Monolithic Integration of Individually Addressable Light-Emitting Diode Color Pixels

Transcription

1 Monolithic Integration of Individually Addressable Light-Emitting Diode Color Pixels Kunook Chung, Jingyang Sui, Brandon Demory, Chu-Hsiang Teng and Pei- Cheng Ku* Department of Electrical Engineering and Computer Science, University of Michigan, Ann Arbor, Michigan 48109, USA * Monolithic integration of individually addressable light-emitting diode (LED) color pixels is reported. The integration is enabled by local strain engineering. The use of a nanostructured active region comprising of one or more nanopillars allows color tuning across the visible spectrum. In the current work, integration of amber, green and blue pixels is demonstrated. The nanopillar LEDs exhibit an electrical performance comparable to that of a conventional thin-film LED fabricated on the same wafer. The proposed platform uses only standard epitaxy and a similar process flow as a conventional LED. It is also shown that the emission intensity can be linearly tuned without shifting the color coordinate of individual pixels. 1

2 The ability to integrate light emitters of different colors on a single chip platform is highly desirable for applications in display 1 3, lighting 4, 5, biosensors 6, and optogenetics 7. Despite the remarkable efficiency, reliability, and high output brightness achievable by indium gallium nitride (InGaN) based light-emitting diodes (LEDs) 4, 8, 9, the development of chip-scale integration of multi-color emitters has been largely dominated by organic LEDs. Multi-color chips based on InGaN LEDs are attractive for applications such as microdisplays that require high output intensity in an ultracompact form factor. Because light emission from InGaN quantum wells (QWs) exhibit a narrow spectral linewidth, typically around 15 25nm in wavelength, integration of multiple colors across a wide spectral range cannot be easily achieved using a conventional InGaN QW structure. Multi-color integration using pick-andplace 10, 11, multiple epitaxial steps 12 14, selective area epitaxy in molecular beam epitaxy 15, and selectively injecting carriers into one of a few QW regions 16 have been reported. These approaches either have inherently low manufacturing throughput or require non-standard thin-film epitaxy. The goal of this work is to develop a chip-scale multi-color LED platform using existing manufacturing infrastructure. The operating principle is local strain engineering which can simultaneous produces blue, green, and red color emission from standard InGaN QWs The strain relaxation is controlled by patterning the InGaN QW into an array of nanoscale pillars. The emission color is tunable by the nanopillar diameter and can be controlled locally in a sub-micron spatial resolution 20. In this paper, we report individually addressable amber, green and blue LED pixels integrated and processed together on a single chip platform. We also show the intensity of each color channel can be independently controlled without any color shift, a critical requirement for many applications. Figure 1 shows a schematic of the multi-color device comprising of three color pixels, individually wired but sharing a common n-contact. Each color pixel further consists of an array of nanopillars lithographically defined. The diameter of nanopillars is varied in different color pixels to control the emission color. In addition, test pixels with a variety of other nanopillar diameters as well as thin-film active regions were fabricated on the same sample using identical processes to study the dependence of emission wavelength and nanopillar diameter. All LED pixels have the same dimensions of 100µm by 100µm. The sample was grown by metal-organic chemical vapor deposition (MOCVD) on a two-inch 2

3 unpatterned c-plane sapphire substrate. The active region consists of five periods of InGaN/GaN QWs. The QW thickness is 2.5nm. The barrier thickness is 12nm. A 20-nm-thick Mg-doped Al 0.2 Ga 0.8 N electron blocking layer was inserted underneath a 150-nm thick Mg-doped p-type GaN layer. FIG. 1. A schematic of top-down fabrication of nanopillar LED arrays with various diameters. The small-sized individual light emitters can not only generate strain-induced multi-color emissions but also enable the high density integration of RGB color pixels with submicron spatial resolution (L). The LED fabrication followed a similar procedure used in a typical top-emitting LED structure after the nanopillars were defined using electron-beam lithography and a hybrid dry-wet etching process. Nickel was used as the etch mask and was kept throughout the entire LED process to protect the surface of p-gan. Etching of the nanopillars consisted of two steps. First, the pillars were dry etched using inductively-coupled plasma reactive ion etching (ICP-RIE). Wet etch using 2% buffered KOH was then performed to achieve the final diameter as well as to remove any damaged materials at the surface from dry etch. The total etching depth was measured to be 300nm, around 50nm over the active region. After defining the nanopillars, the sample was planarized using spin-on-glass (SOG). To better isolate the p- contact and n-gan, a thin 50nm Si 3 N 4 layer was conformally coated on GaN using plasma-enhanced chemical vapor deposition (PECVD). After planarization, the insulating layers were etched back using ICP-RIE to expose the tips of the nanopillars. The Ni etch mask was also removed by HNO 3 solutions before metallization of p- and n- contacts. The p-contact (Ni/Au) was thermally annealed in air. The sample was first characterized for electroluminescence (EL) using a DC source meter (Keithley 4200) and a spectrometer (Ocean optics USB2000). All measurements were taken at room temperature 3

