Anisotropy and synaptic plasticity in CrSBr/WSe 2 heterojunction for advanced neural network applications

Authors:

Abstract
Employing two‐dimensional (2D) synaptic devices to develop a brain‐inspired neuromorphic computing system is a promising approach to overcoming the limitations of the von Neumann architecture. However, isotropic 2D materials are predominantly utilized to fabricate synaptic devices. Research on inherently anisotropic 2D materials in synaptic devices remains scarce. Here, we report an intrinsically anisotropic material, CrSBr, which exhibits optoelectronic properties with significant angular dependence, achieving a carrier mobility ratio as high as 7.83 between the a ‐axis and b ‐axis. Based on this, we couple the in‐plane anisotropy into the synaptic device and construct CrSBr/WSe 2 multi‐terminal device. This device can be regulated by the gate voltage and laser, exhibiting storage and synaptic behaviors dependent on the a and b axes. Furthermore, we apply the synaptic property to achieve image recognition. Due to the anisotropic response to identical external stimulus, the a ‐axis conductance trend transits from nonlinear to approximately linear within the multi‐terminal conductance framework. This multi‐terminal synapse model achieves a recognition rate of up to 91% on the Fashion‐MNIST database, significantly outperforming single‐terminal recognition performance. Our work introduces a novel approach to anisotropic artificial synapses for simulated image recognition and establishes a foundation for developing AI systems with enhanced recognition rates.
Funding Information
  • National Science Fund for Distinguished Young Scholars (JQ23007)
  • National Natural Science Foundation of China (52450014)
  • National Natural Science Foundation of China (62174013)
  • National Natural Science Foundation of China (92265111)
  • National Science Fund for Distinguished Young Scholars (JQ23007)
  • Key R&D Program of China (2022YFA1203901)
  • National Natural Science Foundation of China (52450014)
  • National Natural Science Foundation of China (62174013)
  • National Natural Science Foundation of China (92265111)
  • Beijing Natural Science Foundation (JQ23007)
  • Beijing Natural Science Foundation (L233003)

This publication has 43 references indexed in Scilit: