Fabrication of a Silicon Nanowire Solar Cell on a Silicon-on-Insulator Substrate

: This study proposes metal-assisted chemical etching (MAE) as a facile method to fabricate silicon nanowire (SiNW) array structures, with high optical conﬁnement for thin crystalline silicon solar cells. Conventional SiNW arrays are generally fabricated on Si wafer substrates. However, tests on conventional SiNW-based solar cells cannot determine whether the photo-current is derived from SiNWs or from the Si wafer. Herein, SiNW arrays were fabricated on a silicon-on-insulator substrate with a 10- µ m-thick silicon layer for measuring the photocurrent of the SiNW only. The 9 µ m-long p-type SiNW arrays were applied to a solar cell structure fabricated using an n-type H-doped amorphous Si layer, thereby conﬁrming the photovoltaic effect. However, the device exhibited a conversion efﬁciency of 0.0017% because of a low short-circuit current ( J sc ) and a low open-circuit voltage ( V oc ). The low J sc resulted from a high series resistance and high absorption loss from the amorphous Si layer, whereas the low V oc resulted from the high surface recombination velocity of the SiNW array structure. Therefore, reducing the surface recombination of SiNW-based solar cells can improve their conversion efﬁciency. results indicate that the SiNW array structures exhibit antireﬂection rather than light-conﬁnement effects.


Introduction
Crystalline silicon (c-Si) solar cells have a high conversion efficiency and low-fabrication cost [1][2][3][4]. However, the efficiency of c-Si solar cells is limited by Auger recombination. Recently, c-Si solar cells, composed of ultra-thin films, have attracted considerable attention because such cells generate a higher open-circuit voltage (V oc ) than cells with thicker films, thereby reducing Auger recombination and increasing the conversion efficiency [5,6]. However, ultra-thin c-Si solar cells have a low absorption coefficient and, therefore, have a low short-circuit current density (J sc ). Typically, the conversion efficiency of ultra-thin c-Si solar cells is limited by the trade-off between V oc and J sc . Light management techniques are required to achieve a high J sc [7][8][9][10][11], while maintaining a high V oc .
Silicon nanowire (SiNW) arrays have garnered considerable research interest as light-trapping structures because of their strong light-confinement effect [12][13][14]. Accordingly, a SiNW array structure can increase the absorption of a thin absorber layer. Several methods have been proposed for developing SiNW arrays, including laser ablation [15,16], thermal evaporation [17], chemical vapour deposition [18,19], reactive-ion etching [20,21], and metal-assisted chemical etching (MAE) [22][23][24][25]. Among these methods, MAE is widely used because it is simple and cost-effective. In the MAE process, silver (Ag) particles are used as a catalyst to develop large-area aligned SiNW arrays on Si-wafer substrates [26,27]. The photovoltaic effect of the solar cell was confirmed. However, a very low efficiency of 0.0017% was obtained because of very low J sc and V oc values. Low J sc was caused by the high series resistance and absorption loss from the n-type a-Si:H layer. The low V oc was caused by the high surface recombination velocity of the uncoated SiNW surfaces. These results demonstrate that reducing surface recombination is vital for improving the efficiency of SiNW-based solar cells.

Fabrication of SiNW Arrays
SiNWs, Si wafers (p-type, [100], 2-10 Ω cm, 500 µm thickness) were dipped into a HF solution for 1 min to remove the native oxide. To cover the surface with Ag particles via electroless silver plating [31], the Si wafers were immersed in a solution of 4.8 M HF and AgNO 3 for 1 min. The concentration of AgNO 3 in the solution was varied as follows: 0.0025 M, 0.005 M, 0.01 M, 0.015 M, and 0.02 M. To optimise the lengths of the SiNWs, the Si wafers were chemically etched using 4.8 M HF and 0.15 M H 2 O 2 , at room temperature for an etching time of 1-60 min. The etched wafers were subsequently placed in an HNO 3 solution to remove Ag particles. Finally, the oxide layer on the fabricated SiNW array was removed by immersing it in an HF solution. However, the surface pores were observed around the SiNWs. To remove the surface pores, the HNO 3 treatment and HF treatment were conducted three times. The morphology of the prepared SiNW arrays was characterised using field-emission scanning electron microscopy (FE-SEM) (JEOL JSM-7001F). The reflectance of the SiNW on the Si wafer was measured using ultraviolet-visible-near-infrared (UV-Vis-NIR) spectrophotometry (Shimadzu Co., Ltd., SolidSpec-3700). The absorption coefficient of the n-type a-Si layer was measured via spectroscopic ellipsometry (J.A. Woolam, M-2000).

