Boosting the Performance of Flexible Perovskite Photodetectors Using Hierarchical Plasmonic Nanostructures

Plasmonic nanostructures can enhance the performance of photodetectors (PDs) owing to their amplification effect in light absorption, leading to overcoming the inherent properties of the photoactive layer. Herein, hierarchical plasmonic nanopatterns have been prepared and used for high‐performance flexible perovskite PDs. The developed hierarchical nanostructures, featuring nanoposts on cross‐nanograting patterns, exhibit a notably enhanced light trapping effect compared to hierarchical nanostructures based on a unidirectional simple nanograting structure. Moreover, hierarchical pattern‐based perovskite PDs show a photoresponsivity of 580 mA W−1 and a specific detectivity of 3.2 × 1012 Jones, which are 420% and 990% higher than those of perovskite PDs without plasmonic nanostructures, respectively. Furthermore, a flexible 10 × 10 PD array has been developed, which enables the accurate mapping of light signals with exceptional operational and mechanical stabilities, under a bending radius as small as 8 mm and after undergoing more than 1000 bending cycles. Comprehensive analyses using finite‐difference time‐domain calculations and photoluminescence mapping reveal the effective light trapping effect of hierarchical plasmonic patterns in the perovskite layers. This work provides an efficient approach to achieve high‐performance perovskite optoelectronic devices.


Introduction
[11][12][13][14] Nevertheless, the intrinsic limitations of perovskites impose an upper limit on the extent of intrinsic property improvement, necessitating the development of novel strategies to enhance the performances of perovskite devices.The plasmonic effect is actively utilized in optoelectronics and bio fields where light Plasmonic nanostructures can enhance the performance of photodetectors (PDs) owing to their amplification effect in light absorption, leading to overcoming the inherent properties of the photoactive layer.Herein, hierarchical plasmonic nanopatterns have been prepared and used for high-performance flexible perovskite PDs.The developed hierarchical nanostructures, featuring nanoposts on cross-nanograting patterns, exhibit a notably enhanced light trapping effect compared to hierarchical nanostructures based on a unidirectional simple nanograting structure.Moreover, hierarchical pattern-based perovskite PDs show a photoresponsivity of 580 mA W À1 and a specific detectivity of 3.2 Â 10 12 Jones, which are 420% and 990% higher than those of perovskite PDs without plasmonic nanostructures, respectively.Furthermore, a flexible 10 Â 10 PD array has been developed, which enables the accurate mapping of light signals with exceptional operational and mechanical stabilities, under a bending radius as small as 8 mm and after undergoing more than 1000 bending cycles.Comprehensive analyses using finite-difference time-domain calculations and photoluminescence mapping reveal the effective light trapping effect of hierarchical plasmonic patterns in the perovskite layers.This work provides an efficient approach to achieve high-performance perovskite optoelectronic devices.][21][22][23][24][25][26] In plasmonics-assisted optoelectronic PDs that induce performance improvements through the plasmonic effect, incorporating plasmonic nanostructures into devices is only feasible within limited spatial constraints within the device structure becasue the significant light absorption of these nanostructures can cause a potential decrease in device performance due to the simultaneous reduction in light absorption within the photoactive layer. [27,28]The utilization of plasmonic nanostructures is confined to specific, limited spaces within the PDs device structure, so achieving significant performance enhancement in plasmonic perovskite PDs requires the utilization of highly efficient plasmonic nanostructures with improved light amplification effects.However, most studies of plasmonic perovskite PDs have focused only on utilizing monotonous plasmonic nanostructures including spherical nanoparticles, [19,22,24] nanogratings, [13] nanosquares, [20] or nano triangles, [21] limiting the benefits of utilization to those structures rather than more effective complex nanostructures exhibiting enhanced light amplification effects.
