Predicting Adiabatic Temperature Changes in Electrocaloric Ceramics for Cooling Electronic Devices Through Machine Learning

Materials with the Electrocaloric effect constitute a promising candidate for new solid-state refrigeration techniques. This effect refers to the adiabatic temperature change (ΔT), which can play an important role in cooling electronic devices. Typically, to estimate ΔT of new materials without characterization, Hamiltonian approaches or Landau theory can be employed. However, these methods are computationally intensive and require many parameters, prompting the use of artificial intelligence techniques as an alternative approach. In this regard, we employ a data-driven machine learning approach to predict the adiabatic temperature change (ΔT) in BaTiO3-based ceramics. The data, extracted from a survey of published electrocaloric measurements in previous studies, represents the evolution of ΔT as a function of chemical composition, temperature, applied electric field, and other features. Several models have been investigated to predict ΔT, and computational results suggest that the Decision Trees model is slightly better than the Support Vector Machines model and significantly outperforms the Ridge Regression model.