- Email: [email protected]
Effect of gas flow on the selective area growth of gallium nitride via metal organic vapor phase epitaxy
ARTICLE IN PRESS
Journal of Crystal Growth 306 (2007) 75–79 www.elsevier.com/locate/jcrysgro
Effect of gas flow on the selective area growth of gallium nitride via metal organic vapor phase epitaxy L.E. Rodak, K.R. Kasarla, D. Korakakis Lane Department of Computer Science and Electrical Engineering, West Virginia University, Morgantown, WV 26506, USA Received 7 March 2007; received in revised form 22 March 2007; accepted 26 March 2007 Communicated by R.M. Biefeld Available online 16 May 2007
Abstract The effect of gas flow on the selective area growth (SAG) of gallium nitride (GaN) grown via metal organic vapor phase epitaxy (MOVPE) has been investigated. In this study, the SAG of GaN was carried out on a silicon dioxide striped pattern along the GaN h1 1¯ 0 0i direction. SAG was initiated with the striped pattern oriented parallel and normal to the incoming gas flow in a horizontal reactor. The orientation of the pattern did not impact cross section of the structure after re-growth as both orientations resulted in similar trapezoidal structures bounded by the (0 0 0 1) and f1 1 2¯ ng facets (n 1:7 2:2). However, the growth rates were shown to depend on the orientation of the pattern as the normally oriented samples exhibited enhanced vertical and cross-sectional growth rates compared to the parallel oriented samples. All growths occurred under identical conditions and therefore the difference in growth rates must be attributed to a difference in mass transport of species. r 2007 Elsevier B.V. All rights reserved. PACS: 81.15.Gh; 71.55.Eq Keywords: A3. Metalorganic chemical vapor deposition; A3. Selective area growth; B1. Gallium nitride
1. Introduction Gallium nitride (GaN) is a promising semiconductor for use in a wide range of applications. The 3.4 eV band gap allows for emission of blue and ultraviolet (UV) wavelengths while the binary and ternary alloys in (Al, In, Ga)N can be engineered for emission over the entire visible spectrum. This makes the material useful for a number of optical applications such as solid-state lighting, biological sensing, and optical storage. Furthermore, the high electron saturation velocity and high breakdown field of GaN make it suitable for high-frequency and high-power applications [1]. GaN is typically grown heteroepitaxially on foreign substrates such as sapphire or silicon carbide [1]. The lattice mismatch between these materials leads to a large density of threading dislocations that reduce the lifetime [2] Corresponding author. Tel.: +1 (304) 293 0405x3587.
E-mail address: [email protected] (L.E. Rodak). 0022-0248/$ - see front matter r 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2007.03.063
and efficiency of devices fabricated from this material by forming nonradiative recombination centers [3]. Typical films grown via metal organic vapor phase epitaxy (MOVPE) have dislocation densities ranging from 109 to 1011 cm3 [1,2]. Extensive research efforts have focused on reducing the dislocation density through methods such as epitaxial lateral overgrowth (ELOG) [1] and pendeoexpitaxy [4]. ELOG has been demonstrated to reduce the dislocation density a few orders of magnitude to 107 cm3 [1,5]. The ELOG process involves partially masking a GaN layer with a striped dielectric pattern and then continuing GaN growth. Because GaN selectively grows only on the exposed GaN areas, growth begins in the pattern openings and extends laterally over the pattern. The ELOG process is heavily dependent on the selective area growth (SAG) of GaN, and therefore a complete physical model of GaN SAG would allow us to further utilize this technique by examining the dependence of morphology on growth parameters. A comprehensive model would also extend
ARTICLE IN PRESS 76
L.E. Rodak et al. / Journal of Crystal Growth 306 (2007) 75–79
