A Single Volatile VO 2 Memristor Enabling Neuromorphic Perception, Learning, and Computing
Bingkai Ding, Bohao Ding, Omesh Kapur, Yisong Han, Richard Beanland, Kai Sun, C. H. (Kees) de Groot, Ruomeng HuangABSTRACT
Neuromorphic computing requires compact hardware devices capable of supporting multiple brain‐inspired functions. Vanadium dioxide (VO 2 ), with its thermally and electrically driven insulator–metal transition (IMT), offers an attractive materials platform for such applications owing to its intrinsic volatility and fast switching dynamics. Here, we demonstrate a vertical Pt/VO 2 /Pt memristor that integrates multiple neuromorphic functions within a single volatile device platform. VO 2 thin films are deposited by atomic layer deposition (ALD) using the TEMAV precursor, yielding semiconducting insulating and metallic phases with a resistivity contrast of approximately two orders of magnitude. Electrical transport analysis indicates that back‐to‐back Schottky barriers at the Pt/VO 2 interfaces play a dominant role in device conduction and enable reproducible volatile threshold switching. Leveraging the interfacial transport and thermally driven switching dynamics, the same device emulates leakyintegrate‐and‐fire neuronal behavior, Pavlovian‐like associative conditioning, reservoir‐computing‐based temporal signal recognition, and artificial nociceptor responses. In particular, the device‐driven reservoir system is applied to electrocardiogram heartbeat classification, demonstrating its capability for processing real physiological temporal signals. These results establish ALD‐grown vertical VO 2 memristors as compact and multifunctional building blocks for neuromorphic and sensory hardware, highlighting a route toward single‐device platforms that couple perception, learning, and temporal information processing.