【Device Papers】Enhanced electrical performances with HZO/β-Ga₂O₃ 3D FinFET toward highly perceptual synaptic device application

      Researchers from the Jeonbuk National University have published a dissertation titled "Enhanced electrical performances with HZO/β-Ga₂O₃ 3D FinFET toward highly perceptual synaptic device application" in Materials Science in Semiconductor Processing.

Abstract

      In this study, we have demonstrated a novel high-performance ferroelectric FinFET (Fe-FinFET) that integrates ultra-wide-bandgap (UWBG) beta-gallium oxide (β-Ga₂O₃) channel with an atomic layer deposited (ALD) hafnium zirconium oxide (HZO) ferroelectric layer for the first time, and experimentally validated the effectiveness of the robust β-Ga₂O₃ platform as a memory application. Compared with conventional planar ferroelectric FET (FeFET), the Fe-FinFET exhibits a markedly wider counter-clockwise hysteresis loop in its transfer characteristics, achieving a large memory window (MW) of 13.9 V with a single HZO layer. When normalized to the actual channel width, the Fe-FinFETs show an improved I_ON/I_OFF ratio of 2.3 × 10⁷ and a subthreshold swing value of 110 mV/dec; Y-function method further indicates that the intrinsic mobility is improved to 4.25 × 10² cm²/Vs. The enhanced polarization due to larger electric fields across the ferroelectric layer in Fe-FinFET is validated using Sentaurus TCAD, and this result is further supported by the energy-dependent distribution of interface trap density (D_it) extracted in both forward and reverse sweep directions. After 5 × 10⁶ program/erase (PGM/ERS) cycles, the MW was maintained at 9.2 V, and the retention time was measured up to 3 × 10⁴ s with low degradation. To verify its potential as an artificial synapse, we trained a convolutional neural network (CNN) and achieved an accuracy of 91.7 %. These results establish the HZO/β-Ga₂O₃ Fe-FinFET as a promising candidate for high voltage-enduring, non-volatile memory devices that also offer synaptic functionality for neuromorphic applications.

DOI：

https://doi.org/10.1016/j.mssp.2025.110104