Embedding Operant Conditioning Functions in a Four-Terminal Memristor
We present an all-in-one operant conditioning (OC) implementation in a single four-terminal memristor, requiring no external circuits or logic. Unlike conventional two-terminal devices, our multi-terminal architecture enables true co-localization of learning functions, supporting both heterosynaptic plasticity and synaptic competition. The operant training process can be scheduled via controlled dopant-ion migration, precisely modulated by a bulk-switching mechanism of our device. This process is visualized via electrochromism, enhancing the interpretability of brain-inspired computation in our device. Our work offers a compact, higher-order functional hardware platform for neuromorphic computing, paving the way toward large-scale neuromorphic systems with embedded learning functionality.
Authors
- Yusuke Hayashi (ORCID: https://orcid.org/0000-0001-5672-1497)
- Tetsuya Tohei (ORCID: https://orcid.org/0000-0002-4113-2566)
- Ryohei Yamamoto (ORCID: https://orcid.org/0000-0001-9789-3690)
- Zhuo Diao (ORCID: https://orcid.org/0000-0001-5358-444X)
- Akira Sakai
- Hayato Kamitsuna
- Zijie Meng
Institutions
- Toneyama National Hospital (JP)
- The University of Osaka (JP)
Publication Details
- Journal
- ACS Applied Electronic Materials
- Published
- 2026-06-27
- DOI
- https://doi.org/10.1021/acsaelm.6c00586
- Primary Topic
- Advanced Memory and Neural Computing
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- Kyoto Technoscience Center
- Japan Society for the Promotion of Science