Open Access
Add to BibSonomy
Abstract
The fabrication and feasibility assessment of n-side up, thin-epilayer, AlGaInP-based vertical light-emitting-diodes (LEDs; emitting area: 1 mm × 1 mm) with a copper-invar-copper-composite metal (CIC) substrate was obtained by wafer bonding and epilayer transferring technologies. The structure of CIC substrate is a top Cu layer of 20 μm, a middle Invar layer of 64 μm, and a bottom Cu layer of 20 μm. The invar layer consists of Fe and Ni at a ratio of 70% to 30%. The coefficient of thermal expansion for CIC is about 6.1 × 10−6 /K, which is similar to that of the GaAs substrate (5.7 × 10−6 /K) and AlGaInP epilayers. Due to the high thermal conductivity (160 W/m-K) of 104-μm-thick CIC, the high performances of the packaged LEDs are obtained. They present a low red shift phenomenon (from 623 to 642 nm for 100 mA to 1 A) and a high output power 212 mW at 800 mA. The CIC substrate can be extended to fabricate high-efficiency thin film LEDs with conventional vertical electrodes.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Light-emitting diodes (LEDs) are described as solid-state self-emissive devices. The inherent physical characteristics of LEDs allow their operation at low voltages/currents with high efficiency, reliability, and a long lifespan [1]. Besides, LEDs can be operated under extreme conditions, such as extremely high or low temperatures [1]. Therefore, LEDs have a wide range of applications including indoor/outdoor illumination, automotive lighting, traffic signals, and displays. The red-light LEDs, ideal for numerous applications, are highly mature given their years of extensive development. High-quality red LEDs are based on (AlxGa1−x)0.5In0.5P whose internal quantum efficiencies exceed 90% due to being grown on precisely lattice-matched GaAs substrate by metal organic chemical vapor deposition (MOCVD) without the generation of excess dislocations [1–5]. Nevertheless, GaAs substrate is a light-absorbing material for AlGaInP LEDs, thus limiting the light extraction efficiency of red and yellow LEDs [1,5,6]. Various studies have been performed to fabricate the brightness of thin film LEDs by wafer bonding technology [1,7,8] in which light-absorbing GaAs substrate can be replaced to improve the efficiency of the LEDs [1]. Due to the problem of the light absorbing of LEDs from GaAs substrate, epilayers with a mirror structure are always transferred to Si substrate and CuW metal substrates to achieve a high brightness by wafer bonding and epilayer transferring technologies. Nevertheless, there exists a difference in thermal expansion coefficient (TEC) between Si (2.6 × 10−6 /K) or CuW (6.5 × 10−6 /K) and GaAs substrate (5.7 × 10−6 /K) [9–11]. Moreover, a thickness of about 600 μm of Si substrate with 4 inch in diameter is required to prevent the Si from breaking during processing. After that, the permanent Si substrate must experience thinning before the LEDs are diced into chips. On the other hand, concerning the CuW substrate, it is difficult to thin the CuW substrate below 150 μm and hard to dice. In addition, the CuW substrate is very expensive. High-powered AlGaInP LEDs with patterned copper substrates created by electroplating have been fabricated in our previous study [12]. Due to the large TEC difference between the Cu and AlGaInP epilayer, the epilayer is easily cracked during thermal annealing for Ohmic electrode contact processing after the AlGaInP epilayers were transferred to the Cu substrate. Thus, the parallel electrodes were fabricated before electroplating the Cu substrate [12]. Based on the above concerns, a new metal substrate, copper-invar-copper (CIC) substrate, was proposed. The TEC of CIC can be tuned to match those of GaAs and AlGaInP. The thermal conductivity of CIC is about 168 W/mK. It can compete with those of Si (130 W/mK) and CuW (180 W/mK) [10,11]. Moreover, the thickness of CIC is only 100 μm. This means that when the epilayer is transferred to the CIC substrate, it can be directly fabricated into an LED chip without the thinning process. In this study, the CIC was evaluated for the possibility to transfer an AlGaInP epilayer to replace the Si and CuW substrates for high power, thin-film red light LED applications.
