Porous perovskite films integrated with Au–Pt nanowire-based electrodes for highly flexible large-area photodetectors

Flexible, large-area, and stable perovskite photodetectors have drawn increasing widespread research attention for next-generation wearable and portable optoelectronic devices. However, high mechanical durability coupled with large device area and enhanced environmental stability has not been demonstrated yet to attain practical viability. Herein, a highly bendable, stable, and large-area (3 cm2) flexible polystyrene incorporated perovskite photodetector is presented. Due to the formation of a porous polystyrene-perovskite composite film in a single step it allows unprecedented mechanical stability, maintaining 85% of its original photocurrent value after 10,000 bending cycles at a bending angle of 120°. Equally crucial, the solution-processed self-assembled Pt–Au nanochains were developed to provide a simple and fast method of patterning the conductive and flexible electrodes onto the filter substrate. The optimized polystyrene-perovskite photodetector exhibits a high responsivity up to 2.73 A W−1, a maximum specific detectivity of 6.2 × 1013 Jones, and a superior switching ratio of 1.0 × 104. In addition, the polystyrene-perovskite photodetector yields excellent stability under the combined stresses of moisture, ambient air, and room light, and retains 92% of its original performance for over 30 days. All these results demonstrate that this work provides a facile and cost-effective approach that paves the way to develop high-performance, stable, and highly flexible optoelectronic devices.


INTRODUCTION
Large-area, cost-effective, and flexible photodetectors are required in multiple applications such as next-generation wearable optoelectronics, robotics, bio-imaging, illumination monitoring systems, and remote sensing [1][2][3][4][5] . Organometal halide perovskites, specifically CH 3 NH 3 PbI 3 (or MAPbI 3 ) has emerged as outstanding light-harvesting material in the optoelectronic field due to its long charge carrier diffusion length, low exciton binding energy, broadband absorption, and direct bandgap [6][7][8][9] . The added advantage of their solution based low-temperature fabrication process makes them suitable for application in flexible electronics and photonic devices 10,11 . Flexibility is primarily defined by the ability to bend these devices to achieve a low form factor or easy integration with a curved surface for mounting on non-traditional spaces. The common strategy in perovskite photodetectors (and others) for reducing the maximum strain on bending has been based on two common themes, 1. By making nano/micro-wire arrays to limit the dimensions of the material 12-14 , or 2. Using thin perovskite films to reduce the thickness 4,15,16 . Combined with the inherent challenge of stability in perovskites, either of these strategies has limited success and as a result, most of the previous reports on flexible perovskite photodetectors, either conducted only a few hundred bending cycles test at large bending radii or suffer from poor environmental stability 4,12,15,16 . Progress has been made recently in improving stability by encapsulation of the perovskite layer, along with achieving ultra-fast response time in photodetectors 17,18 . It is also worth noting that most of the reported performances for flexible perovskite photodetectors are on the very small area < 0.4 cm 2 and very few reports exist on the large area (>1 cm 2 ) devices 4,19,20 . Therefore, there is a need to perform further studies and improve on the repeatable bending durability, while stabilizing the device performance. Further to ensure cost-effectiveness facile manufacturing processes resulting in large-area photodetectors, without significantly affecting the performance characteristics of the material are required.
Here, we demonstrate a highly bendable (10,000 cycles), stable, and large-area (3 cm 2 ) flexible polystyrene (PS) incorporated MAPbI 3 photodetectors. The high bendability is achieved by having a thin porous film of perovskite made by spin casting. This combines the effect of lowering the bending strain due to, the small thickness of the film and a lower modulus from the porous structure. The specific interaction between PbI 2 and MAI in the perovskite precursor solution and PS chains results in the formation of a cross-linked polymer-perovskite network, which assists in enhancing the performance and improving the stability of the PS-MAPbI 3 photodetectors. In this work, we used microporous polyamide filters as the flexible substrate for fabricating large-area photodetectors, owing to their excellent bendability, durability, good recyclability, and low-cost. The other significant challenge that needs to be addressed for making flexible devices is the requirement of equally durable and flexible high-performance electrodes 21 . The mechanical and electrical properties of electrodes fabricated on soft and bendable substrates degrade gradually due to fatigue damage during repeated deformations. To address this key issue, we synthesized self-assembled platinum-gold (Pt-Au) nanoparticle chains and deposited them on the polyamide filters by simple vacuum filtration to fabricate the patterned Pt-Au electrodes. This approach shows significant advantages in terms of conductivity, ease of fabrication, cost-effectiveness, and compatibility with the soft and flexible substrates used for the devices. In addition, the solution-processed self-assembled Pt-Au nanoparticles chains patterned as electrodes are highly bendable and provide a simple and fast method for fabricating flexible metal electrodes.
