- Email: [email protected]
Study on sapphire removal for thin-film LEDs fabrication using CMP and dry etching
Applied Surface Science 255 (2009) 9469–9473
Contents lists available at ScienceDirect
Applied Surface Science journal homepage: www.elsevier.com/locate/apsusc
Study on sapphire removal for thin-film LEDs fabrication using CMP and dry etching Shengjun Zhou a, Sheng Liu a,b,c,* a
Research Institute of Micro/Nano Science and Technology, Shanghai Jiao Tong University, Shanghai, 200240, PR China Wuhan National Laboratory for Optoelectronics, Huazhong University of Science & Technology, Wuhan, 430074, PR China c Institute of Microsystems, Huazhong University of Science & Technology, Wuhan, 430074, PR China b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 15 May 2009 Received in revised form 15 July 2009 Accepted 17 July 2009 Available online 25 July 2009
Mechanical grinding, chemical mechanical polishing (CMP) and dry etching process are integrated to remove sapphire substrate for fabricating thin-film light-emitting diodes. The thinning of sapphire substrate is done by fast mechanical grinding followed by CMP. The CMP can remove or reduce most of the scratches produced by mechanical grinding, recovering both the mechanical strength and wafer warpage to their original status and resulting in a smoother surface. The surface morphology and surface roughness on grinded and polished sapphire substrate are measured by using atomic force microscopy (AFM). The etch rates of sapphire by BCl3-based dry etching are reported. Pattern transfer to the physical and chemical stability of sapphire is made possible by inductively coupled plasma (ICP) etch system that generates high density plasma. The patterning of several microns period in sapphire wafer by using a combination of BCl3/Ar plasma chemistry and SiO2 mask is presented. The anisotropic etch profile formed on sapphire wafer is obtained from scanning electron microscopy (SEM) images. ß 2009 Elsevier B.V. All rights reserved.
Keywords: Sapphire substrate Mechanical grinding CMP Dry etching
1. Introduction Recently, the Group-III nitride based semiconductors have emerged as the leading materials for realization of high performance light emitters from ultraviolet (UV) to the blue and green spectral regions [1–4]. Generally, GaN-based light-emitting diodes (LEDs) are grown on the sapphire (Al2O3) substrate. Although the deposition of low temperature nucleation layer such as GaN and AlN on sapphire substrate can improve the crystal quality of the subsequent GaN epitaxial layers, the threading dislocation density between 108 and 1010 cm 2 will still remain due to the large mismatch in lattice constants and thermal expansion coefficients between the nitride epi-layer and sapphire substrate. Thus, we need to reduce the threading dislocation density in order to improve the performance of GaN-based LED. GaN-based LED grown on the patterned sapphire substrate can improve the internal quantum efficiency by the reduction of threading dislocation density [5]. It is well known that sapphire is chemically inert and insoluble in most substances. Therefore, it is extremely difficult to etch or pattern the sapphire substrate with
* Corresponding author at: Research Institute of Micro/Nano Science and Technology, Shanghai Jiao Tong University, No. 800, Dongchuan Road, Shanghai, 200240, PR China. Tel.: +86 27 87542604; fax: +86 27 87557074. E-mail address: [email protected] (S. Liu). 0169-4332/$ – see front matter ß 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.apsusc.2009.07.063
wet chemical etching at room temperature. Compared with wet etching, dry etching can provide an anisotropic profile and a fast etching rate. Much of the previous work has investigated several etch techniques such as chemical wet etching after ion implantation [6], reactive ion etching [7], laser-induced etching [8,9], and inductively coupled plasma etching (ICP) [10,11]. BCl3-based gas chemistry is widely used to etch sapphire because B scavenges oxygen, and it forms BOClx volatile etch products [12]. BCl3/Cl2 and BCl3/Cl2/Ar gas combinations have been reported to have high etching rates for sapphire etching but poor etch selectivities over a photoresist. Therefore, to use hardmasks such as SiO2 instead of photoresist as etch mask and to achieve more anisotropic etch profiles were required. Thin-film LEDs with vertical structure electrodes are widely employed for fabricating high power LED [13]. To fabricate thinfilm LEDs, sapphire substrate must be removed due to its nonelectrical conductivity and low thermal conductivity. Generally, sapphire substrate is removed by laser lift-off (LLO) process [14,15]. However, the LLO method has some drawbacks. During the LLO process, the temperature should be above 900 8C in the GaN/ sapphire interface due to the absorbed photon energy, leading to the destruction of the GaN. Another drawback of this approach lies in that the bonding layer can be affected, because the bonding layer is only several microns away from GaN/sapphire interface [16]. In this paper, we combine mechanical grinding with chemical mechanical polishing (CMP) and dry etching to remove sapphire
9470
S. Zhou, S. Liu / Applied Surface Science 255 (2009) 9469–9473
substrate. The BCl3/Ar gas combinations are used to etch and pattern (0 0 0 1) oriented sapphire substrate. The etch characteristics of sapphire wafers are investigated.
