1. Introduction

In visible wavelengths, the light absorption efficiency of crystalline silicon (c-Si) base has only 60% due to severe optical reflection loss in the front surface [1]. This considerable weakness seriously limits the photoelectrical performance of c-Si solar cells (SCs). Researchers are therefore continuing to seek more efficient light-trapping strategies, especially advanced surface-textures, to suppress the reflection loss by effectively coupling incident light into devices [2–7].

In bulk c-Si SCs, anisotropic wet chemical etching is usually used to form micro-sized random pyramids as light trapping structure by taking advantage of the different etching rates in the <100> and <111> surfaces of silicon crystals. This is the most mature production process applied in the photovoltaic industry for decades because of its low process threshold and cost-effective performance gain. An overall averaged reflectivity ranging from 11% ~15% for the random pyramids can be obtained by process control of experiment conditions, such as solution temperature and etching time [8]. Although micro-sized pyramid leads to a near Lambertian scattering, it doesn’t absolutely suppress reflection losses especially in the visible wavelength range due to its larger feature size. Advanced light-trapping structures that focusing on light absorption enhancement in sub-wavelength range is therefore of great importance. In addition, single-, double- or multi-layers of anti-reflection coatings (ARCs) can straightforwardly reduce the reflection by gradient refractive index designs [9]. However, the quarter-wavelength (λ/4) single ARC layer performs well only in a limited spectral range. Double or multiple ARCs can suppress reflection over a broad range of wavelengths, but such gain mostly lost after module encapsulation [10].

Recently, more focuses have been devoted on nano-sized two-dimensional photonic crystals, such as inverted nanopyramids [11], nanocones [12,13] and nanopencils [14,15] etc, with the feature size comparable or even smaller than the wavelength of the desired spectral range [16], which show superior light harvesting properties in visible wavelengths compared to random micropyramids [17–20]. However, owing to the difficulties of fabrication and other inherent deficiencies [21–27], these nanoscale textures have been rarely applied to industrial practices. For an example, like nanopencil, which has a vertical morphology leading to pronounced deviations from conformality of the silicon layers and isn’t suitable for the conformal growth of the subsequential passivating films.

In this work, we focused on periodic upright nanopyramids (UNPs) that can be regarded as an excellent-performing nanophotonic light-trapping design. By combining a colloidal lithography and a wet etching technique, we report a simple process for the fabrication of well-ordered periodic nanopyramids in large area. With the results of experiments and simulations, a contrast between surface textures of conventional random pyramids and periodic UNPs has been raised. The UNPs resulted in an excellent light absorption than random pyramids over the whole wavelength range, with an average reflectance of 9.7%. Moreover, the reflectivity was further reduced down to 1.4% in the range of 375-850 nm after the deposition of an ARCs with 85 nm SiN_x, providing an outstanding optical gain for sufficient improvements in solar efficiencies.

2. Experimental details

2.1 Fabrication monolayer of polystyrene (PS) nanospheres

Highly ordered and hexagonally close-packed PS nanosphere templates were fabricated using the Micro-Propulsive Injection method invented by our lab, where the schematic was shown in Fig. 1 [28]. The process can readily reach a throughput of 3000 wafer/h. Aqueous suspension with 5 wt% PS nanosphere colloidals (synthesized by dispersion polymerization) was further diluted in ethanol with a 1:1 ratio, which served as a dispersant. The homogeneous mixture was slowly and steadily injected on the water surface of a bath until a little bit overfilling, resulting in a self-assembly of close-packed hexagonal arrays on the air/water interface. After several minutes waiting, the monolayer of PS nanospheres was transferred to the preset silicon wafers by declining the water level slowly as shown in Fig. 1(b).

 

Fig. 1 Schematic of the fabrication flow for periodic UNP arrays.

Download Full Size | PDF

2.2 Fabrication of two-dimensional periodic UNP arrays

4-inch CZ n-type (100) c-Si wafers (270 ± 10 μm,1~3 Ω·cm) were used in this work. The diameter of the PS nanosphere is 1400 nm and the size was narrowed down to about 1000 nm by a reactive ion etching (RIE) process with O₂/Ar (processes a-c in Fig. 1). The array consisted of reduced nanospheres was subsequently served as a mask for anisotropic etching in a mixture of 20 wt% sodium hydroxide and 20 wt% isopropyl alcohol at 60 °C for 8~12 min to form periodic nanopyramids (processes d-e in Fig. 1). In order to obtain an uniform UNP patterning over large scale, it is important to fast and steadily immerse the Si together with PS nanospheres into the solution before bubbles generating. The random-pyramids textured samples used for comparison were prepared by immersing the polished silicon wafer in a mixture of 2.5 wt% sodium hydroxide and 1.25 wt% flocking additive aqueous solution at 80 °C for 10~15 min. ARC layers with 85 nm-SiN_x were deposited by a plasma enhanced chemical deposition (PECVD) system (PECVD-4200) and passivation layers of AlO_x were obtained by an atomic layer deposition (ALD) system (Lucida D200B).

