© F. Riesebeck et al., Published by EDP Sciences, 2024

1 Introduction

Combining perovskite based solar cells with conventional silicon solar cells in a tandem cell setup enables higher efficiencies due to covering a higher range of wavelengths in the solar spectrum. Perovskite-silicon tandem solar cells have strongly advanced and recently even reached 33.9% efficiency [1]. To fully exploit the efficiency potential of this technology, textured interfaces are required to minimize optical losses. These textures must be compatible with the processing of perovskite, which is challenging, particularly when the perovskite is processed from solution. Pioneering works in monolithic perovskite-silicon tandem solar cells used silicon bottom cells with a nanopyramidal surface texture and successfully combined them with a blade-coated [2] or spin-coated [3] perovskite top cell. It was also shown that perovskite solar cells can be fabricated via spin-coating on various other textured substrates, such as periodic sinusoidal nanotextures and inverted pyramids, without deterioration of the electronic performance [4]. Later, we showed that the implementation of sinusoidally nanotextured interfaces enables 29.8% efficiency in perovskite-silicon tandem solar cells by not only reducing optical losses but also improving the optoelectronic properties in terms of open-circuit-voltage and increasing the fabrication yield by improved wetting of the perovskite on textures [5]. The interplay between perovskite solution processing and texture silicon bottom cell received more attention [6]. A non-ideal wetting of perovskite precursor solution was found to be an issue particularly if self assembling monolayers (SAMs) are used, as reported by various groups in the recent past [7–10]. However, several studies have also been published indicating a better wettability of the perovskite solution and a better layer growth if a textured substrate was used [5,11,12]. While the correlation between textures and reduced optical losses is well understood, there are only a few works investigating the in-situ perovskite layer growth from solution on textured surfaces. Here in this study, we investigate the perovskite layer growth from solution on textured substrates using in-situ photoluminescence (PL) measurements and various microscopic characterization techniques.

2 Experimental

2.1 Nanoimprint lithography

The texturing of the glass substrate covered by a resist layer is done by nanoimprint lithography (NIL), as structuring substrates with conventional lithography methods is very time- and cost intensive. The NIL-process enables lithography with a resolution down to 5 nm while keeping the costs low and the output volume high.

The NIL-process starts with a so-called “master structure” that has the desired texture, which has been produced by other lithography processes such as interference lithography or photolithography or a combination of different methods.

First, an inverted version of the master structure is prepared using a polydimethylsiloxane (PDMS) polymer. For this purpose, a commercially available PDMS polymer precursor and a suitable catalyst are mixed in a ratio of 9:1 and degassed through a desiccator to avoid air entrapment in the later stamp. Figure 1a shows how the viscous PDMS is then poured onto the master structure and thermally annealed at 70 °C for 30 min. Subsequently, the negative stamp can be removed from the master structure without leaving any residue. Since this is a non-destructive process, the original structure remains intact, allowing the process to be performed multiple times as needed. The stamp can now be used in the second process step to transfer the structure to any substrate. The substrates used in this work are 5 × 5 cm² of Schott T-eco 264® glass with a thickness of 1.1 mm covered by a UV-hardening resist. Here we use the hybrid polymer OrmoComp® (Micro Resist Technologies), which is mixed with the thinning agent OrmoThin® (Micro Resist Technologies) in a ratio of 1:2 to adjust the film thickness. With this ratio, a film thickness of ≈ 1 μm is achieved.

Usually, a layer thickness in the height of the texture is sufficient. However, since different textures with different texture heights are processed in this work, a uniform thickness of the resist layer is chosen for comparison purposes. This has no difference in the optical properties of the material. Figure 1b shows the steps to transfer the inverted texture in the stamp to the substrate. The resist is applied to the glass substrate in a spin coating process at 3000 rotations per minute (rpm) for 30 s. Afterward, the substrate must be heated for 2 min at 80 °C on a hotplate in a pre-bake process to remove any remaining solvents after spin coating. The stamp is now carefully rolled onto the target substrate with the aid of a metal cylinder and pressed in. For curing, the stack of substrate, resist and stamp is illuminated for 5 min under a commercial UVB broadband lamp. PDMS is a transparent polymer, which allows the curing of the underlying resist. The PDMS stamp is then carefully removed from the substrate. The stamp can be used for up to ten substrates before a new stamp is recommended.

