EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Ultralow‐Thermal‐Threshold CsPbBr 3 Quantum Dot Memristor for Thermal‐Aware In‐Memory Computing With Physical Pruning

Shuaidi Wu, Yuezhong Lu, Wenkang Zhong, Qingwei Shi, Song He, Qinghua Sun, Xiaofeng Wang, Baochang Cheng, Jie Zhao

Advanced Computing · 2026-12 · 1 (1)

DOI: 10.1002/adco.70009 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Advanced Computing, online erschienen am 07.10.2026. Verfasst von Shuaidi Wu und 8 weiteren Autor:innen. Die Arbeit nennt 72 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China, Natural Science Foundation of Jiangxi Province.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Thermal reliability is a critical bottleneck for compute‐in‐memory (CIM) based on resistive switching devices, as Joule heating degrades switching ratio, introduces weight uncertainty, and compromises accuracy. Here we report a CsPbBr 3 perovskite quantum dot thermal‐threshold memristor (TTM) with an ultralow writing threshold of 50°C, an on/off ratio of 10 5 –10 6 , a read voltage of 0.05 V, and a reversible 0.28 V erase capability, with retention exceeding 104 s. Leveraging these characteristics, we develop a saliency‐driven smart mapping algorithm that allocates high‐importance weights to cold array regions and low‐importance weights to hotspots. System‐level simulations on a 2048 × 2048 virtual array show that smart mapping maintains 78.63% accuracy under 30.4% device flip, a 64.3 percentage point advantage over random mapping, while achieving 30% power reduction. This initiates a positive feedback loop where reduced power lowers Joule heating, suppressing further failures and enabling self‐adaptive thermal regulation. A 0.28 V erase pulse fully restores accuracy to baseline, confirming reversibility. This work establishes perovskite quantum dot thermal‐threshold memristors as a promising platform for thermal‐aware CIM, offering a pathway toward reliable, self‐adaptive, and energy‐efficient accelerators in thermally constrained environments.

Förderung

  • National Natural Science Foundation of China
  • Natural Science Foundation of Jiangxi Province

Zitieren

APA

Wu, S., Lu, Y., Zhong, W., Shi, Q., He, S., Sun, Q., Wang, X., Cheng, B., & Zhao, J. (2026). Ultralow‐Thermal‐Threshold CsPbBr 3 Quantum Dot Memristor for Thermal‐Aware In‐Memory Computing With Physical Pruning. Advanced Computing, 1(1). https://doi.org/10.1002/adco.70009

Vancouver

Wu S, Lu Y, Zhong W, Shi Q, He S, Sun Q, et al. Ultralow‐Thermal‐Threshold CsPbBr 3 Quantum Dot Memristor for Thermal‐Aware In‐Memory Computing With Physical Pruning. Advanced Computing. 2026;1(1). doi:10.1002/adco.70009

BibTeX
@article{Wu2026,
  author = {Wu, Shuaidi and Lu, Yuezhong and Zhong, Wenkang and Shi, Qingwei and He, Song and Sun, Qinghua and Wang, Xiaofeng and Cheng, Baochang and Zhao, Jie},
  title = {Ultralow‐Thermal‐Threshold CsPbBr 3 Quantum Dot Memristor for Thermal‐Aware In‐Memory Computing With Physical Pruning},
  journal = {Advanced Computing},
  year = {2026},
  volume = {1},
  number = {1},
  doi = {10.1002/adco.70009}
}
Kostenlos starten →