ANTIFERROMAGNETIC SKYRMION BASED ENERGY-EFFICIENT NEURON DEVICES

Abstract

Artificial intelligence (AI) has the potential to revolutionize the way we work and live, but the high energy consumption of existing computing technology limits its application. To address this issue, researchers have been exploring ways to emulate the brain’s capability for processing the information in energy-efficient manner. Unlike modern computers that rely on complex algorithms and high-precision calculations, the brain performs many low-precision calculations in parallel to complete tasks like image and speech recognition. Additionally, computers consume significant amount of energy in shuttling the information between storage and the processor whereas the brain stores and processes the information locally. To overcome the limitations of modern computing technology, one promising approach is neuromorphic computing that uses hardware designed to mimic the structure and function of the brain. Over the past few decades, neuromorphic computing has primarily relied on complementary metal oxide semiconductor (CMOS) technology. However, due to the disparity between the computing units of the brain and the architecture of CMOS transistors, CMOS-based neuromorphic computing demands more energy and resources. To bridge this gap, alternative devices are necessary, where neural and synaptic functionalities correspond to the device's operation. Fortunately, recent advances in nanotechnology have led to the development of spintronic devices that use the spin of electrons to process the information instead of their charge. Magnetic skyrmion, in particular, is a potential candidate for neuromorphic computing owing to its topological stability, small size, and ultra-low depinning current density. To date, the majority of studies have been focused on ferromagnetic (FM) skyrmions based devices. However, due to skyrmion Hall effect (SkHE) experienced by FM skyrmions, there is a displacement of such skyrmions in the in-plane direction, perpendicular to the direction it is driven along the track using spin-transfer torque (STT)/spin-orbit torque (SOT) mechanism. This pushes the skyrmion towards the nanotrack edge, causing it to annihilate. Hence, there is a requirement to eliminate the SkHE intrinsically that allows efficient and ultrafast skyrmion motion along the racetrack. Moreover, FM skyrmions are also vulnerable to stray fields. To overcome these issues intrinsically, antiferromagnetic (AFM) skyrmion comes into picture that is formed as a pair of strongly coupled sub-lattices that are aligned antiparallel to each other. The AFM skyrmionics for the implementation of neuromorphic computing is yet to be explored and the challenge of designing high-speed and energy-efficient neuron devices based on AFM skyrmions remains. Hence, in this context, the requirement of designing the AFM skyrmion based high-speed energy-efficient neuron devices is a great concern.