Fig. 1 Fabrication of textured glass substrates using nanoimprint lithography. (a) Replication of the master structure in an inversed PDMS-stamp. (b) Processing steps needed to transfer the inverted texture of the stamp into the resist on the substrate.

2.2 Hole transport layer

To ensure the results are applicable to real solar cell applications, a well-known stack previously utilized for tandem perovskite-silicon solar cells was employed to demonstrate its transferability. The first layer after the texturing is an indium-tin-oxide (ITO) layer that is RF sputtered from a ceramic target with 0.2% oxygen in the Ar/O₂ sputter gas mixture in a sputtering tool from Roth&Rau. For this work a sputtering process was chosen that would produce a layer of thickness of 150 nm on a planar substrate. The sputtering process was conducted at room temperature and with a rotating sample holder underneath the sputter target to ensure an even ITO layer. Due to the different surface enhancement factors the actual layer thickness of the ITO varies depending on the texture. Since the goal was not to produce a properly working solar cell this does not pose any difficulties. Afterwards, 100 µl of a 1 mM solution of MeO-2PACz molecule dissolved in ethanol was spin-coated for 30 s at 3000 rpm with a windup time of 5 s and afterwards cured for 10 min at 100 °C to bond it to the surface and evaporate remaining solvents. This self-assembling-monolayer (SAM) works as a hole transport layer [13].

2.3 Perovskite layer

After the preparation of the substrates with NIL, and the application of an ITO and SAM layer, the perovskite layer is deposited in a solvent-based spin-coating process. The Cs_0.05(FA_0.83MA_0.17)_0.95Pb(I_0.83Br_0.17)₃ perovskite was prepared following Saliba et al. [14] by mixing precursor solutions containing formamidinium iodide (FAI) and lead iodide (PbI₂), as well as lead bromide (PbBr₂) in the ratio of 5:1. The precursor contained 1 M of the organic and 1.1 M of the lead salts and was dissolved in anhydrous dimethyl formamide/dimethyl sulfoxide (DMF/DMSO) 4:1 (v/v). Further on, 5% vol cesium iodide (CsI) from a 1.5 M stock solution in DMSO was added to the precursor solution. The ink is then spin-coated in a two-step program at 1000 rpm for 10 s and afterward with 6000 rpm for 20 s. 15 seconds after the beginning of the spin coating, 250 μl of anisole was deposited onto the spinning substrate as an antisolvent drop. During the PL measurements, the second step duration was increased to 45 seconds. This was done to further visualize the PL data over a longer time and get PL data even after the actual formation of the film. This was necessary, as the PL setup was specifically designed for measuring during spin coating opposed to a conventional PL setup. After the spin coating, the substrates were hard-baked for 30 min at 100 °C to remove all the leftover solvents.

2.4 Characterization

In order to display the investigated textures with feature size around and below one micrometer, that were used in this work, atomic force microscopy (AFM) measurements of the textured substrates were conducted in a non-contact mode on a Park Systems NX12 device using 160AC-NA (OPUS by MikroMasch) cantilevers. The Gaussian Circular Diffusor texture with micrometer sized features was imaged with a Keyence VK-X 200 confocal laser scanning microscope (CLSM).

After measuring the PL spectrum during crystallization, the substrates with the newly deposited perovskite layer are imaged using a LEXT OLS5100 3D laser scanning microscope (Olympus) in optical imaging mode at 50x magnification. To get an overview of the whole surface, the substrates are scanned in a 5 × 5 grid and images are taken at the respective positions. Unlike the SEM, these images do not yet show the crystal structure on the surface. However, they can be used to make an initial assessment of whether the perovskite has formed a conformal coating and count the pinholes that might have formed during crystallization.

For further characterization, a UV/Vis/NIR spectroscopy measurement was conducted using a Perkin Elmer Lambda 1050+ UV/Vis/NIR spectrometer. The reflection, as well as the transmission of the produced samples, were then measured in a wavelength range between 250 nm and 1200 nm in 5 nm steps. Since the sum of transmission, reflection and absorptance always equals 100%, the absorptance could then be determined from the values obtained.

To visualize the perovskite growth on the textured layers, the substrates were examined under the scanning electron microscope. Both cross-sectional and top views were taken of the samples for this purpose.

