Efficient light trapping of quasi-inverted nanopyramids in ultrathin c-Si through a cost-effective wet chemical method

Abstract

In this paper, we report quasi-inverted nanopyramids (QIP) for light-trapping in ultrathin c-Si by a cost-effective wet chemical method. The QIP is fabricated by a well-known two-step Ag assisted chemical etching method followed by a post nanostructure rebuilding (NSR) process, lowering the surface area to ∼3.0 times for suppressing surface recombination losses. The comparable average absorptance value of 43 μm c-Si with double-sided QIP to that of 182 μm c-Si with double-sided conventional pyramid in the spectral range of 300–1100 nm demonstrates an over 4.2-fold reduction in material usage. Finally, a simulation model is proposed to explain the superiority of our QIP compared with the periodic inverted pyramid (IP) structure of the same size, presenting a promising method to the mass production of high-efficiency ultrathin c-Si HIT solar cells.

---

Introduction

Until recently, the bulk crystalline silicon (c-Si) solar cells are still the mainstream products in photovoltaic (PV) market.¹ Typically, 160–200 μm c-Si wafers are used to ensure the adequate absorption of incident sunlight, and the c-Si material accounts for 25% or more of the solar cell module cost.² Therefore, there is a continuous effort toward ultrathin c-Si (≤50 μm) solar cells in order to reduce the cost per generated power from a PV module.³ Besides, as ultrathin c-Si solar cells can be encapsulated to light-weight and mechanically flexible PV panels, they can be easily integrated with infrastructure of various shapes and sizes, consequently broaden the application scope of the energy-generating products.⁴ Most importantly, ultrathin c-Si solar cells have potentially higher open circuit voltage (V_oc) and efficiency limits imposed by Auger recombination compared with bulk wafers,⁵ making ultrathin c-Si highly attractive for high performance flexible solar cells.

Generally, specific cell structure and preparation techniques have to be chosen for ultrathin c-Si solar cells to avoid bowing effect and to reduce yield loss due to breakages. Heterojunction with Intrinsic Thin layer (HIT) solar cell, as its abundant advantages including symmetrical cell structure, low temperature process and planar technology, can eliminate cell bowing and increase the yield, thus realizing high efficiency bifacial glass modules.^6,7 Indeed, a lot of research groups have realized HIT solar cells on thin wafers (98 μm (ref. 7 and 8) and 58 μm (ref. 9)) and the obtained bowing-free cells all possess high efficiency, which further demonstrates the potential ability of HIT solar cells in ultrathin c-Si solar cells. However, one big challenge accompanied with the thickness reduction is the inadequate absorption of sunlight due to the indirect bandgap of c-Si.¹⁰ As a result, efficient double-sided surface nanotexture structures should be designed to enhance light absorption for ultrathin c-Si HIT solar cells. For conventional diffusion based silicon solar cells, nanostructure textures tend to worsen the Auger recombination channel by altering the doping profile, and also dead layer can be easily formed during diffusion process, which makes the fabrication of nanostructure based high efficiency solar cells difficult. Besides, the relatively large particle size (several micro-meters) of silver paste makes its intimate contact to nanostructure textures difficult, as a result of which the fill factor (FF) can be not that large. As the bifacial HIT solar cells are always fabricated to realize high efficiency bifacial glass modules, double-sided surface textures (normally the micro-scale upright pyramids) are adopted to enhance the light trapping ability of the HIT solar cells. Its components including emitter and back side field (BSF) can be all directly deposited on the texture surfaces by PECVD or HWCVD. Since the non-diffusion process is adopted to form the junctions in HIT solar cells, the severe Auger recombination caused by diffusion can be eliminated, and the low FF problem induced by the non intimate contact between textures and silver paste can be also alleviated by the direct deposition of the transparent conductive oxide (TCO) film on the surfaces of emitter and BSF. As a result, bifacial HIT solar cell is a promising candidate with our double-sided QIP structures.

