Artificial synaptic device and neural network based on the FeGa/PMN-PT/FeGa memtranstor

Abstract

The memtranstors employing the magnetoelectric effects have the great potential to develop artificial synaptic devices. We have fabricated a memtranstor made of the FeGa/PMN-PT/FeGa multiferroic heterostructure and used it to mimic the functions of synapses. The magnetoelectric voltage of the device can be continuously tuned by applying a train of electric-field pulses. Consequently, synaptic plasticity, including the long-term potentiation, long-term depression, and spiking-time-dependent plasticity, has been demonstrated in the memtranstor at room temperature. Simulations on a neural network made of an array of the memtranstors reveal the capability of pattern learning with a high accuracy.