Effects of Al addition to Si-based flux on the growth of 4H-SiC films by vapour–liquid–solid pulsed laser deposition

Abstract

The stabilizing effect of Al addition to Si-based flux on 4H-type SiC has recently become a well-known phenomenon as it is commonly found in solution growth for bulk wafer use and in the vapor–liquid–solid (VLS) mechanism for epitaxial thin films. To understand its mechanism, in this study, we investigated the Al additive effects on VLS pulsed laser deposition (PLD) of SiC films on 4° off 4H-SiC (000−1) substrates, systematically varying the Al content in the Si flux. The inclusion of 3C-type SiC in the films tended to decrease as the Al content in the flux increased. On the other hand, a sizable amount of Al, ∼10²⁰ cm⁻³, was necessarily incorporated as a dopant homogeneously into the SiC films, giving a good linear relationship between the Al density in the SiC films and the Al content in the flux. In situ observation of a flux/SiC growth interface with a confocal laser scanning microscope technique revealed that the Al addition induced a well-regulated step-flow growth even at a temperature as low as 1300 °C, which is favoured for the selective growth of the 4H-type SiC, succeeding to the 4H stacking sequence of the seed substrate.

---

1. Introduction

SiC is a semiconductor material with more than 200 different polytypes, many of which have varied physical properties, e.g. band gap, electron mobility and breakdown electric field. Among such a variety of polytypes, 4H-type SiC is one of the most promising candidates for power device applications owing to both its high breakdown electric field and large saturation drift velocity, as well as its high thermal conductivity, superior to those of existing Si. Therefore, there have been many intensive efforts so far to develop growth techniques for high-quality 4H-type SiC bulk single crystals for wafer use and epitaxial thin films as a channel layer in MOS-FET devices. However, further innovations in growth techniques, not only to improve the quality of 4H-type SiC bulk single crystals and epitaxial thin films, but also to reduce their production costs, are still necessary to accelerate the use of 4H-type SiC-based power devices in our daily lives.

For the reasons discussed above, recently much progress has been made in the development of solution growth of 4H-type SiC bulk single crystals.^1,2 It is a new alternative growth process, potentially operating at temperatures lower than the sublimation process, the only industrial production process for SiC bulk wafers that requires operating temperatures of over 2000 °C.³ The high temperature in the sublimation process can be a drawback with regards to the reduction of its production costs. In the solution process, Si-based alloys are used as a flux, to which various metal elements, such as Ni,⁴ Cr,^5,6 Al,^6,7 Ge,⁸ Ti (ref. 9) and Fe,¹⁰ have often been added to improve the solution growth behaviour of SiC. Among the additives of these metal elements, the additive effects of Al have recently drawn much attention. The study of the Al effect on the crystal growth behaviour of SiC can, at least, date back to a report in 1970 by Mitomo et al.¹¹ that the most stable phase was the 4H polytype in SiC ceramics sintered with the addition of Al, though it was not in the solution process. In the solution growth of SiC bulk crystals, a flattening effect on the growing surface⁶ as well as a stabilizing effect on the 4H-polytype have been reported.¹² As one of the reasons for this flattening effect, Komatsu et al. recently proposed, based on their experimental results and thermodynamic calculations, that the addition of Al increases the liquid/solid interfacial energy, and consequently the two-dimensional nucleation energy increases, lowering the frequency of the two-dimensional nucleation on the growing surface.¹² Furthermore, our recent in situ observation of the SiC solution growth interface with our original confocal laser scanning microscope (CLSM) technique has revealed that the Al addition can effectively suppress the step-bunching, rendering the growing surface relatively flat and uniform.¹³

On the other hand, it is a challenge to apply the solution growth method to the vapour growth of SiC films. Ferro et al. first reported a successful growth of epitaxial SiC layers in chemical vapour deposition (CVD) by a vapour–liquid–solid (VLS) mechanism with Si melt flux in 2002.¹⁴ His group also found the stabilizing effect of Al addition on the 4H-poltype in the CVD-based VLS process.¹⁵ Furthermore, even in the pulsed laser deposition (PLD)-based VLS process, we have recently found a similar Al effect.¹⁶ The preferential growth of the 6H-polytype, but not the 4H-polytype, was caused by 4 at% Al addition at an early stage of the growth, otherwise it much more likely resulted in the dominant growth of the 3C polytype, the most thermodynamically favoured phase at a temperature as low as 1160 °C.¹⁷

Motivated by the common or similar effects of Al addition on the crystal growth behaviours of SiC irrespective of the crystal growth processes as stated above, we have further investigated the Al additive effects on the SiC films grown in the PLD-based VLS process. In this study, we fabricated a series of SiC films with Al-added Si-based flux by systematically varying the Al content from 0 to 75 at% in the Si flux, and the resultant polytypes, surface morphologies and incorporated Al content were characterized in the SiC films. In addition, we have attempted to observe the SiC solution growth interface with the CLSM technique and to find any sign of Al effects on the initial stage of the SiC growth.