Here, we report novel efficient hierarchical plasmonic structures composed of cross-nanograting patterns (CGs) and nanoposts (NPs) and their utilization for high-performance flexible perovskite PDs.The hierarchical nanostructures were successfully fabricated by combining a block copolymer lithography for NPs and a nanoimprint lithography for cross-grating patterns.In a previous study, we reported multiple plasmonic patterns (GNs) combined with conventional simple nanograting and NP patterns, which exhibited a high light amplification effects compared to single grating patterns or single NP patterns.However, the light amplification effects of the proposed multiple plasmonic structure mainly occur with incident light with polarization direction nearly perpendicular to the grating pattern due to the conventional unidirectional shape of the single nanograting pattern. [28,29]In contrast, the hierarchical nanostructures in this study based on cross-nanogratings with NPs (CNs) induce strong light amplification effects regardless of the polarization direction of the incident light, significantly improving the performance of perovskite PDs.Our CN pattern-based perovskite PDs achieved a photoresponsivity (R) of 580 mA W À1 and a specific detectivity (D*) of 3.2 Â 10 12 Jones, which were 420% and 990% higher than those of flat perovskite PDs without plasmonic effects, respectively.In addition, the CN patterns exhibited more effective light amplification effects and higher PDs performances compared with other types of plasmonic patterns, including onlyunidirectional nanogratings (grating), GNs, and only-crossnanogratings.Moreover, comprehensive analyses of the electric field enhancement and far-field scattering were conducted using the finite-difference time-domain (FDTD) method and photoluminescence analyses were also conducted to validate the robust plasmonic effects of the hierarchical CN-patterned electrodes.Furthermore, flexible 10 Â 10 arrays of the CN pattern-based perovskite PDs, fabricated on a flexible plastic substrate, demonstrated exceptional mechanical stability by detecting incoming photonic signals at a high resolution even when subjected to a bending radius as small as 8 mm, and maintained their performance throughout over 1000 cycles of repeated bending tests.These results highlight the significant potential of our hierarchical plasmonic patterns in a wide range of perovskite optoelectronic devices, encompassing photovoltaics, lightemitting diodes, lasers, as well as organic and quantum dotbased optoelectronic devices.

Results and Discussion
To clarify the benefits of our hierarchical CN-patterned electrodes, we also prepared flat, grating, CG, and GN-patterned electrodes for comparison.Figure 1a shows schematic illustrations of these patterns.The CG pattern has a crossed grating pattern where the grating patterns cross each other, and the GN and CN patterns have a hierarchical structure in which smaller NP patterns are additionally present on the grating and CG patterns, respectively.The strong light trapping effect of the single grating patterns is almost effective only for incident light with an electric field perpendicular to the direction of the grating patterns while the presence of two gratings in the CG pattern, where the grating patterns cross each other perpendicularly, results in a stronger light trapping effect for incident light regardless of the polarization direction of incident light.In addition, the hierarchical CN pattern has further enhanced the effective light trapping effect owing to the surface plasmon resonance effects from NP pattern (i.e., LSPR and SPP modes) and grating pattern (i.e., SPP mode), as a consequence of the presence of the NP patterns on the CG patterns. [28,29]Thus, these hierarhical CN-patterned electrodes are expected to greatly enhance the light absorption in the perovskite photoactive layer compared with other types of patternes including flat, grating, CG, and CN patterns.
Hierarchical CN patterns were prepared by combining block-copolymer and nano-imprinting lithography (Figure 1b).Initially, polystyrene-block-poly(methyl methacrylate) (PS-b-PMMA) films were prepared on Si/SiO 2 substrates by spincoating and were annealed for the phase separation of PS and PMMA.Then, nano-imprinting was performed using a CGpatterned polydimethylsiloxane (PDMS) stamp and PMMA domains were selectively removed by conventional UV/acid treatments.Figure 1c shows scanning electron microscope (SEM) images of CN-patterned PS films, which show clear CG patterns by nanoimprinting and circular nanoholes by PMMA removing.SEM images of PS films without nanoimprinting process are also shown in Figure S1, Supporting Information.After that, CNpatterned SiO 2 films were prepared by dry etching process using CN-patterned PS films as a mask.During the dry etching process, the nanohole patterns of the PS film led to the formation of bumpy NP patterns of SiO 2 (Figure 1d).Finally, CN-patterned PDMS stamps were prepared by conventional PDMS molding process using CN-patterned SiO 2 as a master mold.The plasmonic CN-patterned Au films were fabricated by Au deposition on the CN-patterned PS film prepared through the nanoimprinting process for PS films on the Si/SiO 2 substrates using CN-patterned PDMS as a stamp (Figure 1e).
