A cross-layer 2D photodetection framework for standardized readout and signal fusion

Beyond improving device responsivity, a central challenge in photodetectors based on two-dimensional (2D) materials is to resolve the information-interface mismatch and delayed system translation caused by trap-governed persistent photoconductivity (PPC). Here we present a cross-layer engineering route oriented towards practical implementation that integrates defect-mechanism analysis, non-stationary signal definition, dynamic distortion compensation and readout-interface engineering, thereby establishing a conversion pathway from physical response to system output in 2D MoS2, WSe2 and PtSe2 photodetectors. In this context, drift, hysteresis and fatigue accumulation in time-varying photoresponses are no longer treated merely as non-ideal perturbations, but are reformulated as state-dependent information that can be defined, compensated and interfaced. Through interface restructuring and stable readout, nW-level state-dependent photocurrents are systematically mapped onto a mixed-signal readout interface. We further extend this strategy to array-level unified output and multisensor fusion, with its frequency-guided representation evaluated in a machine-vision task. Our results show that the history-dependent nonlinear response governed by PPC is not simply a local device non-ideality, but a state-encoded signal linking defect dynamics, readout pathways and system-level tasks. This work provides an engineering route for translating 2D photodetectors from device-level studies to system-level applications. Persistent photoconductivity usually complicates the readout of 2D photodetectors, limiting the stability and speed of the signal representation and processing. Here, the authors introduce a cross-layer signal engineering approach to analyse and stabilize the dynamic outputs of 2D MoS2, WSe2 and PtSe2 photodetectors, showing its extension to multi-sensor image recognition tasks.