The crystal quality of a 4H-silicon carbide (4H-SiC) epitaxial layer is crucial to the development of high-performance 4H-SiC-based electronic power devices. However, the quality assessment of 4H-SiC homoepitaxial thin film is problematic because the same bulk material interferes with the probe of the epilayer. In this paper, we propose a simple and straightforward strategy to assess the quality of a homoepilayer using ultraviolet (UV) Raman spectroscopy (RS). Rather than focusing on the normally allowed modes, we shift our attention to the forbidden modes instead. We demonstrate that forbidden modes, which were usually ignored, are more sensitive to the crystalline imperfection and can be an effective quality probe. Our approach analyzes the crystal quality swiftly, without the need for the data fitting involved in the conventional method, and therefore makes the quality assessment much more efficient. The new method may also be applied to the other thin film materials.
© 2017 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
Single crystalline silicon carbide (SiC) has excellent material properties such as wide-band gap, high thermal conductivity, and high electron mobility especially suitable for high temperature and high power devices. Among the various types of SiC, the 4H polytype is considered to be one of the most promising materials for the new generation of power and ultraviolet (UV) devices [1–8]. As the crystalline quality of the SiC seriously affects the performance of the devices, a good-quality 4H-SiC wafer is critical to the device applications. Unfortunately, it still takes strenuous efforts to grow a high-quality 4H-SiC wafer [1, 4, 9–12], since SiC has more than 200 polytypes, depending on the crystal stacking structure. 4H-SiC wafers can thus have varying quality which affects the quality of the epilayers. This, in turn, affects the subsequent device performance. Quality assessment of the epilayer thus becomes essential and imperative before the material can be used for device fabrication.
However, nondestructive quality analysis of a homoepitaxial 4H-SiC epilayer remains as a challenge today, because the homo-substrate interferes with the probing of epilayer. There are many restrictions in commonly used techniques. For example, transmission electron microscopy (TEM) is destructive, and the atomic force microscopy (AFM) is time consuming. The conventional non-destructive methods, such as x-ray diffraction (XRD) and visible Raman spectroscopy, are less effective because of the strong signals arising from the substrate, interfering with those from the thin surface layers. Besides, it is difficult to single out an individual layer from the whole. All these leave us with no choice but the ultraviolet (UV) and deep ultraviolet (DUV) Raman scattering spectroscopy (RSS), which is capable of probing individual layers from a few nanometers to several thousands of nanometers, by varying laser excitation wavelengths, which results in variable penetration depths [13–19]. However, in addition the ordinary allowed modes in UV/DUV Raman scattering spectra, there appear extraordinary, or forbidden modes from time to time [20–23]. Normally, attention has solely been given to the allowed Raman scattering modes, but not to the forbidden modes. Thus crystal quality has usually been analyzed by the allowed modes, producing limited amount of information, despite the fact that, the forbidden modes are more sensitive to the crystalline imperfection found in most of these thin films. As a result, little research has been done to study the forbidden mode, at least its relationship to the quality assessment has not yet to be explored.
In searching for a better probe of the quality of homoepitaxial 4H-SiC thin films, we found that rather than paying attention only to the allowed modes, it is more effective to do the opposite, viz., to start with the “bad samples”. Based on the Raman scattering (RS) theory, under normal scattering conditions, certain lattice vibrations cannot be observed due to the selection rules [21, 24]. But in reality, besides the allowed Raman modes, some unconventional, forbidden Raman modes occur in the Raman scattering spectra [20–23], which were often attributed to a “bad case”, and then ignored. If, however, we take the opposite approach, the appearance of forbidden modes can actually be considered as a signal of revealing the crystal quality. In other words, the forbidden modes may just be the better quality assessment probe, and thus a novel strategy for rapid quality analysis can be explored.
