Ultralow Power Optoelectronic Reconfigurable Hf_0.2Zr_0.8O_x-Based Antiferroelectric Device for Adaptive Image...

Reconfigurable electronics offer the ability to build compact and efficient neuromorphic systems. Existing reconfigurable devices suffer from complex switching between synapses and neurons. In this work, we demonstrated the on-demand creation of artificial neurons, synapses, and memory capacitors based on antiferroelectric Hf_0.2Zr_0.8O_x thin films, which can be reversibly reconfigured for a specific purpose by optical/electrical hybrid stimuli without chronological delay. The reconfigurable device exhibits significant advantages in raising the operating speed and reducing the complexity of neuron circuits. The neuron exhibits low energy consumption (90 aJ/spike), high endurance (>10¹¹), high uniformity, and high stability. Based on reconfigurable artificial neurons and synapses, adaptive image recognition outperformed existing preprocessing steps. The on-chip power consumption was reduced by 42.1% without degradation of the recognition accuracy. This work opens positive directions to achieve highly efficient neuromorphic systems with multichannel hybrid stimuli, which provides a promising approach to building energy-efficient neuromorphic hardware.