Abstract

Inspired by the human brain, neuromorphic computing uses artificial synapses and the network of neurons to realize the ability for parallel computing, progressive inference, and learning. This evolving technology can transcend the current limitations of von Neumann computing architectures. Memristors are nonlinear resistance switching devices with memory functions that can mimic brain-like synapses. The in-memory computation feasibility of memristors is the major advantage. They have potential application in AI and neural computing due to their rapid switching speed, storage density, low power consumption, superior data processing capabilities, and can be simulated on a biological scale. Owing to their excellent conductivity, hydrophilic surface, fast charge response, high stacking density, and durability, 2D MXenes are emerging as potential materials for memristors. Brain-inspired parallel computing ‘neuromorphic computing’ is one of the most promising technologies for efficiently handling large amounts of information data, which operates based on a hardware-neural network platform consisting of numerous artificial synapses and neurons. Memristors, as artificial synapses based on various 2D materials for neuromorphic and data storage technologies with low power consumption, high scalability, and high speed, have been developed to address the von Neumann bottleneck and limitations of Moore’s law. The 2D MXenes have strong potential application in memristors due to their ultrahigh conductivity, fast charge response, high stacking density, and high hydrophilicity. Here, we discuss how MXenes are emerging as a potential material towards artificial synapses. Recent progress in research on artificial synapses, fabricated particularly using MXenes and their composite materials, is comprehensively discussed with respect to mechanism, synaptic characteristics, power efficiency, and scalability. Finally, we present an outlook of the future development of MXenes for artificial intelligence and challenges in integrating memristors with MXenes are briefly discussed. Brain-inspired parallel computing ‘neuromorphic computing’ is one of the most promising technologies for efficiently handling large amounts of information data, which operates based on a hardware-neural network platform consisting of numerous artificial synapses and neurons. Memristors, as artificial synapses based on various 2D materials for neuromorphic and data storage technologies with low power consumption, high scalability, and high speed, have been developed to address the von Neumann bottleneck and limitations of Moore’s law. The 2D MXenes have strong potential application in memristors due to their ultrahigh conductivity, fast charge response, high stacking density, and high hydrophilicity. Here, we discuss how MXenes are emerging as a potential material towards artificial synapses. Recent progress in research on artificial synapses, fabricated particularly using MXenes and their composite materials, is comprehensively discussed with respect to mechanism, synaptic characteristics, power efficiency, and scalability. Finally, we present an outlook of the future development of MXenes for artificial intelligence and challenges in integrating memristors with MXenes are briefly discussed. a state of material with high resistance under the influence of external force. the weight change in synapse for long duration after applied voltage spikes. a state of material with low resistance under the influence of an external field. layered, hexagonal carbide and nitride materials that bridge the gap between properties typical of metals and ceramics. a physical phenomenon where a dielectric changes its resistance with respect to the applied electric field or current. weight change in synapse for short duration after voltage spikes. synapse weight change with respect to frequency of applied voltage spikes. synapse weight change with respect to time duration of applied voltage spikes. traditional computer architecture, consists of separate memory and processor units. limitation on throughput caused by the separate memory and processor units.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call