1. Introduction

Cost pressure is the driving force in silicon photovoltaics. Therefore, all technical innovations aiming at improvements of photovoltaic modules have to be achieved without significantly raising overall fabrication costs.

Untreated front cover glass for photovoltaic modules implies reflection losses of around 4% on each air/glass interface. Two different schemes to enhance the transmission of the cover glass are currently under investigation: antireflective layers and antireflective surface structures. Cover glasses are directly exposed to harsh environmental conditions, and must meet the requirements for a module life time of 25 years.

While in a single antireflective layer the antireflective condition only holds exactly for a certain wavelength and incident angle, multiple layers may be designed to obtain omnidirectional and broadband enhancement of cover glass transmission [1,2]. Long-term stability, however, remains a critical issue for antireflective layers [3]. Thus, a solution based on an antireflective structure (ARS) fabricated directly in the glass surface may be preferable.

The antireflective properties of the moth-eye are based on corneal nipple arrays with sub-wavelength lateral dimensions [4,5]. This surface corrugation creates a gradient of the effective refractive index between air and the bulk material, which reduces reflections at the interface over a broad range of wavelengths and angles of incidence. Due to their smooth index transition, surface structures with high aspect ratios are desirable to obtain low reflectivities, but are often technologically difficult to manufacture.

ARSs based on the moth-eye effect have been fabricated in various materials, e.g. silicon [6–11], polymers [12,13], silica [14,15], and glass [16–18]. However, their fabrication typically employs processing steps such as high-resolution lithography [7,8,11,16,17] or dry-etching [7–11,14–18] which requires expensive and sophisticated equipment and contradicts the demand for low fabrication costs. The fabrication of ARSs in polymers using nanoimprint lithography [12] or hot embossing [13] show promising results in terms of cost efficiency and large-area production, but the long term-stability of polymeric materials as protective covers in outdoor use is questionable. In order to avoid these problems, we chose a fabrication method allowing the formation of ARS in a glass surface by wet chemical etching in hydrofluoric acid (HF). The sub-wavelength lateral structures are created by colloidal lithography, which is recognized as a powerful technique to inexpensively fabricate nanopatterned arrays [19,20].

Most studies on ARSs have been focused on their antireflective properties. In the context of photovoltaic applications of ARSs, both the spectral distribution of the transmission as well as its dependence on the angle of incidence is of interest. Investigations covering both these aspects are relatively rare [13,16]. To characterize the potential improvements in photovoltaic modules, the transmission of ARS glass is compared to that of untreated glass in a wide range of wavelengths and angles of incidence.

2. Fabrication

In this section, the fabrication is described in detail. Low-iron content glass slides (Schott Microcrown glass) were cleaned for 30 minutes in a detergent (5 vol% CQ55, qteck GmbH, in deionized water). Strong ultrasonic agitation was applied, and the bath temperature was set to 65°C. After cleaning, the samples were thoroughly rinsed in deionized (DI) water.

We synthesized spherical SiO2 nanoparticles following the description given by Wang et al. [21]. To obtain 100 g nanoparticle suspension, we mixed 3.2 g tetraethyl orthosilicate (TEOS, 99.999%, Sigma Aldrich) with 87.77 g ethanol (96%, Merck) in a beaker, and 4.25 g of 32% ammonia solution (GPR Rectapur) with 4.78 g DI-water in a second beaker. While agitating the TEOS/ethanol solution with a magnetic stirrer, the H2O/NH3 solution was added. The resulting solution was stirred over night with a closed lid. The nanoparticles were observed in a scanning electron microscope (SEM) revealing a diameter of 250 nm ± 8%. Both numerical simulations and experimental results indicate that diffraction can be expected to be small in ARSs with this periodicity [16].

For the deposition of self-assembled nanoparticle monolayers via dip coating (Bungard dip coater model RDC 15), the nanoparticle concentration was increased by heating the dispersion and evaporating the solvent until the volume of the suspension had decreased to 20% of its original value.