The SEM image of the plasmonic CN-patterned Au film is shown in Figure 1f, which exhibits a surface morphology very similar to the CN-patterned SiO 2 with NP patterns along with the CG pattern.Figure S2, Supporting Information, shows the diameter distribution of the NP patterns, with an average size of 22.6 nm.Similarly, other plasmonic patterns including grating, GN, and CG Au layers were created in a comparable manner to the CN Au layer.This entailed substituting PS for PS-b-PMMA in the case of CGs, using a single grating-patterned PDMS stamp instead of a CG-patterned PDMS for GNs, or using both for grating patterns.The SEM images of other plasmonic Au films are shown in Figure S3, Supporting Information.Figure 1g shows photograph images of plasmonic Au layers.The diffraction colors of light are clearly shown for the grating and CG-patterned Au layers, whereas in the GN and CN-patterned Au layers containing NP patterns, the diffraction colors are not visible due to the interference of light diffraction by the NPs.The flat Au layers also did not exhibit any diffraction colors (Figure S4, Supporting Information).Figure 1h shows the measured absorption of flat and all types of nanopatterned 20 nm thick Au layers.All types of nanopatterns showed higher absorption characteristic than the flat Au layers owing to enhanced plasmonic resonance effect engendered by nanopatterned Au layer.
The GN and CG-patterned Au layers showed higher absorption than the grating-patterned Au layers, while the CN-patterned Au layers demonstrated the highest absorption across the entire visible light spectrum owing to the high density of localized surface plasmon resonance.These high absorptions indicate the exceptional light trapping capabilities of plasmonic CN patterns. [30,31]urthermore, the results closely resemble the absorption spectrum obtained through numerical calculations using the Lumerical FDTD method (Figure 1i and Figure S5, Supporting Information).Here, the NPs of the GN and CN patterns were randomly placed without overlapping.
The fabrication process of perovskite layers on CN-patterned Au layers is shown in Figure 2a.We selected polyvinyl alcohol (PVA) as an organic insulating and planarization layers for the nanopatterned Au layers.PVA solution was spin-coated on the Au layers and dried in a vacuum chamber to evaporate the solvent.Subsequently, methylammonium lead iodide (MAPbI 3 ) solution was spin-coated on the PVA films and annealed for crystalization.Figure 2b shows an atomic force microscopy (AFM) image depicting the surface of the PVA film on the hierarchical CN-patterned Au layer showing the planarity with 0.5 nm root mean squar (rms) roughness.AFM images of the PVA films fabricated on the other types of Au layer also showed the high planarity with low roughness values of 0.9, 1.6, 0.7, and 0.6 nm rms, respectively (Figure S6, Supporting Information).The roughness of the PVA film based on CG patterns decreased compared to the grating pattern, and there was a further decrease in roughness when NP patterns were additionally added.Figure 2c and Figure S7a-e, Supporting Information, show SEM images of perovskite films on PVA layers.All perovskite films were uniformly formed with dense structures, with a moderate excess PbI 2 distributed along the grain boundaries as passivating point defects.Interestingly, the perovskite layers based on the CG and CN patterns exhibited larger grain sizes as compared with other perovskite layers, probably owing to the surface energy change of the PVA layer in the heterogeneous nucleation of the perovskite film (Figure S7f-j, Supporting Information). [32]The average grain sizes perovskite layers based on flat, grating, CN, CG, and CN patterns are 259, 251, 263, 301, and 353 nm, respectively.It is noted that the perovskite films grown on the PVA layers exhibited a larger grain sizes than those prepared on SiO 2 films (average grain size: 202 nm), indicating that PVA layer contributes not only to the functions of planarization and insulation but also to the formation of larger perovskite grains (Figure