Roller Nanoimprinted Honeycomb Texture as an Efficient Antireflective Coating for Perovskite Solar Cells

The properly chosen light management strategy in perovskite solar cell devices is indispensable in achieving high power conversion efficiency. To diminish the reflection losses, texturization of the front surface, similar to what is used in established solar cell technologies, shall be taken into consideration. Within this paper, a honeycomb‐like textured SU‐8 photoresist layer is applied using a roller nanoimprint technique onto a planar perovskite solar cell to minimize reflection losses. The results show that the applied honeycomb pattern reduces the solar‐weighted reflectance from 13.6% to 2.7%, which enhances the current density of the unmodified cell by 2.1 mA cm−2, outperforming the commonly used planar MgF2 antireflective coating by 0.5 mA cm−2. The experimental results are combined with optical modeling to find optimized structures and predict the optical behavior within a solar module. The process used within this work can be transferred to perovskite‐silicon tandem solar cells, providing a promising pathway for the reflection reduction in future devices.

Excellent light management is crucial for the highest performance in PSCs, similarly to other PV technologies. Light management can be separated into decreasing reflection losses and maximizing absorption by optical path length enhancement. The latest results show excellent broadband absorptance and internal quantum efficiency respectively, as even a several hundred nanometers thick layer of perovskite material absorbs sunlight very effectively. [11][12][13] Thus, the focus of recent works is reflection reduction, which remains an issue in the multilayer stacks of perovskite solar cells. A common approach to reduce reflection losses in planar PSC is the deposition of single-layer antireflective coating (ARC) with a typical refractive index between 1.3 and 1.4. A prominent candidate is magnesium fluoride (MgF 2 ). Alternatively, structured antireflective surfaces such as random nanorods, [14,15] biomimetic moth-eye patterns, [16][17][18][19] semi-spheroidal cavities, [20] or microlens arrays [21] were investigated. Apart from this, direct patterning of the perovskite active layer, glass texturization or applying texturized foils lead to decreased reflection. [22−29] However, preserving the perovskite morphology during the application process of those structures remains a major challenge, due to the general sensitivity of the perovskite absorber.
To solve this challenge, we used a new approach to achieve extremely low reflectance by using honeycomb structures on top of the active PSC. To this end, we utilized thermally assisted UV roller nanoimprint lithography (TUV-Roller-NIL). [30−33] This technique meets the criteria of reliable repeatability of high aspect ratio structures with high throughput and long stamp lifetimes. [34] We showed that this process chain was harmless enough to not deteriorate the performance of the PSC. To the best of our knowledge such or a similar structure has never been directly applied to a perovskite solar cell, especially not with a scalable process like TUV-Roller-NIL. The presented process may be implemented onto perovskite-silicon tandem cells in the future. In addition, this study discusses the limits of light harvesting improvements using a honeycomb texture with optimized indexmatching material along with the potential enhancement of the full module stack.

Perovskite Solar Cells Fabrication
The realized single-junction perovskite cells were manufactured using the procedure described in detail by Schulze et al. [35] and the Supporting Information. The resulting perovskite absorber exhibits a high energy bandgap of around 1.66 eV, suitable for perovskite-silicon tandem cell applications. As top contacts, 200 nm Ag lines were evaporated around each ITO top electrode pad to create a U-shape conductive busbar that defines four-times 0.25 cm 2 active cell area per glass substrate (Figure 1). A 100 nm MgF 2 layer was evaporated on top of one sample as a reference ARC. On the top of the other cell, the honeycomb texture was imprinted using the procedure described in Section 2.2. Both samples were characterized optically and electrically before and after these treatments. Figure 1 schematically shows the investigated samples before and after antireflective systems applications. Also, the real appearance of the sample after the imprinting procedure is shown. The sample shows a uniform color and no visible damage to the cell.

Honeycomb Front Side Structure Fabrication
To fabricate the honeycomb front side structure, the process steps illustrated in Figure 2a are followed: First, a PDMS stamp replicated from a master shim (1-2) as described in detail by Hauser et al. [31] Then, a SU-8 2002 photoresist (micro resist technology) was spin-coated on the sample (3). The resulting photoresist thickness was determined with an SEM measurement to be around 5 μm. The prepared sample was then placed in the TUV-Roller-NIL system [31] and heated to 90°C to let the thermoplastic photoresist become formable. Afterward, the stamp is slowly pressed against the sample by the weight of the rotating roller, forming the negative of the stamp pattern on the sample. While the photoresist is being pressed and formed, UV light triggers its cross-linking (4). After the exposure, the sample was postbaked for 3 min to completely crosslink the SU-8 (5). In Figure 2b, representative SEM images (characterization details in Section 2.3) of the imprinted texture obtained on a glass substrate are shown. A cross-section reveals that the slopes of the imprinted valleys are uniform and well reproduced. At the bottom, there is a thin (≈100 nm) residual layer.

