Study and Design of a Skyrmion based Leaky Integrate-and-Fire Neuron

Among all the neurons proposed in the literature to imitate brain’s functioning in terms of performing complex computational operations efficiently, spintronic neurons, leveraging both the magnetic and electrical properties of electrons are one of the most favored types because they can bypass the Moore’s law limitations for scaling and energy efficiency present in CMOS technology. The focus of this thesis is to study and design a Leaky Integrate-and-Fire neuron based on the motion of a skyrmion along the magnetic nano-track consisting of W/CoFeB/Mgo/Ta stack. A previous study, examined the LIF with the domain wall motion on a Ta/CoFeB/Mgo track. The inputs of the neuron which are pulses applied to the heavy metal layer, generate Spin Orbit Torque (SOT) induced current on the top ferromagnetic layer to realize the motion of skyrmion over the track. The effect of Voltage Controlled Magnetic Anisotropy (VCMA) creates a localized and configurable energy barrier for the skyrmion, inhibiting its motion only if the applied current exceeds a defined threshold. Firing property of the neuron is reached by using a Magnetic Tunnel Junction (MTJ) placed at the end of the track. An intrinsic leaking property, without additional circuitry, is achieved with applying an anisotropy gradient along the track. Transitioning from a lower anisotropy to a higher one at the other end. All simulations in this thesis were conducted using Mumax3 micromagnetic simulator, and the results were post-processed by MATLAB.