Low-Power, Electrochemically Tunable Graphene Synapses for Neuromorphic Computing

Journal: Advanced Materials

Authors: M. T. Sharbati and F. Xiong et al.

Affiliation: The University of Pittsburgh (USA)

Publication date: 2018.07.23

Summarized by Inyeol Yun


– Background
v. The neural network in a human brain has ≈1011 neurons. Each neuron is typically   connected to ≈5,000 to 10,000 other neurons through synapse.
v. Synapse plasticity is the ability of synapses to strengthen or weaken over time, in   response to increases or decreases in their activity (Wikipedia). It related to memory   function.
v. Typical artificial synapses cannot mimic the analog behaviors of biological synapses.
v. In this paper, authors demonstrate the synapse plasticity using “nano-battery”   technology. (Fig. 1)


<Fig. 1>

– Structure & Fabrication
v. Graphene layers – Solid electrolyte (LiClO4 in poly(ethylene oxide) (PEO) – Lithium   iron phosphate (LFP) (Fig. 2)


<Fig. 2>

 v. Metal contacts (80 nm of Cu) were defined by e-beam lithography and deposited   through e-beam evaporation.
v. Real image (Fig. 3)


<Fig. 3>

– Results
v. Raman spectroscopy shows weakening of bond between graphene layers as   increasing Li-ion. (Fig. 4)


<Fig. 4>

 v. Resistance change (Figure. 5)


<Fig. 5>