Machine learning identification of experimental conditions for the synthesis of single-phase white phosphors

Summary

Single-phase white phosphors for solid-state lighting are commonly designed using different dopants responsible for emissions in different spectral regions. However, the phenomena of energy transfer and concentration quenching often prevent any clear prediction of the accurate experimental conditions to be selected, leading to a time-consuming trial-and-error discovery process. In this article, a high-throughput experimental approach equipped with machine learning (ML) enabling an efficient identification of the experimental conditions for designing a white phosphor is demonstrated. Li₂BaSiO₄:Eu,Ce was selected to illustrate this strategy. A total of 88 samples were prepared from the initial synthesis of eight compounds with different concentrations of dopants followed by a post-treatment under a gradient of temperature. The decision tree model identified the experimental conditions for designing a white emission. The analysis of the experimental conditions to obtain other colors of emission, which were also identified by ML, enabled rationalization of the different mechanisms of energy transfer between dopants.