All Spin Logic device as a compact artificial neuron

Abstract

In our recent work we have proposed the possibility of using magnets interacting via spin currents to implement all-spin logic (ASL) devices for information processing, fundamentally different from the standard charge-based architecture. Information is stored in the state of magnetization of the magnets and is communicated between magnets through pure spin currents. Simulations of these multi-magnet networks with our experimentally benchmarked model [2] indicate that such spin-magnet circuits can mimic many of the attributes of standard charge-based circuits such as inverters, universal NAND/NOR gates, ring oscillators etc .. However, these circuits do not take full advantage of the natural hybrid analog/digital character of spin currents and magnets, the possibility of which we discuss in this paper. Indeed, our contribution in this paper is twofold: We show that a simple model of an ASL device, maps on to the well-known 'leaky integrate and fire' equation of neuron dynamics, highlighting the analogy of the magnet to a neuron. We introduce LLGz for the first time in this paper and relate it to our full rigorous model. We also show that LLGz captures the steady state behavior of magnets predicted by the full model.