Instrument Technology Research Center, National Applied Research Laboratories, Hsinchu 300, Taiwan, ROC
Received 17 September 2006; accepted 29 September 2006
Abstract
The InGaN/GaN light emitting diodes (LEDs) with a nano-roughened top p-GaN surface fabricated by using Ni nano-masks and laser etching methods were demonstrated and analyzed. The experiment results observed the maximum light output power of the InGaN/GaN LED with a nano-roughened top p-GaN surface etched with the laser energy of 300 mJ/cm2was 1.55 times higher than that of a conventional LED, and the
wall-plug efficiency was enhanced 68% at 20 mA. The series resistance of InGaN/GaN LED was reduced by 32% by the increase in the contact area of the nano-roughened surface.
© 2006 Elsevier B.V. All rights reserved.
Keywords: Gallium nitride (GaN); Light emitting diode (LED); Laser etching
GaN-based materials have attracted considerable interest due to their potential use in optoelectronic devices, such as light emitting diodes (LEDs) and laser diodes (LDs)[1–4]. Recently, as the brightness of GaN-based LEDs has increased, applica-tions such as displays, traffic signals, backlights for cell phones, exterior automotive lighting and printers, have become possi-ble. However, the internal quantum efficiency of GaN-based LEDs is much less than 100% at room temperature because of the presence of non-radiative defects. Furthermore, because the refractive index of the nitride epitaxial layer differs greatly from that of the air, for which the refractive indexes of GaN and air are 2.5 and 1.0, respectively, the critical angle at which light generated in the InGaN–GaN active region can escape, is approximately [θc= sin−1(nair/nGaN)]∼23◦, which limits the external quantum efficiency of conventional GaN-based LEDs to only a few percent[5,6]. The light from LEDs can be enhanced either through the device surface or through the side walls of the device. Research on improving the light extraction efficiency (external quantum efficiency) and brightness in the LEDs[5–12]
∗Corresponding author.
E-mail address:[email protected](S.-C. Wang).
has been intense. Recently, Chang et al. reported that cap layers grown at low temperature (800◦C) increased the power output in the InGaN–GaN MQW LEDs by 10%[13]. Fujii reported an increased extraction efficiency in the GaN-based LEDs by surface roughening[12]. These processes all allow the photons generated within the LEDs to find the escape cone, by multiple scattering from a rough surface. We had reported that nano-roughened top surface of an InGaN/GaN LED using Ni clusters as a wet etching mask increased the wall-plug efficiency by 45% [5]. The large improvement in the light output power we have shown indicated that the use of metal clusters to fabricate a roughened p-GaN surface is an excellent means of making a high-power LED. In comparison to the wet etching, the excimer laser etching for p-GaN surfaces have advantages of the easy control and high etching rate. This investigation reported the fabrication of GaN LEDs with nano-roughened p-GaN surfaces using a KrF laser etching and self-assembled Ni metal clus-ters as the laser etching masks. The dimensions and density of the self-assembled Ni clusters can be controlled by the rapid thermal annealing (RTA) at temperatures from 750 to 900◦C. Details of Ni cluster formation have been recently reported[14]. Fig. 1 depicts a schematic diagram of the LED with nano-roughened surface. The light output efficiency of LED with a 0921-5107/$ – see front matter © 2006 Elsevier B.V. All rights reserved.
Fig. 1. Schematic diagram of the nano-roughed LED structure.
nano-roughened surface fabricated by the laser etching through the nano-masks was significantly higher than that of a con-ventional LED without a roughened surface. Additionally, the current–voltage (I–V) measurements showed that the forward voltage of the LED with a nano-roughened surface was lower than that of a conventional LED.
We simulated light propagation and reflection using the ray tracing method provided by Advanced System Analysis Pro-gram (ASAP). For simplicity, we employed a two-dimensional model which is similar on GaN-based LED structures. The top mesa area is 300m × 300 m and depth is 1 m with and with-out rough top surface.Fig. 2shows light extraction efficiency of conventional and rough top surface LED structures as a function of absorption coefficient of the p-GaN layer. The light extraction efficiency here was defined as the ratio of the collected power outside the LED structure and the total power generated from the active region. The result clearly shows that by adding a rough top surface, the extraction efficiency can be greatly enhanced. The extraction efficiency can be enhanced about a factor of 1.29–1.67 times compared with conventional and rough top surface LED structures considering the absorption coefficient of p-GaN layer changed from 10,000 to 100 cm−1.
