Speaker
Description
Plasmonic solid–molecule systems enable light-driven processes that couple electromagnetic field localization with interfacial energy transfer. However, the complexity arising from heterogeneous structures, dynamic interactions, and multiple coupled parameters limits quantitative understanding and predictive design. To quantitatively access and generalize these processes across plasmonic solid–molecule systems, machine learning (ML) methodologies are employed to probe dynamic processes and resolve chemical heterogeneity. These approaches were initially developed for surface-enhanced Raman spectroscopy sensing, enabling differentiation and quantitative analysis of multiplexed spectral signals in complex matrices. This is exemplified by the sensing of microplastics in wastewater and DNA fragments in biosamples. Building on this, ML is extended to catalytic systems with multiple coupled parameters (e.g., catalyst type, dosage, pH, etc.), where algorithms enable acceleration of catalytic discovery. Supported by high-throughput experimental platforms, this approach enables minimization of experimental effort while maximizing catalytic performance.