Quick and accurate optical thermometry based on chemometrics model strategy in Na0.5Gd0.5TiO3: Er3+

Abstract

Current optical temperature measurements are confined to materials with high sensitivity, while probes with fluorescence intensity ratio (FIR) sensitivity above 12% K−1 still confront severe challenges. Therefore, we propose a novel strategy to obtain temperatures in one step from the temperature-dependent emission spectra by establishing a chemometrics model instead of calculating FIR values to improve temperature measurement accuracy of materials with a low FIR sensitivity. Here, Er3+-doped Na0.5Gd0.5TiO3 with low sensitivity and the monotonic variation of FIR is used as a temperature sensing material. The temperature-dependent spectra are preprocessed by the moving average filter method, and the useful information is extracted from the preprocessed spectra by the synergy interval partial least squares (siPLS) algorithm. The extracted spectra are then combined with the partial least squares regression (PLSR), the principal component regression (PCR), the multiple linear regression (MLR), and the support vector machine regression (SVM) respectively to build chemometrics models. The errors between the predicted and the actual temperatures of different models are evaluated to verify the feasibility of the proposed strategy and to select the optimal temperature measurement model for Na0.5Gd0.5TiO3: Er3+. The results show small errors and high correlation coefficients over 0.98 between the predicted and the actual temperatures of all models, indicating the applicability of the chemometrics model strategy in optical thermometry. Finally, the best model (a combination of siPLS-PCR + SVM) and the periodic testing of FIR with temperature may make Na0.5Gd0.5TiO3: Er3+ a candidate for optical thermometer with rapidity, durability and accuracy.