Effect of hydrogen on low temperature epitaxial growth of polycrystalline silicon by hot wire chemical vapor deposition

    Corresponding author: Fengzhen Liu, liufz@ucas.ac.cn
  • University of Chinese Academy of Sciences, Beijing 100049, China

Key words: polycrystalline siliconhot-wire chemical vapor depositionlow temperature epitaxial growth

Abstract: Polycrystalline silicon (poly-Si) films were prepared by hot-wire chemical vapor deposition (HWCVD) at a low substrate temperature of 525 ℃. The influence of hydrogen on the epitaxial growth of ploy-Si films was investigated. Raman spectra show that the poly-Si films are fully crystallized at 525 ℃ with a different hydrogen dilution ratio (50%—91.7%). X-ray diffraction, grazing incidence X-ray diffraction and SEM images show that the poly-Si thin films present (100) preferred orientation on (100) c-Si substrate in the high hydrogen dilution condition. The P-type poly-Si film prepared with a hydrogen dilution ratio of 91.7% shows a hall mobility of 8.78 cm2/(V·s) with a carrier concentration of 1.3 × 1020 cm-3, which indicates that the epitaxial poly-Si film prepared by HWCVD has the possibility to be used in photovoltaic and TFT devices.


1.   Introduction
  • Polycrystalline silicon film with high mobility and an absence of the Staebler-Wronski effect[1] has many potential applications in photovoltaic[2, 3, 4] and electronic devices[5, 6]. Recently, the epitaxial growth of polycrystalline silicon (ploy-Si) thin film on crystalline seed layer substrates has attracted a lot of attention[7, 8] due to the capability in reducing the material usage and eliminating the costly and polluting wafer production process in the photovoltaic industry. Many techniques such as thermal chemical vapor deposition (CVD)[9], plasma enhanced CVD (PECVD) and hot wire CVD (HWCVD), have been applied to obtain high quality epitaxial growth of Si films.

    Among these techniques, HWCVD has the advantages of a high decomposition rate of the reaction gases, a high film growth rate[10], and no ion bombardment during the deposition process[11]. Furthermore, the abundance of hydrogen atoms[12] generated in the HWCVD process is beneficial for the crystallization of silicon thin films at a relatively low substrate temperature[13]. Large area uniformity of the thin films prepared by HWCVD has been an area of much concern for a long time, so we are happy to know that this problem has finally been solved. In our lab, amorphous silicon thin films with a thickness nonuniformity of smaller than $\pm $5 % (10 $\times$ 10 cm$^{2})$ was successfully prepared on glass substrate by HWCVD[14]. Large area uniformity ($\pm $7.5 %, $\sim$1 m$^{2})$ of the thin films has been realized through the design of the hot filament configuration[15]. Now, HWCVD technology has already been used in the large scale commercial production of Si or C related thin films.

    Considering the advantages, low-temperature microcrystalline and poly-Si thin film deposited by HWCVD have been investigated. During the deposition of high-quality microcrystalline silicon thin film, the substrate temperature is usually below 300 C. In this case, the high hydrogen dilution technique[16, 17, 18] is widely used to promote crystallization, modulate grain orientation[17], and passivate grain boundaries. Unlike the microcrystalline silicon film deposition, the substrate temperature for the epitaxial growth of ploy-Si by HWCVD is normally above 500 C, which is higher than the temperature of the dehydrogenation on the growth surface of silicon film. In the latter case, whether hydrogen is conducive to the epitaxial growth is not fully understood. Further study is needed to reveal the role of hydrogen in the epitaxial growth of poly-Si films by HWCVD.

    In this paper, poly-Si films were prepared using the HWCVD technique. Combined with the analysis of film microstructure and optoelectronic properties, the influences of hydrogen on the epitaxial growth of ploy-Si films are discussed.

2.   Experimental details
  • Poly-Si films were prepared on polished CZ (100) P-type silicon wafer using HWCVD. Before being loaded into the HWCVD reactor, the silicon substrates were treated by 4 % HF solution dipping for 60 s to remove the native oxide layer. The temperature of the hot filaments (3 tungsten wires, 99.95 % purity, 0.5 mm diameter) and the substrate temperature were set at 2000 C and 525 C, respectively. The SiH$_{4}$ and H$_{2}$ mixture was used as the reaction gas. The deposition pressure was fixed at 0.09 Torr. By varying the flow rates of SiH$_{4}$ ($F_{\rm SiH_4})$ and H$_{2}$ ($F_{\rm H_2})$, two series of poly-Si thin films with different hydrogen dilution ratios ($D_{\rm H_2}$ $=$ $F_{\rm H_2}$/($F_{\rm SiH_4}$ $+$ $F_{\rm H_2}))$ were prepared. For series-1 (samples: 1-1, 1-2, 1-3), the total flow rate of the mixing gas (SiH$_{4}$ and H$_{2})$ was fixed at 12 sccm. For series-2 (samples: 2-1, 2-2, 2-3), the flow rate of SiH$_{4}$ was set at 6 sccm. The gas flow rates, hydrogen dilution ratio, and deposition time ($t_{\rm dep})$ are shown in Table 1.

