Vertical Light-Emitting Diode Fabrication by Controlled Spalling - icorlab
Applied Physics Express 6 (2013) 112301 http://dx.doi.org/10.7567/APEX.6.112301
Vertical Light-Emitting Diode Fabrication by Controlled Spalling Stephen W. Bedell , Can Bayram, Keith Fogel, Paul Lauro, Jonathan Kiser, John Ott, Yu Zhu, and Devendra Sadana IBM T. J. Watson Research Center, Yorktown Heights, NY 10598, U.S.A. E-mail: [email protected] Received July 30, 2013; accepted October 1, 2013; published online October 18, 2013 A fracture-based layer transfer technique referred to as controlled spalling was used to separate a conventional InGaN/GaN multiple quantum well light-emitting diode (LED) structure from a 50 mm sapphire wafer enabling the formation of vertical spalled LEDs (SLEDs). A 25-m-thick tensile Ni layer was electrodeposited on the surface of the wafer, followed by the application of a polyimide tape layer. By mechanically guiding the tape layer, a 3-m-thick layer of the LED epitaxy was removed. Transmission electron microscopy imaging indicated that spalling preserved the quality of the epitaxial layers, and electroluminescence verifies the operation of the SLED. # 2013 The Japan Society of Applied Physics
ramatic advances have been made in solid state lighting within the past few decades.1) In addition to improvements in the starting material quality and light extraction techniques, thermal management has played an important role specifically in the development of highpower light emitting diode (LED) lighting.2) Sapphire or patterned sapphire substrates (PSS) are used extensively as epitaxial growth templates for GaN/InGaN based materials. Although relatively cost-effective, the low thermal conductivity of sapphire restricts heat flow from the device thereby limiting the maximum operating power. The superior heat conductivity of silicon carbide has made it the substrate of choice for high-power applications; however, the cost is significantly higher as well. The concept of a vertical LED (VLED) structure improves many aspects of these devices including current spreading and thermal management. In order to fabricate a VLED device, the epitaxial layers must be removed from the sapphire substrate. There are essentially three different ways this has been accomplished: laser liftoff (LLO), chemical liftoff (CLO), and backside grinding of the sapphire. In laser liftoff,3–5) an ultraviolet excimer laser is projected through the backside of the sapphire wafer and results in decomposition of the GaN at the GaN/Al2 O3 interface. In chemical liftoff, a specific layer within the epitaxial stack is preferentially etched and subsequently releases the overlying device layers. This has been demonstrated using selective etching of layers composed of CrN,6) GaN:Si,7) ZnO8) or using photoelectrochemical (PEC) etching techniques.9) The drawback to laser liftoff is that damage to the GaN layers can occur during irradiation resulting in reduced device performance and yield.10) The chemical liftoff processes that incorporate new materials into the epitaxial stack are faced with the challenge of maintaining high-quality GaN growth, and large-area CLO processes are often difficult to implement in practice. We have reported previously on a layer transfer technique referred to as controlled spalling.11,12) This approach uses a stressed layer deposited on the surface of a substrate to induce lateral fracture within the underlying substrate and ‘‘peel off’’ a controlled thickness from the surface. We have applied controlled spalling to the fabrication of flexible, high-efficiency GaAs-based photovoltaics13) as well as highperformance nanoscale Si circuits14) and in all cases the defectivity of the material is unchanged by this process. In this work, we use controlled spalling to remove multiquantum well (MQW)-based InGaN/GaN epitaxial LED layers grown on PSS substrates and demonstrate successful
D
optical emission from the resulting vertical spalled LED (SLED) layers. The physical origin of spalling mode fracture comes from the combination of opening stress (mode I) and shear stress (mode II) due to the presence of a surface stressor layer acting on a subsurface crack tip.15,16) For a tensile surface stressor layer, a crack will propagate downward into the substrate to a depth where the shear component is zero. If the crack continues below this point, it will be deflected back towards the surface by the change in sign of the shear stress component. Therefore, the fracture depth is stable within the substrate as the crack tip trajectory oscillates about the zero shear position due to the corrective action of the shear field. For fracture to be energetically possible within the substrate, the magnitude of the opening mode stress field (given by the stress intensity factor KI ) must be equal to, or greater than, the material fracture toughness KIC at the zero shear position (KII ¼ 0). For a particular choice of stressor material (e.g., Ni) there will be a combination of stress and thickness above which fracture is possible for a given substrate type (using KIC from literature17)). In order to apply the controlled spalling process to the epitaxial GaN material system using Ni as the stressor layer, the required loading conditions (stress and thickness) were determined experimentally. If the loading conditions are too high, spalling can occur spontaneously during Ni deposition and generally leads to unusable layers. Therefore, it is important to determine the combinations of stress and thickness that energetically permit fracture, but do not lead to spontaneous spalling. PSS substrates (c-plane), 50 mm in diameter, were used as growth templates for Inx Gað1xÞ N/GaN MQW LED structures using a commercial metal–organic chemical vapor deposition (MOCVD) reactor. The epitaxial structure consisted of a thin low-temperature GaN buffer layer, followed by a 2-m-thick undoped GaN layer (u-GaN), a 3-m-thick Si doped n-GaN layer, InGaN/GaN MQW layers, and Mgdoped p-GaN capping layers. After growth, the wafers were annealed in a N2 ambient to activate the p-GaN layer. Figure 1 shows a schematic of the controlled spalling process used in this work. A Ni stressor layer was deposited onto the surface of the wafers by first depositing a thin (75 nm) adhesion layer of Ti followed by 500 nm of Ni by DC magnetron sputtering. Electroplating using a NiCl2 based solution at room temperature resulted in approximately 400 MPa tensile stress as measured by wafer bowing. We then determined that approximately 25 m of Ni deposited onto the sputtered Ni was required to permit controlled spalling of the GaN. After deposition of the thick Ni layer,
