Silver-carbon nanotube heterostructures for photothermoelectric artificial synaptic devices

Abstract

In this thesis, a silver-carbon nanotube (Ag-CNT) heterostructure artificial synaptic device based on the photothermoelectric effect is proposed, aiming to break through the limitations of traditional von Neumann architectures in real-time computation and energy efficiency, and to mimic the dynamic plasticity echanism of biological synapses. By optimising the light-absorbing properties of carbon nanotubes and the plasma-enhancing effect of silver, combined with the asymmetric photothermal gradient-driven Seebeck current, the device achieves an efficient conversion of optical to electrical signals. Experiments show that the device can accurately simulate the short-time potentiation (STP) and long-time potentiation (LTP) properties of biological synapses under the modulation of light intensity, light duration and pulse to-pulse. Further studies have verified the potential of the device in spatiotemporal signal processing: temporal signal differentiation is achieved by 4-bit binary optical coding, and the "NTUEEE" light pattern is successfully recognised using a 4×5 synaptic array, which demonstrates its potential for applications in time-series recognition, light-trajectory tracking, light-pattern recognition, and neuromorphic computation. This study provides new ideas for the design of low-power light controlled synaptic devices and ultimodal signal processing and promotes the evolution of biomimetic computing to an efficient and adaptive system.