4 (~300K) with no heat sink or active cooling. Figure 2 shows current density voltage (J V) curves of various LED pixels with different nanopillar diameters. The I V characteristics show good rectifying behaviors for all pixels and low leakage current, measured to be A at 5 V, suggesting a negligible impact of a nanostructural active region on the electrical performance of the LED. There are two reasons for this. First, the QWs are flat and therefore current crowding typically expected in nanostructural LEDs is minimal in our case 21, 22. Second, the strain induced potential confinement helps keep the carriers toward the center of the nanopillar. The increased current density through a smaller nanopillar device is attributed to the enhanced radiative efficiency due to the reduction of the quantumconfined Stark effect (QCSE) 23. FIG. 2. J V curves of 70-nm nanopillar, 100-nm nanopillar, 800-nm nanopillar, and thin film LEDs. Figure 3 shows the EL spectra and how the peak emission wavelengths change with the bias voltage. Figure 3(a) demonstrates wavelength tuning from amber to blue by changing the diameter of the constituent nanopillars. Figure 3(b) shows a more detailed dependence of the EL peak emission versus the nanopillar diameter. The data was taken by biasing each LED pixel with the lowest possible bias voltage while still large enough for a good signal-to-noise ratio to be obtained. This is to avoid any ambiguity due to piezoelectric field screening which will be discussed below. Note that for larger nanopillars, there is more variation in the data shown in Fig. 3(b). This is likely due to QW thickness 4

5 fluctuation across the wafer. But overall, the experimental results in Fig. 3(b) can be fitted well by the equation E ph = E o B m [1 sech( R)] which was derived from a simple 1D strain relaxation model 23. In the equation, E ph is the photon energy of the nanopillar output, R is the nanopillar radius, E 0 is the photon energy corresponding to the unstrained InGaN QW, and E 0 B m is the photon energy of a fully strained InGaN QW. 1/ is the characteristic length for strain relaxation which has been shown to be dependent on the indium composition in the QW 19. In our sample, 1/ was found to be 24nm, a trend consistent with previous results 20, 23. FIG. 3. EL characteristics of the top-down nanopillar LEDs with various diameters. (a) Roomtemperature EL spectra and photograph images of blue (487nm), green (512 nm), and red (600 nm) EL emissions obtained from 50nm, 100nm, 800nm, and thin-film LED pixels, respectively. (b) The measured EL emission wavelength as a function of nanopillar diameter (black circles) and the fitting curve obtained by a 1D strain relaxation theory. (c) Dominant EL peak positions at various applied bias 5

6 voltages. The data points for 100nm and 50nm pixels are lacking for bias voltage 3V as the emission intensities were too low due to smaller active region areas (8.7% and 2.2% of the active region area of the thin film pixel for the 100nm and 50nm pixels, respectively). The drastic reduction of the active layer area is due to the fixed spacing chosen for the current device design, mainly to simplify the planarization process. The stronger QCSE in a larger diameter nanopillar is expected to make the emission wavelength more sensitive to the applied bias voltages. Figure 3(c) shows how the emission wavelengths of different LED pixels vary with the bias voltage. It can be observed that while the smaller diameter (50nm and 100nm) pixels show relatively stable emission wavelength versus the bias, the larger diameter ( 800nm) pixels show a significant blue shift, over 40nm as the bias is increased from 2.8 to 4V which is attributed to the screening of the piezoelectric field when the bias voltage increases. Figure 3(c) also confirms that the wavelength tuning from amber to blue is due to localized strain relaxation instead of from the screening of the piezoelectric field. In applications, when multiple color LED pixels are monolithically integrated on the same chip, it is important to be able to adjust the relative intensity of different color pixels without changing the color coordinate of each pixel. To demonstrate the feasibility of the proposed LED pixels for practical applications, we used pulse frequency modulation while fixing the magnitude of the bias voltage. The width of each individual pulse was fixed at 400µs while the time interval between two pulses is modulated using a function generator (Agilent 33250A). The frequency range was tuned between 200 and 2000 Hz, corresponding to a duty cycle between 8% and 80%, respectively. Figure 4 shows the EL intensity recorded from each LED pixel as a function of the modulation frequency. All EL intensities were normalized with respect to the intensity value at 80% duty cycle, after subtracting the background and integrating over the spectrum. It can be observed that the relative EL intensities for all colors were nearly linear with respect to the duty cycle of the applied current signal (solid line). Importantly, the emission color remained very stable at different intensities as shown in the inset of Fig. 4. 6