112301-1
# 2013 The Japan Society of Applied Physics
Appl. Phys. Express 6 (2013) 112301
S. W. Bedell et al.
Tape
(a)
Ni (tensile) GaN Epitaxy Sapphire
Fig. 2. Image of a spalled 50-mm-diameter InGaN/GaN MQW LED structure (SLED) mounted onto a stainless steel frame using the handling tape.
(b)
Controlled Spalling 200 Schematic illustration of the controlled spalling process used to remove InGaN/GaN MQW LED layers from a sapphire substrate. The GaN/Al2 O3 substrate was prepared for spalling by (a) first depositing a thin Ti (75 nm)/Ni (500 nm) seed layer by sputtering, followed by electrodeposition of a 25-m-thick Ni layer (400 MPa tensile), then application of a tape layer. Removal of the epitaxial layers from the sapphire substrate was performed by (b) peeling the tape layer away from the surface. Fig. 1.
the wafers were rinsed and dried and held flat using a custom-made vacuum chuck table with an integrated tape dispenser. A layer of polyimide tape (25 m) was rollapplied to the surface of the Ni, and fracture occurred upon gently pulling the tape away from the wafer surface, as illustrated in Fig. 1(b). To prevent excessive bending of the thin film after spalling, the outer portions of the tape are used to secure the sample to a stainless steel frame. The entire process, from metal deposition to layer removal, was performed at room temperature. Figure 2 shows an entire 50-mm-diameter InGaN/GaN epitaxial structure removed from the sapphire wafer by controlled spalling and mounted in a handling frame. The surface artifacts observed on the sample shown in Fig. 2 are due to roughness caused by fracture depth oscillations around the equilibrium spalling depth and are not cracks in the film. Surface profilometry was performed and the results are shown in Fig. 3. Although the resulting surface roughness after spalling varies with length scale, the maximum amplitude occurs with a spatial period of about 700 m and an amplitude of about 70 nm. Fracture propagation speed directly affects this spatial period, with faster fracture resulting in a longer oscillation period. In this work, fracture propagation across the 50 mm wafer surface took on the order of one second leading to a fracture velocity of about 50 mm/s. In the case of LED structure spalling, this roughness may offer a natural means of surface texturing for improved light extraction. Figure 4 shows a cross-sectional scanning electron microscopy (XSEM) image of the InGaN/GaN LED structure before [Fig. 4(a)] and after [Fig. 4(b)] controlled spalling. Coloration has been added to the figure to indicate the various regions of the epitaxial structure. By comparison of the two
150
Height (nm)
100 50 0 -50 -100 -150 -200 0
500
1000 1500 2000 Distance (μm)
2500
3000
Fig. 3. Surface profilometry data showing the characteristic roughness associated with the as-fractured surface after spalling. The largest amplitude is seen at a spatial period of about 700 m with a magnitude of about 70 nm.
