Machine learning-based prediction of fluorescence lifetimes of zero-dimensional antimony halides

Abstract

The fluorescence lifetime is a critical parameter for evaluating the fluorescence efficiency of zero-dimensional antimony halides (ZDAH) which showed broadband emissions and large Stokes shifts for optical applications. Current methods to control fluorescence lifetimes rely on adjusting experimental synthesis methods and face limitations including long-time experimental processes and high costs. To resolve these challenges, this work proposes an innovative machine learning-based approach for predicting fluorescence lifetimes of ZDAH, which is for the purpose of establishing relationships between material structures and emission properties. By constructing two types of feature datasets and training eight machine-learning models respectively, it has been found that models with only structure parameters can achieve prediction with four models, with a low prediction error of 0.7%. After adding PLQY and emission intensity, the prediction error drops further to 0.11%. This method enables rapid optimization of luminescence properties without requiring experimental synthesis, providing guidance for efficiently screening high-performance ZDAH. The study establishes an intelligent research pathway for developing and evaluating novel luminescent materials.