Modeling Self‐Rectifying Bipolar Resistive Switching Memory Through Machine Learning

ABSTRACT Resistive switching memory has emerged as a promising candidate for next‐generation data storage and is regarded as a key enabler for the advancement of artificial intelligence technologies. However, conventional resistive memory architectures face critical challenges, including sneak‐path current interference, poor scalability, and limited integration density, which hinder their practical deployment in high‐density neuromorphic systems. In this study, a self‐rectifying 3D 2 × 2 crossbar array based on a diode–memristor–diode (DMD) architecture using a Pt/SrMnO 3 /Pt configuration is demonstrated for the first time, exhibiting excellent electrical performance and outstanding operational stability. The device effectively suppresses sneak‐path currents, thereby enabling the integration of high‐density memory arrays. The reset process is quantitatively analyzed through machine learning applied to electrical transport characteristics, while the high‐resistance state is accurately described using the Poisson–Boltzmann formalism. These results highlight the significant potential of the stacked Pt/SrMnO 3 /Pt structure for next‐generation non‐volatile memory and provide a solid foundation for its application in neuromorphic computing and artificial neural networks.