Machine learning assisted electrocaloric and pyroelectric performance of Ba_0.85Ca_0.15Ti_0.9Zr_0.1O₃ceramic for solid state refrigeration

In the present study, a thorough investigation was carried out to assess the electrocaloric performance of Ba₀.₈₅Ca₀.₁₅Zr₀.₁₀Ti₀.₉O₃(BCZT) ceramics synthesized via the mixed oxide route, incorporating machine learning tools for enhanced analysis. The Rietveld refinement of the ceramics' X-ray diffraction (XRD) patterns verifies the simultaneous existence of two phases (orthogonal and tetragonal symmetries). The micrograph obtained from the scanning electron microscope indicates a well-defined and dense grain. The temperature variant dielectric exhibits two phase transformations associated with orthorhombic symmetry to tetragonal symmetry, and tetragonal symmetry to cubic symmetry. The electric field variant ferroelectric hysteresis was measured at various temperatures, and the remnant polarization and coercive field decreased with Temperature. The temperature variant polarization values at different electric fields were modelled using various machine learning approaches. With minimal experimental input, the expected outcomes enable efficient and reliable prediction of electrocaloric behaviour. Thermodynamic Maxwell relations were employed to indirectly evaluate variations in isothermal entropy, adiabatic Temperature, and electrocaloric strength. Different figures of merit, like relative refrigerant capacity, cooling power, and entropy change aggregated over Temperature, are analyzed across varying electric fields, highlighting the material's promise for environmentally friendly cooling technologies.