The etch rate of sapphire as a function of BCl3 concentrations in BCl3/Ar gas chemistry was shown in Fig. 1. During the etching process, the ICP power and RF power were kept at 1500 W and 150 W, respectively. The total gas flow was 60 sccm. The DC bias voltage decreased from 325 V for 90% BCl3/10%Ar to 259 V for 10% BCl3/90%Ar. As shown in Fig. 1, the increase of BCl3 in BCl3/Ar generally increased the sapphire etch rate. The increase of BCl3 in BCl3/Ar rapidly increased sapphire etch rates until 30% BCl3 was reached due in part to higher concentration of reactive radicals such as BCl, Cl which increased the chemical etching of sapphire. However, sapphire etching rates did not increase much faster above 30% BCl3
in BCl3/Ar according to the slope of line segment shown in Fig. 1. During the etching process, when BCl3 percent increased from 10% to 30%, 30% to 50%, 50% to 70%, and 70% to 90%, the increment of DC bias voltage was 25 V, 12 V, 15 V, and 14 V, respectively. This indicated that the increment of DC bias voltage in high BCl3 percent region was much smaller than that in low BCl3 percent region. Therefore, the increment of sputtering yield in high BCl3 percent region could be much smaller than that in low BCl3 percent region due to the smaller increment of DC bias voltage in high BCl3 percent region. Meanwhile, the increment of chemical etching yield in high BCl3 percent region could be much smaller than that in low BCl3 percent region due to the smaller increment of reactive radicals density in high BCl3 percent region [17]. Besides, the Langmuir probe diagnostics of the BCl3/Ar gas plasma indicated that the ion density in BCl3/Ar gas plasma increased with increasing Ar fraction [17–19]. In other words, the ion density in BCl3/Ar gas plasma decreased with increasing BCl3 fraction, which would cause the decrease of ion-assisted etching rate in high BCl3 percent region. Accordingly, the reason that sapphire etching rates did not increase much faster above 30% BCl3 in BCl3/Ar was related to the smaller increment of DC bias voltage and reactive radicals density in high BCl3 percent region, and to the decrease of ion density in high BCl3 percent region. Fig. 2 showed that the sapphire etch rates as a function of BCl3 gas flow in pure BCl3 gas chemistry while the ICP power, RF power, operating pressure were 600 W, 150 W, and 5 mTorr, respectively. The etch rates of sapphire increased with the increasing BCl3 gas flow up to 50 sccm and further increase of BCl3 decreased the etch rates slightly at specific ICP/RF power. It is noted that the DC bias voltage showed only small change, from 430 V at 10 sccm to 432 V at 70 sccm, and the small change in the DC bias voltage was not expected to be a critical factor in the change in etch rate. The exact reasons for the slight decrease in sapphire etching rate when the pure BCl3 gas flow up to 70 sccm were not clear at this moment. It might be related to either saturation of reactive species at the sapphire surface or sputter desorption of reactive species from the sapphire surface before the occurrence of chemical reactions [20]. The effect of operating pressure on the sapphire etch rates was investigated while maintaining the BCl3/Ar mixture ratio at 90%/ 10%, ICP power and RF power at 2000 W and 100 W, respectively. As a function of operating pressure, plasma condition including the mean free path can change leading to changes in both ion energy and plasma density. Sapphire etch rates were plotted as a function of operating pressure in Fig. 3. As shown in Fig. 3, the increase of operating pressure from 3.6 mTorr to 9 mTorr decreased the sapphire etch rates from 1831 A˚/min to 1549 A˚/min. In general, the
Fig. 1. Etch rate of sapphire as a function of %BCl3 in BCl3/Ar gas chemistry.
Fig. 2. Etch rate of sapphire as a function of BCl3 gas flow in pure BCl3 gas chemistry.
2. Experiments For etch rate experiments, we used (0 0 0 1) oriented sapphire as etch samples. A layer of SiO2 mask was deposited by plasmaenhanced chemical vapor deposition (PECVD) on the top of sapphire. The etch depth of sapphire was measured by profilometry. The etch profile of sapphire was obtained from scanning electron microscopy (SEM) images. All etching was carried out in an ICP etcher (Oxford Plasma Lab System 100). The plasma was generated by a radio frequency (13.56 MHz) glow discharge. To conduct the experiment for separating sapphire substrate from GaN epitaxial layer, LED wafer was used as the experimental sample. The LED structure consists of a 375-mm-thick sapphire substrate, a 2-mm-thick unintentionally doped GaN, a 2-mm-thick n-GaN layer, an active region with ten periods of InGaN/GaN muitiple quantum well, and a 0.2-mm-thick p-GaN. GaN epitaxial layer was grown on the sapphire substrate by metal organic chemical vapor deposition (MOCVD). To detach sapphire substrate from GaN epitaxial layers, the LED wafer was first flip bonded to silicon substrate by Au–Sn eutectic bonds. Afterwards sapphire substrate was thinned from 375 mm to around 100 mm by mechanical grinding including coarse and fine grinding, then sapphire substrate was thinned from 100 mm to about 8 mm by CMP, finally dry etching was used to remove the remaining sapphire substrate. The surface morphology and surface roughness were measured on sapphire substrate using atomic force microscopy (AFM). 3. Results and discussion
S. Zhou, S. Liu / Applied Surface Science 255 (2009) 9469–9473
Fig. 3. Etch rate of sapphire as a function of operating pressure.