Solar Cell Fabrication
Fabrication procedure of the solar cell structure are in the Figure 1A,B schematises the fabricated solar cell structure. The SOI substrate included a 10 µm-thick c-Si layer (p-type, [100], 2-10 Ω cm) on an oxide insulator layer. The SiNW array structure was etched according to the optimised procedure to provide an array length of 9 µm. To create a p-n junction, a thick n-type a-Si:H layer was deposited onto the p-type SiNW array structure, using radio-frequency (RF) plasma-enhanced chemical vapour deposition. A mixture of 20 sccm SiH 4 and 30 sccm PH 3 was used as the precursor material at 300 • C for 30 min. Subsequently, an 80nm-thick indium tin oxide (ITO) layer, with an area of 0.25 mm 2 , was deposited at the centre of the device via RF sputtering at 100 • C for 10 min. The area outside the 0.25 mm 2 ITO layer was then etched via reactive-ion etching using CF 4 gas to remove the n-type a-Si:H and SiNW array structures. Finally, a grid Al electrode was evaporated onto the ITO layer surface. The solar cells were characterised by I-V measurements under air mass (AM) 1.5G simulated solar illumination at 100 mW/cm 2 and 25 • C and by quantum efficiency measurements (Bunkoukeiki, CEP-25BX). The etched wafers were subsequently placed in an HNO3 solution to remove Ag particles. Finally, the oxide layer on the fabricated SiNW array was removed by immersing it in an HF solution. However, the surface pores were observed around the SiNWs. To remove the surface pores, the HNO3 treatment and HF treatment were conducted three times. The morphology of the prepared SiNW arrays was characterised using field-emission scanning electron microscopy (FE-SEM) (JEOL JSM-7001F). The reflectance of the SiNW on the Si wafer was measured using ultraviolet-visible-near-infrared (UV-Vis-NIR) spectrophotometry (Shimadzu Co., Ltd., SolidSpec-3700). The absorption coefficient of the n-type a-Si layer was measured via spectroscopic ellipsometry (J.A. Woolam, M-2000).

Solar Cell Fabrication
Fabrication procedure of the solar cell structure are in the Figure 1(A) and Figure 1(B) schematises the fabricated solar cell structure. The SOI substrate included a 10 μm-thick c-Si layer (ptype, [100], 2-10 Ω cm) on an oxide insulator layer. The SiNW array structure was etched according to the optimised procedure to provide an array length of 9μm. To create a p-n junction, a thick n-type a-Si:H layer was deposited onto the p-type SiNW array structure, using radio-frequency (RF) plasmaenhanced chemical vapour deposition. A mixture of 20 sccm SiH4 and 30 sccm PH3 was used as the precursor material at 300°C for 30 min. Subsequently, an 80nm-thick indium tin oxide (ITO) layer, with an area of 0.25 mm 2 , was deposited at the centre of the device via RF sputtering at 100°C for 10 min. The area outside the 0.25 mm 2 ITO layer was then etched via reactive-ion etching using CF4 gas to remove the n-type a-Si:H and SiNW array structures. Finally, a grid Al electrode was evaporated onto the ITO layer surface. The solar cells were characterised by I-V measurements under air mass (AM) 1.5G simulated solar illumination at 100 mW/cm 2 and 25°C and by quantum efficiency measurements (Bunkoukeiki, CEP-25BX). (A)

