Transmittance, T(λ), and reflectance, R(λ), spectra were recorded in the wavelength range between 325 and 850 nm with a step resolution of 2.5 nm, employing a Lambda 950 double-beam UV/Vis/NIR spectrophotometer equipped with an integrating sphere and a Spectralon reflectance standard. Both total and diffuse reflectance, R(λ), were measured with 8° incidence angle. The absorptance, A(λ) is then obtained as

Instrument corrected values of T(λ) and R(λ) were used in Equation (1) (Roos, 1993). In order to estimate the fraction of blocked UV light and the solar spectrum weighted T% and R% data for given single values throughout the paper, we calculate a Figure of Merit (FoM). The solar spectrum AM 1.5 G-173 from NREL, ϕ_{AM 1.5 G}(λ), is used as the solar light intensity distribution function, and the weighted average is between 320 and 350 nm for estimating UV blocking and between 350 and 1200 nm for estimating T% and R% as single values, according to

where λ and λ₀ is the wavelength range that is evaluated and P(λ) represents T(λ) or R(λ).

The photoluminescence of the ZnO- and TiO₂-coated samples was measured using a Renishaw Invia Raman microscope employing a 40 × objective lens. The samples were irradiated with a He-Cd laser with a wavelength of 325 nm, and photoluminescence of wavelengths between 330 and 720 nm were recorded. The maximum effect of the laser was 6 mW, but 99 and 90% of the laser flux was filtered away for the ZnO and TiO₂ samples, respectively.

Energy calibration was performed by Raman spectroscopy on non-coated float glass with the same instrument settings. Prior to the measurements, the instrument was calibrated by measuring the 1,332 cm⁻¹ peak of the diamond.

From the spectrophotometry data it is possible to estimate the optical bandgap, i.e., at which wavelength the coated glass starts to absorb most of the radiation acting as a short wavelength cutoff filter. In this study the cutoff wavelength was defined as the wavelength below which <10% of the incoming light is transmitted, i.e., at a somewhat larger wavelength compared to the optical bandgap.

The optical bandgap, E_g, of the coated glasses was analyzed according to a simplified demarcation point analysis which is very similar to the Tauc analysis (Tauc, 1968). The demarcation point can be graphically extracted from A(λ) by approximating the bandgap region below E_g with a fairly horizontal straight line and approximate the region above E_g (the actual UV-edge) with another straight line. The point where these two straight lines intersect is the demarcation point and is a good approximation of the E_g.

Figure 1 shows the transmittance at normal incidence for the glass samples coated with ZnO and TiO₂, respectively. The insets in Figure 1 show optical microscopy images acquired with the Raman instrument, showing that the thin films consist of particulate features with a root-mean-square roughness of about 4–8 nm for the ZnO and 2–10 nm for the TiO₂ samples (see Table 3) which is quite common for the spray pyrolysis technique (Perednis and Gauckler, 2005). The ZnO coatings show the expected rough surface, see Figure 2. The highest roughness is for the thinnest layer, Zn2, while the others have comparable roughness values. In comparison, for the TiO₂ coated samples, see Figure 3, there are larger structures with peak to valley heights in the range 60–150 nm. Also, most samples show clear pits/openings in the film. The depths of these openings are in the range 10–40 nm. Both ZnO- and TiO₂-coated glass show a decrease of the transmittance compared to the uncoated reference sample. The ZnO-coated samples showed larger transmittance in the visible regime than the TiO₂-coated. While the sample coated with the largest amount of ZnO (sample Zn6) showed a transmittance of 74.7%, the transmittance for the sample coated with the largest amount of TiO₂ (sample Ti6) was 66.7%, see Table 3. Glass with ZnO coatings showed a decreased transmittance in the UV regime <350 nm, exhibiting a plateau between 320 and 370 nm. A small absorption peak can also be observed for the reference samples at 380 which can be denoted to tetrahedral configuration of Fe³⁺ (Volotinen et al., 2008). Therefore, we attribute the plateau between 320 and 370 nm for the ZnO-coated samples to be caused by Fe³⁺ but sensitized by the presence of ZnO thin film.

Figure 4 shows the reflectance in the glass samples coated with ZnO and TiO₂. The TiO₂ coated glasses show more reflection as expected due to a larger refractive index mismatch between the glass and the coating. The refractive index for TiO₂ is 2.65 for rutile-TiO₂ (Jellison et al., 1997) or 2.56 anatase-TiO₂ (Schröder, 1928) and for ZnO 2.0 (Jellison and Boatner, 1998), all reported at 589.3 nm as reported by Shannon et al. (2002). The peaks at 375 nm for the ZnO samples coincide with their observed absorption edge in Figure 6 as expected for localized transitions, signaling the presence of defect states, possibly with contribution from Fe³⁺ charge transfer reactions, since the reference sample also exhibits small absorption in this region. It has only been possible to find a zone without that reflection behavior for Zn5. That zone without the mentioned peak corresponded to a place which was far from the internal crack that the sample presented. It has been seen that the samples with that behavior show more spot to spot deviance in reflection measurements. Figure 5 shows the diffuse reflection in the ZnO of the TiO₂-coated samples, respectively. The diffuse reflection of the coatings is <4% for both coatings. The low diffuse reflection is likely to be caused by non-uniform deposited thin films. This means that parts of the diffuse transmitted light that reaches the glass back end will be reflected back into the glass instead of being transmitted through it, similar to a conventional light trapping and back-reflecting cover glass structure (Deubener et al., 2009).

The measured transmittance/reflectance spectra do not exhibit any interference patterns as is expected for thin films with thicknesses >> 100 nm.

