From the discovery of giant magnetoresistance (GMR) to tunnel magnetoresistance (TMR), their subsequent application in large capacity hard disk drives (HDDs) greatly speeded up the information era over the past decades. However, the growing demand for big-data storage and processing is limited by the von-Neumann architecture due to the memory bottleneck and power dissipation. Taking advantage of nonvolatility, high speed, and low power, magnetic random access memory (MRAM) becomes a promising candidate to overcome this limitation through processing-in-memory (PIM) architectures. In this article, we provide an overview of existing technology and give a roadmap of spintronic devices for future energy-efficient computing and its relevant integration architectures. We begin with the fundamentals of Toggle-MRAM and spin-transfer torque (STT)-MRAM, which already have commercial applications. We then introduce spin-orbit torque (SOT), a critical mechanism to realize low-power data manipulation in the next generation of MRAM and summarize the recent experimental breakthroughs of field-free SOT switching schemes. Finally, we present MRAM-based PIM architectures and novel spintronic devices, provide an application outlook, and deliver the future development potential of energy-efficient computing systems.

We present a NAND-like spintronics memory (NAND-SPIN) device for high-density non-volatile memory applications. Fast erasing and programming of magnetic tunnel junction (MTJ) are implemented with two unidirectional currents generating spin orbit torque (SOT) and spin transfer torque (STT), respectively. The asymmetric switching drawback of STT mechanism can be definitively overcome as only anti-parallel to parallel operation happens for NAND-SPIN programming, which allows lower switching current, smaller access transistor and reduced maximum write voltage across the MTJ. By sharing the SOT-induced erase operation in a nanowire, the area overhead due to the three-terminal structure can be also eliminated. Simulation results show that NAND-SPIN can achieve 3~5X reduction in write energy compared to STT-MRAM, and 2~4X less bit-cell area than SOT-MRAM at 28 nm node.

Neuromorphic computing, which relies on a combination of a large number of neurons massively interconnected by an even larger number of synapses, has been actively studied for its characteristics such as energy efficiency, intelligence, and adaptability. To date, while the development of artificial synapses has shown great progress with the introduction of emerging nanoelectronic devices, e.g., memristive devices, the implementation of artificial neurons, however, depends mostly on semiconductor-based circuits via integrating many transistors, sacrificing energy efficiency and integration density. Here, we present a novel compact neuron device that exploits the current-driven magnetic skyrmion dynamics in a wedge-shaped nanotrack. Under the coaction of the exciting current pulse and the repulsive force exerted by the nanotrack edges, the dynamic behavior of the proposed skyrmionic artificial neuron device is in analogy to the leaky–integrate–fire (LIF) spiking function of a biological neuron. The tunable temporary location of the skyrmion in our artificial neuron behaves like the analog membrane potential of a biological neuron. The neuronal dynamics and the related physical interpretations of the proposed skyrmionic neuron device are carefully investigated via micromagnetic and theoretical methods. Such a compact artificial neuron enables energy-efficient and high-density implementation of neuromorphic computing hardware.

We propose a toggle spin torques magnetic random access memory (TST-MRAM) for ultrafast computing. The write operation of the TST-MRAM is achieved by applying two currents to the MTJ and a heavy metal layer, respectively, in the toggle-like manner. These two currents generate spin transfer torque (STT) and spin orbit torque (SOT) to switch the magnetization. Thanks to the interplay effect between STT and SOT, low-energy and ultrafast write operation can be implemented without the need of the additional magnetic field. It is important to mention that both the amplitude and duration of the current flowing through the tunnel barrier can be decreased to enhance significantly the write endurance of the MTJ. The results demonstrate that the TST-MRAM is a good candidate for the upper level caches where the ultrafast operation and high write endurance are indispensable.

The interplay of spin-transfer torque (STT) and spin-orbit torque (SOT) highlights the potential of current-induced magnetization reversal for the ultrahigh-speed and ultralow-power memory applications. However, the ultrafast field-free switching and its dynamic mechanism have not been clarified yet. Here, we experimentally demonstrate the sub-nanosecond field-free switching by the interplay of STT and SOT in perpendicular magnetic tunnel junctions (MTJs). The dynamic investigations of the STT and SOT switching reveal that the applied SOT current can highly decrease the incubation time and current density of STT, which leads to the achievement of ultra-fast switching. Besides, the timing scheme investigations of SOT and STT pulses are performed. The firstly applied SOT current further decreases the transition time of the magnetization switching, which could result from the different nucleation and motions of domains. The interplay and the timing operations of STT and SOT provide more flexible solutions for high-speed, large-scalability and energy efficient memory applications.

A comprehensive study of the magnetic skyrmion dynamics in terms of size, velocity, energy, and stability in width-varying nanotracks is reported by micromagnetic simulations. We find that the diameter of a skyrmion reduces with the decrease in the nanotrack width in the spin Hall effect (SHE)-induced skyrmion motion. Accordingly, the skyrmion energy increases giving rise to the growing instability of the skyrmion. It is also numerically demonstrated that the velocity of the skyrmion varies during the motion, since the repulsive force of the nanotrack edges acting on the skyrmion as well as the driving force created by the SHE associated with the size of the skyrmion have a joint impact on the skyrmion motion dynamics in the width-varying nanotrack. In addition, one interesting finding reveals that skyrmions with small sizes, which may be inaccessible to typical approaches by means of directly injecting a spin-polarized current, could be obtained by utilizing this structure. This finding is potential for generating nanoscale skyrmions in skyrmionic applications with ultra-dense density. Finally, inspired by the skyrmion dynamics in the width-varying nanotrack, a general summary on the tradeoff between the nanotrack width (storage density) and the skyrmion velocity (data access speed) is given by further analyzing the skyrmion dynamics in parallel nanotracks with different widths, which may provide guidelines in designing racetrack-type skyrmionic applications.