In commercial bulk crystalline Si solar cells, light trapping is typically provided by random pyramid texture¹¹ which is obviously not appropriate for ultrathin c-Si for its comparable characteristic size (3–10 μm) to ultrathin c-Si thickness (≤50 μm). The conventional micro-scale pyramid texture tends to behave like hidden cracks to lower the yield of ultrathin c-Si solar cells. Therefore, significant effort has been made to enhance light absorption without sacrificing the yield by nanoscale structures including nanowires,^12–14 nanocones,^15,16 nanoholes,^17,18 nanopillars,^19,20 nanopits,²¹ nanohemispheres,²² nanopores,²³ and inverted nanopyramids.^24,25 Of all these structures, inverted nanopyramids are mostly adopted for its excellent anti-reflection property and small surface area increase (only 1.7 times compared to flat surface), and thus high efficiency silicon solar cells can be obtained by the increase of light absorption and easy surface passivation. However, techniques including lithography²⁴ and laser processes^26,27 usually have to been involved in inverted nanopyramids fabrication, the complexity and additional cost of which hinders its further implement in mass production. Our previous work have demonstrated the ability of cost-effective wet chemical method in realizing nanotexture-based large area (156 × 156 mm²) multi-crystalline silicon (mc-Si) solar cells, the efficiency of which (18.03%) is 0.64% higher than that of conventional mc-Si solar cells (17.39%) at that time.²⁸ Therefore, cost-effective wet chemical method should be developed to realize mask-less inverted nanopyramids fabrication on ultrathin c-Si. Though Cu assisted chemical etching method has recently realized the mask-less fabrication of inverted pyramid and ultralow reflection has been achieved, the large size (1–5 μm) of the obtained structure makes it not suitable for ultrathin c-Si textures.²⁹

In this work, we report a cost-effective wet chemical method for the fabrication of double-sided surface quasi-inverted nanopyramids (QIP) on wafers with different thickness (ranging from 182 μm to 43 μm). The QIP is realized by the well-known metal-assisted chemical etching (MACE) method followed by a post nanostructure rebuilding (NSR) process. The obtained 43 μm c-Si with double-sided surface QIP can achieve an average absorptance value comparable to that of 182 μm c-Si with double-sided surface conventional micro-scale pyramid in the spectral range of 300–1100 nm, amounting to an over 4.2-fold reduction in material usage. The surface area increase is calculated to be only ∼3.0 times compared to flat surface, resulting in the alleviation of surface recombination losses in comparison with as-prepared black silicon (BS). With the NSR time prolonging to 10 min or 15 min, inverted pyramid (IP) with a size larger than 1 μm can be obtained. Finally, a simulation model is proposed to demonstrate that position disorders accompanied with the presence of spikes in our QIP can increase the absorptance compared with the periodic IP structure of the same size, indicating that our wet chemical method possesses more advantages than usually used lithography process.

Experimental section

The QIP fabrication is achieved by NSR treating MACE-based nanohole structure. The schematics for the silicon QIP fabrication is shown in Fig. 1 in a cross-section view. Nanohole structure fabrication is conducted by the two-step MACE method which includes the silver (Ag) particle deposition (Fig. 1(a)), Ag-assisted chemical etching (Fig. 1(b)) and Ag particle removal (Fig. 1(c)). For a solar cell suitable texture fabrication, a post NSR treatment is of great significance for surface area reduction without sacrificing its light trapping effect too much. Typically, p-type (100) Czochralski (CZ) Si wafers (resistivity, 1–5 Ω cm) with different thickness ranging from 182 μm to 43 μm are used as the substrates. Wafer thickness is controlled by NaOH solution etching of 200 ± 10 μm wafers at 90 °C for different time. The etching rate is about 75 μm h⁻¹. For MACE method, the wafers are firstly dipped in a 4.0 M HF solution containing 1.0 mM AgNO₃ for 10 s for electroless Ag plating. Then, the silver coated wafers are immersed in an etching solution consisting of deionized (DI) water, HF (7.52 M) and H₂O₂ (4.60 M) for 60 s to form the nanopore structure. Finally, the as-etched wafers are immersed into the solution containing ammonia and H₂O₂ for 90 s to remove the silver particles remaining in the c-Si. For QIP formation, nanopore structure is restructured in NSR solution (containing H₂O₂ and additive with certain ratio) for different time (5–15 min) at 50 °C after being rinsed in 5 wt% HF aqueous solution for 10 s. The optimized NSR treatment parameters are applied onto c-Si substrate with different thicknesses.

Fig. 1 Schematics of the silicon quasi-inverted nanopyramids fabrication process. (a)–(c) Silicon nanopore structure is achieved by two-step MACE process. (d) Quasi-inverted nanopyramids are realized by a post NSR treatment.

