Preparation of silicon carbide nanowires via a rapid heating process

Preparation of silicon carbide nanowires via a rapid heating process

Materials Science and Engineering B 176 (2011) 87–91 Contents lists available at ScienceDirect Materials Science and Engineering B journal homepage:...

803KB Sizes 0 Downloads 68 Views

Materials Science and Engineering B 176 (2011) 87–91

Contents lists available at ScienceDirect

Materials Science and Engineering B journal homepage: www.elsevier.com/locate/mseb

Preparation of silicon carbide nanowires via a rapid heating process Xintong Li, Xiaohong Chen, Huaihe Song ∗ State Key Laboratory of Chemical Resource Engineering, Beijing University of Chemical Technology, 100029, Beijing, PR China

a r t i c l e

i n f o

Article history: Received 4 May 2010 Received in revised form 22 July 2010 Accepted 10 September 2010 Keywords: Silicon carbide Nanowire Carbothermal reduction Rapid heating process Aerogel

a b s t r a c t Silicon carbide (SiC) nanowires were fabricated in a large quantity by a rapid heating carbothermal reduction of a novel resorcinol-formaldehyde (RF)/SiO2 hybrid aerogel in this study. SiC nanowires were grown at 1500 ◦ C for 2 h in an argon atmosphere without any catalyst via vapor–solid (V–S) process. The ␤-SiC nanowires were characterized by field-emission scanning electron microscope (FE-SEM), X-ray diffraction (XRD), transmission electron microscope (TEM), high-resolution transmission electron microscope (HRTEM) equipped with energy dispersive X-ray (EDX) facility, Fourier transformed infrared spectroscopy (FTIR), and thermogravimetric analysis (TGA). The analysis results show that the aspect ratio of the SiC nanowires via the rapid heating process is much larger than that of the sample produced via gradual heating process. The SiC nanowires are single crystalline ␤-SiC phase with diameters of about 20–80 nm and lengths of about several tens of micrometers, growing along the [1 1 1] direction with a fringe spacing of 0.25 nm. The role of the interpenetrating network of RF/SiO2 hybrid aerogel in the carbothermal reduction was discussed and the possible growth mechanism of the nanowires is analyzed. © 2010 Elsevier B.V. All rights reserved.

1. Introduction Recently, one-dimensional (1D) nanostructures have become one of the most important research focuses because of their potential applications in mesoscopic research and nanostructured composite materials [1]. Silicon carbide (SiC) is an ideal candidate for 1D materials owing to its unique virtues. In recent decades, the synthesis, formation mechanism and properties of SiC continue to attract growing attention due to its excellent electrical mechanical and chemical properties, such as high electric field at breakdown (2 × 106 V/cm), high electron velocity (2 × 107 cm/s), large band gap (2.3 eV for cubic 3C-SiC (␤-SiC) and 3.2 eV for the hexagonal 4H polytype at room temperature), high thermal stability and high thermal conductivity (ca. 400 W/(K m)) [2]. SiC materials in 1D form possess shape-induced unique electrical and optical properties, as well as better elasticity and strength than those of the bulk SiC. Due to these qualities, SiC is considered as a kind of promising materials for applications in biomaterials, high-temperature semi-conducting devices, light weight/high strength structure and catalysis fields, especially for use in harsh environments [3]. To date, 1D SiC nanostructures with various shapes, including SiC nanobelts [4], nanorods [5–7], nanowires and nanotubes [1,8–19], have been successfully synthesized. A widely used

method to synthesize 1D SiC is carbothermal reduction because of its simpleness and convenience, especially the broad availability of precursors containing Si and carbon [20]. Since the vapor–liquid–solid (VLS) mechanism was proposed by Wagner et al. [21] in 1964, some metals (e.g., Fe, Mg and Na) have been used as the catalysts to improve the synthesis amount of SiC [2,8,9,14]. Yang et al. [22] reported the formation of SiC nanorods by rapid thermal processing of thin films containing a mixture of carbon and iron over silicon wafers. However, complicated equipment, a long control process, and special conditions (e.g., high pressure) are still needed for these methods. It is necessary to explore a simpler and more effective way to prepare 1D-SiC nanostructures for further practical applications. In this work, we demonstrate a new simple but effective synthesis route, rapid heating process, to prepare well-crystallized SiC nanowires at 1500 ◦ C. Compared to above mentioned routes, it requires a relative short time, simple equipment and manipulation. In this system, a novel resorcinol-formaldehyde (RF)/SiO2 hybrid aerogel was used as the precursor, and no catalysts were used. 2. Experimental section 2.1. Preparation of RF/SiO2 hybrid aerogel