Fig. 2. Simulation describes the light extraction efficiency of LEDs with and without rough top surface as a function of absorption coefficient of the p-GaN.
Fig. 3. The schematic diagram of laser etching process setup.
Fig. 3shows the schematic diagram of the setup for conduct-ing the laser etchconduct-ing experiment. A KrF excimer laser (Lambda Physick LPX210) operated at wavelength of λ = 248 nm with a pulse width of 25 ns was used for laser etching technique. The laser output energy can be varied from 10 to 25 mJ. The laser beam was reshaped and homogenized using a special opti-cal system to form a highly uniform (±5% RMS) beam profile of 12 mm× 12 mm after the mask plane. A beam splitter then splits the laser beam into a laser etching beam and a monitor beam. The laser etching beam passed through a projection sys-tem of 10× magnification with a 0.2 numerical aperture, and then focused on the LED sample with a squared spot size of 1.0 mm× 1.0 mm. The monitor beam was incident on a beam analyzer for real-time monitoring of the laser beam quality. The LED samples were placed on the top of a working station which can be moved 1.0 mm step by step to scan a typical sample size of 2 cm× 2 cm by using the computer controlled stepper motor. The CCD camera was used to in situ monitor the laser etching process.
The GaN LED samples were grown by metal-organic chem-ical vapor deposition (MOCVD) with a rotating-disk reactor (Emcore D75TM) on a c-axis sapphire (0 0 0 1) substrate at a growth pressure of 200 mbar. Trimethylgallium, trimethylalu-minum, ammonia, CP2Mg and Si2H6were used as sources of Ga, Al, N, Mg and Si. The LED structure includes a 30 nm-thick GaN low-temperature buffer layer, a 4.0m-thick highly con-ductive Si-doped GaN layer (grown at 1050◦C), an active region of undoped multiple quantum wells (MQW) that includes five periods of 2/5 nm-thick In0.21Ga0.79N/GaN (grown at 750◦C), a 50 nm-thick Mg-doped AlGaN layer (grown at 1050◦C) and finally a 0.1m-thick Mg-doped GaN layer (grown at 1050◦C). The top surface of LED, which is a p-GaN surface, was rough-ened by the laser etching after the formation of Ni nano-mask on the top surface. The surface roughness of the LED cap layer was measured by the tapping mode atomic force microscopy (Veeco).
The nano-masks were formed by depositing a Ni thin film with a thickness of 5 nm on a p-GaN surface by electron beam evaporation. RTA was then performed at 750◦C for 1 min to change the Ni layer to the metal Ni nano-masks on the top p-GaN surface. Then, a KrF excimer laser at wavelength of 248 nm with a pulse width of 25 ns and the incident laser
flu-Fig. 4. The etching rate as a function of laser fluence.
ence were from 250 to 800 mJ/cm2was used to etch the p-GaN surface in the air atmosphere. In this process, the beam size of KrF laser (1 mm× 1 mm) was larger than the size of LEDs (300m × 300 m) to avoid non-uniformity of the laser irradia-tion on the surface of p-GaN.Fig. 4shows the etching rate of the p-GaN layer as a function of laser fluence in the air atmosphere. The etching rate increased with the increasing laser fluence. The etching rate of the p-GaN layer from 250 to 800 mJ/cm2were determined to be approximately 15 to 50 nm/pulse. After the laser etching process, the nano-roughened LED samples were dipped into HCl solution for 5 min to remove the residual Ga and Ga oxide on the p-GaN and then dipped into a nitric acid solution (HNO3) for 5 min to remove the Ni nano-mask from a nano-roughened LED. Afterwards, the conventional LED and the LED with a nano-roughened surface were fabricated using the standard process (four mask steps) having a mesa area of 300m × 300 m. First, the 0.5 m SiO2was deposited onto the sample surface by plasma enhanced chemical vapor deposi-tion. Photo-lithography was used to define the mesa pattern after wet etching of SiO2by a buffer oxide etching solution. The mesa etching was then performed with Cl2/Ar as the etching gas in an ICP-RIE system (SAMCO ICP-RIE 101iPH) which the ICP source power and bias power were operated at the frequency of 13.56 MHz. The metal contact layers, including transparent con-tact and pad layers, were patterned by a lift-off procedure and deposited onto samples by electron beam evaporation. Ni/Au (3/5 nm) was used for the transparent electrode and Ti/Al/Ni/Au (20/150/20/200 nm) was used for the n-type electrode. Finally, Ni/Au (20/150 nm) was deposited onto the p-type electrode.
