Novel Solution‐Processed Fe2O3/WS2 Hybrid Nanocomposite Dynamic Memristor for Advanced Power Efficiency in Neuromorphic Computing
Abstract Non‐volatile memory (NVM) based neuromorphic computing, which is inspired by the human brain, is a compelling paradigm in regard to building energy‐efficient computing hardware that is tailored for artificial intelligence. However, the current state of the art NVMs are facing challenges wit...
Saved in:
| Main Authors: | , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2025-05-01
|
| Series: | Advanced Science |
| Subjects: | |
| Online Access: | https://doi.org/10.1002/advs.202408133 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Non‐volatile memory (NVM) based neuromorphic computing, which is inspired by the human brain, is a compelling paradigm in regard to building energy‐efficient computing hardware that is tailored for artificial intelligence. However, the current state of the art NVMs are facing challenges with low operating voltages, energy efficiencies, and high densities in order to meet the new computing system beyond Moore's law. It is therefore necessary to develop novel hybrid materials with controlled compositional dynamics is crucial for initiating memristor devices capable of low‐power operations. This study validates the effectiveness of Ag/Fe90W10/Pt hybrid nanocomposite memristor devices, demonstrating superior performance including ultra‐low voltage operation, high stability, reproducibility, exceptional endurance (105 cycles), environmental resilience, and low energy consumption of 0.072 pJ. Moreover, the memristor exhibits the ability to emulate essential biological synaptic mechanisms. The resistive switching phenomenon is primarily attributed to the controlled filament formation along unique heterophase grain boundaries. Furthermore, the hybrid nanocomposite synaptic device achieved an image recognition accuracy of 94.3% in Artificial Neural Network (ANN) simulations by using the Modified National Institute of Standards and Technology (MNIST) dataset. These results imply that the device's performance has promising implications for facilitating efficient neuromorphic architectures in the future. |
|---|---|
| ISSN: | 2198-3844 |