Abstract
Since the experimental discovery of magnetic skyrmions achieved one decade ago, there have been significant efforts to bring the virtual particles into all-electrical fully functional devices, inspired by their fascinating physical and topological properties suitable for future low-power electronics. Here, we experimentally demonstrate such a device: electrically-operating skyrmion-based artificial synaptic device designed for neuromorphic computing. We present that controlled current-induced creation, motion, detection and deletion of skyrmions in ferrimagnetic multilayers can be harnessed in a single device at room temperature to imitate the behaviors of biological synapses. Using simulations, we demonstrate that such skyrmion-based synapses could be used to perform neuromorphic pattern-recognition computing using handwritten recognition data set, reaching to the accuracy of ~89 percents, comparable to the software-based training accuracy of ~94 percents. Chip-level simulation then highlights the potential of skyrmion synapse compared to existing technologies. Our findings experimentally illustrate the basic concepts of skyrmion-based fully functional electronic devices while providing a new building block in the emerging field of spintronics-based bio-inspired computing.
Introduction
To build computers that think like the human brain (neuromorphic computing), we need hardware "synapses" that can store analog weights efficiently. While technologies like Resistive RAM (RRAM) and Phase-Change Memory (PCM) exist, they often struggle with high operating voltages, limited endurance, or non-linear behavior during training. Spintronics—using the spin of electrons rather than just their charge—offers a low-power alternative. Magnetic skyrmions are particularly exciting because they are topologically protected and behave like discrete particles. This study experimentally demonstrates the first all-electrical artificial synapse based on magnetic skyrmions, integrating the entire synaptic lifecycle—creation, motion, detection, and deletion—into a single ferrimagnetic device.
Experimental Design and Methods
The team fabricated a multilayer thin-film stack consisting of [Pt/GdFeCo/MgO]₂₀. This specific ferrimagnetic material was chosen because it supports Néel-type skyrmions at room temperature and allows for efficient electrical manipulation.
Key technical components included:
- STXM Imaging: Scanning Transmission X-ray Microscopy was used to visually confirm the presence and number of skyrmions during operation.
- SOT Manipulation: Spin-Orbit Torques (SOT) and Joule heating were used to generate and move skyrmions into an "active" synaptic area.
- Electrical Readout: The "weight" of the synapse was read via the Hall resistance, which changes linearly based on how many skyrmions are present in the device's cross-section.
Synaptic Performance and Results
A major advantage of skyrmions is their particle-like behavior, allowing multiple skyrmions to accumulate in a single area to create a multi-level analog memory state. Crucially, the Hall resistance is remarkably linear with respect to the number of skyrmions, making the training process stable.
The researchers demonstrated 24 distinct resistance states during potentiation (adding skyrmions) and 16 states during depression (removing them). In simulations of a three-layer Artificial Neural Network (ANN) using the MNIST dataset, the skyrmion-based synapses achieved ~89% accuracy, close to the ~94% accuracy of ideal software-based training. Chip-level simulations suggest these synapses could operate at significantly lower voltages (~30mV) compared to RRAM (~1.8V).
Discussion and Outlook
While a major breakthrough, challenges remain. The current ON/OFF ratio is relatively low (<200%) compared to RRAM, and device footprints must be scaled down from the micrometer range to ~10nm to compete with CMOS. Future work will focus on using Magnetic Tunnel Junctions (MTJs) to boost the signal and moving toward large-scale crossbar arrays to test the scalability of skyrmion-based neuromorphic chips.
Summary updated