S8, Supporting Information).These perovskite films with larger particle size and fewer grain boundaries also contributes to the improvement of PDs performance, owing to the reduction in trap/defect density and recombination rate, which results in an extended carrier lifetime for photo-generated carriers and an increased amount of photocurrent. [33,34]The average sizes of perovskite grains are summarized in Table S1, Supporting Information.Figure 2d and Figure S9, Supporting Information, show colored and gray-scale cross sectional SEM images of the perovskite film based on CN patterns showing well-stacked films of Si/SiO 2 /PS/Cr/Au/PVA/perovskite structures.Figure 2e shows the X-Ray diffraction (XRD) patterns obtained from perovskite layers, revealing well-matched perovskite crystal structures that include strong Au peaks originating from the plasmonic Au layer and small PbI 2 peaks.A grazing incidence wide-angle X-Ray diffraction (GIWAXS) analysis was also performed to investigate the influence of plasmonic patterns on the crystalline structure of perovskite films (Figure S10, Supporting Information).The observed diffraction patterns revealed an isotropic signal from the (110) plane of 3D MAPbI 3 at q z ≈ 1.0 Å À1 , as well as a smaller peak at q z ≈ 0.9 Å À1 , indicating the presence of small amounts of PbI 2 . [35,36]Furthermore, the (110) diffraction peak of the MAPbI 3 perovskite in all 2D GIWAXS patterns shows a consistent intensity scale, indicating a similar degree of crystallization among the different perovskite films based on flat and various plasmonic patterns.
To study the influence of the light amplification effect resulting from our plasmonic patterns on the exciton generation characteristics of perovskite films, we conducted photoluminescence (PL) measurements (Figure 2f ).The PL intensities exhibited a significant enhancement as patterns with more complex and hierarchical structures became more pronounced, resembling the reflectance results.The PL intensity of the perovskite films based on grating, GN, CG, and CN patterns showed respective increases of 1.2, 1.6, 1.5, and 1.8 times compared to perovskite films based on flat patterns, indicating the enhanced exciton generation rate owing to the effective light trapping effects of our plasmonic patterns, as well as effective suppression of the PL quenching of PVA via prevention of the charge transfer between the perovskite film and the Au layer. [21,37,38] Next, we fabricated PDs using perovskite films based on various types of plasmonic patterns to investigate and compare their effects on electronic devices.The perovskite PDs had a lateral device structure with Si/SiO 2 /PS/Cr/Au/PVA/MAPbI 3 /Au electrodes as shown in Figure 3a.In this design of device structures, the plasmonic Au layer acts as a back reflector, enhancing the light absorption properties of the perovskite light harvesting layer.In addition, the PVA layer serves as an insulating layer to prevent the significant resistance decrease that may occur when the Au plasmonic pattern comes into contact with the perovskite film.This approach allows us to investigate the performance improvement of the perovskite PD solely due to the plasmonic light amplification effect.To investigate the overall performance enhancements of the PDs in the visible light range, the performance of the perovskite PDs were measured under various light intensity conditions for red (670 nm), green (532 nm), and blue (450 nm) light.The current-voltage (I-V ) characteristics of the CN pattern-based perovskite PDs under red light are shown in Figure 3b.Under dark conditions, these CN patternbased perovskite PDs exhibited a mere current of only 0.13 nA while the current increased significantly more than 25 times under light illumination, indicating high photosensitivities of our perovskite PDs with low dark currents and high light currents.The I-V curves for all types of plasmonic pattern-based perovskite PDs under the illumination of red, green, and blue light at various intensities are shown in Figure S12, Supporting Information.
