Efficiency of silane gas generation in high-rate silicon etching by narrow-gap microwave hydrogen plasma

First, to confirm the effect of cooling the Si sample on the etching rate, we conducted etching experiments with and without a water-cooled stage at W_μw  =  100 W, P_H2  =  8.0 kPa (60 Torr), and Q_H2  =  10 SLM. The Si temperatures measured by the thermocouple attached to the backside of the Si sample with and without the cooling stage were 50 °C and 350 °C, respectively. The actual Si surface temperature would be higher than these temperatures. The etching rates at 50 °C and 350 °C were 10.6 μm min⁻¹ and 2.1 μm min⁻¹, respectively. The etching rate at 50 °C was five times higher than that at 350 °C. This remarkable effect of cooling is consistent with previous results, which were characterized by a negative activation energy [14–18]. This tendency is characteristic behavior of the chemical etching reaction between Si and atomic hydrogen. Therefore, the Si etching phenomena observed in this study are not dominated by thermal evaporation owing to plasma heating but by chemical reaction, although the high input power is supplied to the small volume plasma. The temperature dependence of the negative activation energy has been explained by the presence of the reverse reaction to Si etching – namely, the heterogeneous recombination reaction of surface H atoms to release H₂ molecules [15–17]. For this phenomenon, an effective substrate cooling system is indispensable to obtain high Si etching rates under constant input power. In the reaction studied here, atomic hydrogen impinging on the Si surface terminates unsaturated Si dangling bonds to form silicon hydride species (−SiH, −SiH₂, and  −SiH₃) on the surface. In the case of a high flux of H atoms on the Si surface, breaking the Si bond between the substrate Si atom and the  −SiH₃ group can be induced by a H atom [19, 20]. Moreover, some of the H atoms reaching the Si surface abstract the hydrogen from the H-terminated Si surface by forming a H₂ molecule, and some of the H atoms dissolve and diffuse into the Si bulk. By supplying a high flux of H atoms from plasma, a hydrogen-enriched layer is formed in the Si bulk near the surface, that is, in the subsurface region. Cleavage of Si bonds in the Si bulk crystal is induced by excess incorporated hydrogen and leads to platelet defect creation [21, 22]. This hydrogen-enriched layer near the Si surface seems to play an important role in generating SiH₄ gas. The SiH₄ formation process on the H-terminated Si surface has been reported to go through the following pathways [18, 23, 24]. (1) Reaction between H atoms and surface silyl (−SiH₃) groups: H  +  −SiH₃  →  SiH₄. (2) Reaction between the silyl group and the Si hydride group on the Si surface:  −SiH_x  +  −SiH₃  →  SiH₄  +  −SiH_x−1 (x  =  1, 2, or 3). Formation of the  −SiH₃ group on the Si surface is an important factor for Si etching. In pathway (1), it seems that not only H atoms existing in the plasma gas phase but also H atoms incorporated in the Si subsurface are important for SiH₄ generation. Furthermore, it has been reported that the volatile SiH_x (x  ⩽ 3) radical is directly generated from the Si surface by atomic hydrogen etching [18, 24, 25]. However, the volatile SiH_x (x  ⩽ 3) radical has been observed in low hydrogen flux or high vacuum conditions. Moreover, the radicals have been observed at an elevated temperature of more than 300 °C [18, 24, 25]. However, in this study, hydrogen plasma etching is conducted in a high-pressure hydrogen atmosphere and at a low substrate temperature around room temperature. Therefore, Si etching seems to proceed mainly through the formation of SiH₄ molecules, and SiH_x radical generation in the Si etching process can be ignored, although SiH_x radicals are generated by SiH₄ decomposition in the plasma.

The Si etching phenomenon was not observed when pure He and Ar plasmas were generated instead of H₂ plasma at W_μW  =  100 W and a total process pressure of 20 kPa. From this result, the Si etching reaction observed in this high-pressure atmosphere does not occur by physical sputtering because of high-energy ions in the plasma. The higher process pressure makes the mean free path of the particles shorter. In addition, the collision frequency among the particles becomes higher with increasing process pressure. For these reasons, because the ion energy given by the electric field could be rapidly dispersed on the neutral particles, the high-energy ions that can induce physical sputtering of Si are less likely to be generated in the high-pressure plasma. Therefore, the chemical reaction induced between H atoms and solid Si is the dominant factor in this Si etching process.

