- Email: [email protected]
Influence of AlN buffer layer thickness and deposition methods on GaN epitaxial growth
Thin Solid Films 517 (2009) 5057–5060
Contents lists available at ScienceDirect
Thin Solid Films j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / t s f
Influence of AlN buffer layer thickness and deposition methods on GaN epitaxial growth J.H. Yang a, S.M. Kang b, D.V. Dinh a, D.H. Yoon a,b,⁎ a b
Sungkyunkwan Advanced Institute of Nanotechnology (SAINT), Sungkyunkwan University, Suwon 440-746, Republic of Korea School of Advanced Materials Science & Engineering, Sungkyunkwan University, Suwon 440-746, Republic of Korea
a r t i c l e
i n f o
Available online 20 March 2009 Keywords: MOCVD GaN AIN Raman spectroscopy Photo luminescence X-ray diffraction
a b s t r a c t A gallium nitride (GaN) epitaxial layer was grown by metal-organic chemical vapor deposition (MOCVD) on Si (111) substrates with aluminum nitride (AlN) buffer layers at various thicknesses. The AlN buffer layers were deposited by two methods: radio frequency (RF) magnetron sputtering and MOCVD. The effect of the AlN deposition method and layer thickness on the morphological, structural and optical properties of the GaN layers was investigated. Field emission scanning electron microscopy showed that GaN did not coalesce on the sputtered AlN buffer layer. On the other hand, it coalesced with a single domain on the MOCVD-grown AlN buffer layer. Structural and optical analyses indicated that GaN on the MOCVD-grown AlN buffer layer had fewer defects and a better aligned lattice to the a- and c-axes than GaN on the sputtered AlN buffer layer. © 2009 Elsevier B.V. All rights reserved.
1. Introduction With its direct and wide bandgap of 3.39 eV at room temperature, high thermal stability, high breakdown field voltage and high saturation drift velocity, gallium nitride (GaN) is considered a promising material for optoelectronic applications in the blue and UV wavelengths, as well as in high-power and high-temperature electronics [1]. Commercially available devices, such as light emitting diodes, are usually grown by metal-organic chemical vapor deposition (MOCVD) on sapphire or SiC substrates due to the lack of large and inexpensive GaN substrates. Compared to these substrates, Si substrates have the advantage of significantly lower cost, and good electrical and thermal conductivities [2]. However, it is difficult to grow single crystalline GaN directly on Si substrates due to the large lattice mismatch (17%) between GaN and Si and the large difference in thermal expansion coefficients (54%) [3]. Therefore, a range of buffer layers, including SiC [4], low temperature-deposited GaN [5], aluminum arsenide [6] and AlN [7], has been investigated as an intermediate layer to minimize the lattice and thermal expansion mismatch between the GaN layer and Si substrate. AlN is the most universal buffer layer because it supports a high-quality GaN epitaxial layer due to the good wettability of GaN, which produces twodimensional (2D) growth [8], thereby preventing a meltback-etching reaction of Si with Ga [5,9]. In addition, it reduces the lattice and thermal mismatch between GaN and Si.
⁎ Corresponding author. Sungkyunkwan Advanced Institute of Nanotechnology (SAINT), Sungkyunkwan University, Suwon 440-746, Republic of Korea. Tel.: +82 31 290 7388; fax: +82 31 290 7371. E-mail address: [email protected] (D.H. Yoon). 0040-6090/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2009.03.089
AlN is easily deposited by radio frequency (RF) magnetron sputtering but the properties of sputtered AlN buffer layers are different from those of MOCVD-grown AlN buffer layers. This study examined the effect of the AlN buffer layer deposition method and thickness, which controls the stresses and crystallinity of the GaN epitaxial layer, on the optical and structural properties of GaN.
