Realization of a temperature sensor using both two- and three-dimensional photonic structures through a machine learning technique

A temperature sensor based on photonic crystal structures with two- and three-dimensional geometries is proposed, and its measurement performance is estimated using a machine learning technique. The temperature characteristics of the photonic crystal structures are studied by mathematical modeling. The physics of the structure is investigated based on the effective electrical permittivity of the substrate (silicon) and column (air) materials for a signal at 1200 nm, whereas the mathematical principle of its operation is studied using the plane-wave expansion method. Moreover, the intrinsic characteristics are investigated based on the absorption and reflection losses as frequently considered for such photonic structures. The output signal (transmitted energy) passing through the structures determines the magnitude of the corresponding temperature variation. Furthermore, the numerical interpretation indicates that the output signal varies nonlinearly with temperature for both the two- and three-dimensional photonic structures. The relation between the transmitted energy and the temperature is found through polynomial-regression-based machine learning techniques. Moreover, rigorous mathematical computations indicate that a second-order polynomial regression could be an appropriate candidate to establish this relation. Polynomial regression is implemented using the Numpy and Scikit-learn library on the Google Colab platform.