2. Experiments
An AlGaInP LED epitaxial structure was grown by MOCVD. The structure consisted of a 0.5 μm thick GaInP etching stop layer, a 100 nm thick n+-GaAs contact layer, a 3.5 μm thick n-cladding AlGaInP, 0.5 μm thick GaInP-AlGaInP MQWs, a 0.6 μm thick p-cladding AlGaInP, a 2.5 μm thick p-GaP: Mg window layer and a 100 nm p+-GaAs Ohmic contact layer. After surface cleaning, metal Au/AuBe/Au was deposited with 0.1/0.2/2 μm thicknesses by a thermal evaporator. These metals not only create the Ohmic contact with the p+-GaAs, but are also the mirror and bonding metal.
On the other hand, the CIC metal was first cleaned, and then bonding metal Sn with 2 μm thickness was immediately deposited by the thermal deposited system after the CIC treatment. Then, the CIC with Sn metal and AlGaInP LED epiwafers with Au/AuBe/Au metal was bonded together at 300°C using thermal pressure with 400 Kg for 1 hour. The GaAs substrate was removed and stopped at the GaInP epilayer by wet etchant NH4OH:H2O2. After, GaInP was removed using HCl:H3PO4, and Au/AuGe/Au metals were deposited on the n + GaAs contact layer. Then, the whole wafer was thermally annealed at 340°C for 1 min by rapid thermal annealing for the electrode Ohmic contact. Finally, the epilayers were etched into isolation and diced into 1mm × 1mm dice by laser cutting. It is worth mentioning that the LED epilayer with 104 μm CIC substrate can stand for thermal annealing processing without epilayer cracking.
In this study, the TEC of CIC was measured by a Thermo Mechanical Analyzer (TMA). Measurements of the linear TECs of the CIC substrate and reference material were performed in a TMA, Model 7100, Hitachi, interfaced to a PC for data acquisition and data analysis [13–15]. The measurements were performed in expansion mode with an applied force of 50 mN [14]. The minimal 50 mN applied static load required by measurement system did not affect the TEC results. Dry nitrogen gas flowed through the TMA furnace at a rate of 100 ml/min. The curve of the CIC with an epilayer was measured by the thin film stress measurement system (FXL 2320-S, Toho Technology). The current-voltage (I–V) characteristics of the LED with a CIC substrate were measured at room temperature using an Agilent 4155B semiconductor parameter analyzer. All of the LED chips were packaged for the output power measurement using an integrating sphere detector (CAS 140B, Instrument Systems). All measured LEDs with CIC substrate data were averaged from 50 different samples.
3. Results and discussion
Figure 1 shows the cross-section measured by SEM for CIC substrate. The thickness of the CIC is about 104 μm. From the cross-section, it is obvious that the structure of the CIC substrate is a sandwich structure; the top layer is 20 μm thick Cu, the middle layer is 64 μm thick Invar, and the bottom layer is 20 μm thick Cu. The invar consists of Fe and Ni with a ratio of 70% to 30%. The top and bottom Cu layers were used to tune the TEC of the CIC to match with the GaAs substrate.
The TEC of CIC was measured by a TMA. Figure 2 shows the expansion displacement of CIC as a function of temperature. If the TEC difference is large between the GaAs substrate and the CIC, the CIC with the LED epilayer will bend after removing the GaAs substrate. CIC substrate was subjected to a constant load of 50 mN, and measurements were taken for temperatures from 25°C to 300°C for the heating portion of the cycle and from 300°C to 25°C for the cooling portion of the cycle at a sweep rate of 5°C. It was found that the deflection from 25°C to 75°C presented as nonlinear. This resulted from the thermal stress being less than 50 mN. Nevertheless, the TEC was calculated from 30°C to 300°C because the temperature of the wafer bonding was 300°C and the wafer pair cooled down to 30°C. The mean linear TEC, , can be calculated from the slope of the curve shown in Fig. 2, as follows:
where L0 is the initial thickness of CIC at 25°C, ΔL is the change in thickness and ΔT is the change in temperature between the two equilibration temperatures. This α will be the average TEC for this particular temperature interval [13,14]. Here, L0 is about 104 μm, the corresponding deflections for 25°C and 300°C are −0.03025 and 0.144302 μm, respectively. This means that ∆L and ∆T are 0.174552 μm and 275°C, respectively. After calculation, the TEC of CIC is about 6.1 × 10−6/K. This is very close to the TECs of GaAs (5.7 × 10−6/K) and AlGaInP (5 × 10−6/K). Correspondingly, the TECs of Si and CuW are 2.6 × 10−6/K and 6.5 × 10−6/K, respectively. The inset in Fig. 2 shows the side view of the epilayer/CIC. Obviously, the CIC with an epilayer is very flat, and most of the epilayer can be transferred to the CIC substrate. These results suggest that the TEC of CIC is matched to the GaAs substrate even if the CIC thickness is only 104 μm.