2. Experimental

Thin film fabrication

4° off 4H-SiC (000−1) (5 × 5 × 0.3 mm³) was used as a seed substrate after ultrasonic cleaning with ethanol, acetone and ultrapure water. Si (6N) and Al (4N) in an arbitrary bulk amount were then placed on the substrate as the flux. The Al content was varied from 0 to 75 at% with a total thickness of about 400 μm. Fig. 1 shows the vacuum chamber (base pressure: 10⁻⁷–10⁻⁶ Torr) used for depositing SiC in the present experiments. On heating the substrate at a rate of 30 °C min⁻¹ under a 0.1% H₂ contained Ar gas flow (7.5 Torr in total) up to a temperature of 1250 °C or 1530 °C, the flux becomes a liquid phase. At that temperature, an excimer laser (λ: 248 nm, 6.9 J cm⁻²) was then introduced into the chamber to deposit SiC, irradiating a SiC (3N) target placed inside the chamber at a repetition rate of 100 Hz for 10 min. In order to homogeneously ablate the SiC target, the laser was scanned two-dimensionally over the target during the deposition. The supply rate of the SiC constituent was about 30 μm h⁻¹ in thickness, and the resultant VLS growth rate of the SiC films was ∼25 μm h⁻¹, the 10 min deposition giving SiC films of 4 μm in thickness (Fig. S1†). After the deposition, the sample was cooled down to room temperature at a rate of 40 °C min⁻¹, and the flux remaining on the sample was removed by wet etching with a mixed solution of nitric acid and hydrofluoric acid (1 : 1).

Fig. 1 Schematic illustration of the VLS-PLD chamber system. SiC is deposited on a 4H-SiC (000−1) substrate via Si–Al flux.

The surface morphology of the SiC thick films was observed by a differential interference contrast microscope (DIC, ECLIPSE LV100ND, Nikon) as well as by an atomic force microscope (AFM, SP400, Seiko Instrument) and a scanning electron microscope (SEM, S-4800, Hitachi Ltd.). The Al content in the SiC films was determined by energy-dispersive X-ray spectroscopy (EDX) with which the SEM system was equipped. The polytypes of SiC in the films were identified by Raman spectroscopy (NRS-5100 JASCO, Ltd.), based on a 532 nm laser (6.9 mW), lens: ×20 and grating: 1800 l mm⁻¹. The Raman mapping images were given by the linear interpolation of the spectral data between measured points with an interval of 100 μm for a spot size of 5 μm. The micro- and nano-structures of the SiC films were characterized by cross-sectional transmission electron microscopy (TEM, JEM-ARM200F, JEOL) with energy-dispersive X-ray spectroscopy (EDX, JED-2300T, JEOL). The thickness of the SiC films was determined by a surface profiler (Surfcoder ET 4000A, Kosaka Lab. Ltd.) as well as by the cross-sectional TEM, which was almost equal to that estimated when SiC was deposited without flux as already discussed.

For the high-temperature confocal laser scanning microscope (CLSM) observation under vacuum: on-axis 4H-SiC (000−1) crystals (Cree Inc. 5 × 5 × 0.35 mm3) were used for in situ CLSM (VK-X120/130, KEYENCE) observation of a flux/SiC growth interface in a high-vacuum chamber (Kitano Seiki Co., LTD.). The detailed experimental setup of the in situ CLSM observation was reported in our previous literature.¹³

3. Results and discussion

Prior to deposition of SiC with the Si–Al flux, it was deposited without any flux under almost the same conditions as those with flux. As a result, the obtained film is a carbon-like film rather than a SiC film, as confirmed from a set of D and G bands clearly found in its Raman spectrum and from the fiber-like features observed in its SEM image (Fig. S2†). This is probably because the deposited Si at the high temperature was ready to re-evaporate under the vacuum conditions. Then, in order to investigate the Al additive effects on the SiC films grown in the PLD-based VLS process, we attempted to fabricate SiC films with Si–Al flux whose Al content was varied from 0 to 75%.

