A GaN-based light-emitting diode (LED) with a direct-Ohmic contact structure, formed by an indium-tin-oxide (ITO) transparent film and Au thermal-diffused and removed layer, is studied. By depositing an Au metallic film on the Mg-doped GaN layer followed by thermal annealing and removed processes, an ITO direct-Ohmic contact at p-GaN/ITO interface is formed. An enhanced light output power of 18.0% is also found at this condition. This is mainly attributed to the larger and more uniform light-emission area resulted from the improved current spreading capability by the use of an ITO direct-Ohmic contact structure.
© 2011 OSA
InGaN/GaN multiple-quantum-well (MQW) light-emitting diodes (LEDs) with lateral geometries grown by metal-organic chemical vapor deposition (MOCVD) are very attractive and give variable applications covering visible emission from yellow-green to blue spectra. Recently, due to the significant improvement on epitaxial technique of nitride-based material system, the perfect crystalline quality of InGaN/GaN-based LEDs has become possible. However, the current-crowding phenomenon  still limits their light output efficiency. This drawback is known to be mainly originated from the high-resistive p-GaN contact layer , due to the high activation energy of Mg dopants and low acceptors (holes) mobility. A higher operation voltage of a GaN-based LED is inevitable because of the Schottky-like characteristic of the p-GaN contact layer, and the induced joule-heating effect underneath n- and p- metal pads deteriorates optical, electrical properties, and the reliability of LEDs .
Up to date, the indium tin oxide (ITO) layer has been reported to enhance the current spreading performance of GaN-based LEDs . However, the poor Ohmic nature between the ITO current-spreading layer and high-resistive p-GaN contact layer [3, 4] seems still remain a miserable problem. Recently, many approaches have been raised trying to enhance this Ohmic contact property. For example, one could apply an InGaN/GaN superlattice (SL) structure above (increases hole concentration for contact) /underlying (modulate current and release strains)  the p-GaN contact layer, or an InGaN capping layer on the p-GaN layer  to lower down the contact resistivity. However, these methods cause a relatively complicated growth process and high cost of LEDs. Without complicating epitaxial growth conditions, a lower contact resistivity was achieved by an In2O3/ITO structure  studied by our group. Nevertheless, the p-GaN Ohmic contact property needs to be further improved.
One could find some significant progresses related to enhancing p-GaN Ohmic contact by means of a Ni/Au current-spreading layer  or a Pd contact layer with a surface treatment by aqua regia solution . However, many extracted photons from active region will be sacrificed underlying these opaque metal contact/current-spreading layers. In this work, based on an Au thermal-diffused and removed thin layer, an easy and feasible method to form an ITO direct-Ohmic contact characteristic on the p-GaN layer is implemented, without significant photon loss within blue spectra. The corresponding mechanism will be explored in detail.
