Field-free spin–orbit torque devices for logic and neural network applications
Robust field-free spin–orbit torque (SOT) switching plays an essential role in realizing practical SOT-based magnetic memories and logic devices. Leveraging the strong interlayer Dzyaloshinskii–Moriya interaction, field-free SOT switching is achieved with a threshold switching current density of 5.4 × 1010 A/m2. By exploiting the advantage of separated read–write paths, we demonstrate fundamental logic operations such as AND, OR, and NOT in a dual-channel SOT Hall bar device. We further explore the application of a digital SOT magnetic random access memory (MRAM) design in addressing the issue of the low ON/OFF ratio of the conventional spin-transfer torque MRAM, highlighting its potential for neural network applications. Our simulation results show that using digital SOT MRAM can achieve a classification accuracy of 92.2% for the MNIST dataset, outperforms the analog SOT MRAM case (<10%), and presents remarkable improvements in power and time efficiencies compared to the SRAM case. These findings establish digital SOT MRAM as a frontrunner for edge artificial intelligence applications and offline inference.
Authors
- Chi‐Feng Pai (ORCID: https://orcid.org/0000-0001-6723-8302)
- Po‐Chuan Wang (ORCID: https://orcid.org/0009-0002-4353-2974)
- Chun‐Yi Lin (ORCID: https://orcid.org/0009-0005-7246-3535)
- Jui-Yu Hsieh (ORCID: https://orcid.org/0009-0004-9056-1828)
- Chia-Chin Tsai (ORCID: https://orcid.org/0009-0009-8078-360X)
Institutions
- National Taiwan University (TW)
Publication Details
- Journal
- APL Machine Learning
- Published
- 2024-11-21
- DOI
- https://doi.org/10.1063/5.0226135
- Citations
- 3
- Primary Topic
- Advanced Memory and Neural Computing
- Type
- article
- Field-Weighted Citation Impact
- 0.63
Funders
- National Taiwan University
- National Science and Technology Council