images it is evident that fracture propagated approximately 3 m below the surface within the n-GaN region. This has the twofold benefit of having immediate access to the n-type contact area without etching, and the possibility of reusing not only the original sapphire wafer, but also the u-GaN buffer region as well. Spalling depth is generally controlled by the Ni thickness. We have observed a spalling depth change of approximately 500 nm per micron of Ni thickness change for the GaN on PSS system. Also, due to the extremely high fracture toughness of sapphire, GaN epitaxy layers could be removed entirely from non-PSS (planar) sapphire substrates. Figure 5 shows cross-sectional transmission electron microscopy (XTEM) images of the SLED structure. The low magnification image [Fig. 5(a)] shows the full thickness of the SLED structure indicating a total layer thickness of about 3.1 m and no spalling-related defectivity was observed in the sample. Figures 5(b) and 5(c) show high-resolution imaging of the InGaN/GaN MQW structure and the 10-period active layer of the device, respectively. Figure 6 shows the electroluminescence spectrum taken from direct probing (no metallization) of the exposed n-GaN
112301-2
# 2013 The Japan Society of Applied Physics
Appl. Phys. Express 6 (2013) 112301
(a)
S. W. Bedell et al. 1.2
(b)
Bulk
1
Normalized Intensity
p-GaN MQW
~ 3μm
n-GaN
n-GaN
u-GaN
u-GaN
SLED
0.8 0.6 0.4
λp=529.2 nm (SLED) λp=533.9 nm (Bulk)
0.2 0 475
Sapphire
(b)
Spalled n-GaN
10 nm
(c)
Ti 0.5 μm
Ni
575 Wavelength (nm)
625
675
Fig. 6. Electroluminescence curves for the SLED structure and the conventional (bulk) structure. The inset shows the green optical emission by directly probing the exposed n-GaN surface of the SLED device. The inhomogeneous emission is likely caused by localized conduction regions due to the use of a non-ohmic Ti contact to p-GaN.
Fig. 4. Cross-sectional SEM image showing the InGaN/GaN device structure (a) before and (b) after controlled spalling. The image has been colorized to indicate the various regions of the structure. Approximately 3 m of the surface was removed, with fracture occurring near the center of the n-GaN layer.
(a)
525
100 nm
Fig. 5. Cross-sectional TEM image showing (a) entire spalled layer (b) high-resolution image of the MQW structure, and (c) image of the 10-period active device region. The few defects observed in the sample were related to GaN growth suggesting that the material quality is preserved during controlled spalling.
surface of the vertical SLED (inset), as well as a conventional device formed on a small non-spalled portion from the edge of the same wafer (Bulk). The optical emission from the SLED films was inhomogeneous, as seen in the image (Fig. 6 inset). This is because the Ti adhesion layer formed a rectifying contact with the p-GaN surface and con-
duction during forward bias occurred primarily at localized breakdown regions. The peak emission p was measured to be 529.2 nm from the SLED device and 533.9 for the bulk device (with annealed ohmic NiAu contacts). This corresponds to an energy shift Eg of 21 meV. Although changes in the stress state of the SLED device occur due to the presence of the tensile Ni and are known to result in Eg shift,18) the energy shift in green InGaN/GaN LED structures such as these is also very sensitive to biasing conditions.19,20) Owing to the inhomogeneous conduction observed in the SLED devices, the biasing conditions between the two devices cannot be properly compared and the origin of the observed Eg is likely a combination of both effects. In addition to the direct probing of the as-spalled surface described above, a portion of the SLED layer was bonded to a Si wafer using In. After bonding the as-spalled n-GaN surface to Si, the polyimide tape was removed and the Ni stressor and Ti adhesion layers were chemically etched revealing the original p-GaN surface. Ni/Au contacts were evaporated; however, no annealing could be performed due to the low melting point of the In bonding material. Figure 7 shows the resulting current density versus bias of the bulk InGaN/GaN device with annealed Ni/Au contacts compared to the In-bonded SLED device with as-deposited Ni/Au contacts. The bonded SLED device showed an expected increase in parasitic resistance as well as a 200 mV increase in turn-on voltage due to the non-annealed contact layers compared to the bulk device. One of the most important differences between controlled spalling and other liftoff technologies is the generality associated with spalling. In other words, because there is no need for specialized layers, transfer can be performed on any optimized epitaxial structure. Therefore, layer transfer of a variety of GaN-based structures (LED, laser, HEMT, etc.) grown on either sapphire, SiC, free-standing GaN, or Si is straightforward. The other attractive aspect of layer transfer by spalling is that it can be accomplished using conventional laboratory equipment; no lasers or other specialized tools are required.
112301-3
# 2013 The Japan Society of Applied Physics
Appl. Phys. Express 6 (2013) 112301
S. W. Bedell et al.
0.4
Current Density (A/cm2)
Bulk 0.3
SLED (bonded) Ni/Au (no anneal)
Ni/Au (annealed) In
Acknowledgments The authors would like to thank Drs. G. Shahidi and T. C. Chen for their encouragement and support of this work.
In
SLED
GaN
0.2
due to a combination of mechanical stress and dissimilar biasing conditions.
1) S. Nakamura, M. Senoh, and T. Mukai: Appl. Phys. Lett. 62 (1993) 2390. 2) M. R. Krames, O. B. Shchekin, R. Mueller-Mach, G. O. Mueller, L. Zhou,
Si
PSS
G. Harbers, and M. G. Craford: J. Disp. Technol. 3 (2007) 160.
3) W. S. Wong, T. Sands, and N. W. Cheung: Appl. Phys. Lett. 72 (1998)
599.
