Abstract
The hydrogenated amorphous silicon (a-Si:H)/hydrogenated microcrystalline silicon (c-Si:H) double p-type window layer has been developed and applied for improving microcrystalline silicon-germanium p-i-n single-junction thin-film solar cells deposited on textured SnO2:F-coated glass substrates. The substrates of SnO2:F, SnO2:F/c-Si:H(p), and SnO2:F/a-Si:H(p) were exposed to H2 plasma to investigate the property change. Our results showed that capping a thin layer of a-Si:H(p) on SnO2:F can minimize the Sn reduction during the deposition process which had H2-containing plasma. Optical measurement has also revealed that a-Si:H(p) capped SnO2:F glass had a higher optical transmittance. When the 20 nm c-Si:H(p) layer was replaced by a 3 nm a-Si:H(p)/17 nm c-Si:H(p) double window layer in the cell, the conversion efficiency () and the short-circuit current density () were increased by 16.6% and 16.4%, respectively. Compared to the standard cell with the 20 nm c-Si:H(p) window layer, an improved conversion efficiency of 6.19% can be obtained for the cell having a-Si:H(p)/c-Si:H(p) window layer, with = 490 mV, = 19.50 mA/cm2, and FF = 64.83%.
1. Introduction
Hydrogenated microcrystalline silicon (μc-Si:H) has attracted attentions as a promising material for an absorbing layer in Si-based thin-film solar cells [1–3]. Compared to hydrogenated amorphous silicon (a-Si:H), μc-Si:H has a higher resistance to Staebler-Wronski effect [4]. The effect generally found in amorphous materials could lead to the light-induced degradation [5–7] which deteriorates the long-term film quality as well as the efficiency in solar cells. Moreover, in contrast to the wider bandgap of 1.73 eV for a-Si:H [8], an extended near-infrared (NIR) response arising from the narrower bandgap of 1.1 eV [9, 10] of μc-Si:H film can be attained. However, μc-Si:H has a low absorption coefficient due to its indirect bandgap. A relatively thick μc-Si:H absorber is required for generating sufficient photon-excited carriers. For reducing the thickness of μc-Si:H absorber, μc-:H has been employed as an absorber. Matsui et al. [11] reported that adding Ge into microcrystalline Si-Si network effectively enhanced NIR spectral response. For a μc-:H film having Ge content of 50 at.%, approximately one order of absorption coefficient greater than that of μc-Si:H was observed. The absorption coefficient can achieve 104 cm−1 at 1.5 eV for μc-:H. Matsui et al. [12] have later revealed that the μc-:H single-junction solar cell achieved a cell efficiency of 6.3% with Ge content of approximately 20 at.% in the absorber.
For Si-based thin-film solar cells, the quality of the front transparent conducting oxide (TCO) also significantly influences the cell performance. The textured SnO2:F-coated glass substrates have been widely applied. To promote the crystallization of μc-Si:H films, a highly H2-containing gas mixture of H2 and SiH4 is generally utilized. Although there is a p-type layer on the TCO surface, the energetic hydrogen atom impinging on the surface can further penetrate inyo subsurface growth zone (up to 20 nm) [13–16]. When the SnO2:F is directly or indirectly exposed to H2-containing plasma, Sn reduction could appear and degrade cell performance due to the decreased light absorption [17–19]. In contrast to μc-Si:H film, we have found that adding GeH4 for μc-:H growth had an adverse effect on crystallization. A much higher H2 dilution is required to maintain the crystallization of μc-:H films. Thus, to alleviate unfavorable Sn reduction of SnO2:F surface is one of the key issues for achieving high-efficiency μc-:H cells.
Previous works [20, 21] have indicated that zinc oxide (ZnO) has a higher resistance to H2-containing plasma environment. A thin aluminum-doped zinc oxide (AZO) layer deposited onto SnO2:F surface has been proposed as a protection layer [22, 23]. However, a magnetron sputtering and a post-annealing treatment may generally be required for reducing the defects of the sputtered AZO and improving AZO/SnO2:F interface. In this contribution, we introduced a simple in situ PECVD method to protect the SnO2:F from Sn reduction. The double p-type window layer of a-Si:H/μc-Si:H has been developed to improve cell performance of μc-:H p-i-n single-junction solar cells. We have investigated the effect of H2 plasma on the transmittance and the surface morphology of the SnO2:F. The results demonstrated that capping a thin p-type amorphous silicon (a-Si:H(p)) on SnO2:F can minimize unfavorable Sn reduction during the deposition of microcrystalline films.