the possible applications of SAG techniques beyond dislocation reduction to directly implementing threedimensional structures in device fabrication. For example, it has been demonstrated that GaN SAG structures with a square cross section could be implemented as low loss waveguides [6]. The effects of the various growth parameters on the morphology of GaN SAG on a striped dielectric pattern in rotational environments have been previously investigated. The majority of these studies investigated the effect of growth conditions for ELOG. Before coalescence, ELOG growths typically result in three-dimensional structures bounded by some combination of the (0 0 0 1), f1 1 2¯ 2g, and f1 1 2¯ 0g facets and exhibit square, triangular, or trapezoidal cross sections. Growth parameters investigated include temperature, pressure, V/III ratio, and the pattern fill factor. Low growth temperatures result in f1 1¯ 0 1g sidewalls and poor surface morphology due to the decreased migration of gallium species. Increased growth temperature results in f1 1 2¯ 2g sidewalls and improved morphology on the (0 0 0 1) surface. Even higher growth temperatures lead to the formation of f1 1 2¯ 0g sidewalls. Sidewall formation can be attributed to the number of dangling bonds on each surface. The f1 1 2¯ 0g plane has fewer dangling bonds when compared to the f1 1 2¯ 2g plane and therefore is energetically favored under high temperatures [7,8]. The f1 1 2¯ 2g facet is comprised of rows of Ga–N dimmers with a probable configuration of two dangling bonds on the gallium atom and one dangling bond on the nitrogen atom indicating that this facet is most stable under low V/III ratios. Conversely, an equal number of gallium and nitrogen atoms, with one dangling bond each, exist on the f1 1 2¯ 0g facet causing this facet to be stable under high V/III ratios [9]. The local V/III ratio increases as the ELOG growth laterally extends over the mask and the concentration of gallium species diffusing from the mask to the growth location decreases resulting in the transformation from f1 1 2¯ 2g facets to f1 1 2¯ 0g facets [10]. Furthermore, it has been demonstrated that flow modulation of ammonia (NH3) can be used to control the morphology [11]. As the NH3 interruption time increased, the lateral growth rate increases and is credited to the enhanced diffusion of the gallium species during the NH3 interruption. All of the studies mentioned have investigated the growth in rotational environments. It is the goal of this
work to expand our initial investigation of the effect of gas flow [12,13] on the morphology of the SAG of the GaN in a nonrotating environment in order to further develop a complete physical model. In this study, the SAG of GaN was carried out on a silicon dioxide striped pattern via MOVPE. 2. Experimental procedure Approximately a 1.5 mm thick GaN film was grown on a 2 in. diameter c-plane sapphire wafer via MOVPE using a two step process involving a 30 nm aluminum nitride nucleation layer followed by a high-temperature GaN main layer in an AIXTRON 200/4 RF-S horizontal reactor. In this system, the metal organic and hydride source gases enter at the center of one end of the reactor and are separated by a quartz plate until mixing directly before the susceptor. Trimethylgallium and ammonia were used as the source gasses while hydrogen was used as the carrier gas. Next, a 130 nm layer of silicon dioxide was deposited via plasma enhanced chemical vapor deposition (PECVD) by an Oxford Plasma Lab 80 Plus system. Standard photolithography and wet etching techniques were used to create a striped pattern in the silicon dioxide layer along the GaN h1 1¯ 0 0i direction as shown in Fig. 1a. Both the window opening and the stripe width were 5 mm. Two similarly sized pieces, each approximately 1/6 of a 2 in. wafer, were placed side by side in a line normal to the incoming gas flow in the reactor for re-growth. One sample was placed with the striped pattern oriented parallel to the incoming gas flow and the other with the striped pattern oriented normal to the incoming gas flow as shown in Fig. 1b. In roughly half of the growths, the samples with the striped pattern oriented parallel to the incoming gas flow were located on the left-hand side of the incoming gas, as shown in Fig. 1b. In the other growths, the samples with the striped pattern oriented parallel to the incoming gas flow were located on the right-hand side. This configuration was used in order to simultaneously fit two samples on the 2 in. susceptor and ensure similar growth conditions by eliminating the effects of uneven gas flow and substrate heating. The samples were grown with no rotation at 1115 1C, 200 mbar, and a V/III ratio around 1400. The ammonia and the carrier gas flow rates were around 1.2 and 4.0 slm, respectively. Table 1 summarizes the various growth parameters during the different stages of SAG.
Fig. 1. (a) Striped pattern along GaN h1 1¯ 0 0idirection; (b) Samples oriented parallel and normal to the incoming gas flow during SAG of GaN.
