Failure Mechanism for GaN-Based High-Voltage Light ... - IEEE Xplore

3 downloads 0 Views 885KB Size Report
Shoou-Jinn Chang, Fellow, IEEE, Chung-Ying Chang, Chun-Lung Tseng,. Ching-Shing Shen ... Index Terms—GaN, high-voltage, light-emitting diodes, metal,.
IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 26, NO. 11, JUNE 1, 2014

1073

Failure Mechanism for GaN-Based High-Voltage Light-Emitting Diodes Shoou-Jinn Chang, Fellow, IEEE, Chung-Ying Chang, Chun-Lung Tseng, Ching-Shing Shen, and Bing-Yang Chen

Abstract— We report a detailed reliability test study on GaN-based high-voltage light-emitting diodes. Under high temperature (i.e., 80 °C) and high current injection (i.e., 100 mA) conditions, it was found that Al metal whiskers were formed from the sidewall of the Cr/Al/Ti/Pt/Au p-finger metal after 120-h burn-in test. It was also found that the whiskers became longer as we increased the burn-in time. Furthermore, it was found that the formation of Al whiskers is directly related to Al migration. Index Terms— GaN, high-voltage, light-emitting diodes, metal, Al, whisker.

operation. Instead of high current and low-voltage, one can drive these micro-LEDs with high-voltage and low current. As compared to conventional high power LEDs, it has been shown that such high-voltage LEDs (HV-LEDs) are safer and could provide larger wall-plug-efficiency. However, no report on the failure mechanism of such HV-LEDs could be found in the literature, to our knowledge. In this letter, we report detailed reliability test results for GaN-based HV-LEDs. II. E XPERIMENTS

I. I NTRODUCTION

N

ITRIDE-BASED compound semiconductors, such as GaN, InGaN, and AlGaN, have become the most important material system for short wavelength light emitters in recent years. Indeed, GaN-based blue and green light emitting diodes (LEDs) are extensively used in our daily life [1]–[3]. GaN-based high power white LEDs with large chip-size are also used for solid-state lighting. However, these large-sized LED chips suffer from poor current spreading. Such poor current spreading can result in enhanced “efficiency droop” when the LEDs are injected with high current [4]. This could be solved by forming multiple micro-LEDs on a large area chip. By properly connecting these micro-LEDs, one can realize self-rectified alternating-current LEDs (AC-LEDs) [5]. However, the output power of AC-LEDs is relatively small since only half of the active area is driven. It has also been shown that GaOx oxidation grains will be generated when the micro-LEDs are negatively biased [6]. This could degrade the performance and eventually results in failure of the AC-LEDs. Alternatively, one can form series-connected micro-LEDs on a large area chip. These series-connected micro-LEDs could result in high forward voltage under direct current (DC)

Manuscript received January 21, 2014; revised March 23, 2014; accepted March 29, 2014. Date of publication April 4, 2014; date of current version April 29, 2014. This work was supported in part by the Advanced Optoelectronic Technology Center, National Cheng Kung University, through the Ministry of Education, Taiwan, and in part by the Bureau of Energy, Ministry of Economic Affairs of Taiwan, under Contract 102-E0603. S.-J. Chang and C.-Y. Chang are with the Advanced Optoelectronic Technology Center, Institute of Microelectronics and Department of Electrical Engineering, National Cheng Kung University, Tainan 70101, Taiwan (e-mail: [email protected]). C.-L. Tseng, C.-S. Shen, and B.-Y. Chen are with Epistar Corporation, Tainan 744, Taiwan (e-mail: [email protected]; [email protected]; [email protected]). Color versions of one or more of the figures in this letter are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/LPT.2014.2314701

