Crystallization-Driven Synaptic Behavior in Hafnia-based Ferroelectric Tunnel Junction Devices
Year: 2025
Type: article
Abstract: We present a device-to-system evaluation of ferroelectric tunnel junction synapses based on hafnium zirconium oxide (HZO), zirconium dioxide (ZrO 2 ), and morphotropic-phase-boundary (MPB) hafnia-zirconia films integrated in tungsten (W)/oxide/W metal-ferroelectric-metal stacks.HZO shows the most system-friendly behavior [dynamic range (ΔG) ≈ 1.15 𝜇S, symmetry factor (𝛼) ≈ 1.17], yielding accuracy 0.9659 → 0.9663 (linearized) versus the software baseline 0.9665.ZrO 2 exhibits small ΔG (≈ 9.1 nS) and strong curvature yet good 𝛼 (≈ 0.98); Linearization markedly improves accuracy 0.9475 → 0.9647.MPB exhibits intermediate behavior (ΔG ≈ 8.9 nS, 𝛼 ≈ 1.17) with a modest gain (0.9640 → 0.9650).The gap between measured and linearized curves correlates with confusion-matrix off-diagonals, linking update physics to class-wise errors.These results provide a compact guideline: HZO needs minimal calibration, ZrO 2 benefits strongly from write-verify, and MPB offers low-energy fine quantization with light calibration.
Author: Hyeon Jun Hwang
Institution: Mokpo National University
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: diamond
Funder: Mokpo National University
OpenAlex ID: W4416939667