The optimum 1 wt % PS-MAPbI 3 device exhibits a low dark current of 1.1 nA at 2 V, a high responsivity of 2.73 A W −1 , a large detectivity of 6.2 × 10 13 Jones, and a superior switching ratio of 1 Department of Chemistry, Waterloo Institute for Nanotechnology, University of Waterloo, Waterloo, ON N2L 3G1, Canada. ✉ email: vmaheshw@uwaterloo.ca 1.0 × 10 4 as compared to plain MAPbI 3 device. Furthermore, the 1 wt % PS-MAPbI 3 flexible photodetector shows outstanding bending stability and retains a photocurrent of 85% of its initial value after 10,000 bending cycles at a bending angle of 120°. This also illustrates the robustness of the nanoparticle electrodes for bending. Under the same bending conditions, a substantial decrease in photocurrent to 50% was observed for plain MAPbI 3 device. Finally, the 1 wt % PS-MAPbI 3 flexible photodetector shows excellent stability and retains 92% of its original value even after 30 days (720 h) of continuous exposure of moisture, ambient air, and under room light, which was far better than that of the plain MAPbI 3 photodetector whose performance significantly deteriorates to almost zero after 8 days (192 h). This facile fabrication process of large-area flexible photodetectors with superior bending robustness coupled with improved environmental stability and reduced cost will pave the way for nextgeneration wearable and smart optoelectronics.

RESULTS AND DISCUSSION
Fabrication and opto-electronic performance A schematic representation of the fabrication process of flexible PS-MAPbI 3 photodetector is shown in Fig. 1a, b. First, the highly conductive Pt-Au electrodes were deposited on the flexible polyamide filter membrane (see I-V curves of Pt-Au electrodes in Supplementary Fig. 1) by solution filtration of micron size longchain networks of Au nanoparticles self-assembled in a solution using Pt 4+ ions and then reduced to form continuous Pt-Au nanonetworks. The TEM and high-resolution TEM (HRTEM) images of Pt-Au nanochains are presented in Supplementary Fig. 2. The TEM image illustrates the typical Pt-Au nanoparticles (~microns in size, with local nanowire width of~5-10 nm) chain structure. The pore size of the polyamide filter is around 1-2 µm as observed in the field emission scanning electron microscopy (FESEM) image (Fig. 1a). Then, the PS containing MAPbI 3 (or plain MAPbI 3 ) precursor solution was spin-coated and annealed on 2.0 × 1.5 cm 2 area Pt-Au-patterned polyamide substrate (shown in Fig. 1b). The cross-sectional FESEM images shown in Supplementary Fig. 3 illustrate that the thickness of the plain MAPbI 3 and 1 wt % PS-MAPbI 3 films are around 1.010 and 1.038 µm, respectively. The porous structure of the polyamide filter would lead to the formation of a porous perovskite structure on the filter during the spin-coating process. A FESEM image of the resulting porous PS-MAPbI 3 films is shown in Fig. 1b. A similar approach was used to fabricate the porous plain MAPbI 3 (without PS) films on the polyamide filter. A photograph of the Pt-Au electrodes on the flexible polyamide filter is displayed in Fig. 1c. Figure 1d and e shows the photograph of the resulting PS-MAPbI 3 device and its flexibility. The details of the fabrication process are in the experimental section. The Lewis-acid characteristic of PbI 2 leads to the interaction between PS and MAPbI 3 precursors and crosslinking of the polymer chains. While the cation-π electron interactions are the basis of coupling between the MAI and PS chains. We have confirmed and reported these interactions in our previous work 22,23 . The impact of PS incorporation into the perovskite was evaluated by Xray diffraction (XRD), and UV-Vis absorption spectroscopy. The diffraction patterns (Fig. 1f) showed no obvious changes with and without PS, and all the samples exhibit the same tetragonal perovskite crystal structure with the dominant (110) lattice plane. As displayed in Fig. 1g, the UV-Vis absorption spectra show a broad absorption in the range of 300-780 nm, indicating high light-harvesting capabilities over the ultraviolet to the visible spectrum. There is a small difference in the absorption value that can be due to the slight variation (±50 nanometers, Supplementary Fig. 3) in the thickness of the perovskite films, which revealed that the absorption properties of MAPbI 3 films with and without PS are preserved.
