Fast Bienenstock–Cooper–Munro rule with tunable threshold realized in ferroelectric tunnel junction for neuromorphic computing

Abstract

Ferroelectric tunnel junctions remain great prospects for neuromorphic computing due to their significant nondestructive readout and ultra-low-energy consumption. In this work, versatile synaptic functions including paired-pulse facilitation, paired-pulse depression, spike-rate-dependent plasticity, and Bienenstock–Cooper–Munro learning rules have been simulated by electrical stimulation in ferroelectric tunnel junctions of the Au/La0.1Bi0.9FeO3/Nb:SrTiO3/In system. A potentiation at high frequency and a depression at low frequency are found under positive bias, while an opposite behavior is observed at negative bias. Furthermore, a fast Bienenstock–Cooper–Munro learning rule in nanosecond order with the sliding threshold rate and enhanced depression effect is demonstrated by only applying a series of presynaptic pulse trains. These findings provide a method for simulating fast Bienenstock–Cooper–Munro learning rules based on ferroelectric tunnel junction.