A low-power Si:HfO2 ferroelectric tunnel memristor for spiking neural networks

As key components of the human brain's neural network, synapses and neurons are important processing units that enable highly complex neuromorphic systems. Spiking neural network (SNN) is more powerful and efficient in terms of neuromorphic computing. Moreover, memristor-based neuromorphic computers can implement neural network algorithms more effectively than conventional hardware. However, the investigation on spiking neural network (SNN) based neuromorphic computing is still in the exploratory stage. Herein, a SNN based on ferroelectric Si:HfO2 film (-6.8 nm) memristor was realized. The Si:HfO2-based memristor exhibits lower switching voltage (1.55/-1.50 V) and super low power consumption (-32.65 fJ). Additionally, it also shows superior conductance tunability and reliable realization of multiple synaptic functions. Especially, the highly linear conductance modulation of the Si:HfO2-based memristor results in a high accuracy of -96.23 % for handwritten digits. Spatiotemporal model recognition and unsupervised synaptic weight update functions were successfully implemented with the SNN constructed by these synaptic devices and artificial neuron models, which demonstrates the excellent adaptability and versatility of this SNN and paves the way for future neural network studies.

Automated summary

A spiking neural network (SNN) based on ferroelectric Si:HfO2 film (-6.8 nm) memristor was successfully realized in this study. The Si:HfO2-based memristor exhibited lower switching voltage (1.55/-1.50 V) and super low power consumption (-32.65 fJ), and reliably implemented multiple synaptic functions. The SNN constructed by these synaptic devices and artificial neuron models successfully implemented spatiotemporal model recognition and unsupervised synaptic weight update functions, demonstrating the excellent adaptability and versatility of this SNN.