Effect of flux composition ratio on the coalescence growth of GaN crystals by the Na-flux method

Effect of flux composition ratio on the coalescence growth of GaN crystals by the Na-flux method

Optical Materials xxx (2016) 1e4 Contents lists available at ScienceDirect Optical Materials journal homepage: www.elsevier.com/locate/optmat Effec...

801KB Sizes 2 Downloads 13 Views

Recommend Documents

No documents
Optical Materials xxx (2016) 1e4

Contents lists available at ScienceDirect

Optical Materials journal homepage: www.elsevier.com/locate/optmat

Effect of flux composition ratio on the coalescence growth of GaN crystals by the Na-flux method Masatomo Honjo*, Masayuki Imanishi, Hiroki Imabayashi, Kosuke Nakamura, Kosuke Murakami, Daisuke Matsuo, Mihoko Maruyama, Mamoru Imade, Masashi Yoshimura, Yusuke Mori Division of Electrical, Electronic and Information Engineering, Graduate School of Engineering, Osaka University, 2-1 Yamadaoka, Suita, Osaka, 565-0871, Japan

a r t i c l e i n f o

a b s t r a c t

Article history: Received 29 April 2016 Received in revised form 4 September 2016 Accepted 6 September 2016 Available online xxx

Previously, we demonstrated that the Na-flux coalescence growth technique had high potential for the fabrication of large-diameter, high-quality GaN crystals. This present study investigates the relation between the flux composition (Ga/Na) and void formation in GaN crystals grown by this technique. It was found that void formation decreases with a decrease in the Ga composition of the flux and that stable coalescence with no voids at the GaN grain boundaries occurred for a Ga composition of 15 mol%. Bandedge emission peaks were clearly observed for a crystal grown at 15 mol% Ga composition, while other peaks were hardly observed. © 2016 Elsevier B.V. All rights reserved.

Keywords: Gallium nitride Coalescence growth Flux composition Void formation

1. Introduction Large-diameter, high-quality GaN wafers are highly demanded for low-cost and high-performance GaN-based optical and electronic devices such as ultraviolet light-emitting diodes, laser diodes, and high-power and high-frequency transistors [1e4]. GaN bulk crystals are conventionally grown by various methods; hydride vapor phase epitaxy (HVPE) [5,6], high-pressure solution growth [7], ammonothermal growth [8e11], and Na-flux growth [12e21]. HVPE is advantageous for its high growth rate (over 100 mm/h), and GaN substrates produced by this method are now commercially available. However, because this technique fabricates the GaN layers on lattice-mismatched substrates such as sapphire, the high threading-dislocation density (at least 106 cm 2) degrades their efficiency and reliability in optoelectronic devices. In addition, the thermal expansion coefficients differ between the GaN crystals and their underlying substrates, causing severe bowing and bending [6]. In the past decade, we have attempted to produce high-quality GaN crystals by the Na-flux method. In our previous study, we grew

GaN crystals with low dislocation density from a tiny GaN seed known as a “point seed” [18]. In addition, we showed that the coalescence growth technique, in which GaN crystals grow and coalesce on the GaN point seeds, can potentially fabricate largediameter, high-quality GaN wafers [19,20]. This growth technique requires the stable coalescence of GaN crystals to prevent cracking of the grown crystals during post-process cooling and handling. Therefore, the boundaries of the coalesced crystals should contain no voids, but this ideal situation is not easy to achieve in practice. The growth morphology of GaN crystals grown by the Na-flux method is known to depend on the composition ratio of the Ga/ Na flux. Yamane et al. reported that changing the flux composition ratio alters the spontaneous nucleation growth behavior [12e14]. Imade et al. reported that the growth habit and growth mode depends on the flux composition ratio in the growth on GaN seeds [16,17]. In this study, we investigated how the flux composition ratio affects the void formation in GaN coalescence growth by the Na-flux method.

2. Experimental procedure * Corresponding author. E-mail address: [email protected] (M. Honjo).

GaN crystals were grown to coalesce on 1-cm2 multipoint-seedGaN substrates (MPS-GaN sub.) in which GaN point seeds arranged

http://dx.doi.org/10.1016/j.optmat.2016.09.017 0925-3467/© 2016 Elsevier B.V. All rights reserved.