ARTICLE IN PRESS L.E. Rodak et al. / Journal of Crystal Growth 306 (2007) 75–79
77
Table 1 Growth conditions for selective area growth of GaN
Nucleation layer Main layer Step 1-base structure Step 2-SAG
Growth time (min)
Temperature (1C)
TMAl (mmol/min)
TMGa (mmol/min)
V/III
10 45 30 15–160
1125 1125 1115 1115
12 — — —
— 100 38 38
5500 700 1400 1400
The cross sections of the resulting three-dimensional structures were observed using a Hitachi S4700 scanning electron microscope (SEM) and conventional image processing techniques were used to determine the size of the structures. For practical reasons, one piece per growth per orientation was evaluated and each piece characterized was about one quarter of the size of the grown sample. In order to eliminate experimental artifacts in the analysis of the results, such as temperature nonuniformity, images and measurements were taken from areas of the two samples that were closest during growth. Any variations in the samples are summarized by the error bars shown in each figure. Occasionally, pieces from other parts of the sample, both off center and further downstream, were evaluated. Based on SEM images, the growth results were consistent across the samples and fall within the shown error bars.
Fig. 2. SEM image of SAG base structure. Structure was identical for samples with stripes oriented normal or parallel to incoming gas flow.
3. Results and discussion Figs. 2 and 3 show the SEM images of the cross section for the samples with stripes oriented parallel and normal to the incoming gas flow. After 30 min of growth both orientations exhibited similar growth modes resulting in the same trapezoidal structure with the (0 0 0 1), f1 1 2¯ 2g, and f1 1 2¯ 0g facets exposed shown in Fig. 2. Therefore, all subsequent growths were performed on this base structure and the growth rates were measured as an increase in size from this base structure. It should be noted that no asymmetry in the structures was observed for either orientation as might be expected. Throughout later growth stages, both samples maintained a similar trapezoidal cross section with (0 0 0 1) and f1 1 2¯ ng facets (n 1:7 2:2) exposed. The samples oriented normal to the flow experienced larger growth rates, as is indicated in Figs. 2–5. Fig. 4 illustrates the height as a function of growth time. Clearly, the normally oriented sample experiences a larger vertical growth rate when compared to the parallel oriented sample. Fig. 5 illustrates the crosssectional area of the three-dimensional structures per pattern period. Once again, the normally oriented sample is displaying an enhanced cross-sectional growth rate. In both Figs. 4 and 5, the line is a linear fit using the error bars as weight. Also shown in Figs. 4 and 5 is the height and cross section of the structures when SAG is carried out for 120 min while rotating the sample. The height of the rotating sample is similar to that of the normal oriented
sample, while the cross section is similar to the parallel oriented sample. Since the growth conditions are identical for both orientations, the increased cross-sectional growth rate of the normal oriented sample indicates that more species are being incorporated into the growth. Both orientations maintain exposed (0 0 0 1) and f1 1 2¯ ng facets (n 1:7 2:2), indicating that there is no difference in surface growth kinetics between the samples. In MOVPE, it is assumed that the gallium species sticking coefficient is unity on the GaN surface [10] and there is efficient lateral transport across the oxide mask allowing all reactants reaching the surface to be incorporated into the SAG growth [9,14,15]. Therefore, it can be concluded that there is a difference in mass transport between the normal and parallel oriented sample to account for the difference in growth rate. The normal orientation experienced enhanced growth rate because more species were diffusing to the growth site. The size of the structures is negligible when compared to the boundary layer thickness, which is proportional to the square root of the kinematic viscosity of the gas and the distance from the edge of the susceptor and inversely proportional to the square root of the free stream gas velocity [16,17]. Therefore, it can be assumed that the height of the structure does not have a significant impact on the boundary layer, which is consistent with the fact that both orientations maintain constant growth rates even as the structures grow taller. These findings are
ARTICLE IN PRESS 78
L.E. Rodak et al. / Journal of Crystal Growth 306 (2007) 75–79
Fig. 3. SEM images of SAG selective area growth of GaN with h1 1¯ 0 0i striped pattern oriented parallel and normal to the incoming gas flow.
Fig. 4. Height as a function of growth time for parallel and normal oriented samples. The lines are a linear fit of the data using the error bars as weight. The y-axis intercept corresponds to the height of the base structure.
Fig. 5. Cross section as a function of time for normal and parallel oriented samples. The lines are a linear fit of the data using the error bars as weight. The y-axis intercept corresponds to the cross-sectional area of the base structure.
contrary to the results obtained when the effect of gas flow was investigated on cubic GaN [18]. Cubic GaN was reported to demonstrate only a slight dependence of the cross-sectional geometry and growth rates on the orientation of the pattern with respect to the incoming gas flow. Future work is needed to further explain the mechanism responsible for dependence of growth rates on the orientation of the striped pattern with respect to the incoming gas flow.
