Plasma etching of AlN/AlGaInN superlattices for device fabrication

K. Zhu, V. Kuryatkov, B. Borisov, G. Kipshidze, S. A. Nikishin, H. Temkin, M. Holtz
2002 Applied Physics Letters  
We report a study of plasma etching of GaN, AlN, and AlN/AlGaN superlattices for the processing of deep ultraviolet light emitting diodes. Etching was carried out using inductively coupled plasma of chlorine diluted with argon under reactive ion etching conditions. Using parameters selected for etch rate, anisotropy, and surface smoothness, we study etching of n-and p-type superlattices. The former etches at a rate of 250 nm/min, which is intermediate to that of AlN and GaN, while the latter
more » ... while the latter exhibits a slower etch rate of 60 nm/min. Based on these studies, we prepare low-leakage p -n junctions and mesa light emitting diodes with peak emission at 280 nm. There has been considerable recent activity in preparing ultraviolet light emitting diodes ͑LEDs͒ and detectors, based on AlGaN, 1 operating at wavelengths below 300 nm. Short period superlattices ͑SLs͒ based on AlGaN/GaN and AlGaInN/AlGaInN 2,3 have been used to overcome some of the difficulties of growing high quality p-type Al-rich layers. Our recent work has shown that diodes based on SLs of AlN/AlGaInN can be used to produce LEDs with light emission below 300 nm. 4 This is accomplished using SLs of AlN and AlGaInN with well and barrier thicknesses of 1.25 and 0.5 nm, respectively. The average AlN mole fraction is very high in these structures, over 0.6, providing the necessary large optical gap, and the wells are heavily doped to produce a high average carrier concentration. An essential step in the fabrication of these devices is mesa etching. Etching by chlorine has become the most popular method for processing GaN-based devices. 5 However, little work has focused on plasma etching for processing Al-rich device structures or superlattices. We describe Cl 2 -based plasma etching experiments for preparing wide band-gap devices from Al-rich AlGaN. The GaN, AlN, and AlN/AlGaN SLs studied were grown on silicon and sapphire substrates by gas-source molecular-beam epitaxy with ammonia as the nitrogen source. An AlN buffer layer ͑ϳ40 nm thick͒ was first grown at 830-860°C. A second buffer layer of GaN was then grown, with thickness of 1.35 m. This layer is known to reduce the density of threading dislocations, which originate at the substrate-nitride interface. 6,7 The n-and p-type SLs were grown next, with the wafer held at temperatures from 760 to 800°C. Silicon and magnesium were used for n-and p-type doping, respectively. Details of the growth can be found elsewhere. 8,9 A commercial system was used for plasma etching with Cl 2 /Ar gas mixture. Samples are mounted on 200 mm silicon wafers coated on the processing side by SiO 2 . The baseline pressure of the etching system is 2ϫ10 Ϫ6 Torr. The gases are injected through a showerhead located ϳ100 mm above the sample. A high-density inductively coupled plasma ͑ICP͒ discharge is generated by applying 13.56 MHz rf power to the water cooled inductive coil. The wafer electrode is powered separately by a 600 W generator ͑13.56 MHz͒ for reactive ion etching ͑RIE͒. As the ion energy and plasma density are effectively decoupled for the ICP-RIE system, uniform density, and energy distributions are transferred to the sample while keeping the ion and electron energies low. 5 The substrate was cooled by He gas flowing from the electrostatic chuck, with no temperature control. We estimate the wafer temperature to remain below ϳ100°C during our experiments. Etch depths were determined using a stylus profilometer. Scanning electron microscopy ͑SEM͒ and atomic force microscope ͑AFM͒ were used to determine etch anisotropy and surface properties. To establish baseline etching parameters for the SLs, we etched test layers of GaN and AlN. Figure 1 shows etch rates versus plasma parameters of ICP power, RIE power, and pressure. The Cl 2 to Ar flow ratios were kept constant at 20-5 sccm in all cases. The etch rate of AlN is consistently lower than that of GaN. Etching is faster with higher ICP and RIE powers. The former is primarily due to increased chlorine radical concentration in the plasma, while the latter stems from enhancing the volatility of etch byproducts at the target surface. We use ICP power of 300 W and RIE power of 150 W for etching the SLs, because these parameters give us a high etch rate while producing anisotropic features and smooth etched surfaces. Figure 1͑c͒ shows that the etch rate decreases with increased chamber pressure, particularly for GaN. This is plausibly due to reduced mean-free paths of the etching species. Consequently, a lower pressure is preferred; we use 8 mTorr for the SL etching experiments. The etch rate results in Fig. 1 are in good agreement with what has been previously reported. 10,11 A typical SEM micrograph is shown in Fig. 2 , verifying the desired anisotropic etching. Our AFM studies of etching induced roughness also show that these parameters lead to smooth postetch surfaces. For example, the rms roughness over the 10 mϫ10 m scan area is found to be ϳ4.2 nm for the p-type SL ͑to be discussed͒ following a 165 nm deep etch. Using the above parameters, and prior to etching of LED a͒ Electronic
doi:10.1063/1.1527986 fatcat:b7bdq4ht4vgxpi2ponrycvfbdu