With pulling speeds of around 125 mm/min, the SiO₂ nanoparticles formed monolayers weakly adhered to the glass surface on both sides of the substrate. Contrary to dry etching, wet etching requires a strongly adhering etch mask. Therefore, in a second dip coating process, the nanoparticles were covered with an additional thin SiO₂ layer which partially filled the space between the nanoparticles and fixes their base to the glass surface. The dipping solution was prepared by mixing 10.6 g TEOS in 73.4 g 2-Propanol with an aqueous solution of HNO₃ (3 g of 1 molar HNO₃, diluted with 3 g DI water). This solution was further diluted with 2-propanol in a volume ratio of 1:1. The amount of space filling between the nanoparticles is largely determined by the pulling speed. Higher pulling speeds result in thicker films and, thus, in an increased planarization of the surface. A pulling speed of 125 mm/min was found to give the best results in the subsequent wet etching. Figure 1(a) shows a cross section of a typical compound etch mask.

 

Fig. 1 Fabrication of antireflective structures. (a) Etch mask after the fixation dip. A monolayer of nanoparticles fixed by an additional thin SiO₂ layer covers the glass surface. (b) Etch rate of fixation layer after thermal annealing. The dashed baseline indicates the etch rate of the glass substrate. (c) Etch mask and glass surface after 120 seconds HF-etching. The beginning of nanoparticle detachment and etched protrusions are visible. (d) ARS in glass surface after 135 s HF-etch.

Download Full Size | PDF

In a separate set of experiments, we determined the etch rate of the fixation layer in HF (48 wt.%) diluted with DI water in a volume ratio of 1:100. Etching was stopped by rinsing the samples in DI water. Samples were thermally annealed for 10 minutes at various temperatures in order to optimize the etch resistance. The results are shown in Fig. 1(b). The etch rate drops from 1125 nm/min for unannealed films to 300 nm/min for annealing temperatures above 400°C, which is close to the etch rate of the glass substrate (250 nm/min).

Etching of the glass surface is done in diluted HF as described above. Figure 1(c) illustrates the progress during the etching process. The etching of the glass surface begins after the fixation layer in the regions between the silica nanoparticles is removed. The spherical nanoparticles and the remaining fixation layer at their base form pillars which are isotropically underetched. After 120 s, some of the pillars are already completely underetched. Finally, the nanoparticles detach leaving nano-features reminding of sand dunes in the glass surface [Fig. 1(d)].

Remaining mask residues can be easily removed in an ultrasonic bath. Atomic force microscopy reveals a peak-to-valley height of 120 nm which is close to the expected optimum. For an idealized pinhole mask with distance d between the pinholes, the maximum peak-to-valley height for isotropic etching will be d/2, or 125 nm in the case of our 250 nm particles. We found 135 s to be the optimum etch time. A further increase of the etch time results in flattened peaks and a degradation of the optical properties.

3. Transmission measurement

The transmission of glass can be enhanced by 8% at maximum. Thus, the characterization of the optical properties of the nanostructured glass requires a high-precision measurement of the transmission. Figure 2 shows schematically the setup used in our experiments. By placing the glass samples directly on the surface of a monocrystalline solar cell (Model RSC-M125XL, Sol Expert), we closely mimicked the situation in a real solar module and ensure a high collection efficiency for scattered light. The solar cell was shunted with a low-ohmic resistor and illuminated with chopped light from the output of a monochromator (Triax 550, Jobin-Yvon). A part of the light was directed on a reference silicon photodiode (NT53-377, Edmund Optics) by means of a beam splitter to compensate for drifts in the light power of the halogen light source. The beam was chopped to allow the measurement of the photocurrents through the solar cell and the reference detector with lock-in amplifiers. The experimental error of the transmission measurements with this setup was 0.1%.

 

Fig. 2 Experimental setup for transmission measurement.