2. Experimental details AlN was selected as a buffer layer to overcome the difference in thermal expansion coefficient and lattice mismatch between GaN and Si (111). RF magnetron sputtering and MOCVD were used to grow the buffer layer. An AlN buffer layer with three different thicknesses (20 nm, 50 nm, 100 nm) was deposited by RF magnetron sputtering. AlN was reactively sputtered on Si substrates at room temperature using Al metal target and nitrogen gas. After sputtering, the samples were annealed in a horizontal quartz tube furnace at temperature up to 950°C for 30 min in an ammonia (NH3) atmosphere. An AlN buffer layer was also prepared by MOCVD. For comparison with the sputtered AlN buffer, AlN buffer layers with four different thicknesses (4 nm, 45 nm, 60 nm, 100 nm) were deposited by MOCVD at 1070 °C using an AIXTRON AIX2400G3 HT MOCVD system. The precursors for Al and N were trimethyl-aluminum (TMAl) and NH3, respectively. The flow of TMAl and NH3 was 45 sccm and 2000 sccm, respectively, and the group V/III ratio was 2320. The same MOCVD system was used to grow the GaN epitaxial layer (1 μm) using trimethyl-gallium (TMGa) and NH3 as the Ga and N precursors, respectively. The growth temperature was 1045 °C. The TMGa and NH3 flow rates were 100 sccm and 4500 sccm, respectively, and the group V/III ratio was 450.
5058
J.H. Yang et al. / Thin Solid Films 517 (2009) 5057–5060
Field emission scanning electron microscopy (FE-SEM) was used to examine the morphological changes to the GaN layer. High-resolution X-ray diffraction (HRXRD) and Raman spectroscopy were used to determine the structural properties and crystallinity of the GaN layer. Photoluminescence (PL) was used to investigate the optical properties of the GaN layer. All characterization experiments were carried out at room temperature. 3. Results and discussion Fig. 1 shows the FE-SEM images of the GaN layer grown on the sputtered AlN buffer layer with different thicknesses. The GaN layers in Fig. 1(b) show no coalescence between the GaN islands. M.A. Moram et al. reported that low crystallinity of the buffer layer affects the disturbance of the coalescence between GaN islands [10].
Fig. 2. GaN (002) ω rocking curves of sputtered AlN buffer layer samples.
Fig. 1. FE-SEM images of GaN layer according to sputtered, AlN buffer layer thickness: (a) AlN 20 nm, (b) AlN 50 nm, and (c) AlN 100 nm.
Therefore, it was concluded that these AlN buffer layers have low crystallinity. Fig. 1(a) and (c) shows non-uniform GaN islands formed by the residual stress of the buffer layer. It was concluded that annealing after sputtering could not improve the quality of the buffer layer for GaN epitaxial growth. Fig. 2 shows the HRXRD GaN (002) ω rocking curves of the GaN layer on the sputtered AlN buffer layer. The (002) ω scan full width at half maximum (FWHM) of GaN on the sputtered AlN buffer layer varied from 5.7–9.7°, which is much larger than that on the MOCVDgrown AlN buffer layer, indicating strong out-of-plane misorientation in the GaN layer on the sputtered AlN layer. The θ–2θ scan shows only (002) and (004) peaks but the (102) phi scan showed no six-fold symmetry. These results showed that an annealing process after sputtering only assists in recrystallization of the c-plane preferred orientation of the AlN buffer layer, which makes an a-direction tilt and twist, and increases the FWHM value of the GaN epitaxial layer. Fig. 3 shows the FE-SEM image of the GaN layers grown on the MOCVD-grown AlN buffer layer with different thicknesses. Fig. 3(a) shows severe V-shaped pits, which were formed by a high density of threading dislocations bending to reduce the surface energy [11], indicating that the thickness of the buffer layer is inadequate. Fig. 3(d) shows cracks arising from a failure of stress control due to the excessive thickness of the 100 nm-thick buffer layer. These results suggest that higher crystallinity and adequate thickness of the buffer layer are essential for ensuring high-quality GaN epitaxial layer growth. Fig. 4 shows the HRXRD GaN (002) ω rocking curves of the GaN on MOCVD-grown