Evaluation of Structure and rReflectance of SiNWs on Si Wafers
To measure the length and reflectance of the SiNWs, the SiNW arrays were fabricated on Si wafers. Figure 2 shows the SEM images of Ag particles fabricated via electroless plating from solutions with different AgNO3 concentration onto c-Si wafer substrates. When the AgNO3 concentration was 0.0025 M (Figure 2A), the Ag particles were deposited onto the Si wafer. When the AgNO3 concentration was increased to 0.005 M ( Figure 2B), Ag was formed over the entire substrate, with an island structure, rather than a particle structure; large gaps were observed between the Ag islands. When 0.0025 M and 0.005 M AgNO3 ( Figure 2C and 2D) were used, Ag particles formed with non-uniform size and shape, and with large gaps among them. When 0.01 M AgNO3 was used ( Figure  2E), Ag nano-particles were formed along with Ag islands. For AgNO3 concentrations >0.02 M, many dendrites formed, precluding the control of the density of Ag particles, as shown in Figure 2E.   Figure 3A and 3B), Si nano-walls were formed, rather than a SiNW structure. These nano-wall formations strongly resembled the gaps among the Ag particles observed in Figure 2A and 2B, indicating that the density

Evaluation of Structure and rReflectance of SiNWs on Si Wafers
To measure the length and reflectance of the SiNWs, the SiNW arrays were fabricated on Si wafers. Figure 2 shows the SEM images of Ag particles fabricated via electroless plating from solutions with different AgNO 3 concentration onto c-Si wafer substrates. When the AgNO 3 concentration was 0.0025 M (Figure 2A), the Ag particles were deposited onto the Si wafer. When the AgNO 3 concentration was increased to 0.005 M ( Figure 2B), Ag was formed over the entire substrate, with an island structure, rather than a particle structure; large gaps were observed between the Ag islands. When 0.0025 M and 0.005 M AgNO 3 ( Figure 2C,D) were used, Ag particles formed with non-uniform size and shape, and with large gaps among them. When 0.01 M AgNO 3 was used ( Figure 2E), Ag nano-particles were formed along with Ag islands. For AgNO 3 concentrations >0.02 M, many dendrites formed, precluding the control of the density of Ag particles, as shown in Figure 2E.