4. Conclusions In this work, we investigated the effect of gas flow on the SAG of GaN on a dielectric striped pattern by placing the pattern parallel and normal to the incoming gas flow. Our results show that there is a strong dependence of growth rates on the orientation of the pattern and therefore on the gas flow. Initially, the parallel and normal orientations exhibit similar growth rates and for the same structures.
ARTICLE IN PRESS L.E. Rodak et al. / Journal of Crystal Growth 306 (2007) 75–79
However, in later growth stages the samples with the pattern oriented normal to the incoming gas flow showed enhanced vertical and cross-sectional growth rates. While current models have accounted for the effect of various growth parameters including temperature, pressure, V/III ratio, and fill factor, most have assumed the samples were rotating during growth. This work indicates that gas flow plays an important role in the SAG and therefore the rotation of the sample needs to be incorporated into a complete physical model. Acknowledgments This work was supported in part by a grant from the West Virginia Graduate Student Fellowship in Science, Technology, Engineering and Math (STEM) program to L.E.R. and by the National Science Foundation (Grant EPS-0554328), The State of West Virginia Research Challenge Fund, and the West Virginia University Research Corporation. Partial funding was also provided by AIXTRON Inc. References [1] P. Gibart, Rep. Prog. Phys. 67 (2004) 667. [2] B. Beaumont, Ph. Venne´gue´s, P. Gibart, Phys. Status Solidi (b) 227 (1) (2001) 1.
79
[3] S.J. Rosner, E.C. Carr, M.J. Ludowise, G. Girolami, H.I. Erikson, Appl. Phys. Lett. 70 (1997) 4. [4] T.S. Zheleva, S.A. Smith, D.B. Thomson, T. Gehrke, K.J. Linthicum, P. Rajagopal, E. Carlson, W.M. Ashmawi, R.F. Davis, MRS Internet J. Nitride Semicond. Res. 4S1 (1999) G3.38. [5] P. Vennegues, B. Beaumont, V. Bousquet, M. Vaille, P. Gibart, J. Appl. Phys. 87 (2000) 4175. [6] T. Tanaka, K. Uchida, A. Watanabe, S. Minagawa, Appl. Phys. Lett. 68 (1996) 7. [7] K. Hiramatsu, J. Phys.: Condens. Matter 13 (2001) 6961. [8] K. Hiramatsu, K. Nishiyama, A. Motogaito, H. Miyake, Y. Iyechika, T. Maeda, Phys. Status Solidi (a) 176 (1999) 535. [9] H. Marchand, J.P. Ibbetson, P.T. Fini, S. Keller, S.P. DenBaars, J.S. Speck, U.K. Mishra, J. Crystal Growth 195 (1998) 328. [10] G. Feng, Y. Fu, J.S. Xia, J.J. Zhu, X.M. Shen, D.G. Zhao, H. Yang, J.W. Liang, J. Phys. D: Appl. Phys. 35 (2002) 2731. [11] X. Zhang, P.D. Dapkus, D.H. Rich, Appl. Phys. Lett. 77 (2000) 10. [12] L.E. Rodak, N.J. Berry Ann, K. Kasarla, Y. Yang, Mat. Res. Soc. Symp. Proc. 892 (2006) 665. [13] N.J. Berry Ann, L.E. Rodak, K. Kasarla, Y. Yang, Proc. SPIE 6017 (2005) 60170. [14] K. Hiramatsu, H. Miyake, Mat. Res. Soc. Symp. Proc. 639 (2001). [15] M.E. Coltrin, C.C. Mitchel, III-Nitride Based Semiconductor Electronic and Optical Devices and State-of-the-Art Program on Compound Semiconductors XXXIV, in: F. Ren, D. N. Buckley, S.N.G. Chu, S.J. Pearton (Eds) The Electrochemical Society Proceedings Series, Pennington, NJ, 2001, PV 2001-01, 1p. [16] G.B. Stringfellow, Organometallic Vapor-Phase Epitaxy, Academic Press, San Diego, CA, 1999, p. 303. [17] K. Harafuji, Jpn. J. Appl. Phys. 40 (2001) 6263. [18] X.M. Shen, G. Feng, B.S. Zhang, L.H. Duan, Y.T. Wang, H. Yang, J. Crystal Growth 252 (2003) 9.
