A nonpolar edge emitting thin film InGaN laser diode has been separated from its native substrate by mechanical tearing with adhesive tape, combining the benefits of Epitaxial Lateral Overgrowth (ELO) and cleavability of nonpolar GaN crystal. The essence of ELO is mainly to weakening strength between native substrate and the fabricated laser device on top of it. We report a 3 mm long laser bar removed from its native GaN substrate. We confirmed edge emitting lasing operation after cleaving facets on a separated thin bar. Threshold current density of the laser was measured to be as low as 2.15 kA/cm2.
© 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
There has been significant development of c-plane GaN laser diode technology since the first demonstration by Nakamura et al. . GaN laser diodes have been quickly adopted into several applications such as optical storage, lighting, projectors, and displays. However, even after nearly two decades in market, GaN based laser diodes are still expensive due to the limited size and cost of the bulk GaN native substrates commonly used. Currently, gallium is used in many applications such as GaAs-based vertical cavity surface emitting lasers (VCSEL) for smart phones, GaN power devices for electronics, GaN and AlInGaP LEDs for signage and displays. Recently, GaN based micro-LEDS and automotive laser lighting seem to be gathering momentum. If this demand continues to grow at this rate gallium metal will face a problem. One must take measures to protect the use of gallium [2,3].
As a potential solution to this cost problem we report successful operation of a thin edge emitting GaN laser after removal from its native substrate, preserving the substrate for reuse. There have been several reports on removing devices from a free standing GaN substrate [4–6]. However, due to the involved complexity and scaling limitations no one has previously demonstrated removal of working edge-emitting lasers.
Most commercial GaN edge emitting lasers have individual die sizes of 100-200 µm wide, 400-1600 µm long and 60-100 µm thick. They are grown on bulk c-plane GaN substrates typically 350 µm thick, which are later thinned to enable facet formation by cleaving and to reduce the thermal impedance. It is this thinning that accounts for the excessive waste of the expensive growth substrate. If industry moves to nonpolar or semipolar substrates to benefit from the superior performance afforded by these alternative orientations , the small size and very high cost of these substrates will make recycling even more important. Accordingly, we have chosen m-plane lasers to demonstrate die removal as the first step toward substrate recycling. In an industrial setting, the m-plane may also be cleaved for reliable facet formation . Our technique utilizes cleavability of m-plane GaN crystal and a commercially available adhesive tape with changing the temperature.
2. Experimental procedure
Bulk m-plane substrates with a −1 degree miscut orientation to the (0001) plane were used. A 1µm thick SiO2 mask with window areas for selective growth was formed on the surface of the substrate. The width of the window areas and wing areas varied from 20 to 45µm and 55-80µm respectively. The length of the stripe along the substrate was 4mm and was directed parallel to the c-axis <0001>.
On this masked native substrate ELO was performed using MOCVD at 100 Torr, V/III ratios above 2000, and at a temperature between 1010 and 1210°C. Growth was stopped after reaching ELO layer thickness about 15-20 µm. The SiO2 mask was designed such that at this thickness there was still a gap between adjacent ELO stripes. In other words, the epitaxially grown ELO stripes were discretely distributed along the a-axis without coalescing.
Next, laser structures were grown on these epitaxially grown ELO features. We designed our laser structure without an AlGaN cladding, as reported previously by UCSB in nonpolar InGaN/GaN laser diodes . The laser comprised an n-GaN cladding layer (n = 2x1019 cm−3), consisting 10 periods of an In0.06Ga0.94N/GaN (2nm/2nm) superlattice as a lower waveguide, a 3 period In0.12Ga0.88N/In0.06Ga0.93N/GaN (5nm/7nm/3nm) multiple quantum well active layer, a 10 period InGaN/GaN (2nm/2nm) superlattice as upper waveguide, a p-Al0.26Ga0.74N electron blocking layer (EBL) (p = 1.0x1019 cm−3, 7nm), a p-GaN cladding layer (p = 2.5x1019 cm−3) and p + -GaN contact layer (p = 1x1020 cm−3). Ridge height was about 250nm and the thickness from the active layer to the surface was 300µm. Doping level was measured by SIMS (Secondary Ion Mass Spectroscopy).
The ridge structure is fabricated using a conventional photolithography process. A 200nm thick SiO2 mask was used as current limitation current limitation. The SEM image of the fabricated laser ridge structure with a p-electrode (ITO/Ti/Au = 150/250/3000nm) on the epitaxially grown ELO feature is shown in Fig. 1. Adhesive tape was then applied to the p-electrode side of the wafer and the laser stripes were removed from the growth substrate by cleaving the GaN in the relatively narrow window region as shown in Fig. 2. The back sides of the removed die were then coated with an n-electrode Al/Ni/Au = 50/100/300nm using electron beam deposition while the dies were still attached to tape. After that two facets were cleaved using a tweezer, one facet at each end of the die.
