-
Voltage Gated Domain Wall Magnetic Tunnel Junction-based Spiking Convolutional Neural Network
Authors:
Aijaz H Lone,
Hanrui Li,
Nazek El-Atab,
Xiaohang Li,
Hossein Fariborzi
Abstract:
We propose a novel spin-orbit torque (SOT) driven and voltage-gated domain wall motion (DWM)-based MTJ device and its application in neuromorphic computing. We show that by utilizing the voltage-controlled gating effect on the DWM, the access transistor can be eliminated. The device provides more control over individual synapse writing and shows highly linear synaptic behavior. The linearity depen…
▽ More
We propose a novel spin-orbit torque (SOT) driven and voltage-gated domain wall motion (DWM)-based MTJ device and its application in neuromorphic computing. We show that by utilizing the voltage-controlled gating effect on the DWM, the access transistor can be eliminated. The device provides more control over individual synapse writing and shows highly linear synaptic behavior. The linearity dependence on material parameters such as DMI and temperature is evaluated for real-environment performance analysis. Furthermore, using skyrmion-based leaky integrate and fire neuron model, we implement the spiking convolutional neural network for pattern recognition applications on the CIFAR-10 data set. The accuracy of the device is above 85%, proving its applicability in SNN.
△ Less
Submitted 19 December, 2022;
originally announced December 2022.
-
Skyrmion-Magnetic Tunnel Junction Synapse with Mixed Synaptic Plasticity for Neuromorphic Computing
Authors:
Aijaz H. Lone,
Arnab Ganguly,
Selma Amara,
Gobind Das,
H. Fariborzi
Abstract:
Magnetic skyrmion-based data storage and unconventional computing devices have gained increasing attention due to their topological protection, small size, and low driving current. However, skyrmion creation, deletion, and motion are still being studied. In this study, we propose a skyrmion-based neuromorphic magnetic tunnel junction (MTJ) device with both long- and short-term plasticity (LTP and…
▽ More
Magnetic skyrmion-based data storage and unconventional computing devices have gained increasing attention due to their topological protection, small size, and low driving current. However, skyrmion creation, deletion, and motion are still being studied. In this study, we propose a skyrmion-based neuromorphic magnetic tunnel junction (MTJ) device with both long- and short-term plasticity (LTP and STP) (mixed synaptic plasticity). We showed that plasticity could be controlled by magnetic field, spin-orbit torque (SOT), and the voltage-controlled magnetic anisotropy (VCMA) switching mechanism. LTP depends on the skyrmion density and is manipulated by the SOT and magnetic field while STP is controlled by the VCMA. The LTP property of the device was utilized for static image recognition. By incorporating the STP feature, the device gained additional temporal filtering ability and could adapt to a dynamic environment. The skyrmions were conserved and confined to a nanotrack to minimize the skyrmion nucleation energy. The synapse device was trained and tested for emulating a deep neural network. We observed that when the skyrmion density was increased, the inference accuracy improved: 90% accuracy was achieved by the system at the highest density. We further demonstrated the dynamic environment learning and inference capabilities of the proposed device.
△ Less
Submitted 30 May, 2022;
originally announced May 2022.
-
Skyrmion-based Leaky Integrate and Fire Neurons for Neuromorphic Applications
Authors:
Aijaz H. Lone,
Selma Amara,
Fernando Aguirre,
Mario Lanza,
H. Fariborzi
Abstract:
Spintronics is an important emerging technology for data storage and computation. In this field, magnetic skyrmion-based devices are attractive due to their small size and energy consumption. However, controlling the creation, deletion and motion of skyrmions is challenging. Here we propose a novel energy-efficient skyrmion-based device structure, and demonstrate its use as leaky integrate (LIF) a…
▽ More
Spintronics is an important emerging technology for data storage and computation. In this field, magnetic skyrmion-based devices are attractive due to their small size and energy consumption. However, controlling the creation, deletion and motion of skyrmions is challenging. Here we propose a novel energy-efficient skyrmion-based device structure, and demonstrate its use as leaky integrate (LIF) and fire neuron for neuromorphic computing. Here we show that skyrmions can be confined by patterning the geometry of the free layer in a magnetic tunnel junction (MTJ), and demonstrate that the size of the skyrmion can be adjusted by applying pulsed voltage stresses. A spiking neural network (SNN) made of such skyrmion-based LIF neurons shows the capability of classifying images from the Modified National Institute of Standards and Technology (MNIST) dataset.
△ Less
Submitted 30 May, 2022;
originally announced May 2022.