3 Results and discussion

3.1 Surface textures

In this work, three different substrate textures (Fig. 2) were investigated and compared to a planar reference. The first texture is a sinusoidal texture with a peak distance of around 750 nm (manufacturor's information) and a texture depth of 300. The second texture is a periodic structure consisting of inverted pyramids with a periodicity of around 1250 nm (manufacturor's information), a depth of around 500 nm and a pyramid base of 1000 nm. The last structure examined is a commercially available Gaussian circular diffusor texture, produced in foils by the company temicon GmbH. This diffusor has a vertical feature size of 1.6 μm ± 0.2 μm and a horizontal distance between features of 10–30 μm. The sinusoidal and inverted pyramidic textures were chosen as their compatibility with high quality perovskite solar cell processing via spin-coating had been proven in the past [4,5]. The Gaussian circular diffusor texture had been added to the portfolio to investigate if its smoothness and large-area compatibility might be beneficial solution processed perovskite solar cells. The texture parameters were determined by a minimum of 16 measurement points with corresponding standard deviations provided in Table 1. AFM was deployed for its high z-axis resolution capabilities in nano-scale texture analysis, while CLSM was selected to overcome the z-axis limitations of AFM for larger micro-scale texture characterization.

Fig. 2 Atomic force (a and b) and confocal laser (c) microscope images of the three textures investigated in this study. (a) Sinusoidal texture in hexagonal geometry, with peak distance and texture depth of around 750 nm (manufacturer's information) and 300 nm, respectively. (b) Periodic inverted pyramids with periodicity and depth of around 1250 nm (manufacturer's information) and 500 nm, respectively, and a base side length of 1000 nm. (c) Commercially available Gaussian circular diffusor texture, produced in foils by temicon GmbH, Germany.

3.2 In situ photoluminescence

During the deposition of the perovskite layer, the photoluminescence signal was measured in time steps of 0.5 s in the wavelength range between 250 nm and 1050 nm. Plotting the recorded spectra over time creates a heat map (Fig. 3a) visualizing the formation process of the wet perovskite film onto the substrate. We see a PL-signal onset after 25 s that is consistent in time with setting antisolvent drop. The intensity of the PL signal changes over time to higher intensities and longer wavelengths, as expected from the perovskite growth process of triple cation perovskites [15].

Figure 3b shows an example spectrum taken at 47 s of the map shown in part (a) (dashed line). The red curve shows the actual measured intensity of the spectrum. The background radiation of the measurement was calibrated before the start of the measurement, so the background noise was offset to get more clear data. To analyze the data, a Gaussian fit is applied to each of the recorded PL spectra (blue curve). This filters out the noise of the measured values from the results. From this fit, the peak position and the full width at half maximum (FWHM) of the recorded spectrum can be calculated.

As soon as the antisolvent drop is released, a yellowish haze can be seen on the heat map. The more a point moves towards yellow, the higher the measured intensity of the PL spectrum at that point.

Since a Gaussian fit has been performed on each time-resolved spectrum recorded, the position of the peak is also known at each time point. Accordingly, a curve of peak position, full width at half maximum (FWHM) and the peak intensity over time can be displayed for each measured textured substrate. Figure 4a shows this curves for all textured surfaces examined shortly after the time of adding the antisolvent drop, which is added after exactly 25 s of the spin-coating process. Shortly after the drop is added, the formation of the perovskite begins.

It can be seen in Figure 4a that the course of the position of the peak is very similar for all textured substrates. Likewise, all textures show a start of crystallization with a peak position around 710 nm ± 5 nm. A fast shift in the position of the PL peak from 710 nm to 725 nm within 3 seconds is seen, which we refer to the growth of perovskite crystallites, initiated by the antisolvent-drop. At the end of the measurement the peak wavelength is around 741 nm ± 1.5 nm, which is in line with other published triple cation data [16]. Due to the lower DMF content and the increasingly progressive crystallization of the perovskite, the peak of the PL spectrum shifts to higher wavelengths between 739 nm and 742.5 nm. Since the measurement is started manually and the antisolvent drop is added by hand, an error in the time of addition of the antisolvent drop to time of ± 1 s is assumed. It can be seen that the perovskite layers on all textures show the first peaks in the PL measurement at the time of 26.5 s ± 0.5 s, i.e. they start to crystallize. There are no remarkable differences between the different textures regarding PL starting time and PL peak position at the end, only the curve of the Gaussian circular diffusor (green dash dotted line) varies in its trajectory.

