Persistent Photoconductivity Enables In-Sensor Binarization and Reservoir Computing in a Single Organic Phototransistor

Abstract

The persistent photoconductivity (PPC) effect in organic phototransistors (OPTs) offers a promising platform for in-sensor reservoir computing (RC). However, the non-uniform wavelength-dependent spectral response of OPTs, while a valuable resource for RC dynamics, poses a challenge for traditional binarization encoding. In this work, we reframe this non-uniformity as a computational resource. We demonstrate that the device’s non-uniform PPC decay dynamics can be directly harnessed to resolve this contradiction. This physical process, which we term time-dependent binarization (TDB), intrinsically performs channel-adaptive thresholding by optimizing a single decay time (Td) instead of multiple current thresholds. Using a PDVT-10 device, we demonstrate the first system that leverages the same PPC dynamics for both sensor-level preprocessing and physical reservoir computing. This integrated system achieved 96.54% accuracy on the Human Protein Atlas image recognition task, surpassing traditional hardware-based and software-based methods. This work redefines physical ‘defects’ as tunable computational resources, paving the way for developing minimalist and energy-efficient in-sensor computing systems.