Download Full Size | PDF

To evaluate the angular dependence of the transmission, the solar cell was mounted on a rotary stage. The transmission spectra for transversal electric (TE) and transversal magnetic (TM) polarization were measured separately employing a Glan Thompson polarizer. For the characterization of the glass transmission we used the transmission change ΔT defined as ΔT = T_ARS-T, where T_ARS and T are the transmissions of ARS and unstructured glass, respectively, both measured in percent. Data for non-polarized light was obtained by averaging the values for the two polarizations. Figure 3(a) shows the spectral distribution of ΔT for non-polarized light measured at various angles of incidence.

 

Fig. 3 (a) Spectral transmission difference between ARS glass and unstructured glass for non-polarized light at various angles of incidence. (b) Angular dependence of transmission difference for non-polarized light.

Download Full Size | PDF

The maximum value of transmission increase ( + 0.9%) is obtained for normal incidence at wavelengths around 700 nm. The transmission remains increased over the entire spectral range. The drop at short wavelengths indicates the onset of diffraction at the glass surface. An improvement of the transmission may be achieved by decreasing the periodicity of the structure [16]. This may reduce diffraction and result in a higher ΔT at short wavelengths. A general improvement of ΔT requires surface structures with significantly higher aspect ratios. The isotropic HF- etch used in our current fabrication process, however, does not allow higher aspect ratios, and the introduction of a low-cost anisotropic glass etching process may prove difficult. ΔT drops for higher incident angles, and at 30° it changes its sign for short wavelengths indicating a reduced transmission. This tendency is continued at an angle of incidence of 45°, where the transmission is further decreased and the crossover, at which the transmission change is zero, shifts to longer wavelengths. These findings are qualitatively consistent with simulated angular dependencies [16]. Contrary to these simulations, however, a reversal of this trend is found for higher angles. The transmission difference rises again, and the crossover of the transmission change shifts back to shorter wavelengths broadening the spectral range in which the transmission of the ARS glass is enhanced. For 65° the transmission difference is again positive in the whole range from 400 to 1100 nm. The angular dependence is more pronounced for shorter wavelengths as depicted in Fig. 3(b).

Chuang and associates recently pointed out [9] that TE- and TM-polarized light experiences different effective index profiles on the interface of the ARS. The smoothness of the index profile may be significantly reduced for TM-polarized light at higher angles resulting in a degradation of the antireflective properties of the surface. As a consequence, the angular dependence of ΔT is even more pronounced for the different polarizations. This is illustrated in Fig. 4. For TM-polarization, ARS glass shows enhanced transmission only for small angles [Fig. 4(a)]. For higher angles, ΔT is reduced at short wavelengths and approaches zero in the long wavelength region [Figs. 4(b) and 4(c)]. The opposite tendency is observed for TE-polarization, where ΔT continuously increases with the angle of incidence. The net transmission gain of the ARS glass for non-polarized light and high angles [Fig. 3(a)] can thus be entirely ascribed to the contribution of the TE-polarization. For both polarizations, the transmission change is more pronounced at short wavelengths. At an angle of incidence of 65°, the decrease of ΔT in the short wavelength range vanishes completely for TE-polarization [Fig. 4(c)] leaving an entirely positive average transmission change.

 

Fig. 4 Polarization dependent transmission difference between ARS glass and unstructured glass for 15° (a), 45° (b), and 65° (c) incidence.

Download Full Size | PDF

4. Conclusion

In summary, we presented the fabrication of ARS in glass surfaces by wet chemical etching. The formation of the etch mask involves the formation of a of a self- assembled SiO₂ nanoparticle monolayer, which is fixed to the glass surface in a second sol-gel coating process. The employed techniques are scalable, low-cost, and, thus, suitable for high volume production in the cost driven photovoltaic market.

The transmission of ARS glass was characterized with respect to its intended use as cover glass for photovoltaic modules. Compared to untreated glass, the ARS glass shows enhanced transmission over a wide spectral and angular range. The maximum transmission enhancement of 0.9% was found for normal incidence. The angular dependence of the transmission reveals an unexpected increase of the transmission for non-polarized light at large angles. Polarization dependent measurements show that this increase can be entirely ascribed to TE-polarized light and that this increase is counterbalanced by a reduced transmission for TM-polarization.