Journal of Crystal Growth 306 (2007) 75–79 www.elsevier.com/locate/jcrysgro
Effect of gas flow on the selective area growth of gallium nitride via metal organic vapor phase epitaxy L.E. Rodak, K.R. Kasarla, D. Korakakis Lane Department of Computer Science and Electrical Engineering, West Virginia University, Morgantown, WV 26506, USA Received 7 March 2007; received in revised form 22 March 2007; accepted 26 March 2007 Communicated by R.M. Biefeld Available online 16 May 2007
Abstract The effect of gas flow on the selective area growth (SAG) of gallium nitride (GaN) grown via metal organic vapor phase epitaxy (MOVPE) has been investigated. In this study, the SAG of GaN was carried out on a silicon dioxide striped pattern along the GaN h1 1¯ 0 0i direction. SAG was initiated with the striped pattern oriented parallel and normal to the incoming gas flow in a horizontal reactor. The orientation of the pattern did not impact cross section of the structure after re-growth as both orientations resulted in similar trapezoidal structures bounded by the (0 0 0 1) and f1 1 2¯ ng facets (n 1:7 2:2). However, the growth rates were shown to depend on the orientation of the pattern as the normally oriented samples exhibited enhanced vertical and cross-sectional growth rates compared to the parallel oriented samples. All growths occurred under identical conditions and therefore the difference in growth rates must be attributed to a difference in mass transport of species. r 2007 Elsevier B.V. All rights reserved. PACS: 81.15.Gh; 71.55.Eq Keywords: A3. Metalorganic chemical vapor deposition; A3. Selective area growth; B1. Gallium nitride
1. Introduction Gallium nitride (GaN) is a promising semiconductor for use in a wide range of applications. The 3.4 eV band gap allows for emission of blue and ultraviolet (UV) wavelengths while the binary and ternary alloys in (Al, In, Ga)N can be engineered for emission over the entire visible spectrum. This makes the material useful for a number of optical applications such as solid-state lighting, biological sensing, and optical storage. Furthermore, the high electron saturation velocity and high breakdown field of GaN make it suitable for high-frequency and high-power applications [1]. GaN is typically grown heteroepitaxially on foreign substrates such as sapphire or silicon carbide [1]. The lattice mismatch between these materials leads to a large density of threading dislocations that reduce the lifetime [2] Corresponding author. Tel.: +1 (304) 293 0405x3587.
E-mail address: [email protected] (L.E. Rodak). 0022-0248/$ - see front matter r 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2007.03.063
and efficiency of devices fabricated from this material by forming nonradiative recombination centers [3]. Typical films grown via metal organic vapor phase epitaxy (MOVPE) have dislocation densities ranging from 109 to 1011 cm3 [1,2]. Extensive research efforts have focused on reducing the dislocation density through methods such as epitaxial lateral overgrowth (ELOG) [1] and pendeoexpitaxy [4]. ELOG has been demonstrated to reduce the dislocation density a few orders of magnitude to 107 cm3 [1,5]. The ELOG process involves partially masking a GaN layer with a striped dielectric pattern and then continuing GaN growth. Because GaN selectively grows only on the exposed GaN areas, growth begins in the pattern openings and extends laterally over the pattern. The ELOG process is heavily dependent on the selective area growth (SAG) of GaN, and therefore a complete physical model of GaN SAG would allow us to further utilize this technique by examining the dependence of morphology on growth parameters. A comprehensive model would also extend
ARTICLE IN PRESS 76
L.E. Rodak et al. / Journal of Crystal Growth 306 (2007) 75–79