Evaluation of Structure and rReflectance of SiNWs on Si Wafers
To measure the length and reflectance of the SiNWs, the SiNW arrays were fabricated on Si wafers. Figure 2 shows the SEM images of Ag particles fabricated via electroless plating from solutions with different AgNO3 concentration onto c-Si wafer substrates. When the AgNO3 concentration was 0.0025 M (Figure 2A), the Ag particles were deposited onto the Si wafer. When the AgNO3 concentration was increased to 0.005 M ( Figure 2B), Ag was formed over the entire substrate, with an island structure, rather than a particle structure; large gaps were observed between the Ag islands. When 0.0025 M and 0.005 M AgNO3 ( Figure 2C and 2D) were used, Ag particles formed with non-uniform size and shape, and with large gaps among them. When 0.01 M AgNO3 was used ( Figure  2E), Ag nano-particles were formed along with Ag islands. For AgNO3 concentrations >0.02 M, many dendrites formed, precluding the control of the density of Ag particles, as shown in Figure 2E.    nano-walls were formed, rather than a SiNW structure. These nano-wall formations strongly resembled the gaps among the Ag particles observed in Figure 2A,B, indicating that the density of Ag particles had likely increased. Accordingly, when the AgNO 3 concentration was 0.01M, a SiNW structure emerged, and Si nano-walls were divided into several nanowires ( Figure 3C). resembled the gaps among the Ag particles observed in Figure 2A and 2B, indicating that the density of Ag particles had likely increased. Accordingly, when the AgNO3 concentration was 0.01M, a SiNW structure emerged, and Si nano-walls were divided into several nanowires ( Figure 3C).
The images in Figure 3C-3E indicate, that Ag particles were deposited in the gaps between the Ag islands, with increasing AgNO3 concentration because higher AgNO3 concentrations led to the involvement of Ag + ions in dendrite growth. Figure 4A-4C shows cross-sectional SEM images of SiNW arrays formed with AgNO3 concentrations of 0.0025 M, 0.005 M, and 0.01 M. These images indicate, that the SiNW arrays were fabricated vertical to the substrate, and that the concentration of AgNO3 affected only the horizontal structure of the SiNW arrays. Therefore, we selected an AgNO3 concentration of 0.01 M to form SiNW arrays.  The SiNW array length increased with increasing etching time, and a 12 μm-long SiNW array was obtained after 60 min. However, the length of the SiNW array did not linearly increase with respect to the etching time. The SiNW formation mechanism can be interpreted according to the following electrochemical reactions of Si in an HF and H2O2 solution: The images in Figure 3C-E indicate, that Ag particles were deposited in the gaps between the Ag islands, with increasing AgNO 3 concentration because higher AgNO 3 concentrations led to the involvement of Ag + ions in dendrite growth. Figure  of Ag particles had likely increased. Accordingly, when the AgNO3 concentration was 0.01M, a SiNW structure emerged, and Si nano-walls were divided into several nanowires ( Figure 3C). The images in Figure 3C-3E indicate, that Ag particles were deposited in the gaps between the Ag islands, with increasing AgNO3 concentration because higher AgNO3 concentrations led to the involvement of Ag + ions in dendrite growth. Figure   The SiNW array length increased with increasing etching time, and a 12 μm-long SiNW array was obtained after 60 min. However, the length of the SiNW array did not linearly increase with respect to the etching time. The SiNW formation mechanism can be interpreted according to the following electrochemical reactions of Si in an HF and H2O2 solution: 4Ag Si6F -→4Ag SiF 6 2-, After the consumption of HF and H2O2, new HF and H2O2 must diffuse to the bottom of the forming SiNWs. However, the length of etchant migration increases with increasing etching time After the consumption of HF and H 2 O 2 , new HF and H 2 O 2 must diffuse to the bottom of the forming SiNWs. However, the length of etchant migration increases with increasing etching time because the lengths of the SiNWs increase. Hence, the SiNW length depended on the diffusion of the etching solution. Since the MAE reaction is complicated, a slight difference in wire length may occur depending on the experimental environment. These results reveal that the SiNW array length can be reasonably controlled by varying the etching time. However, to obtain a SiNW of arbitrary length, each laboratory should conduct its own preliminary experiments.
To confirm the light-confinement effect of SiNWs, the reflectance of SiNW arrays of different lengths was measured, as shown in Figure 5B. Compared with the reflectance of a bare Si wafer, the optical reflectance of all of the SiNW arrays, for wavelengths <1000 nm, was reduced because of a decrease in the effective refractive index of the SiNW layer and light scattering of the SiNWs. With shorter SiNWs (0.12-7.5 µm), the reflectance decreased to 5-25% compared with 35% for the bare Si wafer in the wavelength range from 600 nm to 1000 nm. Additionally, when the lengths of SiNWs were less than the wavelength of light, the refractive index of the SiNW array could be considered as a middle refractive index layer between the Si and air. Therefore, the reflectance was reduced because the effective refractive index of the SiNW layer decreased. However, the optical reflectance from SiNWs longer than 1.4 µm decreased drastically. In particular, for 7.5 µm-long SiNWs, the optical reflectance below 1000 nm decreased to less than 1%. These results indicate that the SiNW array structures exhibit antireflection rather than light-confinement effects.
Appl. Sci. 2019, 8, x FOR PEER REVIEW 6 of 10 etching solution. Since the MAE reaction is complicated, a slight difference in wire length may occur depending on the experimental environment. These results reveal that the SiNW array length can be reasonably controlled by varying the etching time. However, to obtain a SiNW of arbitrary length, each laboratory should conduct its own preliminary experiments.
To confirm the light-confinement effect of SiNWs, the reflectance of SiNW arrays of different lengths was measured, as shown in Figure 5B. Compared with the reflectance of a bare Si wafer, the optical reflectance of all of the SiNW arrays, for wavelengths <1,000 nm, was reduced because of a decrease in the effective refractive index of the SiNW layer and light scattering of the SiNWs. With shorter SiNWs (0.12-7.5 μm), the reflectance decreased to 5-25% compared with 35% for the bare Si wafer in the wavelength range from 600 nm to 1000 nm. Additionally, when the lengths of SiNWs were less than the wavelength of light, the refractive index of the SiNW array could be considered as a middle refractive index layer between the Si and air. Therefore, the reflectance was reduced because the effective refractive index of the SiNW layer decreased. However, the optical reflectance from SiNWs longer than 1.4 μm decreased drastically. In particular, for 7.5 μm-long SiNWs, the optical reflectance below 1,000 nm decreased to less than 1%. These results indicate that the SiNW array structures exhibit antireflection rather than light-confinement effects.