In order to evaluate the superiority of our NSR treatment process in reducing the surface recombination, the samples including polished silicon, as-prepared BS and NSR treated BS are all passivated with 70 nm of Al₂O₃ deposited by Atomic Layer Deposition (ALD) method at 230 °C. A thermal ALD layer is used, using H₂O and trimethylaluminum (TMA) as the oxidant and metal precursor respectively. Prior to ALD deposition, samples are cleaned in standard RCA solutions to remove any organic and metallic contaminant, after which samples are immersed in 5 wt% HF aqueous solution for 10 s to remove any native oxide formed on the surface.

The surface morphology of the samples is examined by scanning electron microscope (SEM, Zeiss Sigma) and atomic force microscope (AFM, Shimadzu SPM9600). The effective minority carrier lifetime is measured using Sinton WCT-120 tester. The reflectance and transmittance measurements are carried out by UV-Vis-NIR spectrophotometer (Shimadzu, UV-3600, with an integrating sphere) in the wavelength range of 300–1100 nm. The absorptance (A) of the samples is calculated from the A = 1 − R − T, where R and T represent the reflectance and transmittance, respectively.

Results and discussion

Fig. 2 compares SEM images of typical BS with and without 5 min NSR treatment. The untreated pores are vertically aligned pores with diameters ranging from ∼15 to ∼70 nm (Fig. 2(a) and (c)), and the pore depth appears to be random, with an average depth of ∼704 nm. When 5 min NSR treatment is adopted, the original small pores turn into randomly arranged large square pits (Fig. 2(b)) and lots of the original small structures are removed. The cross-section SEM image of 5 min NSR treated BS shown in Fig. 2(d) reveals that the angle between the structure facet and the Si (100) surface is 54.7°, indicating the Si (111) plane termination of the facets. The untreated pore cross section becomes QIP due to anisotropic etching character of our NSR solution. Technically, increased photocarrier recombination resulting from drastically increased surface area of nanostructure can deteriorate the cell's performance by reducing short-current density (J_sc) and open-circuit voltage (V_oc), as a result of which the surface area of the obtained QIP should be presented. The surface area calculation here is based on the structure model shown in the inset of Fig. 2(e). Three parameters including length of the QIP (L), height of the spikes (H) and depth of the inverted nanopyramid (D) are needed for the calculation. Neglecting the small amount of irregular structure on the surface, L is distributed in the range of 520–620 nm, with an average value of ∼540 nm based on which the D value is calculated to be ∼381 nm (2D = L × tan(54.7°)). The depth of the QIP (H + D) is estimated from SPM image (Fig. 2(e)) with an average value of 560 nm. Therefore, H can be obtained by simple subtraction (H = 179 nm). The calculated surface area increase is 3.0 times compared to flat surfaces, which is low enough for high efficiency silicon solar cells.³⁰ Generally, dense and deep nanowires or nanoholes reported by many research groups have been adopted to realize the efficient light trapping process in silicon solar cells due to their low reflective property. However, the dense and deep nature of nanowires or nanoholes not only increase the surface area of the nanostructures but also make the intimate contact between the passivation (or emitter) materials and the nanostructure surface difficult during the following deposition process. Because of this shortcoming, a post treatment is always adopted to increase the size of nanostructures to ensure good passivation effect, which at the same time increases reflectance again. Compared with the nanowires or nanoholes, our QIP possesses much larger size and lower surface area, which enables its good surface passivation possible. A lower average reflectance of our QIP than that of commercial available micro-scale pyramids (discussed below) accompanied with an easier surface passivation than as-prepared BS (discussed below) further confirmed this viewpoint.

Fig. 2 Comparison of as-prepared BS and 5 min NSR treated BS. (a) and (b) Surface SEM images of the samples; (c) and (d) cross-section SEM images of the samples; (f) typical macro pictures of different samples; (e) 3-dimensional AFM image of 5 min NSR treated BS. The inset in (b) represents the enlarged surface SEM image of 5 min NSR treated BS, while the inset in (e) shows the structure model of a quasi-inverted nanopyramid.

