Two-dimensional (2D) semiconductors are revolutionizing optoelectronic devices, finding applications in machine vision, bio-inspired systems, and in-sensor computing. 2D perovskite oxides, such as Ca2Nb3O10 and Sr2Nb3O10, are particularly attractive due to their tunable bandgap, superior photoconductive gain, efficient carrier extraction, and stability compared to metal halide perovskites. They address the scarcity of high-performance UV detectors among common 2D semiconductors. However, their widespread adoption is hampered by challenging synthesis conditions, leading to difficulties in large-area preparation with controlled thickness, and the persistent photoconductivity effect stemming from abundant traps that results in slow response speeds. This creates a trade-off between high responsivity and fast response. Current research largely focuses on individual nanosheet performance enhancement, neglecting large-area integration which is crucial for practical applications. This work aims to overcome these limitations by developing a scalable and universal approach for wafer-scale integration of 2D perovskite oxides, enabling advanced optoelectronic applications such as motion recognition.

The successful integration of 2D semiconductors like MoS2, WS2, and WSe2 in optoelectronics is linked to their suitability for large-area fabrication and heterogeneous integration with microelectronics. In contrast, 2D perovskite oxides suffer from harsh synthesis conditions and defect chemistry. While impressive photodetection capabilities have been demonstrated in previous studies, these often focus on individual nanosheets or utilize top-down strategies, which may not be suitable for large-scale applications. Therefore, there is a critical need for scalable synthesis and integration methods that simultaneously improve response time and maintain high photosensitivity to facilitate commercialization and research.

This study introduces a charge-assisted oriented assembly film-formation (COAF) process for large-area, ordered spatial orientation of Ca2Nb3O10 (CNO) nanosheets. CNO nanosheets were synthesized via a top-down process involving high-temperature calcination, ion exchange, and liquid-phase exfoliation. The COAF process uses a water-ethanol cosolvent system to control nanosheet dispersion and volatilization rate, promoting layer-by-layer stacking and minimizing aggregation. Spraying the solution onto a substrate creates a large-area, highly ordered film with controllable thickness. The process was shown to be scalable, producing wafer-scale (6-inch) films on n-type Si wafers. Characterization techniques such as SEM, AFM, XRD, TEM, XPS, and GIWAXS confirmed the high-quality and ordered orientation of the resulting films (S-CNO) compared to conventionally prepared films (D-CNO). Photodetectors were fabricated using the S-CNO films, and their optoelectronic properties were assessed, including responsivity, detectivity, and response speed. A 256-pixel array was integrated onto a flexible Parylene C substrate to demonstrate the versatile material-to-substrate integration and conformal imaging capabilities. Finally, a convolutional neural network (CNN) model was trained to recognize motion trajectories using spatiotemporal images obtained from the flexible array. The synthesis of Sr2Nb3O10 films using the COAF method demonstrated its universality for other 2D perovskite oxides. Space-charge-limited current (SCLC) measurements were performed to quantify trap density and explain the improved speed in S-CNO devices.

The COAF process successfully produced wafer-scale, highly oriented films of CNO nanosheets. The S-CNO films exhibited significantly improved properties compared to D-CNO films, including enhanced uniformity, smoother surface morphology, and preferred (001) crystal plane orientation. Steady-state and time-resolved photoluminescence (PL) measurements confirmed reduced carrier recombination rates in S-CNO films due to fewer defects. GIWAXS analysis further validated the highly ordered orientation of nanosheets at different depths in the S-CNO films. The fabricated S-CNO photodetectors achieved high responsivity (11.9 A W⁻¹ at 280 nm) and detectivity (3.71 × 10¹⁴ Jones), surpassing state-of-the-art perovskite UV devices. Crucially, the S-CNO devices showed a dramatic improvement in response speed (τr = 42.4 ± 14.8 μs, τd = 1.77 ± 0.40 ms), about 10 and 20 times faster than D-CNO devices, respectively. SCLC measurements revealed significantly lower trap density in S-CNO films compared to D-CNO films, which explains the improved response speed. The flexible 256-pixel array demonstrated uniform photocurrents and excellent imaging capabilities, even under bending. A trained CNN model effectively recognized motion trajectories with over 99.8% accuracy, overcoming the ghosting issue typically present in devices with slow response times. The successful fabrication of Sr2Nb3O10 films using the COAF method indicated the universality of the approach for various 2D perovskite oxides.

This work addresses the critical challenges associated with the large-scale integration and application of 2D perovskite oxides in optoelectronics. The COAF method offers a scalable and efficient pathway to fabricate high-quality, highly oriented films, leading to significantly improved device performance. The achieved balance of high photosensitivity and fast response speed is a significant breakthrough, overcoming a major limitation of previous 2D perovskite oxide-based photodetectors. The successful demonstration of a flexible, high-performance 256-pixel array expands the potential applications of these materials in wearable and flexible electronics. Moreover, the ability to perform efficient motion recognition highlights the potential of these devices in advanced machine vision systems. These findings provide valuable insights for future research and development of 2D perovskite oxide-based optoelectronic devices.

This study successfully demonstrated the wafer-scale integration of 2D perovskite oxide nanosheets using a novel COAF process. This approach yielded devices with superior performance, exhibiting high photosensitivity, fast response times, and excellent flexibility. The integration into a 256-pixel array enabled accurate motion recognition, showcasing the potential for advanced applications. Future research could explore different 2D perovskite oxides, optimize the COAF process for even higher throughput, and investigate further integration with microelectronics for more sophisticated sensing and computing capabilities.

While the COAF process shows great promise, further optimization is needed to completely eliminate larger particles and incompletely exfoliated materials observed in the SEM images. The study primarily focused on Ca2Nb3O10 and Sr2Nb3O10; investigations into other 2D perovskite oxides are needed to assess the generalizability of the COAF process. The CNN model was trained on a specific dataset; its performance on more diverse datasets needs to be evaluated. Finally, long-term stability tests on the flexible devices under various environmental conditions are required to fully assess their robustness for practical applications.