Spin–Orbit Torque-Driven Memristor in L1₀ FePt Systems with Nanoscale-Thick Layers for Neuromorphic Computing

Abstract

In this study, a memristor driven by spin−orbit torque (SOT) is realized in the nanoscale thickness L1 0 FePt systems with high perpendicular magnetization anisotropy (PMA). Due to the domain nucleation and expansion driven by current pulses, multilevel Hall resistance states can be continuously tuned by current density, where the memristive states are retained by the domain wall pinning effects. The properties of multilevel resistance states for samples with different structures are associated with the magnitude of field-like torque, and the larger efficiency of field-like torque enhances multiple resistance characteristics. Furthermore, the stable memristive behavior is obtained in the FePt/MgO/NiFe heterostructure. Finally, a three-layer multilayer perceptron (MLP) neural network is built to perform the MNIST handwritten digit recognition task based on the device's memristive behaviors, and the accuracy of weight update can reach up to 88.55%. These results pave the way for the application of L1 0 FePt nanomaterials in neuromorphic computing.

Field-Free Magnetization Switching in A1 CoPt Single-Layer Nanostructures for Neuromorphic Computing

“…The network structure has three layers, as shown in Figure b. This neural network is based on the back propagation (BP) structure, which has strong nonlinear mapping ability and a trainable network structure and has been widely used in the field of artificial intelligence . The resistance state of the SOT synaptic device was used to update the synaptic weights between the hidden layer and the output layer in the network, and the weights were randomly assigned before training.…”

“…This neural network is based on the back propagation (BP) structure, which has strong nonlinear mapping ability and a trainable network structure and has been widely used in the field of artificial intelligence. 29 The resistance state of the SOT synaptic device was used to update the synaptic weights between the hidden layer and the output layer in the network, and the weights were randomly assigned before training. Training was carried out using the BP gradient-descent technique, and the synaptic weights were updated by mapping the ideal weights determined theoretically with the resistance state of our apparatus.…”

Neuromorphic Computing with Emerging Antiferromagnetic Ordering in Spin–Orbit Torque Devices

“…Hereafter, this scenario is termed ideal . The second scenario was implemented by using an experimentally constructed synapse based on 150 R Hall states (Figure d–f) for synaptic weight , and fitted sigmoidal functions (Figure b) as the hidden layer activation function . The second scenario is experimental and therefore termed exp .…”

“…This can be attributed to massive parallel processing and high energy efficiency . Several reports have suggested that SOT-based devices could be used as artificial synapses for NC. ,,− Yadav et al experimentally constructed LTD/LTP responses for the Pt/Co/SiO 2 SOT system and utilized it as an artificial synapse for NC . Kurenkov et al realized an artificial neuron and synapse in an AFM/FM-based SOT device for NC .…”

“…In addition to the switching induced by current-driven SOT, we observed an analog change in the magnetization of the Co layer when the devices were exposed to a pulsed electric current, as shown in Figure . This trait could potentially be used as a synapse ,,− ,,− for information transmission in high-efficiency NC applications. Figure illustrates the neuromorphic characteristics of NiO-based SOT devices under an applied pulse current.…”