Junction-engineered halide perovskite neuronal arrays with on-chip physical time surfaces for spatiotemporal computing
Event-driven neuromorphic hardware requires reliable neuronal arrays that retain recent sensory history over a controlled temporal window. Here, we report on-chip physical time surfaces in two-dimensional Dion–Jacobson halide perovskite neuronal arrays. Nanoscale sidewall junctions confine ionic migration, enabling low-voltage volatile switching with reduced variability and no external current-compliance circuitry. Junction widths are scaled down to 5 nm, and processing below 100 °C enables fabrication on 4-inch wafers. The films retain crystallographic stability for more than two years. Residual ionic states support temporal integration and history-dependent firing. A 32 × 32 crossbar array represents changes in object position and apparent size through self-decaying conductance traces. Computational evaluation shows that controlled relaxation improves vehicle-motion prediction by preserving recent motion history and limiting interference from earlier activity. The results link nanoscale ionic confinement to reliable neuronal arrays and physical temporal processing for spatiotemporal computing.
Authors
- Ho Won Jang (ORCID: https://orcid.org/0000-0002-6952-7359)
- Yongwoo Kwon (ORCID: https://orcid.org/0000-0002-9779-2512)
- Hyeon Ji Lee (ORCID: https://orcid.org/0009-0002-3299-4604)
- Yuan Li (ORCID: https://orcid.org/0000-0001-7452-1149)
- Kwan Woo Song
- Seung Ju Kim (ORCID: https://orcid.org/0009-0000-0405-803X)
- Sung Woo Park
- Jinwoo Oh
- Youngmin Kim
Institutions
- University of Southern California (US)
- Seoul National University (KR)
- Huazhong University of Science and Technology (CN)
- Hongik University (KR)
Publication Details
- Journal
- Research Square
- Published
- 2026-11-18
- DOI
- https://doi.org/10.21203/rs.3.rs-11442592/v1
- Primary Topic
- Advanced Memory and Neural Computing
- Type
- preprint