Fig. 2 shows a typical set of DIC images of the SiC films grown with Si_100−x–Al_x flux (x = 0, 25, 50 and 75), together with a binary phase diagram of the Si–Al alloy system,¹⁸ showing the liquidus temperature of each flux composition for reference. Note that only for the SiC film grown with the Si flux, the growth temperature was set to be 1530 °C, above the melting temperature of Si, 1414 °C, while for all the other SiC films grown with the Al-contained flux it was 1250 °C, just close to or above the liquidus temperatures of the Si–Al alloys tested as the flux. On the surface of the film grown with the pure Si flux, there is a mixture of two kinds of step lines with step edge angles of 120° and 60°, which are considered to come from 4H(6H)-SiC and 3C-SiC polytypes, respectively. In contrast, a uniform steps-and-terraces structure with a step edge angle of 120°, suggestive of having (11̄0n) facets, spreads all over the film surfaces grown with the Si₇₅Al₂₅ and Si₅₀Al₅₀ flux. When the amount of Al in the flux was further increased, the terrace widths became much larger with round-shaped steps.⁶

Fig. 2 Typical set of DIC images of the SiC films grown with the Si_100−x–Al_x flux (x = 0, 25, 50 and 75), together with a binary phase diagram of the Si–Al alloy system. The arrows indicate the direction of dip of the substrate.

High magnification observation by AFM (Fig. 3) revealed that the surface structure of the SiC film grown with the pure Si flux was rougher than expected from the surface morphology as macroscopically observed in its DIC image. On the other hand, the step-and-terrace structures for the SiC films grown with the Al-contained Si flux were clearly observed in the AFM images comparable to those in the corresponding DIC images, and furthermore it was found that the round-shaped macro-step on the SiC film grown with the Si₂₅Al₇₅ flux also consisted of many small steps with a step edge angle of 120°, while the step lines were macroscopically observed to run perpendicular to the direction of dip in the DIC image.

Fig. 3 Typical set of AFM images (20μm × 20μm) of the SiC films grown with the Si_100−x–Al_x flux (x = 0, 25, 50 and 75) in Fig. 2.

Fig. 4(a) shows a set of Raman spectra of the SiC films grown with different Al content in the flux, from which the polytypes of SiC can be identified. 3C- and 4H-polytpes were mixed in the SiC film grown with the pure Si flux. Meanwhile, no apparent 3C-polytype peaks were found in any other SiC films grown with the Al-contained Si flux, and the spectra were almost identical to that of the single phase 4H-polytype. In more detail, the spectra were expanded in a Raman shift range from 100 to 220 cm⁻¹ to more closely look into the FTA(+)/(−) peaks. There are two features to note as the Al content in the flux increases in this spectral region. One is an increase of the background, and the other is a shift of the FTA(+) peak as shown in the inset, both of which are known as a phenomenon of Fano interference.^15,19–21 Taking into account a gradual broadening of the A₁ LOPC peak around 1000 cm⁻¹ with the Al content in the flux as well (Fig. 4(a)), a substantial amount of Al should be incorporated as a dopant into the SiC films grown with the Al-contained Si flux, as will be discussed later.

Fig. 4 (a) Set of Raman spectra of the SiC films grown with different Al content in the flux. (b) The expanded Raman spectra in a Raman shift range from 100 to 220 cm⁻¹, along with the peak shift of FTA(+) plotted against the Al content in the flux (inset). (c) Series of micro-Raman spectral mappings of the intensity ratio of two bands at 796 cm⁻¹ and 776 cm⁻¹ for different Al content in the flux. The warm color in the color scale, i.e. the intensity ratio >0.03, means higher inclusions of the 3C-polytype in the scanned area.

Fig. 4(c) is a series of micro-Raman spectral mappings of the intensity ratio of two bands at 796 cm⁻¹ and 776 cm⁻¹. The warm color, i.e. the intensity ratio >0.03 in the color scale, means higher inclusions of the 3C-polytype in the scanned area (cf. it is below 0.03 for the 4H-SiC single crystal substrate used in the present experiments). As expected from Fig. 4(a), a large portion of the 3C-polytype could be detected in the SiC film grown with the pure Si flux. In contrast, though the mapping results often depend on the scanned areas even in one sample, a trend was found where the inclusions of 3C-SiC decreased as the Al content in the flux increased, and finally, when using the Si₂₅Al₇₅ flux, almost perfect single phase 4H-SiC could be obtained.

