Ensuring Excitation: Machine Learning a Phosphor's Excitation Band Position

The creation of innovative lanthanide-activated inorganic phosphors is pivotal for advancing energy-efficient LED lighting and backlit flat panel displays. The most fundamental property these materials must possess is the effective absorption of photon's produced by a blue InGaN LED chip for conversion into white light. The lanthanide 5d₁ excited state energy level that sets the excitation peak position is governed by the physical and chemical interactions between the lanthanide and inorganic host structure, including local environment, crystal structure, and compositional characteristics, making it challenging to predict a priori. This study introduces a new extreme gradient boosting machine learning method capable of quantitatively determining a phosphor's excitation wavelength. The initial focus is on Ce³⁺ due to its well-defined 5d₁ energy level observed in excitation and diffusion reflectance spectra. The model, constructed using data from 357 cation substitution sites across 337 Ce³⁺ unique phosphors sourced from literature and in-house experiments, was trained using leave-one-group-out cross-validation and, more importantly, experimentally validated through the successful synthesis of a novel blue-excited phosphor. The compound's excitation under commercial blue LED wavelengths aligned remarkably well with the model's prediction. These results demonstrate the transformative potential of data-driven approaches for expediting the discovery of blue-absorbing phosphors essential for next-generation LED lighting.