Figure 2 shows a typical emission spectrum of the H₂ plasma during Si etching with P_H2  =  5.3 kPa, W_μw  =  600 W, and Q_H2  =  20 SLM. Strong emission lines from H atoms are present at 656 nm (H_α) and 486 nm (H_β). Moreover, the Fulcher bands, which originate from excited H₂ molecules [26], are present at around 600 nm. Emission bands around 414 nm from SiH radicals and an emission line at 288 nm from Si atoms are also present, which show that the SiH₄ molecules generated by Si etching decompose in the H₂ plasma. This re-decomposition in the plasma should be suppressed to obtain higher utilization efficiency of the etched Si.

Zoom In Zoom Out Reset image size

Figure 3 shows that R_m (the weight etching rate) increases almost linearly with increasing W_μw (the input power). In figure 3, P_H2 (the hydrogen pressure) and Q_H2 (the hydrogen flow rate) were 5.3 kPa and 20 SLM, respectively. The input power dependence of the H_α emission intensity obtained from the optical emission spectrum is also plotted in figure 3. The H_α intensity increases almost linearly with increasing W_μw, suggesting that the high etching rate may be related to the high H atom density in the present H₂ plasma.

Zoom In Zoom Out Reset image size

Figure 4 shows R_d (the depth etching rate) and R_m as a function of H₂ pressure (P_H2) with W_μw  =  100 W and Q_H2  =  10 SLM. Although R_d monotonically increases with P_H2, R_m is almost independent of P_H2. Under this serial experiment conditions, the optimum P_H2 to obtain the maximum R_m is about 8.0 kPa. Furthermore, R_m gradually decreases with increasing P_H2 above 8.0 kPa. This conflicting behavior between R_d and R_m with increasing P_H2 seems to be mainly because of the variation of the plasma volume. Figure 5 shows surface profiles of etched Si obtained at P_H2  =  5.3 kPa and 21.3 kPa, and the plasma appearance at each process pressure is also shown. As shown in this figure, the area of the etched Si surface and the bright region of the plasma clearly decrease with increasing P_H2. This means that the plasma becomes more localized around the pipe electrode with increasing P_H2. The power density of the plasma seems to increase with increasing P_H2. Therefore, the maximum etching depth increases by the plasma localization effect.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In this experiment, the plasma volume at W_μw  =  100 W was estimated to be at most 0.4 cm³ from the etching profile and plasma appearance observation. The plasma volume estimation was conducted by assuming that the plasma shape is a circular truncated cone with a hole in the middle of the basal plane as shown in figure 6. The height was determined by the plasma appearance (shown in figure 5(b)). The diameters of the basal plane and the hole were determined from the etching profile (shown in figure 5(a)). The power density of the plasma reached more than 250 W cm⁻³ at W_μw  =  100 W. In high-pressure hydrogen plasma usually used to synthesize diamonds, it has been reported that the gas temperature easily exceeds more than 2000 K, even at input microwave power density of less than 30 W cm⁻³, and increases with increasing input power [27–30]. Considering this increase of gas temperature for the diamond synthesis plasma, the temperature of our Si sample surface may have greatly increased and the etching rate may not have increased because of the negative activation energy for the etching reaction despite the high atomic hydrogen density. However, as shown in figure 3, the etching rate almost linearly increased with increasing W_μw. This linear increase is considered to be because of the high flow rate of H₂ gas and the narrow plasma gap. The narrow-gap plasma used in this study has a relatively large specific surface area compared with conventional diamond synthesis plasma. Therefore, the gas temperature is easily influenced by the boundaries, such as the water cooled electrode. This seems to be one of the characteristics of narrowing the plasma gap.