Laser measurements were then performed at room temperature with a pulse width of 500ns and a repetition rate of 10 kHz, corresponding to a duty cycle of 0.5%. The output power was measured by collecting the light emitted from uncoated facets facing toward the (0001) direction with a calibrated broad-area Si photodiode. Similarly, electroluminescence (EL) spectra was measured by coupling the light emitted from the facet into an optical fiber connected to a spectrometer.
3. Results and discussion
In our experiment ELO was mainly used to weaken the link between the fabricated laser bars and their native substrate. The additional benefit from ELO to reduce the defect density on bulk substrates was not carefully explored. We measured the dislocation density on an epi-ready m-plane substrate and on an epitaxially grown ELO feature by cathodoluminescence (CL) and the results are shown in Figs. 3(a) and 3(b), respectively. The mask pattern for this sample was a 20µm window and 60µm masked region. The width of the epitaxial ELO layer was about 55µm. As can be seen in Fig. 3(a), the substrate had a fluctuation of in-plane distribution of the dislocation density, ranging from 1 to 5 x106cm-2. We also observed defect clusters distributed throughout the surface. Defect densities at these defect clusters was very large compared to the reported value in the non-clustering region. In the figure, dislocations parallel to the c-direction appear as dark lines and dislocations with a component parallel to the m-direction as black dots. We were not able to observe the presence of Stacking Faults using CL investigation.
The CL data on the ELO feature can be divided into two regions as shown in Fig. 3(b). The window region directly above the native substrate had a defect density of 5x106 cm−2 - 2x107 cm−2 and no defect clusters were observed. The defects are distributed uniformly throughout the window region. On the other hand, the dislocation density on the wing area was less than 2x105cm−2. We consider that further reduction in defect density is possible by optimizing the growth condition. We think that the reduction of defect number on the ELO feature was caused by bending of dislocations during the growth. Since the wing region is epitaxial lateral overgrowth (ELO), the defect density on the wing region is smaller than in the window region. Thus, we propose by implementing ELO technique one can easily manage to place the ridge on a low defect wing region without having pre-knowledge on defect density map of the substrate.
The laser bars were grown for laser process on the pattern which is set a 45µm window and 55µm masked region. After growing, the laser bar thickness, bar bottom and top widths were respectively 16.8 µm, 67.2 µm and 51.6 µm as shown in Fig. 1. The top flat surface of the bar was wide enough to accommodate a 6 µm ridge structure. After removal from the substrate, the reported lasers had the dimensions of only 16.8 µm thick, 67.2 µm wide, with cavity lengths of 1.6 mm or 3 mm.
For example, commercial 2-inch laser wafer fabrication typically employs a polishing technique from the backside of the substrate after fabricating laser devices on top surface of the substrate. To facilitate easy handling without breakage the processed wafer is thinned to a thickness of 80 µm to 60 µm. With the process reported here the thickness will be determined by the ELO feature height, and we expect that by automating the entire process a thickness under 10 µm can be handled without difficulty. In this report we chose a little larger thickness for our device demonstration, to accommodate human handling with tweezers.
The growth of InGaN or AlGaN layers on patterned c-plane substrate, such as occurs during ELO, is challenging as edge growth from each ELO feature will ruin the thickness uniformity along the width of ELO wing . As a result, restrictions will be imposed on the location and size of the ridge. However, in our case, no edge growth was seen at the edge of the ELO features, as shown in Fig. 2. It might be attributed to AlGaN cladding free laser design and optimization of the growth condition. This result can be translated into a further reduction in die width, increasing the number of devices gained from one wafer, and improving die yield by further weakening the link between the native substrate and the fabricated device. Additionally, reduced footprint devices will find applications in future devices for augmented/virtual reality applications, pico-projectors etc. In this article we could successfully remove a 67.2 µm die containing a laser ridge and contact pads.
Figure 4(a) shows the laser die after removing it from m-plane GaN was 1.2 mm long without breaking. Position of the cleavage is around the interface of the ELO layer and the native substrate. The image of the backside of the laser bar and surface profile of the surface of the substrate after removing the laser bar are shown in Figs. 4(b) and 5, respectively. The window area between line A and line B is about 45µm wide. We assume GaN cleavage in the window region started at line A and extended toward line B. The area near line A is smoother than the area near line B. The area near line B shows wavy lines, which we assume result from a non-uniform stress applied by the adhesive tape, although this is not yet well understood. The wavy lines on the substrate surface have been measured after removing laser bars, as shown in Fig. 5. The difference in the height of wavy lines is about ± 1.5 µm.
Figure 6 shows a High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image of the laser structure at the center of the laser die. As can be seen, precisely controlled and flat epitaxial layers were grown despite having a nonplanar ELO base layer of height 16.8µm.