etching rate of sapphire is related to ion energy and density of radicals and ions at specific ICP/rf power. During the etching process, the DC bias voltage increased from 198 V to 251 V with the increase of operating pressure from 3.6 mTorr to 9 mTorr, which caused higher sputtering yield. The ion induced etching rate would increase provided that ion density increased at high operating pressure. Thus, both increase of ion energy and increase of ion density would lead to the increase of sapphire etch rates,
9471
which was contrary to experimental results. Accordingly, the decrease of sapphire etch rates with increasing operating pressure could be related to the decreased ion density at higher operating pressure because the ion induced etching rate would decrease provided that ion density decreased at high operating pressure. The Langmuir probe and quadrupole mass spectroscopy (QMS) diagnostics of the BCl3/Ar inductively coupled plasma showed that the increasing operating pressure caused the decrease of ion density [21,22]. It is well known that patterned sapphire substrate technology can improve the light extraction efficiency compared with the conventional sapphire substrate LEDs due to the angled facets that redirected light into the escape cone [23]. Therefore, the periodic pattern formed on sapphire substrate is often required to increase light extraction efficiency. Fig. 4 showed the sapphire sample etch profiles for (a) and (b) 80% BCl3/20%Ar, 5 mTorr, 2000 W/100 W, for (c) and (d) 90% BCl3/ 10%Ar, 5 mTorr, 1500 W/200 W. The sapphire samples were etched for 15 min at a flow rate of 50 sccm. As shown in Fig. 4, it can be seen that the sapphire sample was etched to a depth of about 2.69 mm. The periodic pattern was transferred into sapphire substrate by using SiO2 mask, which was opened by RIE with fluorine-based gas chemistry. The more anisotropic etch profile was obtained in Fig. 4(c) and (d). In other words, we could achieve profiles that were almost vertical by using these etching conditions. LEDs have evolved from simple top-emitting structure to flipchip structure, and now to thin-film structure [24]. In order to fabricate thin-film LEDs, the sapphire substrate must be removed. CMP can produce high material removal rates, little or no surface
Fig. 4. SEM cross-sectional micrograph of etched sapphire sample with: (a and b) 80% BCl3/20%Ar, 5 mTorr, 2000 W/100 W and (c and d) 90% BCl3/10%Ar, 5 mTorr, 1500 W/ 200 W.
9472
S. Zhou, S. Liu / Applied Surface Science 255 (2009) 9469–9473
Fig. 5. AFM scans of sapphire substrate surfaces. (a) Grinded sapphire substrate by mechanical grinding. (b) Polished sapphire substrate by CMP.
Fig. 6. Surface microscopic images of sapphire substrates. (a) Grinded sapphire substrate. (b) Polished sapphire substrate.
and subsurface damage, CMP therefore has been widely applied to various materials in last two decades [25]. The thinning of sapphire substrate was done by fast mechanical grinding followed by CMP. The surface morphology and roughness of grinded and polished sapphire substrates were measured using AFM. The results were shown in Fig. 5, (a) for the grinded sample, and (b) for the polished sample by CMP. As shown in Fig. 5, the root-mean-square (RMS) roughness of the grinded sapphire substrate was 20.8 nm over a 20 mm 20 mm area, while after the sapphire substrate was polished by CMP, the roughness in that region was 0.738 nm. Fig. 6 demonstrated the results of surface microscopy of grinded and polished sapphire substrate. As shown in Fig. 6, the sapphire surface presented some scratches, which were produced by mechanical grinding. After being polished by CMP, the scratches could be reduced or eliminated. It was worth noting that grinding generated wafer warpage because of damaged layer created during the grinding process. The CMP removed or reduced most of the damage produced by coarse and fine grinding, resuming both the mechanical strength and wafer warpage to their original status and resulting in a smoother surface. After the sapphire substrate was thinned to about 8 mm by grinding and CMP, we used the 2000 W/150 W, 5 mTorr, 90% BCl3/ 10%Ar condition at a total flow rate of 60 sccm to etch the remaining sapphire substrate. The etch rate was up to 1809 A˚/min. As the total etch time can take up about 45 min, the sapphire wafer
temperature should be controlled and maintained by backside helium cooling during the etch process. 4. Conclusions In this study, the BCl3/Ar plasma was used to investigate the etch characteristics of sapphire substrate. The increase of BCl3 in BCl3/Ar generally increased the etch rates for sapphire substrate. The increase of BCl3 gas flow in pure BCl3 gas chemistry rapidly increased sapphire etch rates until 50 sccm gas flow was reached and further increased BCl3 gas flow slowly decreased the etch rate slightly. The etch rates increased with the decrease of operating pressure. The anisotropic etch profile and periodic pattern on sapphire substrate were obtained using the 1500 W/200 W, 5 mTorr, 90% BCl3/10%Ar condition at a total flow rate of 50 sccm. By integrating mechanical grinding with CMP and dry etching process, the sapphire substrate can be detached from GaN-based LED epi-layer. Acknowledgments This work was supported by a key project of the National Natural Science Foundation of China under Project 50835005. The authors would like to acknowledge AquaLite Optoelectronics Co., Ltd. for the support in some experimental processes.
