Neuron- and Axon-like behavior in metal-to-insulator transition material.

Neuromorphic computing aims to mimic the human brain and its nervous system to overcome the limitations of traditional computing architectures. Materials exhibiting resistive switching lend well to the development of biomimetic artificial neural networks. Their transitions between insulating and metallic states enable the emulation of key neuronal functionalities such as neurons, synapses or axons.

La_0.7Sr_0.3MnO₃ (LSMO) is a strongly correlated oxide that shows a temperature-driven metal-to-insulator transition (MIT). This transition can be triggered electrically allowing fast, reversible and repeatable switching. The ability to trigger the LSMO MIT electrically allows us to observe self-sustained oscillations under a constant bias similarly to what has been reported for insulator-to-metal transition compounds commonly used to mimic neuron-like spiking. Furthermore, axon-like behavior (i.e., signal amplification) is also achievable. Indeed, biasing the system to the onset of negative differential resistance while applying a small sinusoidal perturbation allows us to observe amplification without the use of an amplifier or complex electronic circuitry.

These findings demonstrate that LSMO is a promising candidate for implementation of neuronal functionalities, paving the way for integration in the next generation of artificial neural networks.