Research on the ability of high-power LED chips to resist overcurrent

　　As a new type of lighting technology, LED has the advantages of low energy consumption, long life, small size, dimmable, flexible control and environmental protection. Its application prospects have attracted worldwide attention. As the price of LED drops, the market gradually opens up, and more and more lighting products use LED as the light source. Especially in the field of road lighting, high-power LED products have become the protagonists of the market, and LED has shined in the field of outdoor lighting [2]. However, withthe increase in the application of LED lamps , the number of outdoor LED lamps that fail due to lightning surges is also increasing. According to surveys, most of the LED outdoor lamps that are damaged within the normal service life are due to the over-electrical stress generated by lightning surges that fails the lamp power supply and LED light source . This not only affects the service life of the lamp, but also increases the maintenance cost of the enterprise. In view of this, the lightning surge resistance of LED outdoor lamps should be given enough attention.

　　The lightning surge resistance of LED lamps mainly depends on two aspects: (1) the lightning surge resistance and protection mechanism of the LED driver power supply. (2) the over-voltage stress resistance of the LED chip . For LED driver power supplies, the lightning surge resistance should be judged by two points: (1) the lightning surge resistance of its own components to ensure that the power supply can still work normally after the lightning surge. (2) The power supply's ability to attenuate the surge current and voltage waveform to ensure that the peak current and voltage attenuated after the surge passes through the power supply is within the range that the LED chip can withstand. Zhang Jinjian et al. [3] of Tianjin University studied the lightning surge resistance of LED driver power supplies. According to the characteristics of lightning surge, they designed a surge protection circuit suitable for LED power supplies using surge devices such as gas discharge tubes, varistors, and transient suppression diodes. They also used a lightning surge generator to conduct an immunity test to test its lightning resistance performance. The experimental results show that it can withstand differential mode 1kV and common mode 2kV lightning high voltages to ensure the normal operation of the LED power supply. However, the lightning surge resistance requirement of the LED driver power supply is determined by the over-voltage stress resistance of the LED chip. Therefore, it is necessary to study the over-voltage stress resistance of the LED chip.

　　Based on this, this paper conducts lightning surge experiments on several different types of high-power LED chips to explore the ability of different high-power LED chips to resist over-electrical stress, which provides a reference for the selection of driving power supplies and LED chips for LED outdoor lamps, as well as the design and development of lightning surge resistance, which has important practical significance.

　　2. Factors affecting the LED chip's ability to resist over-electrical stress

　　First, the current density that an LED chip can withstand determines its ability to resist over-electrical stress. The greater the current per unit cross-sectional area that an LED chip can withstand, the stronger its ability to resist over-electrical stress. For conventional conductors, the current density must be low enough to prevent the conductor from melting or fusing, or the insulating material from being broken down [4]. Electromigration will occur inside the LED chip at high current density. When a conductive metal material passes through a high current density, metal atoms will migrate and diffuse along the direction of electron movement. Electromigration in an LED causes metal atoms to diffuse freely from one lattice to another lattice vacancy. Taking a flip-chip structure chip as an example, when the electron flow flows from the interconnect lead to the eutectic alloy bump, due to the sudden change in the geometry of the interconnect lead to the bump, current density concentration and local Joule heating effect will occur at the interface [5]. Current density concentration makes the current density distribution in the bump and the chip and substrate leads uneven, resulting in complex electromigration forces locally at the current density concentration point, accelerating the electromigration process and accelerating the failure of the LED.

　　Secondly, the current concentration effect affects the chip's ability to resist over-electric stress. Current concentration is the uneven distribution of current density on the chip, especially near the chip contact points and above the PN contacts. The current concentration phenomenon of the LED chip causes local overheating and hot spots on the chip, aggravating the electromigration effect and making the local current density unevenly distributed. The uneven current density causes the local temperature of the chip to rise, and the temperature rise causes the resistivity to decrease, which leads to an increase in the Auger recombination of local carriers [6], affecting the internal quantum efficiency of the chip. When minority carriers leak through the charge region of the heterojunction, the current injection efficiency will decrease, resulting in uneven local light emission and overheating of the LED chip, affecting the chip's light-emitting performance and service life, and ultimately causing the LED chip to short-circuit or open-circuit. This phenomenon is particularly serious when the chip size and injection current are large.

　　Finally, the current carrying capacity of the LED chip bonding wire is a factor that affects the LED chip's ability to resist over-electrical stress. Although LED failure due to the melting of the bonding wire is not common in actual applications, the diameter, length, bonding type, physical material properties of the metal, and resistance of the bonding wire all have an impact on the current carrying capacity of the gold wire. When the over-electrical stress is large, the conductor melts and the LED opens.

