Machine learning investigation to predict the relationship between photoluminescence and crystalline properties of blue phosphor Ba0.9-xSrxMgAl10O17:Eu2+

Abstract

The promising optical and photophysical behavior of inorganic phosphors need to be optimized by highly accurate and trustworthy predictions. To control the wavelength converting mechanism of phosphors, the host lattice materials and crystalline morphology should be thoroughly investigated. In this work, we established a relationship between photoluminescence (PL) and crystalline properties by coupling the machine learning (ML) model with our experimental dataset from the synthesized Ba0.9-xSrxMgAl10O17:Eu0.1 (x = 0–0.9). The phosphors (Ba0.9-xSrxMgAl10O17:Eu0.1) were synthesized by the combustion solution method and extensively characterized by morphological, compositional, crystalline, and optical properties. By varying the Ba/Sr ratio, the experimental dataset was coined from the XRD parameters (FWHM, d(Å), crystallite size) and PL wavelengths, which were used to train the ML algorithms. After validating the trained models, we selected five high performing ML models, which blended to build the high performing ensemble Voting Regressor (VR) model with R2 = 0.9987 and RMSE = 0.1327. The best VR model accorded a good relation between crystalline and PL properties of phosphor by controlling the Ba/Sr ratio. These findings suggest that ML prediction models can be excellent preliminary investigations to guide the research of novel inorganic phosphors.

Keywords

• Artificial intelligence
• Machine learning
• Phosphor
• Crystalline property
• Photoluminescence
• Wavelength.