What governs phosphorus adsorption on biochar-supported layered double hydroxides? A data-driven analysis of parameter interactions

Abstract

The rational design of advanced functional materials, such as adsorbents for tackling water eutrophication, is often challenged by inconsistent experimental findings and a lack of generalizable design principles. This difficulty stems from the complex nature of these materials, where performance is dictated by the non-linear interplay of numerous synthesis and operational parameters that traditional trial-and-error methods cannot adequately resolve. This study presents a systematic explainable AI (XAI) framework to move beyond simple prediction and instead deconstruct the complex structure-performance relationships in biochar-supported layered double hydroxide composites (LDH@BC). Leveraging a 2057-point literature dataset, our high-accuracy CatBoost model (R² = 0.956) serves as a reliable foundation for this in-depth analysis. The analysis identified a set of optimal operating conditions, including a solution pH of 4–7 and an adsorbent dosage of 1.0–2.0 g·L⁻¹. Crucially, the model also revealed a key design principle: a low LDHs/BC ratio (0–1.0), combined with sufficient adsorbent dosages, maximizes adsorption capacity by promoting optimal nanoparticle dispersion. This work provides not only specific, data-driven guidelines for the rational design of LDH@BC adsorbents but also demonstrates a transferable XAI framework for accelerating the transition from empirical screening to intelligent, data-informed design for other complex functional materials.