Inspired by biological neural and sensory systems, the in-memory computing and in-sensor computing paradigms have emerged, which integrate computation with memory and processing with sensor respectively, offering a promising solution to address latency and power bottlenecks of traditional von Neumann architectures. Neuromorphic devices such as artificial synapse and neuromorphic sensors are the core components of these innovative computing paradigms. Among various types of neuromorphic devices, ferroelectric devices can not only emulate synaptic weight updates via electric-field-induced dynamic modulation of conductance, but also sense and process multimodal physical stimuli, such as light, mechanical force, and heat. Such multifunctionality enables the integration of sensing, memory, and computation at the device level, positioning ferroelectric devices as an ideal platform for neuromorphic computing and sensing systems. Herein, we present a systematic review of ferroelectric neuromorphic devices for in-memory computing and in-sensor computing. The applications of ferroelectric devices as artificial synapses in in-memory computing are summarized. Furthermore, we examine the applications of ferroelectric devices as sensing elements in in-sensor computing systems and summarize the latest research advances in these fields. Finally, this review outlines the key challenges faced by ferroelectric neuromorphic devices and proposes future development directions to promote their practical applications.
Fang et al. (Thu,) studied this question.