Transition-Metal Oxides under Strong Electric Fields, from Resistive Switching to Artificial Synapses and Neurons

The information age we live in is supported on a physical under-layer of electronic hardware, which originates in condensed matter physics research. The mighty progress made in silicon based technology seemed endless. However, with the smallest feature size of transistors reaching down to a mere 10 nm, this technology is finding an unavoidable physical limit. This calls for exploration of new alternatives. Neuromorphic inspired systems are making fast progress. But this is based either on dedicated hardware made with conventional electronics, or in software, such as Deep Neural Networks, running on conventional computers. Resistive switching phenomena opens the way to explore a radical solution, namely, the implementation of simple devices with the required bio-inspired functionalities to directly build neuromorphic systems. In this talk we shall describe recent efforts towards this goal using transition metal oxides, including Mott strongly correlated systems. We shall describe two types of resistive switching, volatile and non-volatile. The latter is now much better understood and is surprisingly ubiquitous among transition metal oxides. This functionality may serve to implement analog synapses. On the other hand, the former, is observed in strongly correlated Mott systems and rather remarkably may realize the electric function of a "leaky-integrate-and-fire" neuron. We shall describe our recent modeling efforts along these lines.

References

E.Janod et al. Resistive switching in Mott insulators and correlated systems Adv. Func. Mat 25 (2016) 6287 - a review on volatile Mott systems.

M. Rozenberg, Resistive switching, Scholarpedia 6 :11414 (2011) - a brief introduction to volatile systems.