Another way to illustrate the crystallization of perovskite on different textured substrates is the full width at half-maximum (FWHM) value (Fig. 4b). This value can be used to show whether the width of the peak changes over time. In the specific case of the PL measurement, higher wavelength and smaller FWHM values correspond to bigger and more similar crystallites. As with the PL peak positions, the FWHM can also be shown as a time-resolved curve for the different textures examined. Figure 4b shows the course of the half-widths of the PL measurement during the spin coating of the perovskite layer onto the different textured substrates. With the deposition of the antisolvent drop a first crystallization of the perovskite layer occurs with smaller crystallites, recognizable in the lower wavelength peak and wider FWHM values. Due to the still remaining solvents in the film, these first crystals are being dissolved again only to reform into bigger and more similar crystallites, recognizable in the higher wavelength peak closer to the desired perovskite bandgap and the narrower FWHM values ≈ 3–5 s later. Here, for all substrates, the half-width drops to 48 nm ± 2 nm and remains close to this value until the end of the measurement.

Due to the PL measurement being in-situ, we average the rotations of the spin coater, thus we also average locally over a large area. The measured signal intensity (Fig. 4c) of the perovskite deposition on the sinusoidal structure (red dashed line) and the inverted pyramidal structure (blue dotted line) resembles the measurement of the planar reference (black solid line) until 40 s, with the intensity slightly decreasing afterwards. The increase of the signal after 52 s arises from the stop of the spin-coating process and will not be considered here. In contrast, the PL intensity of perovskite on the Gaussian circular diffusor (green dash dotted line) strongly deviates from the respective values of the planar reference with almost a factor of 7 lower values. The reasons for the lower PL intensity can be manifold, for example a lower coverage of the substrate with perovskite, PL light being scattered away from the detecting fiber, and quenching of the PL by non-radiative recombination channels. Disassembling the different contributions will be a topic of future work.

Summarizing the in situ PL measurements, there is no indication for a substantially different crystallization behavior of spin-coated perovskite films on planar substrates and on the investigated sinusoidal and inverted pyramid textured substrates with texture feature sizes of around one micrometer. The PL data of spin-coated perovskite films on the Gaussian circular diffusor texture indicate a quite different crystallization behavior. In the following paragraphs, possible reasons for that will be analyzed.

Fig. 3 (a) Heat map of the photoluminescence (PL) measurement of the planar perovskite reference film. After 25 seconds of spin-coating of the perovskite precursor solution, the antisolvent-drop was added. (b) Example PL spectrum taken at 47 seconds of the map shown in part (a) (dashed line) to visualize the parameters PL peak position, full width at half maximum (FWHM) and intensity.

Fig. 4 (a) Center position, (b) full width at half max (FWHM), and (c) intensity of the PL peak over time right after the deposition of the antisolvent drop.

3.3 Scanning electron microscopy

To visualize more precisely how the perovskite is finally crystallized on the textured substrates, the layer stacks are viewed in a cross-section and a top view under the scanning electron microscope (SEM).

The cross-section view is a very good way of displaying the individual layers of the layer stack. Just before the image is taken, the back of the substrate is scored with a glass cutter and then broken at the edge. Figure 5a shows the cross-sectional view of the planar reference substrate. All layers of the layer stack deposited on the glass substrate are visible, with exception of the SAM, as it is a monolayer and therefore cannot be seen in SEM images. The lowest layer is the imprint resist with a thickness of around 1.5 μm. Directly above is the slightly lighter ITO layer followed by the perovskite film. In case of the sinusoidal (Fig. 5b) and the inverted pyramidal (Fig. 5c) textures the perovskite has filled the valleys of texture and created a surface like that of the perovskite layer on the planar substrate. On the Gaussian diffusor texture, the perovskite film is very unevenly distributed on the texture and there are different layer thicknesses at all places where the layer was formed. As the sample break is not directly crossing tips and valleys of the inverted pyramidal structure (Fig. 5c) the cross-section image gives the impression that the ITO has not conformally coated. However, previous studies indicate that the ITO coating should be nearly conformal [4]. Under the SEM the crystal structure of the perovskite with the individual crystals can be seen.