AlN buffer layer. The FWHM value of GaN was affected by the AlN buffer layer thickness. As the AlN buffer layer thickness was increased from 4 nm to 60 nm, the FWHM value of GaN decreased from 1.466° to 1.027°, but then increased to 1.188° for the 100 nmthick AlN buffer layer. These values were lower than the (002) ω scan FWHM of the GaN grown on the sputtered AlN buffer layer (5.7–9.7°), indicating higher crystallinity in the GaN layer grown on the MOCVDgrown AlN buffer layer. A slight increase in the FWHM with the 100 nm-thick AlN indicates that the excessive thickness of the buffer layer caused cracks and reduced the GaN crystallinity, as mentioned above. The θ–2θ scan shows only (002) and (004) peaks, and the (102) phi scan shows six-fold symmetry. The FWHM values of the (102) ω rocking curve of the GaN on MOCVD sample were lower than 0.6, which means the edge dislocation toward 〈002〉 direction is effectively eliminated by applying an AlN buffer layer. This suggests that a single crystalline GaN epitaxial layer was grown successfully on Si. Fig. 5 shows the Raman spectra of the GaN on the sputtered and MOCVD-grown AlN buffer layers. The strong peak at 520 cm− 1 was
J.H. Yang et al. / Thin Solid Films 517 (2009) 5057–5060
5059
Fig. 3. FE-SEM images of GaN layer on MOCVD-grown AlN buffer layer: (a) AlN 4 nm, (b) AlN 45 nm (c) AlN 60 nm, and (d) AlN 100 nm.
attributed to the Si substrate, and there were two Raman peaks that were indicative of the wurtzite GaN crystal: phonon modes for high E2 at 567 cm− 1 and A1(LO) at 736 cm− 1 [12]. The latter has atomic displacement parallel to the c-axis, and propagates along the c-axis [13]. Therefore, the crystallinity can be measured by the relative intensities of the E2 and A1(LO) modes. The Raman spectra of GaN on the sputtered AlN buffer layer showed a larger E2/A1(LO) ratio range of 15.805–63.641 than on the MOCVD-grown AlN buffer layer (range = 8.461–12.914). In agreement with the HRXRD results of the GaN (002) ω rocking curves, it indicates that the GaN on the sputtered AlN buffer layer had poorer crystallinity than that on the MOCVDgrown AlN buffer layer. The E2/A1(LO) ratio decreased from 12.914 to
8.461 with increasing AlN buffer layer thickness from 4 nm to 60 nm, but then increased to 8.906 for the 100 nm-thick AlN buffer layer. This indicates that an AlN buffer layer thickness N60 nm can relieve the stress originating from lattice mismatch. However, an excessively thick AlN buffer layer cannot relieve the stress originating from thermal coefficient mismatch. Fig. 6 shows the PL spectrum for GaN on the sputtered and MOCVDgrown AlN buffer layers. The main emission peak was observed at 363 nm for all samples but the intensity and FWHM were affected by the AlN buffer layer. A lower emission intensity and larger FWHM were observed for the GaN layers grown on the sputtered AlN buffer layer
Fig. 4. GaN (002) ω rocking curves of MOCVD-grown AlN buffer layer samples.
Fig. 5. Raman spectra for the GaN epitaxial layer on Si with the AlN buffer layer.
5060
J.H. Yang et al. / Thin Solid Films 517 (2009) 5057–5060
nesses. FE-SEM showed that GaN did not coalesce on the sputtered AlN buffer layer but coalesced with a single domain on the MOCVDgrown AlN buffer layer. The GaN islands deposited on the sputtered AlN buffer layer with a c-axis preferred orientation did not show any coalescence. The buffer layer thickness was one of the key factors for growing high-quality GaN. An excessively thin AlN buffer layer could not reduce the dislocation density, which led to V-shaped pits on the GaN surfaces. On the other hand, an excessively thick AlN buffer layer could not relieve the stress arising from thermal coefficient mismatch, which led to cracking in the GaN layer. HRXRD and Raman spectroscopy showed that the GaN on the MOCVD-grown AlN buffer layer had better crystalline quality and fewer defects than the GaN layer on the sputtered AlN buffer layer. The GaN on the MOCVD-grown AlN buffer layer had a higher PL emission efficiency and narrower FWHM of emission light than the GaN layer on the sputtered AlN layer.