Please cite this article in press as: M. Honjo, et al., Effect of flux composition ratio on the coalescence growth of GaN crystals by the Na-flux method, Optical Materials (2016), http://dx.doi.org/10.1016/j.optmat.2016.09.017

2

M. Honjo et al. / Optical Materials xxx (2016) 1e4

in sapphire substrates. The MPS-GaN sub. was produced by patterning a 30-mm-thick c-plane GaN layer grown on a c-plane sapphire substrate by hydride vapor-phase epitaxy. The GaN point seeds were hexagonally arrayed such that coalescence proceeded in the m-direction of the GaN. Each point seed was 250 mm in diameter, and the centers of neighboring points seeds were separated by 550 mm as shown in Fig. 1(b). Coalescence growth on the MPS-GaN sub. was performed as follows. Within an Ar-filled glove box, MPS-GaN sub. was placed in a ceramic crucible (with a diameter and height of 17 and 25 mm, respectively), followed by the starting materials of metallic Ga (purity: 6 N), metallic Na (purity: 4 N), and granular graphite (purity: 6 N). The granular graphite suppressed the growth of polycrystalline material on the crucible wall [15]. The crucible was then transferred to a stainless-steel tube and placed in an electric furnace. Within the furnace, the tube was heated to its growth temperature (870  C) in a 1-h period. Next, the tube was filled with N2 gas until its internal pressure reached 3.6 MPa. During the growth period (72e96 h), the temperature and N2 pressure in the tube were maintained at 870  C and 3.6 MPa, respectively. The stainlesssteel tube was unmoved throughout the growth, removing the effects of solution stirring. After natural cooling of the stainless-steel tube, the crucible was removed and immersed in cold ethanol to dissolve the residual flux. Once the residual flux had completely dissolved, the grown GaN crystals were removed from the crucible. The growth method is detailed in our previous report [21]. The grown GaN crystals were separated from the MPS-GaN sub., and their void formations were investigated after polishing their cand a-faces, as shown in Fig. 1(a). The polished a-faces were located at the centers of the GaN point seeds. The as-grown and polished surfaces of the grown crystals were observed by scanning electron microscopy (SEM) (VHX-D510, KEYENCE, Japan). The area ratio of voids on the surface covered by the polished c-face (here called the void ratio) was calculated from surface SEM images at growth thicknesses of 500 and 700 mm. The photoluminescence (PL) spectra of the grown crystals were measured in the polished c-face at the growth thickness of 1 mm using a HeeCd laser (325 nm, 20 mW) as the extinction light source at room temperature. For the PL measurement, the crystal surfaces were treated with chemical mechanical polishing after the mechanical polishing had been done.

3. Results and discussion Fig. 2 shows the as-grown and polished-surface SEM images of the crystals grown at various Ga compositions (15 mol%, 27 mol%, and 40 mol%). The polished c-faces were located at a growth thickness of 500 mm in Fig. 2. The crystals in Fig. 2 (a), (b), and (c) were grown to an approximate thickness of 1.3 mm. The average growth rates along the c-direction for the growths at 15, 27, and 40 mol% Ga compositions were approximately 13.4, 17.4, and 13.0 mm/h, respectively. The average growth rate was obtained by dividing the growth thickness by the growth period. We note there is a possibility that the growth almost stopped during the middle or latter growth period in the growth at 15 mol% Ga composition because of the depletion of Ga source. The GaN grains grown on the GaN point seeds were mainly faceted by their c-faces and inclined {10e11}-faces. In the surface SEM images of the as-grown surfaces, hexagonal pyramidal grains are observed at 15 and 27 mol% Ga composition (see Fig. 2 (a) and (b)), whereas inhomogeneous grains appear at 40 mol% Ga composition (Fig. 2(c)). At 15 mol% Ga composition, the as-grown surface, polished cface and polished a-face were all free of voids (see Fig. 2 (a), (d) and (g)). Under this growth condition, the GaN grains were tightly coalesced. Conversely, in the crystals grown at 27 and 40 mol% Ga composition, voids appeared on the as-grown surface and on both polished faces. At 27 mol% Ga composition, the voids were concentrated at the grain boundaries. Although the hexagonal GaN grains were evenly sized, they were almost separated by voids on the as-grown surface and polished c-face (see Fig. 2 (b) and (e)). Voids were sparse at the grain boundaries in the initial growth region but became prominent in the middle of the growth stage and extended in the c-direction, as shown in Fig. 2 (h). This indicates that although the grains grew to similar size on each GaN point seed, this growth condition induced numerous voids at the grain boundaries during most of the growth stage. The crystals grown at 40 mol% Ga composition were separated not only at their grain boundaries but also in the grain surfaces (see Fig. 2 (c)). Moreover, the grains on the as-grown surface overlapped under this growth condition. Many inhomogeneous voids appeared on the polished c-face and a-face, as shown in panels (f) and (i) of Fig. 2. To quantify the void formation in the grown crystals, we calculated the void ratio on the polished c-faces at growth

