Emerging higher-order memristors for bio-realistic neuromorphic computing: A review

• Published in: Materials today (Kidlington, England) Vol. 68; pp. 356 - 376
• Main Authors: Chaurasiya, Rajneesh, Shih, Li-Chung, Chen, Kuan-Ting, Chen, Jen-Sue
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 01.09.2023
• Subjects: Bio-realistic neuromorphic computing; Complex dynamics; Higher-order memristors; Neuronal and synaptic properties; State variables; Neuronal and synaptic properties; Bio-realistic neuromorphic computing; Complex dynamics; Higher-order memristors; State variables
• ISSN: 1369-7021; 1873-4103
• DOI: 10.1016/j.mattod.2023.08.002

[Display omitted] Memristor devices offer an alternative route to transistor scaling and the exponential growth of computational resources. Numerous memristors have the potential to be used in neuromorphic computing or to complement the von Neumann architecture-based digital computation. However, most of the available reports on memristor are based on the first-order devices, which are unable to mimic all the biological properties. In contrast, recent studies have exposed that the conduction mechanism of the memristors is influenced by the multiple internal state variables, making them higher-order memristors, that exhibit complex dynamical behavior for encoding temporal information. In this review paper, we discuss the different types of higher-order memristors, including their mechanism and applications. We also highlight the material properties that enable the higher-order complexity in these devices. The presence of multiple internal state variables in higher-order memristors enable the dynamic complexity and adaptive properties, mimicking the neuronal and synaptic properties, similar to biological system. These higher-order memristors show the potential for processing temporal information and performing analogue computing. Further efforts are needed to optimize higher-order memristor devices and develop new architectures for future bio-realistic neuromorphic computing.