2.3 Measurements

The morphologies of the samples were observed using a Field Emission Scanning Electron Microscope (FE-SEM, Hitachi S-4800). The optical properties of the sample were carried out using a spectrophotometer (Helios LAB-rc, with an integrating sphere) in 375-1000 nm wavelength range. The minority carrier life-time of the samples were measured by microwave photoconductivity decay method obtained by a lifetime monitor (Semilab, WT-2000).

3. Results and discussions

Figures 2(a) and 2(b) show the cross-sectional scanning electron microscope (SEM) images of the random and periodic pyramids, respectively, clearly revealing a comparison between them. It is easy to find that the periodic UNPs are arranged in a hexagonal manner with the periodicity of about 1400 nm. Besides, the interlinked property of lateral dimension and depth related to the etching process causes an angle of 54.7°, which defines a calculated height of 988 nm for the periodic UNPs. Figure 2(c) shows a top-viewed SEM image of the UNPs in large area. It demonstrates that our method has been highly controlled except for few point defects. The presence of the iridescent color gradient of the 4-inch sample in Fig. 2(d) is caused by the strong light interaction with the periodic sub-wavelength structure under a particular inclined angle. The lateral dimension between the adjacent pyramids can be controlled by varying the etching time, as shown in Figs. 2(e) to 2(h). The 9 min etching produces the close connected UNPs (Fig. 2f) while the insufficient (Fig. 2e) and over-etching (Fig. 2h) cause shrinking in height and pitch of the resultant UNPs.

 

Fig. 2 Top-viewed SEM image for (a) random pyramids and (b) hexagonal UNP arrays, (c) an overview image with a large area of UNP arrays, (d) the optical image of 4-in. c-Si wafer textured with UNP arrays, the SEM images of time-to-time wet etched UNP arrays formation for (e) 5 min, (f) 9 min, (g) 12 min, and (h) 16 min.

Download Full Size | PDF

The antireflection properties of periodic UNP structures are characterized in Fig. 3. Using the same nanosphere array as etching template, periodic UNPs with different heights and pitches were fabricated by tuning the etching durations. The reflectance spectra of the periodic UNPs under different etching times as well as the random pyramids are compared in Figs. 3(a) and 3(b). The spectra indicate that the reflectivity of Si wafer can be greatly suppressed by the presence of UNPs, and it decreases with the increase of the etching time from 5 to 9 min. Depending on the reduced space of nanopyramids and the increased depth of the textures, the 9 min-etched samples with a smaller fraction of planar surface area exhibit superior absorption performance. Once the etching time longer than 9 min, the separation between two adjacent UNPs starts to become larger, i.e., from 50 nm to 600 nm with the etching time ranges from 9 to 16 min, which is because the etching is beyond the polishing critical point as shown in Figs. 2(f)-2(h), so the reflectance is increased. Therefore, the optimal etching time in our experiments is in between 8 to 10 min. Figure 3(c) compared the reflectivity dependence on wavelengths of periodic UNPs and random pyramid structures. Absolutely, the UNPs show a large light absorption enhancement contrasted to the random pyramid in broadband, with the averaged reflection decreasing from 14.5% to 9.7%. This can be explained by that a much higher proportion of light be trapped into the periodic UNP arrays for six passes than random pyramidal scheme [29,30]. Note that the texturing depth of periodic UNPs is only 988 nm, which is much shallower than the random one (3-5 μm) [31]. This advantage contributes to maximized absorption in silicon slab and also offers an unique application in c-Si thin-film SCs. The antireflection property of Si pyramid samples were further enhanced by adopting top-sided SiN_x layer coating. The reflectance of periodic UNPs together with SiN_x is down to 2.5%, in comparison with 4.9% for the random pyramids with an optimized SiN_x coating, as shown in Fig. 3(d). The enhancement via pyramidal structures coated by SiN_x is mainly caused by gradual change of the refractive index and better impedance-matched behaviors [32]. This signifies that the SiN_x coating effect plays a primary role in the antireflection enhancement in the near ultraviolet and visible regions due to the lower reflectance at its centre wavelength [33]. More importantly, the performance of the UNP arrays is typically extraordinary because of the exceptional 1.4% reflectance throughout the sub-wavelength range, in which the usage is especially important as energy sources in photovoltaic application. For UNPs, the period and height are in the same scale of wavelength, so the incident rays are trapped among nanopyramids, which may result in multiply effects including the waveguide resonance modes in the period arrays, diffraction resonance in the case of oblique incidence, coupling effect of light rays between adjacent UNPs, and F-P resonances caused by the reflection from the top and bottom.

 

Fig. 3 Comparison of the reflection curves of the periodic UNP arrays with etching time from 5 to 12 min (a) and column diagram of the average reflectance of periodic UNPs in the range of 375~1100 nm compared with random pyramids marked with blue dashed line (b). The best reflectance spectra of periodic UNP arrays and random pyramids before (c) and after (d) coating with 85 nm SiN_x film.

