An artificial synaptic transistor using an α-In 2 Se 3 van der Waals ferroelectric channel for pattern recognition

Despite being widely investigated for their memristive behavior, ferroelectrics are barely studied as channel materials in field-effect transistor (FET) configurations. In this work, we use multilayer α-In₂Se₃ to realize a ferroelectric channel semiconductor FET, i.e., FeS-FET, whose gate-triggered and polarization-induced resistive switching is then exploited to mimic an artificial synapse. The FeS-FET exhibits key signatures of a synapse such as excitatory and inhibitory postsynaptic current, potentiation/depression, and paired pulsed facilitation. Multiple stable conductance states obtained by tuning the device are then used as synaptic weights to demonstrate pattern recognition by invoking a hidden layer perceptron model. Detailed artificial neural network (ANN) simulations are performed on binary scale MNIST data digits, invoking 784 input (28 × 28 pixels) and 10 output neurons which are used in the training of 42000 MNIST data digits. By updating the synaptic weights with conductance weight values on 18000 digits, we achieved a successful recognition rate of 93% on the testing data. Introduction of 0.10 variance of noise pixels results in an accuracy of more than 70% showing the strong fault-tolerant nature of the conductance states. These synaptic functionalities, learning rules, and device to system-level simulation results based on α-In₂Se₃ could facilitate the development of more complex neuromorphic hardware systems based on FeS-FETs.