Year: 2025
Type: article
Abstract: Brain-inspired machine learning is now ubiquitous in daily life. However, understanding the mechanisms of the brain and developing optimal systems for neuromorphic computing remain open challenges. This work targets these obstacles by developing a semiconductor chip platform for interdisciplinary neuroscience and machine learning research. Silicon nanosheets (SiNSs) possess two-dimensional properties, including hierarchical stacking, a quantum-confined quasi-direct band gap, and a large surface-to-volume ratio that induces indirect recombination. Exploiting these properties, we develop SiNS devices with multi-lifetime leaky-integration responses. Constructing a device-based prototype chip for a spiking neural network (SNN), we demonstrate on-chip simulations of neuron postsynaptic potential and achieve pattern recognition. Further, the interpretable quasi-rectified linear unit activation function of the SNN is found, demonstrating the intrinsic connection mapping artificial neural networks to brain functionalities.
Source: Cell Reports Physical Science
Authors: Chenhao Wang, Xinyi Xu, Anzhe Chen, Feng Tian, Zhenyi Ni, Xin Yuan, Bodepudi Srikrishna Chanakya, Yang Xu, Wei Sun, Deren Yang
Institutions: Zhejiang University, Westlake University
Language: English
Subfield: Electrical and Electronic Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $5,620
OpenAlex ID: W4415424040