Full Electrical Switching of a Freestanding Ferrimagnetic Metal for Energy-Efficient Bipolar Neuromorphic Computing
Flexible electronics and neuromorphic computing face key challenges in material integration and function retention. In particular, freestanding membranes suffer from slow sacrificial layer removal and interfacial strain, while neuromorphic hardware often relies on area-intensive dual-device schemes for bipolar synaptic weights. Here, we present a universal strategy based on water-soluble Sr<sub>4</sub>Al<sub>2</sub>O<sub>7</sub> sacrificial layers, enabling the rapid release of freestanding ferrimagnetic metal membranes, which exhibit deterministic spin-orbit torque switching characteristics with well-preserved perpendicular magnetic anisotropy and are potential for next-generation ultrafast information technology. Extending this approach, we realize single-device ferrimagnetic synapses exhibiting intrinsic bipolar resistive switching. When implemented in a ResNet-18 architecture, these devices achieve 92% accuracy on CIFAR-10─comparable to floating-point software models─while halving device counts relative to differential-pair implementations. These results establish a scalable platform linking flexible spintronics with compact, high-performance neuromorphic systems, offering foundational advances for next-generation electronics and brain-inspired hardware.
Authors
- Shaoxuan Zhang (ORCID: https://orcid.org/0000-0003-0366-9992)
- Qiong Wu (ORCID: https://orcid.org/0000-0002-6156-1077)
- Ziang Meng
- Yu Liu (ORCID: https://orcid.org/0000-0002-4191-5974)
- Peixin Qin (ORCID: https://orcid.org/0009-0009-8889-3205)
- Xiaoning Wang (ORCID: https://orcid.org/0000-0002-5305-0824)
- Li Liu (ORCID: https://orcid.org/0000-0002-5641-446X)
- Zhiqi Liu (ORCID: https://orcid.org/0000-0001-7538-4879)
- Guojian Zhao (ORCID: https://orcid.org/0009-0004-7868-5094)
- Xiang Wang (ORCID: https://orcid.org/0009-0005-8461-9082)
- Zhiyuan Duan (ORCID: https://orcid.org/0009-0007-0138-7437)
- Xiaobo She
- Qinghua Zhang (ORCID: https://orcid.org/0000-0001-5591-9776)
- Hongyu Chen (ORCID: https://orcid.org/0000-0002-5325-9249)
Institutions
- Chinese Academy of Sciences (CN)
- Beijing University of Technology (CN)
- Nanjing University of Posts and Telecommunications (CN)
- FZU ‒ Institute of Physics of the Academy of Sciences of the Czech Republic (CZ)
- Kementerian Pendidikan Malaysia (MY)
- Beihang University (CN)
Publication Details
- Journal
- Nano Letters
- Published
- 2025-09-15
- DOI
- https://doi.org/10.1021/acs.nanolett.5c04033
- Citations
- 1
- Primary Topic
- Advanced Memory and Neural Computing
- Type
- article
- Field-Weighted Citation Impact
- 0.73