Fig. 6. PL spectrum for the GaN epitaxial layer on Si with the AlN buffer layer.
than for the GaN layers grown on the MOCVD-grown AlN buffer layer, confirming the higher quality of GaN on the MOCVD-grown AlN buffer layer. A lattice defect-related, deep-level transition was also observed. GaN deposited on the 60 nm- and 100 nm-thick, MOCVD-grown AlN buffer layers had a similar intensity and similar FWHM of PL emission of 8.8 nm. However, GaN on the 100 nm-thick AlN buffer layer had a much deeper level transition than GaN on the 60 nm-thick AlN buffer layer. Overall, the buffer layer thickness and deposition method have significant effects on the properties of the GaN layer. GaN on the sputtered AlN buffer layer did not coalescence and showed a larger FWHM value of the GaN (002) ω rocking curves than the GaN on MOCVD-grown AlN buffer layer. The GaN layers on the thin and thick AlN buffer layers showed many v-pits and cracks, respectively. Among the samples, the GaN grown on the 60 nm-thick, MOCVD-grown AlN buffer layer showed the highest quality. 4. Conclusion GaN layers were grown on Si (111) substrates by MOCVD using sputtered and MOCVD-grown AlN buffer layers with various thick-
References [1] H. Morko G., S. Strite, G.B. Gao, M.E. Lin, B. Sverdlov, M. Burns, J. Appl. Phys. 76 (3) (1997) 1363. [2] L. Liu, J.H. Edgar, Mater. Sci. Eng., R 37 (2002) 61. [3] S. Pal, C. Jacob, Bull. Mater. Sci. 27 (6) (2004) 501. [4] H.M. Liawa, R. Venugopal, J. Wan, R. Doyle, P.L. Fejes, M.R. Melloch, Solid-State Electron. 44 (2000) 685. [5] Hiroyasu Ishikawa, Kensaku Yamamoto, Takashi Egawa, Tetsuo Soga, Takashi Jimbo, Masayoshi Umeno, J. Cryst. Growth 189/190 (1998) 178. [6] A. Strittmatter, A. Krost, M. Straßburg, V. Turck, D. Bimberg, J. Blasing, J. Christen, Appl. Phys. Lett. 74 (9) (1999) 1242. [7] Yuan Lu, Xianglin Liu, Da-Cheng Lu, Hairong Yuan, Zhen Chen, Tiwen Fan, Yufeng Li, Peide Han, Xiaohui Wang, Du Wang, Zhanguo Wang, J. Cryst. Growth 236 (2002) 77. [8] P. Waltereit, O. Brandt, A. Trampert, M. Ramsteiner, M. Reiche, M. Qi, K.H. Ploog, Appl. Phys. Lett. 74 (24) (1999) 3660. [9] A. Krost, A. Dadgar, Mater. Sci. Eng., B 93 (2002) 77. [10] M.A. Moram, M.J. Kappers, T.B. Joyce, P.R. Chalker, Z.H. Barber, C.J. Humphreys, J. Cryst. Growth 308 (2007) 302. [11] J. Grilhe, Europhys. Lett. 23 (1993) 141. [12] V.Yu. Davydov, Yu.E. Kitaev, I.N. Goncharuk, A.N. Smirnov, J. Graul, O. Semchinova, D. Uffmann, M.B. Smirnov, A.P. Mirgorodsky, R.A. Evarestov, Phys. Rev. B 58 (1998) 12899. [13] H. Harima, J. Phys., Condens. Matter 14 (2002) R967.