Figure 3c shows the photocurrent (light currentdark current) values for various plasmonic patterns and flat pattern-based perovskite PDs under red light at 10 μW cm À2 light intensity.All plasmonic patterns-based PDs exhibited higher current levels than flat pattern-based PDs, with CN-based PDs showing the highest value at 3.22 nA.The GN and CG pattern-based PDs showed a similar values of around 1.75 nA, followed by the grating-based PDs at 1.40 nA.As the complexity of plasmonic patterns increased, there was a tendency toward higher photocurrent values.This trend was also observed under different wavelengths, including green and blue illumination (Figure S13, Supporting Information).
To quantitatively compare the performances of the different types of perovskite PDs, the R, and external quantum efficiency (EQE) were calculated using the following equations where I light is the current under light illumination, I dark is the current under dark condition, P inc is the incident illumination power, P int is the incident light intensity (i.e., the incident optical power density) in the effective area, h is Planck's constant, c is the speed of light, e is the charge of an electron, A is the active area, and λ is the wavelength.In addition, D* is a key performance parameter for PDs, typically indicating the smallest detectable signal, as follows where A is the effective area of the detector in cm 2 , B is the bandwidth, NEP is the noise equivalent power, and i n 2 1=2 is the measured noise current.If shot noise from the dark current is the major factor contributing to noise limiting the D*, then D* can be simplified as [39] The R, EQE, and D* values of perovskite PDs based on various Au patterns (flat, grating, GN, CG, CN) under red, blue, and green lights at various intensities of light are shown in Figure 3d,e,f, respectievely.The 2D version graphs of Figure 3d-f are also presented in Figure S14, Supporting Information.The enhancements in R, EQE, and D* significantly increased as the light intensity decreased, which is a commonly observed phenomenon. [29,40]nder a red light intensity of 10 μW cm À2 , R values were significantly improved by up to 4.2-fold, from 140.5 mA W À1 for the flat perovskite PDs to 588.6 mA W À1 for the CN pattern-based perovskite PDs, while those of the grating, GN, and CG pattern-based perovskite PDs were 233.9, 332.4,and 335.0 mA W À1 , respectively.This indicates that the perovskite PDs based on CN patterns demonstrated a higher photocurrent  S2 and Figure S15, Supporting Information.It is noted that these R and D* values are the highest among all reported values for plasmonic perovskite PDs utilizing plasmonic structures in perovskite PDs without using additional conductive materials.Furthermore, the D* value is highly competitive with those for PDs incorporating conductive materials including graphene and conducting polymers as shown in Table S2, Supporting Information, particularly surpassing the highest reported D* value of 7.1 Â 10 11 Jones for conventional MaPbI 3 -based plasmonic perovskite PDs by more than four times.This indicates that our hierarchical CN plasmonic pattern not only exhibits a high light amplification effect but can also be efficiently integrated into perovskite PDs.Furthermore, real-time photoswitching at increasing intensities for each wavelength was conducted using CN pattern-based perovskite PDs. Figure 3g-i demonstrate that our CN pattern-based PDs maintain stable current output signals under repeated light on/off cycles at 10 μW cm À2 of red, green, and blue light, respectively.They exhibited a gradual increase in current with increasing light intensity and show rapid current changes when the light was turned on or off.The rise time (t r , defined as the time required to go from 10% to 90% of the maximum photocurrent) and decay time (t d , defined as the time required to go from 90% to 10% of the maximum photocurrent) of our devices were estimated to be shorter than 122 and 52 ms, respectively, as shown in Figure S16, Supporting Information.
