Artificial neural networks show enhanced performance for key applications such as data mining or pattern recognition, but need to be implemented in hardware to make these applications accessible to everyone. Memristors are the electronic equivalent of synapses, whose variable connecting strength is at the heart of the learning process.
Accurate modelling of memristor dynamics is essential for the development of autonomous learning in artificial neural networks. In this paper, we demonstrate that spike-timing-dependent plasticity can be harnessed from inhomogeneous polarization switching in ferroelectric memristors. Combining time-dependent transport measurements, ferroelectric domain imaging, and effective-Hamiltonian-based atomistic molecular dynamics simulations, we show that the ferroelectric switching underlying resistance changes in these devices can be described by a nucleation-limited model. Using this physical model, we can reliably predict the conductance evolution of ferroelectric synapses with varying neural inputs. These results pave the way toward low-power hardware implementations of billions of reliable and predictable artificial synapses in future brain-inspired computers.
Learning through ferroelectric domain dynamics in solid-state synapses
S. Boyn et al ; Nature Comm. 8, 14736 (2017)