S. Zhou, S. Liu / Applied Surface Science 255 (2009) 9469–9473
References [1] S. Nakamura, T. Mukai, M. Senoh, Appl. Phys. Lett. 64 (1994) 1687. [2] .E.F. Schubert, Light-emitting Diodes, Cambridge University Press, New York, 2003 [3] A. Bell, R. Liu, F.A. Ponce, H. Amano, I. Akasaki, D. Cherns, Appl. Phys. Lett. 82 (3) (2003) 349. [4] F. Dwikusuma, D. Saulys, T.F. Kuech, J. Electrochem. Soc. 149 (2002) G603. [5] K. Tadatomo, H. Okagawa, Y. Ohuchi, T. Tsunekawa, Y. Imada, M. Kato, T. Taguchi, Jpn. J. Appl. Phys. 40 (6B) (2001) L583. [6] X.D. Zhu, D.Z. Zhu, H.C. Pan, H.J. Xu, Z.X. Ren, J. Phys. D: Appl. Phys. 31 (1998) 1647. [7] J.W. Kim, Y.C. Kim, W.J. Lee, J. Appl. Phys. 78 (1995) 2045. [8] S.I. Dolgaev, A.A. Lyalin, A.V. Simakin, G.A. Shafeev, Appl. Surf. Sci. 96–98 (1996) 49l. [9] S. Pissadakis, R. Bo¨hme, K. Zimmer, Opt. Exp. 15 (2007) 1428. [10] Y.P. Hsu, S.J. Chang, Y.K. Su, J.K. Sheu, C.H. Kuo, C.S. Chang, S.C. Shei, Opt. Mater. 27 (2005) 1171. [11] D.P. Kim, J.W. Yeo, C.I. Kim, Thin Solid Films 459 (2004) 122. [12] C.H. Jeong, D.W. Kim, K.N. Kim, G.Y. Yeom, Jpn. J. Appl. Phys. 41 (2002) 6206. [13] K.H. Baik, S.J. Pearton, Appl. Surf. Sci. 255 (2009) 5948.
9473
[14] W.S. Wong, T. Sands, N.W. Cheung, Appl. Phys. Lett. 72 (1998) 599. [15] W.S. Wong, T. Sands, N.W. Cheung, M. Kneissl, D.P. Bour, P. Mei, L.T. Romano, N.M. Johnson, Appl. Phys. Lett. 75 (1999) 1360. [16] S.C. Hsu, C.Y. Liu, Electrochem. Solid-State Lett. 9 (2006) G171. [17] M. Kim, N.K. Min, S.J. Yun, H.W. Lee, A.M. Efremov, K.H. Kwon, ETRI J. 30 (2008) 383. [18] S.J. Yun, K.H. Kwon, Y.E. Lee, C.I. Kim, Jpn. J. Appl. Phys. 43 (2004) 2716. [19] G.H. Kim, C.I. Kim, A.M. Efremov, Vacuum 79 (2005) 231. [20] S.J. Pearton, J.C. Zolper, R.J. Shul, F. Ren, J. Appl. Phys. 86 (1999) 1. [21] S.J. Yun, A. Efremov, M. Kim, D.W. Kim, J.W. Lim, Y.H. Kim, C.H. Chung, D.J. Park, K.H. Kwon, Vacuum 82 (2008) 1198. [22] H.S. Kim, G.Y. Yeom, J.W. Lee, T.I. Kim, J. Vac. Sci. Technol. A 17 (1999) 2214. [23] Y.J. Lee, J.M. Hwang, T.C. Hsu, M.H. Hsieh, M.J. Jou, B.J. Lee, T.C. Lu, H.C. Kuo, S.C. Wang, IEEE Photonics Technol. Lett. 18 (2006) 1152. [24] O.B. Shchekin, J.E. Epler, T.A. Trottier, T. Margalith, D.A. Steigerwald, M.O. Holcomb, P.S. Martin, M.R. Krames, Appl. Phys. Lett. 89 (2006) 071109. [25] H.L. Zhu, L.A. Tessaroto, R. Sabia, V.A. Greenhut, M. Smith, D.E. Niesz, Appl. Surf. Sci. 236 (2004) 120.