Samples used in this letter were all grown by metalorganic chemical vapor deposition (MOCVD) on c-plane cone-shaped patterned sapphire substrate (PSS) [7], [8]. The diameter and height of the cones were 2.6 and 1.55 µm, respectively, while the spacing between the cones was 0.4 µm. The structure of the sample consists of a 25 nm thick low-temperature GaN nucleation layer grown at 560°C, a 2 µm thick undoped GaN layer grown at 1050°C, a 2.5 µm thick Si-doped n-GaN layer grown at 1050°C, a Si-doped strain releasing multiquantum well (SRMQW) grown at 900°C, an undoped light emitting MQW active region grown at 770°C, a 25 nm thick Mg-doped p-Al0.15Ga0.85 N electron blocking layer grown at 1050°C, a 0.22 µm thick Mg-doped p-GaN contact layer grown at 1050°C and a Si-doped n+ -InGaN/GaN short period superlattice tunnel contact structure, (SPS). The SRMQW consists of 20 periods of 1.2 nm thick In0.04Ga0.96 N well layers and 3 nm thick GaN barrier layers. The InGaN/GaN MQW active region consists of 8 periods of 3 nm thick In0.22 Ga0.78 N well layers and 8 nm thick GaN barrier layers. On the other hand, the SPS structure consists of four pairs of 5 Å thick In0.23 Ga0.77 N layers and 5 Å thick GaN layers [9]. These LED wafers were then processed with standard procedures [10], [11]. An inductively coupled plasma (ICP) etcher was used to form the isolation trenches. Cr/Al/(Ti/Pt)x3/Au (3/200/(100/50)x3/3000 nm) was then e-gun evaporated to serve as the connection bridges and metal contacts. Fig. 1(a) shows an optical microscopic image of a fabricated HV-LED chip. It can be seen that 16 micro-LEDs connected in series were formed in the 1125 µm × 1125 µm HV-LED chip. The blue lines shown in fig. 1(a) are the Cr/Al/Ti/Pt/Au metal stack lines. It should be noted that the designated line widths were 5 and 30 µm for the metal contacts and the connection bridges, respectively. Fig. 1(b) shows the electroluminescence (EL) intensity distribution of the HV-LED chip driven at 20 mA. It can be seen that

1041-1135 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

1074

IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 26, NO. 11, JUNE 1, 2014

Fig. 1. (Color online) (a) Optical microscopic images taken from the HV-LED chip. (b) EL intensity distribution of the HV-LED chip driven at 20 mA.

the EL intensity was reasonably uniform across the entire chip. This indicates good current spreading for the fabricated device. With 20 mA current injection, it was found from I-V measurement that the forward voltage of each micro-LED was around 3.0 V while the forward voltage of the entire HV-LED chip was around 49 V. To analyze failure mechanism, a burn in test was performed by continuously driving these HV-LED chips with 100 mA at 80°C. Scanning electron microscopy (SEM), energy dispersive spectrometer (EDS) and focused ion beam (FIB) were then used to evaluate properties of the HV-LED chip before and after the burn-in test.

Fig. 2. (a) SEM image taken from two micro-LEDs separated by an isolation trench. (b) Enlarged SEM image of the same sample showing a whisker protruding from the p-metal.

Fig. 3. (Color online) SEM images taken from FIB treated cross-sections of (a) area 1 in Fig. 2(a) (i.e., away from the trench) and (b) area 2 in Fig. 2(a) (i.e., near the trench).

III. R ESULTS AND D ISCUSSION With 100 mA current injection at 80°C, it was found that a whisker started to appear from the sidewall of the Cr/Al/Ti/Au p-contact metal after 120-hour burn-in test. It was also found that the whisker became longer as we increased the burnin time. Such continuous growth indicates that the whisker was formed during the burn-in test, instead of formed during device processing. To show the results are reproducible, we prepared 10 wafers with the same epitaxial structure. After device processing, we randomly picked 5 HV-LED chips from each of these 10 wafers. It was found that similar whiskers were observed from all these HV-LED chips after the same 100 mA and 80°C burn-in test. It should be noted that the whiskers were all observed at about the same position (i.e., where the p-finger connects to the bridge). Fig. 2(a) shows a SEM image taken from two micro-LEDs separated by an isolation trench. The measurement was performed after 300-hour burn-in test. Fig. 2(b) shows an enlarged SEM image of the same sample. It can be seen clearly that the whisker indeed emerges from the sidewall of the Cr/Al/Ti/Pt/Au p-type metal. It should be noted that similar whiskers were also observed from other micro-LEDs. To understand the origin of the whisker, we used FIB to mill two different areas of the same micro-LED. These two areas are labeled in Fig. 2(a). Fig. 3(a) and 3(b) show SEM images taken from FIB treated cross-sections of area 1 in Fig. 2(a) (i.e., away from the trench) and area 2 in Fig. 2(a) (i.e., near the trench), respectively. As shown in Fig. 3(a), it can be seen that the Cr/Al/Ti/Au metal stack in area 1 (i.e., away from the trench) remained smooth after the 300-hour burn-in test. In contrast, it was found that the Al layer in the metal stack was inflated and a whisker was extended outward from the Al layer in area 2 (i.e., near the trench), as shown in Fig. 3(b).