Characterization Techniques
Reflectance spectra were measured using a Perkin Elmer UV/VIS/NIR Lambda 950 Spectrometer with an integrating sphere. External quantum efficiency measurements were collected with an in-house built setup in 10 nm steps from 300 to 900 nm with no additional bias light. J-V characteristics were collected using a xenon lamp calibrated to a reference silicon cell sample to assert 100 mW cm −2 illumination power density. During the measurement of the given pixel, all the other pixels were covered with a black mask to eliminate the unintentional over-illumination of the sample. Josť et al. [27] employed this measurement technique and showed that no relevant measurement uncertainties were found, even when examining textured samples. A few measurements were taken until the stabilization of the J-V curves, then the latest result is presented. Scanning electron microscopy was performed using a Hitachi S-4700 system with 5 kV accelerating voltage, 12 nA current, and around 16 mm working distance. More details are available in the Supporting Information.

Wave-Optical Modeling with Rigorous Coupled Wave Analysis
The entire optical system, consisting of the thin film solar cell and the antireflective structure, if applicable, was modeled completely coherently using Rigorous Coupled Wave Analysis (RCWA). For this, the "reticolo" implementation as provided by Lalanne, Hugonin et al. [36][37][38] was used. Due to the calculation time of RCWA depending highly nonlinearly on the unit cell size, it was not possible to simulate the honeycomb structure with the exact spatial dimensions as in the experiment. Instead, the structure was scaled down with fixed proportions to allow for feasible calculation times. A convergence analysis was also done for this scaling factor, ensuring a close approximation of the behavior of the experimental system. As only the real part of the refractive index of the structure material was used and the period of the scaled structure was almost four times larger than the longest wavelength used in the simulation, this approach was found to be reasonable in the convergence study. More details regarding the simulation methodology can be found in the Supporting Information.

Comparison of Reflectance Spectra
Reflectance spectra were measured to investigate the effect of both concepts-planar MgF 2 -ARC and honeycomb structureon the reflectance of the PSCs. These spectra are depicted as solid lines in Figure 3. The measurement without any antireflective system serves as a reference.
The sample with the MgF 2 top layer exhibits a reflectance spectrum with wavelength-dependent behavior with constructive interference maximum reaching 15% at around 400 nm. The sample reflectance is reduced below 5% in the 500-700 nm region. It increases to more than 10% in the shorter wavelength region (350-500 nm). The sample with honeycomb texture exhibits  a quite uniform, wavelength-independent reflectance spectrum, with the reflectance signal below 5% across the relevant wavelength range for the PSC (400-740 nm).
To quantify and evaluate the reflective properties of the samples for photovoltaic applications, the AM1.5g photon flux solar spectrum [39] (Figure S5, Supporting Information) weighted reflectance was calculated using the following formula where R( ) is the reflectance of the sample and Φ AM1.5 is the photon flux of the AM1.5g solar spectrum. To estimate the maximum generated current assuming perfect external quantum efficiency, the following formula is used where q is the elementary charge. The planar solar cell without ARC exhibits 13.6% of weighted reflectance in the 300-750 nm range which translates to 3.2 mA cm −2 current density loss. Applying a planar antireflective layer of 100 nm MgF 2 decreases this loss to around 4.9% (1.2 mA cm −2 ). With the honeycomb pattern presented in this work, the reflection loss can be further reduced to a value of 2.7%, which corresponds to a loss of 0.6 mA cm −2 .
In addition, the reflection behavior of all three systems was modeled using RCWA. The dashed lines in Figure 3 depict the simulation results. They show proximity to the measured values, and in particular the relative changes caused by the different approaches are well reproduced. The systematic deviation of the simulation toward higher reflectance can be explained by the assumption of perfectly planar layers and surfaces in the simulation. In reality, relevant microroughness is present which leads to reduced reflection. [40]