Contents lists available at ScienceDirect
Applied Surface Science journal homepage: www.elsevier.com/locate/apsusc
Study on sapphire removal for thin-film LEDs fabrication using CMP and dry etching Shengjun Zhou a, Sheng Liu a,b,c,* a
Research Institute of Micro/Nano Science and Technology, Shanghai Jiao Tong University, Shanghai, 200240, PR China Wuhan National Laboratory for Optoelectronics, Huazhong University of Science & Technology, Wuhan, 430074, PR China c Institute of Microsystems, Huazhong University of Science & Technology, Wuhan, 430074, PR China b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 15 May 2009 Received in revised form 15 July 2009 Accepted 17 July 2009 Available online 25 July 2009
Mechanical grinding, chemical mechanical polishing (CMP) and dry etching process are integrated to remove sapphire substrate for fabricating thin-film light-emitting diodes. The thinning of sapphire substrate is done by fast mechanical grinding followed by CMP. The CMP can remove or reduce most of the scratches produced by mechanical grinding, recovering both the mechanical strength and wafer warpage to their original status and resulting in a smoother surface. The surface morphology and surface roughness on grinded and polished sapphire substrate are measured by using atomic force microscopy (AFM). The etch rates of sapphire by BCl3-based dry etching are reported. Pattern transfer to the physical and chemical stability of sapphire is made possible by inductively coupled plasma (ICP) etch system that generates high density plasma. The patterning of several microns period in sapphire wafer by using a combination of BCl3/Ar plasma chemistry and SiO2 mask is presented. The anisotropic etch profile formed on sapphire wafer is obtained from scanning electron microscopy (SEM) images. ß 2009 Elsevier B.V. All rights reserved.
Keywords: Sapphire substrate Mechanical grinding CMP Dry etching
1. Introduction Recently, the Group-III nitride based semiconductors have emerged as the leading materials for realization of high performance light emitters from ultraviolet (UV) to the blue and green spectral regions [1–4]. Generally, GaN-based light-emitting diodes (LEDs) are grown on the sapphire (Al2O3) substrate. Although the deposition of low temperature nucleation layer such as GaN and AlN on sapphire substrate can improve the crystal quality of the subsequent GaN epitaxial layers, the threading dislocation density between 108 and 1010 cm 2 will still remain due to the large mismatch in lattice constants and thermal expansion coefficients between the nitride epi-layer and sapphire substrate. Thus, we need to reduce the threading dislocation density in order to improve the performance of GaN-based LED. GaN-based LED grown on the patterned sapphire substrate can improve the internal quantum efficiency by the reduction of threading dislocation density [5]. It is well known that sapphire is chemically inert and insoluble in most substances. Therefore, it is extremely difficult to etch or pattern the sapphire substrate with
* Corresponding author at: Research Institute of Micro/Nano Science and Technology, Shanghai Jiao Tong University, No. 800, Dongchuan Road, Shanghai, 200240, PR China. Tel.: +86 27 87542604; fax: +86 27 87557074. E-mail address: [email protected] (S. Liu). 0169-4332/$ – see front matter ß 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.apsusc.2009.07.063
wet chemical etching at room temperature. Compared with wet etching, dry etching can provide an anisotropic profile and a fast etching rate. Much of the previous work has investigated several etch techniques such as chemical wet etching after ion implantation [6], reactive ion etching [7], laser-induced etching [8,9], and inductively coupled plasma etching (ICP) [10,11]. BCl3-based gas chemistry is widely used to etch sapphire because B scavenges oxygen, and it forms BOClx volatile etch products [12]. BCl3/Cl2 and BCl3/Cl2/Ar gas combinations have been reported to have high etching rates for sapphire etching but poor etch selectivities over a photoresist. Therefore, to use hardmasks such as SiO2 instead of photoresist as etch mask and to achieve more anisotropic etch profiles were required. Thin-film LEDs with vertical structure electrodes are widely employed for fabricating high power LED [13]. To fabricate thinfilm LEDs, sapphire substrate must be removed due to its nonelectrical conductivity and low thermal conductivity. Generally, sapphire substrate is removed by laser lift-off (LLO) process [14,15]. However, the LLO method has some drawbacks. During the LLO process, the temperature should be above 900 8C in the GaN/ sapphire interface due to the absorbed photon energy, leading to the destruction of the GaN. Another drawback of this approach lies in that the bonding layer can be affected, because the bonding layer is only several microns away from GaN/sapphire interface [16]. In this paper, we combine mechanical grinding with chemical mechanical polishing (CMP) and dry etching to remove sapphire
9470
S. Zhou, S. Liu / Applied Surface Science 255 (2009) 9469–9473
substrate. The BCl3/Ar gas combinations are used to etch and pattern (0 0 0 1) oriented sapphire substrate. The etch characteristics of sapphire wafers are investigated.
