Dataset. Accelerate Flash Removal of PFAS from Soil by Human-guided Bayesian Optimization and Interpretable Machine Learning

The study hypothesizes that coupling human-in-the-loop decision-making with Bayesian optimization and interpretable neural representations enables rapid identification of optimal FJH parameters and molecular-level insights into PFAS degradation mechanisms. Specifically, we posit that functional group composition, rather than chain length, primarily governs the defluorination efficiency under FJH conditions. The dataset contains 80 experimental records across four representative PFAS compounds. Each record combines: Molecular descriptors: molecular weight, number of fluorine atoms, mean C–F bond energy, polarity indices, and functional group identity. Experimental parameters: applied voltage, resistance, pulse duration, temperature rise rate, and calculated energy input.