2. Experimental Details
In this work, Si-based films were deposited by a 27.12 MHz multichamber plasma-enhanced chemical vapor deposition (PECVD) system with a single chamber process at a substrate temperature of approximately 200°C. The parameters for different growth processes and H2-plasma treatment were added in Table 1. The germane flow ratio and hydrogen ratio for SiGe alloys were defined as = [GeH4]/([GeH4] + [SiH4]) and = [H2]/([GeH4] + [SiH4]), respectively. The hydrogen ratio was varied from 71.4 to 124 with of 0 and 5.06%. The dark and the photoconductivities were measured by an - measurement system under dark and AM1.5G illumination. The standard cell structure was textured SnO2:F-coated glass/μc-Si:H(p)/0.9 μm μc-Si0.88Ge0.12:H/μc-:H(n)/Ag, as shown in Figure 1(a). In our previous work [24], the optimization and details of μc-:H absorber were reported. The optimized μc-:H absorber was deposited at = 5.06% and = 95.2, which corresponded to a Ge content of approximately 12 at.%. The film Ge content was evaluated by an X-ray photoelectron spectrometer. On the other hand, n-type μc-SiOy:H was employed in the cells. N-type μc-:H has been reported for improving cell performance in thin-film silicon solar cells [25, 26], in which there was less parasitic light loss in n-type layer and more long-wavelength reflection at i/n interface. Then, the cells were defined by the metal electrode with a cell area of 0.25 cm2.
For standard cell, 20 nm thick μc-Si:H(p) layer was applied as a window layer. The μc-Si:H(p) layer was deposited by highly hydrogen-diluted SiH4 and B2H6 ([H2]/[SiH4] = 80 and [B2H6]/[SiH4] = 1%) with 1-minute deposition time. The 200 nm thick μc-Si:H(p) layer has a conductivity of S/cm. On the other hand, the 200 nm thick a-SiH(p) deposited with a relatively low H2-to-SiH4 ratio of 2.5 has a conductivity of S/cm. The schematic structure of the improved cell is illustrated as Figure 1(b).
To investigate the change in the optical property, different film stacks on glass substrate including SnO2:F, SnO2:F/μc-Si:H(p) and SnO2:F/a-Si:H(p) were prepared and exposed to the H2 plasma for 1 minute. As can be seen in Table 1, the gas phase concentration of H2 in the process of μc-Si:H(p) is 98.8% which is quite similar to the pure H2 process (100%). Parameters such as pressure and power were kept the same for the H2-plasma treatment and the deposition of μc-Si:H p-layer. This parameter setting should minimize the potential discrepancy between conditions treated by direct H2-plasma exposure and the growth of μc-Si:H p-layer in the cell process. The samples were then measured by an ultraviolet-visible spectrophotometry for optical transmittance. The scanning electron microscope (SEM) was also used to reveal the surface morphologies. The experiments of optical transmittance and characterization of surface morphology changes provided clues for the TCO reduction. Furthermore, we characterized the cell performance by an - measurement system and a solar simulator under AM1.5G illumination. The quantum efficiency (QE) measurement was used to analyze the spectral response in the range of 300–1100 nm.
3. Results and Discussion
3.1. Effect of Hydrogen Ratio on Microcrystalline Si and SiGe Thin Films
Figure 2(a) shows the crystalline volume fraction () of μc-Si:H and μc-:H films as a function of hydrogen ratio. When the hydrogen ratio was increased, an increase in the was observed. With a higher hydrogen ratio in the plasma, more atomic hydrogen promotes the crystallization. In contrast to μc-Si:H film, a higher hydrogen dilution was needed to have the same for μc-:H. With an of approximately 50%, the hydrogen ratios for μc-Si:H and μc-:H growth were 80 and 95.2, respectively. The result suggested that the crystallization of the silicon film is suppressed by adding Ge. The difference in the atomic radius interrupts the ordered crystalline network which reduces the degree of crystallization. Moreover, the GeH3 related species on the film surface during deposition were relatively harder to reach relaxation, which also decreases the crystalline volume fraction. In Figure 2(b), it can be seen that the photo- and dark conductivities of μc-Si:H and μc-:H films increased with raising the hydrogen ratio. With a similar of 50%, μc-Si:H and μc-:H films had the dark conductivities of and S/cm, with the photoconductivities of and S/cm, respectively. Compared to μc-Si:H, the lower photoconductivity and the higher dark conductivity of μc-:H were obtained. The more defective μc-:H films were mainly due to the Ge incorporation which induces Ge-related defects in the films [11, 12].
