Abstract

In this work, artificial neural network (ANN) modeling was used to model ferroelectric hysteresis under the influence of compressive uniaxial stress using the hysteresis data obtained from soft lead zirconate titanate as an application. The main objective is to model the role of external stress, including electric field perturbation, on the complex hysteresis properties, which are hysteresis area, remnant polarization, coercivity and loop squareness. With its false tolerance abilities, ANN was used to predict how the stress direction (on applying and releasing), the stress magnitude (s) the electric field amplitude (E₀), and the electric frequency (f) affect on the hysteresis properties, quantitatively. The best network architecture with highest accuracy was found in the ANN training through extensive architecture search. It was then used to predict hysteresis properties of the unseen testing patterns of input. The predicted and the actual testing data were found to match very well for the whole extensive range of considered input parameters. This well match, even when the stress was applied, certifies the ANN one of the superior techniques, which can be used for the benefit of technological development of ferroelectric applications.

Key words: Artificial neural network, hysteresis properties, soft lead zirconate titanate, uniaxial stress.

Abbreviation

ANN, Artificial neural network; PZT, lead zirconate titanate;  stress magnitude; E_0, the electric field amplitude; f, frequency; MLP, multilayer perceptron; BP,back propagation; A, loop area; P_r,, remnant polarization; E_c, hysteresis coercivity; S,hysteresis loop squareness; MAE, mean absolute error.