Explainable Machine Learning for Solvents-Driven G-quartet Circularly Polarized Luminescence Material in Multilevel Anti-counterfeiting

Circularly polarized luminescence (CPL) materials are highly sought after for their unique chiroptical properties, but their rational design remains challenging. To address this, a machine learning (ML) platform that intelligently screens deep eutectic solvents (DES) for crafting G-quartet-based gels with tailored CPL is established. An ML classifier trained on experimental data predicts CPL activity with 0.916 accuracy. SHapley Additive exPlanations (SHAP) analysis identifies key DES molecular descriptors, guiding the discovery of novel materials that exhibit a record-high luminescence dissymmetry factor (g lum ) of 0.222 for nucleotide-based systems and achieve chirality inversion via DES modulation. The Sure Independence Screening and Sparsifying Operator (SISSO) algorithm further yields interpretable design formulas. Leveraging the distinct polarization of these gels, we demonstrate a fourth-level anti-counterfeiting strategy, decrypting concealed information via Morse and ASCII codes to enhance information security. This ML-driven methodology accelerates the discovery of high-performance CPL materials and opens new design avenues.