The conventional computing method based on the von Neumann architecture is limited by a series of problems such as high energy consumption, finite data exchange bandwidth between processors and storage media, etc., and it is difficult to achieve higher computing efficiency. A more efficient unconventional computing architecture is urgently needed to overcome these problems. Neuromorphic computing and stochastic computing have been considered to be two competitive candidates for unconventional computing, due to their extraordinary potential for energy-efficient and high-performance computing. Although conventional electronic devices can mimic the topology of the human brain, these require high power consumption and large area. Spintronic devices represented by magnetic tunnel junctions (MTJs) exhibit remarkable high-energy efficiency, non-volatility, and similarity to biological nervous systems, making them one of the promising candidates for unconventional computing. In this work, we review the fundamentals of MTJs as well as the development of MTJ-based neurons, synapses, and probabilistic-bit. In the section on neuromorphic computing, we review a variety of neural networks composed of MTJ-based neurons and synapses, including multilayer perceptrons, convolutional neural networks, recurrent neural networks, and spiking neural networks, which are the closest to the biological neural system. In the section on stochastic computing, we review the applications of MTJ-based p-bits, including Boltzmann machines, Ising machines, and Bayesian networks. Furthermore, the challenges to developing these novel technologies are briefly discussed at the end of each section.

The writing performance of the easy-cone magnetic tunnel junction (MTJ) and perpendicularly magnetized MTJ (pMTJ) under various temperatures was investigated based on the macrospin model. When the temperature is changed from 273 K to 373 K, the switching current density of the pMTJ changes by 56%, whereas this value is only 8% in the easy-cone MTJ. Similarly, the temperature-induced variation of the switching delay is more significant in the pMTJ. This indicates that the easy-cone MTJ has a more stable writing performance under temperature variations, resulting in a wider operating temperature range. In addition, these two types of MTJs exhibit opposite temperature dependence in the current overdrive and write error rate. In the easy cone MTJ, these two performance metrics will reduce as temperature is increased. The results shown in this work demonstrate that the easy-cone MTJ is more suitable to work at high temperatures compared with the pMTJ. Our work provides a guidance for the design of STT-MRAM that is required to operate at high temperatures.

The dynamics of a spin torque-driven ferrimagnetic (FiM) system is investigated using the two-sublattice macrospin model. We demonstrate ultrafast switching in the picosecond range. However, we find that the excessive current leads to magnetic oscillation. Therefore, faster switching cannot be achieved by unlimitedly increasing the current. By systematically studying the impact of thermal fluctuations, we find that the dynamics of FiMs can also be distinguished into the precessional region, the thermally activated region, and the crossover region. However, in the precessional region, there is a significant deviation between FiM and ferromagnet (FM), i.e., the FM is insensitive to thermal fluctuations since its switching is only determined by the amount of net charge. In contrast, we find that the thermal effect is pronounced even when a very short current pulse is applied to the FiM. We attribute this anomalous effect to the complex relation between the anisotropy and overdrive current. By controlling the magnetic anisotropy, we demonstrate that the FiM can also be configured to be insensitive to thermal fluctuations. This controllable thermal property makes the FiM promising in many emerging applications such as the implementation of tunable activation functions in the neuromorphic computing.

The field-free spin–orbit torque induced 180° reorientation of magnetization is beneficial for the high performance magnetic memory. The antiferromagnetic material (AFM) can provide a higher operation speed than the ferromagnetic counterpart. In this paper, we propose a trilayer AFM/insulator/heavy metal structure as the AFM memory device. We show that the field-free switching of the AFM with a perpendicular Néel vector can be achieved by using two orthogonal currents, which provide a uniform damping-like torque and uniform field-like torque, respectively. The reversible switching can be obtained by reversing either current. A current density of 1.79 × 1011 A/m2 is sufficient to induce the switching. In addition, the two magnetic moments become noncollinear during switching. This enables an ultrafast switching within 40 ps. The device and switching mechanism proposed in this work offer a promising approach to deterministically switch the AFM with the perpendicular Néel vector. It can also stimulate the development of an ultrafast AFM-based MRAM.