Download Full Size | PDF

Although the average reflectance of periodic UNPs with SiN_x coating is lower than the random one (especially in the short wave-band), the light-trapping ability of random structure with less reflection loss is better than periodic UNPs in the long wave-band as shown in Fig. 3(d). In order to further understand this observation, we revealed the physical mechanism behind the light-trapping performance by examining the electric field distribution. In this study, the random structure is difficult to reproduce or realize via simulation method, so we only choose two representational sizes (i.e., P = 1.2 μm and 2.0 μm), which can express our problem adequately. As shown in Figs. 4(a) and 4(b), at λ = 600 nm, the electrical field intensity on the front surface for P = 2.0 um case is relatively stronger than that of P = 1.2 um, showing a significant reflectance [34]. With the wavelength increases to 1000 nm, the electrical field intensity for P = 2.0 um case is slightly weaker than that of P = 1.2 um one, revealing a better light-trapping performance. By combining with reflection spectra in Fig. 3(d), we can find that the best response wavelength of UNPs with the period of 1.2 um is about 600 nm. With the size of structure increases, the feature size of scattering increases and the response wavelength also increases (about 1000 nm), leading to better optical performance of the random structure in the long wavelength.

 

Fig. 4 Normalized electric field profiles at λ = 600/1000 nm for (a)/(c) P = 1.2 μm and (b)/(d) P = 2μm.

Download Full Size | PDF

Besides of light trapping properties, the effect of the advanced UNP structures on the passivation must also be investigated before utilization in solar devices. For this purpose, we measured the minority carrier lifetime of the textured samples with AlO_x coating on both sides as shown in Figs. 5(a) and 5(b). The minority carrier lifetime of periodic UNPs shows a little bit higher than the random one, representing better capability for passivation. The periodic UNPs possess the same enhancement in surface area to the random one (1.73 times to flat surface), thus the reduced surface recombination losses may come from the more uniform coating of the passivating layer.

 

Fig. 5 The minority carrier lifetime of random pyramids (a) and periodic pyramids (b).

Download Full Size | PDF

External quantum efficiency (EQE) is the ratio of the number of photo-carriers collected by a SC to the number of photons of a given wavelength that strike the surface of a solar cell from outside. Thus, integrating the EQE of a SC over the entire solar energy spectrum makes it possible to evaluate the amount of current that the cell could produce when exposed to sunlight. In this simulation, the coefficients of surfaces, Auger and bimolecular radiative recombinations are 1 × 10² cm/s, 9.9 × 10⁻³² (2.2 × 10⁻³¹) cm⁶/s and 9.5 × 10⁻¹⁵ cm³/s, respectively. More detailed information about the simulation can be referred to previous publications [35–38]. Based on the above preset recombination parameters, the internal quantum efficiency (IQE) of the two kind of solar cells are high enough to be neglected in its effect on the values of EQE. In this premise, the simulated EQEs can be directly correlated to the optical generation or light-trapping design, so that to properly assess the performances of solar cells only by simulation method.

To reveal the effect of UNPs, EQE-measurements were performed in Fig. 6(a). Traditional c-Si SCs with textured structure and SiN_x ARC layer were selected as the simulation model. The cell with UNPs present higher EQE values than the random pyramid solar cell at wavelengths from 400 to 800 nm, corresponding to a simulated photocurrent density gain of 1.72 mA/cm². In comparison with the UNPs, the random structure with SiN_x ARC shows a photocurrent density gain of 0.37 mA/cm² in the near infrared range. To represent full-spectrum absorption capability, the short-circuit current density (J_sc) based on the light absorption generation for random and periodic pyramid structures was calculated by integrating the measured EQE spectra under AM1.5G condition. Besides, the corresponding current density spectra including random pyramid, periodic pyramid and AM1.5G case (full irradiation) were presented in Fig. 6(b) to have a visual comparison. As a result, the J_sc of the periodic and random pyramids are 41.28 and 39.93 mA/cm², respectively, with an enhancement of 1.35 mA/cm². This simulation is especially relevant as it takes both light behavior and electric property, offering further evidence that UNP textures outperform random ones.

 

Fig. 6 (a) The simulation of external quantum efficiency (EQE) curves of traditional c-Si SCs with periodic UNP arrays (red) and random pyramids (blue), respectively. (b) Simulated curves of current density vs wavelength for both SCs. A normalized AM1.5G solar spectrum is overlaid in black for reference.

Download Full Size | PDF

4. Conclusion

In summary, we demonstrated a simple and cost-effective method featuring alkaline etching and nanosphere lithography to produce periodic Si UNP arrays, as a light trapping structure for silicon solar cells. Compared to the state-of-the-art texture of random pyramids, our experiments and simulations show that the UNP arrays facilitate the coupling of incident sunlight into c-Si layer by the complicated resonant modes and lead to optical absorption enhancement across a wide solar spectrum especially in the sub-wavelength range, achieving an outstanding photocurrent density of 41.28 mA/cm². Our work suggests a proper path toward more efficient surface texturing for high-efficiency c-Si solar cells.