Magnetic Stochastic Oscillators: Noise-Induced Synchronization to Underthreshold Excitation and Comprehensive Compact Model

Abstract

In magnetic tunnel junctions specifically designed to be super-paramagnetic (S-MTJs), thermal noise induces stochastic switching between the parallel and anti-parallel states (Figure 1) [1]. As a consequence, they behave like stochastic oscillators, oscillating with non-constant period but with a well-defined average frequency. Although such junctions cannot be used for conventional memory applications, when exposed to spin transfer torque, they exhibit complex behavior that could be harnessed to implement original bio-inspired computing schemes. Notably, spin transfer torque can be used to influence the stability of states and therefore control the natural frequency of the stochastic oscillator. In this work, we investigate how this can lead to the phenomena of “stochastic resonance” and noise-induced synchronization of the stochastic oscillations to weak periodic excitations [2,3]. In order to design the bio-inspired systems exploiting such effects, accurate device models are needed. We therefore introduce a compact model written in the Verilog-A language and that can be used within standard integrated circuit design tools (Cadence platform) for system design, and that reproduces our experimental results.