∗ Corresponding author at: P.O. Box 34, College of Materials Science and Engineering, Beijing University of Chemical Technology, Beijing, 100029, PR China. Tel.: +86 10 64434916; fax: +86 10 64434916. E-mail address: [email protected] (H. Song). 0921-5107/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.mseb.2010.09.007

All the reagents used in this research are analytical grade. The precursor, RF/SiO2 hybrid aerogel, according to the procedures reported in our previous work [23], was fabricated from RF sol and SiO2 sol by the co-polymerization and supercritical drying pro-

88

X. Li et al. / Materials Science and Engineering B 176 (2011) 87–91

Fig. 1. FE-SEM images of the purified SiC samples prepared at 1500 ◦ C for 2 h: (a) using rapid heating process and (b) at a heating rate of 5 ◦ C/min.

cess. Firstly, the RF sol and SiO2 sol were prepared through the traditional methods [24,25], respectively. Then, the hybrid sol was formed by blending the prepared RF sol and SiO2 sol under magnetic stirring at room temperature. The dosages of the two kinds of sols were adjusted to a C/Si atomic ratio = 3. The mixed sol was injected into some ampoules and then cured in a water bath at 358 K for 72 h to form the wet hybrid gel. The obtained gel was immersed in acetone (wet gel:acetone = 3–4, v/v) to exchange the containing water. Finally, the obtained hybrid gel was dried by supercritical drying using petroleum ether as the extraction solvent (Tc = 250 ◦ C, Pc = 7 MPa) for 2 h. Consequently, the dark red, transparent and monolithic RF/SiO2 hybrid aerogel with a bulk density of 0.146 g/cm3 was successfully obtained. The resulting gel was milled into powder for the convenience of SiC synthesis.

Fourier transform (FFT) patterns and energy dispersive X-ray (EDX) patterns were obtained with a JEOL JEM-2010 microscope equipped with EDX facilities. The samples for TEM and HRTEM observations were prepared by dispersing the products in ethanol with an ultrasonic bath for 10 min and a few drops of resulting suspension were placed on the copper grids. Fourier transform infrared absorption (FTIR) spectra were recorded on samples embedded in KBr pellets with a NICOLET NEXUS 870 FTIR spectrophotometer. Thermogravimetric analysis (TGA) of the as-synthesized SiC sample was conducted on a NETZSCH STA449C simultaneous thermal apparatus at a heating rate of 10 ◦ C/min to 1000 ◦ C from room temperature under a flow of air.

2.2. Heat treatment and purification

The as-purified SiC sample prepared using rapid heating process is fluffy and light-green. The total yield of the obtained SiC sample was 79.9% (defined as the ratio of the amount of SiC formed to the theoretical amount calculated from complete conversion of carbon according to reaction 3C + SiO2 → SiC + 2CO). A typical FESEM image shown in Fig. 1a indicates that the morphologies of the sample are almost wire-like structures with diameters in the range of 20–80 nm and lengths of several tens of micrometers, implying a large aspect ratio. The percentage of nanowires is estimated to be over 90% according to the FE-SEM observation. These wires are randomly oriented with straight or a little curved morphologies. For the tens of wires observed, no spherical caps at the tips of the wires, which is a typical feature of VLS mechanism [26], were

First, a GSL-1600X horizontal tube furnace was pumped out and fed back with high-purity Ar gas. Second, the system was heated to 1500 ◦ C in a flowing Ar gas with a rate of 250 ml/min. Then the furnace was opened with care and an alumina boat loading a certain amount of the as-prepared RF/SiO2 powders was slowly pushed into the hot zone of the ceramic tube furnace. After the furnace was sealed again, the system was maintained at 1500 ◦ C for 2 h. After the furnace was cooled down to 200 ◦ C at a rate of 5 ◦ C/min and then to room temperature naturally, the white-gray product was taken out. The as-synthesized sample was calcined in air at 600 ◦ C for 5 h and washed with a dilute mixture of nitric acid (HNO3 ) and hydrofluoric acid (HF) (HNO3 /HF = 1:3, v/v) for 72 h to eliminate the residual carbon and silica, respectively. After washing with deionized water and drying at 80 ◦ C, the final purified SiC product was obtained. Additionally, for comparison, the RF/SiO2 aerogel with the same C/Si atomic ratio as above were also carbonized using the common heating procedure, i.e., the sample was heated to 1500 ◦ C from room temperature at a heating rate of 5 ◦ C/min in the presence of argon gas and maintained at 1500 ◦ C for 2 h. Then by the same purification mentioned above, the purified sample was obtained.