　　The above factors jointly affect the ability of LED chips to resist over-electric stress. Different chip technology processes can improve the electromigration and current concentration effects of the chip. For example, optimized interdigitated electrodes can improve the current crowding phenomenon; vertical structure chips can improve the current concentration phenomenon by allowing the current to flow longitudinally in the chip. At the same time, the number and position of the electrodes and bumps of the flip chip[5], as well as the processing and manufacturing of the ohmic contacts have a significant impact on the current expansion of the chip. By optimizing the geometry and electrical parameters of the electrodes and bumps, the current crowding effect can be greatly reduced, the unevenness of the current density distribution can be improved, the current expansion can be promoted, and the total equivalent resistance of the chip can be reduced.

　　It can be seen that LED chips with different structures and processes have different performances in resisting overvoltage stress under the same surge pulse. The following experiment finds the range of single pulse current peak resistance of common high-power LED chips on the market.

　　3. Experiment on the resistance of different high-power LED chips to single pulse current

　　In order to simulate the impact of actual lightning strikes on LED lamps and chips, the Hangzhou Yuanfang EMS61000-5A[7] intelligent lightning surge generator was used to simulate the surge waveform generated in the power grid during a lightning strike.

　　The output waveform of EMS61000-5A is a standard combination of a voltage composite wave of 1.2/50μs and a current composite wave of 8/20μs [8]. The voltage composite wave (as shown in Figure 1) has a wavefront time of T1=1.67T=1.2μs±0.36μs and a half-peak time of T2=50μs±10μs.

　　The current comprehensive wave (as shown in Figure 2) has a wavefront time of T1 = 1.25T = 8μs ± 1.6μs and a half-peak time of T2 = 20μs ± 4μs.

　　When using the entire LED lamp (including lamp beads and driver power supply) as the surge test object, it is found that the output surge current of the standard surge waveform after passing through the LED power supply is uncertain. Due to the different lightning protection designs of driver power supplies from different manufacturers, the shape of the surge waveform and the peak current at the power supply output end, i.e. the lamp bead input end, are uncontrollable, which increases the uncertainty factors in the experimental process.

　　To solve the above problems, this experiment uses a DC power supply to directly power a single LED and add a surge waveform to the DC circuit. By adjusting the peak value of the pulse voltage output by the device and the resistance value of the resistor in series with a single LED, the peak value of the current pulse at the input end of the lamp bead can be changed. In this way, the surge waveform can be determined and the peak current of the pulse can be controlled.

　　During the experiment, the lamp beads were first lit at a working current of 350mA, and the above pulse waveform was applied to the circuit with a resistor and LED lamp beads in series. The pulse voltage was gradually increased from 250V, with an increase interval of 50V. Surge impact was performed 5 times in each voltage range, with an interval of 10s between each waveform. If the lamp beads work normally after the test, the next voltage range was entered to continue the test. At the same time, an oscilloscope was used to observe the peak current waveform at both ends of the lamp beads. When the lamp beads failed and broke down, their pulse waveforms were recorded to determine the pulse peak current when they failed (in order to eliminate the influence of the lamp bead protection electrode on the experiment, the lamp bead protection electrode was removed before the experiment).

　　3.1 Sapphire substrate horizontal structure chip surge test

　　A surge test was conducted on a 45mil*45mil horizontal structure LED chip on the market. Starting with a 250V surge, the lamp bead failed during the second 300V surge. The current surge waveform recorded at both ends of the lamp bead when it failed is shown in Figure 3.

　　The experiment tested a total of 5 horizontal structure LED lamp beads, and the peak pulse currents when they failed were: 15.55A, 15.88A, 15.00A, 15.62A, and 15.22A respectively.

　　Analysis of the lamp beads after the surge found that the lamp beads were completely short-circuited and the chip electrodes were completely broken down and fused, as shown in Figure 4:

　　Figures 5 and 6 are the surface brightness distribution diagrams of the sapphire substrate upright structure chip at 1mA and 150mA respectively. It can be observed from the figure that the uneven current distribution on the chip surface at 1mA leads to uneven brightness on the chip surface, and as the current increases (150mA), the uneven current distribution phenomenon becomes more severe.

　　3.2 Surge resistance test of SiC flip-chip LED chip

　　A 40mil*40mil SiC substrate flip chip on the market was selected for over-voltage stress test. The lamp bead failed under a 650V surge voltage, and the surge waveform it was subjected to when it failed is shown in Figure 7.

　　The experiment tested a total of 5 SiC substrate flip-chip structure LED chips, and the pulse peak currents when they failed were 29.22A, 29.68A, 33.57A, 35.68A, and 35.39A. By comparison, it was found that the above peak currents were twice as high as the surge peak current of the tested horizontal structure chip.

