ML-Based Forecasting of Temporal Dynamics in Luminescence Spectra of <italic>Ag</italic>₂<italic>S</italic> Colloidal Quantum Dots

Abstract

The study delves into the temporal dynamics of luminescence in colloidal $Ag_{2}S$ quantum dots, utilizing time series forecasting techniques. Through an analysis of intensity measurements taken at different time intervals, it uncovers temporal trends and utilizes predictive models to anticipate future behaviour of luminescence spectra. The outcomes contribute to a more profound understanding of optimizing experimental conditions and foreseeing the evolution of these nanomaterials over time. Among the tested models, the most robust and effective approaches for predicting the decay of integral intensity within the first hour include polynomial features with regressors, particularly ElasticNetCV, Ridge, and Lasso, with $R^{2}$ scores of 0.74, 0.82, and 0.80, respectively. However, upon comparison with the results of additional experiment conducted over a duration of two hours, the Ridge model demonstrated the best performance in predicting the decay of integral intensity.

Keywords

• Ag₂S
• luminescence
• machine learning
• prediction
• quantum dots
• temporal dynamics