Accelerated Design of Near-Infrared-II Molecular Fluorophores via First-Principles Understanding and Machine Learning

Abstract

Organic molecular fluorophores in the second near-infrared window (NIR-II) have attracted much attention in the recent decade due to their great potentials in both fundamental research and practical applications. This is especially true for biomedical research, owing to their deep light penetration depth and low bioluminescence background at the long wavelength. However, the fluorescence quantum yields (QY) of most NIR-II materials are very low, which are not ideal for practical applications. Although there is a growing need to discover new NIR-II fluorophores, most of them were designed based on experience, and the structures were limited to few molecular motifs. Herein, we report the design of high QY NIR-II fluorophores in solutions based on enhancing the rigidity of the conjugated backbones, which could be quantified by the Seminario method. A deep neural network was trained to predict the HOMO-LUMO energy gaps for a chemical library of NIR-II backbone structures. Hundreds of new NIR-II cores with low energy gap were discovered, and eight of them across different acceptor cores are found to have relatively rigid conjugated backbones. With further molecular processing or formulation, the proposed new fluorophores should boost the development of NIR-II materials for applications in a wide range of fields.