Item

Abstract

Spintronics is a growing research field that focuses on exploring materials and devices that take advantage of the electron’s “spin” to go beyond charge based devices. The most impactful spintronic device to date is a highly sensitive magnetic field sensor, the spin-valve, that allowed for a 10,000-fold increase in the storage capacity of hard disk drives since it was first introduced in a magnetic recording read head in 1997. In about 2007, the original spin-valve that was based on spin-dependent scattering in metallic magnetic/non-magnetic interfaces evolved to a closely related device in which the essential physics changed to that of spin-dependent tunneling across ultra-thin insulating layers placed between magnetic electrodes, but the basic spin-engineered structure remained largely unchanged. These latter structures were proposed in 1995 as potential memory elements for a magnetic random-access memory (MRAM) and the first demonstration of this possibility was made in 1999. It was only recently (about 2019) that MRAM became a mainstream foundry technology.

Compared with most conventional charge based electronic devices, spintronic devices have the advantage of non-volatility, low-power consumption, and scalability to smaller dimensions. For these reasons, spintronic devices are highly attractive for next-generation information memory-storage and are promising for advanced applications such as in-memory computing. Furthermore, spintronics allows for a unique high capacity, non-volatile, solid-state memory-storage device that relies on devices that can store multiple digital bits in the form of a series of chiral domain walls that are moved at high-speed using nanosecond long current pulses along magnetic nanowires. These devices also enable synaptic functionalities in neuromorphic computing and are therefore, potential hardware candidates for artificial intelligence.

In this review article, recent advances in multi-state spintronic devices are discussed. The review starts with an introduction followed by a discussion on using domain-walls for achieving multiple states for memory and neuromorphic computing. In the next section, achieving multiple levels based on domain nucleation are discussed. Subsequent discussions review the use of magnetic pillars, and other schemes for achieving high-density memory. The prospects of spintronic devices in neuromorphic computing for artificial intelligence (AI) are also presented. The outlook and directions for new research are provided at the end.