In-Memory and In-Sensor Neuromorphic Computing with 2D Ferroelectrics
The von Neumann architecture is increasingly constrained by energy consumption and data-transfer efficiency as artificial intelligence and data-intensive applications expand. Neuromorphic computing, inspired by the human brain's information-processing mechanisms, offers an alternative paradigm. Two-dimensional (2D) ferroelectric materials are promising candidates due to their intrinsic non-volatility, atomic-scale thickness, ultra-low power consumption, excellent fatigue endurance, and dangling-bond-free surfaces. This review examines recent advances in 2D ferroelectric materials and associated device architectures for neuromorphic applications. It first introduces ferroelectric mechanisms and representative 2D ferroelectrics, then surveys key device architectures including ferroelectric tunnel junctions, diodes, transistors, and photovoltaic devices. Their applications in in-memory computing and in-sensor neuromorphic systems are discussed, with emphasis on artificial neural networks, spiking neural networks, reservoir computing, and neuromorphic perception for efficient information processing and intelligent sensing. The unique properties of 2D ferroelectrics enable integrated sensing, memory, and computing functionalities, demonstrating potential for future neuromorphic and brain-inspired intelligent systems.