Zoom In Zoom Out Reset image size

And now, as shown in figure 4, in contrast to R_d, R_m gradually decreases with increasing P_H2 above 8.0 kPa. This gradual decrease of R_m is considered to be because of the decrease of the plasma-exposure area and the increase of the gas residence time in the plasma reaction chamber. The Si temperature may also increase with increasing P_H2, because the plasma gas temperature increases with increasing gas residence time at constant Q_H2. From this result, high P_H2 is important to obtain a locally deeper etching profile and an optimum P_H2 exists to obtain the highest throughput for SiH₄ generation. The fact that the maximum etching depth increases with increasing P_H2 means that the atomic hydrogen density at this local position can be increased by increasing P_H2. According to the above results, a high local etching rate is expected by simultaneously increasing P_H2 and W_μw. High-rate local etching of a Si sample was performed at P_H2  =  33.3 kPa, W_μw  =  500 W, and Q_H2  =  20 SLM. A maximum R_d of about 38 μm min⁻¹ was achieved. Although the etching was localized near the pipe electrode, this etching rate is high and supposed to be related to the high H atom density generated by the narrow-gap high-pressure microwave H₂ plasma [11, 31, 32]. The H atom density (n_H) near the Si surface required to obtain a local etching rate of 38 μm min⁻¹ was estimated by using the following equation:

$$\begin{eqnarray}&&{{\Gamma}_{\text{Si}}}={{R}_{\text{d}}}{{n}_{\text{Si}}}=\eta {{n}_{\text{H}}}\bar{v},\end{eqnarray} \tag{ 2 }$$

where η, $\bar{v}$, Γ_Si, and n_Si are the reaction probability, average velocity of hydrogen molecules, Si flux away from the sample surface, and Si atom density per unit volume, respectively. The reaction probability is defined as the contributing ratio of H atoms impinging on the Si surface for SiH₄ gas formation. In other words, this reaction probability is the etching probability. Based on the literature, the etching probability of Si per H atom is of the order of 10⁻² [17, 18, 33, 34]. Additionally, the average velocity of H₂ molecules is assumed to be 3700 m s⁻¹. Therefore, the estimated H density near the sample surface in the present experiment is of the order of 10¹⁶ cm⁻³. It should be noted that this value is not necessarily equal to the atomic hydrogen density in the plasma bulk. Because the process pressure of our plasma is high and the mean free path is less than a few micrometers, atomic hydrogen transport in the plasma is not only affected by the molecular flow but it is also strongly affected by diffusion. Thus, other measurement methods, such as actinometry [35] and laser absorption [36], should be used to confirm the atomic hydrogen density in the plasma bulk.

In section 3.1, we confirmed that the local etching rate can be increased by simultaneously increasing P_H2 and W_μw. However, from the viewpoint of using this system as an on-site silane generator, it is important to access not only the SiH₄ generation rate but also the efficiency of the reaction system in terms of energy, Si consumption, and H₂ usage.

First, the energy efficiency of Si etching, which is defined by the etching rate per unit input energy, was investigated. Figure 7 shows the input power dependence of the energy efficiency of etching. As shown in figure 7, the energy efficiency decreases with increasing W_μw. In figure 3, the weight etching rate (R_m) almost linearly increases with increasing W_μw, and their relationship can be represented as R_m  =  aW_μw  +  b, where a is a constant and b is the y intercept obtained by linear extrapolation. The energy efficiency (R_m/W_μw) is then inversely proportional to W_μw such that R_m/W_μw  =  a  +  b/W_μw. This result suggests that the energy efficiency is the highest when W_μw is close to the plasma breakdown power. The reasons for this result are unclear because various plasma parameters, such as gas and electron temperatures and plasma density, can vary with increasing W_μw. Moreover, plasma volume expansion is also considered to be one of the reasons, because the plasma part away from the Si surface increases with volume expansion. From the above result, higher input power is a simple way to obtain a larger amount of etching. However, if we require high energy efficiency, it may be better to divide high-power plasma into a number of small-volume plasmas, namely low-power microplasmas. Moreover, the SiH₄ partial pressure in the plasma will not continue to increase with increasing gas residence time but should saturate at an equilibrium SiH₄ partial pressure because the deposition reaction, which is the back reaction of the SiH₄ generation, becomes significant with increasing SiH₄ partial pressure. Therefore, if a large-area plasma is generated that produces a long gas residence time, a part of the input power is used to maintain the plasma, which does not contribute to increasing the SiH₄ concentration. This wasted energy is mainly caused by the excessive gas residence time in the plasma. Hence, using small-volume plasma seems to be one way to improve the energy efficiency of SiH₄ generation.

