A caloritronics-based Mott neuristor (1903.01062v1)

Abstract: Machine learning imitates the basic features of biological neural networks to efficiently perform tasks such as pattern recognition. This has been mostly achieved at a software level, and a strong effort is currently being made to mimic neurons and synapses with hardware components, an approach known as neuromorphic computing. CMOS-based circuits have been used for this purpose, but they are non-scalable, limiting the device density and motivating the search for neuromorphic materials. While recent advances in resistive switching have provided a path to emulate synapses at the 10 nm scale, a scalable neuron analogue is yet to be found. Here, we show how heat transfer can be utilized to mimic neuron functionalities in Mott nanodevices. We use the Joule heating created by current spikes to trigger the insulator-to-metal transition in a biased VO2 nanogap. We show that thermal dynamics allow the implementation of the basic neuron functionalities: activity, leaky integrate-and-fire, volatility and rate coding. By using local temperature as the internal variable, we avoid the need of external capacitors, which reduces neuristor size by several orders of magnitude. This approach could enable neuromorphic hardware to take full advantage of the rapid advances in memristive synapses, allowing for much denser and complex neural networks. More generally, we show that heat dissipation is not always an undesirable effect: it can perform computing tasks if properly engineered.