Properties of Solar Cells with the SiNW Array Structure on an SOI Substrate
The SiNW over 7.5 μm length can achieve strong light confinement. To reduce the influence of the silicon layer as much as possible, 9 μm-long SiNW array structure was adopted in the solar cell. Figure 6A shows the dark and illuminated I-V characteristics of the solar cell with an optimum SiNW absorber layer fabricated on an SOI substrate. A Voc of 110 mV was obtained, indicating that the fabricated solar cell provided a photovoltaic effect. However, the conversion efficiency of the solar cell was extremely low (0.00017%). Therefore, the solar cell parameters were analysed to determine the cause of low efficiency.
A low Jsc of 6.2 μA/cm 2 was obtained because of the high series resistance of the device layer, whose sheet resistance was 5 kΩ/sq. Figure 6B presents a cross-sectional SEM image of the SiNW array structure after deposition of the n-type a-Si:H layer. Herein, a very thick n-type a-Si:H layer

Properties of Solar Cells with the SiNW Array Structure on an SOI Substrate
The SiNW over 7.5 µm length can achieve strong light confinement. To reduce the influence of the silicon layer as much as possible, 9 µm-long SiNW array structure was adopted in the solar cell. Figure 6A shows the dark and illuminated I-V characteristics of the solar cell with an optimum SiNW absorber layer fabricated on an SOI substrate. A V oc of 110 mV was obtained, indicating that the fabricated solar cell provided a photovoltaic effect. However, the conversion efficiency of the solar cell Appl. Sci. 2019, 9, 818 7 of 10 was extremely low (0.00017%). Therefore, the solar cell parameters were analysed to determine the cause of low efficiency.
formed around the surfaces of SiNWs between the n-type a-Si:H and the SiNW array structure. The negative bias enhanced the width of the depletion region and the electric fields, thereby increased the carrier collection, even though the uncoated SiNWs have a high surface recombination velocity. These results reveal that reducing the surface recombination is vital for improving the efficiency of solar cells, fabricated using SiNW array structures as the absorber layer. Our previously reported results indicated that Al2O3 fabricated by atomic-layer deposition is a good passivation material for a SiNW array [30]. If SiNW with Al2O3 is used in a solar cell, the Voc can be improved. However, the insulating Al2O3 film prevents the carriers from migrating to the external circuit. Thus, the inclusion of a passivation film without the SOI substrate is necessary to improve the efficiency of the device. A low J sc of 6.2 µA/cm 2 was obtained because of the high series resistance of the device layer, whose sheet resistance was 5 kΩ/sq. Figure 6B presents a cross-sectional SEM image of the SiNW array structure after deposition of the n-type a-Si:H layer. Herein, a very thick n-type a-Si:H layer was deposited onto the SiNW array structure surface to ensure that a connection formed between each p-type SiNW and n-type layer. In a high-efficiency heterojunction Si solar cell, the thickness of a-Si is on the order of tens of nanometres. Therefore, the very thick a-Si leads to a solar cell, with high resistivity. Moreover, reaming the Si layer under the SiNW arrays contributes substantially to the high resistivity. To improve the I sc , we must consider a new structure. Figure 6C shows the quantum efficiency (QE) of the solar cell under a negative applied bias voltage with respect to the optical wavelength. The QE of the device is extremely low in the short-wavelength region (300-700 nm). Figure 6D shows the absorption coefficient of n-type a-Si:H. The band gap of n-Si:H is 1.7 eV. The optical absorption coefficient of n-type a-Si is