Fast Electronic Memristors with Improved Retention and Synaptic Functionality Enabled by Tri-TaOx Engineering
Artificial synaptic memristors require fast switching, stable retention, low variability, and linear conductance modulation, but achieving these properties simultaneously is difficult in electronic bipolar resistive switching (e-BRS) devices. In this study, a tri-TaOx memristor is demonstrated by interposing a defective oxygen-deficient TaOx switching layer (Tx, x ≈ 1.8) between insulating near-stoichiometric TaOy interfacial layers (Ty, y ≈ 2.5) at the electrode interfaces. The optimized Ty/Tx/Ty device exhibits forming-free electronic switching, an on/off ratio of 138, a rectification ratio of ∼1.1 × 103, and low cycle-to-cycle and device-to-device variations. Stable retention is maintained in both resistance states, with an on/off ratio greater than 10 over 103 s at room temperature and 120 °C. Under alternating-current pulse operation, the device is programmed by 10 ns SET/RESET pulses and exhibits reliable conductance modulation. The memory window is maintained up to approximately 109 cycles, followed by gradual degradation and final failure near 2 × 109 cycles. The Ty/Tx/Ty tri-TaOx memristor also exhibits gradual potentiation/depression under an optimal identical-pulse condition of ±2.6 V and 400 ns, with low nonlinearity, low-conductance variation, and symmetric spike-timing-dependent plasticity behavior. These results demonstrate that the Ty/Tx/Ty tri-TaOx device outperforms previous e-BRS devices as a high-performance electronic synaptic memristor for both artificial deep neural network accelerators and low-power spiking-driven neuromorphic systems.
Authors
- Cheol Seong Hwang (ORCID: https://orcid.org/0000-0002-6254-9758)
- Kyung Seok Woo (ORCID: https://orcid.org/0000-0001-9184-7255)
- Han Yong Jeong
- Jong Hoon Shin (ORCID: https://orcid.org/0000-0002-5120-2716)
- Janguk Han (ORCID: https://orcid.org/0009-0000-1588-1005)
- Hyungjun Park (ORCID: https://orcid.org/0009-0004-6482-1024)
- Hyun Wook Kim (ORCID: https://orcid.org/0009-0003-3345-9264)
Institutions
- Seoul National University (KR)
- Ulsan National Institute of Science and Technology (KR)
Publication Details
- Journal
- ACS Applied Materials & Interfaces
- Published
- 2026-09-18
- DOI
- https://doi.org/10.1021/acsami.6c14083
- Primary Topic
- Advanced Memory and Neural Computing
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Research Foundation of Korea