0.1
Area = 1 mm2
4) T. Fujii, Y. Gao, R. Sharma, E. L. Hu, S. P. DenBaars, and S. Nakamura:
Appl. Phys. Lett. 84 (2004) 855. 5) C. A. Tran, C.-F. Chu, C.-C. Cheng, W.-H. Liu, J.-Y. Chu, H.-C. Cheng,
F.-H. Fan, J.-K. Yen, and T. Doan: J. Cryst. Growth 298 (2007) 722.
0 0
1
2
3
4
5
Bias (V)
6) K. Fujii, S. Lee, J.-S. Ha, H.-J. Lee, H.-J. Lee, S.-H. Lee, T. Kato, M.-W.
Cho, and T. Yao: Appl. Phys. Lett. 94 (2009) 242108. 7) J. Park, K. M. Song, S.-R. Jeon, J. H. Baek, and S.-W. Ryu: Appl. Phys.
Lett. 94 (2009) 221907.
Current density versus bias for the bulk InGaN/GaN LED device compared to a SLED device which has been bonded to a Si wafer. The asspalled n-GaN surface was bonded using In, and the tape and Ni layers removed revealing the original p-GaN surface. Ni/Au contacts were deposited but not annealed due to the low melting point of the In bond. The data show an expected increase in parasitic resistance and turn-on voltage due to the non-annealed contacts. Fig. 7.
8) D. J. Rogers, F. H. Teherani, A. Ougazzaden, S. Gautier, L. Divay, A.
20)
Lusson, O. Durand, F. Wyczisk, G. Garry, T. Monteiro, M. R. Correira, M. Peres, A. Neves, D. McGrouther, J. N. Chapman, and M. Razeghi: Appl. Phys. Lett. 91 (2007) 071120. C. Hsieh, H.-S. Chen, C.-H. Liao, C.-Y. Chen, C.-H. Lin, C.-H. Lin, S.-Y. Ting, Y.-F. Yao, H.-T. Chen, Y.-W. Kiang, and C.-C. Yang: IEEE Photonics Technol. Lett. 24 (2012) 1775. Y. S. Wu, J.-H. Cheng, W. C. Peng, and H. Ouyang: Appl. Phys. Lett. 90 (2007) 251110. S. W. Bedell, D. Shahrjerdi, B. Hekmatshoar, K. Fogel, P. Lauro, J. Ott, N. Sosa, and D. Sadana: IEEE J. Photovoltaics 2 (2012) 141. S. W. Bedell, K. Fogel, P. Lauro, D. Shahrjerdi, J. A. Ott, and D. Sadana: J. Phys. D 46 (2013) 152002. D. Shahrjerdi, S. W. Bedell, C. Bayram, C. C. Lubguban, K. Fogel, P. Lauro, J. A. Ott, M. Hopstaken, M. Gayness, and D. Sadana: Adv. Energy Mater. 3 (2013) 566. D. Shahrjerdi and S. W. Bedell: Nano Lett. 13 (2013) 315. M. D. Thouless, A. G. Evans, M. F. Ashby, and J. W. Hutchinson: Acta Metall. 35 (1987) 1333. Z. Suo and J. W. Hutchinson: Int. J. Solids Struct. 25 (1989) 1337. M. D. Drory, J. W. Ager, T. Suski, I. Grzegory, and S. Porowski: Appl. Phys. Lett. 69 (1996) 4044. C. Kisielowski, J. Kru¨ger, S. Ruvimov, T. Suski, J. W. Ager III, E. Jones, Z. Liliental-Weber, M. Rubin, E. R. Weber, M. D. Bremser, and R. F. Davis: Phys. Rev. B 54 (1996) 17745. C. Bayram, J. L. Pau, R. McClintock, and M. Razeghi: Appl. Phys. B 95 (2009) 307. S. Nakamura, T. Mukai, and M. Senoh: J. Appl. Phys. 76 (1994) 8189.