Fig. 1. Schematics of (a) GaN coalescence growth and polishing on the c- and a-faces, and (b) the GaN point-seed arrangement.

Please cite this article in press as: M. Honjo, et al., Effect of flux composition ratio on the coalescence growth of GaN crystals by the Na-flux method, Optical Materials (2016), http://dx.doi.org/10.1016/j.optmat.2016.09.017

M. Honjo et al. / Optical Materials xxx (2016) 1e4

3

Fig. 2. Surface SEM images of the as-grown, polished c-faces and polished a-faces of crystals grown at Ga compositions of 15 mol% (left), 27 mol% (center), and 40 mol% (right). The crystals grown at 15 mol% Ga were almost void-free, whereas those grown at 27 and 40 mol% Ga were separated by many voids.

thicknesses of 500 and 700 mm. Fig. 3 plots the relationship between the flux composition and the void ratio in the polished cfaces of the grown crystals. At both growth thicknesses, the void ratio decreased with decreasing Ga composition of the flux. Moreover, the average void ratio of crystals grown at Ga compositions of 27 and 40 mol% was smaller at 700-mm growth thickness than at 500 mm. This indicates that the void ratio decreases as the growth thickness increases above 500 mm. A previous study reported that the nitrogen solubility of the flux increases with decreasing Ga composition of the flux, both experimentally and in first-principle calculations [22]. Voids in the crystals grown at 40 mol% Ga composition likely result from the poor nitrogen source in the flux. Moreover, supersaturation

Fig. 3. Relationship between flux composition and void ratio on the polished c-faces of crystals grown to thicknesses of 500 and 700 mm.

homogeneity in the flux would be compromised between neighboring grains under this growth condition. Thus, the grains will easily overlap and many inhomogeneous voids will form in the crystal. Reducing the Ga composition to 15 or 27 mol% improves the flux homogeneity, forming homogeneous hexagonal grains on each GaN point seed. Under these growth conditions, the c-face occupies a smaller area than the point seeds, as shown in Fig. 2 (a) and (b), indicating that the grain habit changes from plate-like shape to pyramidal during the growth. The pyramidal grains formed at 27 mol% Ga composition will more readily form boundary voids than plate-like grains. On the other hand, the GaN grains completely coalesced at 15 mol% Ga composition despite their habit of forming pyramidal grains. In Fick's law, the diffusion velocity of a solute molecule is proportional to the concentration gradient of the molecules: in this scenario, this means that a higher nitrogen concentration in the flux is expected to improve the supply of nitrogen to the grain boundaries where the nitrogen source is poor. Kawahara et al. reported that the nitrogen solubility of the flux increases by more than one order of magnitude from 27 to 10 mol% Ga composition [22]. Thus, void formation would be suppressed at low Ga compositions when there is a high nitrogen concentration in the flux. In the present experimental system, the Ga composition of the flux gradually decreased as the Ga was consumed during growth. This phenomenon explains the decreasing void ratio in the polished c-face as the growth thickness increased under Ga compositions of 27 and 40 mol% (see Fig. 3). Fig. 4 shows the PL spectra of the crystal grown at Ga composition of 15 mol%. The spectra were measured for the c-face growth region, {10e11}-face growth region, and around the grain boundary. While the band-edge emission peaks of the GaN crystal were clearly observed, other peaks were hardly observed in the evaluation regions. It was confirmed that the PL spectrum observed around the grain boundary was nearly same as that for the {10e11}-

Please cite this article in press as: M. Honjo, et al., Effect of flux composition ratio on the coalescence growth of GaN crystals by the Na-flux method, Optical Materials (2016), http://dx.doi.org/10.1016/j.optmat.2016.09.017

4

M. Honjo et al. / Optical Materials xxx (2016) 1e4

[3]

[4] [5] [6]

[7]

[8] [9] Fig. 4. Photoluminescence spectra in the polished c-face of the crystal grown at Ga composition of 15 mol%.

