State Key Laboratory of Silicon Materials & School of Materials Science and Engineering, Zhejiang University, 310027, Hangzhou, PR China.
ZJU-Hangzhou Global Scientific and Technological Innovation Centre, 310027, Hangzhou, PR China.
Nat Commun. 2022 Sep 5;13(1):5216. doi: 10.1038/s41467-022-32884-y.
Silicon is vital for its high abundance, vast production, and perfect compatibility with the well-established CMOS processing industry. Recently, artificially stacked layered 2D structures have gained tremendous attention via fine-tuning properties for electronic devices. This article presents neuromorphic devices based on silicon nanosheets that are chemically exfoliated and surface-modified, enabling self-assembly into hierarchical stacking structures. The device functionality can be switched between a unipolar memristor and a feasibly reset-able synaptic device. The memory function of the device is based on the charge storage in the partially oxidized SiNS stacks followed by the discharge activated by the electric field at the Au-Si Schottky interface, as verified in both experimental and theoretical means. This work further inspired elegant neuromorphic computation models for digit recognition and noise filtration. Ultimately, it brings silicon - the most established semiconductor - back to the forefront for next-generation computations.
硅因其高丰度、大规模生产以及与成熟的 CMOS 处理工业的完美兼容性而至关重要。最近，通过精细调整电子器件的性质，人为堆叠的分层二维结构引起了极大的关注。本文提出了基于化学剥离和表面修饰的硅纳米片的神经形态器件，使其能够自组装成分层堆叠结构。该器件的功能可以在单极忆阻器和可重置的突触器件之间切换。该器件的存储功能基于部分氧化 SiNS 堆叠中的电荷存储，然后通过 Au-Si 肖特基界面的电场激活放电，这在实验和理论手段中都得到了验证。这项工作进一步激发了用于数字识别和噪声过滤的优雅神经形态计算模型。最终，它使硅——最成熟的半导体——重新回到下一代计算的前沿。
