Direct photo-patterning of halide perovskites toward machine-learning-assisted erasable photonic cryptography.

The patterning of perovskites is significant for optical encryption, display, and optoelectronic integrated devices. However, stringent and complex fabrication processes restrict its development and applications. Here, we propose a conceptual methodology to realize erasable patterns based on binary mix-halide perovskite films via a direct photo-patterning technique. Controllable ion migration and photochemical degradation mechanism of iodine-rich regions ensure high-fidelity photoluminescence images with different patterns, sizes, and fast self-erasure time within 5 seconds, yielding erasable photonic cryptography chip, which guarantees the efficient transmission of confidential information and avoids the secondary leakage of information. The ultrafast information encryption, decryption, and erasable processes are attributed to the modulation of the crystallographic orientation of the perovskite film, which lowers the ion migration activation energy and accelerates the ion migration rate. Neural network-assisted multi-level pattern encoding technology with high accuracy and efficiency further enriches the content of the transmitted information and increases the security of the information. This pioneering work provides a strategy and opportunity for the integration of erasable photonic patterning devices based on perovskite materials.