually, the reverse leakage current Ir is measured at 5 V in the reverse direction. To compare the results, a more stringent condition is selected and the measurement is made at 10 V in the reverse direction. Figure 1 shows the IV characteristic curves of the three samples before aging, after aging for 80, 150 and 200 hours, as shown in Figures 1(a) to (d) respectively. Figure 1(a) shows that the three samples A, B and C have good IV characteristics before aging, with a turn-on voltage of about 2.5 V and a current of 10-9 A at the reverse direction of 10 V. After aging for 200 hours, the leakage current Ir of the three samples at the reverse direction of 10 V increases significantly compared with that before aging. Table 1 shows that after aging with high current for 200 hours, the leakage current of sample B is the smallest under the same reverse voltage (-10V), followed by sample A, and sample C is the largest. Moreover, as the aging time goes by, the difference in leakage current of the three samples under the same reverse voltage becomes larger and larger. The forward voltage of InGaNMQWLED increases slightly after aging because the long-term aging with high current causes the exposed n-electrode (aluminum) to be locally oxidized, resulting in an increase in contact resistance. The reason why the leakage current increases after aging is that the width of the InGaNLEDpn junction depletion layer is mainly determined by the carrier concentration of the p-type layer. After the chip has been aged for a long time with a large current, the acceptor Mg is reactivated due to the decomposition of the Mg-H complex, which increases the p-type carrier concentration, resulting in a narrower depletion layer, thinner barrier region under reverse bias, more tunnel breakdown components, and increased reverse current. In addition, after the chip has been aged for a long time with a large current, the defect density in the quantum well region increases, and the defect and trap-assisted tunneling causes leakage current under reverse bias. The thermal conductivity of the three samples B, A, and C decreases in turn, so the defects and trap density generated during aging decrease in turn. Therefore, under the same reverse voltage, the leakage current of the three samples increases in turn (as shown in Table 1 and Figure 1).

　　Figure 1 IV characteristic curves of three samples before and after aging

　Table 1 Vf and Ir values ​​of the three samples before and after aging

　　EL spectrum analysis

　　Figure 2 shows the electroluminescence (EL) spectra of the three samples at 1, 10, 100, 500, 800, 1000 and 1200 mA before and after continuous aging at 900 mA for 168 hours at room temperature [Figure 2 (a1) ~ (a3)], as well as the relationship between the EL wavelength and the current before and after aging of the three samples [Figure 2 (b1) ~ (b3)]. The solid line in the figure represents the spectrum before aging, and the dotted line represents the spectrum after aging. Figure 2 (a1) ~ (a3) ​​shows the EL spectra before and after aging after normalization. The EL spectrum waveforms of the three samples at different currents before and after aging have no obvious changes except that the peak wavelength at high current has a red shift. Figure 2 (b1) ~ (b3) shows that there are obvious differences in the wavelength of the three samples before and after aging with the current. Among them, the relationship between the wavelength and the current of sample B before and after aging is almost the same, except that its wavelength increases slightly at the same current after aging. Because the thermal conductivity of the substrates of samples A, B, and C is different, the junction temperature of each sample is different during aging, so the wavelength drift of sample C at the same current after aging is the largest, followed by sample A, and the smallest for sample B. In addition, due to the different substrate materials and chip transfer methods of the three samples, the stress conditions of the GaN epitaxial film on the new substrate after transfer are different. Literature research shows that after GaN is transferred from a silicon substrate to a new silicon substrate by pressure welding and chemical corrosion, the tensile stress of the entire GaN layer is reduced, and the compressive stress of the quantum well InGaN layer is increased. The stress relaxation of GaN transferred by thin film by electroplating is more thorough, so that the compressive stress of the quantum well is greater, and the polarization electric field generated is greater, resulting in a greater energy band tilt, so the energy of the photons released when the carriers recombine is reduced, which is manifested as a longer EL wavelength. Therefore, in the EL spectrum before and after aging, the wavelength of sample A, which is pressure welded on the silicon substrate, is the shortest, followed by sample C, and sample B is the longest, and samples B and C are very close. Figure 2 also shows that the wavelength red shift of sample B is the largest from small current to large current before and after aging. This may be related to the following aspects. On the one hand, the increase in junction temperature makes the GaN bandgap width smaller, causing the wavelength red shift. On the other hand, since the stress relaxation of sample B is the most thorough, the compressive stress on the quantum well of sample B is the largest, so the polarization effect of the multi-quantum well region of sample B is the strongest. The polarization effect produces a strong built-in electric field, which leads to a significant quantum-confined Stark effect, causing the red shift of the emission wavelength.

　Figure 2 EL spectra of the three samples before and after aging at 900 mA for 168 hours at room temperature [(a1)~(a3)] and the relationship between the wavelength and current of the three samples before and after aging [(b1)~(b3)]

　　Power-current (LI) relationship analysis

　　Figure 3 shows the relationship between the relative light intensity of each sample and the aging time at 350mA current. The light intensity before aging is 100% for all three samples. As can be seen from Figure 3, the light intensity of samples A, B, and C first increases and then decreases with the increase of aging time. Among them, the light intensity of sample A increases the most after aging for 2 hours, and then the light intensity begins to decrease as the aging progresses. The light intensity of samples B and C begins to decrease after aging for 32 hours and 10 hours, respectively, and the decreasing trend is slower than that of sample A. It can also be seen that after aging at room temperature and 900mA, the light intensity of samples A, B, and C at 350mA passes through a maximum value and then decreases. Sample C decreases the most, followed by A. Although the light intensity value of sample B is decreasing, it is still greater than the value before aging. The reason for this phenomenon is that some of the acceptor Mg in GaN grown by the MOCVD method is passivated due to the formation of Mg-H complex with H, and the activation rate of Mg is very low, resulting in a low hole concentration. During high current aging, some Mg-H bonds are broken and the acceptor Mg is activated, thereby increasing the hole concentration, and the carrier concentration may become more matched, and the luminescence efficiency becomes higher. On the other hand, aging increases the density of non-radiative recombination centers such as dislocations and defects in GaN materials, thereby reducing the luminescence efficiency and the light intensity. These two mechanisms compete with each other. In the early stage of aging, the Mg acceptor activation mechanism is dominant, so the light intensity of the three samples increases at the same current. As aging proceeds, the non-radiative recombination center proliferation mechanism such as dislocations and defects gradually dominates, so the light intensity of the three samples decreases after a period of high current aging. The different speeds of light decay of the three samples may be caused by the different stress states of the quantum wells of the three samples and the different thermal conductivity of the supporting substrate, which causes different degrees of proliferation of non-radiative recombination centers.

　　Figure 3. Relationship between relative light intensity at 350mA and time after aging at 900mA at room temperature (with the light intensity before aging as 100%)

　　in conclusion

　　By conducting comparative aging studies on GaN-based blue LEDs grown epitaxially on silicon substrates and transferred to silicon substrates, copper substrates, and copper-chromium substrates, the results show that the EL wavelength of the device on the copper substrate is the longest at the same current because the stress relaxation of the GaN epitaxial film is more thorough after electroplating and transfer to the copper substrate. The aging of LED devices on three different substrates shows that the main factor affecting the reliability of LEDs may be their stress state. The IV characteristics, LI characteristics, and EL spectra of the three substrate LEDs before and after aging were studied, and the comparison shows that the copper substrate device has better aging performance.