In order to demonstrate that our NSR process can ease the difficulty of surface passivation, effective minority carrier lifetime at the surface (S_eff) should be presented. The S_eff is usually calculated using the well-known relationship as below:

where τ_eff is the measured effective minority carrier lifetime and τ_bulk is the bulk minority carrier lifetime, and W represents the substrate thickness. As the silicon substrates used here are same, the τ_bulk and W are the same for all the samples, as a result of which the value of τ_eff can be directly used to evaluate the surface passivation effect of the samples. Before the Al₂O₃ passivation, the effective minority carrier lifetimes are almost the same for all the samples (1.59 μs for polished silicon, 1.41 μs for our NSR treated BS and 1.13 μs for as-prepared BS) (shown in Fig. S1(b)†). Surprisingly, the passivation effect of the ALD Al₂O₃ on our NSR treated BS surface (25 μs) is comparable with that on the polished silicon surface (30.8 μs), while the effective minority carrier lifetimes of as-prepared BS is still as low as 1.44 μs, which demonstrates that our NSR process makes the easy surface passivation of the nanostructure possible (shown in Fig. S2(b)†). Fig. 2(f) shows the photographic images of four different samples. To the naked eye, the NSR process alters the color of BS from its original blackish green to dark blue which still obviously possesses a lower reflectance than that of the light blue pyramid texture. After the deposition of Al₂O₃ layer, the color of the samples shown in Fig. S2(a)† is much darker than that of the original corresponding samples shown in Fig. S1(a),† indicating the anti-reflection property of the Al₂O₃ layer.

To evaluate the light trapping ability of BS structure before and after NSR treatment, we measure the surface reflectance as a function of NSR treatment durations in the wavelength range of 300–1100 nm. The average reflectance (R_avg) and average absorptance (A_avg) in this wavelength range can be defined as:³¹

where R(λ) – total reflectance, A(λ) – total absorptance, N(λ) – the solar flux under AM 1.5 standard condition. Fig. 3(a) shows the reflectance spectra of the resulting samples under different treatment time varying from 0 to 15 min. The untreated BS surface shows a low R_avg of ∼5.4%, which indicates that the vertically aligned pores can effectively suppress the reflection in the whole spectra. This is caused by the gradual change of the silicon nanostructure density from air to silicon, resulting in a gradual change of the refractive index of the effective medium from 1 (air) to 3.5 (Si bulk). With the increasement of NSR treatment time, the R_avg rises gradually, which is due to the decrease of depth and density of original pores. When 5 min NSR treatment is adopted, BS possesses a higher R_avg of ∼8.9%, which is still ∼21.2% relatively lower than that of conventional micro-scale pyramid ∼11.3%. With the treatment time extending to 10 min, the size of the original QIP increases to ∼1 μm (shown in Fig. 4(a)) accompanied with a R_avg of 11.2% which is still lower than that of micro-scale pyramid with a tiny gap. A further prolongation of NSR treatment time to 15 min increases the R_avg to 14.8%, which is mainly caused by the removal of the left spikes, as is emphasized by the red dotted circle regions in Fig. 4(a) and (b), with the structure morphology almost unchanging. That indicates the morphology obtained by 5 min NSR treatment possesses more advantages for photo capturing than that of the other NSR treated samples. The variation of the R_avg and the related nanostructure depth with the NSR treatment time shown in Fig. 3(b) clearly reveals that the average reflectance increases in a nearly linear way along with the NSR treatment time, while the average depth of the related nanostructure changes linearly in an opposite way.

Fig. 3 (a) Reflectance spectra of conventional pyramid and BS with NSR durations. (b) The relations of the average reflectance and the nanostructure depth with NSR durations. (c) The absorptance spectra of different thick wafers with and without the QIP on both sides and the absorptance spectra of a 182 μm thick wafer with double-sided surface conventional micro-scale pyramid. (d) Thick dependence of absorptance for different samples.

Fig. 4 (a) Surface SEM images of 10 min NSR treated BS and (b) surface SEM images of 15 min NSR treated BS. The red dotted circles in both images denote the spike structures in the quasi-inverted nanopyramids.