The surface of the grown perovskite layer is shown in Figure 6a with a top-view image for the planar reference substrate. A typical surface structure of perovskite can be seen. The perovskite layer consists of many single crystals crystallized together to form a closed and conformal layer. Table 2 shows the results of a grain size analysis from the SEM images shown in Figure 6. The determined grain sizes are in a similar range for all substrates with a tendency for larger grains if a textured substrate is used. In Figures 6b and 6c growth artifacts can be seen on top of the sinusoidal as well as the inverted pyramidal structure. These artifacts could be so-called dendritic structures originating from an intermediate phase of the DMF [17]. As there are no additional peaks or shoulders in the PL measurements, we assume that these structures have no optical influence on the PL measurements. As these structures are not visible in the cross cross-section, it could be assumed that they are a rather thin layer and don't influence the growth of the perovskite below it. Nevertheless, irregularities in the perovskite film are undesirable and should be avoided if possible. If the structures are dendritic, the cause is a not optimally adjusted antisolvent drop or a not yet optimized solution composition. Holes in the texture of the Gaussian diffusor are visible under the SEM (Fig. 6d) and under the optical microscope (Inset). The holes in the texture have round shapes and are distributed throughout the perovskite film. The texture has heights in the micrometer range, but due to the particularly flat slopes within the texture, one might have expected a particularly well-grown perovskite film. The much lower PL intensities measured on the perovskite film grown on Gaussian diffusor texture (Fig. 4c) cannot be fully explained by the macroscopic holes and the non-complete coverage of the film. Further reasons might be scattering of the light by the texture and/or a higher share of non-radiative recombination in the film. To differentiate the impact of texture induced light scattering and reduced electrical material quality further experiments are necessary.

Fig. 5 Cross-section SEM images of a perovskite/SAM(not visible)/ITO layer stack on differently textured substrates. (a) Planar reference, (b) sinusoidal texture, (c) inverted pyramids, and (d) Gaussian circular diffusor.

Fig. 6 Top-view SEM images of perovskite films grown on the different textures (a–d). (a) Planar reference, (b) sinusoidal texture, (c) inverted pyramids, and (d) Gaussian circular diffusor. The inset in (d) shows a larger area of the latter taken by optical microscopy.

3.4 Optical spectrometry

The most important optical parameter for a solar cell is the absorptance in the perovskite layer, which directly correlates with the energy yield. Figure 7 shows the absorptance (A) spectra for the selected textured substrates compared to the spectrum of the planar reference substrate after perovskite deposition. These values were derived from reflectance (R) and transmittance (T) measurements, as is shown in equation (1).

(1)

Of particular interest is the wavelength region near the band gap of the perovskite, which is 1.68 eV. This corresponds to a photon wavelength of 738 nm. In this region, the influence of the texture beneath the perovskite layer can be seen. Compared to the planar reference sample, the textures with particularly high absorptance are the inverted pyramids and the periodic sinusoidal structure. Although the wavelength range near the bandgap is of particular interest, the textures also provide improved absorptance at nearly all wavelengths shorter than 738 nm. From finite-element method optical simulations on perovskite solar cells containing the sinusoidal and the inverted pyramidal texture we know, that the enhanced absorptance close to the perovskite bandgap arises from texture-induced light trapping in the perovskite layer, which can lead to up to 1 mA/cm² increased short circuit current density in the solar cell device [4]. In contrast, the use of a Gaussian diffusor texture seems to reduce the absorptance considerably. However, this is not due to unwanted optical effects or measurement errors, but to the fact that the perovskite did not form a uniform layer on this texture and that very large macroscopic holes were formed in the perovskite layer during spin coating (see Fig. 6d). With an irregular perovskite layer, more of the light instead hits the glass, where it is either reflected or transmitted, but not absorbed to the same extent.

Fig. 7 Absorptance spectra of perovskite films spin-coated on the investigated textured glasses.