[10] [11]

face growth region. The intensity of the band-edge emission peak in the {10e11}-face growth region and around the grain boundary is more than double of that observed for the c-face growth region. The intensity of the band-edge emission depends on the carrier concentration of GaN crystals [23]. Moreover, it has previously been reported that the incorporation of impurities that behave as donors is something that strongly depends on the growth face [24,25]. Thus, the difference in the emission intensity is something that would be attributed to the differences in carrier concentrations brought about by the difference in impurity concentrations. 4. Conclusions We investigated the dependence of the flux composition ratio on the void formation during GaN coalescence growth. The void formation in the grown crystals was an increasing function of the Ga composition of the flux. Moreover, the grown GaN grains coalesced with no voids at a Ga composition of 15 mol%, and band-edge emission peaks were clearly observed not just for the GaN grains but also around the grain boundaries. We conclude that low Ga composition is advantageous for fabricating void-free GaN crystals by coalescence growth using the Na-flux method. Acknowledgement This study was partly supported by The Japan Science and Technology Agency (Project No. J121052565), Ministry of the Environment, Japan (Project No. J141057005) and Panasonic Corporation. We thank E. Sawai (Frontier Alliance, LLC) for polishing the samples. References [1] S. Nakamura, T. Mukai, M. Senoh, High-power GaN P-N junction blue-lightemitting diodes, Jpn. Appl. Phys. 30 (1991) L1998eL2001. [2] W. Saito, Y. Takeda, M. Kuraguchi, K. Tsuda, I. Omura, T. Ogura, H. Ohashi, High

[12]

[13]

[14]

[15]

[16]

[17] [18]

[19] [20]

[21]

[22]

[23]

[24]

[25]

