Moth eye-inspired highly efficient, robust, and neutral-colored semitransparent perovskite solar cells for building-integrated photovoltaics

Abstract


| INTRODUCTION
The power consumption of buildings is tremendous. 1 Moreover, a vast amount of energy is wasted during distribution. For instance in the United States, 25 TWh of electricity is consumed by buildings each year, accounting for nearly 40% of the total energy demand. 2 Researchers are exploring solutions to this issue. Currently, traditional solar panels are opaque and can only be mounted in the limited space of rooftops. In modern densely packed cities, a more efficient utilization of solar energy is the need of the hour. One of the feasible solutions is to install semitransparent solar panels onto the windows, facades, and skylights. [2][3][4] The emerging semi-transparent perovskite solar cells (ST-PSCs) have gradually come to the fore for buildingintegrated photovoltaics (BIPV). 5 Particularly in the application for power generation windows, ST-PSCs must harvest sunlight and convert it to electricity while transmitting part of daylight to illuminate indoor space. Besides efficiency, which is the primary figure-of-merit for most photovoltaic devices, attributes such as average visible transmittance (AVT), color, and longevity are also key specifications for ST-PSCs. The AVT can be modified by tuning the perovskite thickness or surface coverage. 6,7 Generally, an AVT of 20%-30% can satisfy the transmittance requirement of power-generating windows. 7 The conventional structure for ST-PSCs requires continuous ultrathin perovskites; nevertheless, one of its limitations is the undesired hue of reddish-brown. [8][9][10][11][12][13][14][15] Eperon et al. attempted to reduce perovskite surface coverage and attained neutral-colored ST-PSCs. 16,17 But this strategy rendered low efficiency, with a 3.5% power conversion efficiency (PCE) achieved at AVT 30%. The loss of PCE can be ascribed to the shunt current from the direct contact of the electron transport layer (ETL) with the hole transport layer (HTL). 18 Besides, the non-uniform surface coverage resulted in unfavorable scattering.
Apart from the intrinsic optical properties of materials, colors are also determined by the surface architectures. [19][20][21][22][23] For instance, the moth-eye surface is covered by hexagonally closely packed microcavities, which is one of the intriguing photonic structures invented by mother nature ( Figure 1A). 24 This evolutionary advancement provides moth night vision capability and other unique functions. Notably, the MEIS is a broadband anti-reflecting layer, 25,26 and its surface geometry offers a possibility to alter reflection spectra and optical appearance. [27][28][29] Additionally, human visual perception is different to photons at a different wavelength. 30 Thus, harvesting solar radiation in the low responsive regions is a logical approach toward a highly visual-transparent and efficient ST-PSCs which the incorporation of MEIS realized. 30 To achieve efficient light-harvesting in the desired wavelength range, herein, we report a biomimetic MEIS ST-PSCs. The MEIS demonstrates an appealing light manipulation, which benefits the light-harvesting in the wavelength range where the human eye is less sensitive. As a result, the PCE of MEIS-based devices is significantly improved without compromise AVT. Moreover, due to the alteration of the absorption spectra, a desired near-neutral color is achieved in MEIS based device. Besides, the perovskites with robust performance were investigated in this work. Figure 1A shows a schematic device structure of MEIS ST-PSCs. Figure 1B shows the fabrication process of the device. Briefly, the polystyrene (PS) spheres are assembled on planar-ITO using the Langmuir-Blodgett technique, which is a template used to pattern MEIS-ITO. Then PS spheres are covered with ITO by sputtering. Next, the PS spheres are removed, and the biomimetic MEIS-ITO is obtained.
The tilted-angle view and top view scanning electron microscopy (SEM) images of as-assembled PS spheres and MEIS-ITO are shown in Figure 1C1, C2, D1, D2, respectively. The PS spheres, with a diameter of 900-950 nm, are uniformly distributed on the ITO. The distance of two neighboring spheres is 50-100 nm. The hexagonally arranged MEIS shows a depth of 250 nm and a diameter of 900-950 nm. The cross-sectional SEM image of the whole device based on MEIS and planar counterpart is shown in Figures 1E and S1A, respectively. Figure 1E confirms a conformal and uniform coverage of SnO 2 on MEIS-ITO with a controlled thickness. Close contact is formed at the interface between perovskite and SnO 2 . Although similar MEIS structures were reported in a few works fabricated with solution methods, 18,31 obtaining a crack-free MEIS structure has been a challenge. The presence of cracks is detrimental to a cell. Because these regions with a direct perovskites/electrode contact may result in a short circuit due to the wretched hole blocking. 