112301-4
# 2013 The Japan Society of Applied Physics
9)
10) 11)
In summary, a previously described layer transfer method called controlled spalling was used to separate a conventional InGaN/GaN MQW LED structure from a 50-mmdiameter sapphire wafer enabling the formation of a vertical spalled LED (SLED). A 25-m-thick Ni layer possessing 400 MPa tensile stress was electrodeposited on the surface of the wafer, followed by the application of a polyimide tape layer. By securing the wafer using a vacuum chuck, and pulling on the tape layer, a continuous 3-m-thick layer of the LED epitaxy was removed. XTEM imaging indicated that spalling preserved the quality of the epitaxial layers, and electroluminescence verifies the operation of the SLED. An electroluminescence energy shift Eg of 21 meV was observed as compared to a conventional device and is likely
12) 13)
14) 15) 16) 17) 18)
19)
Vertical Light-Emitting Diode Fabrication by Controlled Spalling Stephen W. Bedell , Can Bayram, Keith Fogel, Paul Lauro, Jonathan Kiser, John Ott, Yu Zhu, and Devendra Sadana IBM T. J. Watson Research Center, Yorktown Heights, NY 10598, U.S.A. E-mail: [email protected] Received July 30, 2013; accepted October 1, 2013; published online October 18, 2013 A fracture-based layer transfer technique referred to as controlled spalling was used to separate a conventional InGaN/GaN multiple quantum well light-emitting diode (LED) structure from a 50 mm sapphire wafer enabling the formation of vertical spalled LEDs (SLEDs). A 25-m-thick tensile Ni layer was electrodeposited on the surface of the wafer, followed by the application of a polyimide tape layer. By mechanically guiding the tape layer, a 3-m-thick layer of the LED epitaxy was removed. Transmission electron microscopy imaging indicated that spalling preserved the quality of the epitaxial layers, and electroluminescence verifies the operation of the SLED. # 2013 The Japan Society of Applied Physics
ramatic advances have been made in solid state lighting within the past few decades.1) In addition to improvements in the starting material quality and light extraction techniques, thermal management has played an important role specifically in the development of highpower light emitting diode (LED) lighting.2) Sapphire or patterned sapphire substrates (PSS) are used extensively as epitaxial growth templates for GaN/InGaN based materials. Although relatively cost-effective, the low thermal conductivity of sapphire restricts heat flow from the device thereby limiting the maximum operating power. The superior heat conductivity of silicon carbide has made it the substrate of choice for high-power applications; however, the cost is significantly higher as well. The concept of a vertical LED (VLED) structure improves many aspects of these devices including current spreading and thermal management. In order to fabricate a VLED device, the epitaxial layers must be removed from the sapphire substrate. There are essentially three different ways this has been accomplished: laser liftoff (LLO), chemical liftoff (CLO), and backside grinding of the sapphire. In laser liftoff,3–5) an ultraviolet excimer laser is projected through the backside of the sapphire wafer and results in decomposition of the GaN at the GaN/Al2 O3 interface. In chemical liftoff, a specific layer within the epitaxial stack is preferentially etched and subsequently releases the overlying device layers. This has been demonstrated using selective etching of layers composed of CrN,6) GaN:Si,7) ZnO8) or using photoelectrochemical (PEC) etching techniques.9) The drawback to laser liftoff is that damage to the GaN layers can occur during irradiation resulting in reduced device performance and yield.10) The chemical liftoff processes that incorporate new materials into the epitaxial stack are faced with the challenge of maintaining high-quality GaN growth, and large-area CLO processes are often difficult to implement in practice. We have reported previously on a layer transfer technique referred to as controlled spalling.11,12) This approach uses a stressed layer deposited on the surface of a substrate to induce lateral fracture within the underlying substrate and ‘‘peel off’’ a controlled thickness from the surface. We have applied controlled spalling to the fabrication of flexible, high-efficiency GaAs-based photovoltaics13) as well as highperformance nanoscale Si circuits14) and in all cases the defectivity of the material is unchanged by this process. In this work, we use controlled spalling to remove multiquantum well (MQW)-based InGaN/GaN epitaxial LED layers grown on PSS substrates and demonstrate successful
D
optical emission from the resulting vertical spalled LED (SLED) layers. The physical origin of spalling mode fracture comes from the combination of opening stress (mode I) and shear stress (mode II) due to the presence of a surface stressor layer acting on a subsurface crack tip.15,16) For a tensile surface stressor layer, a crack will propagate downward into the substrate to a depth where the shear component is zero. If the crack continues below this point, it will be deflected back towards the surface by the change in sign of the shear stress component. Therefore, the fracture depth is stable within the substrate as the crack tip trajectory oscillates about the zero shear position due to the corrective action of the shear field. For fracture to be energetically possible within the substrate, the magnitude of the opening mode stress field (given by the stress intensity factor KI ) must be equal to, or greater than, the material fracture toughness KIC at the zero shear position (KII ¼ 0). For a particular choice of stressor material (e.g., Ni) there will be a combination of stress and thickness above which fracture is possible for a given substrate type (using KIC from literature17)). In order to apply the controlled spalling process to the epitaxial GaN material system using Ni as the stressor layer, the required loading conditions (stress and thickness) were determined experimentally. If the loading conditions are too high, spalling can occur spontaneously during Ni deposition and generally leads to unusable layers. Therefore, it is important to determine the combinations of stress and thickness that energetically permit fracture, but do not lead to spontaneous spalling. PSS substrates (c-plane), 50 mm in diameter, were used as growth templates for Inx Gað1xÞ N/GaN MQW LED structures using a commercial metal–organic chemical vapor deposition (MOCVD) reactor. The epitaxial structure consisted of a thin low-temperature GaN buffer layer, followed by a 2-m-thick undoped GaN layer (u-GaN), a 3-m-thick Si doped n-GaN layer, InGaN/GaN MQW layers, and Mgdoped p-GaN capping layers. After growth, the wafers were annealed in a N2 ambient to activate the p-GaN layer. Figure 1 shows a schematic of the controlled spalling process used in this work. A Ni stressor layer was deposited onto the surface of the wafers by first depositing a thin (75 nm) adhesion layer of Ti followed by 500 nm of Ni by DC magnetron sputtering. Electroplating using a NiCl2 based solution at room temperature resulted in approximately 400 MPa tensile stress as measured by wafer bowing. We then determined that approximately 25 m of Ni deposited onto the sputtered Ni was required to permit controlled spalling of the GaN. After deposition of the thick Ni layer,
