Correlation Between Polymorph Phases and Memristive Behavior in Epitaxial Hafnia-Based Devices

Memristors are considered vital components for the development of neuromorphic computing hardware, which aims to outperform traditional computers (based on the Von Neumann architecture) in data-intensive tasks such as machine learning. Memristive behavior typically involves the electromigration of charged defects, such as oxygen vacancies (OV) in oxides, which primarily affect the device’s two-point resistance. In ferroelectric oxides, the polarization direction can control the tunneling current in ultrathin layers or modify the Schottky barrier height in thicker films, thereby regulating interface resistance. These mechanisms, being electronic in nature, are expected to be faster and more reliable than those dependent on defect migration.

Hafnium oxide (HfO₂, or hafnia) has recently gained significant attention as a lead-free ferroelectric material. Hafnia-based ferroelectrics offer technological advantages over traditional perovskites, including CMOS compatibility, enhanced ferroelectric properties at ultra-thin thicknesses (<10 nm), and the ability to be fabricated using scalable techniques such as atomic layer deposition and chemical solution deposition.

In this study, we fabricated by pulsed laser deposition and characterized the memristive properties of SrTiO₃/La_2/3Sr_1/3MnO₃/Hf_0.5Zr_0.5O₂ heterostructures with Pt top electrodes. By controlling the fabrication conditions, we were able to adjust the balance between polar and non-polar polymorphs, as confirmed through x-ray diffraction and high-resolution transmission electron microscopy. The latter also provided local insights into the chemistry and nanostructure of the layers. Our findings show that films containing only the polar phase, with negligible non-polar monoclinic aggregates, exhibit the most reliable memristive response. This work contributes to advancing the development of hafnia-based memristors for in-memory and neuromorphic computing applications.