Optically-modulated and mechanically-flexible MXene artificial synapses with visible-to-near IR broadband-responsiveness

Abstract

The inherent limitations of von Neumann computing associated with its inefficient parallel-processing of massive data become increasingly pronounced in state-of-the-art digital device technologies. Artificial synapses of the human brain-inspired neuromorphic computing are emerging as a viable solution, which demands to explore unconventional materials responsive to a variety of electrical and/or optical stimuli. Herein, we report that solution-processed titanium carbide MXene (Ti₃C₂T_x) exhibits essential characteristics for optoelectronic synapses-based neuromorphic computing. Specifically, it presents optically-triggered synaptic plasticity with memory effects in a broad spectral range, as well as accompanying a large degree of mechanical deformability. By leveraging the optoelectronics-mechanics coupling, we demonstrate that MXene-based devices can simulate vital functionalities demanded in artificial neural networks (ANNs) such as associative learning behaviors and high-accuracy pattern recognition. Furthermore, the operational principle of the MXene optoelectronic synapses is unveiled in the context of the charge trapping/de-trapping mechanism enabled by its processing-introduced bandgap opening.