According the above results, our 5 min NSR treated nanostructure possesses excellent antireflection property accompanied with minor surface area increase, which are two of the key factors for high-efficiency silicon solar cells. In order to estimate the potential application of the 5 min NSR treated nanostructure in ultrathin c-Si wafers, the absorptance spectra of different thick wafers (ranging from 182 μm to 43 μm) with and without the nanostructure fabricated on both sides is shown in Fig. 3(c) and the absorptance spectra of a 182 μm thick wafer with double-sided surface conventional micro-scale pyramid are also adopted for comparison. As the light in the spectral range of 300–1100 nm can all be used to excite photocarriers, the A_avg can be adopted to evaluate the absorptance performance of the samples, and the A_avg of the samples with and without the 5 min NSR treated nanostructure as a function of wafer thickness is shown in Fig. 3(d). Overall, both absorptance of the wafers with and without our nanostructures rise with increasing wafer thickness, which is mainly due to the long absorptance depth of the long wavelength light. Compared with flat surface wafers, the nanotextured wafers have 32–37% higher absorptance with the same corresponding wafer thickness. Generally, two-side micro pyramid texture is applied in HIT solar cells to enhance light trapping performance. Here, 43 μm c-Si with double-sided surface quasi-inverted nanopyramids (A_avg = 85.79%) can achieve an average absorptance value comparable to that of 182 μm thick silicon wafer with double-sided surface conventional micro-scale pyramid (A_avg = 85.70%) in the spectral range of 300–1100 nm, amounting to an over 4.2-fold reduction in material usage, as a result of which a comparable high efficiency may be realized on ultrathin c-Si wafer by using our nanotexture.

Compared with the widely researched periodic IP structures which are technologically challenging to achieve and costly to realize over a large area, our mask-less wet chemical method tend to fabricate structures of high position disorders with a much lower cost. As our randomly positioned QIP exhibits a much lower reflectance value in comparison to with that of periodically positioned IP of a nearly same size,²⁵ a numerical finite difference time-domain (FDTD) is used to help understanding this phenomena. The horizontal plane of the silicon is divided into nine square cells and the position of the IP and QIP is periodically or randomly arranged. Fig. 5(a) shows the side and top view of the models. The thickness of the silicon is chosen to be 1 μm. The randomly arranged arrays are randomly generated by Matlab software. The values of the desired parameters including L₁, L₂, L₃ and L₄ are all obtained from the experiment results (L₁ = 707 nm, L₂ = 179 nm, L₃ = 381 nm, L₄ = 540 nm). Fig. 5(b)–(d) show the wavelength dependence of reflectance, transmittance and absorptance spectra of periodically or randomly positioned IP and QIP, respectively. It can be seen in Fig. 5(b) that both random IP and QIP show a lower average reflectance in the wavelength range of 300–1100 nm compared to the corresponding periodic arrays (listed in Table 1), indicating that the presence of randomness can influence the optical characteristics of light-trapping structures. In terms of transmittance, it is almost zero for λ < 450 nm for all the structures, meaning that about half of the light is absorbed before reaching bottom of the silicon absorber. As seen in Fig. 5(d), the absorptance spectrum of the periodic IP and QIP is oscillatory for λ > 500 nm, which suggests the presence of guided modes in the structures.³² The presence of guided mode is always accompanied by the light trapping and lateral propagation of the trapped light. However, the absorptance spectrum of the random IP and QIP is much flatter and stronger, as the structural symmetry is broken by position disorders. Therefore, the structural disorders introduce additional resonances and broadens the existing resonance, and consequently more light coupled into the guiding modes of the randomly positioned structures and hence enhanced light absorption.³³ In addition to position randomness, the spikes appearing in our QIP can also make the optical behavior of the structure different from that of IP under the same position arrangement. Table 1 shows the average values of R (T and A) for periodic (random) IP and QIP in the wavelength range of 300–1100 nm. It is clear that the presence of spikes increase the absorptance from 70.49% to 74.17% for periodic structure and from 70.97% to 74.72% for random structures, which is the result of deeper features due to the appearance of spikes. Therefore, it is the position randomness accompanied with the presence of spikes that finally results the higher absorptance of our random QIP compared with the usually prepared periodic IP, indicating the potential application of our QIP in the mass production of high-efficiency ultrathin c-Si HIT solar cells. Though the simulation results is consistent with the experiment, it is difficult for the simulation results to completely agree with the experimental result because the simulation is on a very thin silicon film while the experiment is on relatively thick silicon.

Fig. 5 (a) Schematic illustration of side views of inverted pyramid and quasi-inverted pyramid and top views of the position of periodically positioned and randomly positioned arrays. (b) Light reflectance, (c) transmittance and (d) absorptance of periodic inverted (quasi-inverted) pyramid and random inverted (quasi-inverted) pyramid.