To further investigate the optical performance of each nanopattern structures, we have calculated the electric field enhancement of each nanograting patterns, for x-, and y-polarization incidence, respectively.Demonstrated by the electric field of each nano patterns under 532 nm light (Figure 4a), the grating confines the electric field within the gap when subjected to x-polarized light.In contrast, the CG pattern exhibits field enhancement within the grating gap under both xand y-polarized light.Furthermore, the GN and CN patternes exhibits notably amplified field enhancements in comparison to the grating and CG patterns, where the electric field was significantly enhanced between the NP patterns.However, the electric field at the grating gaps of GN and CN each followed the tendency of grating and CG, respectively, where the field enhancement of the gap of GN was enhanced under x-polarized light, and CN under both xand y-polarized light.The calculated electric field for 450 and 670 nm are illustrated in Figure S17 and S18, Supporting Information.In addition, we calculated the maximum electric field for each nano pattern structures under 0°, 15°, 30°, 45°, 60°, 75°, and 90°polarized light, respectively, indicating that the CN pattern dominates the electric field across various polarization angles.(Figure S19, Supporting Information).The average maximum electric field for each pattern are shown in Figure 4b.The electric field enhancement intensity at 450, 532, and 670 nm, exhibited the following order: CN, GN, CG, and grating, where CN achieved highest field enhancement in each RGB wavelength of light.
Moreover, to study the light trapping performance of our nano pattern, we have calculated the far-field scattering of each nano pattern structure as shown in Figure 4c.Under normal incident light, the scattering intensity depending on the scattering angle (φ) was calculated.Note that the scattering intensity at high φ was largest for CN under x-polarized light (670 nm).We have illustrated the far-field scattering in 2-dimensional (2D) graph under x-and y-polarization light (670 nm), respectively, as shown in Figure 4d,e.Under illumination of x-polarized light, the scattering intensity at high φ shows similar tendency with average maximum electric fields, while the difference between each nano patterns was small.In contrast, for y-polarized light, the scattering intensity at high φ differed significantly between each The standard deviation.These parameters were averaged from at least five devices.
structures, which can be attributed to the strong plasmonic resonance existing of CN.The far-field scatterings in 2D graph under 532 and 670 nm wavelengths of light are also shown in Figure S20, Supporting Information.The experimental confirmation of increased absorbance and perovskite photoluminescence, along with the theoretically confirmed order of increased electric field and scattering associated with the growing complexity of the pattern's structure, aligns with the performance order of photodetector devices based on these patterns.Therefore, the performance enhancement in the photodetector is reasonably inferred to be facilitated by the plasmonic light amplification effects induced by plasmon resonance.
To investigate the potential utility as a flexible sensor platform, a 10 Â 10 flexible perovskite PD arrays based on CN patterns were fabricated on polyethylene naphthalate (PEN) substrates using the same device configuration and fabrication process of the unit device.To assess the mechanical flexibility and operational stability of the perovskite PDs, electrical characterization was conducted at various bending radii.Figure 5c presents the average photocurrent and dark current values of the CN-patterned PDs measured at six different bending radii (20, 18, 16, 14, 10, and 8 mm), including the initial and recovered states.The flexible perovskite PDs showed a stable and consistent photocurrent and dark current performance, closely resembling the values obtained in the initial state without any bending, even when the bending radius was reduced to as small as 8 mm.This stability persisted in the recovered state as well.Moreover, the devices showed high mechanical stability, with consistently stable average photocurrent and dark current values monitored at 200 intervals over a maximum of 1000 cycles at a bending radius of 8 mm (Figure 5d).
Furthermore, the photoimaging sensing performance of the 10 Â 10 PD array was also investigated by exposing the star-shaped red laser to the center of the sensor matrix (Figure 5e).The photonic signal was illustrated by 2D mapping as a function of the ratio of photocurrent/dark current values (Figure 5f ).As depicted in Figure 5f, the 2D pattern of changes in the photocurrent/dark current values were distributed in a star shape.The color discrepancy between the center and edges resulted from the non-uniform intensity of the light source, but the performance of the device was uniform.These demonstrations with the high-performance, flexible, and multifunctional light-detection capability of perovskite PDs using our novel plasmonic hierarchical CN patterns indicate that CN pattern based perovsktie PDs and their associated fabrication processes can potentially be used to make high-performance flexible optoelectronic devices.

