SOLID-STATE lighting thrives on the efficient energy conversion

Transcription

1 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 57, NO. 10, OCTOBER Boosting Green GaInN/GaN Light-Emitting Diode Performance by a GaInN Underlying Layer Yong Xia, Wenting Hou, Liang Zhao, Mingwei Zhu, Theeradetch Detchprohm, and Christian Wetzel Abstract The light output of 530 nm green GaInN/GaN light-emitting diodes on sapphire has been nearly doubled by the insertion of a 130-nm GaInN underlayer (UL) between the n-gan electron injection layer and the quantum-well (QW) active region. Under variation of the alloy composition, best results were obtained for an x = 6.3% Ga 1 x In x N UL. By low-temperature depth-resolved cathodoluminescence spectroscopy, an interplay of the impurity-related donor-acceptor pair recombination, the UL, and the QW emission has been observed. We propose that the resonance and level alignments between the defect and UL levels reroute excitation toward radiative recombination in the QWs. Index Terms Cathodoluminescence (CL), GaN, light-emitting diode (LED), underlayer (UL). I. INTRODUCTION SOLID-STATE lighting thrives on the efficient energy conversion in group-iii nitride light-emitting diodes (LEDs) [1], [2]. The benefit of a narrow-band emission has widely been employed in colored display lighting; yet, for full-spectrum white light generation, usually, a combination of blue LED and yellow phosphor is employed. Green-emitting LEDs have yet to live up to their theoretical potential as the dominant component in the red green blue combination white LED light sources. Here, we propose to boost the green LED efficiency in Ga 1 y In y N/GaN multiple-quantum-well (MQW) LEDs by implementing a Ga 1 x In x N thin-film underlayer (UL) between the n-gan electron injection layer and the MQW active region [3], [4]. The concept has been used in blue Ga 1 y In y N/GaN LEDs [5], yet, its use in green LEDs requires new approaches and offers the opportunity to elucidate the processes at work. Various mechanisms have been proposed to explain its benefits. Akasaka et al. proposed that, in 395-nm LEDs, a 50-nm thick x = 4% UL separates the active region from any centers responsible for nonradiative recombination [4]. Niu et al. proposed that, in 464-nm LEDs, a 25-nm x = 9% UL with a short- Manuscript received May 17, 2010; accepted July 5, Date of publication August 12, 2010; date of current version September 22, This work was supported in part by the Department of Energy/National Energy Technology Laboratory Solid-State Lighting Contract of Directed Research under Grant DE-FC26-06NT42860 and in part by the National Science Foundation Smart Lighting Engineering Research Center under Grant EEC The review of this paper was arranged by Editor L. Lunardi. Y. Xia, W. Hou, L. Zhao, T. Detchprohm, and C. Wetzel are with the Future Chips Constellation and the Department of Physics, Applied Physics, and Astronomy, Rensselaer Polytechnic Institute, Troy, NY USA ( wetzel@rpi.edu). M. Zhu is with Applied Materials, Inc., Santa Clara, CA USA. Color versions of one or more of the figures in this paper are available online at Digital Object Identifier /TED period superlattice reduces the strain in the QW stack [6], while Törmä et al. proposed that, in 440 nm LEDs, a 20 nm, x = 2% UL acts to inject electrons into the QWs at a lower energy than a GaN barrier would [7]. Here, we successfully expand the concept to green emitting LEDs and find an explanation that combines the aspects of the previous models, yet reveals the UL s particular suitability to the longer wavelength region of green LEDs. In the electroluminescence (EL) of 530 nm LEDs, we find a relevant 85% enhancement of the light output power (LOP) as measured in fabricated LED dies. In order to reveal the mechanism of efficiency enhancement, we characterize the luminescence properties by a depth-resolved cathodoluminescence (CL) spectroscopy and a low-temperature photoluminescence (PL). II. EXPERIMENTAL A set of four different