•연구자: 물리학과 오홍석, 화학공학과 강범구
•발표일: 2026.09
•DOI: https://doi.org/10.1016/j.cap.2026.06.005
•Sehyeon Jeong et al., Current Applied Physics (Q3), Volume 89, Pages 151-161 (2026)
•Abstract
Physical reservoir computing (PRC) is emerging as a new paradigm for neuromorphic processing of dynamic information. Here, the inherent nonlinear dynamics and fading memory of a physical system map input signals into a higher-dimensional space, enabling complex tasks only through the trained linear readout. Nanodevices based on nanomaterials can be an ideal platform for PRC, because the quantum confinement effect, structural stochasticity, and high surface to volume ratio collectively provide rich nonlinear responses and tunable memory characteristics. We categorize the strategies for achieving the state richness, or the dimensionality of the mapping, into ‘in materia’ computing and device-level PRC, based on the degree of controllability and stochasticity. Across 0D, 1D, and 2D nanomaterial systems, we systematically review representative implementations. This review provides a general landscape of how nanodevices can be utilized for PRC, discusses current limitations, and presents an outlook toward energy-efficient neuromorphic computing.