Conclusion
In summary, we successfully fabricated hierarchical plasmonic CN nanopatterns composed of CG and NP patterns, and utilized them to demonstrate high-performance perovskite PDs.In comparison with flat pattern based perovskite PDs, the CN pattern based perovskite PDs exhibited significant improvements in R perovskite 10 Â 10 matrix exhibited highly sensitive spatiotemporal photomapping behavior with exceptional mechanical durability, highlighting its strong potential for application in flexible image sensors.Additionally, PL mapping and theoretical calculations were employed to further explore the advantages of CN patterns, confirming their more effective light amplification effects compared to other monotonous patterns.
These results will pave the way for the high-performance perovskite optoelectronic devices utilizing plasmonic structures, not only for enhancing PDs but also for advancing solar cells, light-emitting diodes, and lasers to new levels of performance.

Experimental Section
Fabrication and Characterization of Hierarchical CN Plasmonic Patterns: As illustrated in Figure 1b, a PS-b-PMMA solution (2 wt% in toluene) was spin-coated (≈70 nm thickness) onto a Si/SiO 2 wafer substrate.Then the substrate was annealed for 48 h at 180 °C in a vacuum oven.After the annealing process, the nanoimprinting process of a PS-b-PMMA film was carried out using a cross nanograting patterned PDMS stamp for 10 min at 130 °C.After cooling the substrate at room temperature and lifting off the PDMS stamp, PMMA was selectively etched with a traditional UV/acetic acid treatment to make hierarchical CN-patterned PS film.Subsequently, inductively coupled plasma (ICP) etching was performed for 5 min to remove the SiO 2 layer using a CN-patterned PS pattern as the mask.(CF 4 /CHF 3 /O 2 /Ar flow rate was 10/30/10/10 sccm).Following that, the CN-patterned SiO 2 wafer was cleaned and treated with a fluorinated self-assembled monolayer (F-SAM).To make a CN-patterned PDMS stamp, conventional liquid PDMS precursor solution (1:10 ratio of curing agent to silicone elastomer) was poured consecutively onto the substrate.For making cross nanograting patterned PDMS stamp, instead of using a PS-b-PMMA solution, a PS solution (2 wt% in toluene) was spincoated, and then the nanoimprinting process was carried out on the PS film using a nanograting-patterned PDMS stamp, once each in the horizontal and vertical directions.The grating-patterned PDMS stamp was obtained by pouring the PDMS precursor solution onto a grating mold (Thorlab, GH13-36U, Periodicity = 278 nm) and curing at 60 °C for 2 h.Afterwards, following the ICP etching and F-SAM surface treatment steps, the PDMS precursor solution was poured and cured.The ICP etching was conducted using a Sntek Dry Etcher_ICP.The SEM images were obtained using a Hitachi cold SEM microscope (S-4800).The reflectance spectra were measured on a spectroscopy (JASCO/V-770).
Fabrication and Characterization of Perovskite Films and Perovskite PDs Based on CN Patterns: The PS solution (2 wt% in toluene) was spin-coated (≈70 nm thickness) onto a Si/SiO 2 wafer substrate.Then, the nanoimprinting process on the film was performed utilizing an CN-patterned PDMS stamp.Afterwards, Cr and Au were thermally deposited with thicknesses of 4 and 100 nm, respectively.On top of that, a PVA solution (30 mg mL À1 in distilled water) was spin-coated at 2000 rpm for 30 s and dried in a vacuum chamber for 24 min.After the annealing process, the MAPbI 3 perovskite precursor (0.75 M) solution was spin-coated at 5000 rpm for 60 s, followed by annealing at 100 °C for 30 min in the N 2 -filled glovebox.Finally, Au electrodes were thermally deposited with a thickness of 30 nm on perovskite film.The AFM images were obtained using a Park System NX10 in a tapping mode.The SEM images were obtained using a Hitachi cold SEM microscope (S-4800).XRD measurements were carried out on a Panalytical Empyrean X-Ray diffractometer.The PL spectra were obtained using a spectrometer (SpectraPro HRS-300).The bright-field and PL optical images were collected by a custom microscope (Olympus BX53).Samples were excited with a light source (012-63 000; X-CITE 120 REPL LAMP).The current-voltage characteristics and real time photo switching of perovskite PDs were measured in a vacuum chamber using a Keithley 4200-SCS semiconductor parametric analyzer.