Contents lists available at ScienceDirect
Thin Solid Films j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / t s f
Influence of AlN buffer layer thickness and deposition methods on GaN epitaxial growth J.H. Yang a, S.M. Kang b, D.V. Dinh a, D.H. Yoon a,b,⁎ a b
Sungkyunkwan Advanced Institute of Nanotechnology (SAINT), Sungkyunkwan University, Suwon 440-746, Republic of Korea School of Advanced Materials Science & Engineering, Sungkyunkwan University, Suwon 440-746, Republic of Korea
a r t i c l e
i n f o
Available online 20 March 2009 Keywords: MOCVD GaN AIN Raman spectroscopy Photo luminescence X-ray diffraction
a b s t r a c t A gallium nitride (GaN) epitaxial layer was grown by metal-organic chemical vapor deposition (MOCVD) on Si (111) substrates with aluminum nitride (AlN) buffer layers at various thicknesses. The AlN buffer layers were deposited by two methods: radio frequency (RF) magnetron sputtering and MOCVD. The effect of the AlN deposition method and layer thickness on the morphological, structural and optical properties of the GaN layers was investigated. Field emission scanning electron microscopy showed that GaN did not coalesce on the sputtered AlN buffer layer. On the other hand, it coalesced with a single domain on the MOCVD-grown AlN buffer layer. Structural and optical analyses indicated that GaN on the MOCVD-grown AlN buffer layer had fewer defects and a better aligned lattice to the a- and c-axes than GaN on the sputtered AlN buffer layer. © 2009 Elsevier B.V. All rights reserved.
1. Introduction With its direct and wide bandgap of 3.39 eV at room temperature, high thermal stability, high breakdown field voltage and high saturation drift velocity, gallium nitride (GaN) is considered a promising material for optoelectronic applications in the blue and UV wavelengths, as well as in high-power and high-temperature electronics [1]. Commercially available devices, such as light emitting diodes, are usually grown by metal-organic chemical vapor deposition (MOCVD) on sapphire or SiC substrates due to the lack of large and inexpensive GaN substrates. Compared to these substrates, Si substrates have the advantage of significantly lower cost, and good electrical and thermal conductivities [2]. However, it is difficult to grow single crystalline GaN directly on Si substrates due to the large lattice mismatch (17%) between GaN and Si and the large difference in thermal expansion coefficients (54%) [3]. Therefore, a range of buffer layers, including SiC [4], low temperature-deposited GaN [5], aluminum arsenide [6] and AlN [7], has been investigated as an intermediate layer to minimize the lattice and thermal expansion mismatch between the GaN layer and Si substrate. AlN is the most universal buffer layer because it supports a high-quality GaN epitaxial layer due to the good wettability of GaN, which produces twodimensional (2D) growth [8], thereby preventing a meltback-etching reaction of Si with Ga [5,9]. In addition, it reduces the lattice and thermal mismatch between GaN and Si.
⁎ Corresponding author. Sungkyunkwan Advanced Institute of Nanotechnology (SAINT), Sungkyunkwan University, Suwon 440-746, Republic of Korea. Tel.: +82 31 290 7388; fax: +82 31 290 7371. E-mail address: [email protected] (D.H. Yoon). 0040-6090/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2009.03.089
AlN is easily deposited by radio frequency (RF) magnetron sputtering but the properties of sputtered AlN buffer layers are different from those of MOCVD-grown AlN buffer layers. This study examined the effect of the AlN buffer layer deposition method and thickness, which controls the stresses and crystallinity of the GaN epitaxial layer, on the optical and structural properties of GaN.