Table 1 Calculated average value of R (T, A) in the wavelength range of 300–1100 nm

	R (%)	T (%)	A (%)
Periodic IP	25.94	5.19	70.49
Random IP	25.10	3.69	70.97
Periodic QIP	20.64	3.57	74.17
Random QIP	20.08	2.61	74.72

---

Conclusions

In conclusion, we experimentally investigate the light-trapping ability of the QIP on ultrathin c-Si fabricated by the two-step MACE method followed by a post NSR process. The surface area increase is calculated to be only ∼3.0 times compared to flat surface, which is low enough for the fabrication of high-efficiency ultrathin c-Si HIT solar cells. The comparable passivation effect of the ALD Al₂O₃ on our NSR treated BS surface (25 μs) to that on the polished silicon surface (30.8 μs) demonstrates the easy surface passivation of our QIP. The obtained 43 μm c-Si with double-sided surface QIP can achieve an average absorptance value comparable to that of 182 μm c-Si with double-sided surface conventional micro-scale pyramid in the spectral range of 300–1100 nm, amounting to an over 4.2-fold reduction in material usage. With the NSR time prolonging to 10 min or 15 min, inverted pyramid with a size larger than 1 μm can be obtained. A simulation model is proposed to demonstrate that position disorders accompanied with the presence of spikes in our QIP can increase the absorptance compared with the periodic IP structure of the same size, indicating that our wet chemical method possesses more advantages than usually used lithography process. The method we propose here paves a road for the design and fabrication of high-efficiency ultrathin c-Si HIT solar cells with lower cost.

Acknowledgements

This work has been financially supported by National Nature Science Foundation of China (61176062), a Project Funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions, Funding of Jiangsu Innovation Program for Graduate Education KYLX15_0304 (the Fundamental Research Funds for the Central Universities), and the research fund of Jiangsu Province Cultivation base for State Key Laboratory of Photovoltaic Science and Technology (SKLPSTKF201506).