Fig. 5(a) and (b) are a set of bright-field cross-sectional TEM images of the SiC films grown with the pure and 75 at% Al-contained Si flux, respectively, after removing the remaining flux. Within a range of the sample observable by TEM, the greater portion of the SiC film grown with the pure Si flux is of the 3C-polytype. Looking more closely into the interface of the film and the substrate as shown in the high-resolution TEM (HRTEM) image of Fig. 5(c), the interface is rough and the 3C-type SiC is ready to appear even at the initial stage of growth. In contrast, the SiC film grown with the 75 at% Al-contained Si flux is almost of the 4H-polytype, and the HRTEM image of its interface as shown in Fig. 5(d) is too smooth to distinguish the interface, as the crystal quality of the grown film is almost identical to that of the substrate. It should be noted that the SiC film grown even with the 75 at% Al-contained Si flux was partially covered with some 3C-type SiC thin layers near the film surface, which might be precipitated from the SiC containing flux at a low temperature during the sample cooling process after stopping the deposition.

Fig. 5 (a) & (b) Set of bright-field cross-sectional TEM images of the SiC films grown with the pure and 75 at%-contained Si flux, respectively, after removing the remaining flux. (c) & (d) The corresponding HRTEM images near the film-substrate interface of (a) and (b), respectively.

As already pointed out in the Raman spectra of the SiC films, a sizable amount of Al should be incorporated into the growing SiC film from the Al-containing flux during the film growth. A set of EDX elemental mappings of Si, C and Al in the film grown with the 75 at% Al-contained Si flux is displayed in Fig. 6(a). Al, Si and C are found to homogeneously distribute throughout the film. If the VLS process proceeds under conditions very close to the thermodynamic equilibrium, the Al content in the films should essentially result from the partition equilibrium reached between the liquid flux and the solid growing film at the growth temperature. In Fig. 6(b), the Al density (cm⁻³), evaluated for each SiC film by EDX, is plotted as a function of the content of Al in the flux, giving a good linear relationship between them, with its partition equilibrium coefficient K_d at the temperature of 1250 °C calculated to be (1.55 ± 0.2) × 10⁻². This linearity is evidence that the present VLS-PLD process should proceed under conditions almost close to the thermodynamic equilibrium. Fig. 6(c) shows a line depth profile of Al density in the TEM image of Fig. 5(b). The Al density seems homogeneous throughout the film, though it is relatively high in a region near the interface. Since the bulk amount of Si and Al may not be completely mixed yet just after it becomes liquid, the partition equilibrium will not be reached during the etching back process before starting the deposition. As a result, Al can be transiently more incorporated into the SiC film near the interface.

Fig. 6 (a) Set of EDX elemental mappings of Si, C and Al in the film grown with the 75 at%-contained Si flux. (b) Al density (cm⁻³), evaluated for each SiC film by EDX, is plotted as a function of the content of Al in the flux. (c) The line depth profile of Al density in the TEM image of Fig. 5(b).

As shown in the comparison of the TEM images of the SiC films grown with the pure and 75 at% Al-contained Si flux, the effect of Al addition can be found even at the initial stage of growth. Therefore, it would be informative to directly observe how different the initial growth interface is during solution growth of SiC depending on whether or not Al is contained in the flux. Fig. 7(a) and (b) are a set of sequential CLSM images of the initial growth interface for the pure and 10 at% Al-contained Si fluxes, respectively (see Video S1 and S2†). In the case of the pure Si flux, the flux completely became liquid around the melting temperature of Si, 1414 °C, and the etching back process seemed dominant along the polishing marks of the seed SiC surface at a temperature of 1450 °C. As the temperature increased up to 1500 °C, it was then followed by the appearance of step-and-terrace structures, but the step advance was too random to form straight steps. The step edge angles of 60° as well as 120° indicate inclusion of the 3C-polytype even from the initial stage of the growth. In contrast, the Al 10 at%-contained Si flux was ready to become liquid, homogeneously spreading over the seed substrate, at around 1300 °C, just close to its expected liquidus temperature. Even at this low temperature the step-advance was observed with step lines almost parallel to each other, characteristic to the step-flow growth mode. Above 1600 °C, the terrace width becomes large, accompanied by step-bunching. Since 4° off 4H-SiC (000−1) substrates were used for the growth of the SiC films in the present experiment, such a well-regulated step-flow growth around the growth temperature of 1250 °C should be favoured in succeeding to the 4H-stacking sequence of the seed substrate. This might be one of the possible reasons for the stabilizing effect of Al addition on the 4H-polytype in the growth of SiC films as well as bulk single crystals.

Fig. 7 Set of sequential CLSM images of the initial growth interface for the pure (a) and 10 at% Al-contained (b) Si fluxes.

