A research team led by Prof. ZENG Zhongming from the Nano Fabrication Facility of Suzhou Institute of Nano-Tech and Nano-Bionics (SINANO), Chinese Academy of Sciences (CAS) proposed a voltage-controlled spintronic device that enables low-energy neuromorphic computing, with the stochastic behavior of the device being managed by an applied electric field via voltage-controlled magnetic anisotropy. This work was published in Physical Reviews Applied.
Neuromorphic computing is utilizing the hardware to mimic the functionalities involved in the neurons and synapses in the human brain. Recently, neuromorphic computing based on stochastic spintronic units has attracted intense attention, but controlling such a stochastic system with high energy efficiency remains a challenge.
In this work, the researchers developed a spintronic device to mimic artificial neuron for neuromorphic computing. This spintronic neuron device can produce stochastic spiking signal with an ultra-low power (< 1 nW) while the frequency of the spiking signal is controllable by magnetic field or voltage bias. They applied this spintronic neuron device to construct an artificial neural network, and the recognition of MNIST handwritten digits is realized, with a recognition rate reaching 95%.
The above technology enables to mimic neurons using spintronic devices with low energy consumption and multi-control methods which will advance the quest to create energy-efficient spintronic systems for brain-like cognitive computing.
This work was supported by National Natural Science Foundation of China and China Postdoctoral Science Foundation.
Prof.ZENG Zhongming, Suzhou institute of Nano-Tech and Nano-Bionic, Chinese Academy of Sciences.