112301-1
# 2013 The Japan Society of Applied Physics
Appl. Phys. Express 6 (2013) 112301
S. W. Bedell et al.
Tape
(a)
Ni (tensile) GaN Epitaxy Sapphire
Fig. 2. Image of a spalled 50-mm-diameter InGaN/GaN MQW LED structure (SLED) mounted onto a stainless steel frame using the handling tape.
(b)
Controlled Spalling 200 Schematic illustration of the controlled spalling process used to remove InGaN/GaN MQW LED layers from a sapphire substrate. The GaN/Al2 O3 substrate was prepared for spalling by (a) first depositing a thin Ti (75 nm)/Ni (500 nm) seed layer by sputtering, followed by electrodeposition of a 25-m-thick Ni layer (400 MPa tensile), then application of a tape layer. Removal of the epitaxial layers from the sapphire substrate was performed by (b) peeling the tape layer away from the surface. Fig. 1.
the wafers were rinsed and dried and held flat using a custom-made vacuum chuck table with an integrated tape dispenser. A layer of polyimide tape (25 m) was rollapplied to the surface of the Ni, and fracture occurred upon gently pulling the tape away from the wafer surface, as illustrated in Fig. 1(b). To prevent excessive bending of the thin film after spalling, the outer portions of the tape are used to secure the sample to a stainless steel frame. The entire process, from metal deposition to layer removal, was performed at room temperature. Figure 2 shows an entire 50-mm-diameter InGaN/GaN epitaxial structure removed from the sapphire wafer by controlled spalling and mounted in a handling frame. The surface artifacts observed on the sample shown in Fig. 2 are due to roughness caused by fracture depth oscillations around the equilibrium spalling depth and are not cracks in the film. Surface profilometry was performed and the results are shown in Fig. 3. Although the resulting surface roughness after spalling varies with length scale, the maximum amplitude occurs with a spatial period of about 700 m and an amplitude of about 70 nm. Fracture propagation speed directly affects this spatial period, with faster fracture resulting in a longer oscillation period. In this work, fracture propagation across the 50 mm wafer surface took on the order of one second leading to a fracture velocity of about 50 mm/s. In the case of LED structure spalling, this roughness may offer a natural means of surface texturing for improved light extraction. Figure 4 shows a cross-sectional scanning electron microscopy (XSEM) image of the InGaN/GaN LED structure before [Fig. 4(a)] and after [Fig. 4(b)] controlled spalling. Coloration has been added to the figure to indicate the various regions of the epitaxial structure. By comparison of the two
150
Height (nm)
100 50 0 -50 -100 -150 -200 0
500
1000 1500 2000 Distance (μm)
2500
3000
Fig. 3. Surface profilometry data showing the characteristic roughness associated with the as-fractured surface after spalling. The largest amplitude is seen at a spatial period of about 700 m with a magnitude of about 70 nm.