Figure 6 shows the change in absorptance near the UV-cutoff in the glass samples coated with ZnO and TiO₂. A trend can be observed of a shift of the UV-cutoff to longer wavelength as more ZnO was deposited. This trend is less clear for TiO₂. The estimated optical bandgaps, using the demarcation point analysis, are shown in Table 3. In the ZnO spectra, two UV edges can be observed: one at about 330 to 340 nm and another one at 375 to 380 nm. The latter gave optical bandgap values in the range of 3.21–3.22, i.e., very similar to the wurtzite form of ZnO. The optical band gap of ZnO in wurtzite and zinc-blende form is 3.22 and 3.12 eV, respectively (Lee et al., 2002). The TiO₂-coated glasses also show a small absorption peak at 380 nm which is caused by the tetrahedral configuration of Fe³⁺ (Volotinen et al., 2008). TiO₂ in rutile, anatase, and brookite form has optical band gaps at 3.00, 3.21, and 3.13 eV, respectively (Dhar et al., 2013). The estimated optical bandgaps in Table 3 are in the range around 3.55 eV, i.e., much influenced by the float glass and its conventional impurity Fe₂O₃. Fe₂O₃ has a considerable effect on the UV absorption and probably overshadows some of the effects that would otherwise be seen by our coatings (Volotinen et al., 2008). This is discussed more in section Discussion of the Effect for PV Modules.

Table 3 shows the UV-blocking characteristics as well as the transmittance of light convertible for Si-PV modules. The-ZnO coated samples show larger percentages of blocked UV light with larger transmission in the visible range compared with TiO₂ coatings. Compared to the reference float glass, the coatings show a relative increase of the FoM (c.f. Equation 2) up to 54.3 and 36.0% of UV-blocking characteristics and a relative decrease of the transmittance by up to 12.3 and 21.8% for ZnO and TiO₂ coatings, respectively.

Figure 7 shows the photoluminescence spectra between 330 and 720 nm of the ZnO- and TiO₂-coated glass substrates. All samples as well as the reference shows a distinct peak around 695 nm. This can probably be attributed to the Fe³⁺ present in the glass since it has shown a peak in photoluminescence at wavelengths around 700 nm when present in silicate glass (Bingham et al., 2007). The reference sample also shows a peak at 390 nm, this could be due to photoluminescence from SiO₂ (Nagata et al., 2004).

All Zn-samples, except Zn2, possess a strong narrow emission peak that can be observed at approximately 377 nm, which is not present in the reference sample. This corresponds well to the photoluminescence observed in ZnO nanorods by Wu et al. (2006) and which can be attributed to a common excitation recombination of ZnO (Studenikin et al., 1998; Kong et al., 2001). The 377 nm PL emission can be an explanation for the sensitization of Fe³⁺ absorption at 320–380 nm, c.f. Figure 6. Sample Zn4 shows a different pattern compared to the reference and the other coated samples in the wavelength interval 550–650 nm. When the photoluminescence from the reference sample was subtracted from sample 4, a peak at 640 nm could be distinguished; this is shown in the inset of Figure 7. This also corresponds well to the photoluminescence observed for ZnO nanorods (Wu et al., 2006), which was also observed by Studenikin et al. (1998) and has been attributed to oxygen interstitial defects in ZnO, i.e., over-stoichiometric ZnO. The opposite, under-stoichiometric ZnO, is indicated to give a green emission at about 510 nm.

The effect that our deposited ZnO and TiO₂ thin films have on the photoluminescence emission makes us confident that these thin films are mostly amorphous. The photoluminescent bands that we interpret as defects and the Fe³⁺ sensitization emissions for the samples are strong and dominate the photoluminescence spectra, see Figure 7. For the ZnO-coated samples it is despite that similar literature data, but with somewhat different experimental circumstances, refer to the ZnO films as mostly crystalline (Kamata et al., 1994; Hosseinmardi et al., 2012; Villegas et al., 2018).

We have shown that UV blocking can be achieved with the cost of reducing the transmittance. Specifically, there was a 75.7% transmittance and a simultaneous reduction of 83.4% integrated transmittance for the sample Zn4, c.f. Table 3. This opens the possibility for maintaining UV protection and gaining useful energy for the PV by lowering the Fe₂O₃ content in the glass without compromising the service lifetime of the PV module. The energy balance for transmitted and useful light for PV modules will be possible to model and optimize in future studies based on information as for instance possible limits for Fe₂O₃ content, cost and efficiency. Furthermore, photon energy down-conversion, i.e., photoluminescence, can be an advantage and a route to utilizing UV light while still not exposing the PV cells to UV light. As for disadvantages, we can list higher reflectivity and scattering. If the surface coating is properly structured, it might not be a serious disadvantage or perhaps even an advantage (Brongersma et al., 2014), as the diffused light contains in fact more photons than the direct light of normal incidence. This is especially valid for façade-applied PV modules where there is in fact very little solar radiation of normal incidence. Another parameter not previously mentioned is the factor of heat. A photon's energy that is not converted into electricity is transformed into heat that in fact lowers the efficiency of the PV module. Beyond the scope of the current paper we would also like to draw the attention to making crystalline ZnO or TiO₂ coatings having similar beneficial properties but with the added value of photocatalysis (Gao and Nagai, 2006; He et al., 2012) and hydrophilic behavior with UV exposure (Watanabe et al., 1999; Sun et al., 2001), thereby giving PV-covered glasses reduced maintenance. Doped ZnO also offers another dimension as a transparent conductive coating offering possible IR reflection for wavelengths non-convertible to energy for PV modules (Deubener et al., 2009).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.