Physics-Guided Active Learning Enables Energy-Filtering Flexible Thermoelectric Composites for Self-Powered Wound...

Thermoelectric stimulation offers a self-powered intervention for accelerated wound healing by converting physiological thermal gradients into sustained micro-electric fields. However, the development of flexible thermoelectric materials is constrained by the coupled dependence of electrical conductivity and the Seebeck coefficient, which empirical screening alone is inefficient for exploring the large number of possible interfaces. We developed a small-data, literature-derived, physics-guided screening and design framework that integrates Bayesian optimization with eXtreme Gradient Boosting (XGBoost) regression and subsequent experimental validation. Using 18 physics-informed descriptors selected by recursive feature elimination, the framework predicts interfacial energy barriers within the literature-derived data domain with a test R² of 0.87, enabling targeted barrier engineering that partially decouples the thermoelectric parameters. The model and practical screening criteria were used to select five composite systems for synthesis and experimental evaluation, supporting the applicability of the algorithm. Among the experimentally validated systems, poly(3,4-ethylenedioxythiophene): poly(styrene sulfonate)/eutectic gallium-indium (PEDOT:PSS/EGaIn) flexible composite exhibited a measured interfacial energy barrier of 0.24 eV, close to the predicted effective filtering range, and achieved a Seebeck coefficient of 40.0 µV K^-1. Under a mild physiological-to-ambient temperature gradient of 25 K, the flexible thermoelectric device generated an open-circuit voltage of 8.80 mV. In the scratch assay, the stimulated group showed 34.5% closure compared with 13.9% in the untreated control. In the rat wound model, the thermoelectric group reached near-complete closure approximately two days earlier than the sham and untreated groups. This data-driven approach provides a physical framework for screening energy-filtering composites and evaluating their potential in self-powered bioelectronic therapeutics.