In the brain learning is achieved through the ability of synapses to reconfigure the strength by which they connect neurons (synaptic plasticity). In promising solid-state synapses called memristors, conductance can be finely tuned by voltage pulses and set to evolve according to a biological learning rule called spike-timing-dependent plasticity (STDP). Future neuromorphic architectures will comprise billions of such nanosynapses, which require a clear understanding of the physical mechanisms responsible for plasticity. Here we report on synapses based on ferroelectric tunnel junctions comprising ultrathin films of super-tetragonal BiFeO3. The information is encoded through the non-volatile ferroelectric polarization that can be switched with an electric field. Switching the ferroelectric polarization induces large variations of the tunneling current, enabling a simple non-destructive readout of the polarization state. We show that spike-timing-dependent plasticity can be harnessed from intrinsically inhomogeneous ferroelectric polarization switching [1]. Through combined scanning probe imaging, electrical transport experiments, and atomic-scale dynamic simulations, we demonstrate that conductance variations can be accurately controlled and modeled by the nucleation-dominated switching of domains with opposite polarizations [2]. Our results show that ferroelectric nanosynapses learn in a reliable and predictable way, opening the path towards unsupervised learning in spiking neural networks.
References
[1] B. Xu, V. Garcia, S. Fusil, M. Bibes, L. Bellaiche, Phys. Rev. B 95, 104104 (2017)
[2] S. Boyn, J. Grollier, G. Lecerf, B. Xu, N. Locatelli, S. Fusil, S. Girod, C. Carrétéro, K. Garcia, S. Xavier, J. Tomas, L. Bellaiche, M. Bibes, A. Barthélémy, S. Saïghi, V. Garcia, Nature Comm. (2017) DOI: 10.1038/ncomms14736