2. Experimental details AlN was selected as a buffer layer to overcome the difference in thermal expansion coefficient and lattice mismatch between GaN and Si (111). RF magnetron sputtering and MOCVD were used to grow the buffer layer. An AlN buffer layer with three different thicknesses (20 nm, 50 nm, 100 nm) was deposited by RF magnetron sputtering. AlN was reactively sputtered on Si substrates at room temperature using Al metal target and nitrogen gas. After sputtering, the samples were annealed in a horizontal quartz tube furnace at temperature up to 950°C for 30 min in an ammonia (NH3) atmosphere. An AlN buffer layer was also prepared by MOCVD. For comparison with the sputtered AlN buffer, AlN buffer layers with four different thicknesses (4 nm, 45 nm, 60 nm, 100 nm) were deposited by MOCVD at 1070 °C using an AIXTRON AIX2400G3 HT MOCVD system. The precursors for Al and N were trimethyl-aluminum (TMAl) and NH3, respectively. The flow of TMAl and NH3 was 45 sccm and 2000 sccm, respectively, and the group V/III ratio was 2320. The same MOCVD system was used to grow the GaN epitaxial layer (1 μm) using trimethyl-gallium (TMGa) and NH3 as the Ga and N precursors, respectively. The growth temperature was 1045 °C. The TMGa and NH3 flow rates were 100 sccm and 4500 sccm, respectively, and the group V/III ratio was 450.
5058
J.H. Yang et al. / Thin Solid Films 517 (2009) 5057–5060
Field emission scanning electron microscopy (FE-SEM) was used to examine the morphological changes to the GaN layer. High-resolution X-ray diffraction (HRXRD) and Raman spectroscopy were used to determine the structural properties and crystallinity of the GaN layer. Photoluminescence (PL) was used to investigate the optical properties of the GaN layer. All characterization experiments were carried out at room temperature. 3. Results and discussion Fig. 1 shows the FE-SEM images of the GaN layer grown on the sputtered AlN buffer layer with different thicknesses. The GaN layers in Fig. 1(b) show no coalescence between the GaN islands. M.A. Moram et al. reported that low crystallinity of the buffer layer affects the disturbance of the coalescence between GaN islands [10].
Fig. 2. GaN (002) ω rocking curves of sputtered AlN buffer layer samples.
Fig. 1. FE-SEM images of GaN layer according to sputtered, AlN buffer layer thickness: (a) AlN 20 nm, (b) AlN 50 nm, and (c) AlN 100 nm.
Therefore, it was concluded that these AlN buffer layers have low crystallinity. Fig. 1(a) and (c) shows non-uniform GaN islands formed by the residual stress of the buffer layer. It was concluded that annealing after sputtering could not improve the quality of the buffer layer for GaN epitaxial growth. Fig. 2 shows the HRXRD GaN (002) ω rocking curves of the GaN layer on the sputtered AlN buffer layer. The (002) ω scan full width at half maximum (FWHM) of GaN on the sputtered AlN buffer layer varied from 5.7–9.7°, which is much larger than that on the MOCVDgrown AlN buffer layer, indicating strong out-of-plane misorientation in the GaN layer on the sputtered AlN layer. The θ–2θ scan shows only (002) and (004) peaks but the (102) phi scan showed no six-fold symmetry. These results showed that an annealing process after sputtering only assists in recrystallization of the c-plane preferred orientation of the AlN buffer layer, which makes an a-direction tilt and twist, and increases the FWHM value of the GaN epitaxial layer. Fig. 3 shows the FE-SEM image of the GaN layers grown on the MOCVD-grown AlN buffer layer with different thicknesses. Fig. 3(a) shows severe V-shaped pits, which were formed by a high density of threading dislocations bending to reduce the surface energy [11], indicating that the thickness of the buffer layer is inadequate. Fig. 3(d) shows cracks arising from a failure of stress control due to the excessive thickness of the 100 nm-thick buffer layer. These results suggest that higher crystallinity and adequate thickness of the buffer layer are essential for ensuring high-quality GaN epitaxial layer growth. Fig. 4 shows the HRXRD GaN (002) ω rocking curves of the GaN on MOCVD-grown