32,33 To address this bottleneck challenge described in previous reports, we utilized the sputtering of ITO for the first time to form the crack-free MEIS structure. Additionally, the micro-scale MEIS-ITO offers a possibility to achieve patterned high-crystallinity perovskites. Revealed in Figures 1F and S1B, compact and singlecrystalline Cs 0.05 FA 0.83 MA 0.12 PbBr 0.33 I 2.67 with grain size as large as 1 μm is acquired. Shown in Figures 1E and S1B, the patterned perovskite is compact, rendering the grain boundaries hard to discern. As most perovskite boundaries are perpendicular to the substrate, the undesired impacts of defects at the grain boundaries are minimized. 34,35 As shown in Figure 2A, all the XRD patterns match well with the cubic phase perovskites (Space-group: Pm 3 m, α = 90 ). The miller indices (h k l) and peak position (2θ) are listed in Table S1. No impurity diffraction peak is observed. With the increase of Br concentration (from x = 0 to x = 0.46), a peak position shift is observed, reflected from the 2θ of (100) from 14.032 to 14.096 . The lattice constant and full-width half maximum (FWHM) are listed in Table S2. The FWHM of the (100) plane shows no noticeable change. This means Br incorporation does not play a vital role in controlling grain size, further confirmed in the top-view SEM ( Figure S1). Figures 2B and S2 show the steady-state photoluminescence (PL) spectra and absorbance spectra of Cs 0.05 FA 0.83 MA 0.12 PbBr x I 3-x on quartz, respectively. The PL emission is blue-shifted with the enhancement of Br incorporation. The lattice constant and bandgap both vary linearly with the increase of Br concentration, suggesting that the variation precisely follows Vegard's law 36 and a solid solution of the I/Br mixtures are formed ( Figure S3). The complex refractive index spectra (n, k) of Cs 0.05 FA 0.83 MA 0.12 PbBr 0.333 I 2.667 is plotted in Figure 2C. The real part n value lies in the range from 2.1 to 2.4 in the wavelength range from 400 to 900 nm. The imaginary part k is the extinction coefficient consistent with the measured absorbance spectra ( Figure S2). The original data and more discussion can be found in Figure S4. The carrier lifetime of perovskites and the charge injection from the perovskites to electron quenchers are investigated through the time-resolved photoluminescence analysis (TRPL). The TRPL measurements Figure 2D reveals a lifetime of τ e ¼ 54.48 ns for the neat Cs 0.05 FA 0.83 MA 0.12 PbBr 0.33 I 2.67 , where τ e is the time taken for the PL to plummet to 1/e of its initial intensity. 37,38 The quenching samples were prepared by one-step spinning coating deposition of Cs 0.05 FA 0.83 MA 0.12 PbBr 0.33 I 2.67 on SnO 2 / planar-ITO and SnO 2 /MEIS-ITO, respectively. A substantial quenching is observed in Cs 0.05 FA 0.83 MA 0.12 PbBr 0.33 I 2.67 on SnO 2 /planar-ITO reducing the lifetime to τ e ¼ 8.94 ns. This is further shorted to τ e ¼ 3.04 ns for Cs 0.05 FA 0.83 MA 0.12 PbBr 0.33 I 2.67 on SnO 2 /MEIS-ITO. Improved PL quenching bears evidence that MEIS facilitates a more efficient carrier separation and collection at the interface than the planar counterpart.
The device optical property is investigated in this work. Figure 3A shows the reflection spectra of SnO 2 / MEIS-ITO and the human luminosity curve. Human eyes are sensitive to green light but less responsive to the photon in the blue and red regions. 39 Owing to the MEIS geometry-assisted light-manipulation, the SnO 2 /MEIS-ITO manifests a unique broadband anti-reflection mimicking the moth-eye. 25,28,29 Notably, the reflection spectra of our SnO 2 /MEIS-ITO precisely matches well with the human luminosity curve. This is further resolved in the transmission and absorption spectra ( Figure S5) and photographs of SnO 2 /MEIS-ITO (the inset in Figure 3A). Even though MEIS-ITO is well-designed for ST-PSCs, the geometry still possesses variation ( Figure S6), and the reflection spectra will be altered along with the modification of MEIS geometry ( Figure S7). Figure 3B is the plotted color coordinates of Cs 0.05 FA 0.83 MA 0.12 PbBr x I 3-x on SnO 2 /planar-ITO, SnO 2 / MEIS-ITO under AM1.5 illumination, on the CIE xy 1931 chromaticity diagram. The original data are listed in  The photographs ( Figure 4D) further affirm this color perception management. Even though the perovskites with a low Br concentration (x = 0, 0.17) show desired color-neutrality, the aging test demonstrated that longevity would significantly decrease by reducing Br concentration.
