Artificial synapses with ideal functionalities are essential
in
hardware neural networks to allow for energy-efficient analog computing.
Electrolyte-gated transistors (EGTs) are promising candidates for
artificial synaptic devices due to their low voltage operations supported
by large specific capacitances of electrolyte gate insulators (EGIs).
We investigated the synapse transistor employing an In–Ga–Zn–O
channel and a Li-doped ZrO2 (LZO) EGI so as to improve
the short-term plasticity (STP) and long-term potentiation (LTP).
The LZO EGIs showed distinct differences in characteristics depending
on the Li doping concentration, and we adopted the optimum doping
concentration of 10%. Based on the strong electric double layer effect
secured from the LZO, we successfully demonstrated excellent synaptic
operations with gradual modulations of excitatory synaptic plasticity
with variations in amplitude, width, and number of applied pulse spikes.
The introduction of the LZO EGI was verified to improve typical short-term
plasticity such as paired-pulse facilitation. Furthermore, by minutely
controlling the pulse spike conditions, the conversion to LTP from
STP was clearly accomplished while implementing the anti-Hebbian spike
timing-dependent plasticity. Finally, the array configuration of synaptic
devices, which is essential for realizing neuromorphic computing,
was also demonstrated. In a 3 × 3 array architecture, the weighted-sum
operation was well emulated to assign multilevels in seven states
with the pulse width modulation scheme.