Luminescence Thermometry with Eu-Doped YMoO: Comparison of Performance of Intensity Ratio and Machine Learning Temperature Read-Outs.

Accurate temperature measurement is critical across various scientific and industrial applications, necessitating advancements in thermometry techniques. This study explores luminescence thermometry, specifically utilizing machine learning methodologies to enhance temperature sensitivity and accuracy. We investigate the performance of principal component analysis (PCA) on the Eu-doped YMoO luminescent probe, contrasting it with the traditional luminescence intensity ratio (LIR) method. By employing PCA to analyze the full emission spectra collected at varying temperatures, we achieve an average accuracy (ΔT) of 0.9 K and a resolution (δT) of 1.0 K, significantly outperforming the LIR method, which yielded an average accuracy of 2.3 K and a resolution of 2.9 K. Our findings demonstrate that while the LIR method offers a maximum sensitivity (Sr) of 5‱ K⁻ at 472 K, PCA's systematic approach enhances the reliability of temperature measurements, marking a crucial advancement in luminescence thermometry. This innovative approach not only enriches the dataset analysis but also sets a new standard for temperature measurement precision.