images it is evident that fracture propagated approximately 3 m below the surface within the n-GaN region. This has the twofold benefit of having immediate access to the n-type contact area without etching, and the possibility of reusing not only the original sapphire wafer, but also the u-GaN buffer region as well. Spalling depth is generally controlled by the Ni thickness. We have observed a spalling depth change of approximately 500 nm per micron of Ni thickness change for the GaN on PSS system. Also, due to the extremely high fracture toughness of sapphire, GaN epitaxy layers could be removed entirely from non-PSS (planar) sapphire substrates. Figure 5 shows cross-sectional transmission electron microscopy (XTEM) images of the SLED structure. The low magnification image [Fig. 5(a)] shows the full thickness of the SLED structure indicating a total layer thickness of about 3.1 m and no spalling-related defectivity was observed in the sample. Figures 5(b) and 5(c) show high-resolution imaging of the InGaN/GaN MQW structure and the 10-period active layer of the device, respectively. Figure 6 shows the electroluminescence spectrum taken from direct probing (no metallization) of the exposed n-GaN
112301-2
# 2013 The Japan Society of Applied Physics
Appl. Phys. Express 6 (2013) 112301
(a)
S. W. Bedell et al. 1.2
(b)
Bulk
1
Normalized Intensity
p-GaN MQW
~ 3μm
n-GaN
n-GaN
u-GaN
u-GaN
SLED
0.8 0.6 0.4
λp=529.2 nm (SLED) λp=533.9 nm (Bulk)
0.2 0 475
Sapphire
(b)
Spalled n-GaN
10 nm
(c)
Ti 0.5 μm
Ni
575 Wavelength (nm)
625
675
Fig. 6. Electroluminescence curves for the SLED structure and the conventional (bulk) structure. The inset shows the green optical emission by directly probing the exposed n-GaN surface of the SLED device. The inhomogeneous emission is likely caused by localized conduction regions due to the use of a non-ohmic Ti contact to p-GaN.
Fig. 4. Cross-sectional SEM image showing the InGaN/GaN device structure (a) before and (b) after controlled spalling. The image has been colorized to indicate the various regions of the structure. Approximately 3 m of the surface was removed, with fracture occurring near the center of the n-GaN layer.
(a)
525
100 nm
Fig. 5. Cross-sectional TEM image showing (a) entire spalled layer (b) high-resolution image of the MQW structure, and (c) image of the 10-period active device region. The few defects observed in the sample were related to GaN growth suggesting that the material quality is preserved during controlled spalling.
surface of the vertical SLED (inset), as well as a conventional device formed on a small non-spalled portion from the edge of the same wafer (Bulk). The optical emission from the SLED films was inhomogeneous, as seen in the image (Fig. 6 inset). This is because the Ti adhesion layer formed a rectifying contact with the p-GaN surface and con-
duction during forward bias occurred primarily at localized breakdown regions. The peak emission p was measured to be 529.2 nm from the SLED device and 533.9 for the bulk device (with annealed ohmic NiAu contacts). This corresponds to an energy shift Eg of 21 meV. Although changes in the stress state of the SLED device occur due to the presence of the tensile Ni and are known to result in Eg shift,18) the energy shift in green InGaN/GaN LED structures such as these is also very sensitive to biasing conditions.19,20) Owing to the inhomogeneous conduction observed in the SLED devices, the biasing conditions between the two devices cannot be properly compared and the origin of the observed Eg is likely a combination of both effects. In addition to the direct probing of the as-spalled surface described above, a portion of the SLED layer was bonded to a Si wafer using In. After bonding the as-spalled n-GaN surface to Si, the polyimide tape was removed and the Ni stressor and Ti adhesion layers were chemically etched revealing the original p-GaN surface. Ni/Au contacts were evaporated; however, no annealing could be performed due to the low melting point of the In bonding material. Figure 7 shows the resulting current density versus bias of the bulk InGaN/GaN device with annealed Ni/Au contacts compared to the In-bonded SLED device with as-deposited Ni/Au contacts. The bonded SLED device showed an expected increase in parasitic resistance as well as a 200 mV increase in turn-on voltage due to the non-annealed contact layers compared to the bulk device. One of the most important differences between controlled spalling and other liftoff technologies is the generality associated with spalling. In other words, because there is no need for specialized layers, transfer can be performed on any optimized epitaxial structure. Therefore, layer transfer of a variety of GaN-based structures (LED, laser, HEMT, etc.) grown on either sapphire, SiC, free-standing GaN, or Si is straightforward. The other attractive aspect of layer transfer by spalling is that it can be accomplished using conventional laboratory equipment; no lasers or other specialized tools are required.
112301-3
# 2013 The Japan Society of Applied Physics
Appl. Phys. Express 6 (2013) 112301
S. W. Bedell et al.
0.4
Current Density (A/cm2)
Bulk 0.3
SLED (bonded) Ni/Au (no anneal)
Ni/Au (annealed) In
Acknowledgments The authors would like to thank Drs. G. Shahidi and T. C. Chen for their encouragement and support of this work.
In
SLED
GaN
0.2
due to a combination of mechanical stress and dissimilar biasing conditions.
1) S. Nakamura, M. Senoh, and T. Mukai: Appl. Phys. Lett. 62 (1993) 2390. 2) M. R. Krames, O. B. Shchekin, R. Mueller-Mach, G. O. Mueller, L. Zhou,
Si
PSS
G. Harbers, and M. G. Craford: J. Disp. Technol. 3 (2007) 160.
3) W. S. Wong, T. Sands, and N. W. Cheung: Appl. Phys. Lett. 72 (1998)
599.