Therefore, the purpose of this report is to develop a better strategy to analyze effectively the quality of SiC thin films using the forbidden modes. In combination with UV and DUV excitations for RS probing, of a series of undoped 4H-SiC epilayers grown on n + 4H-SiC substrates, we have investigated the features of the RS forbidden modes, including intensity, width, peak frequency, and their dependence on the crystal quality. Based on the resulting correlation to the forbidden modes in UV RS spectra, the new strategy can successfully be demonstrated, to evaluate effectively the crystal quality of homoepitaxial 4H-SiC thin films with much efficiency.
Firstly, in this report, a series of 4H-SiC homoepitaxial thin films have been prepared by the chemical vapor deposition (CVD), the most established technique for growth of epitaxial layers of 4H-SiC. Five undoped 4H-SiC epilayers were grown on n-type (~5x1018 cm−3) 4H-SiC substrates, using SiH4/C3H8 as sources, H2 as carrier gas, a reactor pressure at 100 Torr and substrate growth temperature of 1450-1500°C. Four samples, namely H1-H4, were grown with a SiH4 source concentration of 150 ppm in hydrogen and different C3H8 source concentrations of 114, 260, 227 and 152 ppm, to obtain different Si:C ratios of 0.44, 0.19, 0.22 and 0.33, respectively. The fifth sample, H5, had a Si:C ratio of 0.44 but higher concentrations of SiH4 (240 ppm) and C3H8 (182 ppm). Growth time was 3-hour for all samples on the same n + 4H-SiC substrate of ~350 μm thick. The thicknesses of undoped 4H-SiC epilayers were measured by Spectroscopic Ellipsometry (SE) with H1/3.9μm, H2/2.9μm, H3/3.2μm, H4/3.7μm and H5/5.9μm, respectively.
And then five samples with the size of 45mm (length) × 45mm (width) × 2mm (thickness), have been measured in backscattering geometry at room temperature (RT), which is illustrated in Fig. 1, by employing two Raman spectrometers: a Renishaw Raman micro-spectroscope with 325/360/514 nm excitations and a Zolix three-stage spectrometer with 266 nm excitation. The spectral resolution of Raman measurements is 0.2 cm−1. The diameter of focused laser beam on samples is about 3 μm. The penetration depth dp≃1/α = l/4πk, where α is the absorption coefficient and k the extinction coefficient , in 4H-SiC is ~200/2000/5000 nm at wavelength of 266/325/360 nm, respectively. And at 514 nm, the light goes through the whole structure including entire substrate for its penetration depth is beyond 30,000 mm . These Raman spectra have been analyzed by taking an unconventional approach.
3. Results and discussion
3.1 The forbidden mode E1(TO)
Based on the above consideration, in this section, we turn our attention on, not the allowed modes, but the forbidden modes. Figure 1 shows the Raman spectra obtained in backscattering geometry from five undoped 4H-SiC/n + 4H-SiC samples H1~H5 excited by the 325 nm HeCd laser, respectively. The 4H-SiC characteristic phonon modes are E2(TA) at 205.5 cm−1, A1(LA) at 612.1 cm−1, E2(TO) at 777.3 cm−1, E1(TO) at 798.6 cm−1, and A1(LO) at 967.3 cm−1 [17, 20]. Here the E1(TO) mode is just a forbidden mode, which is not allowed in this geometry and should not be observed. It appears in the RS spectra of five samples with different features. Most impressively, in the spectrum of H1 sample, there is a very strong E1(TO) mode which appears rarely, its intensity is even stronger than the allowed mode E2(TO). This strong mode E1(TO) is usually regarded as “the bad case” which was explained simply to be caused by the poor quality, and then ignored. As for those weak E1(TO) modes from samples H2-H5, they are also neglected because they are considered to have little effect on the allowed mode E2(TO) which were extensively studied. As a result, the forbidden mode E1(TO), whether it is the strong signal or weak signal, has attracted little interest in most research. But, if we take the opposite approach, the forbidden mode E1(TO) can actually be considered as the characterization of the crystalline imperfection, and its properties and dependence on crystalline imperfection can be explored to be a novel quality probe. So we start our investigation with this “bad sample”, and study the forbidden modes in the following measurements.