    The thicknesses of samples were determined by a step profiler (DEKTAK XT). Raman scattering spectra were measured using a RM2000 (Renishaw) with a 514 nm Ar ion laser line to evaluate the film crystallization degree. X-ray diffraction (XRD) measurements with the configuration of $\theta $-2$\theta $ and the grazing incidence X-ray diffraction (GIXRD) with an incidence angle of 0.1$^\circ$ was carried out to explore the crystal orientation of the poly-Si films. The film morphologies were observed by SEM (Hitachi S-4800). The reflectance spectra were measured using UV-4000 spectroscopy.

3.   Results and discussion
  • The normalized Raman spectra of the series-1 samples with a constant total gas flow rate and different hydrogen dilution ratio are shown in Figure 1 (a). For comparison, the Raman spectrum of the reference silicon wafer is also included. The transverse optical (TO) phonon mode for crystalline silicon locates at 520 $\pm $ 0.5 cm$^{-1}$ (accurately 519.84 cm$^{-1})$. The film thicknesses are 6.0 $\mu $m, 3.1 $\mu $m, and 3.06 $\mu $m corresponding to $D_{\rm H}$ of 50 %, 75 %, and 91.7 %, respectively. So, the influence of the c-Si substrate on Raman scattering can be ignored considering the penetration depth (0.77 $\mu $m) of 514 nm excitation light in single crystal silicon[19].

    There is no TO mode related to the amorphous Si phase (480 cm$^{-1})$ in all of the Raman spectra, which indicates that series-1 samples are finely crystallized. It implies that the substrate temperature is the main factor affecting the crystallization at our experiment conditions. All the Raman spectra (470-540 cm$^{-1})$ of the series-1 samples can be fitted using three Lorentzian peaks[20] located at $\sim $498 cm$^{-1}$, $\sim $517.5~cm$^{-1}$, and $\sim $520 cm$^{-1}$ as Figure 1(a) shows. The Raman peak located at $\sim $498 cm$^{-1}$ is usually attributed to the grain boundary or small grains[21, 22]. Both of the 517.5 cm$^{-1}$ and 520~cm$^{-1}$ Raman peaks correspond to crystalline silicon TO mode scattering. Nickel et al.[23] pointed out that the 517~cm$^{-1}$ Raman peak relates to tensile stress in un-doped or doped poly-Si. The area percentages of the lower wave number Raman peaks ($\sim $498 cm$^{-1}$ and $\sim $517.5 cm$^{-1})$ decrease with the increase of the hydrogen dilution. However, the area percentage of peak $\sim $520 cm$^{-1}$ increases from 31 % to 38 % as the $D_{\rm H}$ increases from 50 % to 91.7 %. Furthermore, the values of the full width at half maximum (FWHM) of the Raman peak at 520 cm$^{-1}$ decrease with the increase of the $D_{\rm H_2}$ (Figure 1(b)). As the $D_{\rm H}$ increases to above 90 %, the value of the FWHM for the 520~cm$^{-1}$ TO mode scattering peak decreases to 4.5 cm$^{-1}$, which is close to that of the reference crystal silicon in our test setup (4.23 cm$^{-1})$. The increase of the area percentage and the decrease of the FWHM of the 520 cm$^{-1}$ peak indicate that high hydrogen dilution is beneficial for enlarging the grain size.

    Figure 2(a) shows the XRD patterns of the series-1 samples. The inset is an enlarged view of the diffraction peak (400). Sample 1-2 and 1-3 have similar thicknesses. It can be seen from Figure 2(a) that the (400) diffraction peak intensity gradually increases with the hydrogen dilution ratio increasing. However, the intensities of other diffraction peaks gradually decrease. For the sample 1-3 with the highest $D_{\rm H_2}$ of 91.6 %, the 2$\theta $ of the (400) diffraction peak is 69.2$^\circ$, which is consistent with that of the standard crystal silicon.

    Figure 2(b) shows the GIXRD patterns of the series-1 samples. Three peaks of (111), (220), and (311) can be observed for all of the samples. The intensities of the three peaks decrease with the increasing of $D_{\rm H_2}$. The irregular sharp and splitting of the peaks are caused by twins or a slight deviation of the crystal orientation. There is an absence of the (400) peak for all samples, which is due to the extremely narrow width of the silicon (400) XRD peak. The XRD and GIXRD results indicate that the preferential growth orientation of polycrystalline silicon thin film is gradually identical to the crystal silicon substrate (100) and show an epitaxial growth character at the high hydrogen dilution ratio.

