Realizing an Isotropically Coercive Magnetic Layer for Memristive Applications by Analogy to Dry Friction

Designing hardware specifically for neural-network algorithms is crucial for low power consumption in ``big data'' analysis tasks. The challenge is to realize reliability, endurance, and scalability in the same synaptic device. Here a spintronic memristor, based on an isotropically coercive free layer in a magnetic tunnel junction, is modeled by the Landau-Lifshitz-Gilbert equation with an additional dissipative term reminiscent of dry friction. Simulations indicate subnanosecond data writing, and synaptic behavior in an in-plane magnetic field. Meeting all three of the criteria above, this device would be a good building block for neuromorphic circuitry.