Effect of the Antireflection Systems on the External Quantum Efficiency
External quantum efficiency (EQE) measurements were obtained to further investigate the enhancement of the optical and electrical properties of the solar cell achieved by using the proposed structure (Figure 4). The raw EQE data were calibrated with the short-circuit current density (J SC ) obtained from the J-V measurements discussed in Section 3.3. The spectra were taken over the range of 300-800 nm to fully cover the perovskite absorption range.
Both modified samples exhibit an improvement in EQE over the 400-750 nm range, yet the honeycomb structure outperforms the planar MgF 2 alternative. The MgF 2 EQE spectrum has a noticeably visible maximum around 550 nm wavelength which is also present as the reflectance minimum in Figure 3. In contrast, the honeycomb texture has more uniform, broadband EQE enhancement in the 370-750 nm range, reaching more than 10% in the 430-560 nm range, and above 5% in the region up to 740 nm. This improvement is comparable to the reduction in the reflection in Section 3.1. In both types of structures, there is some loss compared to the case without ARC in the 300-370 nm wavelength region. Across this region, the honeycomb-treated solar cell suffers from parasitic absorption in the SU-8 layer ( Figure S4, Supporting Information) that prevents the light from entering the active layer, while the MgF 2 ARC exhibits a constructive interference, enhancing reflection in this region. The loss of the EQE is shown as a red-filled curve in Figure 4 and is equal to 0.12 and 0.30 mA cm −2 for MgF 2 and honeycomb samples, respectively.

Comparison of J-V Curves
J-V characteristics of the cells were measured before and after the application of the MgF 2 layer and the honeycomb texture, respectively ( Figure 5). The shapes of the curves are very similar in both measurement directions, indicating minimal hysteresis in the J-V curves. Table 1. summarizes the solar cell parameters. Figure 5. J-V curves of the reference sample without any ARC, the sample with the MgF 2 layer and the imprinted honeycomb texture measured under standard AM1.5g solar spectrum illumination conditions. Solid lines represent measurements in the forward direction, whereas dashed lines the reverse direction.
After the ARC application the J SC is increased in both cases: the honeycomb structure exhibits an increase of 2.1 mA cm −2 , whereas 1.6 mA cm −2 increase is observed for the planar MgF 2 layer (2.0 and 1.5 mA cm −2 for reverse direction scan). Due to this current increase, open-circuit voltage (V OC ) is enhanced by 12-16 mV and the filling factor (FF) by 2%-3%. The final power conversion efficiency (PCE) is increased by an absolute 2.2% (2% for reverse direction scan) for the honeycomb structure and by 1.6% (1.8% for reverse direction scan) for the MgF 2 coating. The experimentally obtained current density enhancement aligns with reflectance and EQE measurements, from which the estimated current density is calculated by the AM1.5G solar photon flux integration. In both cases, the current density enhancement was equal to 2.4 mA cm −2 which is a very comparable result to 2.1 mA cm −2 increase observed in the experiment. This coherency leads to conclusion that the honeycomb structure achieved the most J SC and PCE improvement due to exhibiting the minimal reflection losses. Note that the cells efficiency is not a reliable indicator for the cell quality, as the wide bandgap  [41] for such cells.

Honeycomb Textures Potential in PSC Module Stacks
To investigate the full potential of the proposed honeycomb structure, additional optical simulations were performed: 1. Modeling of the same geometric structure, but assuming an idealized photoresist with zero absorption and a real part of the refractive index equal to the one of the top layer ITO (Figure 6a). This assumption removes the reflection at the structure-cell interface. Therefore, this simulation provides an outlook on what can be achieved by further optimizing the process demonstrated within this work. 2. Modeling the optimized structure from case 1 above with a module glass (attached, e.g., with EVA) on top of the stack, rather than air to closely approximate the actual use case in the real world. In the module, the individual solar cells along with the honeycomb structure are protected from detrimental environment conditions, such as dust particles and water, by the front glass. The results of this simulation are shown in Figure 6b.
For the optimized case (1), a very promising R W value of 1.6% has been calculated, which would lead to a significant fur-ther improvement of ≈0.3% in the PCE. In the module stack (2) values for R W for the glass-solar cell interface as low as 2.5% compared to 6.8% for a standard planar interface are obtained, leading to an estimated PCE advantage of 0.9% for the antireflective structure. This shows that the honeycomb structure can also play a beneficial role when integrated into the module stack.

Conclusions
Within this paper, we successfully applied a honeycomb structure to a perovskite solar cell using the TUV-Roller-NIL technique, resulting in excellent antireflection properties of the device. We demonstrated a J SC increase of 2.1 mA cm −2 and a PCE increase of 2.2% absolute as compared to the planar reference sample, beating the commonly used MgF 2 planar layer. Using the modeling results, we showed that such a structure can be beneficial in module stacks as well as with optimized refractive index material, leading to extremely low weighted reflectance equal to 1.6%. Among other advantages of using the TUV-Roller-NIL technique, one of the foremost is possible applicability with perovskite-silicon tandem cell systems. This possibility is of sheer importance as tandem cells represents one of the most rapidly developing technology in photovoltaic research. The results presented here pave the way toward constructing tandem devices with exceptional light harvesting properties in future research. In conclusion, TUV-Roller-NIL was found to be a suitable and scalable process for the structurization of perovskite solar cells.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.