Fig. 2 Expansion displacement of CIC as function of temperature. The inset is the side view of the AlGaInP transferring on the CIC substrate.
It is well known that Cu is an easily oxidized material, and the surface is always coated with a protective layer to prevent oxidization. Thus, the surface treatment of CIC is very important for the adhesion strength of the bonding metal. There are five methods for surface treatment. In the first method, CIC is cleaned with acetone (ACE) for 10 min to remove the protective layer, after which, the measured surface contact angle of the CIC substrate is 76.00°. In the second method, CIC is cleaned by ACE for 10 min and etched with a dilute HF solution for 10 min. The obtained surface contact angle is 55.71°. In the third method, treated CIC is cleaned by ACE for 10 min and etched with dilute HCl for 5 min, after which the measured surface contact angle is 43.70°. In the fourth method, CIC is cleaned by ACE for 20 min and etched with dilute HCl for 5 min, obtaining a contact angle of 27.47°. The optimized method is CIC being cleaned by ACE for 30 min and etched with dilute HCl (HCl:DI water = 1:10) solution for 10 min to remove the protected layer. The measured surface contact angle of the CIC substrate is 14.41°. Details on the treatments for the CIC and the corresponding contact angles after the surface treatments are shown in Fig. 3.
In order to evaluate the effect of surface treatments on the adhesive strength of the bonding metal, the bonding metal Sn was deposited on the CIC substrate with and without surface treatment. It is worthy to mention that the Sn was deposited on CIC substrate immediately after the surface treatment using ACE for 30 min and HCl for 10 min (shown in Fig. 3(e)). After that, the 3M tape was applied to the Sn/CIC and used to measure the adhesive strength of Sn on CIC. Figures 4(a) and 4(b) show the adhesive strength of the bonding metal Sn with CIC substrate. Obviously, the Sn metal can easily be removed from the CIC without surface treatment. If the organic contamination can be removed, it will affect the adhesive strength of the bonding metal Sn on CIC. The adhesive strength of Sn deposited on the CIC after optimized cleaning with a small contact angle is strong enough to resist the 3M tape test without any peeling off which can represent LED processing. A 3M scotch magic tape was used to perform the peeling test. The adhesion is found to be strong enough for the epilayer transferring to the treated CIC substrate with contact angles less than 30°.
Fig. 4 (a) Low adhesive strength of bonding metal Sn with CIC substrate without surface treatment and (b) Strong adhesion of bonding metal Sn with CIC substrate with surface treatment.
The measurement of the radius and curve of the CIC before and after wafer bonding can be used to demonstrate whether the TEC of CIC is as effective as that of the GaAs substrate. The radius and curve of the CIC with a 4-inches diameter before wafer bonding compared with those of the CIC after wafer bonding and the GaAs substrate being removed are shown in Figs. 5(a) and 5(b). The radius and bow are 84.131 m and −9.57 μm for a 4-inch CIC substrate. Correspondingly, the radius and bow are −6.132 m and 128.40 μm for a 4-inch CIC substrate with an AlGaInP LED epilayer structure. The radius with 84.131 m for the bare CIC substrate means the metal substrate is very flat. The changed radius from 84.131 m to −6.132 m for the CIC before and after wafer bonding suggests that there still exists a small TEC between the CIC and GaAs substrate. Because the TEC of CIC is greater than that of an AlGaInP epilayer, it results in a concave curvature. Even though, it only results in a 128.40 μm bow for CIC with AlGaInP epilayers. The bowing is very small and should not affect the LED processing.