To obtain further convincing evidence for the forbidden mode E1(TO), we have taken measurements from the center to the edge of samples. Figure 2 shows the TO modes from H1 by moving the probing positions from the center to the edge of the sample, using 325 nm laser excitation, in where the parameter D is the distance between the probing position and the edge of H1, where D1 = 20000 μm, D2 = 10000 μm, D3 = 5000 μm, D4 = 2000 μm, D5 = 500 μm, D6 = 100 μm, respectively. As seen, the forbidden mode E1(TO) what we are looking for, its intensity gradually increases as the distance D reduces, and becomes very strong at the edge of H1. To the contrary, the ordinary allowed mode E2(TO) becomes weak step by step. By comparing the E1(TO) modes in Fig. 1 and Fig. 2, the strong E1(TO) mode from H1 must be detected when the probing position just right the edge region. As expected, a weak E1(TO) mode appears in the position of D1 = 20000 μm. The “bad sample” is not really bad. Moreover, Fig. 3 exhibits the variation of intensity ratio E1(TO)/E2(TO) with the distance D for five samples H1-H5. Likewise, the intensity ratio E1(TO)/E2(TO) increases gradually while the probing region is near to the edge of sample. As known, the crystalline quality of the center of a thin film is better than that of the edge. This quality change has reflected accordingly in the variation of E1(TO)/E2(TO).
In addition, we have investigated the forbidden modes E1(TO)s from different individual layers of sample H1 using different excitation lasers with wavelengths of 266 nm, 325 nm, 360 nm, and 514 nm. The laser lights probe the layers which the penetration depths are ~200/2000/5000/300000 nm, which are near twentieth of, half of the whole 4H-SiC epi-film, a layer including the whole epilayer and thin substrate of 1000 nm, and the entire wafer with the substrate, respectively. Figure 4 presents their Raman spectra, in where the forbidden modes we are searching, E1(TO)s, also appear with different features.
It is evident that, from our experiments, a forbidden mode E1(TO) is not “the anomaly”, but a real signal of revealing the imperfect feature which correlate closely with the crystalline quality. Therefore, we focus on the forbidden mode E1(TO), and develop a new approach to analyze the quality of these thin films.
3.2 Quantitative analysis on the E1(TO) mode
Under normal scattering geometry conditions, the appearance of a forbidden mode E1(TO) reveals that the orientation of at least parts of the epi-film is different from the expected direction. The stronger the E1(TO) mode is, the more crystalline imperfection has, and thus the poorer crystalline quality is. Even in the case of imperfect scattering geometry, the intensity differences between the forbidden modes evidence the quality differences, because the system errors are the same. We have defined here the intensity ratio E1(TO)/E2(TO) as the important quality-performance parameter, due to the crystalline imperfection and perfection contribute to the E1(TO) and E2(TO) mode respectively. The more the imperfection, the larger the intensity ratio E1(TO)/ E2(TO), and thus the worse quality. Table 1 lists the line shape parameters of both E1(TO) and E2(TO) modes, the intensity ratios E1(TO)/ E2(TO)s for five samples H1-H5, are 1:0.82/H1, 0.067:1/H2, 0.083:1/H3, 0.052:1/H4, 0.062:1/H5, respectively. Among five samples, H1 possessing the maximum E1(TO)/ E2(TO), H4 with the minimum one, and H2, H5 in middle, that reveals, H1, H4, H2 and H5 have the worst, the best, and the better crystalline quality in the probing region, respectively. Figure 2 shows, the E1(TO) mode becomes gradually stronger while the E2(TO) mode decreases, as the distance D reduces. This means that the epilayer have worse and worse crystal quality in where is gradually close to the edge of the sample. Figure 3 illustrates, the intensity ratios E1(TO)/E2(TO)s increase seriously at the edge of five samples and their quality are terrible. Near the center of epi-films, H1 has the strongest intensity ratio E1(TO)/ E2(TO) and H4 owns the weakest value. It also shows that the sample H1 has the worst quality and H4 is the best film near the center region. As seen, it is quick and effective to assess the quality of five 4H-SiC epi-films by using the forbidden modes.
