A multiscale computational framework for the design and screening of graphene oxide in wastewater remediation

Freshwater pollution caused by Emerging Contaminants (ECs) represents a major global challenge due to the increasing release of anthropogenic compounds such as pharmaceuticals, PFAS, pesticides, microplastics, and industrial chemicals into aquatic environments. Conventional wastewater treatments are often ineffective at removing these pollutants, motivating the development of advanced adsorption materials. Among these, graphene oxide (GO) and its derivatives have emerged as promising candidates because of their high adsorption capacity and tunable surface chemistry. This doctoral thesis developed a multiscale computational framework to investigate interactions between ECs and GO-based materials, combining virtual screening, molecular dynamics (MD), and thermodynamic analyses. An in-house database of approximately 2000 ECs was constructed, including stereochemically accurate 3D structures and environmentally relevant protonation states. In parallel, scalable graphene oxide models with different oxidation levels were generated based on experimental data. A large-scale virtual screening campaign, followed by all-atom MD simulations, produced over 86000 adsorption poses and approximately 29 microseconds of cumulative simulation time. Binding affinities were evaluated using MM-GBSA calculations combined with entropy corrections through Normal Mode analysis. Results demonstrated that adsorption is mainly governed by molecular orientation on the heterogeneous GO surface rather than stereochemistry. Hydrophobic and π-stacking interactions were identified as the dominant driving forces, favoring polycyclic aromatic hydrocarbons and plasticizers, while polar and anionic compounds showed weaker affinities due to desolvation penalties. Reduced graphene oxide exhibited even stronger adsorption, particularly for neutral and positively charged compounds, although the study highlighted the limitations of representing GO through a single structural microstate. Consequently, the work emphasizes the need for ensemble-based models to achieve reliable affinity predictions. Finally, a Martini 3 coarse-grained library of PFAS molecules was developed to simulate complex contaminant mixtures in the presence of graphene surfaces, enabling the investigation of competitive adsorption phenomena under environmentally realistic conditions at scales inaccessible to all-atom simulations alone.