Ga 1 y In y N/GaN MQW LED structures were grown on a c-plane sapphire in a vertical flow metal organic vapor phase expitaxy [8]. Trimethylgallium, trimethylindium, and ammonia were used as gallium, indium, and nitrogen precursors, respectively. Silane and biscycolpentadienyl magnesium were used as the n-type and p-type dopant sources. The active region contains eight periods of QWs (3 nm thick) and barriers ( 23 nm thick). A nominally undoped 130 nm thick Ga 1 x In x N UL was grown before the QWs. Structures contain ULs of different compositions: x UL = 0%, 3.8%, 6.3%, and 8.8%, as determined by a high resolution ω 2θ X-ray diffraction. Here, x UL = 0% is a shorthand for omitting the UL. The respective sample serves as a reference of a conventional sample structure. For the UL and the QW growth, a particular effort was paid to avoid inhomogeneities of alloy composition and thickness as detailed elsewhere [9], [10]. The MQW structures are topped off with a 20 nm p-algan electron-blocking layer and a 200 nm p-gan layer. Si doping levels of cm 3 and Mg doping levels of cm 3 are typically achieved. To minimize process fluctuations, the samples have been grown in consecutive epitaxy runs using a proven base recipe that delivered reproducible results before and after the growth sequence. For spectroscopic analysis, the CL, at variable low temperature, was collected in a scanning electron microscope. For a depth-resolved analysis, the acceleration voltage was swept from 3 to 25 kv. Using the Kanaya -Okayama relation [11] and the calibration runs, we correlate this to a mean excitation depth of μm. To maintain a constant carrier generation rate, the beam power was kept at 750 nw by reducing the beam current. For EL, epi wafers have been processed to (700 μm) 2 LED dies with /$ IEEE

2 2640 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 57, NO. 10, OCTOBER 2010 Fig. 1. Reciprocal space mapping in X-ray diffraction around the (10 15) maximum of GaN for the three samples with different UL compositions x UL. Horizontal maxima for the UL and the MQW are ideally matched with that of GaN. No in-plane relaxation can be observed. Fig. 2. (a) EL spectra at 100 ma peaking at 530 nm. (b) (Left axis) LOP of the green LED dies at 20 and 100 ma as a function of InN fraction in the Ga 1 x In xn UL. (Right axis) IQE at room temperature. Both reach a clear maximum in the x UL = 6.3% sample. amesaareaof(600 μm) 2. The LOP of the bare unencapsulated die was measured by a power-calibrated spectrometer inside a 10-in integrating sphere. All data is representative for the 1-in center portion of the 2-in wafer structures with a variation of not more than 5%. III. RESULTS AND DISCUSSION Reciprocal space mapping around the (10 15) diffraction maximum of GaN is shown in Fig. 1 for the samples with x UL = 3.8%, 6.3%, and 8.8%. The horizontal axis corresponds to the reciprocal a-lattice constant, while the vertical axis corresponds to the reciprocal c-lattice constant. In all three samples, the maximum corresponding to the GaInN UL is in very close alignment with the reciprocal a-lattice constant of the GaN maximum. This proves that the biaxial strain is not relaxed within the GaInN layer to a degree more than 3%. The EL spectra of all LED structures show a single peak around 530 nm at a current of 100 ma [see Fig. 2(a)]. Peak wavelengths are 534 nm (x UL = 0%), 536 nm (x UL = 3.8%), 529 nm (x UL = 6.3%), and 529 nm (x UL = 8.8%). Line widths are 38 nm, 40 nm, 36 nm, and 42 nm full-width at halfmaximum (FWHM), respectively. Apparently, the insertion of Fig. 3. Low-temperature CL spectra of all four LED structures with an acceleration voltage for the maximum excitation in the UL. The UL peak shows a strong variation with the InN fraction, and the DAP recombination seems to interact with the UL emission. the UL with different alloy compositions has no significant effect on the peak wavelength and the line widths of the QWs. This, furthermore, is an evidence that the UL does not substantially