Dynamic Multicolor Upconversion Through Excitation Pulse Duration Modulation toward Machine Learning Assisted Optical Encryption

Abstract

Optical encryption technology demonstrates significant potential for advanced information security applications due to its inherent advantages in high-speed operation, multidimensional processing, and parallel computation capabilities. However, current research in this field has predominantly focused on elementary optical anti-counterfeiting techniques and binary coding systems, with limited exploration of sophisticated encryption methodologies. In this study, a novel strategy is presented that employs NaYF₄ multilayer core–shell nanocrystals that enable dynamic full-color upconversion (UC) emission modulation under single-wavelength excitation, thereby facilitating high-capacity optical encryption through machine learning (ML)-assisted processing. Through systematic investigation of UC photophysical mechanisms, it is revealed that the full-color tunability originates from both excitation power dependence and excitation pulse width sensitivity mediated by rare earth ion cross-relaxation processes. The rich optical information generated through these mechanisms has been systematically organized into a comprehensive ML-constructed database, functioning as an optical “codebook” for encryption protocols. The developed ML framework demonstrates exceptional capability in identifying subtle optical signature differences, achieving over 98% recognition accuracy through database pattern matching. This system theoretically enables the encryption and decryption of 18⁸ distinct optical encryption patterns. These findings establish a new paradigm for optical encryption development and provide critical insights for advancing next-generation optical security systems.