We further investigated the impact of MEIS-ITO on the absorption spectrum of perovskites. The corresponding Tauc plot is shown in Figure S8. As revealed in Figure 3C, the absorption of Cs 0.05 FA 0.83 MA 0.12 PbBr x I 3-x on SnO 2 /planar-ITO increases monotonically from the bandgap. This can be ascribed to the trap states near the conduction band minimum (CBM) and valence band maximum (VBM) in the ultra-thin polycrystalline perovskite. The loss of the light absorption in the red region is the cause for the undesired reddish-brown hue of the ultra-thin perovskite. Besides, a steeper slope is observed in the absorption spectra as the Br doping reaches the point of x = 0.46, attributed to crystalline structure breakage caused by the excess amount of Br. 40 In Figure 3D, compared with the planar counterpart, there is enhanced broadband absorption of perovskites on SnO 2 /MEIS-ITO, especially for the longerwavelength photons. [41][42][43] MEIS shape are used in the back to couple the photons into the perovskites and thereby enhance the absorption without compromising AVT. 44 Most photons (400-500 nm) are absorbed in the single path of perovskites, making the light-coupling effect of these wavelengths less observable. In contrast, a boosted lightharvesting is witnessed for the photons at the wavelength of 600-800 nm. The photons at these wavelengths are less absorbed in a single path. The light-coupling increases the optical length in perovskites and enhances the lightharvesting of these wavelengths, consequently. The unique optical property of MEIS results in the alteration of visual appearance to the near color-neutrality as illustrated in color coordinates Figure 3B and the photograph of device Figure 4D. This result is further confirmed by the finitedifference time-domain (FDTD) simulations ( Figure S9).
The photovoltaic devices with ITO/SnO 2 /Cs 0.05 FA 0.83 MA 0.12 PbBr x I 3-x /spiro-OMeTAD/MoO 3 /IZO structure is  Figure 4A shows the energy-band diagram of the complete device. Band energy levels of SnO 2 , 45 spiro-OMeTAD, 46 MoO 3 /IZO 47 are obtained from previously published literature. The VBM of Cs 0.05 FA 0.83 MA 0.125 PbBr x I 3-x is À 6.14 eV, derived by ultraviolet photoelectron spectroscopy (UPS) measurement ( Figure S10), whereas the CBM is obtained by combining VBM with optical bandgap ( Figure S8). The CBM of Cs 0.05 FA 0.83 MA 0.12 PbBr x I 3-x (x = 0, 0.17, 0.33, 0.46) is estimated to be À4.52, À4.55, À4.58, and À4.61 eV, respectively. The Cs 0.05 FA 0.83 MA 0.125 PbBr x I 3-x all demonstrate a proper energy level alignment with SnO 2 and spiro-MeOTAD. Figure 4B shows  Figure 4B shows the simulated photo-carriers generation rates (EHP/sec/m 3 ) on the fabricated ST-PSCs. The MEIS structure has less perovskite materials, however showing higher absorption ( Figure S9) compared to the planar counterpart. As revealed in Figure 4C, a higher external quantum efficiency (EQE) in the red region is observed in the MEIS device, producing an enhanced integrated current density of 13.52 mA cm À2 for MEIS compared with the planar counterpart (12.36 mA cm À2 ). Figure 4D are photographs of the planar and MEIS devices based on Cs 0.05 FA 0.83 MA 0.12 PbBr 0.33 I 0.27 , taken under outdoor sunlight. The device based on planar and MEIS both demonstrate high visual transparency. Compared with the reddish-brown hue of the planar device, a preferred near-color-neutrality was obtained in the MEIS device. In addition, the incident angle of sunlight varies over time, the angular-dependent performance of a solar cell is a concern for its real application. [48][49][50] Shown in Figure S11, in contrast to the planar counterpart, the MEIS ST-PSCs possess an appealing capability of harvesting photons from a wide angle with a broad wavelength range. Figure 5A shows the J-V curves of MEIS ST-PSCs with different Br concentrations. The AVT and luminosity of devices are listed in Table S5. The level of Br doping does not affect the efficiency dramatically, and J sc is only slightly decreased with the increase of Br concentration in perovskite. Whereas the degree of Br incorporation remarkably impacts long-term stability which are major concerns for practical applications.
In Figure 5B, we performed a long-term stability test on the devices by placing the unencapsulated MEIS ST-PSCs in a desiccator with humidity below 10% and tested in an ambient environment 70%-90% humidity every week. Moisture tolerance is observed in the device based on Cs 0.05 FA 0.83 MA 0.12 PbBr 0.33 I 0.27 , as the unpackaged device retained above 80% of its original PCE after 224 days. In contrast, the device based on Cs 0.05 FA 0.83 MA 0.12 PbBr x I 3-x with a low Br concentration (x = 0, 0.17) decomposed within 20 min when exposed at ambient conditions (humidity = 70%-90%). To further ascertain the photostability, we implemented a light-soaking stability test by exposure the encapsulated device based on Cs 0.05 FA 0.83 MA 0.12 PbBr 0.33 I 0.27 under continuous AM1.5G illumination in the ambient condition at~40 C. We recorded the photovoltaic metrics every To evaluate the figure-of-merit of ST-PSCs, Earlier reported works and our results are presented in Figure 4D, and Table S4 for comparison. 10,18,30,[53][54][55][56] As it can be seen, our MEIS-based ST-PSCs shows the best performance to our best knowledge, which can be interpreted by the highest PA value (Table S4). This illustrates that the incorporation of our MEIS template is a practical approach toward highly visually transparent, efficient, and robust ST-PSCs.

