Spin-torque diodes (STDs) offer the possibility of using spin torque to generate rectification voltage with promising applications in microwave detecting, energy harvesting, and neuromorphic computing. Here, we demonstrate a highly sensitive STD with ultralow current density based on a magnetic tunnel junction with perpendicular magnetic anisotropy. At zero magnetic field, a high sensitivity exceeding 3785 V/W is obtained with a low current of −20 μA, corresponding to a current density of ∼105 A/cm2, which is one order lower than the previously reported. When a weak external magnetic field is applied, the sensitivity can be further increased by five times to 20 000 V/W. Furthermore, we construct an artificial neural network with STD neurons to perform recognition of handwritten digits in the Mixed National Institute of Standards and Technology database, where a produced accuracy of up to 94.92% is obtained. Our work provides a route to develop low-power consumption high-sensitivity STDs for Internet of Things applications and neuromorphic computing.
We investigate the highly sensitive spin torque diode (STD) effect in a magnetic tunnel junction (MTJ) with an in-plane polarizer and an in-plane free layer. Under injection locking mechanisms, a high rectification voltage of 12 mV is obtained with an input radio frequency power of 1 μW under direct current bias current and a weak magnetic field, corresponding to a high sensitivity of 12 000 mV/mW. In addition, we use the nonlinear rectification characteristics of STD to mimic a neuron with a ReLU-like activation function to perform the recognition of handwritten digits in the Mixed National Institute of Standards and Technology database, where a produced accuracy of up to 93.53% is obtained. These findings suggest that the MTJ with in-plane magnetized electrodes holds promising potential in developing high sensitivity STDs for Internet of Things applications and neuromorphic computing.
Recently, it’s proposed that spin torque oscillators and spin torque diodes can be used as artificial neurons and synapses to directly process the microwave signal, which can lower latency and power consumption greatly. However, one critical challenge is to make the microwave emission frequency of the STO stay constant with the varying input current. In this work, we study the microwave emission characteristics of STO based on magnetic tunnel junction with MgO cap layer. By applying a small magnetic field, we realize invariability of the microwave emission frequency of STO, making it qualified to act as artificial neuron. Furthermore, we have simulated an artificial neural network using the STO neuron to recognize the handwritten digits in Mixed National Institute of Standards and Technology (MNIST) database and obtained a high accuracy of 92.28%. Our work paves the way for the development of RF-oriented neuromorphic computing system.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.