Neuromorphic tunnel junctions using atomic-layer-deposited Al2O3 for emulating highly reliable synaptic functions
Year: 2026
Type: article
Abstract: A frequent outcome of neuromorphic transitions induced by ion migration through a medium is unpredictable resistive switching, arising from the formation of conducting filaments within the dielectric. Random filament formations abruptly alter the devices' analog nature, which is not desired for neuromorphic applications. In this study, we demonstrate that ultrathin dielectric neuromorphic tunnel junctions (NTJs) can effectively address this issue by restricting filament formation, as the tunneling effect dominates in these devices. This mechanism enhances device reliability, even under substantially higher electric fields than those typically associated with intrinsic breakdown. We utilize 1–2 nm thick Al2O3 layers, deposited on fluorine-doped tin oxide via atomic-layer deposition, paired with silver electrodes as the NTJs. Compared with their thicker counterparts, devices with these ultrathin Al2O3 layers exhibited reliable, stable neuromorphic performance. They successfully emulated biological synaptic characteristics, including paired-pulse facilitation/depression, spike-timing-dependent plasticity, and spike-number-dependent plasticity. Furthermore, the NTJs were effectively trained and evaluated for pattern recognition tasks using the Modified National Institute of Standards and Technology database. This research paves the way for the development of highly reliable ultrathin layered devices for neuromorphic applications.
DOI: 10.1116/6.0005462
Authors: J. Jyothish Raj, T. Archana, K. B. Jinesh
Institution: Indian Institute of Space Science and Technology
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
Awards: G3880831940, G8807009784
OpenAlex ID: W4406711294