Fig. 5 (a) Radius and curve of CIC with 4-inch diameter before wafer bonding and (b) Radius and curve of CIC with 4-inch diameter after wafer bonding.
The most important factor is the LED performance after the epilayer is transferred to a CIC substrate. The forward and reverse currents of an LED chip as a function of the voltage are shown in Fig. 6. The voltages of LED chips are about 1.8 V and 2.25 V with the injection current at 20 mA and 350 mA, respectively. The reverse current is about 10−7 A with the reverse voltage at −5V. These results suggest that the LED epilayer transferred to a CIC substrate can offer good electrical properties. Concerning the optical performance of the LED/CIC, the output power of the LED/CIC chip as a function of the current was measured and is shown in Fig. 7. Obviously, the output power presents a linear relation with the injection current up to 800 mA and then, saturation. The saturated output power resulted from the current density being too high for the 40 mil chip LEDs. Correspondingly, the saturated currents of LEDs with Si substrate or LEDs with Cu substrate occurred at about 500 mA. It could be due to the CIC being only 100 um and Si with 150 μm for the thin film AlGaInP LEDs. Thermal dissipation is related to the chip substrate. If the thickness is thinner, it would result in the higher thermal dissipation as the thermal conductivity of CIC and Si are very similar. This suggests that the thermal dissipation of CIC can compete with those of Si and Cu substrates. Furthermore, a demonstration of the CIC offers good thermal dissipation in the wavelength variation as shown in the inset of Fig. 7. It was found that the wavelength changed from 623 to 642 nm as the current injected from 100 mA to 1A. These behaviors are almost the same or better than previous reports on LEDs with Si substrate.
Fig. 6 I-V characteristics of LEDs/CIC. Inset shows the reverse currents of an LED chip as a function of the voltage.
Fig. 7 Output power of LEDs/CIC as function of current. Inset shows the wavelength-current characteristics.
4. Conclusion
In this paper, AlGaInP LED epitaxial films were successfully transferred on composite metal substrate CIC by wafer bonding and epilayer transferring technologies. From the properties of CIC measurement, the CIC substrate showed that the TEC matched to the GaAs substrate and AlGaInP epilayer and resulted in low bending for the epilayer/CIC substrate. On the other hand, the CIC substrate should be cleaned by ACE and acid to improve the mechanical strength of bonding metal on CIC. The mechanical strength exceeds that required to withstand the LED backend processes, such as chemical etching, Ohmic contact processing, and chip dicing. Using CIC to replace the original GaAs substrate not only improved the heat dissipation of the LED itself, but also prevented the red light from being absorbed by the GaAs substrate. After AlGaInP LED transferring to the CIC substrate, the thin film LEDs not only presented normal electrical properties, but also presented a high performance of optical properties. The thin film AlGaInP LEDs with CIC substrate can be operated at high injected current, a low red shift phenomenon (from 623 to 642 nm for 100 mA to 1 A) and a high output power (212 mW at 800 mA). These results suggest that the CIC substrate can replace the Si substrate and CuW substrate and be applied to thin film AlGaInP LED applications.
Funding
Ministry of Science and Technology, Taiwan (107-2221-E-009-117-MY3, 107-2262-E-009-018-CC2, 107-2218-E-009-056); Ministry of Education; Hsinchu Science Park, Taiwan (106A03).
Acknowledgments
This study was supported by the Research Team of Photonic Technologies and the Intelligent System at NCTU within the framework of the Higher Education Sprout Project by the Ministry of Education.
References
1. R. H. Horng, H. Y. Chien, K. Y. Chen, W. Y. Tseng, Y. T. Tsai, and F. G. Tarn Tair, “Development and Fabrication of AlGaInP-Based Flip-Chip Micro-LEDs,” IEEE J. Electron Devi. 6, 475–479 (2018).