AlN buffer layer. The FWHM value of GaN was affected by the AlN buffer layer thickness. As the AlN buffer layer thickness was increased from 4 nm to 60 nm, the FWHM value of GaN decreased from 1.466° to 1.027°, but then increased to 1.188° for the 100 nmthick AlN buffer layer. These values were lower than the (002) ω scan FWHM of the GaN grown on the sputtered AlN buffer layer (5.7–9.7°), indicating higher crystallinity in the GaN layer grown on the MOCVDgrown AlN buffer layer. A slight increase in the FWHM with the 100 nm-thick AlN indicates that the excessive thickness of the buffer layer caused cracks and reduced the GaN crystallinity, as mentioned above. The θ–2θ scan shows only (002) and (004) peaks, and the (102) phi scan shows six-fold symmetry. The FWHM values of the (102) ω rocking curve of the GaN on MOCVD sample were lower than 0.6, which means the edge dislocation toward 〈002〉 direction is effectively eliminated by applying an AlN buffer layer. This suggests that a single crystalline GaN epitaxial layer was grown successfully on Si. Fig. 5 shows the Raman spectra of the GaN on the sputtered and MOCVD-grown AlN buffer layers. The strong peak at 520 cm− 1 was
J.H. Yang et al. / Thin Solid Films 517 (2009) 5057–5060
5059
Fig. 3. FE-SEM images of GaN layer on MOCVD-grown AlN buffer layer: (a) AlN 4 nm, (b) AlN 45 nm (c) AlN 60 nm, and (d) AlN 100 nm.
attributed to the Si substrate, and there were two Raman peaks that were indicative of the wurtzite GaN crystal: phonon modes for high E2 at 567 cm− 1 and A1(LO) at 736 cm− 1 [12]. The latter has atomic displacement parallel to the c-axis, and propagates along the c-axis [13]. Therefore, the crystallinity can be measured by the relative intensities of the E2 and A1(LO) modes. The Raman spectra of GaN on the sputtered AlN buffer layer showed a larger E2/A1(LO) ratio range of 15.805–63.641 than on the MOCVD-grown AlN buffer layer (range = 8.461–12.914). In agreement with the HRXRD results of the GaN (002) ω rocking curves, it indicates that the GaN on the sputtered AlN buffer layer had poorer crystallinity than that on the MOCVDgrown AlN buffer layer. The E2/A1(LO) ratio decreased from 12.914 to
8.461 with increasing AlN buffer layer thickness from 4 nm to 60 nm, but then increased to 8.906 for the 100 nm-thick AlN buffer layer. This indicates that an AlN buffer layer thickness N60 nm can relieve the stress originating from lattice mismatch. However, an excessively thick AlN buffer layer cannot relieve the stress originating from thermal coefficient mismatch. Fig. 6 shows the PL spectrum for GaN on the sputtered and MOCVDgrown AlN buffer layers. The main emission peak was observed at 363 nm for all samples but the intensity and FWHM were affected by the AlN buffer layer. A lower emission intensity and larger FWHM were observed for the GaN layers grown on the sputtered AlN buffer layer
Fig. 4. GaN (002) ω rocking curves of MOCVD-grown AlN buffer layer samples.
Fig. 5. Raman spectra for the GaN epitaxial layer on Si with the AlN buffer layer.
5060
J.H. Yang et al. / Thin Solid Films 517 (2009) 5057–5060
nesses. FE-SEM showed that GaN did not coalesce on the sputtered AlN buffer layer but coalesced with a single domain on the MOCVDgrown AlN buffer layer. The GaN islands deposited on the sputtered AlN buffer layer with a c-axis preferred orientation did not show any coalescence. The buffer layer thickness was one of the key factors for growing high-quality GaN. An excessively thin AlN buffer layer could not reduce the dislocation density, which led to V-shaped pits on the GaN surfaces. On the other hand, an excessively thick AlN buffer layer could not relieve the stress arising from thermal coefficient mismatch, which led to cracking in the GaN layer. HRXRD and Raman spectroscopy showed that the GaN on the MOCVD-grown AlN buffer layer had better crystalline quality and fewer defects than the GaN layer on the sputtered AlN buffer layer. The GaN on the MOCVD-grown AlN buffer layer had a higher PL emission efficiency and narrower FWHM of emission light than the GaN layer on the sputtered AlN layer.