Optical Calculation: The absorbance and electric field density were calculated using a commercial FDTD solver (Lumerical).The simulation setup had periodic boundary conditions in the x-and y-directions with a periodicity of 278 nm and a perfectly matched layer in the z-direction.The planewave source was illuminated to the structure.
The far-field scattering was calculated using a commercial finite element methods (Multiphysics, COMSOL v6.1).The simulation used periodic boundary conditions in the x-and y-directions and perfectly matched layer in the z-direction.

Figure 1 .
Figure 1.a) Schematic illustrations of grating, GN (grating pattern with NPs), CG (crossed grating pattern), and CN (CG pattern with NPs) plasmonic patterns.b) Schematic diagram of fabrication process of CN-patterned PDMS stamp.c,d) SEM images of CN-patterned c) PS and d) SiO 2 layers.e) Schematic diagram of fabrication process of CN-patterned Au layer.f ) SEM image of the CN-patterned Au layer.g) Photograph images of grating, CN, CG, and CN-patterned Au layer-based substrates.h) Absorbance spectra of 20 nm thick Au layers as a type of pattern.i) Calculated absorbance spectra of five types of 20 nm thick Au pattern systems under light polarized at 45°to the direction of the nanograting patterns (i.e., flat, grating, GN, CG, and CN patterns).

Figure 2 .
Figure 2. a) Schematic diagram of fabrication process of PVA and MAPbI 3 perovskite films on a GN-patterned Au layer.b) AFM image of PVA film on GN-patterned Au layer.c,d) c) Top view and d) colored cross-sectional view SEM images of perovskite layers based on CN-patterned Au layers.e) XRD spectra of perovskite films based on flat, grating, GN, CG, and CN-patterned Au layers.f,g) f ) PL spectra and g) PL-OM images of perovskite films based on flat, grating, GN, CG, and CN-patterned Au layers.
Figure 2g and Figure S11, Supporting Information, show PL-optical microscopy (OM) images of perovskite films based on each pattern, along with their corresponding OM images, clearly demonstrating the differences in PL intensities despite the similar morphology of the perovskite films.

Figure 3 .
Figure 3. a) Schematic diagram of flexible perovskite PD structure based on CN-patterned Au layer.b) I-V characteristics of CN-based perovskite PDs under red light illumiation and dark conditions (inset: enlarged I-V characteristics of dark conditions).c) photocurrent-V characteristics of perovskite PDs based on flat, grating, GN, CG, and CN patterned Au layers under red light at 10 μW cm À2 .d-f ) d) R, e) EQE, and f ) D* of perovskite PDs based on flat, grating, GN, CG, and CN patterned Au layers under various intensities of red, green, and blue light (5 V bias) g-i) photoswitching characteristics of perovskite PDs based on CN-patterned Au layer under various intensities of g) red, h) green, and i) blue light (5 V bias).
Figure 5a illustrates schematic images of the single flexible CN pattern-based perovskite PD within the PD arrays, featuring the PEN/SiO 2 /PS/Au/PVA/MAPbI 3 /Au electrodes structures.Figure 5b shows a photograph of the fabricated PD array, showing the overall substrate exhibiting a yellowish coloration owing to the plasmonic CN Au back reflectors.

Figure 4 .
Figure 4. a) Electric field of grating, GN, CG, and CN for x-(top), and y-(bottom) polarized light (670 nm).b) Maximum electric field for each nano pattern structures.c) 3D far-field scattering for each nano pattern under x-polarized light (670 nm).d,e) Normalized far-field scattering intensity for each nano pattern under d) x-, and e) y-polarized light (670 nm), respectively.