@article{Huang2026, 
author = {Yongqing Huang and Runhua Zhang and Jiaxiang Chen and Liuqi Cheng and Zhihao Li and Chao Hu and Jiahong Yang and Ning Li and Binbin Zhang and Qi Da and Xiangcheng Li and Zhenrong Zhang and Qijun Sun and Jinran Yu and Chaofan Zhang},
title = {A CuCrP2S6-based lateral memristor for in-materia reservoir computing in temporal information processing},
year = {2026},
journal = {Nano Research},
keywords = {reservoir computing, CuCrP2S6 lateral memristor, Cu+ ion migration, in-materia computing, temporal information processing},
url = {https://www.sciopen.com/article/10.26599/NR.2026.94908805},
doi = {10.26599/NR.2026.94908805},
abstract = {Silicon-based accelerators deliver high computational precision through the von Neumann architecture, yet incur substantial energy costs due to frequent data movement and discrete logic switching. In contrast, in-materia reservoir computing harnesses the intrinsic nonlinear dynamics of materials to enable energy-efficient temporal information processing, offering a promising route toward neuromorphic hardware. Here, we report a two-terminal lateral memristor based on 2D ferroelectric CuCrP₂S₆, where electric-field-driven Cu+ ion migration yields continuously tunable nonlinear conductance, short-term memory, and rich relaxation dynamics—properties that closely match the physical requirements of reservoir computing. On this basis, pattern recognition and chaotic prediction were implemented. On the MNIST handwritten digit benchmark, the system achieves 88.91% accuracy. Furthermore, the reservoir achieved normalized root-mean-square errors (NRMSE) of 0.02732 and 0.3716 for autonomous prediction of the Hénon map (steps 500–550) and the Mackey-Glass (steps 500–600) time series, respectively. These results establish CuCrP₂S₆ lateral memristors as an in-materia reservoir platform for temporal information processing and highlight their potential for advancing post-Moore neuromorphic computing systems.}
}