Figures 7(a) and 7(b) show the current-light (I-L) characteristics and the lasing spectrum of a removed laser die, without coated facets. The cavity length and the ridge width are 3.0mm and 6µm respectively. The cavity length was intentionally chosen longer compared to commercially available laser diodes to demonstrate that long laser bars can be removed using this technique. The threshold current density was 2.15 kA/cm2 and the slope efficiency was 0.11 W/A with pulsed operation. The threshold voltage was approximately 6.5 V. We also confirmed lasing in a laser bar having a cavity length of 1.6 mm. The threshold current for this short cavity laser was 3.32 kA/cm2 (data not shown.) The lasing wavelength was 408 nm as shown in Fig. 7(b) and the inset show a far field pattern of the lased die. Facets on these laser bars were made by simply breaking the die with a tweezer, which resulted in uncontrollable mirror loss. We believe with a complete automated setup at industrial scale and an optimized design will extract more promising characteristics from lasers fabricated using this technique.
In conclusion, we have demonstrated for the first time an edge emitting GaN laser removed from its native free-standing GaN substrate by mechanical cleaving technique, allowing recycling of the substrate. In this paper, a threshold voltage as low as 6.5V and a threshold current density as low as 2.15kA/cm2 was obtained, with a cavity length of 3.0mm. Our removal method uses the advantages of ELO to weaken the bond between the laser and its substrate. The reported laser die has a smaller footprint than commercially available laser devices, for lower cost. Further cost reduction is obtained by reducing material waste generated by conventional substrate thinning. The small footprint of these devices may benefit emerging applications related to VR/AR or pico-projectors via new packaging methods. This technique can be even more valuable when implemented on an expensive high performance non-polar and semi polar lasers.
This work was supported by the Solid State Lighting and Energy Electronics Center (SSLEEC) at UCSB. A portion of this work was done in the UCSB nanofabrication facility.
1. S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa, T. Yamada, T. Matsushita, H. Kiyoku, and Y. Sugimoto, “InGaN-based multi-quantum-well-structure laser diodes,” Japanese Journal of Applied Physics , 35(2), L74–L76 (1996).
2. L. Zhan, F. Xia, Q. Ye, X. Xiang, and B. Xie, “Novel recycle technology for recovering rare metals (Ga, In) from waste light-emitting diodes,” J. Hazard. Mater. 299, 388–394 (2015). [CrossRef] [PubMed]
3. M. Bergamos and R. Höltig, “Recycling options for LEDs and LED products,” 2012 Electronics Goes Green 2012+ conference paper (2012).
4. J. T. Leonard, D. A. Cohen, B. P. Yonkee, R. M. Farrell, T. Margalith, S. Lee, S. P. DenBaars, J. S. Speck, and S. Nakamura, “Nonpolar III-nitride vertical-cavity surface emitting lasers incorporating an ion implanted aperture,” Appl. Phys. Lett. 107(1), 011102 (2015). [CrossRef]
5. S. W. Bedell, C. Bayram, K. Fogel, P. Lauro, J. Kiser, J. Ott, Y. Zhu, and D. Sadana, “Vertical Light-Emitting Diode Fabrication by Controlled Spalling,” Appl. Phys. Express 6(11), 112301 (2013). [CrossRef]
6. C. Youtsey, R. McCarthy, R. Reddy, K. Forghani, A. Xie, E. Beam, J. Wang, P. Fay, T. Ciarkowski, E. Carlson, and L. Guido, “Wafer-scale epitaxial lift-off of GaN using bandgap-selective photo enhanced wet etching,” Phys. Status Solidi, B Basic Res. 254(8), 1600774 (2017). [CrossRef]
7. P. Waltereit, O. Brandt, A. Trampert, H. T. Grahn, J. Menniger, M. Ramsteiner, M. Reiche, and K. H. Ploog, “Nitride semiconductors free of electrostatic fields for efficient white light-emitting diodes,” Nature 406(6798), 865–868 (2000). [CrossRef] [PubMed]
8. D. A. Stocker, E. F. Schubert, W. Grieshaber, K. S. Boutros, and J. M. Redwing, “Facet roughness analysis for InGaN/GaN lasers with cleaved facets,” Appl. Phys. Lett. 73(14), 1925–1927 (1998). [CrossRef]
9. D. F. Feezell, M. C. Schmidt, R. M. Farrell, K. C. Kim, M. Saito, K. Fujito, D. A. Cohen, J. S. Speck, S. P. Denbaars, and S. Nakamura, “AlGaN-Cladding-Free Nonpolar InGaN/GaN Laser Diodes,” Jpn. J. Appl. Phys. 46(4), L284–L286 (2007). [CrossRef]
10. I. Mahaboob, J. Marini, K. Hogan, E. Rocco, R. P. Tompkins, N. Lazarus, and F. S. Sandvik, “Selective Area Epitaxial Growth of Stretchable Geometry AlGaN-GaN Heterostructures,” J. Electron. Mater. 47(11), 6625–6634 (2018). [CrossRef]