high, and the thickness of the n-a-Si layer is approximately 300 nm. Therefore, light with wavelengths shorter than 720 nm was absorbed by the n-a-Si. Because n-type a-Si has a high carrier concentration and high defect density, it did not contribute to power generation. To improve the QE in the short-wavelength region, the thickness of the n-a-Si:H layer should be reduced. Although p-SiNW can absorb light longer than 720 nm, because of sheet resistance, the QE is very low in the absence of a negative bias voltage. Additionally, the QE at each wavelength increased with increasing negative bias voltage. The n-type a-Si:H layer was deposited only on the top of the SiNW array structure because of the high density of SiNWs ( Figure 3C); this configuration strongly limited the concentrations of the n-type layer precursors at the bottom of the SiNWs. Moreover, a 467 nm-wide depletion region, as indicated by the difference between the donor and acceptor concentrations (Na (p-type SiNW) = 5×10 15 cm −3 , Nd (n-type-a-Si) = 1×10 19 cm −3 ), formed around the surfaces of SiNWs between the n-type a-Si:H and the SiNW array structure. The negative bias enhanced the width of the depletion region and the electric fields, thereby increased the carrier collection, even though the uncoated SiNWs have a high surface recombination velocity. These results reveal that reducing the surface recombination is vital for improving the efficiency of solar cells, fabricated using SiNW array structures as the absorber layer. Our previously reported results indicated that Al 2 O 3 fabricated by atomic-layer deposition is a good passivation material for a SiNW array [30]. If SiNW with Al 2 O 3 is used in a solar cell, the V oc can be improved. However, the insulating Al 2 O 3 film prevents the carriers from migrating to the external circuit. Thus, the inclusion of a passivation film without the SOI substrate is necessary to improve the efficiency of the device.

Conclusions
SiNW array structures were fabricated on p-type c-Si wafer substrates, using electroless Ag plating from solutions with various AgNO 3 concentrations and MAE, over different etching times. With increasing AgNO 3 concentration, the resulting Si nano-structures changed from the nano-walls to SiNWs, with decreasing average diameters. The length of the SiNWs depended on the diffusion of the etching solution, such that the length of the SiNWs could be reasonably controlled by adjusting the etching time. The reflectance of the SiNW array structures decreased with increasing SiNW length, indicating that the SiNW array structures with SiNW lengths >7.5 µm exhibited an anti-reflection effect and could be expected to realise solar cells, with a high J sc , while maintaining a high V oc . Using the optimum etching process, we fabricated a solar cell, using an absorber layer composed of a SiNW array structure with a SiNW length of 9 µm etched within a 10 µm-thick p-type c-Si layer on an insulator substrate. A hetero-junction solar cell structure was fabricated by depositing an n-type a-Si:H layer onto the SiNW array surface. The photovoltaic effect of the solar cell was confirmed. However, a low efficiency of 0.0017% was obtained because of the low J sc and V oc values. The low J sc was caused by the high series resistance and by the absorption loss from the n-type a-Si:H layer. Additionally, the low V oc was caused by a high surface recombination velocity of the uncoated SiNW surfaces. These results demonstrate that reducing the surface recombination is vital for improving the efficiency of solar cells with SiNW array structures as the absorber. Thus, the inclusion of a passivation film, without the SOI substrate from the structure, is necessary to improve the efficiency of the device.