| CONCLUSION
In conclusion, we demonstrate MEIS for highly efficient, robust ST-PSCs with an optimal combination of AVT, PCE, and near-neutral color. The incorporation of MEIS combines versatile merits, such as light-trapping and benefiting the carrier collection at the interface. The alteration of the absorption spectra of perovskites based on MEIS resulted in the significantly enhanced J sc without sacrificing the AVT, as well as a modification of desired near color neutrality. Additionally, the ST-PSCs exhibited appealing resistance against moisture and solar irradiation after engineering the halide blend ratio in perovskites.

| Fabrication of SnO 2 /MEIS-ITO
The ITO glass was ultrasonically cleaned by detergent, acetone, isopropanol, and deionized water in sequence after blow-drying the substrate with N 2 .
The monodispersed polystyrene solution was prepared by mixing polystyrene colloidal suspensions, ethanol, and phosphoric acid at the ratio of 1:3:1 (vol%: vol%: vol%) followed by 15 min supersonic dispersion. Next, the as-prepared solution was steadily injected on a clean silicon wafer via a syringe pump. And then PS spheres slowly distributed on the top surface of DI water. Notably, a monolayer of the PS spheres floating on water was acquired. Therefore, it was necessary to gently and gradually slide the PS spheres on the top surface of DI water to prevent PS spheres from sinking. After putting it to stand for 3 h, sodium dodecyl sulfate (100 μl) was dropped into the container to form closely packed and well-ordered PS spheres. Thereafter, the water level was gradually reduced by utilizing the siphoning effect, and the PS spheres were transferred to the ITO glass. After that, the PS spheres coated ITO glass was dried in ambient conditions for 6 h. To alter the diameter of the PS spheres, the oxygen plasma treatment (at the power of 30 W, with the oxygen flow of 30 s.c.c.m.) was carried out.
The highly ordered PS spheres were filled by sputter deposition of 250 nm thick ITO at a deposition rate of 0.6 Å s À1 . The D.C. sputtering uses an ITO target and a gas flow rate of 0.5 O 2 and 50 s.c.c.m Ar. Afterward, PS spheres were removed by soaking the sample in chloroform ultrasonic for 15 min, followed by ultrasonic cleaning by acetone for 5 min.