The etch rate of sapphire as a function of BCl3 concentrations in BCl3/Ar gas chemistry was shown in Fig. 1. During the etching process, the ICP power and RF power were kept at 1500 W and 150 W, respectively. The total gas flow was 60 sccm. The DC bias voltage decreased from 325 V for 90% BCl3/10%Ar to 259 V for 10% BCl3/90%Ar. As shown in Fig. 1, the increase of BCl3 in BCl3/Ar generally increased the sapphire etch rate. The increase of BCl3 in BCl3/Ar rapidly increased sapphire etch rates until 30% BCl3 was reached due in part to higher concentration of reactive radicals such as BCl, Cl which increased the chemical etching of sapphire. However, sapphire etching rates did not increase much faster above 30% BCl3
in BCl3/Ar according to the slope of line segment shown in Fig. 1. During the etching process, when BCl3 percent increased from 10% to 30%, 30% to 50%, 50% to 70%, and 70% to 90%, the increment of DC bias voltage was 25 V, 12 V, 15 V, and 14 V, respectively. This indicated that the increment of DC bias voltage in high BCl3 percent region was much smaller than that in low BCl3 percent region. Therefore, the increment of sputtering yield in high BCl3 percent region could be much smaller than that in low BCl3 percent region due to the smaller increment of DC bias voltage in high BCl3 percent region. Meanwhile, the increment of chemical etching yield in high BCl3 percent region could be much smaller than that in low BCl3 percent region due to the smaller increment of reactive radicals density in high BCl3 percent region [17]. Besides, the Langmuir probe diagnostics of the BCl3/Ar gas plasma indicated that the ion density in BCl3/Ar gas plasma increased with increasing Ar fraction [17–19]. In other words, the ion density in BCl3/Ar gas plasma decreased with increasing BCl3 fraction, which would cause the decrease of ion-assisted etching rate in high BCl3 percent region. Accordingly, the reason that sapphire etching rates did not increase much faster above 30% BCl3 in BCl3/Ar was related to the smaller increment of DC bias voltage and reactive radicals density in high BCl3 percent region, and to the decrease of ion density in high BCl3 percent region. Fig. 2 showed that the sapphire etch rates as a function of BCl3 gas flow in pure BCl3 gas chemistry while the ICP power, RF power, operating pressure were 600 W, 150 W, and 5 mTorr, respectively. The etch rates of sapphire increased with the increasing BCl3 gas flow up to 50 sccm and further increase of BCl3 decreased the etch rates slightly at specific ICP/RF power. It is noted that the DC bias voltage showed only small change, from 430 V at 10 sccm to 432 V at 70 sccm, and the small change in the DC bias voltage was not expected to be a critical factor in the change in etch rate. The exact reasons for the slight decrease in sapphire etching rate when the pure BCl3 gas flow up to 70 sccm were not clear at this moment. It might be related to either saturation of reactive species at the sapphire surface or sputter desorption of reactive species from the sapphire surface before the occurrence of chemical reactions [20]. The effect of operating pressure on the sapphire etch rates was investigated while maintaining the BCl3/Ar mixture ratio at 90%/ 10%, ICP power and RF power at 2000 W and 100 W, respectively. As a function of operating pressure, plasma condition including the mean free path can change leading to changes in both ion energy and plasma density. Sapphire etch rates were plotted as a function of operating pressure in Fig. 3. As shown in Fig. 3, the increase of operating pressure from 3.6 mTorr to 9 mTorr decreased the sapphire etch rates from 1831 A˚/min to 1549 A˚/min. In general, the
Fig. 1. Etch rate of sapphire as a function of %BCl3 in BCl3/Ar gas chemistry.
Fig. 2. Etch rate of sapphire as a function of BCl3 gas flow in pure BCl3 gas chemistry.
2. Experiments For etch rate experiments, we used (0 0 0 1) oriented sapphire as etch samples. A layer of SiO2 mask was deposited by plasmaenhanced chemical vapor deposition (PECVD) on the top of sapphire. The etch depth of sapphire was measured by profilometry. The etch profile of sapphire was obtained from scanning electron microscopy (SEM) images. All etching was carried out in an ICP etcher (Oxford Plasma Lab System 100). The plasma was generated by a radio frequency (13.56 MHz) glow discharge. To conduct the experiment for separating sapphire substrate from GaN epitaxial layer, LED wafer was used as the experimental sample. The LED structure consists of a 375-mm-thick sapphire substrate, a 2-mm-thick unintentionally doped GaN, a 2-mm-thick n-GaN layer, an active region with ten periods of InGaN/GaN muitiple quantum well, and a 0.2-mm-thick p-GaN. GaN epitaxial layer was grown on the sapphire substrate by metal organic chemical vapor deposition (MOCVD). To detach sapphire substrate from GaN epitaxial layers, the LED wafer was first flip bonded to silicon substrate by Au–Sn eutectic bonds. Afterwards sapphire substrate was thinned from 375 mm to around 100 mm by mechanical grinding including coarse and fine grinding, then sapphire substrate was thinned from 100 mm to about 8 mm by CMP, finally dry etching was used to remove the remaining sapphire substrate. The surface morphology and surface roughness were measured on sapphire substrate using atomic force microscopy (AFM). 3. Results and discussion
S. Zhou, S. Liu / Applied Surface Science 255 (2009) 9469–9473
Fig. 3. Etch rate of sapphire as a function of operating pressure.