relax the strain. The presence and the composition of the UL, however, has a strong impact on the LOP [see Fig. 2(b)]: Both, for the 20 ma and 100 ma operations, the LOP increases with x UL and shows a maximum for x UL = 6.3%, before it drops in the x UL = 8.8% sample. The maximum is 4.9 mw at 100 ma for x UL = 6.3%, while the reference sample, x UL = 0%, shows an LOP of 2.7 mw. This is a significant enhancement of 85% due to the insertion of the UL. We estimate the upper limit of the internal quantum efficiency (IQE) by the ratio of the peak-integrated PL intensity at room temperature and that at 4.2 K. At room temperature, the PL is highest in the x UL = 3.8% sample when excited at 325 nm above the GaN band gap. When excited with 20 mw at 405 nm directly in the active region, the x UL = 6.3% sample has the strongest PL. We choose the latter excitation to avoid absorption in the p-layers. Possible resonance-enhanced absorption is unlikely to affect the result due to the small energy variation of the transition energies. Under the assumption of the negligible non-radiative recombination at low temperature, an IQE value at room temperature as high as 66% is found in the x UL = 6.3% sample versus 32% without the UL [see Fig. 2(b)]. There is a high chance that the absolute measurements of efficiency will reveal values that could be lower than those upper limit values; their relative increase, however, is quite meaningful. This trend in IQE closely follows that of the LOP in the EL, revealing a maximum at x UL = 6.3%. To elucidate the mechanism of efficiency enhancement, a CL spectroscopy at 77 K and an acceleration voltage of 11 kv, equivalent to a predominant excitation in the UL, of the four LED structures is shown in Fig. 3. Besides the dominant QW emission (530 nm), a donor-acceptor-pair (DAP) recombination (380 nm) with a phonon replica (390 nm) and a near-band-edge emission (NBE) of GaN (363 nm) are seen in the reference sample (x UL = 0%). At 370 nm, we identify the corresponding free electron-to-acceptor transition as a shoulder to the DAP transition. Three samples show strong emission at 388 nm (x UL = 3.8%), 410 nm (x UL = 6.3%), and

3 XIA et al.: BOOSTING GREEN GaInN/GaN LED PERFORMANCE BY A GaInN UNDERLYING LAYER 2641 Fig. 4. (a) Peak energies of relevant transitions at 77 K as a function of InN fraction in the UL. (b) The same energies separated into VBE and CBE contributions. In both cases, near-crossing points are seen. (a) UL transition energy crosses the phonon replica of the DAP transition. (b) UL VBE crosses with the acceptor level. 435 nm (x UL = 8.8%), which we assign to the UL. All identified energies are summarized in Fig. 4, versus the InN-fraction of the UL. While Fig. 4(a) lists the transition energies directly, Fig. 4(b) separates them into the valence band edge (VBE) and conduction-band edge (CBE) contributions, according to the relative band offset between GaN and InN [13]. Included are the effective mass donor (Don), assumed at 30 mev, and the resulting acceptor (Acc) levels as concluded from the peak positions. Also shown are the anticipated absorption edges (abs) for electrons (e) and holes (h) based on the QW-emission (em) peak and extrapolated data on the reported Stokes shift in similar structures [13]. A peculiarity is seen for the sample with x UL = 3.8% [see Figs. 3 and 4(a)]. Here, the pattern typical for the DAP recombination appears distorted by the UL emission. The 370-nm shoulder of the free electron-to-acceptor transition can still be seen, but the DAP transition expected at 380 nm is missing. The UL peak appears at 388 nm, and the phonon replica at 400 nm appears to follow the UL recombination band instead of the DAP transition. A possible reason is the energetic near-coincidence of the DAP transition and the UL band-edge transition. Of even higher interest is the structure with x UL = 6.3%. Fig. 5 shows the CL peak intensities as a function of the effective excitation depth, together with the layers indicated