Fig. 6. PL spectrum for the GaN epitaxial layer on Si with the AlN buffer layer.
than for the GaN layers grown on the MOCVD-grown AlN buffer layer, confirming the higher quality of GaN on the MOCVD-grown AlN buffer layer. A lattice defect-related, deep-level transition was also observed. GaN deposited on the 60 nm- and 100 nm-thick, MOCVD-grown AlN buffer layers had a similar intensity and similar FWHM of PL emission of 8.8 nm. However, GaN on the 100 nm-thick AlN buffer layer had a much deeper level transition than GaN on the 60 nm-thick AlN buffer layer. Overall, the buffer layer thickness and deposition method have significant effects on the properties of the GaN layer. GaN on the sputtered AlN buffer layer did not coalescence and showed a larger FWHM value of the GaN (002) ω rocking curves than the GaN on MOCVD-grown AlN buffer layer. The GaN layers on the thin and thick AlN buffer layers showed many v-pits and cracks, respectively. Among the samples, the GaN grown on the 60 nm-thick, MOCVD-grown AlN buffer layer showed the highest quality. 4. Conclusion GaN layers were grown on Si (111) substrates by MOCVD using sputtered and MOCVD-grown AlN buffer layers with various thick-
References [1] H. Morko G., S. Strite, G.B. Gao, M.E. Lin, B. Sverdlov, M. Burns, J. Appl. Phys. 76 (3) (1997) 1363. [2] L. Liu, J.H. Edgar, Mater. Sci. Eng., R 37 (2002) 61. [3] S. Pal, C. Jacob, Bull. Mater. Sci. 27 (6) (2004) 501. [4] H.M. Liawa, R. Venugopal, J. Wan, R. Doyle, P.L. Fejes, M.R. Melloch, Solid-State Electron. 44 (2000) 685. [5] Hiroyasu Ishikawa, Kensaku Yamamoto, Takashi Egawa, Tetsuo Soga, Takashi Jimbo, Masayoshi Umeno, J. Cryst. Growth 189/190 (1998) 178. [6] A. Strittmatter, A. Krost, M. Straßburg, V. Turck, D. Bimberg, J. Blasing, J. Christen, Appl. Phys. Lett. 74 (9) (1999) 1242. [7] Yuan Lu, Xianglin Liu, Da-Cheng Lu, Hairong Yuan, Zhen Chen, Tiwen Fan, Yufeng Li, Peide Han, Xiaohui Wang, Du Wang, Zhanguo Wang, J. Cryst. Growth 236 (2002) 77. [8] P. Waltereit, O. Brandt, A. Trampert, M. Ramsteiner, M. Reiche, M. Qi, K.H. Ploog, Appl. Phys. Lett. 74 (24) (1999) 3660. [9] A. Krost, A. Dadgar, Mater. Sci. Eng., B 93 (2002) 77. [10] M.A. Moram, M.J. Kappers, T.B. Joyce, P.R. Chalker, Z.H. Barber, C.J. Humphreys, J. Cryst. Growth 308 (2007) 302. [11] J. Grilhe, Europhys. Lett. 23 (1993) 141. [12] V.Yu. Davydov, Yu.E. Kitaev, I.N. Goncharuk, A.N. Smirnov, J. Graul, O. Semchinova, D. Uffmann, M.B. Smirnov, A.P. Mirgorodsky, R.A. Evarestov, Phys. Rev. B 58 (1998) 12899. [13] H. Harima, J. Phys., Condens. Matter 14 (2002) R967.