LL
Long Liu
Author with expertise in Memristive Devices for Neuromorphic Computing
Achievements
Open Access Advocate
Key Stats
Upvotes received:
0
Publications:
1
(100% Open Access)
Cited by:
1
h-index:
4
/
i10-index:
1
Reputation
Biology
< 1%
Chemistry
< 1%
Economics
< 1%
Show more
How is this calculated?
Publications
0

Domain wall magnetic tunnel junction-based artificial synapses and neurons for all-spin neuromorphic hardware

Long Liu et al.May 28, 2024
Abstract We report a breakthrough in the hardware implementation of energy-efficient all-spin synapse and neuron devices for highly scalable integrated neuromorphic circuits. Our work demonstrates the successful execution of all-spin synapse and activation function generator using domain wall-magnetic tunnel junctions. By harnessing the synergistic effects of spin-orbit torque and interfacial Dzyaloshinskii-Moriya interaction in selectively etched spin-orbit coupling layers, we achieve a programmable multi-state synaptic device with high reliability. Our first-principles calculations confirm that the reduced atomic distance between 5 d and 3 d atoms enhances Dzyaloshinskii-Moriya interaction, leading to stable domain wall pinning. Our experimental results, supported by visualizing energy landscapes and theoretical simulations, validate the proposed mechanism. Furthermore, we demonstrate a spin-neuron with a sigmoidal activation function, enabling high operation frequency up to 20 MHz and low energy consumption of 508 fJ/operation. A neuron circuit design with a compact sigmoidal cell area and low power consumption is also presented, along with corroborated experimental implementation. Our findings highlight the great potential of domain wall-magnetic tunnel junctions in the development of all-spin neuromorphic computing hardware, offering exciting possibilities for energy-efficient and scalable neural network architectures.