3. Results and discussion

2.3. Characterization The X-ray diffraction (XRD) measurement was carried out at a Rigaku D/max-2500B2+/PCX system operated at a step size of ˚ over the range of 20–80◦ 0.02◦ using Cu K␣ radiation ( = 1.5406 A) (2) at room temperature. The scanning electron microscope (SEM) images of the products were obtained with a field-emission scanning electron microscope (FE-SEM, Hitachi S-4700) using an accelerating voltage of 20 kV. The transmission electron microscope (TEM) images were taken with a Hitachi H-800 microscope. The high-resolution transmission microscope (HRTEM) images, fast

Fig. 2. XRD pattern of the purified SiC sample.

X. Li et al. / Materials Science and Engineering B 176 (2011) 87–91

89

Fig. 3. (a) A low resolution TEM image of the SiC nanowires, (b) HRTEM image of a single nanowire without any defects, (c) HRTEM image of a nanowire with defects, and (d) the EDS spectrum taken from the product (the insets of the (b) and (c): the magnified images of interplanar spacing and their corresponding FFT patterns).

found. So it appeared that the wires were formed by the VS mechanism. The surface of the nanowires is smooth and clean without any nanoparticles (see the inset of Fig. 1a). However, the comparative SiC sample prepared using the common heating procedure contains more particles, besides many needle-like nanowires with an average diameter of 200 nm (Fig. 1b). It is suggested that the heating process plays an important role for the morphology and size of SiC. Phase identification of the as-purified sample was characterized using the XRD pattern. The typical XRD pattern in Fig. 2 indicates that the intensive characteristic peaks located at 2 = 35.6, 41.4,

60.0, 71.7, and 75.4◦ can be ascribed to (1 1 1), (2 0 0), (2 2 0), (3 1 1), and (2 2 2) diffractions of face-centered cubic (fcc) ␤-SiC (space group F-43m (2 1 6)), respectively, possessing a cell constant of a = 0.4362 nm predicted by Bragg formula, which is in good agreement with the reference data (JCPDS No.: 29-1129, a = 0.4359 nm). These sharp peaks imply that the sample have a highly crystalline structure. It should be noted that the small shoulder diffraction peak at 2 = 33.6◦ (marked with “SF”) comes from the stacking faults, commonly observed in SiC material, which agrees with the previous report [27]. There is no other impurity such as crystalline silica and carbon detected in the product.

90

X. Li et al. / Materials Science and Engineering B 176 (2011) 87–91

Fig. 4. FTIR spectra of the SiC sample before and after purification. Fig. 5. TGA curve of the un-purified SiC sample.

