Switching plasticity in compensated ferrimagnetic multilayers for neuromorphic computing

Li, Weihao; Lan, Xiukai; Liu, Xionghua; Zhang, Enze; Deng, Yongcheng; Wang, Kaiyou

doi:10.1088/1674-1056/ac89dd

Cited by 7 publications

(5 citation statements)

References 47 publications

(56 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The non-volatility of magnetic elements naturally allows for data storage, while ultra-low-power control mechanisms, such as spin-polarised currents or applied strain [27,28] offer routes towards energy-efficient logic-in-memory computing. Recent demonstrations of spin-orbit torque based devices have shown how magnetic materials can be used as both binary and multi-level synapses for efficient neuromorphic systems [29][30][31][32][33]. Meanwhile, ongoing developments have shown how to manipulate magnetic domains to both move data and process it [22,24,34,35].…”

Section: Introductionmentioning

confidence: 99%

Machine learning using magnetic stochastic synapses

Ellis

Welbourne

Kyle

et al. 2023

Neuromorph. Comput. Eng.

View full text Add to dashboard Cite

The impressive performance of artificial neural networks has come at the cost of high energy usage and CO2 emissions. Unconventional computing architectures, with magnetic systems as a candidate, have potential as alternative energy-efficient hardware, but, still face challenges, such as stochastic behaviour, in implementation. Here, we present a methodology for exploiting the traditionally detrimental stochastic effects in magnetic domain-wall motion in nanowires. We demonstrate functional binary stochastic synapses alongside a gradient learning rule that allows their training with applicability to a range of stochastic systems. The rule, utilising the mean and variance of the neuronal output distribution, finds a trade-off between synaptic stochasticity and energy efficiency depending on the number of measurements of each synapse. For single measurements, the rule results in binary synapses with minimal stochasticity, sacrificing potential performance for robustness. For multiple measurements, synaptic distributions are broad, approximating better-performing continuous synapses. This observation allows us to choose design principles depending on the desired performance and the device's operational speed and energy cost. We verify performance on physical hardware, showing it is comparable to a standard neural network.

show abstract

Section: Introductionmentioning

confidence: 99%

Machine learning using magnetic stochastic synapses

Ellis

Welbourne

Kyle

et al. 2023

Neuromorph. Comput. Eng.

View full text Add to dashboard Cite

show abstract

“…[17][18][19][20] Recently, there was some related hardware-accelertation of SNN or ANN. [21][22][23][24][25][26][27] The most obvious advantage of SNN is the ultra-low energy cost for the event-triggered characteristic. In the hardware implementation of SNN, the main components, synapses and neurons, are usually based on different materials and structures.…”

Section: Introductionmentioning

confidence: 99%

Reconfigurable Mott electronics for homogeneous neuromorphic platform

Yang 杨,

Lu 路,

Yang 杨

2023

Chinese Phys. B

View full text Add to dashboard Cite

To simplify the fabrication process and increase the versatility of neuromorphic systems, the reconfiguration concept has attracted much attention. Here, we developed a novel electrochemical VO2 (EC-VO2) device, which can be reconfigured as synapses or LIF neurons. The ionic dynamic doping contributes to the resistance changes of VO2, which enables the reversible modulation of device states. The analog resistance switching and tunable LIF functions were both measured based on the same device to demonstrate the capacity of reconfiguration. Based on the reconfigurable EC-VO2, the simulated spiking neural network model exhibited excellent performances by using low-precision weights and tunable output neurons, whose final accuracy reached 91.92%.

show abstract

“…The Pt/Co/HfOx/Ti stack exhibits a gradual switching behavior, accompanied by an increase in the number of resistance states. Moreover, the linear variation region of the magnetization reversal curve for the two samples, Δ J = J max – J min , is quantitatively calculated, which is the difference between the minimum current required to change the R xy ( J min ) value, and the maximum current within the linear change of the R xy ( J max ) value, , as shown in the insets of Figure c and d. The Pt/Co/HfOx/Ti sample has a wider linear variation region, up to 5.81 × 10 10 A/m 2 , compared to that of the Pt/Co/Ti sample, with Δ J = 2.91 × 10 10 A/m 2 , indicating more multiresistance states. The critical switching current density of the two samples has been summarized, and the details are presented in Supporting Information S6.…”

mentioning

confidence: 99%

Granular Magnetization Switching in Pt/Co/Ti Structure with HfOx Insertion for In-Memory Computing Applications

Jin,

Zhang,

Tan

et al. 2024

Nano Lett.

View full text Add to dashboard Cite

Exploring multiple states based on the domain wall (DW) position has garnered increased attention for in-memory computing applications, particularly focusing on the utilization of spin−orbit torque (SOT) to drive DW motion. However, devices relying on the DW position require efficient DW pinning. Here, we achieve granular magnetization switching by incorporating an HfOx insertion layer between the Co/Ti interface. This corresponds to a transition in the switching model from the DW motion to DW nucleation. Compared to the conventional Pt/Co/Ti structure, incorporation of the HfOx layer results in an enhanced SOT efficiency and a lower switching current density. We also realized stable multistate storage and synaptic plasticity by applying pulse current in the Pt/Co/HfOx/Ti device. The simulation of artificial neural networks (ANN) based on the device can perform digital recognition tasks with an accuracy rate of 91%. These results identify that DW nucleation with a Pt/Co/HfOx/Ti based device has potential applications in multistate storage and ANN.

show abstract

Switching plasticity in compensated ferrimagnetic multilayers for neuromorphic computing

Cited by 7 publications

References 47 publications

Machine learning using magnetic stochastic synapses

Machine learning using magnetic stochastic synapses

Reconfigurable Mott electronics for homogeneous neuromorphic platform

Granular Magnetization Switching in Pt/Co/Ti Structure with HfOx Insertion for In-Memory Computing Applications

Contact Info

Product

Resources

About