posted on 2024-01-10, 15:04authored byMing Li, Mingjun Li, Haoqun An, Jun Seop An, Pengyu Gu, Dae Hun Kim, Kwan Kyu Park, Tae Whan Kim
The metallic conductive filament
(CF) model, which serves as an
important conduction mechanism for realizing synaptic functions in
electronic devices, has gained recognition and is the subject of extensive
research. However, the formation of CFs within the active layer is
plagued by issues such as uncontrolled and random growth, which severely
impacts the stability of the devices. Therefore, controlling the growth
of CFs and improving the performance of the devices have become the
focus of that research. Herein, a synaptic device based on polyvinylpyrrolidone
(PVP)/graphene oxide quantum dot (GO QD) nanocomposites is proposed.
Doping GO QDs in the PVP provides a large number of active centers
for the reduction of silver ions, which allows, to a certain extent,
the growth of CFs to be controlled. Because of this, the proposed
device can simulate a variety of synaptic functions, including the
transition from long-term potentiation to long-term depression, paired-pulse
facilitation, post-tetanic potentiation, transition from short-term
memory to long-term memory, and the behavior of the “learning
experience”. Furthermore, after being bent repeatedly, the
devices were still able to simulate multiple synaptic functions accurately.
Finally, the devices achieved a high recognition accuracy rate of
89.39% in the learning and inference tests, producing clear digit
classification results.