Temperature-driven co-optimization of IGZO/HZO ferroelectric field-effect transistors for optoelectronic neuromorphic computing

Abstract

The integration of ferroelectricity and the photoelectric effect offers substantial potential for the realization of optoelectronic-based artificial neural networks (ANNs) and biomimetic systems. Despite their potential, the application of ferroelectric field-effect transistors (FeFETs) in optoelectronic neuromorphic devices, where light regulates synaptic activation, remains largely unexplored. Here, we report InGaZnO/Zr-doped HfO₂ (IGZO/HZO) artificial synapses for the emulation of optoelectronic ANNs and optogenetic-inspired neural functions. Particularly, by simultaneously optimizing the post-deposition annealing (PDA) process for the stabilization of HZO ferroelectric phase and activation of IGZO channel, high-performance FeFETs exhibiting a memory window of ∼1 V and endurance up to 10⁴ cycles were achieved. Spectroscopic analyses correlated PDA temperature-dependent dynamic transitions in electrical properties with hydrogen relocation, oxygen vacancy formation, zinc vaporization, and film densification. The IGZO/HZO synapses successfully emulated synaptic functions such as spike-amplitude and spike-duration-dependent plasticity, using both electrical (E) and optical (O) stimulation. Furthermore, dual E/O stimulus activation of IGZO/HZO synapses, combined with polarization state modulation, was employed to emulate optogenetic-inspired neural functions. We also demonstrated that dual E/O stimulus activation of artificial synapses enhanced ANN image recognition accuracy from 86 % to 90 %, outperforming the performance achieved with a single E-stimulation. We envision that these findings can provide a process protocol for IGZO/HZO FeFETs to achieve stable phasic control and realization of optoelectronic neuromorphic devices.