as designed. We distinguish the GaN NBE emission (362 nm), the UL emission (410 nm), the MQW emission (532 nm), and the DAP transitions (380 nm). Starting from the sample surface, the first intensity rise is seen in the QW emission at a depth of 0.1 μm. At this depth, the excitation reaches the QWs. At 0.30 μm, the emission reaches a maximum. This marks the center of the MQWs, which, per design, should lie at 0.32 μm. The DAP transition peaks at 0.23 μm, which corresponds to the interface of the p-gan layer and the MQW region per design. At higher excitation depth, the DAP transition drops in intensity, while the UL emission rises to a maximum at 0.40 μm, corresponding to the UL per design. At even larger depth, the GaN NBE rises, and the DAP transition again follows the suit at similar rate. Fig. 5. Depth profiling of the CL peak intensities in the LED with x = 6.3% UL at low temperature. The layers, as designed, are indicated. DAP transitions are highest when the QW reaches its maximum. For larger depth, the QW peak seems to follow the UL intensity, indicating a possible excitation-transfer process. The excitation here is predominant within the n-layers of the structure. Notice that, while the DAP transition exhibits a clear maximum at the depths of the p-gan layer interface to the MQW region, the QW emission does not drop off any faster with depth than the UL emission. This close correlation of the UL and QW emissions and the anticorrelation of the UL and DAP emission intensities in the x UL = 6.3% structure, as well as the energetic interaction of the UL and DAP transitions in the x UL = 3.8% structure, can be explained by an excitation transfer mechanism between the UL and the QWs. Apparently, whenever the UL is excited, while it produces a strong CL, it also transfers excitation to the QWs and enhances that recombination. On the other hand, the GaN DAP luminescence is reduced whenever the excitation reaches the UL. While the DAP luminescence can only be observed at low temperature in the CL, the relevant transitions are still active at room temperature but are dominated by nonradiative recombination through the same defects. The DAP luminescence, therefore, indicates the existence of a major shunt-recombination process that likely limits the overall device performance. The ability to suppress the GaN DAP luminescence by a resonance with an x = 3.8% UL suggests the possibility of recovering the excitation from such DAP recombination and redirecting it to the radiative centers of the QWs. This is a likely reason for the LOP gains in the x = 3.8% UL structure over the reference device. The observation that the LOP gains are even higher in the x = 6.3% UL sample can be explained by the following additional considerations. The proposed excitation transfer from the DAP to the UL should gain further if the respective levels line up, particularly those for the low-mobility holes. According to the established relative band offsets between GaN and InN, only some 40% appear on the valence band side. The VBE of the Ga 1 x In x N UL should, therefore, vary with ΔE v = 0.40(E GaN E UL ) and amount to ΔE v = 160 mev in the x = 6.3% UL sample. This value, indeed, is close to or just below the acceptor level participating in the DAP transition observed near E GaN E DAP E Don = 130 mev. Here,

4 2642 IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 57, NO. 10, OCTOBER 2010 E D = 30 mev is the effective mass value for a donor [see Fig. 4(b)]. Therefore, it appears possible that the holes initially poised for the radiative and predominantly nonradiative recombination through DAP processes may be rerouted to the UL and the QW as part of a resonant excitation transfer. The same analysis for the x = 8.8% UL sample leads to ΔE v = 220 mev, which lies an entire LO phonon energy (90 mev) below the suspected acceptor level and may not provide an attractive route for holes to transfer to. This is in line with a significantly lower LOP observed in such a sample in the EL. It