| ALD Depositions of ETL
SnO 2 was deposited by the ALD method (MNT Micro and Nanotech Co., Ltd) using TDMASn and DI water as precursors. The TDMASn, DI water sources were kept at 80 C and room temperature, respectively. The deposition was carried out at 150 C. Each cycle consisted of Sn pulse (30 ms), Ar purge (30 s), H 2 O pulse (300 ms), and Ar purge (30 s). The growth rate of the SiO 2 per cycle under this deposition condition was found to be 0.095-0.1 nm, which was estimated by SEM studies. In the present work, 200 cycles were run to yield 20 nm thick (actual thickness was between 18 and 22 nm). After that, the asdeposited SnO 2 was annealed at 400 C in the air for 4 h, and then furnace cooling. To enhance the perovskite precursor solution's wettability on the as-grown sample, the oxygen plasma treatment (at the power of 30 W, oxygen flow of 30 s.c.c.m.) for 10 min was applied.  Table S6. The precursor solution was stirred at 1000 rpm in the N 2 filled glovebox at room temperature for exactly 2 h. Then 60 μl precursor solution was spincoated on as-grown substrates with 1500 rpm for 5 s. Then spin rate was accelerated up to 5500 rpm for another 30 s. Antisolvent chlorobenzene (60 μl) was dropped at the 30th second. After the spin-coating of the precursor solution, the substrates were transferred to a hot plate and heated at 100 ± 5 C for 20 min. The spin coating and annealing process were operated in the N 2 filled glovebox.

| HTL and Top-contact depositions
The HTM precursor solution was prepared by dissolving 80 mg spiro-OMeTAD in chlorobenzene (1 ml) and 4tert-butylpyridine (28.8 μl) and Li salt solution (520 mg ml À1 lithium bis-(trifluoromethylsulfonyl)imide in acetonitrile, 17.5 μl) was then added. The solution was spin-coated on the perovskite films with 3500 rpm for the 30 s. The films were left in a dry desiccator for 4 h. To protect the spiro-OMeTAD from sputtering damage, a buffer layer of 5 nm of MoO 3 was thermally evaporated at a rate of 1 Å s À1 on the HTM layer before IZO deposition. The evaporation was operated in an evaporator that connected with the N 2 filled glovebox. Finally, 100 nm IZO was sputtered as a transparent electrode under a pressure of 8.2 Â 10 À2 Pa of Ar at 50 W at the rate of 2 nm min À1 at room temperature.

| Encapsulation
Devices were encapsulated in a nitrogen-filled glove box. The epoxy resin was dropped over the top electrode, and a small glass coverslip was placed on top. And then, the sample was exposed to ultraviolet illumination for 15 min.

| Characterization
The as-grown perovskites were characterized by the Xray diffraction (XRD) (Bruker D8 X-ray Diffractometer, United States) utilizing Cu-Kα radiation and scanning electron microscopy (FESEM, JEOL JSM-7100F, Tescan VEGA3). Ultraviolet-visible spectroscopy (Varian Cary 500) was used to obtain the absorption and transmission spectra. The UV-vis Spectrophotometer-Perkin Elmer (Model: Lambda 20) was used to measure the absorbance spectra. The Ocean Optics spectrometer (Model: FLAME-T-VIS-NIR-ES) obtained the color coordinate on the CIE xy 1931 chromaticity diagram. Photoluminescence (PL) and time-resolved photoluminescence (TRPL) were measured with an FS5 fluorescence spectrometer. Carrier lifetime was determined by mono-exponential decay fittings. The device performance was measured with an AM© solar simulator (Abet Class AAB Sun 2000, calibrated by a KG5-filtered Si reference cell). The current density-voltage (J-V) curves were measured with a Keithley 2400 under simulated AM1.5G illumination from a calibrated solar simulator with an irradiation intensity of 100 mW cm À2 . Ultraviolet photoelectron spectroscopy (UPS) was performed with Axis Ultra DLD multi-technique surface analysis system. The complex refractive index was measured with an ellipsometer (Model: Alpha-SE). To calculate the EQE, the photocurrent under monochromatic illuminations at different wavelengths was measured using a 150 W xenon lamp and a monochromator.

| FDTD simulations
FDTD method was conducted to study the absorption properties of the devices based on MEIS and planar counterpart. Plane-wave light source with a wavelength from 300 to 800 nm was normally illuminated onto the simulated structures. Periodic boundary condition was imposed at vertical boundaries, and only one unit cell was simulated to reduce the computational demand. The optical constants for ITO and SnO 2 were taken from the website refractiveindex.info. The complex refractive index of Cs 0.05 FA 0.83 MA 0.12 PbBr 0.33 I 2.67 was calculated from the ellipsometer measurement in this work. We calculated the generation rate profile, namely, generated electron-hole pairs per unit volume per second G opt ( r ! , λ) where k is the imaginary part of the permittivity and E r ! , λ is the electric field.
and Optoelectronics Technologies at HKUST and Foshan Innovative and Entrepreneurial Research Team Program (2018IT100031). Yiyi Zhu and Lei Shu contributed equally to this work.