Zoom In Zoom Out Reset image size

Next, from the viewpoint of material usage, it is important to evaluate the Si utilization efficiency, which is defined as the ratio of the amount of SiH₄ obtained to the amount of etched Si. Figure 8 shows a photograph of the Si surface etched with a relatively low H₂ flow rate (Q_H2  =  5 SLM). In this image, Si particles cover the outside of the circular etched area, which appears dark brown to yellow on the Si surface. The circular etched area is almost coincident with the plasma area. The etched Si surface is mirror-like and almost free of Si particles. This result suggests that particle coarsening very rarely occurs in the narrow-gap hydrogen plasma because of the short residence time and hydrogen termination of Si dangling bonds, even if the nucleation of Si particles is occurred in the plasma. The amount of Si particles is high just outside the etched area (plasma area) and decreases further away from the plasma area. Particle formation outside the plasma may be related to the coarsening of fine particles because of the lack of termination of dangling bonds by atomic hydrogen. In the high-pressure atmosphere, the three-body recombination reaction of atomic hydrogen [37, 38] becomes too significant to ignore. Thus, atomic hydrogen recombines to form H₂ molecules immediately after leaving the plasma. Because the sticking coefficient of the H₂ molecule on the bare Si surface is very small [39], the dangling bond on the fine Si particle surface becomes hard to terminate outside the plasma. Therefore, it seems that the coarsening of the particle becomes significant outside the plasma. Moreover, the vortex gas flow around the pipe electrode and vacuum ultraviolet (VUV) radiation from the H₂ plasma might also influence particle formation in this area. Because the hydrogen pressure of our plasma is high, a continuous VUV radiation spectrum, which induces photodissociation of SiH₄ [40], may be observed – similar to deuterium lamp irradiation [41, 42]. From this typical photograph of the etched Si sample, it is confirmed that the etched Si transforms not only to SiH₄ gas but also to solid Si particles by the plasma reaction, although the amount of generated particles depends on the experimental conditions.

Zoom In Zoom Out Reset image size

To determine the utilization efficiency of the etched Si as SiH₄ gas, we measured the etched Si weight, the weight of generated Si particles, and the SiH₄ concentration in the process atmosphere. Si particles on the Si surface were collected by the swab method and trapped by a fine mesh filter, which was installed in between the plasma chamber and the vacuum pump. The etching products in the gas phase were transported through 1 m-long gas tubing to a gas cell and analyzed by FT-IR spectroscopy. The gas cell had a 3 m optical path by multiple reflections. Figure 9 shows a typical FT-IR absorption spectrum of the etching product generated by the interaction of H₂ plasma with Si wafer. The plasma was generated at P_H2  =  5.3 kPa (40 Torr), W_μw  =  100 W, Q_H2  =  10 SLM, and a plasma gap of 0.6 mm. As shown in figure 9, two predominant peaks are present at 913 cm⁻¹ and 2189 cm⁻¹, corresponding to the ν₃ and ν₄ bands of SiH₄ [43, 44], respectively. In this spectrum, the ν₆ band [45] of Si₂H₆ at around 843 cm⁻¹ is not present. Hence, the density of stable higher order silane molecules, which can drift to 1 m downstream from the plasma, is below the determination limit of FT-IR spectroscopy and the main etching product is determined to be SiH₄ gas. By changing the tube length from 1 m to 5 m it was also confirmed that the SiH₄ concentration detected by FT-IR measurement was independent of the transportation distance. This means that the FT-IR absorption peak used to measure the SiH₄ concentration is not influenced by the reactive hydrogenated silicon species SiH_x (x  =  1, 2, or 3). The SiH₄ molar concentration was determined by the FT-IR absorption peak intensity of the 913 cm⁻¹ peak using a calibration curve obtained by measuring commercial SiH₄ gas with known concentration at the same total pressure as the plasma etching experiment. The commercial SiH₄ and H₂ were supplied to this in-line FT-IR system through each mass flow controller to adjust each mass flow rate. Furthermore, the total pressure in the FT-IR gas cell was also measured by the capacitance manometer attached to the cell. The relationship between the SiH₄ mass flow rate and the infrared absorbance can be revealed by this calibration. The typical calibration curves obtained at the FTIR spectrometer resolution of 4 cm⁻¹ are shown in figure 10. These curves are obtained at Q_H2 of 5 SLM and 10 SLM while keeping the total pressure of the plasma reactor at 8.0 kPa (60 Torr). This calibration should be done whenever there is a change in conductance of the exhaust system, FTIR spectrometer resolution, or the H₂ flow rate. The measured absorbance does not necessarily increase linearly with increasing SiH₄ mass flow rate – mainly due to the spectrometer resolution [46]. In the experiment for SiH₄ generation, by measuring the absorbance at 913 cm⁻¹, the SiH₄ mass flow rate in the FT-IR gas cell can be estimated by comparing with the obtained calibration curve. Therefore, the SiH₄ generation rate can be found in the form of mass flow rate. The utilization efficiency of the etched Si as SiH₄ gas was determined from the amount of generated SiH₄ and the amount of etched Si for an etching time of 10 min. We define the particle yield as the ratio of the particle weight to the etched Si weight. The particle weight includes all of the Si particles accumulated on the etched Si wafer and those trapped by the fine mesh filter. Figure 11 shows the etched Si weight and the weight of generated Si particles as a function of H₂ flow rate (Q_H2). From figure 11, particle generation decreases with increasing Q_H2, while the etched Si weight increases with increasing Q_H2. The particle yield drastically decreases with increasing H₂ flow rate and becomes less than 10% at Q_H2  =  20 SLM, which indicates that more than 90% of the etched Si is converted to SiH₄ gas. From this result, it is suggested that the gas flow rate is an important factor to achieve high utilization efficiency of etched Si.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