(a)
(b)
3.2. Effect of H2 Plasma on SnO2:F-Coated Glass Substrate
As discussed in the previous section, silicon film with Ge incorporation requires a relatively higher hydrogen ratio to have appropriate crystallization. To suppress the Sn reduction of SnO2:F due to hydrogen plasma during the deposition of the window layer is beneficial for the development of p-i-n μc-:H single-junction solar cells.
Table 2 shows the optical transmittance of the different film-stacked glass substrates with or without the H2-plasma treatment. In order to quantify the difference, the transmittance at the wavelength of 400 nm and 600 nm was compared. When the textured SnO2:F-coated glass was treated by the H2-plasma treatment for 1 minute, the transmittance decreased by 2.9% and 1.6% at 400 nm and 600 nm, respectively, compared to the fresh SnO2:F-coated glass. This transmittance loss of SnO2:F after H2-plasma treatment has also been demonstrated by Wallinga et al. [18]. For the SnO2:F underwent H2-plasma treatment, the binding energies of Sn in 3d5/2 orbit shifted to 486.5 eV and 484.8 eV, related to suboxides of tin and metallic tin [18, 19]. Therefore, the suboxides and the metallic Sn reduce the transmittance.
Compared to μc-Si:H(p), a much lower H2-to-SiH4 ratio was used for the deposition of a-Si:H(p) layer. As shown in Table 2, after being treated by H2 plasma for 1 minute, the sample having structure of SnO2:F/a-Si:H(p) had higher transmittance, compared to the raw SnO2:F substrate. When the thickness of a-Si:H(p) on SnO2:F increased from 1 nm to 3 nm, the transmittance at 400 nm increased from 69.7% to 71.8% and the transmittance at 600 nm increased from 80.3% to 81.5%. On the contrary, the transmittance decreased to 70.3% and 80.2% at 400 nm and 600 nm, respectively, as the thickness of a-Si:H(p) increased to 5 nm. Considering the trade-off between SnO2:F protection and optical transmission, a 3 nm thick a-Si:H(p) layer was suited for SnO2:F substrate. Moreover, the H2-plasma treated SnO2:F/μc-Si:H(p) had the worst transmittance, compared to the H2-plasma treated SnO2:F-coated glass and the H2-plasma treated SnO2:F/a-Si:H(p). This should be due to the higher hydrogen dilution during the deposition of μc-Si:H(p) and the less dense μc-Si:H film for resisting hydrogen penetration.
Figure 3 shows the optical transmittance of different glass substrates in the wavelength ranged from 300 to 1100 nm. The results show that the transmittance of the H2-plasma treated SnO2:F/3 nm a-Si:H(p) glass substrate was greater than that of the H2-plasma treated SnO2:F glass substrate. For the wavelength shorter than 780 nm, the H2-plasma treated SnO2:F/3 nm a-Si:H(p) glass substrate exhibited a superior transmittance, compared to the H2-plasma treated SnO2:F/3 nm μc-Si:H(p) glass substrate. Depositing a thin layer of a-Si:H(p) could be suitable for a microcrystalline silicon process on SnO2:F based glass substrates.
