Luminescence thermometry driven by a support vector machine: a strategy toward precise thermal sensing

Area of Science:

• Materials Science
• Spectroscopy
• Machine Learning

Background:

• Luminescence thermometry offers non-contact temperature sensing but faces precision and reliability challenges.
• Existing methods like luminescence intensity ratio (LIR) and multiple linear regression (MLR) have limitations in accuracy and robustness.

Purpose of the Study:

• To develop an advanced thermal sensing strategy for luminescence thermometry using machine learning.
• To compare the performance of a support vector machine (SVM) against LIR and MLR methods for temperature measurement.

Main Methods:

• Utilized Gadolinium Gallium Garnet doped with Erbium and Ytterbium (Gd3Ga5O12:Er3+-Yb3+) as the luminescent sensing material.
• Employed a support vector machine (SVM) to correlate upconversion emission spectra with temperature.
• Compared SVM performance with LIR and MLR methods across a broad temperature range (303-853 K).

Main Results:

• The SVM method achieved significantly lower maximum (0.38 K) and mean (0.12 K) errors compared to LIR (3.75 K, 1.37 K) and MLR (1.82 K, 0.43 K).
• SVM-based thermometry demonstrated high robustness against spectral distortions caused by environmental interferences, where LIR and MLR proved ineffective.

Conclusions:

• Support vector machine (SVM) is a powerful tool for advancing luminescence thermometry.
• This machine learning approach enables highly precise, reliable, and robust non-contact temperature measurements.