References

1. G. Masson, M. Latour, M. Rekinger, I.-T. Theologitis and M. Papoutsi, Global Market Outlook for Photovoltaics, European Photovoltaic Industry Association, 2013, pp. 12–32.
2. I. Irena, Concentrating solar power, 2012.
3. H. S. Radhakrishnan, R. Martini, V. Depauw, K. Van Nieuwenhuysen, M. Debucquoy, J. Govaerts, I. Gordon, R. Mertens and J. Poortmans, IEEE Journal of Photovoltaics, 2014, 4, 70–77.
4. Q. F. Lin, H. T. Huang, Y. Jing, H. Y. Fu, P. C. Chang, D. D. Li, Y. Yao and Z. Y. Fan, J. Mater. Chem. C, 2014, 2, 1233–1247.
5. M. A. Green, IEEE Trans. Electron Devices, 1984, 31, 671–678.
6. D. Y. Song, J. F. Xiong, Z. Y. Hu, G. F. Li, H. F. Wang, H. J. An, B. Yu, B. Grenko, K. Borden, K. Sauer, T. Roessler, J. H. Cui, H. T. Wang, J. Bultman, A. H. G. Vlooswijk and P. R. Venema, 2012 38th Ieee Photovoltaic Specialists Conference (Pvsc), 2012, pp. 3004–3008.
7. M. Taguchi, A. Yano, S. Tohoda, K. Matsuyama, Y. Nakamura, T. Nishiwaki, K. Fujita and E. Maruyama, IEEE Journal of Photovoltaics, 2014, 4, 96–99.
8. T. Mishima, M. Taguchi, H. Sakata and E. Maruyama, Sol. Energy Mater. Sol. Cells, 2011, 95, 18–21.
9. S. Tohoda, D. Fujishima, A. Yano, A. Ogane, K. Matsuyama, Y. Nakamura, N. Tokuoka, H. Kanno, T. Kinoshita, H. Sakata, M. Taguchi and E. Maruyama, J. Non-Cryst. Solids, 2012, 358, 2219–2222.
10. S. Fonash, Solar cell device physics, Elsevier, 2012.
11. P. Campbell and M. A. Green, J. Appl. Phys., 1987, 62, 243–249.
12. S. Thiyagu, H. J. Syu, C. C. Hsueh, C. T. Liu, T. C. Lin and C. F. Lin, RSC Adv., 2015, 5, 13224–13233.
13. H. J. Syu, S. C. Shiu and C. F. Lin, Sol. Energy Mater. Sol. Cells, 2012, 98, 267–272.
14. M. Foldyna, L. Yu, S. Misra and P. R. I. Cabarrocas, Light trapping enhancement in ordered and disordered silicon nanowire based solar cells, San Diego, CA, 2013.
15. J. S. Li, H. Y. Yu, S. M. Wong, G. Zhang, G. Q. Lo and D. L. Kwong, J. Phys. D: Appl. Phys., 2010, 43, 255101–255107.
16. S. Jeong, E. C. Garnett, S. Wang, Z. G. Yu, S. H. Fan, M. L. Brongersma, M. D. McGehee and Y. Cui, Nano Lett., 2012, 12, 2971–2976.
17. L. X. Yang, Y. P. Liu, Y. Wang, X. Q. Li, W. Chen, Y. Y. Hua, Q. J. Zhang, J. Q. Fu, H. L. Liang, Z. X. Mei and X. L. Du, RSC Adv., 2014, 4, 24458–24462.
18. T. Subramani, C. C. Hsueh, H. J. Syu, C. T. Liu, S. T. Yang and C. F. Lin, RSC Adv., 2016, 6, 12374–12381.
19. J. S. Li, H. Y. Yu, S. M. Wong, G. Zhang, X. W. Sun, P. G. Q. Lo and D. L. Kwong, Appl. Phys. Lett., 2009, 95, 033102.
20. Y. F. Chang, Q. R. Chou, J. Y. Lin and C. H. Lee, Appl. Phys. A: Mater. Sci. Process., 2007, 86, 193–196.
21. J. S. Li, H. Y. Yu, Y. L. Li, F. Wang, M. F. Yang and S. M. Wong, Appl. Phys. Lett., 2011, 98, 021905.
22. Y. L. Li, H. Y. Yu, J. S. Li, S. M. Wong, X. W. Sun, X. L. Li, C. W. Cheng, H. J. Fan, J. Wang, N. Singh, P. G. Q. Lo and D. L. Kwong, Small, 2011, 7, 3138–3143.
23. F. Wang, H. Y. Yu, J. S. Li, X. W. Sun, X. C. Wang and H. Y. Zheng, Opt. Lett., 2010, 35, 40–42.
24. A. Mavrokefalos, S. E. Han, S. Yerci, M. S. Branham and G. Chen, Nano Lett., 2012, 12, 2792–2796.
25. S. Sivasubramaniam and M. M. Alkaisi, Microelectron. Eng., 2014, 119, 146–150.
26. K. Kumar, K. K. C. Lee, P. R. Herman, J. Nogami and N. P. Kherani, Appl. Phys. Lett., 2012, 101, 222106.
27. H. Y. Chen, G. D. Yuan, Y. Peng, M. Hong, Y. B. Zhang, Y. Zhang, Z. Q. Liu, J. X. Wang, B. Cai, Y. M. Zhu and J. M. Li, Appl. Phys. Lett., 2014, 104, 193904.
28. Z. H. Yue, H. L. Shen, Y. Jiang, W. L. Chen, Q. T. Tang, J. L. Jin, T. Pu, J. W. Luo, F. J. Kong, C. B. Rui and J. B. Cai, Appl. Phys. A: Mater. Sci. Process., 2014, 116, 683–688.
29. Y. Wang, L. X. Yang, Y. P. Liu, Z. X. Mei, W. Chen, J. Q. Li, H. L. Liang, A. Kuznetsov and X. L. Du, Sci. Rep., 2015, 5, 16019.
30. J. Oh, H. C. Yuan and H. M. Branz, Nat. Nanotechnol., 2012, 7, 743–748.
31. R. B. Stephens and G. D. Cody, Thin Solid Films, 1977, 45, 19–29.
32. A. Oskooi, P. A. Favuzzi, Y. Tanaka, H. Shigeta, Y. Kawakami and S. Noda, Appl. Phys. Lett., 2012, 100, 181110.
33. Q. G. Du, C. H. Kam, H. V. Demir, H. Y. Yu and X. W. Sun, Opt. Lett., 2011, 36, 1884–1886.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra19819h

---