Fig. 5show the (a) SEM and (b) AFM images of the Ni nano-mask on p-GaN surface morphology of a nano-roughened LED sample. The SEM image in Fig. 5(a) shows that the dimen-sion and density of the self-assembled Ni masks under RTA conditions of 750◦C for 1 min were approximately 250 nm and 3× 109cm−2, and the height of the Ni clusters was approx-imately 30 nm when the original Ni thickness was 50 ˚A. The AFM image inFig. 5(b) shows a Ni nano-mask with a root mean-square (RMS) roughness of 7.4 nm before the laser etching was performed.
Fig. 5. (a) SEM and (b) AFM images of the Ni nano-mask on p-GaN surface morphology of a nano-roughened LED sample.
Fig. 6(a and b) show the AFM images depicting the change of the surface morphology of the p-GaN surface during the surface roughening.Fig. 6(a) shows that the conventional p-GaN cap has a RMS roughness of 0.7 nm, and a surface depth of approx-imately 2 nm, demonstrating a rather smooth p-GaN surface in the conventional LED.Fig. 6(b–d) show the AFM images of the p-GaN surface roughened by the laser etching energy of 300, 400 and 800 mJ/cm2, respectively. The RMS roughness of p-GaN surface increased drastically to 8.7 nm as the energy of excimer laser increased from 300 to 800 mJ/cm2. The maximum depth of the surface etched by the laser energy of 800 mJ/cm2 was approximately 37 nm after the removal of the Ni nano-masks.
The I–V characteristics of the conventional and nano-roughened LEDs were also measured. Fig. 7 plots the I–V characteristics of conventional and nano-roughened LEDs. The forward voltages of the conventional and nano-roughened LEDs etched by the laser energy of 300 mJ/cm2were 3.54 and 3.27 V at a driving current of 20 mA, respectively. Furthermore, the dynamic resistance (R = dV/dI) of the nano-roughened LED (27) was 32% lower than that of the conventional LED (40 ).
Fig. 6. AFM images of the top surface morphology of a LED sample. (a) Conventional LED p-GaN surface image. Nano-roughened LED top p-GaN surface applied laser etching energy of: (b) 300 mJ/cm2, (c) 400 mJ/cm2and (d) 800 mJ/cm2.
The results indicated that the nano-roughening surfaces could facilitate the p-type contact to form an Ohmic contact with a low specific resistance when the energy of laser etching ranged from 250 to 400 mJ/cm2. However, the energy of laser etching over 600 mJ/cm2could result in a Schottky contact with a high specific resistance, which might be due to that the surface mor-phology was too rough to form good contacts on the top p-GaN surface.
Electroluminescence (EL) was measured by injecting a con-tinuous current into a chip device at room temperature. The light output was detected using a calibrated large-area Si pho-todiode placed 5 mm from the top of the device. This detecting condition covers almost all of the power emitted from the top
of the LEDs. Fig. 8(a) shows the spectra of the conventional and nano-roughened LEDs.Fig. 8(b) plots the intensity–current (L–I) characteristics versus the RMS roughness of the surfaces in nano-roughened LEDs. As shown inFig. 8(a), the EL inten-sity of the nano-roughened LED exceeds the conventional LED. At an injection current of 20 mA, the MQW emission peak was at approximately 452 nm and the light output power of the con-ventional and nano-roughened LEDs etched by laser energy of 300 mJ/cm2were approximately 5.3 and 8.3 mW, respectively (as shown in Fig. 8(b)). Nano-roughening the p-GaN surface with the laser etching energy of 300 mJ/cm2increased the out-put power of the conventional InGaN–GaN MQW LEDs by a factor of 1.55. The enhancement of the wall-plug efficiency
Fig. 7. Forward I–V curves of conventional and nano-roughened LEDs.