4 Conclusion

In this study, the growth process of perovskite layers via spin-coating from solution was investigated on differently textured substrates. Three different nanoimprint-textured glasses and planar reference glasses, all covered with an ITO layer and a hole transporting layer to mimic solar cell fabrication conditions, were used as substrates: A smooth sinusoidal and an inverse pyramidal structure with periodicities of 750 nm and 1250 nm, and structural heights of 300 nm and 500 nm, respectively. Further, a Gaussian circular diffusor structure with horizontal peak-to-peak distances in the range of several tens of micrometers and vertical feature of 1.6 ± 0.2 μm. We monitored the perovskite spin-coating process using in-situ photoluminescence (PL). No significant differences in crystallization time, PL peak position, PL peak width or PL intensity were observed among the investigated sinusoidal and inverted pyramidal textured substrates compared to the planar reference. The much larger and random Gaussian circular diffusor texture varied especially in the measured PL intensity having a factor of roughly 7 lower intensity compared to all other textures. Scanning electron microscopy images revealed a homogenous growth of perovskite film on planar, sinusoidally textured, and inverted pyramidally textured substrates. In the case of the two textures (sinusoidal and inverted pyramids), the texture valleys were found to be well-filled with perovskite. In contrast, the larger diffusor texture yielded macroscopic holes in the perovskite film. Optical experiments showed an increased absorptance in the whole wavelength range below the electronic bandgap in films with sinusoidal and inverse pyramidal textures compared to a planar reference film. However, due to the observed macroscopic holes, the absorptance decreased in the perovskite film on the diffusor texture. We conclude, that within the set of investigated substrate textures the growth of a homogeneous and strongly absorbing perovskite film is possible for textures with horizontal and vertical features with a size of around one micrometer or smaller. In contrast, spin-coating perovskite onto a textured surface with vertical features larger than 1-2 μm resulted in non-uniform films with macroscopic holes. The results add new experimental data on in-situ growth of perovskite films on textured surfaces.

Funding

We thank the Helmholtz association for funding within the project “Zeitenwende - Accelerated transfer of the next generation of solar cells into mass production - future technology tandem solar cells”. We further thank the Deutsche Bundesstiftung Umwelt (DBU) for funding within the TexturedTandem project (grant no. 37908/01), as well as temicon GmbH, Germany – for providing their commercial textured foils free of charge.

Conflicts of interest

The authors have no conflict of interests related to this publication.

Data availability statement

The data that supports the findings of this study are available from the corresponding author, Florian Riesebeck, upon reasonable request.

Author contribution statement

C.B. planned and supervised the project, F.M. supervised the experimental work, F.R. performed the measurements, processed the experimental data, performed the analysis, drafted the manuscript and designed most figures. S.T. performed the AFM measurements, D.Y. analyzed the AFM data and designed the designated figures. C.B., E.U., F.R. and F.M. aided in interpreting the results. C.B. aided in shaping the manuscript.

References

All Tables

All Figures

	Fig. 1 Fabrication of textured glass substrates using nanoimprint lithography. (a) Replication of the master structure in an inversed PDMS-stamp. (b) Processing steps needed to transfer the inverted texture of the stamp into the resist on the substrate.
In the text

Fig. 2 Atomic force (a and b) and confocal laser (c) microscope images of the three textures investigated in this study. (a) Sinusoidal texture in hexagonal geometry, with peak distance and texture depth of around 750 nm (manufacturer's information) and 300 nm, respectively. (b) Periodic inverted pyramids with periodicity and depth of around 1250 nm (manufacturer's information) and 500 nm, respectively, and a base side length of 1000 nm. (c) Commercially available Gaussian circular diffusor texture, produced in foils by temicon GmbH, Germany.

In the text

	Fig. 3 (a) Heat map of the photoluminescence (PL) measurement of the planar perovskite reference film. After 25 seconds of spin-coating of the perovskite precursor solution, the antisolvent-drop was added. (b) Example PL spectrum taken at 47 seconds of the map shown in part (a) (dashed line) to visualize the parameters PL peak position, full width at half maximum (FWHM) and intensity.
In the text

	Fig. 4 (a) Center position, (b) full width at half max (FWHM), and (c) intensity of the PL peak over time right after the deposition of the antisolvent drop.
In the text

	Fig. 5 Cross-section SEM images of a perovskite/SAM(not visible)/ITO layer stack on differently textured substrates. (a) Planar reference, (b) sinusoidal texture, (c) inverted pyramids, and (d) Gaussian circular diffusor.
In the text

	Fig. 6 Top-view SEM images of perovskite films grown on the different textures (a–d). (a) Planar reference, (b) sinusoidal texture, (c) inverted pyramids, and (d) Gaussian circular diffusor. The inset in (d) shows a larger area of the latter taken by optical microscopy.
In the text

	Fig. 7 Absorptance spectra of perovskite films spin-coated on the investigated textured glasses.
In the text