etching rate of sapphire is related to ion energy and density of radicals and ions at specific ICP/rf power. During the etching process, the DC bias voltage increased from 198 V to 251 V with the increase of operating pressure from 3.6 mTorr to 9 mTorr, which caused higher sputtering yield. The ion induced etching rate would increase provided that ion density increased at high operating pressure. Thus, both increase of ion energy and increase of ion density would lead to the increase of sapphire etch rates,
9471
which was contrary to experimental results. Accordingly, the decrease of sapphire etch rates with increasing operating pressure could be related to the decreased ion density at higher operating pressure because the ion induced etching rate would decrease provided that ion density decreased at high operating pressure. The Langmuir probe and quadrupole mass spectroscopy (QMS) diagnostics of the BCl3/Ar inductively coupled plasma showed that the increasing operating pressure caused the decrease of ion density [21,22]. It is well known that patterned sapphire substrate technology can improve the light extraction efficiency compared with the conventional sapphire substrate LEDs due to the angled facets that redirected light into the escape cone [23]. Therefore, the periodic pattern formed on sapphire substrate is often required to increase light extraction efficiency. Fig. 4 showed the sapphire sample etch profiles for (a) and (b) 80% BCl3/20%Ar, 5 mTorr, 2000 W/100 W, for (c) and (d) 90% BCl3/ 10%Ar, 5 mTorr, 1500 W/200 W. The sapphire samples were etched for 15 min at a flow rate of 50 sccm. As shown in Fig. 4, it can be seen that the sapphire sample was etched to a depth of about 2.69 mm. The periodic pattern was transferred into sapphire substrate by using SiO2 mask, which was opened by RIE with fluorine-based gas chemistry. The more anisotropic etch profile was obtained in Fig. 4(c) and (d). In other words, we could achieve profiles that were almost vertical by using these etching conditions. LEDs have evolved from simple top-emitting structure to flipchip structure, and now to thin-film structure [24]. In order to fabricate thin-film LEDs, the sapphire substrate must be removed. CMP can produce high material removal rates, little or no surface
Fig. 4. SEM cross-sectional micrograph of etched sapphire sample with: (a and b) 80% BCl3/20%Ar, 5 mTorr, 2000 W/100 W and (c and d) 90% BCl3/10%Ar, 5 mTorr, 1500 W/ 200 W.
9472
S. Zhou, S. Liu / Applied Surface Science 255 (2009) 9469–9473
Fig. 5. AFM scans of sapphire substrate surfaces. (a) Grinded sapphire substrate by mechanical grinding. (b) Polished sapphire substrate by CMP.
Fig. 6. Surface microscopic images of sapphire substrates. (a) Grinded sapphire substrate. (b) Polished sapphire substrate.
and subsurface damage, CMP therefore has been widely applied to various materials in last two decades [25]. The thinning of sapphire substrate was done by fast mechanical grinding followed by CMP. The surface morphology and roughness of grinded and polished sapphire substrates were measured using AFM. The results were shown in Fig. 5, (a) for the grinded sample, and (b) for the polished sample by CMP. As shown in Fig. 5, the root-mean-square (RMS) roughness of the grinded sapphire substrate was 20.8 nm over a 20 mm 20 mm area, while after the sapphire substrate was polished by CMP, the roughness in that region was 0.738 nm. Fig. 6 demonstrated the results of surface microscopy of grinded and polished sapphire substrate. As shown in Fig. 6, the sapphire surface presented some scratches, which were produced by mechanical grinding. After being polished by CMP, the scratches could be reduced or eliminated. It was worth noting that grinding generated wafer warpage because of damaged layer created during the grinding process. The CMP removed or reduced most of the damage produced by coarse and fine grinding, resuming both the mechanical strength and wafer warpage to their original status and resulting in a smoother surface. After the sapphire substrate was thinned to about 8 mm by grinding and CMP, we used the 2000 W/150 W, 5 mTorr, 90% BCl3/ 10%Ar condition at a total flow rate of 60 sccm to etch the remaining sapphire substrate. The etch rate was up to 1809 A˚/min. As the total etch time can take up about 45 min, the sapphire wafer
temperature should be controlled and maintained by backside helium cooling during the etch process. 4. Conclusions In this study, the BCl3/Ar plasma was used to investigate the etch characteristics of sapphire substrate. The increase of BCl3 in BCl3/Ar generally increased the etch rates for sapphire substrate. The increase of BCl3 gas flow in pure BCl3 gas chemistry rapidly increased sapphire etch rates until 50 sccm gas flow was reached and further increased BCl3 gas flow slowly decreased the etch rate slightly. The etch rates increased with the decrease of operating pressure. The anisotropic etch profile and periodic pattern on sapphire substrate were obtained using the 1500 W/200 W, 5 mTorr, 90% BCl3/10%Ar condition at a total flow rate of 50 sccm. By integrating mechanical grinding with CMP and dry etching process, the sapphire substrate can be detached from GaN-based LED epi-layer. Acknowledgments This work was supported by a key project of the National Natural Science Foundation of China under Project 50835005. The authors would like to acknowledge AquaLite Optoelectronics Co., Ltd. for the support in some experimental processes.