4. Conclusions

We have demonstrated homoepitaxial growth of Al-doped 4H-type SiC films at a growth rate of 25 μm h⁻¹ by the VLS-PLD process with Al-contained Si flux at a temperature as low as 1250 °C. The Al density in the SiC films can be well controlled in proportion to the Al content in the flux, which supports that the VLS process proceeds under conditions almost close to the thermodynamic equilibrium similar to the solution growth process. From in situ CLSM observation of the solution growth interface, the stabilizing effect of Al addition on the 4H-polytype can be attributed to its enhanced well-regulated step-flow growth even at a low temperature of 1300 °C, allowing it to effectively succeed to the 4H-stacking sequence of the seed substrate.

Acknowledgements

This work was supported by the Novel Semiconductor Power Electronics Project Realizing Low Carbon Emission Society under the New Energy and Industrial Technology Development Organization (NEDO), and by the Advanced Low Carbon Technology Research and Development Program (ALCA) of the Japan Science and Technology Agency (JST).

Notes and references

1. Y. Yamamoto, S. Harada, K. Seki, A. Horio, T. Mitsuhashi, D. Koike, M. Tagawa and T. Ujihara, Appl. Phys. Express, 2014, 7, 065501.
2. K. Kamei, K. Kusunoki, N. Yashiro, N. Okada, K. Moriguchi, H. Daikoku, M. Kado, H. Suzuki, H. Sakamoto and T. Bessho, Mater. Sci. Forum, 2012, 45, 717–720.
3. Y. M. Tariov and V. F. Tsvetkov, J. Cryst. Growth, 1978, 43, 209–212.
4. Y. Yonezawa, M. Ryo, A. Takigawa and Y. Matsumoto, Sci. Technol. Adv. Mater., 2011, 12, 054209.
5. K. Danno, H. Saitoh, A. Seki, H. Daikoku, Y. Fujiwara, I. Ishii, H. Sakamoto and Y. Kawai, Mater. Sci. Forum, 2010, 13, 645–648.
6. T. Mitani, N. Komatsu, T. Takahashi, T. Kato, K. Fujii, T. Ujihara, Y. Matsumoto, K. Kurashige and H. Okumura, J. Cryst. Growth, 2014, 401, 681–685.
7. C. Jacquier, G. Ferro, F. Cauwet, J. C. Viala, G. Younes and Y. Monteil, J. Cryst. Growth, 2003, 254, 123–130.
8. O. Filip, B. Epelbaum, M. Bickermann and A. Winnacker, J. Cryst. Growth, 2004, 271, 142–150.
9. K. Kusunoki, S. Munetoh, K. Kamei, M. Hasebe, T. Ujihara and K. Nakajima, Mater. Sci. Forum, 2004, 123, 457–460.
10. T. Yoshikawa, S. Kawanishi and T. Tanaka, Jpn. J. Appl. Phys., 2010, 49, 051302.
11. M. Mitomo, Y. Inomata and M. Kumanomido, Yogyo Kyokaishi, 1970, 78, 224–228.
12. N. Komatsu, T. Mitani, Y. Hayashi, T. Kato, S. Harada, T. Ujihara and H. Okumura, J. Cryst. Growth, 2017, 458, 37–43.
13. A. Onuma, S. Maruyama, N. Komatsu, T. Mitani, T. Kato, H. Okumura and Y. Matsumoto, Cryst. Growth Des., 2017, 17, 2844–2851.
14. G. Ferro, D. Chaussende, F. Cauwet and Y. Monteil, Mater. Sci. Forum, 2002, 389–393, 287.
15. G. Ferro and C. Jacquier, New J. Chem., 2004, 28, 889–896.
16. A. Onuma, S. Maruyama, T. Mitani, T. Kato, H. Okumura and Y. Matsumoto, J. Ceram. Soc. Jpn., 2016, 124, 506–509.
17. A. Onuma, S. Maruyama, T. Mitani, T. Kato, H. Okumura and Y. Matsumoto, CrystEngComm, 2016, 18, 143–148.
18. Binary Alloy Phase Diagrams, ed. T. B. Massalski, et al., ASM International, Materials Park, Ohio, 2nd edn, 1998, vol. 1.
19. H. Harima, T. Hosoda and S. Nakashima, Mater. Sci. Forum, 2000, 338–342, 607–610.
20. R. Hu, C. C. Tin, Z. C. Feng, J. Liu and Y. Vohra, Inst. Phys. Conf. Ser., 1996, 142, 345.
21. T. Mitani, S. Nakashima, M. Tomobe, S. Y. Ji, K. Kojima and H. Okumura, Mater. Sci. Forum, 2014, 778–780, 475–478.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c7ce00945c

---