Figures 4(a), 4(c), and 4(e) show the SEM images of the SnO2:F surface, SnO2:F surface covered with 3 nm thick a-Si:H(p), and SnO2:F surface covered with 3 nm thick μc-Si:H(p) before the hydrogen plasma treatment, respectively. The surface morphologies of SnO2:F surface covered with 3 nm thick films (Figures 4(c) and 4(e)) were both similar to the surface morphology of SnO2:F before hydrogen plasma treatment (Figure 4(a)). To emulate the morphological change after the growth of p-type window layer in the cell process, the samples were treated with 1-minute H2 plasma. As can be seen in Figure 4(b), the surface of the H2-plasma treated SnO2:F had many small particle-like structures with a size of approximately 20 nm, which indicated that the H2 plasma significantly changed the surface morphology. Study had reported that it could be due to the Sn reduction or surface damage by H2 plasma [27]. When the SnO2:F is capped with a 3 nm thick a-Si:H(p) layer followed by the H2-plasma treatment, the nanostructures were effectively decreased, as shown in Figure 4(d). In contrast, Figure 4(f) showed that the H2 plasma still significantly changed the surface morphology of the SnO2:F which was capped with a 3 nm thick μc-Si:H(p) layer. This surface morphology was similar to the surface of the H2-plasma treated SnO2:F. According to these results, a 3 nm thick a-Si:H(p) layer can minimize the effect of H2 plasma on the SnO2:F surface, while maintaining acceptable optical performance. Regarding the surface coverage of the 3-nm thick films on the textured SnO2:F surface, Tsai et al. have reported that the device-quality a-Si:H films were deposited conformally on the substrates with aspect ratio (width/height) ranging from 0.2 to 2 [28]. Since the random pyramidal-like texture of SnO2:F-coated substrates had smoother surface with roughness of approximately 40 nm and correlation length of approximately 175 nm [29, 30], a 3 nm thick a-Si:H(p) or a 3 nm thick μc-Si:H(p) film can effectively cover the SnO2:F surface.
(a)
(b)
(c)
(d)
(e)
(f)
3.3. Improving the Cell Performance of μc-SiGe:H Single-Junction Cells by Capping an a-Si:H(p) Film on SnO2:F
Figures 5 and 6 show the - characteristics and the spectral responses, respectively, of the μc-Si0.88Ge0.12:H p-i-n solar cells with a 0.9 μm active layer. The cell performance of the μc-Si0.88Ge0.12:H p-i-n single-junction solar cells with different p-type window layer is demonstrated in Table 3. The thickness of p-type window layer was kept at 20 nm for comparison. The standard cell with a 20 nm thick single p-type μc-Si:H window layer can achieve a conversion efficiency of 5.31%. Based on the structure, we employed a 3 nm a-Si:H(p)/17 nm μc-Si:H(p) double window layer in the μc-:H p-i-n single-junction solar cell. This cell with the double p-type window layer has an improved cell performance, especially in the short-circuit current (). Compared to the standard cell, the can be significantly enhanced from 16.75 to 19.50 mA/cm2, which was a 16.4% improvement.
As can be seen in Figure 6, the cell with the a-Si:H(p)/μc-Si:H(p) double window layer had a greater quantum efficiency in the wavelength ranging from 300 to 1100 nm. It is also shown in Table 4 that the spectral response of the short wavelength (400 nm) was increased by 19.6% as compared to the cell having only μc-Si:H(p). Moreover, the long-wavelength (800 nm) absorption was increased by 32.4%. The improved spectral response can be due to the less Sn reduction of the SnO2:F surface. More incident light can get into the active layer of the cell and be absorbed to generate photoexcited carriers. Besides, the open circuit voltage () was also enhanced by 10 mV. The larger could be attributed to a lower defect density at the TCO/p interface or in the p-layer, which has less metastable suboxides of tin and metallic tin arising from Sn reduction. The Sn reduction could also decrease work function of SnO2:F which would lead to a larger potential barrier at the TCO/p interface. The SnO2:F can be protected by capping the 3 nm thick a-Si:H(p) layer to minimize the Sn reduction which comes from the sequent films growth of μc-Si:H(p) and μc-Si0.88Ge0.12:H layers with high H2-containing plasma environment. The improved TCO/p interface enhanced the built-in field and facilitated the carrier transport. As a result, the cell with the a-Si:H(p)/μc-Si:H(p) double window layer reached a greater conversion efficiency of 6.19%, which is significantly increased by 16.6% compared to the standard cell structure.
