To improve the sensitivity of the magnetic tunnel junction(MTJ)sensor, a novel architecture for a double-gap magnetic flux concentrator (MFC) was studied theoretically and experimentally in this paper. The three-dimensional finite element model of magnetic flux was established to optimize the magnetic field amplification factor, with different gaps. The simulation results indicate that the sensitivity of an MTJ sensor with a double-gap MFC can be significantly better than that of a sensor with a traditional single-gap MFC, due to the fact that the magnetic magnification sharply increases with the decrease in effective gap width. Besides this, the half-bridge MTJ sensors with the double-gap MFC were fabricated using photolithography, ion milling, evaporation, and electroplating processes. Experimental results show that the sensitivity of the MTJ sensor increased by ten times compared to the sensor without the double-gap MFC, which underlines the theoretical predictions. Furthermore, there is no significant increase in the sensor noise. The work in this paper contributes to the development of high-performance MTJ sensors.

Physical reservoir computing using magnetic tunneling junction leverages the inherent nonlinearity of physical systems for computation, offering advantages such as low energy consumption and low hardware overhead for the time-series data preprocessing. However, parallelism and multiple-signal processing abilities remain challenges for it. In this work, we utilized the synchronized oscillations between coupling magnetic tunneling junction with its spin-transfer torque oscillators (STOs) effect as a nonlinear dynamical resource to construct the PRC unit. Unlike traditional time-multiplexing methods, this network directly processes information through the voltage amplitude and frequency dynamics of magnetic coupled STO units. The results demonstrate that this architecture achieves accuracy and memory capacity in classification tasks compared to existing solutions. Furthermore, we also used it for aeromagnetic compensation. The results show that it significantly enhances the compensation accuracy and network convergence speed. This work validates the potential of magnetic tunneling junction for high-performance parallel computing and provides an innovative path for the design of low-power neuromorphic hardware.

Neuromorphic computing is one of the most promising technologies to solve the von Neumann bottleneck, which has the advantages of fast processing speed and low energy consumption in performing complex tasks. The development of neuromorphic computing is currently driven by several kinds of novel devices. Magnetic tunnel junctions (MTJs) are rich in nonlinear properties and can be regulated by multiple physical fields such as magnetic field, current and temperature. Meanwhile, MTJ has the advantages of good stability and low power consumption, which makes it an ideal device for neuromorphic computing. This paper starts by examining individual MTJ devices and then extends the discussion to full neural networks. First of all, we sorted out the various properties of MTJ, from the structure to physical mechanism and response characteristics. Secondly, the biological neuron model, synaptic properties and related studies on simulating neurons and synapses based on MTJs are introduced. Then, we review the neural network system-level architectures that have been explored with MTJ devices. Finally, the challenges and the future development trend are summarized for advancing MTJ-enabled neuromorphic computing.