In Situ Ion-Exchange-Mediated Dynamic Nanointerface Reconfiguration for Polarity-Switchable Photoelectrochemical–Photothermal Biosensing

Conventional photoelectrochemical (PEC) systems typically rely on static semiconductor interfaces that merely translate biomolecular recognition into amplitude modulation of the pre-existing photocurrent, thus limiting their analytical robustness and self-validation capabilities. Herein, we report in situ ion-exchange-mediated dynamic nanointerface engineering to fabricate an analyte-programmed reconfigurable PEC-photothermal (PT) dual-mode biosensor for the carcinoembryonic antigen (CEA) assay. The Z-scheme heterostructure of a plasmonic Au nanoparticle (NP) layer-supported BiOI/AgI/indium tin oxide glass (BiOI/AgI/Au NPs/ITO) is constructed as the initial photoelectrode. In the sensing process, a CEA recognition event directly reshapes the phase composition of BiOI/AgI, charge-transfer behavior, and photoresponse polarity. The synergistic integration of the BiOI/AgI heterostructure and the close-packed layer of plasmonic Au NPs substantially enhance the light absorption, charge separation, and interfacial charge transfer, making BiOI/AgI/Au NPs/ITO a highly suitable photoelectrode for the PEC mode of the PEC-PT biosensor. This analyte-driven phase evolution reconstructs the semiconductor composition, band alignment, and charge-transfer pathways to switch the photocurrent polarity from cathodic to anodic. In addition, a photothermal imaging mode of the PEC-PT biosensor is triggered by regulating the Bi2S3 formation for intrinsic cross-validation. By integrating an artificial neural network (ANN)-assisted analysis, PEC-PT biosensors enable robust multimodal quantification and improved analytical reliability, achieving > 99.8% prediction accuracy in real-sample analysis. This is the first example of integrating a dynamic nanointerface-driven PEC-PT dual-mode biosensor with artificial intelligence (AI)-enhanced multimodal analysis for CEA detection and provides a dynamic nanointerface-driven strategy for intelligent multimode biosensing.