Machine learning-based prediction and mechanistic insight into PFAS adsorption on carbon-based materials

Area of Science:

• Environmental Chemistry
• Materials Science
• Computational Chemistry

Background:

• Carbon-based materials show promise for removing per- and polyfluoroalkyl substances (PFAS) from the environment via adsorption.
• Elucidating PFAS adsorption mechanisms is complex due to varied material properties, diverse PFAS structures, and differing environmental conditions.
• Experimental methods alone struggle to fully explain the intricate interactions between carbon materials and PFAS.

Purpose of the Study:

• To develop and optimize a machine learning model for predicting carbon-based material adsorption of PFAS.
• To identify key factors governing PFAS adsorption onto carbon materials.
• To enhance understanding of PFAS-carbon interactions for improved environmental remediation strategies.

Main Methods:

• Development and optimization of a gradient boosting decision tree (GBDT) model.
• Utilizing machine learning for systematic prediction of PFAS adsorption performance.
• Employing SHAP and partial dependence plots for model interpretation and factor identification.

Main Results:

• The GBDT model achieved high predictive accuracy (R² = 0.96, RMSE = 0.02) on the test dataset.
• Environmental conditions were the most significant contributors to adsorption, followed by material properties and PFAS molecular features.
• Key influential factors identified include solution pH, fluorine atom count, temperature, and carbon material pore structure.

Conclusions:

• Machine learning, integrated with environmental chemistry, offers a powerful approach to understanding PFAS-carbon interactions.
• Electrostatic interactions and hydrophobic-hydrophilic balance are likely dominant adsorption mechanisms.
• Findings provide valuable insights for environmental risk assessment and the design of effective PFAS remediation materials.