Abstracts

The physical quantity that measures the thermal energy of a body is one of the definitions of temperature, which is the most fundamental thermodynamic state variable. Its fluctuations are central to myriad natural and man-made operations, such as life-sustaining cellular processes and networked devices. In these domains, the traditional contact temperature reading becomes inefficient, as the accuracy of a thermometer is limited by its size. Consequently, remote temperature readings have emerged to meet this need, where luminescence thermometry excels in the pursuit of highly sensitive and accurate thermometers. This field depends on temperature-induced changes in the spectroscopic properties of an ensemble of probes. However, thermal sensitivity may vary substantially depending on the chosen spectroscopic property. Therefore, this study aims to unify the traditional ratiometric approach and the thermal dependence of the emission lifetime to harvest higher sensitivities and lower uncertainties through dimensionality reduction. Accordingly, the principal component analysis (PCA) was employed, a technique often used for dimensionality reduction in machine learning. As a proof-of-concept, the SrY₂O₄:Tb^III/IV(2at.%),Eu^III(5at.%) phosphor was selected, synthesized by the Pechini modified method at 1100 °C/5 h under a partial CO atmosphere. Notably, the Tb^III/IV was used as a co-dopant to distort the host lattice and amplify the intensity of Eu^III emission bands because Tb^IV is spectroscopically inert in the visible region. The XPS analysis confirmed our hypothesis, revealing a 35/75% Tb^III/Tb^IV ratio, being this small amount not detectable in the emission spectrum. To further avoid any Tb^III emission, the sample was excited in the ⁵D₂←⁷F₀ (464 nm) Eu^III 4f transition. Examining the emission spectrum of the phosphor, the characteristic ⁵D₀→⁷F_0-4 emission bands were detected, in addition to the ⁵D₁→⁷F₂ Eu^III band (541 nm). The 5D0/5D1 pair's thermally coupled nature enabled the development of a ratiometric temperature probe, yielding a maximum relative sensitivity (S_r) of 0.95% K^-1 in the 380-470 K range. By switching to the ⁵D₀ lifetime using the ⁵D₀→⁷F₂ (612 nm) emission, and ⁵D₂←⁷F₀ (464 nm) excitation, maximum values of 0.42% K^-1 for S_r were obtained in the same interval. Both values involve individual thermometric approaches, which raises the question of whether joining them would imply an improvement in sensitivity. In this sense, dual-parametric S_r was accomplished by PCA. The overall S_r values reached the 2.5% K^-1 limit in the 350-460 K temperature range, representing a nearly 2.5-fold gain. Furthermore, a temperature uncertainty of 0.006 K was fostered when the dual-parametric approach was employed. Such low uncertainty allows accurate readouts at high temperatures. Hence, these outcomes demonstrate that enhanced sensitivities can be achieved through data analysis rather than solely relying on materials design.