　　Analysis of the lamp beads after the surge found that the lamp beads were completely short-circuited, as shown in Figure 8:

　　FIG9 and FIG10 are the surface brightness distribution diagrams of the SiC substrate upright structure chip at 1 mA and 150 mA respectively.

　　By observing the surface brightness distribution of the flip-chip structured lamp beads with SiC substrates at 1mA and 150mA, it is found that the brightness distribution of the chip is relatively uniform at low currents, indicating that the current distribution of the chip is uniform. At the same time, as the current increases (150mA), there is no obvious current unevenness in the chip.

　　For GaN-based blue LEDs, the lattice mismatch between SiC and GaN is only 3.4%, which is much smaller than the 17% lattice mismatch between the sapphire substrate and GaN. The epitaxially grown GaN film on the SiC substrate has a lower dislocation defect density, which means that the GaN LED on the SiC substrate has a higher internal quantum efficiency and is suitable for operation at high current density. In addition, the thermal conductivity of SiC is very high (420W/mK), which is more than 15 times that of sapphire (23-25W/mK) [9], which is beneficial to the heat dissipation of LED devices and improves the reliability of LEDs.

　　3.3 Sapphire peeling substrate flip-chip LED chip over-electrical stress test

　　A flip-chip structured lamp bead with a size of 55mil*55mil and a sapphire peeling substrate was selected for electrical stress testing. The chip failed under a surge voltage of 350V, and its failure surge waveform is shown in Figure 11.

　　The peak pulse currents of the five sapphire peeled substrate flip-chip structure lamp beads tested in the experiment when they failed were 16.2A, 16.59A, 12.23A, 14.49A, and 14.53A respectively.

　　Figure 12 shows the chip surface of a sapphire flip-chip LED light source after a surge failure. The breakdown caused by the local overtemperature caused by current crowding can be clearly observed on the chip surface.

　　3.4 SiC substrate vertical structure LED chip over-electrical stress test

　　A vertical structure chip with a size of 55mil*55mil and a SiC substrate was selected for lightning surge testing. The chip failed under a surge voltage of 600V. The surge waveform of the failure is shown in Figure 13.

　　In experimental tests, the peak pulse currents of five SiC substrate vertical structure lamp beads when they failed were 24.4A, 28A, 25.2A, 24.6A, and 26.0A respectively.

　　Analysis of the failed lamp beads revealed that the failure area was concentrated near the electrode, as shown in Figure 14.

　　3.5 Si substrate transfer vertical structure LED chip over-electrical stress test

　　A chip with a size of 45mil*45mil and a vertical structure transferred from a Si substrate was used for electrical stress testing. The chip failed under a surge voltage of 350V. The surge waveform it experienced when it failed is shown in Figure 15.

　　The peak pulse currents of the five Si substrate vertical structure chips tested in the experiment when they failed were 16.6A, 16.6A, 16.4A, 16.2A, and 16.5A respectively.

　　By analyzing the failed chip, it can be clearly observed that the metallized electrode near the N electrode is broken down, as shown in Figure 16.

　　The above test results are recorded as shown in Table 1:

　　4 Conclusion

　　Through the test of the over-electric stress resistance of the common high-power LED chips on the market, it is found that the over-electric stress resistance of LED chips with different structures and processes varies greatly. The peak value of the single pulse current it withstands when it fails ranges from 12A to 35A. For the LED driver power supply, when it is subjected to a lightning surge, it is necessary to ensure that the peak current of the surge waveform at its output end is less than 12A on the premise of ensuring its normal operation, so as to protect the LED lamp beads and avoid their immediate failure. Of course, another situation needs to be considered, that is, when the LED lamp beads are subjected to over-electric stress, the initial failure is only manifested as lamp bead leakage, and it takes a period of aging before the luminous flux drops significantly or the lamp is dead. This situation has higher requirements for the lightning resistance of the LED driver power supply. In the later stage, we will add the experimental content of lamp bead leakage detection and accelerated aging after surge impact to improve this part of the research work.

　　Another aspect of improving the lightning resistance of LED lamps is to improve the over-electric stress resistance of the LED light source used in the lamps. In Table 1, it can be seen that the peak pulse current that the flip-chip structure SiC substrate (graphically processed substrate, not completely stripped substrate) can withstand reaches 32A, which has a good performance in over-electric stress resistance. When used in combination with a driving power supply with strong lightning surge resistance, the overall lightning resistance performance of the lamp can be improved. Of course, the aspects of high-power LED chip manufacturing process (such as chip structure, epitaxial manufacturing, etc.) that play a decisive role in the chip's over-electric stress resistance are not fully reflected in Table 1, which is also the direction of our next stage of research.