Dimensionality Reduction Expands the Frontiers of Lanthanide Luminescence Thermometry Beyond Single-Parametric Thermal Sensing

Abstract

Luminescent thermometers (LThs) are particularly suitable for monitoring submicron-scale thermal changes, spurring substantial advances in the field. Compared with the efforts to maximize the performance of LThs by designing novel materials, few advances have been made at the methodological level to exploit classical thermometric essays. Such a viewpoint motivated this study, which sought to combine data analysis algorithms with multiple thermometric parameters to introduce an advanced perspective on postprocessing data methodologies. Specifically, three distinct dimensionality reduction (DR) algorithms were employed: multiple linear regression (MLR), non-negative matrix factorization (NMF), and kernel principal component analysis (k-PCA). These methods were applied to the proof-of-concept SrY₂O₄:Tb^III/IV,Eu^III phosphor using the thermal dependence of the thermally coupled levels of Eu^III (Δ) and the ⁵D₀ lifetime (τ). Compared to traditional fitting and integration analyses, the DR approach provided enhanced thermometric performance, achieving a sensitivity increase from 0.897% K^–1 (using Δ) to 3.68% K^–1 (using k-PCA) and reducing temperature uncertainty below 0.03 K (k-PCA). By moving beyond single parametric thermal sensing with DR, these outcomes enable us to push the limits of luminescence thermometry toward unexplored pathways.