A Resistive Random‐Access Memory With Reduced Current–Voltage Nonlinearity for Reliable Analog Vector–Matrix Multiplication
Advanced Intelligent Systems · 15.11.2026
DOI: 10.1002/aisy.70636 ↗Kurz gefasst
Fachartikel in der Zeitschrift Advanced Intelligent Systems, online erschienen am 15.11.2026. Verfasst von Dohyeong Kim und 4 weiteren Autor:innen. Die Arbeit nennt 45 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Suppressing current–voltage ( I – V ) nonlinearity in resistive random‐access memory (RRAM) is essential for minimizing computational errors in hardware neural networks. In this study, we propose a Pt/Al/TiO x /Al 2 O 3 /AlN/Pt RRAM with reduced nonlinearity in both high‐ and low‐resistance states (HRS and LRS). While the fabricated device maintains reliable multilevel memory characteristics, including pulse endurance over 6 × 10 5 individual programming pulses and multilevel retention for 10 5 s at 120 °C, it exhibits enhanced linearity within the analog read voltage range owing to trap‐assisted tunneling in the HRS and ohmic behavior in the LRS. Extended to the array level, the impact of the proposed device on analog vector–matrix multiplication (VMM) operations was experimentally measured. By incorporating the measurement data into system‐level simulations of fully‐connected network and VGGNet‐11 architectures, the hardware system achieved performance close to the ideal software baseline. Furthermore, comparative simulations based on different conduction mechanisms demonstrate the importance of device I – V linearity for accurate VMM operations. These results demonstrate the potential of the proposed device for reliable hardware neural‐network implementation.
Zitieren
APA
Kim, D., Kwon, S., Na, H., Kim, S., & Hwang, S. (2026). A Resistive Random‐Access Memory With Reduced Current–Voltage Nonlinearity for Reliable Analog Vector–Matrix Multiplication. Advanced Intelligent Systems. https://doi.org/10.1002/aisy.70636
Vancouver
Kim D, Kwon S, Na H, Kim S, Hwang S. A Resistive Random‐Access Memory With Reduced Current–Voltage Nonlinearity for Reliable Analog Vector–Matrix Multiplication. Advanced Intelligent Systems. 2026. doi:10.1002/aisy.70636
BibTeX
@article{Kim2026,
author = {Kim, Dohyeong and Kwon, Seongbin and Na, Hyeokjun and Kim, Sungjoon and Hwang, Sungmin},
title = {A Resistive Random‐Access Memory With Reduced Current–Voltage Nonlinearity for Reliable Analog Vector–Matrix Multiplication},
journal = {Advanced Intelligent Systems},
year = {2026},
doi = {10.1002/aisy.70636}
}
EUVIMED