The photocurrent generation from the flexible PS-MAPbI 3 photodetector is schematically depicted in Fig. 2a. Our photodetector is based on the photoelectric effect in which the applied voltage results in an electric field that leads to the effective separation of electrons and holes generated on light illumination. Due to the close matching of the work function of Pt-Au (5.2 eV) with the highest occupied molecular orbital (HOMO) of PS-MAPbI 3 perovskite (5.4 eV), the collection of holes are favorable by the negative Pt-Au electrode in the symmetric Pt-Au/PS-MAPbI 3 (or MAPbI 3 )/Pt-Au device. Whereas, the electrons are not easily collected (or get trapped) due to the large energy level difference between the Lowest Unoccupied Molecular Orbital (LUMO) of PS-MAPbI 3 (3.9 eV) and work function of Pt-Au (5.2 eV). Figure 2b presents the photoresponse from both plain MAPbI 3 and PS-MAPbI 3 devices (with varying concentrations of PS) at 2 V and under the illumination of 1.0 sun. The plain MAPbI 3 device showed a photocurrent of 2.8 µA, which enhanced with the incorporation of PS. Specifically, the photocurrent for 0.5 wt % PS-MAPbI 3 device is 4.8 µA, and it increased substantially to 11.2 µA for 1 wt % PS-MAPbI 3 device. We attribute this to the fewer defects and suppressed charge carrier recombination by the cross-linking of PS within MAPbI 3 22 . Further, the addition of PS to 3 and 7 wt % caused a drop in photocurrent to 6.9 and 1.1 µA, respectively. The corresponding light switching on/off ratios is illustrated in Supplementary Fig. 4, and it also follows a similar trend. It is worth noting that the light switching ratio for 1 wt % PS-MAPbI 3 device (1.0 × 10 4 ) is 1 order of magnitude higher than that of plain MAPbI 3 device (1.0 × 10 3 ). The comparison reveals that among all the above PS concentrations, 1 wt % PS device gives the best performance due to the highest carrier lifetime and mobility, and simultaneously lowest, charge recombination effects, the density of defect traps, and ion migration, and therefore it was characterized in detail 22 . The dark currents and photocurrents were measured to quantitatively compare the performance of photodetectors made with both plain MAPbI 3 and 1 wt % PS-MAPbI 3 . For both the photodetectors, with increasing voltage the dark currents and photocurrents increase (Fig. 2c) due to greater strength of the electric fields, as it leads to more effective charge separation and collection. The 1 wt % PS-MAPbI 3 device exhibits a better performance with~3 times lower dark current and~4 times higher photocurrent than the plain MAPbI 3 devices. This can be attributed to the presence of PS (insulating material), reduced ionmigration effects, and lower carrier recombination in PS-MAPbI 3 photodetectors 22 . The photoresponse curves of the optimized 1 wt % PS-MAPbI 3 photodetector as a function of different wavelengths (410-710 nm) at a low bias of 2 V are presented in where I ph is the photocurrent, I d is the dark current, P in is the incident light power, q is the electron charge, and A is the effective area of the photodetector. The responsivity and detectivity of the 1 wt % PS-MAPbI 3 photodetector (Fig. 2e) in the spectral range from 400 to 750 nm are 5 and 7.5 times higher than the plain MAPbI 3 photodetector, respectively. We observe that the 1 wt % PS-MAPbI 3 Fig. 2f. The photocurrent increases with the light intensity, which can be ascribed to the increased number of photogenerated charge carriers in response to the higher photon flux 26 . It should be noted that the photocurrent at very low intensity (0.001 mW cm −2 or 1 µW cm −2 ) of 650 nm light is 45 nA, which is more than 1 order of magnitude higher than that under dark conditions (1.1 nA). The corresponding responsivity and detectivity are shown in Supplementary Fig. 6. The response time of the 1 wt % PS-MAPbI 3 photodetector was also determined and found to be <5 ms, which is the detection limit of our setup (Supplementary Fig. 7). The responsivity and detectivity of plain MAPbI 3 and 1 wt % PS-MAPbI 3 photodetectors using 2 V at different light intensities by attenuating 1.0 sun illumination (from solar simulator) using neutral density filters is presented in Fig. 2g. The maximum value of R and D* for 1 wt % PS-MAPbI 3 device is 2.73 A W −1 and 6.2 × 10 13 Jones at illumination intensity of 0.001 mW cm −2 , which is comparable and even higher than the previous reports on the flexible photodetectors (Table 1) 3,12,13,20,[27][28][29][30][31][32][33][34][35][36][37][38] . In contrast, the plain MAPbI 3 device achieves only R (0.61 A W −1 ) and D* (0.86 × 10 13 Jones), which implies that the incorporation of 1 wt % PS in MAPbI 3 greatly improved the photodetector performance. The higher R and D* observed under illumination from the solar simulator (a broad-spectrum source) compared to illumination at 650 nm is attributed to the broad absorption spectrum of MAPbI 3 which leads to greater absorption of photons of higher energies.