S. Zhou, S. Liu / Applied Surface Science 255 (2009) 9469–9473
References [1] S. Nakamura, T. Mukai, M. Senoh, Appl. Phys. Lett. 64 (1994) 1687. [2] .E.F. Schubert, Light-emitting Diodes, Cambridge University Press, New York, 2003 [3] A. Bell, R. Liu, F.A. Ponce, H. Amano, I. Akasaki, D. Cherns, Appl. Phys. Lett. 82 (3) (2003) 349. [4] F. Dwikusuma, D. Saulys, T.F. Kuech, J. Electrochem. Soc. 149 (2002) G603. [5] K. Tadatomo, H. Okagawa, Y. Ohuchi, T. Tsunekawa, Y. Imada, M. Kato, T. Taguchi, Jpn. J. Appl. Phys. 40 (6B) (2001) L583. [6] X.D. Zhu, D.Z. Zhu, H.C. Pan, H.J. Xu, Z.X. Ren, J. Phys. D: Appl. Phys. 31 (1998) 1647. [7] J.W. Kim, Y.C. Kim, W.J. Lee, J. Appl. Phys. 78 (1995) 2045. [8] S.I. Dolgaev, A.A. Lyalin, A.V. Simakin, G.A. Shafeev, Appl. Surf. Sci. 96–98 (1996) 49l. [9] S. Pissadakis, R. Bo¨hme, K. Zimmer, Opt. Exp. 15 (2007) 1428. [10] Y.P. Hsu, S.J. Chang, Y.K. Su, J.K. Sheu, C.H. Kuo, C.S. Chang, S.C. Shei, Opt. Mater. 27 (2005) 1171. [11] D.P. Kim, J.W. Yeo, C.I. Kim, Thin Solid Films 459 (2004) 122. [12] C.H. Jeong, D.W. Kim, K.N. Kim, G.Y. Yeom, Jpn. J. Appl. Phys. 41 (2002) 6206. [13] K.H. Baik, S.J. Pearton, Appl. Surf. Sci. 255 (2009) 5948.
9473
[14] W.S. Wong, T. Sands, N.W. Cheung, Appl. Phys. Lett. 72 (1998) 599. [15] W.S. Wong, T. Sands, N.W. Cheung, M. Kneissl, D.P. Bour, P. Mei, L.T. Romano, N.M. Johnson, Appl. Phys. Lett. 75 (1999) 1360. [16] S.C. Hsu, C.Y. Liu, Electrochem. Solid-State Lett. 9 (2006) G171. [17] M. Kim, N.K. Min, S.J. Yun, H.W. Lee, A.M. Efremov, K.H. Kwon, ETRI J. 30 (2008) 383. [18] S.J. Yun, K.H. Kwon, Y.E. Lee, C.I. Kim, Jpn. J. Appl. Phys. 43 (2004) 2716. [19] G.H. Kim, C.I. Kim, A.M. Efremov, Vacuum 79 (2005) 231. [20] S.J. Pearton, J.C. Zolper, R.J. Shul, F. Ren, J. Appl. Phys. 86 (1999) 1. [21] S.J. Yun, A. Efremov, M. Kim, D.W. Kim, J.W. Lim, Y.H. Kim, C.H. Chung, D.J. Park, K.H. Kwon, Vacuum 82 (2008) 1198. [22] H.S. Kim, G.Y. Yeom, J.W. Lee, T.I. Kim, J. Vac. Sci. Technol. A 17 (1999) 2214. [23] Y.J. Lee, J.M. Hwang, T.C. Hsu, M.H. Hsieh, M.J. Jou, B.J. Lee, T.C. Lu, H.C. Kuo, S.C. Wang, IEEE Photonics Technol. Lett. 18 (2006) 1152. [24] O.B. Shchekin, J.E. Epler, T.A. Trottier, T. Margalith, D.A. Steigerwald, M.O. Holcomb, P.S. Martin, M.R. Krames, Appl. Phys. Lett. 89 (2006) 071109. [25] H.L. Zhu, L.A. Tessaroto, R. Sabia, V.A. Greenhut, M. Smith, D.E. Niesz, Appl. Surf. Sci. 236 (2004) 120.