should be noted that the above explanation should also work for blue-emitting LEDs utilizing the UL concept, as long as the wavelength is above 410 nm. The spectrally larger separation in the green structures, however, simplifies the spectroscopic analysis. The considered model relies on the assumption of an efficient hole-hopping conduction through the acceptor states to add a relevant contribution to the QW luminescence. From this, we may conclude that the density of the residual acceptor states is quite high and might indeed be furthered by defect clusters along threading dislocations. The observation on an enhancement of the hole transport mechanism by a UL located at the electron-injection side of the QWs is surprising. It suggests a more complex exchange of excitation throughout the active region than described in a picture of individual charges. A possible alternate explanation is that of the excitonic or optical excitation transfer between the QWs and the DAP states. It would also well explain other observations like the clear preference of the lowest transition energy and the identical emission wavelength in top and bottom views. Such a model would resemble the processes in organic polymer LEDs. Such evidence for less frequently considered higher order excitation transfer mechanisms may possibly widen the scope for future study to resolve the details of the recombination in the GaInN/GaN QWs in general. [3] J. K. Son, S. N. Lee, T. Sakong, H. S. Paek, O. Nam, Y. Park, J. S. Hwang, J. Y. Kim, and Y. H. Cho, Enhanced optical properties of InGaN MQWs with InGaN underlying layers, J. Cryst. Growth, vol.287,no. 2,pp , Jan [4] T. Akasaka, H. Gotoh, T. Saito, and T. Makimoto, High luminescent efficiency of InGaN multiple quantum wells grown on InGaN underlying layers, Appl. Phys. Lett., vol. 85, no. 15, pp , Oct [5] M. Crawford, D. Koleske, N. Missert, M. Lee, M. Banas, D. Follstaedt, K. Bogart, G. Thaler, K. Cross, Y. Xia, C. Wetzel, and E. F. Schubert, Mechanisms for enhanced quantum efficiency of InGaN quantum wells grown on InGaN underlayers, presented at the 7th Int. Conf. of Nitride Semiconductors (ICNS-7) Conf., [6] N. H. Niu, H. B. Wang, J. P. Liu, N. X. Liu, Y. H. Xing, J. Han, J. Deng, and G. D. Shen, Improved quality of InGaN/GaN multiple quantum wells by a strain relief layer, J. Cryst. Growth, vol. 286, no. 2, pp , Jan [7] P. T. Torma, O. Svensk, M. Ali, S. Suihkonen, M. Sopanen, M. A. Odnoblyudov, and V. E. Bougrov, Effect of InGaN underneath layer on MOVPE-grown InGaN/GaN blue LEDs, J. Cryst. Growth, vol. 310, no. 23, pp , Nov [8] C. Wetzel, T. Salagaj, T. Detchprohm, P. Li, and J. S. Nelson, GaInN/GaN growth optimization for high-power green light-emitting diodes, Appl. Phys. Lett., vol. 85, no. 6, pp , Aug [9] T. Detchprohm, Y. Xia, Y. Xi, M. Zhu, W. Zhao, Y. Li, E. F. Schubert, L. Liu, D. Tsvetkov, D. Hanser, and C. Wetzel, Dislocation analysis in homoepitaxial GaInN/GaN light emitting diode growth, J. Cryst. Growth, vol. 298, pp , Jan [10] T. Detchprohm, M. W. Zhu, Y. F. Li, Y. Xia, C. Wetzel, E. A. Preble, L. H. Liu, T. Paskova, and D. Hanser, Green light emitting diodes on a- plane GaN bulk substrates, Appl. Phys. Lett., vol. 92, no. 24, p , Jun [11] K. Kanaya and S. Okayama, Penetration and energy-loss theory of electrons in solid targets, J.Phys.D,Appl.Phys., vol. 5, no. 1, pp , Jan [12] M. Toth and M. R. Phillips, Monte Carlo modeling of cathodoluminescence generation using electron energy loss curves, Scanning, vol. 20, no. 6, pp , [13] C. Wetzel, T. Takeuchi, H. Amano, and I. Akasaki, Quantized states in Ga 1 x In xn/gan heterostructures and the model of polarized homogeneous quantum wells, Phys. Rev. B, Condens. Matter, vol. 62, no. 20, pp. R R13 305, IV. CONCLUSION The light output performance of green LEDs has been enhanced at 85% by the use of a 130-nm Ga 1 x In x N, x UL = 6.3% thin-film UL inserted between the n-layer and the QW active region. By variation of the alloy composition, the best value of x UL = 6.3% has been