Flexibility and operational stability As mechanical stability is the major concern for the practical feasibility of flexible photodetectors for wearable electronics, we studied the bending durability of both plain MAPbI 3 and 1 wt % PS-MAPbI 3 photodetectors. Figure 3a shows the schematic representation of the bending of the flexible PS-MAPbI 3 photodetector. The device was bent at varied angles as shown in the inset of Fig. 3b, and the photocurrent was measured during the bending to ensure the performance reliability of the flexible device. The photocurrent of both flexible devices remains unchanged with the bending angle varying from 0°to 120° (  Fig. 3b). Even at the maximum bending angle of 180°, there is a slight decrease in photocurrent which can be due to the reduction of the light-receiving area at the bending state. The photocurrent retained to its original value as the device brings back to zero bending angles. Further, we measured the photocurrent of both devices after bending for 10,000 cycles at a fixed bending angle of 120°and a bending radius of 1.5 mm. For clarity, all the photocurrents were normalized to their respective initial values. As can be seen in Fig. 3c, the photocurrent of the 1 wt % PS-MAPbI 3 device maintained 100, 91, and 85% of its initial value after 1000, 5000, and 10000 bending cycles, whereas under the same bending conditions for plain MAPbI 3 devices the photocurrent dropped to 90, 68, and 50% of its initial value. The photodetectors are also evaluated for their long-term stability in operation. The plain MAPbI 3 and 1 wt % PS-MAPbI 3 devices with and without an adhesive layer were tested upon exposure to the ambient environment (in the air at a relative humidity of 35-40% under room light). Figure 3d shows the 1 wt % PS-MAPbI 3 devices maintain 92% of its initial photocurrent value after 720 h (30 days). In contrast, the plain MAPbI 3 devices completely degrade in just 192 h (8 days) with more than 60% loss within 3 days. Similarly, we also achieved a better stability for the 1 wt % PS-MAPbI 3 device without an adhesive layer in the ambient environment as compared to the plain MAPbI 3 device (Supplementary Fig. 8).
The improved stability of 1 wt % PS-MAPbI 3 devices can be attributed to the combined effects of the presence of PS (hydrophobic in nature) which leads to passivation of defects, and reduced ion-migration effects 22,39 . Further to ensure the reliability of the photoresponse measurement, the photoresponse of the plain MAPbI 3 and 1 wt % PS-MAPbI 3 devices were recorded as a function of time at 2 V under 650 nm light irradiation (0.1 mW cm −2 intensity), as presented in Fig. 3e. Both the devices exhibited very stable performances. Measured for 60 min, the photocurrent of the plain MAPbI 3 device was measured to be at 37 nA, whereas the 1 wt % PS-MAPbI 3 device exhibited a photocurrent of 169 nA. These results demonstrate that our 1 wt % PS-MAPbI 3 photodetector has excellent mechanical flexibility and durability as well as outstanding environmental stability. The mechanical stability is also attributed to the porous and chain-like structure of the Pt-Au electrode ( Supplementary Fig. 2) which can endure repeated bending without failure. It also shows that the Pt-Au electrode remains chemically stable for days and under device operating conditions, without any degradation with the perovskite layer. Ag nanowires in comparison can undergo redox reactions with the perovskite layer with relative ease impacting stability [40][41][42] . Ag nanowires also have the challenge of mechanical and thermal stability and junction resistance 43 . Further the Pt-Au electrodes are made using a room temperature based self-assembly process for synthesis and simple vacuum filtration for control of the deposition process. The conductivity/cost ratio of the Pt-Au electrodes is~63,000 Sm −1 /US$, while for Ag nanowire-based electrodes this can range from 8000-123,000 Sm −1 /US $, making them comparable in cost. It is also higher than pure Au electrodes (1023 Sm −1 /US $) 21,44 .