2. I. Schnitzer, E. Yablonovitch, C. Caneau, and T. J. Gmitter, “Ultrahigh spontaneous emission quantum efficiency 99.7% internally and 72% externally from AlGaAs/ GaAs/AlGaAs double hetero-structures,” Appl. Phys. Lett. 62(2), 131–133 (1993). [CrossRef]
3. C. P. Kuo, R. M. Fletcher, T. D. Osentowski, M. C. Lardizabal, M. G. Craford, and V. M. Robbins, “High performance AlGaInP visible light-emitting diodes,” Appl. Phys. Lett. 57(27), 2937–2939 (1990). [CrossRef]
4. H. Sugawara, M. Ishikawa, and G. Hatakoshi, “High-efficiency InGaAlP/GaAs visible light-emitting diodes,” Appl. Phys. Lett. 58(10), 1010–1012 (1991). [CrossRef]
5. T. Gessmann and E. F. Schubert, “High-efficiency AlGaInP light emitting diodes for solid-state lighting applications,” J. Appl. Phys. 95(5), 2203–2216 (2004). [CrossRef]
6. M. R. Krames, M. Ochiai-Holcomb, G. E. Hofler, C. Carter-Coman, E. I. Chen, I.-H. Tan, P. Grillot, N. F. Gardner, H. C. Chui, J.-W. Huang, S. A. Stockman, F. A. Kish, M. G. Craford, T. S. Tan, C. P. Kocot, M. Hueschen, J. Posselt, B. Loh, G. Sasser, and D. Collins, “High-power truncated-inverted-pyramid (AlxGa1-x)0.5In0.5P/GaP, Light emitting diodes exhibiting >50% external quantum efficiency,” Appl. Phys. Lett. 75(16), 2365–2367 (1999). [CrossRef]
7. M. C. Tseng, C. L. Chen, N. K. Lai, S. I. Chen, T. C. Hsu, Y. R. Peng, and R. H. Horng, “P-side-up thin-film AlGaInP-based light emitting diodes with direct ohmic contact of an ITO layer with a GaP window layer,” Opt. Express 22(S7Suppl 7), A1862–A1867 (2014). [CrossRef] [PubMed]
8. Z. Zuo, R. Wang, D. Liu, Q. Yu, Z. Feng, and X. Xu, “Ultra bright AlGaInP light emitting diodes with pyramidally patterned GaP spreading layer,” Appl. Phys. Lett. 99(12), 121104 (2011). [CrossRef]
9. M. R. Krames, O. B. Shchekin, R. Mueller-Mach, G. O. Mueller, L. Zhou, G. Harbers, and M. G. Craford, “Status and future of high-power light-emitting diodes for solid-state Lighting,” J. Disp. Technol. 3(2), 160–175 (2007). [CrossRef]
10. F. L. Wu, S. L. Ou, Y. C. Kao, C. L. Chen, M. C. Tseng, F. C. Lu, M. T. Lin, and R. H. Horng, “Thin-film vertical-type AlGaInP LEDs fabricated by epitaxial lift-off process via the patterned design of Cu substrate,” Opt. Express 23(14), 18156–18165 (2015). [CrossRef] [PubMed]
11. D. Morita, M. Sano, M. Yamamoto, T. Murayama, S. I. Nagahama, and T. Mukai, “High Output Power 365 nm Ultraviolet Light Emitting Diode of GaN-Free Structure,” Jpn. J. Appl. Phys. 41(12B), L1434–L1436 (2002). [CrossRef]
12. R. H. Horng, C. E. Lee, C. Y. Kung, S. H. Huang, and D. S. Wuu, “High-power AlGaInP light-emitting diodes with patterned copper substrates by electroplating,” Jpn. J. Appl. Phys. 43(4B4B), L576–L578 (2004). [CrossRef]
13. S. Pitely, L. Zavalij, S. Zarembo, and E. J. Cotts, “Linear coefficients of thermal expansion of Au0.5Ni0.5Sn4, Au0.75Ni0.25Sn4, and AuSn4,” Scr. Mater. 51(7), 745–749 (2004). [CrossRef]
14. R. E. Taylor and C. Y. Ho, Thermal expansion of solids: CINDAS Data series on material properties, 1–4 (ASM International, 1998).
15. Y. S. Touloukian, Thermal expansion: metallic elements and alloys (IFI/Plenum, 1975).