determined. By the correlation of the depth-resolved CL at low temperature, an excitation-transfer mechanism between the defect-bound recombination, the UL, and the QWs has been proposed as the mechanism of this enhancement. The process, apparently, has worked independent of the mode of excitation, geminate photogeneration or separate charge injection from the opposite ends of the structure. The observations suggest a relevant optical or excitonic excitationtransfer mechanism within the active region of the device. REFERENCES [1] I. Akasaki and H. Amano, Crystal growth and conductivity control of group III nitride semiconductors and their application to short wavelength light emitters, Jpn. J. Appl. Phys., vol. 36, no. 9A, pp , Sep [2] I. Akasaki and C. Wetzel, Future challenges and directions for nitride materials and light emitters, Proc. IEEE, vol. 85, no. 11, pp , Nov Yong Xia received the B.S. degree in physics from the University of Science and Technology of China, Hefei, China, in 2000 and the Ph.D. degree in physics from Rensselaer Polytechnic Institute, Troy, NY, in His main research interests are the spectroscopic characterization and material epitaxy of group-iii nitride light emitters. Wenting Hou received the B.S. degree in physics from the University of Science and Technology of China, Hefei, China, in She is currently working toward the Ph.D. degree in physics with Rensselaer Polytechnic Institute, Troy, NY, in the area of processing of GaN-based green light-emitting diodes and laser diodes.

5 XIA et al.: BOOSTING GREEN GaInN/GaN LED PERFORMANCE BY A GaInN UNDERLYING LAYER 2643 Liang Zhao received the B.S. and M.S. degrees in physics from Tsinghua University, Beijing, China, in 2004 and 2007, respectively. He is currently a Research Assistant with Rensselaer Polytechnic Institute, Troy, NY. His work is mainly focused on the optical spectroscopic characterization of MOVPE-grown GaInN/GaN green multiple-quantum-well LEDs. Theeradetch Detchprohm received the B.S., M.S., and Ph.D. degrees from Nagoya University, Nagoya, Japan, in 1991, 1993, and 1996, respectively, all in electrical and electronic engineering. In 1998, he joined the High Tech Research Center, Meijo University, Nagoya, as a Postdoctoral Researcher. In 2001, he joined the Research and Development Team, Uniroyal Optoelectronics, Tampa, FL, where he developed and managed the production of high-brightness light-emitting diodes. In 2004, he joined the Department of Physics, Applied Physics, and Astronomy, Rensselaer Polytechnic Institute, Troy, NY, where he is currently a Research Associate Professor with the Future Chips Constellation. He has authored or coauthored more than 125 papers published in scientific journals. He has been working on epitaxy and fabrication of AlGaInN-based semiconductor devices (particularly optoelectronic applications) since Mingwei Zhu received the B.S. degree in physics from the University of Science and Technology of China, Hefei, China, in 2004 and the Ph.D. degree in physics from Rensselaer Polytechnic Institute, Troy, NY, in His Ph.D. thesis is on the MOVPE growth and characterization of GaInN/GaN green LEDs in polar and nonpolar orientations. He is currently a Process Engineer with Applied Materials, Inc., Santa Clara, CA. Christian Wetzel received the Ph.D. degree in physics from the Technical University Munich, Munich, Germany, in He was a Visiting Scientist with Lawrence Berkeley National Laboratory in In 1997, he joined the High Tech Research Center, Meijo University, Nagoya, Japan. In October 2000, he joined Uniroyal Optoelectronics as a Senior Epi Scientist and Green Project Manager. He has been the Wellfleet Career Development Constellation Professor, Future Chips Constellation, and an Associate Professor of physics with Rensselaer Polytechnic Institute, Troy, NY, since His research centers on the electronic band and defect structure of piezoelectric group-iii nitrides to realize new concepts of high-efficiency light-emitting devices and solar cells.