Artificial Intelligence-Assisted Ion-Association Design Decouples Linearity and Dynamic Range in Organic Electrochemical Synapses
Abstract Organic electrochemical synaptic transistors (OESTs) are promising neuromorphic devices, but simultaneously achieving highly linear analog weight updates and a wide dynamic range remains fundamentally limited by a severe trade-off in uncontrolled interface doping kinetics. Although multicomponent electrolytes offer a pathway to tune ion behavior, a predictive framework capable of overcoming this strict trade-off remains unexplored due to the highly nonlinear nature of interionic interactions. Here, we introduce an AI-assisted ion-association design strategy utilizing a binary electrolyte architecture to break through the linearity–dynamic range trade-off. GPR-assisted regression analysis maps complex electrolyte formulations to synaptic nonlinearity, identifying an optimized NaCl·Na-TFSI binary configuration. Density functional theory (DFT) calculations suggest distinct intrinsic ion-association tendencies between Na+–Cl– and Na+–TFSI– pairs, consistent with the observed doping and relaxation behaviors: strongly associated Na+–Cl– pairs suppress abrupt doping to optimize linearity, reducing nonlinearity by up to ∼6-fold compared with single-anion controls, while weakly associated Na+–TFSI– pairs facilitate ion retention to expand the dynamic range (Gmax/Gmin ∼ 103). This balanced dual-anion transport enables linear, extensible, and high-dynamic-range multistate conductance modulation. Consequently, the optimized OESTs exhibit improved LTP/D linearity, reproducible cycling, and a 94.88% MNIST recognition accuracy. These results establish predictive binary ion-association design as a general framework for controlling synaptic weight modulation in organic neuromorphic devices.
Authors
- Geun Yeol Bae (ORCID: https://orcid.org/0000-0003-3391-3183)
- Hyoik Jang (ORCID: https://orcid.org/0009-0003-7328-0015)
- Eunho Lee (ORCID: https://orcid.org/0000-0002-8564-6999)
- Donghwa Lee (ORCID: https://orcid.org/0009-0005-0721-7039)
- Junho Sung (ORCID: https://orcid.org/0009-0002-0513-2072)
- Myeongjin An (ORCID: https://orcid.org/0009-0007-1841-0218)
- Yoohyeon Jang
Institutions
- Kumoh National Institute of Technology (KR)
- Seoul National University of Science and Technology (KR)
Publication Details
- Journal
- Journal of the American Chemical Society
- Published
- 2026-11-13
- DOI
- https://doi.org/10.1021/jacs.6c15115
- Primary Topic
- Advanced Memory and Neural Computing
- Type
- article
- Field-Weighted Citation Impact
- 0.00
Funders
- National Research Foundation of Korea