This work demonstrated the versatile SiNSs-based neuromorphic device, which extended Si technology to next-generation computation like SNN, a crucial step towards cognitive integration in artificial intelligence systems. We investigated the unipolar memristor-like behavior based on the NDR and device capacitance, as well as the synaptic response with effective reset ability based on the high capacitance originated from the layered structure and the rectification behavior of SiNSs-Au junctions. Then, we analytically described characteristics, including LTM, STDP, which were essential for neuromorphic calculations. Finally, we demonstrated an SNN inspired by device behaviors, which was proved effective for digit recognition and noise filtration.
In addition, we envision other emerging research areas to encourage in-depth investigation for the potential of SiNSs in SNN and their integration with Si-based electronics. First, photoemission of SiNSs demonstrated the possibility for optoelectronic. Second, given the compatibility with the solution process on soft substrate like PET, our SiNSs demonstrate the possiblility for flexible device fabrication. Third, the CMOS-compatible SNN arrays are attractive for fabrication with existing Si technology. These exciting topics require further study to bridge the gap between technologies of artificial intelligence and the well-established Si industry.
https://www.nature.com/articles/s41467-022-32884-y
Nature Communications volume 13, Article number: 5216 (2022)
Please sign in or register for FREE
If you are a registered user on Research Communities by Springer Nature, please sign in