Figure 4 presents Raman spectra from different individual layers with penetration depths of ~200/2000/5000/300000 nm of sample H1, using 266/325/360/514 nm excitation lasers. The 514 nm excitation light in Raman measurement penetrates through the entire 4H-SiC epi-film and the whole n+ 4H-SiC substrate (about 350 μm thick), the 360 nm and the 325 nm light probe about half of and the whole epi-film, and the 266 nm light detects near surface top layer, near twentieth of the undoped 4H-SiC epi-film. The differences of E1(TO) modes among these four wavelength excitations are revealed, from which useful quality information on homoepitaxial 4H-SiC structure along growing direction is provided. Table 2 lists the line shape parameters of both E1(TO) and E2(TO) modes. Among three individual epilayers, the surface layer has the minimum E1(TO)/E2(TO), which become larger along with increasing penetration depths. That manifests, the layer near the interface of n+ 4H-SiC substrate has worse crystalline quality compared to the middle layer of the 4H-SiC epi-film, and the crystalline quality becomes better with increasing growth thickness in the epilayer. In addition, we noticed, there is a small E1(TO)/E2(TO) with 0.083:1 excited by 514 nm laser, because the RS signals are mainly from the homo-substrate which its thickness is ~70 times of that of 4H-SiC epi-film and the substrate material has a good quality.
We are surprised to find that, the quality assessment has been effectively done through a simple and straightforward method which focuses on the forbidden mode E1(TO) but not the usually allowed mode E2(TO). The forbidden modes associated with the imperfect features are sensitive to the quality change, and can thus be the high-performance quality analysis. In addition, by scanning a sample with different wavelength lasers, the three-dimensional quality evaluation becomes possible to make.
3.3 Spatial correlation model analysis on the E2(TO) mode
For comparison, we adopt the common spatial correlation model (SCM) to analyze the allowed mode E2(TO), and the first-order TO RS intensity I (ω) can be expressed as [27-28]:
By employing Eqs. (1-2) to fit the E2(TO) modes in Fig. 1 under 325 nm excitations, the calculated Raman spectra, including full-width at half-maximum (FWHM), i.e, Γ0, can be obtained. Figures 5 shows the E2(TO) lines shape fit for five 4H-SiC samples H1-H5 under the 325 nm excitation, and the fitted parameters are listed in Table 3. As seen, the fitted results are not good in the low-value parts of E2(TO) modes in where the E1(TO) modes disturb the data fitting. Much worse, the coherence length L, the most important parameter which corresponds to the crystalline quality, is insensitive to the data fitting. The fitted values of L, for H1-H3 and H5, are almost at the same range of 18-22, despite the fact that the sample H1 presents the stronger forbidden mode E1(TO) than the allowed mode E2(TO) which its intensity ratio of E1(TO)/E2(TO) is ~15 times of those of samples H2-H5. In addition, H2, H4, and H5 have roughly the same linewidth Γ0. These make the quality information ambiguous and hard to compare. From Table 3, only the quality of H1 and H4 can be clearly assessed for H1 with the shortest L and the widest Γ0, H4 with the best L value and narrowest linewidth, these indicate H1 and H4 have the worst and the best crystalline quality, respectively. However, this result can be easily obtained based on the straightforward analysis of E1(TO) modes without theoretical fitting, as shown in section 3.2. Besides, the quality of H2, H3, and H5 can also be compared. Obviously, the forbidden modes E1(TO)s are more sensitive to the crystalline quality and the results in quality assessment become more effective.
Finally, a useful conclusion has been obtained here: the forbidden modes which were usually considered as “bad cases” and then ignored were, are actually the signals of revealing the crystalline imperfection. They are more sensitive to the crystal quality and can make a better quality probing. Our finds have demonstrated that, a better strategy for efficient quality assessment can be developed using forbidden modes.