breakdown voltage AlGaNeGaN power-HEMT design and high current density switching behavior, IEEE Trans. Electron Devices 50 (2003) 2528e2531. S. Tomiya, H. Nakajima, K. Funato, T. Miyajima, K. Kobayashi, T. Hino, S. Kijima, T. Asano, M. Ikeda, Dislocations in GaN-based laser diodes on epitaxial lateral overgrown GaN layers, Phys. Status Solidi A 188 (2001) 69e72. T. Paskova, K.R. Evans, GaN substratesdprogress, status, and prospects, IEEE J. Sel. Top. Quantum Electron 15 (2009) 1041e1052. K. Fujito, S. Kubo, H. Nagaoka, T. Mochizuki, H. Namita, S. Nagao, Bulk GaN crystals grown by HVPE, J. Cryst. Growth 311 (2009) 3011e3014. K.M. Chen, H.H. Huang, Y.L. Kuo, P.L. Wu, T.L. Chu, H.W. Yu, W.I. Lee, Effective reduction of bowing in free-standing GaN by N-face regrowth with hydride vapor-phase epitaxy, J. Cryst. Growth 311 (2009) 3037e3039. M. Bockowski, P. Strak, I. Grzegory, B. Lucznik, S. Porowski, GaN crystallization by the high-pressure solution growth method on HVPE bulk seed, J. Cryst. Growth 310 (2008) 3924e3933. T. Fukuda, D. Ehrentraut, Prospects for the ammonothermal growth of large GaN crystal, J. Cryst. Growth 305 (2007) 304e310. T. Hashimoto, F. Wu, M. Saito, K. Fujito, J.S. Speck, S. Nakamura, Status and perspectives of the ammonothermal growth of GaN substrates, J. Cryst. Growth 310 (2008) 876e880. T. Hashimoto, F. Wu, J.S. Speck, S. Nakamura, Ammonothermal growth of bulk GaN, J. Cryst. Growth 310 (2008) 3907e3910. R. Dwilinski, R. Doradzinski, J. Garczynski, L.P. Sierzputowski, A. Puchalski, Y. Kanbara, K. Yagi, H. Minakuchi, H. Hayashi, Bulk ammonothermal GaN, J. Cryst. Growth 311 (2009) 3015e3018. H. Yamane, M. Shimada, T. Sekiguchi, F.J. Disalvo, Morphology and characterization of GaN single crystals grown in a Na flux, J. Cryst. Growth 186 (1998) 8e12. M. Aoki, H. Yamane, M. Shimada, S. Sarayama, F.J. Disalvo, Growth of 5 mm GaN single crystals at 750  C from an NaeGa melt, Crsyt. Growth Des. 1 (2001) 119e122. H. Yamane, M. Aoki, T. Yamada, M. Shimada, H. Goto, T. Goto, H. Makino, T. Yao, S. Sarayama, H. Iwata, F.J. Disalvo, Time dependence of the growth morphology of GaN single crystals prepared in a NaeGa melt, Jpn. J. Appl. Phys. 44 (2005) 3157e3160. F. Kawamura, M. Morishita, M. Tanpo, M. Imade, M. Yoshimura, Y. Kitaoka, Y. Mori, T. Sasaki, Effect of carbon additive on increases in the growth rate of 2 in GaN single crystals in the Na flux method, J. Cryst. Growth 310 (2008) 3946e3949. M. Imade, Y. Hirabayashi, Y. Konishi, H. Ukegawa, N. Miyoshi, M. Yoshimura, T. Sasaki, Y. Kitaoka, Y. Mori, Growth of large GaN single crystals on highquality GaN seed by carbon-added Na flux method, Appl. Phys. Express 3 (2010) 075501. M. Imade, Y. Hirabayashi, N. Miyoshi, M. Yoshimura, Y. Kitaoka, T. Sasaki, Y. Mori, Cryst. Growth Des. 11 (2011) 2346e2350. M. Imanishi, Y. Todoroki, K. Murakami, T. Yamada, D. Matsuo, H. Imabayashi, H. Takazawa, M. Maruyama, M. Imade, M. Yoshimura, Y. Mori, Dramatic reduction of dislocations on a GaN point seed crystal by coalescence of bunched steps during Na-flux growth, J. Cryst. Growth 427 (2015) 87e93. M. Imade, M. Maruyama, M. Yoshimura, Y. Mori, Growth of bulk GaN crystals by the Na-flux point seed technique, Jpn. J. Appl. Phys. 53 (2014) 05FA06. M. Imade, M. Imanishi, Y. Todoroki, H. Imabayashi, D. Matsuo, K. Murakami, H. Takazawa, A. Kitamoto, M. Maruyama, M. Yoshimura, Y. Mori, Fabrication of low-curvature 2 in. GaN wafers by Na-flux coalescence growth technique, Appl. Phys. Express 7 (2014) 035503. F. Kawamura, M. Morishita, N. Miyoshi, M. Imade, M. Yoshimura, Y. Kitaoka, Y. Mori, T. Sasaki, Study of the metastable region in the growth of GaN using the Na flux method, J. Cryst. Growth 331 (2009) 4647e4651. M. Kawahara, F. Kawamura, M. Yoshimura, Y. Mori, T. Sasaki, S. Yanagisawa, Y. Morikawa, A first-principles investigation on the mechanism of nitrogen dissolution in the Na flux method, J. Appl. Phys. 101 (2007) 066106. R.D. Cunningham, R.W. Brander, N.D. Knee, D.K. Wickenden, Variation of photoluminescence with carrier concentration in GaN, J. Lumin. 5 (1972) 21e31. S.R. Xu, Y. Hao, J.C. Zhang, Y.R. Cao, X.W. Zhou, L.A. Yang, X.X. Ou, K. Chen, W. Mao, Polar dependence of impurity incorporation and yellow luminescence in GaN films grown by metal-organic chemical vapor deposition, J. Cryst. Growth (2010) 3521e3524. C. Wetzel, H. Amano, I. Akasaki, J.W. Ager III, I. Grzegory, B.K. Meyer, DX-like behavior of oxygen in GaN, Phys. B 302e303 (2001) 23e38.

Please cite this article in press as: M. Honjo, et al., Effect of flux composition ratio on the coalescence growth of GaN crystals by the Na-flux method, Optical Materials (2016), http://dx.doi.org/10.1016/j.optmat.2016.09.017