0.1
Area = 1 mm2
4) T. Fujii, Y. Gao, R. Sharma, E. L. Hu, S. P. DenBaars, and S. Nakamura:
Appl. Phys. Lett. 84 (2004) 855. 5) C. A. Tran, C.-F. Chu, C.-C. Cheng, W.-H. Liu, J.-Y. Chu, H.-C. Cheng,
F.-H. Fan, J.-K. Yen, and T. Doan: J. Cryst. Growth 298 (2007) 722.
0 0
1
2
3
4
5
Bias (V)
6) K. Fujii, S. Lee, J.-S. Ha, H.-J. Lee, H.-J. Lee, S.-H. Lee, T. Kato, M.-W.
Cho, and T. Yao: Appl. Phys. Lett. 94 (2009) 242108. 7) J. Park, K. M. Song, S.-R. Jeon, J. H. Baek, and S.-W. Ryu: Appl. Phys.
Lett. 94 (2009) 221907.
Current density versus bias for the bulk InGaN/GaN LED device compared to a SLED device which has been bonded to a Si wafer. The asspalled n-GaN surface was bonded using In, and the tape and Ni layers removed revealing the original p-GaN surface. Ni/Au contacts were deposited but not annealed due to the low melting point of the In bond. The data show an expected increase in parasitic resistance and turn-on voltage due to the non-annealed contacts. Fig. 7.
8) D. J. Rogers, F. H. Teherani, A. Ougazzaden, S. Gautier, L. Divay, A.
20)
Lusson, O. Durand, F. Wyczisk, G. Garry, T. Monteiro, M. R. Correira, M. Peres, A. Neves, D. McGrouther, J. N. Chapman, and M. Razeghi: Appl. Phys. Lett. 91 (2007) 071120. C. Hsieh, H.-S. Chen, C.-H. Liao, C.-Y. Chen, C.-H. Lin, C.-H. Lin, S.-Y. Ting, Y.-F. Yao, H.-T. Chen, Y.-W. Kiang, and C.-C. Yang: IEEE Photonics Technol. Lett. 24 (2012) 1775. Y. S. Wu, J.-H. Cheng, W. C. Peng, and H. Ouyang: Appl. Phys. Lett. 90 (2007) 251110. S. W. Bedell, D. Shahrjerdi, B. Hekmatshoar, K. Fogel, P. Lauro, J. Ott, N. Sosa, and D. Sadana: IEEE J. Photovoltaics 2 (2012) 141. S. W. Bedell, K. Fogel, P. Lauro, D. Shahrjerdi, J. A. Ott, and D. Sadana: J. Phys. D 46 (2013) 152002. D. Shahrjerdi, S. W. Bedell, C. Bayram, C. C. Lubguban, K. Fogel, P. Lauro, J. A. Ott, M. Hopstaken, M. Gayness, and D. Sadana: Adv. Energy Mater. 3 (2013) 566. D. Shahrjerdi and S. W. Bedell: Nano Lett. 13 (2013) 315. M. D. Thouless, A. G. Evans, M. F. Ashby, and J. W. Hutchinson: Acta Metall. 35 (1987) 1333. Z. Suo and J. W. Hutchinson: Int. J. Solids Struct. 25 (1989) 1337. M. D. Drory, J. W. Ager, T. Suski, I. Grzegory, and S. Porowski: Appl. Phys. Lett. 69 (1996) 4044. C. Kisielowski, J. Kru¨ger, S. Ruvimov, T. Suski, J. W. Ager III, E. Jones, Z. Liliental-Weber, M. Rubin, E. R. Weber, M. D. Bremser, and R. F. Davis: Phys. Rev. B 54 (1996) 17745. C. Bayram, J. L. Pau, R. McClintock, and M. Razeghi: Appl. Phys. B 95 (2009) 307. S. Nakamura, T. Mukai, and M. Senoh: J. Appl. Phys. 76 (1994) 8189.
112301-4
# 2013 The Japan Society of Applied Physics
9)
10) 11)
In summary, a previously described layer transfer method called controlled spalling was used to separate a conventional InGaN/GaN MQW LED structure from a 50-mmdiameter sapphire wafer enabling the formation of a vertical spalled LED (SLED). A 25-m-thick Ni layer possessing 400 MPa tensile stress was electrodeposited on the surface of the wafer, followed by the application of a polyimide tape layer. By securing the wafer using a vacuum chuck, and pulling on the tape layer, a continuous 3-m-thick layer of the LED epitaxy was removed. XTEM imaging indicated that spalling preserved the quality of the epitaxial layers, and electroluminescence verifies the operation of the SLED. An electroluminescence energy shift Eg of 21 meV was observed as compared to a conventional device and is likely
12) 13)
14) 15) 16) 17) 18)
19)