of the nano-roughened LEDs over the conventional LED was 68% at 20 mA. The nano-roughened LEDs with energy of laser etching over 600 mJ/cm2shows slightly lower light output than the conventional LED. This could be due to a Schottky contact
Fig. 8. (a) Room temperature EL spectra of conventional and nano-roughened LEDs with laser etching energy of 300 mJ/cm2at a current of 20 mA and (b) light output power–current (L–I) and RMS surface roughness characteristics of conventional and nano-roughened LEDs.
ture and has increased the maximum light output and wall-plug efficiency of 55 and 68%, respectively, over the conventional InGaN/GaN LEDs at 20 mA when the energy of laser etching was 300 mJ/cm2. The operating voltage of the InGaN/GaN LED was reduced from 3.54 to 3.27 V at 20 mA and the series resis-tance was reduced by 32% by the increase in the contact area of the nano-roughened surface.
Acknowledgements
The authors would like to thank F. I. Lai and T. H. Hsueh from National Chiao Tung University for useful discussions. This work was supported in part by the National Science Council of Republic of China (ROC) in Taiwan under contract no. NSC 92-2215-E-009-015, NSC 92-2112-M-009-026.
References
[1] S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa, T. Yamada, T. Matsushita, Y. Sugimoto, H. Kiyoku, Appl. Phys. Lett. 70 (1997) 868–870.
[2] S. Nakamura, M. Senoh, N. Iwasa, S. Nagahama, Jpn. J. Appl. Phys. 34 (1995) L797–L799.
[3] S. Nakamura, T. Mukai, M. Senoh, Appl. Phys. Lett. 64 (1994) 1687–1689. [4] T. Mukai, S. Nakamura, Jpn. J. Appl. Phys. 38 (1999) 5735–5739. [5] H.W. Huang, C.C. Kao, J.T. Chu, C.C. Yu, H.C. Kuo, S.C. Wang, IEEE
Photo. Technol. Lett. 17 (2005) 5–7.
[6] T.H. Hsueh, J.K. Sheu, H.W. Huang, J.Y. Chu, C.C. Kao, H.C. Kuo, S.C. Wang, IEEE Photo. Technol. Lett. 17 (2005) 1163–1165.
[7] J.J. Wierer, D.A. Steigerwald, M.R. Krames, J.J. O’shea, M.J. Ludowise, G. Christenson, Y.C. Shen, C. Lowery, P.S. Martin, S. Subramanya, W. Gotz, N.F. Gradner, R.S. Kern, S.A. Stockman, Appl. Phys. Lett. 78 (2001) 3379–3381.
[8] S.X. Jin, J. Li, J.Y. Lin, H.X. Jiang, Appl. Phys. Lett. 77 (2000) 3236–3238. [9] T.N. Oder, J. Shakya, J.Y. Lin, H.X. Jiang, Appl. Phys. Lett. 83 (2003)
1231–1233.
[10] T.N. Oder, K.H. Kim, J.Y. Lin, H.X. Jiang, Appl. Phys. Lett. 84 (2004) 466–468.
[11] H.W. Choi, M.D. Dawson, P.R. Edwards, R.W. Martin, Appl. Phys. Lett. 83 (2003) 4483–4485.
[12] T. Fujii, Y. Gao, R. Sharma, E.L. Hu, S.P. DenBaars, S. Nakamura, Appl. Phys. Lett. 84 (2004) 855–857.
[13] S.J. Chang, L.W. Wu, Y.K. Su, Y.P. Hsu, W.C. Lai, J.M. Tsai, J.K. Sheu, C.T. Lee, IEEE Photo. Technol. Lett. 16 (6) (2004) 1447–1449. [14] H.W. Huang, C.C. Kao, T.H. Hsueh, C.C. Yu, C.F. Lin, J.T. Chu, H.C. Kuo,
S.C. Wang, Mater. Sci. Eng. B 113 (2004) 125–129.
[15] C.J. Youn, T.S. Jeong, M.S. Han, J.W. Yang, K.Y. Lim, H.W. Yu, J. Cryst. Growth 250 (2003) 331–338.
[16] H.W. Huang, C.C. Kao, J.T. Chu, C.C. Yu, C.F. Lin, H.C. Kuo, S.C. Wang, Mater. Sci. Eng. B 113 (2004) 19–23.
