세미나

SMART Metallization for Neuromorphic Edge Computing

  • 일시 2022-04-28 15:00 ~ 18:00
  • 장소 온라인 개최
  • 연사 연한울 교수
  • 소속 광주과학기술원(GIST) 신소재공학과
In the age of information explosion, combination of artificial intelligence (Ai) and big data has revolutionized almost everything in our world. Furthermore, an innovative hardware such as brain inspired Ai chips (i.e., neuromorphic processors) and skin inspired conformable sensor array (i.e., electronic skins) has been intensively developed to realize edge” computing, physically on the edge objects(e.g., human beings) mimicking the human nervous system. Such neuromorphic edges would lead to tremendous benefits over cloud computing including energy consumption, data latency, and data security However, serious reliability issue s exist on the emerging hardware including 1) poor controllability in the neuromorphic process ing and 2) conformability degradation of the electronic skins, impeding commercialization of th e new hardware Here , I describe my novel metallurgical tech nologies, termed SMART metallization , to resolve the challenges in neuromorphic edge computing hardware. First, in terms of memristor synapse array, metal alloy ed electrode can lead to precision synaptic weight control [1]. I will discuss how I designed met al alloy based on the thermodynamic and kinetic interactions of metal s and the memristive medium. Second, in terms of electronic skins, sweat pore inspired perforation engineering on metal interconnects and other electronic components drives long term conformable sensing on human skin over a week, even under profuse sweating in the daily lives [2]. I will discuss how the structural engineering could give rise to reliable health monitoring from the skin. 
Acknowledgments This work was supported by MIT-IBM, Samsung, and Amorepacific Corporation.
References
[1] Yeon, H. et al., Nature Nanotechnology 15, 574 (2020). [2] Yeon, H. et al., Science Advances 7, eabg8459(2021)