An artificial synaptic transistor using an α-In₂Se₃ van der Waals ferroelectric channel for pattern recognition

Area of Science:

• Materials Science
• Neuroscience
• Computer Science

Background:

• Ferroelectric materials are primarily studied for memristive behavior, with limited exploration as channel materials in field-effect transistors (FETs).
• Artificial synapses are crucial components for developing advanced neuromorphic computing hardware.

Purpose of the Study:

• To investigate multilayer α-In2Se3 as a channel material for ferroelectric FETs (FeS-FETs).
• To demonstrate the synaptic functionalities of FeS-FETs and their application in pattern recognition using artificial neural networks (ANNs).

Main Methods:

• Fabrication of FeS-FETs using multilayer α-In2Se3.
• Characterization of gate-triggered and polarization-induced resistive switching for synaptic emulation.
• Implementation of a hidden layer perceptron model for pattern recognition on MNIST data using ANN simulations.

Main Results:

• FeS-FETs exhibited key synaptic signatures including excitatory/inhibitory postsynaptic current, potentiation/depression, and paired pulsed facilitation.
• Multiple stable conductance states were achieved and utilized as synaptic weights.
• A 93% recognition rate was achieved on MNIST data, with over 70% accuracy even with 0.10 variance of noise pixels, highlighting fault tolerance.

Conclusions:

• Multilayer α-In2Se3 is a promising material for FeS-FETs, enabling artificial synaptic functionalities.
• FeS-FETs demonstrate potential for building complex neuromorphic hardware systems.
• The study provides device-to-system level simulation results facilitating future neuromorphic engineering.