The studied LED device (denoted as device B) was grown by a Thomas Swan metal organic chemical vapor deposition (MOCVD) system on a c-plane (0001) oriented sapphire substrate with a 0.2° off toward (1-100) plane. The epitaxial structures consisted of 30 nm GaN buffer layer, 2 μm-thick undoped GaN layer, 2 μm-thick Si-doped n-GaN layer (n=1×1018cm−3), 12-period In0.2GaN/GaN (3 nm/12 nm) MQW as active layers, followed by a 0.3 μm-thick Mg-doped p-GaN layer (p=4×1017cm−3). Then, the as-grown wafers were first cleaned by acetone and deionized water sequentially. An Au metal thin film (10 nm in thickness) was thermally evaporated on the p-GaN contact layer. In order to diffuse Au atoms into p-GaN layer, samples were annealed in N2 ambient at 400°C for 30 min. in a quartz furnace. Referring to the metal thermal-annealing followed by removed processes, different kinds of metal films (such as Ni, Cr, and Au) with the same thickness of 1 nm treated by various annealing temperatures (200, 300, 400, and 470°C) and times (10, 20, 30, and 40 min.) have been carried for choosing a better specific contact resistance for p-GaN/ITO contacts, as shown in Fig. 1 . Generally, it was found that optimized thermal-annealing conditions are 400°C for 10 min., 470°C for 20 min., and 400°C for 30 min. for Ni, Cr, and Au, respectively. Of all these, the better contact resistance is found for Au metal, at thermal-annealing conditions of 400°C for 30 min.. This result may be ascribed to the facts that gallium exhibits higher soluble in Au rather than other metals  including Ni and Cr. The residual Au metal film after annealing process was totally removed by using aqua regia solution (HNO3: HCl = 1:1) for 5 min.. For comparison, the conventional device (device A) which did not need Au film evaporation as well as the followed thermal annealing processes, was dipped in the same aqua regia solution for the same etching time. In fact, for the device B, the residual Au film could be removed completely within etching time of 30 sec., but a longer etching time is chosen to keep p-GaN layers of devices A and B exposing to aqua regia solution for nearly the same time. This progress is very important to verify that the improved p-GaN Ohmic contact property for the device B is not related to a pure surface treatment by the aqua regia solution proposed by Lee et al. . Also note that the epitaxial structures of these two devices need to be grown for the same run. After the wet-etching process, top-view high resolution scanning electron microscope techniques were applied to examine whether all Au residuals were removed or not on the p-GaN layer. Prior to chip processes, X-ray diffraction (XRD), Hall measurements, room temperature micro-Photoluminescence (RT μ-PL, Jobin-Yvon U1000) techniques, secondary ion mass spectrometry (SIMS), and X-ray photoelectron spectroscopy (XPS, PHI Quantera SXM) were used to analyze the exposed p-GaN surface. An inductively coupled plasma (ICP) system was utilized to define mesa regions. A 250 nm-thick ITO layer was deposited on the p-GaN layer by an electronic beam evaporator. Cr/Pt/Au metal was deposited as n- and p- pad Ohmic contacts. The ITO layer and n-p pads were activated for 30 min. in N2 ambient at 470°C and 380°C, respectively. These wafers were diced into individual chips with the dimension of 300 × 300μm2. The current-voltage (I-V) characteristics of devices were measured at room temperature by a semiconductor parameter analyzer (HP 4156). The specific contact resistances of p-pad/ITO/p-GaN interface were obtained by square
transmission line model (s-TLM). The light output power was measured by a Si photopic detector integrated with a current source. Dominant wavelengths are around 457 nm for both devices at 20 mA, determined by electroluminescence spectra.
3. Results and discussion
Figure 2 shows Au4f and Ga3d XPS analyses for the device B before and after 400°C thermal annealing process. A location of “0.0 nm depth” in this figure is defined by the p-GaN contact surface after removal of Au metal film. Clearly, after thermal annealing, Au atoms exhibit an in-diffused distance of around 2 nm. Note that a shallower in-diffusion width of around 1 nm is still found before annealing, probably due to thermal-annealed effect during metal evaporation. Moreover, a gallium out-diffused phenomenon is found with about 1 nm width after annealing. This indicates that an Au in- and Ga out-diffused (AuinGaout) layer with thickness of 1 nm is formed after 400°C annealing, accompanied by presence of many gallium vacancies which could be regarded as deep-level acceptors  on the p-GaN contact layer. Furthermore, the lower Ga3d intensity (20% reduction) on the p-GaN surface for the device B after annealing than that for the device A also suggests this finding. According to Hall measurement results for studied devices A and B summarized in Table 1 , the holes concentration, mobility, resistivity, and sheet resistance of the device A (B) are 3.15×1017 (1.83×1018) cm−3, 13.1 (13.0) cm2/V.s, 1.53 (0.26) Ω-cm, and 5.1×104 (8.6×103) Ω/□. These results demonstrate that a gallium-vacancy-rich p-GaN contact layer is formed after the Au thermal-diffused and removed processes. In other words, these gallium vacancies contribute to the remarkable increase in holes concentration and reduction in resistivity. In Fig. 3 , PL analyses are performed. It is found that the MQW (450 nm) and defect-related (560 nm) bands for both devices are similar, except for a slight reduction in intensity of p-GaN impurity (around 380 nm) band for the device B. This indicates that the AuinGaout layer in the device B has little influence on emission property and epitaxial quality. Only slight light absorption at 360-400 nm wavelength region induced by Au atoms occurs within the shallow (about 2 nm, underlying where less Au atoms can reach (see Fig. 2)) p-GaN layer. Furthermore, based on previous reports [11,12], the formation tendency of this AuinGaout layer possibly could be attributed to the coupling effect of GaAu2 inter-metallic compound, and/or related to the more rapidly diffusing Au species. To confirm that, XRD analyses of θ–2θ scan (Gonio scan) are performed. It is found that both devices show the same satellite peaks of GaN (0002) at 34.9° (2θ), GaN (0004) at 73.2°, and sapphire (0001) at 42.1°. However, only the device B exhibits a crystal peak at 16.5°, which is very close to that of the GaAu2 compound at 16.1° . Note that this peak also could not be observed for the device B before Au thermal-diffused and removed processes. It indicates that this peak detection doesn’t belong to a pure data variation between samples. Previously Lee et al.  also revealed that an Ni/Au Ohmic contact to the p-GaN surface was achieved after thermal annealing, due to the creation of many gallium vacancies, and this could be explained by the fact that Ga exhibits high soluble in Au . Therefore, obviously this mechanism could be directly referred to our result. In fact, according to Waki et al.’s researches, the reason for higher holes concentration on p-GaN contact layer for the device B may also be explained by the metal (Ni) catalytic effect to achieve hydrogen desorption . However, the similar hydrogen SIMS profiles (not shown) for both devices in this work seem don’t support this hypothesis.
Figure 4 depicts p-GaN/ITO/p-pad Ohmic contact properties (derived from I-V curves in Fig. 5 ) of devices A and B from −5V to 5V. These characteristics are measured directly from two p-contacts on the same mesa regions respectively, as referred in the schematic diagram of the device B in the inset of Fig. 4(where the location of AuinGaout layer is also drawn). Due to the lack of an AuinGaout layer which significantly improves Ohmic contact between p-GaN and ITO layers, the device A still exhibits poor voltage-dependent contact resistances ranging from 44.7 to 67.1 KΩ. On the other hand, Ohmic-like contact resistances from 11.8 to 12.2 KΩ are found for the device B. The corresponding specific contact resistances are evaluated to be 6.60×10−3 and 2.30×10−3 for devices A and B respectively, based on s-TLM. In other words, with a help of an AuinGaout layer on the p-GaN contact layer, the higher holes concentration and improved ITO direct-Ohmic contact property have become possible.
Figure 6 shows forward-biased I-V curves of devices A and B. The detailed I-V relations are also depicted in the inset. At 20 mA operation current, the devices A and B exhibit turn-on voltages (Vf) of 3.25 and 3.07 V, respectively. The 0.18 V reduction in Vf for the device B than the device A could be explained by the improved ITO direct-Ohmic contact property which relieves parasitic contact resistance [3, 14] between p-GaN and ITO layers. Basically this is ascribed to the higher hole concentration resulted from increased density of gallium vacancies on the p-GaN contact layer, as mentioned above. In addition, since the current injected from p-pad is effectively spread through the ITO current spreading layer due to the good Ohmic contact behavior, a lower dynamic resistance could also be expected. The corresponding dynamic resistances for both devices are shown in Fig. 7 . Clearly the dynamic resistances of the device B are always lower than those of the device A from 10 to 50 mA current operation region, suggesting that the p-GaN/ITO/p-pad contact resistance plays a significant role in device performance. At 20 mA, the device A exhibits dynamic resistance of 25.6 Ω, while device B only 18.0 Ω. The nearly 30% reduction in dynamic resistance is very important to shrinkage joule heating effect and increase power efficiency of LEDs.