To investigate the effects of the plasma area and the gas flow rate on the utilization efficiency, Si etching experiments were performed by changing the diameter of the SS pipe electrode. By varying the pipe diameter, the plasma area and the gas flow velocity simultaneously changed. Figure 12 shows the relationships between Q_H2 and Si etching rate (figure 12(a)) and Q_H2 and SiH₄ generation rate (figure 12(b)) for three different diameter SS pipe electrodes. The data for the 1/2 inch pipe in figure 12(a) were obtained from the same data in figure 11 divided by the etching time of 10 min. To maintain a constant power density of the plasma, the input power W_μw was adjusted to be proportional to the electrode diameter. The wall thickness of three different outer-diameter pipe electrodes was constant at 1 mm. Therefore, the electrode area (S_e) that directly faces the grounded substrate can be calculated as ${{S}_{\text{e}}}=\pi \left(1\times {{D}_{0}}-1\times 1\right)$ (mm²), where D₀ is the outer diameter of the electrode. Then, we chose the input power by assuming that the plasma is generated in a volume of ${{S}_{\text{e}}}\times {{d}_{\text{gap}}}$, where d_gap is the plasma gap. The etching rate increases with increasing Q_H2 for each pipe diameter, except for the 1/8 inch pipe at 20 SLM. Because the gas residence time in the plasma is the shortest for this plot, it has been suggested that a certain gas residence time is required to generate enough H atoms by plasma dissociation [47]. Figure 12(a) also shows that the Si etching rate increases with increasing pipe diameter. This tendency seems to be simply related to the increase of the plasma etching area. Figure 12(b) shows that the pipe diameter has little effect on the SiH₄ generation rate, except for a 1/8 inch pipe at 20 SLM. To understand these tendencies of the effect of the pipe diameter, we analyzed the experimental data in terms of the average gas residence time (τ) in the plasma. In this study, τ was determined by the plasma length (d_l) along the radial thickness direction of the SS pipe wall divided by the average gas velocity (v):

$$\begin{eqnarray}&&\tau =~\frac{{{d}_{\text{l}}}}{v}={{d}_{\text{l}}}\centerdot \frac{{{l}_{\text{oc}}}\centerdot {{d}_{\text{gap}}}\centerdot {{P}_{\text{H}2}}}{{{Q}_{\text{H}2}}\centerdot 101}.\end{eqnarray} \tag{ 3 }$$

