POS5-1669
Pressure-Light Dual-sensory Ferroelectric Phototransistor Synapse Enabling Pressure-dependent Modulation of Optical Synaptic Plasticity for Sensory-Adaptive Reservoir Computing
Topic
S5. Polymers for Electronics and Photonics
When and Where
Sep 29, 2026
08:30 - 09:30
Room 301 (Grand Ballroom)
Session Chairs
Hae Jung SON
Boseok KANG
Presenter(s)
Seungjun Park (Yonsei university, Department of material science & engineering)
Co-Author(s)
Abstract
Humans perceive and process external information through a multisensory biological system, yet replicating such integrated sensory-cognitive functionality in artificial hardware remains a fundamental challenge. Here, we present a photonic-tactile responsive dual-sensory synapse(PT-DSS) enabling the pressure-dependent modulation of optical synaptic plasticity. Our sensory synapse is based on a top-gate bottom-contact ferroelectric phototransistor consisting of a photoresponsive bulk heterojunction p-n semiconductor channel, a ferroelectric polymer insulator with a pressure sensitive hemispherical gate. When tactile stimuli (0–30 kPa) are imposed via a pressure-sensitive gate electrode, the synaptic plasticity induced by optical stimuli (532 nm) in a photoresponsive semiconductor channel is successfully controlled, with the fast and slow decay time constants (τ1 and τ2) decreasing from 4.67 s to 0.06 s and from 445 s to 19 s, respectively, and their coefficient ratio (A1/A2) increasing from 0.36 to 3.8. This adaptive modulation of synaptic plasticity in response to diverse environmental variables constitutes a key capability for physical reservoir computing (PRC). Critically, this work introduces the first sensory-adaptive reservoir (SAR), in which decay parameters are modulated directly by pressure stimuli rather than electrical voltage, achieving 7 distinct pressure-dependent reservoir states and overcoming the limited temporal tunability prevalent in existing PRC systems. The platform's efficacy is validated through human action recognition tasks at varying speeds (0.5-8 Hz) using the Weizmann dataset (5 action classes), where the SAR achieves a recognition accuracy of 90.1%, outperforming its fixed reservoir counterpart (80.1%). These results establish the PT-DSS as a promising multisensory neuromorphic platform for next-generation adaptive artificial intelligence hardware.