To summarize, a better strategy has been proposed to effectively analyze the quality of homoepitaxial 4H-SiC thin films by taking an unconventional approach, which focuses on the forbidden modes rather than the ordinary allowed modes in UV RS spectra. We have successfully demonstrated that the forbidden modes are actually the signals of uncovering the crystalline imperfection, and are more sensitive to the crystal quality, and can thus be the effective quality probe. This approach – which is very simple and straightforward – makes a rapid quality assessment without the data fitting which was extensively implemented by the conventional method. Therefore, the approach has the advantages of simplification and greater efficiency. It is also possible to make the three-dimensional quality analyses by scanning samples with different wavelength excitation lasers.
National Natural Science Foundation of China (NO.61367004, U1731239).
The work at Guangxi University is supported by National Natural Science Foundation of China and special funding for Guangxi distinguished professors (Bagui Yingcai & Bagui Xuezhe).
References and links
1. Z. C. Feng, SiC Power Materials: Devices and Applications (Springer Berlin, 2004).
3. T. Kimoto and J. A. Cooper, Fundamentals of Silicon Carbide Technology (John Wiley & Sons, 2014).
4. T. Kimoto, “Bulk and epitaxial growth of silicon carbide,” Prog. Cryst. Growth Charact. Mater. 62(2), 329–351 (2016). [CrossRef]
5. C. Bouhafs, A. A. Zakharov, I. G. Ivanov, F. Giannazzo, J. Eriksson, V. Stanishev, P. Kühne, T. Iakimov, T. Hofmann, M. Schubert, F. Roccaforte, R. Yakimova, and V. Darakchiev, “Multi-scale investigation of interface properties, stacking order and decoupling of few layer graphene on C-face 4H-SiC,” Carbon 116, 722–732 (2017). [CrossRef]
6. Y. Ou, V. Jokubavicius, and S. Kamiyama,3C. Liu, R. Berg, M. Linnarsson, R. Yakimova, M. Syvajarvi, and H. Ou, “Donor-acceptor-pair emission characterization in N-B doped fluorescent SiC,” Opt. Mater. Express 1(8), 1439–1446 (2011).
8. J. Hu, R. Jia, B. Xin, B. Peng, Y. Wang, and Y. Zhang, “Effect of Low Pressure on Surface Roughness and Morphological Defects of 4H-SiC Epitaxial Layers,” Materials (Basel) 9(9), 743 (2016). [CrossRef] [PubMed]
9. C. C. Tin, R. Hu, J. Liu, Y. Vohra, and Z. C. Feng, “Raman microprobe assessment of low-pressure chemical vapor deposition-grown 4H-SiC epilayers,” J. Cryst. Growth 158(4), 509–513 (1996). [CrossRef]
10. A. T. S. Wee, K. Li, and C. C. Tin, “Surface chemical states on LPCVD-grown 4H-SiC epilayers,” Appl. Surf. Sci. 126(1–2), 34–42 (1998). [CrossRef]
11. C. Wells, J.-Y. Jiang, T.-F. Chang, C.-F. Huang, J. Ke, W. Luo, G. Xia, K. Y. Cheong, and F. Zhao, “Stress and thermal characterization of 4H-SiC microelectromechanical structures,” Mater. Lett. 191, 196–199 (2017). [CrossRef]
12. S. Nakashima, H. Okumura, T. Yamamoto, and R. Shimidzu, “Deep-Ultraviolet Raman Microspectroscopy: Characterization of Wide-Gap Semiconductors,” Appl. Spectrosc. 58(2), 224–229 (2004). [CrossRef] [PubMed]
13. T. Tomita, S. Matsuo, T. Okada, T. Kimoto, H. Matsunami, T. Mitani, and S.-I. Nakashima, “Deep-ultraviolet micro-Raman investigation of surface defects in a 4H-SiC homoepipaxially grown film,” Appl. Phys. Lett. 87(24), 241906 (2005). [CrossRef]