Fig. 3 shows the representative TEM/HRTEM images of the purified product and its corresponding Fast Fourier transform (FFT) patterns. The TEM image (Fig. 3a) indicates that the product is mainly composed of nanowires as well as some irregular nanoparticles. These nanoparticles are believed to be composed of SiC because only SiC can survive after the raw product was purified. Fig. 3b shows the HRTEM image of a typical single SiC nanowire with a diameter around 25 nm. It exhibits a perfect arrangement of the atomic layers without defects. The fringes perpendicular to the axis with the spacing value of 0.25 nm (indicated by the parallel lines in the inset of Fig. 3b) are clearly seen, which is equal to the interplanar spacing of [1 1 1] plane of ␤-SiC, implying the nanowire growth preferred along the [1 1 1] direction. This is also proved by the FFT patterns in the inset of Fig. 3b. Compared with Fig. 3b, Fig. 3c shows another SiC nanowire consisting of two different parts as highlighted: a region (marked with A) like the structure shown in Fig. 3b corresponding to the regular growth of ␤-SiC oriented in [1 1 1] direction; and a region (marked with B) that exhibits a partially disordered structure where numerous stacking faults perpendicular to the growth direction can be found. The formation of these stacking faults is usually considered to be the result of the thermal stress during the nanowire growth process [28–30]. The semi-quantitative analysis by energy dispersive X-ray spectrum depicted in Fig. 3d demonstrates that there are four elements found in the product: Si (29.3 at.%), C (60.3 at.%), Cu (8.4 at.%) and O (2.0 at.%). It is not surprising because Cu and part of carbon elements arose from the copper TEM grids. The small amount of O may be caused by a trace amount of residual SiO2 even after mixed acid treatment or the absorbed O2 . In other words, the sample is composed mainly of silicon and carbon. The FTIR spectra for a comparison between the as-synthesized and the as-purified samples are shown in Fig. 4. It can be seen that the two FTIR spectra are almost the same expect for the weak peak located at about 1100 cm−1 in the un-purified sample, attributable to the asymmetric Si–O stretching vibration. After purification process, this peak becomes significantly weakened. The obvious absorption peaks of both the two samples at about 830 cm−1 can be indexed as the transversal optic (TO) mode (Si–C stretching vibration) of ␤-SiC crystalline phase, which is consistent with previous report [31]. Meanwhile, each spectrum has a shoulder at around 954 cm−1 due to the longitudinal optic (LO) vibration mode. This FTIR result confirms that the Si–C structure was formed and the SiO2 component was almost eliminated through purification.

The thermal stability (in air) was investigated by TGA. Fig. 5 shows the TGA curve of the as-synthesized SiC sample. We can see that the TGA curve does not show drastic gain or loss of weight at a temperature range of room temperature to 1000 ◦ C. A slight weight loss that commences at ca. 200 ◦ C is mainly attributed to the physically adsorbed water. Another weight loss that commences at ca. 600 ◦ C is due to the combustion of unreacted carbon. The weight starting increasing at ca. 700 ◦ C suggests that SiC is oxidized under the oxidative atmosphere, which was also reported in the early literature [32]. Based on the above observation, the formation mechanism of ␤-SiC nanowires can be explained as follows. It is well known that SiC nanostructures are synthesized by the carbothermal reduction reaction between silica and carbon through the following 0 = overall V–S reaction, which is strongly endothermic with Hm ◦ 618.5 kJ/mol at 1500 C [20,33–35]: SiO2 + 3C(s) → SiC(s) + 2CO(g)

(1)

Here the symbol s and g refer to the solid state and the gaseous state, respectively. In fact, reaction (1) progresses through several step reactions in which a gaseous intermediate, silicon monoxide (SiO) is formed. The gaseous SiO can be easily transported and contacted with the carbon component to offer more contact area between the reactants. In this study, the sol-gel method gives excellent intermixing of the precursor, RF/SiO2 hybrid aerogel. A uniformly distributed intercrossing network built by RF and SiO2 units could be obtained during the sol–gel transformation [23]. When the precursor loaded in a crucible was placed into the 1500 ◦ C-preheated tube furnace, the RF component was carbonized within a very short time. The SiO formation occurs as long as silica and carbon are in intimate contact by solid–solid or solid–liquid process (at 1500 ◦ C SiO2 melts) and further is generated by the reaction between silica and CO according to reactions below: SiO2 (s) + C(s) → SiO(g) + CO(g)

(2)

SiO2 (s) + CO(g) → SiO(g) + CO2 (g)

(3)

As a result, the partial pressure of SiO could be maintained very high in the paths for diffusion provided by the intercrossing network between the original RF and SiO2 components. Apparently higher SiO concentration at a very short time is beneficial for the large generation of SiC nanostructures. Synchronously, CO can be yielded by the reaction of any newly formed CO2 with its sur-

X. Li et al. / Materials Science and Engineering B 176 (2011) 87–91

rounding carbon. Hence the partial pressures of SiO and CO vapor can become supersaturated instantaneously and the growth of SiC nanowires can progress by undergoing a gas-phase interaction according to the reaction (4): SiO(s) + 3CO(g) → SiC(s) + 2CO2 (g)

(4)