Fig. 4. (Color online) (a) SEM images taken from FIB treated cross-section near the trench. (b) EDS line-scanning spectra measured along the arrow shown in Fig. 4(a).

For comparison, we also took the SEM images from the FIB treated samples without burn-in test. Similar to Fig. 3(a), it was found that we observed smooth Cr/Al/Ti/Au metal stack in both area 1 and area 2 for the samples without burn-in test. Such an observation indicates again that the whiskers were formed during the burn-in test. EDS was then used to analyze the whisker along the arrow shown in Fig. 4(a). Fig. 4(b) shows EDS linescanning spectra measured along the arrow shown in Fig. 4(a). It can be seen that the whisker predominately consists of Al. Fig. 5 shows an enlarged SEM image. In this figure, the area with a redcross was the observed whisker. Table I lists the atomic ratios determined from the EDS spectra which were measured from the red-cross marked in Fig. 5. It can be seen again that the whisker was mainly composed of Al. This also agrees well with those observed from Fig. 3(b) and Fig. 4(b). In this letter, the Al layer was deposited by e-gun evaporation. Thus, the deposited Al should be polycrystalline

CHANG et al.: FAILURE MECHANISM FOR GaN-BASED HV-LEDs

1075

Fig. 5. (Color online) An enlarged SEM image. The area with a red-cross was the observed whisker. TABLE I ATOMIC R ATIOS D ETERMINED BY THE EDS S PECTRA M EASURED F ROM THE R ED -C ROSS S HOWN IN F IG . 5

with numerous grain boundaries, instead of single crystalline. During the burn-in test at high temperature and high injection current, the Al atoms might diffuse through these grain boundaries. Such electromigration effect is well-known and often observed in integrated circuit chips [12]. To clarify the origin of the formation of the Al whisker, we also performed burn-in tests under three other conditions. We either continuously injected 20 mA at 25°C, or 100 mA at 25°C, or 20 mA at 80°C. However, no formation of Al whisker was observed under these three burn-in conditions. In other words, the Al whisker was observed only when the HV-LED chips were burn-in tested at high temperature with high injection current. If we assume that the injected current (i.e., 100 mA) was all conducted through the 200 nm thick Al layer, the current density in the 5 µm wide finger should be: 100 mA/(200 nm × 5 µm) = 10 A/cm . . . . 7

2

(1)

At temperature about 100°C, it is known that Al electromigration occurs when the current density is larger than 105 A/cm2 [13]. Thus, the 107 A/cm2 current density calculated in this letter should be able to induce Al electromigration easily at 80°C. As indicated previously, the isolation trenches were formed by ICP etching. Thus, the sidewalls of these trenches should be nearly vertical (i.e., around 80°, as can be seen from Fig. 3(a)). This will inevitably cause step coverage problems for the subsequently e-gun deposited metal stack. In other words, the actual thickness of the Al layer will be smaller than the designated 200 nm at the sidewall of the trenches (i.e., connection bridges). Fig. 6 shows top view SEM image of the fabricated HV-LEDs. The arrows show the direction of electron flow in these chips. As indicated by the arrows,

Fig. 6. (Color online) Top view SEM image of the fabricated HV-LED chips. The arrows show the direction of electron flow in these chips.

