Memcapacitors in neuromorphic architectures: the mechanisms, challenges, and applications

Abstract

The rapid advancement of artificial intelligence and data-intensive processing has intensified the demand for energy-efficient, brain-inspired computing frameworks. To overcome the memory bottleneck inherent in conventional von Neumann-based computing architectures, memristive devices have been explored extensively to enable the capability of parallel in-memory processing. Beyond memristive systems, this paradigm can be further extended to memcapacitive elements. Memcapacitors, a class of passive circuit devices with state-dependent capacitance, have emerged as promising candidates to enhance memory storage. This work begins with a discussion of the four fundamental physical mechanisms underlying memcapacitor operation, followed by an exploration of their diverse biomimetic functionalities and integration into physical neural networks. Furthermore, we evaluate the opportunities and challenges associated with memcapacitors at device and circuit levels, presenting future perspectives for the deployment and applications. By harnessing their unique properties, such as dynamic capacitance modulation, non-volatile memory behavior, and low-power operation, memcapacitors have the potential to revolutionize next-generation computing hardware and intelligent edge devices, paving the way for more efficient and scalable neuromorphic systems.