Zoom In Zoom Out Reset image size

Here, l_oc is the length of the outer circumference of the pipe, d_gap is the plasma gap, and l_oc·d_gap is the area through which H₂ gas flows out. The unit of P_H2 is kPa. d_l was determined by the width of the etched area on the Si sample. Based on this assumption, the relationship between the utilization efficiency of the source materials and the average gas residence time (τ) is plotted in figure 13. The utilization efficiency of H₂ gas is defined as the amount of H₂ gas contributing to SiH₄ formation divided by the amount of supplied H₂ gas. As shown in figure 13, despite the use of different pipe diameters, the relationship between the utilization efficiency of the etched Si and τ can be well-described by a single curve. The utilization efficiency of the etched Si has a strong dependence on τ and greatly increases with decreasing τ, reaching  >90% for τ  <  13 μs. It was found that utilization efficiencies of  >80% can be easily achieved by decreasing τ below about 15 μs. Conversely, the utilization efficiency of H₂ gas shows little change with τ and is around 0.01%. All of the utilization efficiency trends seem to be related to the fact that, as previously mentioned, the SiH₄ partial pressure in the plasma cannot continue to increase with increasing residence time because deposition on the Si substrate and particle formation become significant with increasing SiH₄ partial pressure. Therefore, an equilibrium partial pressure of SiH₄ gas at which deposition balances the etching phenomena would exist in this process. It seems that the following reactions proceed with τ along the gas-flow direction in the plasma. After the dissociation reaction of H₂ molecules, SiH₄ molecules form by the Si etching reaction with H atoms, and the SiH₄ partial pressure increases and approaches an equilibrium pressure. In the case of long τ, some of the SiH₄ molecules decompose in the plasma and reactive SiH_x radicals are produced. Reactive radicals may condense and form higher order silane related species. As a result, Si re-deposition in the plasma and particle formation outside of plasma markedly increase with increasing τ. In particular, if the re-deposition process of generated SiH₄ only occurs significantly when keeping the equilibrium SiH₄ pressure constant, the utilization efficiencies of Si and H₂ would not be affected. But the energy efficiency would decrease because the change of measured weight after Si etching due to this re-deposition process would not be counted as the etched Si weight. Conversely, if the formation processes of higher order silanes and Si particles occur significantly, the utilization efficiency of Si would markedly decrease without a significant decrease in the H₂ utilization efficiency because the higher order silane generation process (i.e. clustering of Si atoms) in the plasma seems to act as a sink for generated SiH₄. Thus, to keep the SiH₄ equilibrium pressure constant in the plasma, Si etching will continue and the measured etched Si weight will increase. In this case, the SiH₄ partial pressure is constant at the equilibrium pressure and the etched Si weight increases. Therefore, the SiH₄ partial pressure is the same as the SiH₄ concentration in H₂ atmosphere. Thus, it seems that the utilization efficiency of H₂ does not decrease, but Si utilization efficiency decreases with increasing τ.

Zoom In Zoom Out Reset image size

The increase of τ is nearly equivalent to increasing the gas residence time in the reaction chamber. The gas residence time in the reaction chamber is solely determined by the total pressure and gas flow rate, and is independent of the plasma area. For example, the residence time in our chamber (125 cm³ volume) was about 117 ms when the hydrogen flow rate and total pressure were 5 SLM and 8.0 kPa (60 Torr), respectively. This residence time is more than 1000 times longer than τ in the plasma. During the residence time in the reaction chamber, generated SiH₄ molecules might be decomposed by VUV irradiation and react with atomic hydrogen from the plasma. In the case of short τ, generated SiH₄ molecules can leave the plasma before they decompose and the probability of particle formation decreases. The decrease of τ most often leads to a decrease of the residence time in the reaction chamber. Therefore, by also considering the residence time in the chamber, the utilization efficiency of etched Si may greatly increase with decreasing τ. If the input power density varied, the plasma parameters, such as electron temperature, electron density, and gas temperature, would change. Because these parameters are strongly correlated with the reaction rate induced in the plasma, the decay behavior of the utilization efficiency of etched Si would be influenced by changing the input power density.

As discussed above, it should be emphasized that the main factor to obtain high utilization efficiency of Si is not the H₂ gas flow rate but the residence time in the plasma, because the utilization efficiency obtained with different electrode diameters cannot be uniquely plotted as a function of H₂ flow rate but can be plotted as a unified curve as a function of the residence time in the chamber.

From the above results, it can be concluded that the gas residence time in the plasma should be short so H₂ gas can be sufficiently decomposed to achieve high utilization efficiency of etched Si. The utilization efficiency of the H₂ gas is constant at 0.01% and remains at a low level. Hence, in the chemical transport system to obtain a high purity Si film, the H₂ gas should be recycled using a closed reaction system of circulation and regeneration. Improvement of the hydrogen utilization efficiency will be the subject of future work.