According to the report with regard to the solid-phase diffusion mechanism reported by Persson et al. [36], the possible growth process can be proposed as follows: SiC molecules in a large quantity can be produced through reaction (4). The SiC molecules are subsequently condensed and deposited on the inner surface and interspaces of micro networks of the precursor to form SiC nuclei. With the supersaturation of the SiO and CO, SiC nanowires grow up via reaction (4) from the nuclei. SiC nanoparticles can also be generated through reaction (5): SiO(g) + 2C(s) → SiC(s) + CO(g)

(5)

Some SiC nanoparticles act as the nuclei for the growth of SiC nanowires, while others are incorporated into irregular particles which are observed in TEM. The diameter of the nanowires is likely to be dictated by the size of the initial SiC nuclei. According to our previous work reported in Ref. [23], the nuclei can be formed at 1300 ◦ C. When the temperature was gradually raised to 1500 ◦ C, those formed nuclei may grow much bigger. However, when the rapid heating method was used, the existence of the supersaturation of the SiO and CO could lead to a different result. According to the von Weimarn rules [37], the supersaturation of the SiO and CO could cause the formation of SiC nuclei via a fast crystallization process. The faster the crystallization process was, the greater the amount of SiC nuclei was. The greater the amount of SiC nuclei was, the smaller size the SiC nuclei showed. So in this study, the diameters of the SiC nanowires were much smaller than before. Because the surface energy of (1 1 1) plane in ␤-SiC is much lower than those of other planes such as the (1 1 0) and (2 1 1) planes, it is generally acceptable that the SiC nanowires should grow preferentially along the [1 1 1] direction to maintain the lowest growing energy [38], and the stacking faults can be inserted in the (1 1 1) planes which can be seen in HRTEM. 4. Conclusion A facile rapid heating process has been developed to synthesize ␤-SiC nanowires in a large quantity via the carbothermal reduction of RF/SiO2 hybrid aerogel at 1500 ◦ C. From FE-SEM, XRD, and TEM measurements, the SiC nanowires possess a diameter of 20–80 nm and a length of several tens micrometers and are identified as a single crystal of ␤-SiC with growth along the [1 1 1] direction. During the growth process, the increased concentration of SiO at a moment from the rapid heating plays important and decisive role for the large-scale formation of SiC nanowires. This novel and simple method contributes a more practical approach to SiC nanostructures than the present ones.

