Fig. 3: Characterization and performance of the artificial synaptic device.
a, b Exploded-view schematic illustration (a) and photograph (b) of the synaptic transistor with dual gates and nanoflake-adsorbed channel. c High-resolution TEM image of a nanoflake with high crystallinity. The measured lattice spacing (d) of 0.27 nm corresponds to the (100) plane of MoS2. The TEM image in the inset (scale bar: 20 nm) shows the nanoflake’s morphology. d AFM image of the semiconductor channel revealing that the nanoflakes are distributed on the smooth surface (root-mean-square roughness Rq = 0.33 nm) of the metal oxide film. e Spike-number-dependent plasticity evaluated by applying spike trains with different spike numbers (N = 10, 20, 30, 40, 50). f History-dependent plasticity evaluated by applying spike trains with changing frequencies (f = 25, 33.3, 50, 33.3 Hz). g Cycle-to-cycle variation evaluated by repeatedly applying potentiation/inhibition spike trains (containing 50 positive and 50 negative pulses at 10 Hz). h Modulation of the sensory memory by applying sustaining spikes (reduced amplitude, same frequency) after stimuli spikes. i Gating effect of the dual gates. j Connection and signal flow of the sensor-device networks in the neuromorphic antennal sensory system imitate the “labeled-line” model of neuronal pathway and signal encoding in biological sensory neurons. Curves in (h) are offset vertically for clarity. SA1 and SA2: slowly adapting spikes from Antenna #1 and #2; FA1 and FA2: fast-adapting spikes from Antenna #1 and #2.