    Figure 3(a) depicts the total reflectance spectra of sample 1-3 and polished crystalline silicon wafer. As a typical feature of crystalline silicon, two reflectance peaks related to inter-band transitions at the $X$ paint (275 nm) and along the $\Gamma $-$L$ axis (365 nm) of the Bril1ouin zone appear for sample 1-3. This proves the fairly good crystallization of the prepared poly-Si films from the other side. The effect of the silicon substrate on the reflectance spectrum of poly-Si film can be ignored due to the short penetration depth ($<$ 10 nm) of the light with the wavelength be1ow 380 nm.

    The total reflectance of sample 1-3 is much 1ower than that of the polished single crystalline wafer. This is due to the light trapping effect of surface roughness. In Figure 3(b), the SEM cross-sectional view shows that sample 1-3 presents a columnar growth with a pyramid-like structure on the top surface. The angle of the pyramid (marked in Figure 3(b), $\sim $71$^\circ)$ is consistent with that of the Si-{\{}111{\}} facets (70.5$^\circ)$ which appeared in the textured surface of the Si (100) wafer. This again indicates the (100) orientation growth of the poly-Si thin film. However, the mechanism of the pyramid formation in the deposition process needs further investigation. The native columnar and pyramid structure can be used as a light trapping structure in photovoltaic devices and also provide a possible application in novel radial pn junction solar cells.

    For series-1 samples, the film growth rates reduce from 100 to 23 nm/min as the $D_{\rm H_2}$ increases from 50 % to 91.7 % (Figure 1(b)). The reduced growth rate may be beneficial for the epitaxial growth, which affects our judgment on the role of hydrogen. To exclude the effect of the growth rate on the epitaxial growth of poly-Si by HWCVD, we fixed the flow rate of silane to control the film growth rate and deposited series-2 samples by varying the hydrogen flow rate to change the hydrogen dilution ratio. The deposition rate maintained at 110-120~nm/min for the series-2 samples. The normalized XRD patterns of (400) peaks are shown in Figure 4(a). The (400) peak position (k$\alpha $1) moves to the standard (400) peak position of c-Si (69.2$^\circ)$ as the hydrogen flow rate increases from 7 to 21 sccm. Figure 4(b) shows the GIXRD patterns of series-2 samples. As the $D_{\rm H}$ increases, all of the GIXRD patterns decrease and approach the reference Si wafer condition. Figures 4(a) and 4(b) indicate that hydrogen plays an important role in regulating the grain orientation of the low-temperature poly-Si epitaxial growth by HWCVD.

    To evaluate the electrical property of the poly-Si films, a boron doped poly-Si film was prepared on n-type c-Si wafer with the flow rates of SiH$_{4}$, hydrogen diluted B$_{2}$H$_{6}$ (0.5 %) and H$_{2}$ of 1 sccm, 2 sccm and 9 sccm, respectively. The Hall measurement was carried out on an Accent HL5500 Hall system (Bio-Rad Laboratories, Hercules, CA, USA). A Hall mobility of 8.78 cm$^{2}$/(V$\cdot $s) is obtained with a carrier concentration of 1.3 $\times$ 10$^{20}$ cm$^{-3}$, which is consistent with the previous reported value (about 10 cm$^{2}$/(V$\cdot $s)) for poly-Si by thermal CVD under 650 $\du^{[24]. The electrical property implies that the epitaxial poly-Si prepared by HWCVD at the relatively low substrate temperature of 525 C is possible for photovoltaic and TFT devices applications.

4.   Conclusion
  • Poly-Si thin film epitaxial growth was carried out at a low substrate temperature of 525 C using the HWCVD technique. The Raman scattering and reflectance spectra indicate that high crystallized poly-Si thin film was achieved under 525 \du{}. XRD, GIXRD and SEM measurements show that H modulates the grain orientation in the growth of polycrystalline silicon thin films and the high $D_{\rm H_2}$ sample ($D_{\rm H_2}$ $=$ 91.7 %) exhibits an epitaxial growth feature. The P type poly-Si film prepared with $D_{\rm H_2}$ of 91.7 % has a hall mobility of 8.78 cm$^{2}$/(V$\cdot $s) with a carrier concentration of 1.3 $\times$ 10$^{20}$ cm$^{-3}$, which indicates that the epitaxial poly-Si film by HWCVD has the possibility to be used in photovoltaic and TFT devices.

    Acknowledgments A portion of this work is based on the data obtained at 1W1A, BSRF. The authors gratefully acknowledge the assistance of scientists of the diffuse X-ray scattering station during the experiments.

Figure (4)  Table (1) Reference (24) Relative (20)

Journal of Semiconductors © 2017 All Rights Reserved