14. S. Nakashima, T. Mitani, J. Senzaki, H. Okumura, and T. Yamamoto, “Deep ultraviolet Raman scattering characterization of ion-implanted SiC crystals,” J. Appl. Phys. 97(12), 123507 (2005). [CrossRef]
15. S. Nakashima, T. Kitamura, T. Kato, K. Kojima, R. Kosugi, H. Okumura, H. Tsuchida, and M. Ito, “Determination of free carrier density in the low doping regime of 4H-SiC by Raman scattering,” Appl. Phys. Lett. 93(12), 121913 (2008). [CrossRef]
16. T. Kitamura, S. Nakashima, T. Mitani, N. Nakamura, K. Furuta, and H. Okumura, “Depth distribution of strain in GaN/AlN/SiC heterostructures by DUV micro-Raman spectroscopy,” Phys. Status Solidi4(7), 2375–2378 (2007) (c). [CrossRef]
17. S. Nakashima, T. Mitani, T. Tomita, T. Kato, S. Nishizawa, H. Okumura, and H. Harima, “Observation of surface polarity dependent phonons in SiC by deep ultraviolet Raman spectroscopy,” Phys. Rev. B 75(11), 115321 (2007). [CrossRef]
18. R. J. Kuhudzai, J. B. Malherbe, N. G. van der Berg, T. T. Hlatshwayo, O. Odutemowo, L. C. Prinshloo, A. V. Buys, R. Erasmus, and E. Wendler, “Deep ultra violet and visible Raman spectroscopy studies of ion implanted 6H-SiC: Recrytallisation behaviour and thermal decomposition/thermal etching of the near surface region,” Nucl. Instrum. Meth. B 365(3), 342–346 (2015). [CrossRef]
19. G. Abstreiter, E. Bauser, A. Fischer, and K. Ploog, “Raman Spectroscopy-A Versatile Tool for Characterization of Thin Films and Heterostructures of GaAs and AlxGa1-xAs,” Appl. Phys. (Berl.) 16(4), 345–352 (1978). [CrossRef]
20. M. Okamoto, R. Kosugi, S. Nakashima, K. Fukuda, and K. Arai, “Deep UV excitation Raman spectroscopy of homoepitaxial 4H-SiC films grown by microwave plasma chemical capor deposition,” Mater. Sci. Forum 457–460, 629–632 (2004). [CrossRef]
21. H. Ashraf, “Investigation of Symmetries of Phonons in 4H and 6H-SiC by Infrared Absorption and Raman Spectroscopy,” Ph, D. thesis, Linköping University, 2005.
22. Y. Peng, X. Hu, X. Xu, X. Chen, J. Peng, J. Han, and A. S. Dimitrijev, “Temperature and doping dependence of the Raman scattering in 4H-SiC,” Opt. Mater. Express 6(9), 2725 (2016). [CrossRef]
23. S. Chen, L. Wan, D. Xie, Z. R. Qiu, X. Jiang, C.-C. Tin, and Z. C. Feng, “Adducing crystalline features from Raman scattering studies of cubic SiC using different excitation wavelengths,” J. Phys. D Appl. Phys. 50(11), 115102 (2017). [CrossRef]
24. R. Loudon, “The Raman effect in crystals,” Adv. Phys. 52, 423–482 (1964). [CrossRef]
25. H. Fujiwara, Spectroscopic Ellipsometry Principles and Applications (John Wiley & Sons, Ltd. 2007).
26. S. Zollner, J. G. Chen, E. Duda, T. Wetteroth, S. R. Wilson, and J. N. Hilfiker, “Dielectric functions of bulk 4H and 6H SiC and spectroscopic ellipsometry studies of thin SiC films on Si,” J. Appl. Phys. 85(12), 8353–8361 (1999). [CrossRef]
27. Z. C. Feng, “Micro-Raman scattering and micro-photoluminescence of GaN thin films grown on sapphire by metalorganic chemical vapor deposition,” Opt. Eng. 41(8), 2022–2031 (2002). [CrossRef]
28. M. Havel, D. Baron, and P. Colomban, “‘Smart’ Raman/Rayleigh imaging of nanosized SiC materials using the spatial correlation model,” J. Mater. Sci. 39(20), 6183–6190 (2004). [CrossRef]