Figure 8 demonstrates light output power as a function of injection current. At 20, 100, and 150 mA, light output power of 10.0 (11.8), 30.5 (37.4), and 38 (47.5) mW for the device A (B) are obtained, and the corresponding output power improvement are 18.0, 22.6, and 25.0%, respectively. It is clear that the improvement is increased with increasing current injection. We reasonably conclude that this mainly originates from the enhanced internal quantum efficiency of the device B, since both devices were designed as the same geometrical structure which confirms tiny variation in light extraction efficiency. In other words, the LED with an ITO direct-Ohmic contact film to p-GaN obtains a more linear and stable light output power increase with increasing current, due to the reduction of current crowding effect nearby the p-metal. It infers that the device B shows an improved current-spreading performance which results in a larger and more uniform light-emission area. On the other hand, the conventional device A easily suffers from current crowding phenomena, i.e., the injection current is mainly confined around the p-metal contact region. This certainly gives the increase of joule heating which results in rapid saturation of light output power .
Figure 9 shows the emission dominant wavelength as a function of current. The wavelength shift for the device B is around 0.9 nm from 40 to 100 mA. This value is better than that of 1.7 nm for the device A. It could be explained by the fact that the studied p-GaN/AuinGaout/ITO structure exhibits a more uniformity of current spreading, giving rise to a relieved joule-heating effect from LED hetero-junctions. Accordingly, it lowers the current-crowding effect and LED junction temperature which results in a smaller wavelength variation. This means that the proposed structure could significantly improve the performance of GaN LEDs.
Figure 10 shows light output power life time behaviors for both studied devices A and B, respectively. During testing conditions, both devices are always biased at 20 mA operation current. There are only around 1.5 – 2% output power degradations after 260 hrs aging time for both devices, indicating that the Au thermal-diffused and removed process for the device B would not severely influence reliability of light behavior.
Figure 11 illustrates the output light images of devices A and B under 20 mA dc current. The output light intensity and uniformity on the device surface are clearly displayed referring to the color bar. From these photographs, a more uniform light emission could be observed for the device B, indicating its superior current spreading ability. Nevertheless, the injected current is always crowded nearby the p-metal pad for the device A. This result agrees well with the discussion in Fig. 8.
In conclusion, an InGaN/GaN based MQW LED with an ITO direct-Ohmic contact to the p-GaN layer is achieved by means of Au thin film thermal-annealing and followed removed processes. An Au in- and gallium out-diffused (AuinGaout) layer is the primary origin to cause the performance improvement. Experimentally, the AuinGaout layer helps improving holes concentration, resistivity, and sheet resistance of the p-GaN contact layer from 3.15×1017, 1.53, and 5.1×104 to 1.83×1018 cm−3, 0.26 Ω-cm, and 8.6×103 Ω/□. At 20 mA operation current, the studied device (device B) exhibits lower dynamic resistance of 18.0 Ω and forward voltage of 3.07 V, as compared with a conventional device (device A) without the AuinGaout layer (25.6 Ω and 3.25 V). An enhanced light output power of 18.0% is also found at this condition. This is mainly attributed to the larger and more uniform light-emission area resulted from the reduced current-crowding effect by means of an ITO direct-Ohmic contact to the p-GaN layer.
This work was supported in part by the National Science Council of Taiwan, R.O.C., under Contract NSC 97-2221-E-006-237-MY3 and Chi Mei Lighting Technology Company.
