| About this Abstract |
| Meeting |
MS&T23: Materials Science & Technology
|
| Symposium
|
Advances in Dielectric Materials and Electronic Devices
|
| Presentation Title |
CMOS-compatible Oxide Memristors Based on SiO2 for Adaptive Neuromorphic Computing |
| Author(s) |
Fei Qin, Han Wook Song, Sunghwan Lee |
| On-Site Speaker (Planned) |
Fei Qin |
| Abstract Scope |
Memristors have emerged as a popular candidate for overcoming the von Neumann architecture bottleneck in the pursuit of neuromorphic computing due to their unique properties, such as low power consumption and integrated memory-processing functions. In this work, a nanoscale dielectric thin film of SiO2 is sputtered to serve as the switching layer in a memristor. I-V sweep measurements are performed to investigate the bipolar switching behaviors. Capacitance measurements are conducted to characterize the inherent properties of the memristor in its high-resistance state, while electrochemical impedance spectroscopy is used to elucidate the switching mechanism between high and low-resistance states. Pulse electrical measurements reveal that SiO2-based memristors possess multi-conductance capabilities, enabling them to mimic synaptic behavior. Lastly, a neural network simulation is conducted using the characteristics of SiO2-based memristors, demonstrating their image recognition abilities when applied to the Fashion Modified National Institute of Standards and Technology database. |