In summary, we have demonstrated the feasibility of using PSincorporated MAPbI 3 networks to develop a porous film on a flexible filter membrane for fabricating large-area ultra-flexible photodetector. Low-cost flexible Pt-Au electrodes were patterned by using the facile solution-processed self-assembled nanochains. The optimum 1 wt % PS-MAPbI 3 device showed extremely high bending stabilities, maintaining a photocurrent of 85% of their original values after 10,000 bending cycles under a large bending angle (θ = 120°). The responsivity and detectivity of 1 wt % PS-MAPbI 3 flexible photodetector reached 2.73 A W −1 and 6.2 × 10 13 Jones, respectively. This is among the best-reported performance with high mechanical durability when compared to other flexible perovskite photodetectors. Moreover, the PS incorporation in MAPbI 3 showed long-term device stability for more than 30 days in the ambient environment. The presented low-cost, ultraflexible, highly stable, and large-area photodetectors will lay a strong foundation for developing next-generation wearable and portable optoelectronic devices.

Preparation of Pt-Au nanoparticles chains
Hundred and thirty microliters of 4.5 mg mL −1 platinum (IV) tetrachloride (PtCl 4 , Sigma Aldrich) were mixed with 1 mL citrate capped gold nanoparticles (~10 nm) solution, purchased from BBI Solutions OEM Limited, and then left on the shaker until the solution color changed to dark blue due to the self-assembly of Au nanoparticles. Seventy-five microliters of 4 mg mL −1 sodium borohydride (NaBH 4 ) was then added to reduce the Pt 4+ ions, the color of the solution changes from initial dark blue to black due to the formation of Pt metal.

Preparation of MAI precursor
Methylammonium iodide (MAI) was synthesized by dropwise addition of 30 mL of hydroiodic acid (Sigma-Aldrich, 57 wt. % in water) to 27.8 mL of methylamine (Sigma-Aldrich, 33 wt. % in absolute ethanol) under constant

Characterization and measurement
The surface morphology of the self-assembled Pt-Au nanoparticles, polyamide filter, and perovskite films was observed by FESEM (FESEM, Zeiss Ultraplus). Transmission electron microscopy (TEM) images were obtained with a LEO 912AB transmission electron microscope. The phase of the perovskite samples was characterized by X-ray diffraction (XRD) using a PANalytical Empyrean diffractometer with Cu-Kα radiation (λ = 1.54 Å). The optical absorption spectra were recorded using a UV-Visible spectrophotometer (Perkin Elmer Lambda 750). The photoelectrical measurements of the device under different light wavelengths and intensities, including I-V, photocurrent, dark current, spectra responsivity, detectivity, and EQE were recorded by a Keysight 3458 A Digital multimeter combined with 6614 C 50 Watt system power supply from Agilent Technologies in ambient condition. All the measurements on Pt-Au/PS-MAPbI 3 (or MAPbI 3 )/Pt-Au devices were conducted on the probing station by using a two-probe method. The voltage was applied on the plain MAPbI 3 or PS-MAPbI 3 porous films by connecting one probe to the first Pt-Au electrode on the filter and another probe connected to the second Pt-Au electrode. Simulated AM 1.5 G irradiation (100 mW cm −2 ) with a xenon-lamp based solar simulator (Newport Oriel Instrument 67005, 150 W Solar Simulator) was used as the illumination source. The Newport optical filters were used to produce incident light at various intensities. The light intensity was calibrated by an NREL calibrated KG5 silicon reference cell to minimize spectral mismatch. All measurements were conducted at room temperature and in ambient air. The active area of the devices was 1.2 cm 2 . For the bending test, an MFA motorized miniature linear stage was used to bend the flexible device at a particular angle and to perform 10,000 bending cycles at a constant speed. The long-term stability test of the devices was tested by directly exposing the devices in ambient air with 35−40% relative humidity and under room light. The humidity was