the possible applications of SAG techniques beyond dislocation reduction to directly implementing threedimensional structures in device fabrication. For example, it has been demonstrated that GaN SAG structures with a square cross section could be implemented as low loss waveguides [6]. The effects of the various growth parameters on the morphology of GaN SAG on a striped dielectric pattern in rotational environments have been previously investigated. The majority of these studies investigated the effect of growth conditions for ELOG. Before coalescence, ELOG growths typically result in three-dimensional structures bounded by some combination of the (0 0 0 1), f1 1 2¯ 2g, and f1 1 2¯ 0g facets and exhibit square, triangular, or trapezoidal cross sections. Growth parameters investigated include temperature, pressure, V/III ratio, and the pattern fill factor. Low growth temperatures result in f1 1¯ 0 1g sidewalls and poor surface morphology due to the decreased migration of gallium species. Increased growth temperature results in f1 1 2¯ 2g sidewalls and improved morphology on the (0 0 0 1) surface. Even higher growth temperatures lead to the formation of f1 1 2¯ 0g sidewalls. Sidewall formation can be attributed to the number of dangling bonds on each surface. The f1 1 2¯ 0g plane has fewer dangling bonds when compared to the f1 1 2¯ 2g plane and therefore is energetically favored under high temperatures [7,8]. The f1 1 2¯ 2g facet is comprised of rows of Ga–N dimmers with a probable configuration of two dangling bonds on the gallium atom and one dangling bond on the nitrogen atom indicating that this facet is most stable under low V/III ratios. Conversely, an equal number of gallium and nitrogen atoms, with one dangling bond each, exist on the f1 1 2¯ 0g facet causing this facet to be stable under high V/III ratios [9]. The local V/III ratio increases as the ELOG growth laterally extends over the mask and the concentration of gallium species diffusing from the mask to the growth location decreases resulting in the transformation from f1 1 2¯ 2g facets to f1 1 2¯ 0g facets [10]. Furthermore, it has been demonstrated that flow modulation of ammonia (NH3) can be used to control the morphology [11]. As the NH3 interruption time increased, the lateral growth rate increases and is credited to the enhanced diffusion of the gallium species during the NH3 interruption. All of the studies mentioned have investigated the growth in rotational environments. It is the goal of this
work to expand our initial investigation of the effect of gas flow [12,13] on the morphology of the SAG of the GaN in a nonrotating environment in order to further develop a complete physical model. In this study, the SAG of GaN was carried out on a silicon dioxide striped pattern via MOVPE. 2. Experimental procedure Approximately a 1.5 mm thick GaN film was grown on a 2 in. diameter c-plane sapphire wafer via MOVPE using a two step process involving a 30 nm aluminum nitride nucleation layer followed by a high-temperature GaN main layer in an AIXTRON 200/4 RF-S horizontal reactor. In this system, the metal organic and hydride source gases enter at the center of one end of the reactor and are separated by a quartz plate until mixing directly before the susceptor. Trimethylgallium and ammonia were used as the source gasses while hydrogen was used as the carrier gas. Next, a 130 nm layer of silicon dioxide was deposited via plasma enhanced chemical vapor deposition (PECVD) by an Oxford Plasma Lab 80 Plus system. Standard photolithography and wet etching techniques were used to create a striped pattern in the silicon dioxide layer along the GaN h1 1¯ 0 0i direction as shown in Fig. 1a. Both the window opening and the stripe width were 5 mm. Two similarly sized pieces, each approximately 1/6 of a 2 in. wafer, were placed side by side in a line normal to the incoming gas flow in the reactor for re-growth. One sample was placed with the striped pattern oriented parallel to the incoming gas flow and the other with the striped pattern oriented normal to the incoming gas flow as shown in Fig. 1b. In roughly half of the growths, the samples with the striped pattern oriented parallel to the incoming gas flow were located on the left-hand side of the incoming gas, as shown in Fig. 1b. In the other growths, the samples with the striped pattern oriented parallel to the incoming gas flow were located on the right-hand side. This configuration was used in order to simultaneously fit two samples on the 2 in. susceptor and ensure similar growth conditions by eliminating the effects of uneven gas flow and substrate heating. The samples were grown with no rotation at 1115 1C, 200 mbar, and a V/III ratio around 1400. The ammonia and the carrier gas flow rates were around 1.2 and 4.0 slm, respectively. Table 1 summarizes the various growth parameters during the different stages of SAG.
Fig. 1. (a) Striped pattern along GaN h1 1¯ 0 0idirection; (b) Samples oriented parallel and normal to the incoming gas flow during SAG of GaN.