the electrons flowed through the internal layers of the microLEDs were collected by the p-type metal and subsequently entered the connection bridge. With a thinner metal layer, the current density will become larger. This could result in enhanced Al electromigration near the trenches. Thus, internal stress will build up as the Al atoms pile up at the p-contact metal. As a result, these extra Al atoms will eventually penetrate through the passivation layers to form the whisker, as shown in Fig. 2(b). Above examination also supports our observation that whiskers prefer to emerge from the position where the p finger connects to the bridge to the n finger of next micro-chip. It should be noted that the continuously growing Al whisker might result in a short circuit eventually. It is also possible that an open circuit will occur in other areas due to the loss of Al atoms. These two phenomena could both result in failure of the GaN-based HV-LEDs. IV. C ONCLUSION In summary, we report a detailed reliability test study on GaN-based HV-LEDs. Under high temperature (i.e., 80°C) and high current injection (i.e., 100 mA) conditions, it was found that Al metal whiskers form from the sidewall of the Cr/Al/Ti/Au p-contact metal during a 120-hour burn-in test. It was also found that the whiskers became longer as we increased the burn-in time. Al migration such as a failure mechanism that has to be taken into account in the design of HV-LEDs. R EFERENCES [1] S. Nakamura, M. Senoh, N. Iwasa, and S. Nagahama, “High-power InGaN single-quantum-well-structure blue and violet light-emitting diodes,” Appl. Phys. Lett., vol. 67, no. 13, pp. 1868–1870, 1995. [2] C. M. Tsai et al., “Enhanced output power in GaN-based LEDs with naturally textured surface grown by MOCVD,” IEEE Electron Device Lett., vol. 26, no. 7, pp. 464–466, Jul. 2005. [3] S. J. Chang et al., “Improved ESD protection by combining InGaN-GaN MQW LEDs with GaN Schottky diodes,” IEEE Electron Device Lett., vol. 24, no. 3, pp. 129–131, Mar. 2003. [4] M. H. Kim et al., “Origin of efficiency droop in GaN-based lightemitting diodes,” Appl. Phys. Lett., vol. 91, no. 18, p. 183507, 2007. [5] H. H. Yen, W. Y. Yeh, and H. C. Kuo, “GaN alternating current lightemitting device,” Phys. Status Solidi (A), vol. 204, no. 6, pp. 2077–2081, 2007. [6] H. H. Yen, H. C. Kuo, and W. Y. Yeh, “Particular failure mechanism of GaN-based alternating current light-emitting diode induced by GaOx oxidation,” IEEE Photon. Technol. Lett., vol. 22, no. 15, pp. 1168–1170, Aug. 1, 2010. [7] H. M. Chang, W. C. Lai, and S. J. Chang, “Effects of initial GaN growth mode on patterned sapphire on the opto-electrical characteristics of GaN-based light-emitting diodes,” J. Display Technol., vol. 9, no. 4, pp. 292–296, Apr. 2013.

1076

[8] C. F. Shen, S. J. Chang, W. S. Chen, T. K. Ko, C. T. Kuo, and S. C. Shei, “Nitride-based high-power flip-chip LED with double-side patterned sapphire substrate,” IEEE Photon. Technol. Lett., vol. 19, no. 10, pp. 780–782, May 15, 2007. [9] S. J. Chang et al., “Highly reliable nitride based LEDs with SPS+ITO upper contacts,” IEEE J. Quantum Electron., vol. 39, no. 11, pp. 1439–1443, Nov. 2003. [10] C. H. Wang et al., “Efficiency and droop improvement in GaN-based high-voltage light-emitting diodes,” IEEE Electron Device Lett., vol. 32, no. 8, pp. 1098–1100, Aug. 2011.

IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 26, NO. 11, JUNE 1, 2014

[11] R. H. Horng, K. C. Shen, Y. W. Kuo, and D. S. Wuu, “Effects of cell distance on the performance of GaN high-voltage lightemitting diodes,” ECS Solid State Lett., vol. 1, no. 5, pp. R21–R23, 2012. [12] I. Blech, “Electromigration in thin aluminum films on titanium nitride,” J. Appl. Phys., vol. 47, no. 4, pp. 1203–1208, 1976. [13] T. Wada, H. Higuchi, and T. Ajiki, “Electromigration in doublelayer metallization,” IEEE Trans. Rel., vol. 34, no. 1, pp. 2–7, Apr. 1985.