Programmable ferroelectric rectifier for reliable and efficient neuromorphic crossbar array.

Kim, Youngmin; Lee, Yoon Jung; Yang, Jiwoong; Kim, Byungsoo; Kim, Seung Ju; Kim, Jaehyun; Im, Inhyuk; Kim, Jae Young et al. · Nat Commun · 2026

basic_science · Level V

Where this comes from

Abstract

The advancement of neuromorphic computing hardware requires energy-efficient operation, scalable device integration, and reliable conduction. These challenges can be effectively addressed by employing functional ferroelectric material as an active layer in memristors, leveraging their electrostatically modulated conduction for reliable switching. In this study, we present memristor devices that achieve a rectifying ratio exceeding 10⁶ and an off-state current below 10⁻¹² A based on epitaxial heterostructures consisting of Pt/Ba<sub>0.2</sub>Bi<sub>0.8</sub>FeO<sub>3</sub> (BBFO)/SrRuO<sub>3</sub>/SrTiO<sub>3</sub> stacks. Substitution of 20% Ba in BiFeO₃ induces a coupled interaction between ferroelectric polarization and oxygen vacancy migration, which under pulsed bias governs vacancy transport and ensures reliable memristive synaptic behavior with near-zero nonlinearity and endurance beyond 10⁷ cycles. Owing to the demonstrated linear synaptic performance and strong memristive rectification, a selector-free crossbar array (CBA) was implemented. By mitigating key CBA challenges such as sneak currents and cell-to-cell variability while maintaining high synaptic performance, BBFO provides a robust material platform for CBA-based neuromorphic systems.