GaInN∕GaN growth optimization for high-power green light-emitting diodes

C. Wetzel, T. Salagaj, T. Detchprohm, P. Li, J. S. Nelson
2004 Applied Physics Letters  
Two different approaches to optimize the growth conditions for high-power green light-emitting diodes (LEDs) using Ga 1−x In x N / GaN metalorganic vapor phase epitaxy are discussed. We compare typical results in terms of morphology, photo-, and electroluminescence properties. We find good results for an optimization of the lateral morphological homogeneity of the active region. An extension of growth conditions for the active layers of blue LEDs was misleading. This suggests that different
more » ... that different emission processes are involved in blue and green LEDs. We achieve die performances of 2.5 mW at 523 nm (526 nm dominant) for low forward voltages of 3.4 V at a typical drive current of 20 mA. Alloys of Ga 1−x In x N provide the opportunity to develop high-power/high brightness light-emitting devices in the near UV, blue, and green spectral range. Significant progress in materials physics and device development have made blue light-emitting diodes (LEDs) a widely available consumer product, and near-UV laser diodes are being implemented in next-generation storage media. 1-3 Full implementation of the technology's capabilities in solid-state lighting, however, requires significant improvements of emission power of green LEDs by an understanding of their light emission processes. 4 The technological challenges lie in the difficulties to grow epitaxially Ga 1−x In x N alloys of high InN fraction x due to contrasting thermodynamical properties of GaN and InN and their respective precursors. Frequently, a phase separation of the alloy is observed, especially for x ജ 0.2. 5 Moreover, large lattice mismatch of GaN and InN leads to strong biaxial strain and huge polarization fields in the uniaxial crystal lattice. The challenges are compounded by a rather limited understanding of the electronic band structure and the light emission processes in the active Ga 1−x In x N alloy layers and their quantum wells (QWs). It has been proposed that Inrelated composition fluctuations could be beneficial for enhanced light emission efficiency. 2,6 Such models, however, do not provide a quantitative description of the electronic band structure and their transition energies. In other approaches, the polarization-induced electric field is used as a free fitting parameter to simulate experimental energies with models of the quantum confined Stark effect. 7-9 Our model instead proposes to limit the energetic contribution of composition fluctuations (i.e., ±20 meV for x = 0.2), to use independently determined electric-field values and instead explore the full extend of other possible mechanisms. In an analysis of multiple QW (MQW) structures optimized for lateral homogeneity, we previously developed a model of the electronic band structure in Ga 1−x In x N / GaN QWs in the presence of electric-field values ͑F͒ directly derived from Franz-Keldysh oscillations. 10 Interband transitions in absorption-type spectroscopy have been identified with their respective transitions in k ϫ p perturbation theory under full consideration of the quantum confined Stark effect. In this way, the strong absorption edge has been asso-ciated with the fundamental interband transition ͑e 1 hh 1 ͒ between the first quantized heavy hole ͑hh 1 ͒ and electron ͑e 1 ͒ states. The photoluminescence (PL) was found to follow e 1 hh 1 by an energetic redshift equivalent to the polarization dipole ͑FeL w ͒ (electron charge e Ͼ 0, well width L w ) across the QW in E͑e 1 hh 1 ͒ -FeL w . 11 For blue LEDs, emission from both transitions, i.e., the fundamental and the redshifted, should be possible. Indeed, some authors describe a competition between two lines. 12 For the challenge of green LEDs ͑515-565 nm͒, an optimization of the redshifted transition holds promise. For L w = 3.0 nm, x Ϸ 0.20 our model predicts emission at 525 nm with a redshift of Ϸ0.3 eV. In contrast, to obtain such an emission wavelength using e 1 hh 1 would require much higher InN fractions. From the perspective of growth optimization to maximize the green emission, we see the following paradigms. The first is a modification to the MQW design of existing high performance blue LED dies ͑470 nm͒ by shifting the emission to longer wavelengths. The parameters known to control the emission wavelength, i.e., In source flow, growth temperature, growth pressure, V/III ratio, and variation of carrier gas flow, are being tuned to extend the emission wavelength into the green. The second approach is motivated by the insight that a different transition might be beneficial for the green emission. We therefore do not make use of the blue conditions and search a specific solution. Here, we optimize each layer in the structure to an utmost homogeneity in terms of spatially resolved PL, x-ray diffraction pattern, and surface morphology on the nanometer scale. We here compare results of atomic force microscopy (AFM), PL, and device performance for optimization under both paradigms. LED epiwafers have been grown by metalorganic vapor phase epitaxy in an Emcore D-180 Spectra GaN rotating disk multiwafer system using trimethyl and diethyl adducts of Ga, In, Al, as well as ammonia. Ga 1−x In x N / GaN MQW structures have been embedded in p-n diodes on a (0001) sapphire substrate. Typical design parameters for the active region are as follows. Five Ga 1−x In x N QWs of nominal well width L w = 3 nm, separated by barriers of nominal L b = 11 nm (Ref. 13) have been grown at a temperature of 650-850°C. Resulting x values are in the range from 0.10 to 0.20 as determined by x-ray diffraction analysis. There is no intentional doping in the active region and carrier injection layers are all nominally identical for n and p, respec-a) Present address: Future Chips Constellation, Rensselaer Polytechnic Institute,
doi:10.1063/1.1779960 fatcat:dqplolg44va4hp6bni432ux2hi