Femtosecond Laser Patterning of Perovskite Quantum Dots for Multifunctional Diffractive Deep Neural Networks

ABSTRACT

Diffractive deep neural networks (D ² NNs) are regarded as promising candidates for breaking through the bottlenecks of traditional electronic computing, owing to their inherent characteristics of light‐speed operation, parallel processing, and low energy consumption. However, most existing D ² NNs rely on passive diffractive layers whose optical responses are essentially non‐tunable, severely limiting their widespread applications. In this work, we propose a multifunctional D ² NN with simultaneous fluorescence image display and optical computing by femtosecond (fs) laser patterning of CsPbI ₃ perovskite quantum dots (PQDs) in a polyacrylonitrile matrix. Specifically, through multiple fs laser exposures, controlled PQDs can be locally produced, enabling both fluorescence generation and phase modulation. As a result, the device displays multiple distinguishable fluorescent images under incoherent green light illumination, serving as a human‐readable visual feedback interface. While the identical device acts as a diffractive computing layer for smile recognition tasks under coherent red light illumination. This work demonstrates the feasibility of integrating active visual display and optical inference within a single material system, providing a viable pathway for multifunctional, human‐machine interactive intelligence photonic systems.