# Artificial Neuron Based on Electrical Anisotropy from WSe2 Field Effect Transistors

## TL;DR

Researchers created an artificial neuron using WSe2 material that mimics brain functions and achieved high accuracy in digit recognition and robot control.

## Contribution

A novel artificial neuron using WSe2's electrical anisotropy with axon-multisynaptic and dendritic functionalities is introduced.

## Key findings

- A six-terminal WSe2 field effect transistor achieved axon-multisynaptic performance with modulating plasticity.
- Dendritic functionality with optoelectronic synergy was demonstrated in the neural device.
- Gesture control of a robot manipulator was achieved using anisotropic synaptic responses from WSe2.

## Abstract

Neuromorphic systems have been recognized as potential computational platforms for overcoming the von Neumann architecture, where neurons are the basic units for neuromorphic computing. However, artificial neurons of two‐dimensional materials suffered from limited stability and complex architecture. In this work, a multi‐terminal neural device was constructed based on the electrical anisotropic properties of WSe2. Axon‐multisynaptic performance with modulating plasticity was realized successfully by a stable six‐terminal WSe2 field effect transistor. And dendritic functionality with optoelectronic synergy was obtained, showcasing the functional integrity of this neural device. Recognition accuracy of handwritten digits was obtained to 97% based on the synaptic weight of an artificial neuron. Additionally, gesture control of a robot manipulator was achieved with anisotropic synaptic responses from a WSe2 sample. These findings not only provide novel material platforms and circuit design solutions for constructing artificial neural networks but also mark a significant advancement in the development of neuromorphic electronics capable of processing complex signals.

An artificial neuron was realized based on the anisotropic mobility of WSe2. Dendritic and axon‐multisynaptic performance of a neuron was achieved with optical and voltage pulses. Recognition accuracy of handwritten digits was obtained to 97% based on the synaptic weight of an artificial neuron. And gesture control of a robot manipulator was succeeded with anisotropic synaptic responses from a WSe2 sample.