ARTICLE IN PRESS L.E. Rodak et al. / Journal of Crystal Growth 306 (2007) 75–79
77
Table 1 Growth conditions for selective area growth of GaN
Nucleation layer Main layer Step 1-base structure Step 2-SAG
Growth time (min)
Temperature (1C)
TMAl (mmol/min)
TMGa (mmol/min)
V/III
10 45 30 15–160
1125 1125 1115 1115
12 — — —
— 100 38 38
5500 700 1400 1400
The cross sections of the resulting three-dimensional structures were observed using a Hitachi S4700 scanning electron microscope (SEM) and conventional image processing techniques were used to determine the size of the structures. For practical reasons, one piece per growth per orientation was evaluated and each piece characterized was about one quarter of the size of the grown sample. In order to eliminate experimental artifacts in the analysis of the results, such as temperature nonuniformity, images and measurements were taken from areas of the two samples that were closest during growth. Any variations in the samples are summarized by the error bars shown in each figure. Occasionally, pieces from other parts of the sample, both off center and further downstream, were evaluated. Based on SEM images, the growth results were consistent across the samples and fall within the shown error bars.
Fig. 2. SEM image of SAG base structure. Structure was identical for samples with stripes oriented normal or parallel to incoming gas flow.
3. Results and discussion Figs. 2 and 3 show the SEM images of the cross section for the samples with stripes oriented parallel and normal to the incoming gas flow. After 30 min of growth both orientations exhibited similar growth modes resulting in the same trapezoidal structure with the (0 0 0 1), f1 1 2¯ 2g, and f1 1 2¯ 0g facets exposed shown in Fig. 2. Therefore, all subsequent growths were performed on this base structure and the growth rates were measured as an increase in size from this base structure. It should be noted that no asymmetry in the structures was observed for either orientation as might be expected. Throughout later growth stages, both samples maintained a similar trapezoidal cross section with (0 0 0 1) and f1 1 2¯ ng facets (n 1:7 2:2) exposed. The samples oriented normal to the flow experienced larger growth rates, as is indicated in Figs. 2–5. Fig. 4 illustrates the height as a function of growth time. Clearly, the normally oriented sample experiences a larger vertical growth rate when compared to the parallel oriented sample. Fig. 5 illustrates the crosssectional area of the three-dimensional structures per pattern period. Once again, the normally oriented sample is displaying an enhanced cross-sectional growth rate. In both Figs. 4 and 5, the line is a linear fit using the error bars as weight. Also shown in Figs. 4 and 5 is the height and cross section of the structures when SAG is carried out for 120 min while rotating the sample. The height of the rotating sample is similar to that of the normal oriented
sample, while the cross section is similar to the parallel oriented sample. Since the growth conditions are identical for both orientations, the increased cross-sectional growth rate of the normal oriented sample indicates that more species are being incorporated into the growth. Both orientations maintain exposed (0 0 0 1) and f1 1 2¯ ng facets (n 1:7 2:2), indicating that there is no difference in surface growth kinetics between the samples. In MOVPE, it is assumed that the gallium species sticking coefficient is unity on the GaN surface [10] and there is efficient lateral transport across the oxide mask allowing all reactants reaching the surface to be incorporated into the SAG growth [9,14,15]. Therefore, it can be concluded that there is a difference in mass transport between the normal and parallel oriented sample to account for the difference in growth rate. The normal orientation experienced enhanced growth rate because more species were diffusing to the growth site. The size of the structures is negligible when compared to the boundary layer thickness, which is proportional to the square root of the kinematic viscosity of the gas and the distance from the edge of the susceptor and inversely proportional to the square root of the free stream gas velocity [16,17]. Therefore, it can be assumed that the height of the structure does not have a significant impact on the boundary layer, which is consistent with the fact that both orientations maintain constant growth rates even as the structures grow taller. These findings are
ARTICLE IN PRESS 78
L.E. Rodak et al. / Journal of Crystal Growth 306 (2007) 75–79
Fig. 3. SEM images of SAG selective area growth of GaN with h1 1¯ 0 0i striped pattern oriented parallel and normal to the incoming gas flow.
Fig. 4. Height as a function of growth time for parallel and normal oriented samples. The lines are a linear fit of the data using the error bars as weight. The y-axis intercept corresponds to the height of the base structure.
Fig. 5. Cross section as a function of time for normal and parallel oriented samples. The lines are a linear fit of the data using the error bars as weight. The y-axis intercept corresponds to the cross-sectional area of the base structure.
contrary to the results obtained when the effect of gas flow was investigated on cubic GaN [18]. Cubic GaN was reported to demonstrate only a slight dependence of the cross-sectional geometry and growth rates on the orientation of the pattern with respect to the incoming gas flow. Future work is needed to further explain the mechanism responsible for dependence of growth rates on the orientation of the striped pattern with respect to the incoming gas flow.
4. Conclusions In this work, we investigated the effect of gas flow on the SAG of GaN on a dielectric striped pattern by placing the pattern parallel and normal to the incoming gas flow. Our results show that there is a strong dependence of growth rates on the orientation of the pattern and therefore on the gas flow. Initially, the parallel and normal orientations exhibit similar growth rates and for the same structures.
ARTICLE IN PRESS L.E. Rodak et al. / Journal of Crystal Growth 306 (2007) 75–79
However, in later growth stages the samples with the pattern oriented normal to the incoming gas flow showed enhanced vertical and cross-sectional growth rates. While current models have accounted for the effect of various growth parameters including temperature, pressure, V/III ratio, and fill factor, most have assumed the samples were rotating during growth. This work indicates that gas flow plays an important role in the SAG and therefore the rotation of the sample needs to be incorporated into a complete physical model. Acknowledgments This work was supported in part by a grant from the West Virginia Graduate Student Fellowship in Science, Technology, Engineering and Math (STEM) program to L.E.R. and by the National Science Foundation (Grant EPS-0554328), The State of West Virginia Research Challenge Fund, and the West Virginia University Research Corporation. Partial funding was also provided by AIXTRON Inc. References [1] P. Gibart, Rep. Prog. Phys. 67 (2004) 667. [2] B. Beaumont, Ph. Venne´gue´s, P. Gibart, Phys. Status Solidi (b) 227 (1) (2001) 1.
79
[3] S.J. Rosner, E.C. Carr, M.J. Ludowise, G. Girolami, H.I. Erikson, Appl. Phys. Lett. 70 (1997) 4. [4] T.S. Zheleva, S.A. Smith, D.B. Thomson, T. Gehrke, K.J. Linthicum, P. Rajagopal, E. Carlson, W.M. Ashmawi, R.F. Davis, MRS Internet J. Nitride Semicond. Res. 4S1 (1999) G3.38. [5] P. Vennegues, B. Beaumont, V. Bousquet, M. Vaille, P. Gibart, J. Appl. Phys. 87 (2000) 4175. [6] T. Tanaka, K. Uchida, A. Watanabe, S. Minagawa, Appl. Phys. Lett. 68 (1996) 7. [7] K. Hiramatsu, J. Phys.: Condens. Matter 13 (2001) 6961. [8] K. Hiramatsu, K. Nishiyama, A. Motogaito, H. Miyake, Y. Iyechika, T. Maeda, Phys. Status Solidi (a) 176 (1999) 535. [9] H. Marchand, J.P. Ibbetson, P.T. Fini, S. Keller, S.P. DenBaars, J.S. Speck, U.K. Mishra, J. Crystal Growth 195 (1998) 328. [10] G. Feng, Y. Fu, J.S. Xia, J.J. Zhu, X.M. Shen, D.G. Zhao, H. Yang, J.W. Liang, J. Phys. D: Appl. Phys. 35 (2002) 2731. [11] X. Zhang, P.D. Dapkus, D.H. Rich, Appl. Phys. Lett. 77 (2000) 10. [12] L.E. Rodak, N.J. Berry Ann, K. Kasarla, Y. Yang, Mat. Res. Soc. Symp. Proc. 892 (2006) 665. [13] N.J. Berry Ann, L.E. Rodak, K. Kasarla, Y. Yang, Proc. SPIE 6017 (2005) 60170. [14] K. Hiramatsu, H. Miyake, Mat. Res. Soc. Symp. Proc. 639 (2001). [15] M.E. Coltrin, C.C. Mitchel, III-Nitride Based Semiconductor Electronic and Optical Devices and State-of-the-Art Program on Compound Semiconductors XXXIV, in: F. Ren, D. N. Buckley, S.N.G. Chu, S.J. Pearton (Eds) The Electrochemical Society Proceedings Series, Pennington, NJ, 2001, PV 2001-01, 1p. [16] G.B. Stringfellow, Organometallic Vapor-Phase Epitaxy, Academic Press, San Diego, CA, 1999, p. 303. [17] K. Harafuji, Jpn. J. Appl. Phys. 40 (2001) 6263. [18] X.M. Shen, G. Feng, B.S. Zhang, L.H. Duan, Y.T. Wang, H. Yang, J. Crystal Growth 252 (2003) 9.