Exploring the adsorption sites and mechanism of biochar towards tetracycline using machine learning

In the past, exploration of the adsorption sites and mechanism of adsorbent towards adsorbate largely rely on adsorption experiments and physico-chemical characterization. The emergence of artificial intelligence techniques, such as machine learning, has provided new approaches for studying adsorption sites and mechanisms. However, most current studies using machine learning to investigate the adsorption performance and mechanisms of materials have not thoroughly extracted or revealed the role and contribution of chemical bonds and chemical states in the adsorption process. In this study, by extracting and separating material XPS analysis data, we exploratively apply machine learning methods to predict and investigate the adsorption sites and mechanisms of antibiotic adsorption on biochar. Specifically, this work employed 4 machine learning models for over 344 datasets to analysis the correlations between the key functional groups (including N–6, π–π and C–O/C–N) and adsorption ability. The results showed N–6, π–π and C–O/C–N structures are key functional groups contributing to adsorption. Among them, π–π interaction have the most pronounced effect, suggesting that this mechanism may play a dominant role in adsorption process. This approach not only helps to identify functional groups that enhance adsorption but also links them to specific mechanisms, offering valuable insights for the rational design of biochar-based adsorbents.