7 FIG. 4. Relative integrated EL intensity and EL emission wavelength of the various nanopillar and thin film LEDs obtained by a pulse frequency modulation technique. The inset shows the emission wavelengths at different modulation frequencies. In summary, monolithic integration of amber, green and blue LED pixels on a single chip platform has been demonstrated. The color tuning is enabled by local strain engineering using a nanopillar structure defined by standard lithography and etching. In our current sample, the color tunes from around 487nm to 602nm. Red-green-blue integration is expected to be possible by increasing the QW emission to 635nm, as previously demonstrated by photoluminescence measurements, 20 while the nanopillar LED performance can be further improved, e.g. with a narrower linewidths for the long-wavelength pixels, with the advances in the epitaxy of red InGaN QWs. The nanopillar LED pixels were fabricated using a typical LED process, including MOCVD growth and standard lithography and etching, making them easily scalable and manufacturable. The electrical performance of the nanostructured active region was shown to be comparable to a conventional thin-film LED. The emission wavelength of the nanopillar LED was shown to be controlled by the nanopillar diameter following a simple hyperbolic function. Despite of QCSE, the colors of the nanopillar LED pixels were shown to be stable when their intensities are linearly tuned, using the pulse frequency modulation scheme. The results suggested the viability of the proposed LED pixels for practical applications such as microdisplays. 7

8 Acknowledgements This work was supported by Samsung (N019887) for fabrication and device design and National Science Foundation (DMR ) for growth and fundamental properties. We thank Prof. Hui Deng at the University of Michigan and her group for fruitful discussions. References [1] S. R. Forrest, Nature 428, 911 (2004). [2] J. Y. Tsao, J. Y., IEEE Circuits Devices 20, 28 (2004). [3] T.-H. Kim, K.-S. Cho, E. K. Lee, S. J. Lee, J. Chae, J. W. Kim, D. H. Kim, J.-Y. Kwon, G. Amaratunga, S. Y. Lee, B. L. Choi, Y. Kuk, J. M. Kim, K. Kim, Nat. Photonics 5, 176 (2011). [4] S. Nakamura, Solid State Commun. 102, 237 (1997). [5] Z. Shen, P. E. Burrows, V. Bulović, S. R. Forrest, M. E. Thompson, Science, 276, 2009 (1997). [6] H. Xu, J. Zhang, K. M. Davitt, Y.-K. Song, A. V. Nurmikko, J. Phys. D Appl. Phys. 41, (2008). [7] F. Wu, E. Stark, P.-C. Ku, K. D. Wise, G. Buzsáki, E. Yoon, Neuron 88, 1136 (2015). [8] F. A. Ponce, D. P. Bour, Nature 386, 351 (1997). [9] M. R. Krames, O. B. Shchekin, R. Mueller-Mach, G. O. Mueller, L. Zhou, G. Harbers, M. G. Craford, J. Disp. Technol. 3, 160 (2007). [10] Y. Huang, X. F. Duan, C. M. Lieber, Small 1, 142 (2005). [11] H. S. El-Ghoroury, C.-L. Chuang, Z. Y. Alpaslan, SID Symp. Dig. Tech. Pap., 46, 371 (2015). [12] H. P. T. Nguyen, K. Cui, S. Zhang, S. Fathololoumi, Z. Mi, Nanotechnology, 22, (2011). [13] K. Kishino, K. Nagashima, K. Yamano, Appl. Phys. Express, 6, (2013). [14] R. Wang, H. P. T. Nguyen, A. T. Connie, J. Lee, I. Shih, Z. Mi, Opt. Express 22, A1768 (2014). [15] Y. H. Ra, R. J. Wang, S. Y. Woo, M. Djavid, S. M. Sadaf, J. Lee, G. A. Botton, Z. Mi, Nano Lett. 16, 4608 (2016). [16] H. S. El-Ghoroury, M. Yeh, J. C. Chen, X. Li, C.-L. Chuang, AIP Adv. 6, (2016). [17] P. Yu, C. H. Chiu, Y.-R. Wu, H. H. Yen, J. R. Chen, C. C. Kao, H.-W. Yang, H. C. Kuo, T. C. Lu, W. Y. Yeh, S. C. Wang, Appl. Phys. Lett. 93, (2008). 8

9 [18] Y. Kawakami, A. Kaneta, L. Su, Y. Zhu, K. Okamoto, M. Funato, A. Kikuchi, K. Kishino, J. Appl. Phys. 107, (2010). [19] Y.-J. Lu, H.-W. Lin, H.-Y. Chen, Y.-C. Yang, S. Gwo, Appl. Phys. Lett. 98, (2011). [20] C.-H. Teng, L. Zhang, H. Deng, P.-C. Ku, Appl. Phys. Lett. 108, (2016). [21] F. Limbach, C. Hauswald, J. Lähnemann, M. Wölz, O. Brandt, A. Trampert, M. Hanke, U. Jahn, R. Calarco, L. Geelhaar, H. Riechert, Nanotechnology, 23, (2012). [22] Y. Tchoe, C.-H. Lee, J. B. Park, H. Baek, K. Chung, J. Jo, M. Kim, G.-C. Yi, ACS Nano, 10, 3114 (2016). [23] L. Zhang, L.-K. Lee, C.-H. Teng, T. A. Hill, P.-C. Ku, H. Deng, Appl. Phys. Lett. 104, (2014). [24] L. Zhang, T. A. Hill, C.-H. Teng, B. Demory, P.-C. Ku, H. Deng, Phys. Rev. B 90, (2014). 9

10

11

12

13