DocumentCode
41533
Title
Low-Temperature Fabrication of Spiking Soma Circuits Using Nanocrystalline-Silicon TFTs
Author
Subramaniam, Anand ; Cantley, K.D. ; Stiegler, Harvey J. ; Chapman, R.A. ; Vogel, Eric M.
Author_Institution
Dept. of Electr. Eng., Univ. of Texas at Dallas, Richardson, TX, USA
Volume
24
Issue
9
fYear
2013
fDate
Sept. 2013
Firstpage
1466
Lastpage
1472
Abstract
Spiking neuron circuits consisting of ambipolar nanocrystalline-silicon (nc-Si) thin-film transistors (TFTs) have been fabricated using low temperature processing conditions (maximum of 250 °C) that allow the use of flexible substrates. These circuits display behaviors commonly observed in biological neurons such as millisecond spike duration, nonlinear frequency-current relationship, and spike frequency adaptation. The maximum drive capacity of a simple soma circuit was estimated to be approximately 9200 synapses. The effect of bias stress-induced threshold voltage degradation of component nc-Si TFTs on the spike frequency of soma circuits is explored. The measured power consumption of the circuit when spiking at 100 Hz was approximately 12 nW. Finally, the power consumption of the soma circuits at different spiking conditions and its implications on a large-scale system are discussed. The fabricated circuits can be employed as part of a compact multilayer learning network.
Keywords
cryogenic electronics; large-scale systems; multilayer perceptrons; nanostructured materials; neural chips; power consumption; substrates; thin film transistors; ambipolar nanocrystalline-silicon thin-film transistors; bias stress-induced threshold voltage degradation; biological neurons; circuit power consumption; large-scale system; low temperature processing conditions; low-temperature spiking Soma circuit fabrication; multilayer learning network; nanocrystalline-silicon TFT; nonlinear frequency-current relationship; spiking neuron circuits; Frequency–current curve; nanocrystalline silicon (nc-Si); soma circuits; thin-film transistor (TFT);
fLanguage
English
Journal_Title
Neural Networks and Learning Systems, IEEE Transactions on
Publisher
ieee
ISSN
2162-237X
Type
jour
DOI
10.1109/TNNLS.2013.2256926
Filename
6510475
Link To Document