We have further investigated the durability of the double p-type window layer against Sn reduction of the SnO2:F surface. When the H2 plasma/a-Si:H(p)/μc-Si:H(p) structure was implemented as the window layer in the cell, the decreased to 465 mV and the decreased to 14.22 mA/cm2. The drop of may be due to more defects at TCO/p interface. The significant absorption loss in the wavelength ranging from 300 to 1100 nm was revealed by the quantum efficiency measurement, which would lead to the decrease in . When the p-i-n cell was prepared on the direct H2-plasma treated SnO2:F surface, the reduction of SnO2 liberated Sn, which could migrate into p-layer [31]. In addition, oxygen could also diffuse to p-layer and form [31–34]. As a result, these defects led to a built-in potential loss which degraded cell performance of the device. On the other hand, using the a-Si:H(p)/H2 plasma/μc-Si:H(p) structure as the window layer in the cell had only slight degradation of and , as compared to the optimized a-Si:H(p)/μc-Si:H(p) structure. This suggested that the 3 nm thick a-Si:H(p) layer reduced the effect of H2 plasma on SnO2:F surface compared to the H2 plasma/a-Si:H(p)/μc-Si:H(p) structure. In contrast to the optimized cell, lower cell efficiency of 5.89% and of 18.19 mA/cm2 were shown. The result indicates that the thin a-Si:H(p) layer cannot completely eliminate the effect of H2 plasma on SnO2:F surface. Certain amount of hydrogen radical could still affect SnO2:F surface during the growth of a-Si:H(p) layer. However, considering the absorption loss arising from the a-Si:H(p) layer, a thickness of 3 nm is suited for optimizing the μc-SiGe:H single-junction cell performance.
On the other hand, the enhancement in EQE between cells with μc-Si:H(p) and a-Si:H(p)/μc-Si:H(p) can only be partly explained by the difference in transmittance observed between the H2-plasma treated SnO2:F and the 3 nm thick a-Si:H(p) capped SnO2:F as shown in Figure 3. This indicated that the H2 plasma also degraded the electrical property of the SnO2:F substrates. Kambe et al. reported that [35] H2-plasma treated SnO2:F had an increased resistivity and a decreased hall mobility. In addition, the liberated Sn may migrate into the p-layer [31], which causes the degradation of the doped layer. In this study, the significant improvement of the cell performance should majorly arise from the improved TCO/p interface, accompanied with minor optical improvement. In comparison with the state-of-art μc-:H cell with an efficiency of 8.2% ( mA/cm2, V, and FF = 0.651) reported by Matsui et al. [36], the reference cell reported in this work exhibited comparable and FF but lower . The reduction in can be mainly attributed to the difference in front TCO layer and antireflection coating. The commercial SnO2:F-coated substrate is much more chemically unstable in the hydrogen-rich plasma than the ZnO:Ga, which limited the of the reference cell. Furthermore, since the surface texture of commercial SnO2:F-coated substrate is not optimized for the μc-:H cell [36], the chemically etched ZnO:Ga should lead to an enhancement in . In our case, the lack of antireflection bilayer in the reference cell also posted a constraint on in our case [37]. We have demonstrated that the protection of SnO2:F surface from the hydrogen-rich plasma significantly enhanced the from 16.75 to 19.50 mA/cm2 in this work. Further improvement on performance of solar cell is expected as light-trapping technique and optimization on the process condition are performed in the current cell.
4. Conclusions
In conclusion, we have shown that H2 plasma significantly degraded the transmittance and changed the surface morphology of SnO2:F. An adequate thickness of a-Si:H(p) layer has been successfully applied to minimize the harmful H2-plasma effect on SnO2:F surface during the sequent deposition of μc-Si:H(p) and μc-SiGe:H layers. In contrast to the standard μc-Si0.88Ge0.12:H p-i-n single-junction cell with a 20 nm thick μc-Si:H(p) window layer, an improved cell performance can be achieved by employing the 3 nm a-Si:H(p)/17 nm μc-Si:H(p) window layer. Due to an improvement in TCO/p interface, the better spectral response at the wavelength of 300–1100 nm was observed. The corresponding increased from 16.75 to 19.50 mA/cm2. As a result, the conversion efficiency was improved from 5.31% to 6.19% which was a marked increase of 16.6%.
Conflict of Interests
The authors do not have any conflict of interests with the content of the paper.
Acknowledgments
This study was sponsored by the National Science Council in Taiwan under contract nos. MOST-103-3113-P-008-001 and MOST-103-2221-E-009-068. The Instrument Center of the National Science Council has provided support to complete this research.