References and links
1. X. Guo and E. F. Schubert, “Current crowding in GaN/InGaN light emitting diodes on insulating substrates,” J. Appl. Phys. 90(8), 4191 (2001). [CrossRef]
2. L. Zhou, W. Lanford, A. T. Ping, I. Adesida, J. W. Yang, and A. Khan, “Low resistance Ti/Pt/Au ohmic contacts to p-type GaN,” Appl. Phys. Lett. 76(23), 3451 (2000). [CrossRef]
3. Y. J. Liu, C. H. Yen, L. Y. Chen, T. H. Tsai, T. Y. Tsai, and W. C. Liu, “On a GaN-based light-emitting diode with a p-GaN/i-InGaN superlattice structure,” IEEE Electron Device Lett. 30(11), 1149–1151 (2009). [CrossRef]
4. K. M. Chang, J. Y. Chu, and C. C. Cheng, “Highly reliable GaN-based light-emitting diodes formed by p–In0.1Ga0.9N–ITO structure,” IEEE Photon. Technol. Lett. 16, 1807 (2004). [CrossRef]
5. C. S. Chang, S. J. Chang, Y. K. Su, C. H. Kuo, W. C. Lai, Y. C. Lin, Y. P. Hsu, S. C. Shei, J. M. Tsai, H. M. Lo, J. C. Ke, and J. K. Sheu, “High brightness InGaN green LEDs with an ITO on n ++ -SPS upper contact,” IEEE Trans. Electron. Dev. 50(11), 2208–2212 (2003). [CrossRef]
6. T. Gessmann, Y. L. Li, E. L. Waldron, J. W. Graff, and J. K. Sheu, “Ohmic contacts to p-type GaN mediated by polarization fields in thin InxGa1-xN capping layers,” Appl. Phys. Lett. 80(6), 986 (2002). [CrossRef]
7. Y. J. Liu, C. H. Yen, C. H. Hsu, K. H. Yu, L. Y. Chen, T. H. Tsai, and W. C. Liu, “Impact of An indium oxide/indium–tin oxide mixed structure for GaN-based light-emitting diodes,” Opt. Rev. 16(6), 575–577 (2009). [CrossRef]
8. S. K. Julita, G. Szymon, L. S. Elzbieta, P. Ryszard, N. Grzegorz, L. Michał, P. Piotr, T. Ewa, K. Jan, and K. Stanisław, “Ni–Au contacts to p-type GaN – Structure and properties,” Solid-State Electron. 54(7), 701–709 (2010). [CrossRef]
9. J. L. Lee, M. Weber, J. K. Kim, J. W. Lee, Y. J. Park, T. Kim, and K. Lynn, “Ohmic contact formation mechanism of nonalloyed Pd contacts to p-type GaN observed by positron annihilation spectroscopy,” Appl. Phys. Lett. 74(16), 2289 (1999). [CrossRef]
10. Y. Koide, T. Maeda, T. Kawakami, S. Fujita, T. Uemura, N. Shibata, and M. Murakami, “Effects of annealing in an oxygen ambient on electrical properties of ohmic contacts to p-type GaN,” J. Electron. Mater. 28(3), 341–346 (1999). [CrossRef]
11. S. Nakahara and E. Kinsbron, “Room temperature interdiffusion study of Au/Ga thin film couples,” Thin Solid Films 113(1), 15–26 (1984). [CrossRef]
12. M. Puselj and J. Schubert, “Kristallstruktur von Au2Ga,” J. Less-Common Met. 38(1), 83–90 (1974). [CrossRef]
13. I. Waki, H. Fujioka, M. Oshima, H. Miki, and A. Fukizawa, “Low-temperature activation of Mg-doped GaN using Ni films,” Appl. Phys. Lett. 78(19), 2899 (2001). [CrossRef]
14. J. S. Jang, “High output power GaN-based light-emitting diodes using an electrically reverse-connected p-Schottky diode and p-InGaN–GaN superlattice,” Appl. Phys. Lett. 93(8), 081118 (2008). [CrossRef]
15. Y. J. Liu, C. H. Yen, K. H. Yu, P. L. Lin, L. Y. Chen, T. H. Tsai, T. Y. Tsai, and W. C. Liu, “Characteristics of an AlGaInP-based light emitting diode with an indium-tin-oxide (ITO) direct Ohmic contact structure,” IEEE J. Quantum Electron. 46(2), 246–252 (2010). [CrossRef]