Interface Element Accumulation-Induced Single Ferroelectric Domain for High-Performance Neuromorphic Synapse

Abstract

Ferroelectric (FE) synapses are promising for neuromorphic computing toward enhanced artificial intelligence systems. Nonetheless, there is a significant gap in understanding how to effectively tailor self-polarization and its implications on synaptic device performance. Here, an approach using interfacial element accumulation is reported to tailor the self-polarization states of BaTiO3 (BTO)/La0.67Sr0.33MnO3 (LSMO) FE heterostructure into a single domain state. This single domain configuration results are demonstrated in a gradient distribution of oxygen vacancies across the film thickness, yielding an extraordinary on/off ratio of 107 in Pt/BTO/LSMO FE diodes. This giant resistive switching enables the long-term potentiation and long-term depression synaptic functions of excellent linearity and symmetry (with a nonsymmetry factor as low as 0.1), leading to a supervised learning ability of the associated artificial neural network with a high pattern recognition accuracy of 95%. This work provides a simple design principle for FE single domain, which is substantial in enhancing the performance of FE synapses for neuromorphic computing.