Harvesting the Elastic Nature of Domain Walls to Emulate the Synaptic Functions for Brain-Inspired Computing.

Neuromorphic computing (NC), based on a network of numerous neurons and synapses, promises to lead the future of artificial intelligence (AI). Among various material platforms, spintronic domain wall (DW) devices emerge as promising energy-efficient candidates for NC. Despite various proposals, the experimental realization of DW devices-based NC is only starting. Therefore, the concept of ladder DW devices, whose 2D layout resembles a conventional ladder, as a synaptic element, is experimentally studied. First, spin-orbit torque-driven DW dynamics is investigated in proposed devices and found DW pinning across a broad range of characteristic parameters (tilt angle (θ)) and current density. Moreover, maximum pinning strength is observed for θ = 30° in the investigated range. Further, micromagnetic simulations suggest that i) Laplace pressure on the DWs and ii) non-uniform distribution of the current in various device segments mainly account for the experimental observations. Further, the current density is optimized and observed a total of 19 multilevel magnetization states, confirming the stability of the design in all the studied devices.