Neuromorphic computing with quantum materials: From Mott insulators to brain-inspired devices

Abstract

The escalating energy demands of modern artificial intelligence have exposed fundamental limitations in conventional computing architectures. Neuromorphic computing, which seeks to emulate the brain’s massively parallel and energy-efficient information processing, offers a compelling alternative, but its realization requires materials that naturally exhibit nonlinear, multi-state, and adaptive behavior characteristic of biological neurons and synapses. This review argues that strongly correlated electron materials, particularly Mott insulators, provide exactly this physical substrate. We present the essential physics of metal-insulator transitions in vanadium oxides (VO2, V2O3) and perovskite manganites in an accessible manner and describe how their switching behaviors map onto neuromorphic primitives: the neuristor (artificial neuron) and the synaptor (artificial synapse). The review draws extensively on the authors’ experimental work: direct imaging of nanotextured phase coexistence in V2O3; an MIT Fingerprinting approach to engineer multi-state resistive switching in VO2; the demonstration of subthreshold firing in Mott nanodevices; and 1 electrically controlled multi-state analog memory in phase-separated manganites. The review concludes by identifying the main challenges and the most promising directions to translate these laboratory demonstrations into practical brain-inspired computing systems.