Quantum Chemical Calculations and Machine Learning Predictions Innovate Synthesis for High-Performance Optical Gold Nanorods

Understanding the optical properties of gold nanorods (GNRs) in the colloidal state is crucial to engineering them for versatile applications in many fields. Concomitant gold nanospheres (GNSs) are easily involved in GNR synthesis, incurring a negative effect on the GNR performance. To unravel the underlying mechanism, we constructed a GNR–GNS heterodimer to imitate their colloidal state and calculated the relevant optical and electronic properties through a quantum chemical approach. The calculations reveal that GNSs prevent certain charge-transfer excitations of adjacent GNRs by affecting the electronic structure and thereby the excitation behavior of the GNR. We synthesized 310 sets of GNR–GNS colloidal solutions with a seed-mediated growth method and then measured their absorption spectra to extract the datasets available for 11 machine learning algorithms. Among them, XGBoost had the best prediction accuracy of over 94%. A direct relevance from the initial synthesis parameters to the final optical properties of GNR–GNS colloids has been successfully identified by the machine learning approach, which could skip the cumbersome step-by-step procedure used for the conventional nanostructure characterization as well as optimize the batch GNR synthesis process with improved GNR performance simultaneously. Methodologically, such a three-in-one approach combining chemical synthesis, quantum chemical calculations, and machine learning predictions can be extended to other chemical synthetic studies, with methodological guidance to chemistry and materials science researchers.