91

Acknowledgements This work was supported by the National Natural Science Foundation of China (50572003) and State Key Basic Research Program of China (2006CB9326022006). References [1] Z.W. Pan, H.L. Lai, F.C.K. Au, X.F. Duan, W.Y. Zhou, W.S. Shi, N. Wang, C.S. Lee, N.B. Wong, S.T. Lee, S.S. Xie, Adv. Mater. 12 (2000) 1186–1190. [2] G.C. Xi, Y.K. Liu, X.Y. Liu, X.Q. Wang, Y.T. Qian, J. Phys. Chem. B 110 (2006) 14172–14178. [3] J. Parmentiera, J. Patarina, J. Dentzerb, C. Vix-Guterl, Ceram. Int. 28 (2002) 1–7. [4] G.C. Xi, Y.Y. Peng, S.M. Wan, T.W. Li, W.C. Yu, Y.T. Qian, J. Phys. Chem. B 108 (2004) 20102–20104. [5] V.G. Pol, S.V. Pol, A. Gedanken, S.H. Lim, Z. Zhong, J. Lin, J. Phys. Chem. B 110 (2006) 11237–11240. [6] E. Munoz, A.B. Dalton, S. Collins, A.A. Zakhidov, R.H. Baughman, W.L. Zhou, J. He, C.J. O’Connor, B. McCarthy, W.J. Blau, Chem. Phys. Lett. 359 (2002) 397– 402. [7] Y.B. Li, S.S. Xie, W.Y. Zhou, L.J. Ci, Y. Bando, Chem. Phys. Lett. 356 (2002) 325–330. [8] J.Q. Hu, Q.Y. Lu, K.B. Tang, B. Deng, R.R. Jiang, Y.T. Qian, W.C. Yu, G.E. Zhou, X.M. Liu, J.X. Wu, J. Phys. Chem. B 104 (2000) 5251–5254. [9] C.H. Liang, G.W. Meng, L.D. Zhang, Y.C. Wu, Z. Cui, Chem. Phys. Lett. 329 (2000) 323–328. [10] S.C. Chiu, C.W. Huang, Y.Y. Li, J. Phys. Chem. C 111 (2007) 10294–10297. [11] X.H. Sun, C.P. Li, W.K. Wong, N.B. Wong, C.S. Lee, S.T. Lee, B.K. Teo, J. Am. Chem. Soc. 124 (2002) 14464–14471. [12] W. Yang, H. Araki, C.C. Tang, S. Thaveethavorn, A. Kohyama, H. Suzuki, T. Noda, Adv. Mater. 17 (2005) 1519–1523. [13] H.H. Ye, N. Titchenal, Y. Gogotsi, F. Ko, Adv. Mater. 17 (2005) 1531–1535. [14] J.J. Niu, J.N. Wang, Eur. J. Inorg. Chem. 25 (2007) 4006–4010. [15] J.J. Niu, J.N. Wang, J. Phys. Chem. B 111 (2007) 4368–4373. [16] G.Q. Jin, P. Liang, X.Y. Guo, J. Mater. Sci. Lett. 22 (2003) 767–770. [17] D.H. Wang, D. Xu, Q. Wang, Y.J. Hao, G.Q. Jin, X.Y. Guo, K.N. Tu, Nanotechnology 19 (2008) 215602. [18] Y.J. Hao, G.Q. Jin, X.D. Han, X.Y. Guo, Mater. Lett. 60 (2006) 1334–1337. [19] N. Keller, C. Pham-Huu, G. Ehret, V. Keller, M.J. Ledoux, Carbon 41 (2003) 2131–2139. [20] J.F. Yao, H.T. Wang, X.Y. Zhang, W. Zhu, J.P. Wei, Y.B. Cheng, J. Phys. Chem. C 111 (2007) 636–641. [21] R.S. Wagner, W.C. Ellis, K.A. Jackson, S.M. Arnold, J. Appl. Phys. 35 (1964) 2993–3000. [22] T.H. Yang, C.H. Chen, A. Chatterjee, H.Y. Li, J.T. Lo, C.T. Wu, K.H. Chen, L.C. Chen, Chem. Phys. Lett. 379 (2003) 155–161. [23] X.T. Li, X.H. Chen, H.H. Song, J. Mater. Sci. 44 (2009) 4661–4667. [24] G.T. Qin, S.C. Guo, Carbon 39 (2001) 1935–1937. [25] H. Tamon, T. Kitamura, M. Okazaki, J. Colloid Inter. Sci. 197 (1998) 353–359. [26] Y.Y. Wu, P.D. Yang, J. Am. Chem. Soc. 123 (2001) 3165–3166. [27] K. Koumoto, S. Takeda, C.H. Pai, T. Sato, H. Yanagida, J. Am. Ceram. Soc. 72 (1989) 1985–1987. [28] B.C. Kang, S.B. Lee, J.H. Boo, Thin Solid Films 464–465 (2004) 215–219. [29] H.J. Li, Z.J. Li, A.L. Meng, K.Z. Li, X.N. Zhang, Y.P. Xu, J. Alloys Compd. 352 (2003) 279–282. [30] G.Z. Shen, Y. Bando, C.H. Ye, B.D. Liu, D. Golberg, Nanotechnology 17 (2006) 3468–3472. [31] L.S Liao, X.M. Bao, Z.F. Yang, N.B. Min, Appl. Phys. Lett. 66 (1995) 2382–2384. [32] R. Moene, M. Makkee, J.A. Moulijn, Appl. Catal. A 167 (1998) 321–330. [33] J.S. Lee, Y.K. Byeun, S.H. Lee, S.C. Choi, J. Alloys Compd. 456 (2008) 257– 263. [34] X.K. Li, L. Liu, Y.X. Zhang, S.D. Shen, S. Ge, L.C. Ling, Carbon 39 (2001) 159–165. [35] H.P. Martin, R. Ecke, E. Müller, J. Eur. Ceram. Soc. 18 (1998) 1737–1742. [36] A.I. Persson, M.W. Larsson, S. Stenström, B.J. Ohlsson, L. Samuelson, L.R. Wallenberg, Nat. Mater. 3 (2004) 677–681. [37] P.P. von Weimarn, Chem. Rev. 2 (1925) 217–242. [38] W. Yang, H. Araki, Q.L. Hu, N. Ishikawa, H. Suzuki, T. Noda, J. Cryst. Growth 264 (2004) 278–283.