Machine Learning Based Distinguishing between Ferroelectric and Non‐Ferroelectric Polarization–Electric Field Hysteresis Loops

The polarization–electric field (P–E) hysteresis loop is one of the most important criteria for identifying ferroelectricity. However, a P–E loop with apparent hysteresis window can be generated from non‐ferroelectric sources such as leakage current. So far distinguishing between ferroelectric and non‐ferroelectric loops is still performed in a manual way, which can be error prone and time consuming, particularly when the loops are not easily distinguishable and the number of loops to be identified is large. Here, two machine learning (ML) approaches are developed, one using the polarization values along the P–E loops as the input dataset (termed as “value‐based” approach) and the other using the loop images as the input dataset (termed as “image‐based” approach), to identify the P–E loops as ferroelectric or non‐ferroelectric. The value‐ and image‐based ML approaches achieve identification accuracies as high as 93.08% and 87.42%, respectively. In addition, it is tested that both approaches complete an identification of about 160 loops in very short time (≈1.0 s). The high accuracy and efficiency therefore demonstrate that the ML approaches significantly outperform the manual way for distinguishing ferroelectric from non‐ferroelectric P–E loops, which may greatly facilitate the research on ferroelectrics. Value‐ and image‐based machine learning (ML) approaches are developed to identify whether polarization–electric field (P–E) hysteresis loops are ferroelectric or non‐ferroelectric, and both approaches achieve high accuracy and efficiency. The ML approaches therefore significantly outperform the manual method for distinguishing ferroelectric from non‐ferroelectric P–E loops, which may greatly facilitate research on ferroelectrics.