The fate and transport of pharmaceutical contaminants in aquatic and terrestrial environments are a growing challenge for water quality and ecosystem health. Unlike traditional pollutants, pharmaceuticals stay biologically active even at the low concentrations found in the environment, where they can drive effects ranging from antimicrobial resistance to shifts in aquatic community structure and biodiversity loss. After entering soils and sediments through many interconnected pathways, their onward mobility is largely set by how strongly they bind to clay minerals, one of the dominant sorbents in these settings. Understanding how a pharmaceutical partitions between water and the clay surface is therefore essential for predicting its environmental mobility, persistence, and bioavailability.
Paper I examines how the pH-dependent s peciation of the antibioticciprofloxacin (CIP) affects its adsorption on sodium-montmorillonite (Na-MMT) clay. All three forms adsorb favourably, but binding strength rises steeply fromanionic to cationic (CIP− < CIP+/− < CIP+). The weakly bound CIP− thereforel eaches readily, whereas the more strongly bound CIP+/− and CIP+ stay fixed to the clay. The two weaker forms sampled several orientations, favouring the poses that minimise electrostatic repulsion from the surface.
Manuscripts II and III form a two-part companion series that extends PaperI. Both extend the Paper I system beyond its idealised single-species picture in two directions, adding background electrolytes and resolving the surface-charge heterogeneity created by clustered octahedral Al3+/Mg2+ substitutions. Manuscript II builds the structural picture of how cation valency (M+ and M2+) and substitution-site clustering steer CIP adsorption. Cations bind strongly at theAl3+/Mg2+ sites and draw coordination water with them, leaving low-density surface patches that offset the cost of displacing water during adsorption. These cations bind in two modes: K+ partially dehydrates into the ditrigonal cavities as inner-sphere complexes, whereas Na+, Mg2+, and Ca2+ stay outer-sphere, with divalent cations bridging CIP− through its carboxylate group. The surface also orders the innermost water layer into a quasi-hexagonal dipole pattern mirroring the ditrigonal siloxane network. Manuscript III quantifies these consequences for the three charge states across the four salts. Raising ionic strength enhances CIP− adsorption by screening the surface charge, with divalent cations (M2+) more effective than monovalent ones (M+), whereas CIP+ adsorption stays largely electrolyte-insensitive. The deepest wells coincide with the highest substitution density, echoing Manuscript II’s structural finding in the free-energy landscape. To handle the many interacting factors, we developed a neural network (NN) framework that learns smooth, differentiable 3D PMF surfaces from the MD trajectories and 1D PMFs, coupling three influences on adsorption:
the location r, the ring tilt of CIP θ, and the type and number of coordinating cations (ncat). It reproduced the 1D PMFs and helped explain features within them. Manuscript V returns to the CIP-Na-MMT system of Paper I, and Manuscripts II and III, but shifts the focus to method, asking how strongly the computed adsorption energy depends on the force-fields used to model it. The MMT force-field set the depth of the PMF global minimum and, more subtly, its shape, while the water model shifted the minimum’s location relative to the surface. The computed adsorption energy is thus as much a property of the chosen models as of the CIP-clay system itself.
Manuscript IV broadens the free-energy methodology from a single antibiotic to 20 pharmaceuticals (15 neutral and 5 charged), testing whether the octanol water partition coefficient (logPo/w) predicts adsorption to Na-MMT. Across the 15 neutral compounds, the correlation challenges the view that “morehydrophobic = stronger adsorption”. We therefore introduce a clay-specific descriptor from free-energy perturbation (FEP): a partition coefficient logPw/c,ads between bulk water and the clay-adsorbed state. Of all descriptors tested, only logPw/c, ads correlated with adsorption free energy, whereas neither logPo/w nor a clay-pore descriptor without surface contact did.
Bringing these findings together, this Ph.D. project builds a molecular-scale picture of how pharmaceuticals partition between water and clay in soils. Paper I established that pH-dependent speciation sets the adsorption strength on Namontmorillonite. Manuscripts II and III showed that pore-water cations and surface-charge heterogeneity reshape that binding both structurally and thermodynamically, and Manuscript IV found that clay adsorption across diverse pharmaceuticals is governed by clay-specific interactions rather than general hydrophobicity. Manuscript V then showed that the computed energies themselves depend strongly on the chosen force-fields. The project therefore demonstrates the power of MD simulation for predicting contaminant retentionand mobility, while making clear that quantitative, standard-state conclusions must be read in light of the sensitivity of the underlying models.