Prediction of metformin adsorption on subsurface sediments based on quantitative experiment and artificial neural network modeling

Abstract

Metformin (MET), a widely employed hypoglycemic pharmaceutical agent, has been frequently detected within groundwater, which has posed a threat to ecosystems and human health. However, the adsorption behavior of MET onto distinct constituent aquitards and aquifers sediments remains shrouded in uncertainty. To reveal the adsorption capacities and mechanisms of diverse sedimentary matrices, we delved into a series of adsorption experiments involving MET on 37 subsurface sediment samples obtained from four boreholes (ranging from 0 to 30 m in depth) in the Jianghan Plain. The quantitative analysis revealed that a majority of the sedimentary compositions consisted of clay minerals (mainly chlorite, montmorillonite and albite), with MET exhibiting considerable variability in across different sediment components (ranging from 15.5 to 489.4 mg/kg). In general, MET adsorption declined in proportion to an increase in quartz composition and depth. Consequently, an artificial neural network model was constructed (R2 = 0.971) to assess the influence of sediment composition on MET adsorption, and thereby elucidating the dominant roles played by chlorite and montmorillonite in this process. Notably, electrostatic attraction, cation exchange, and chemical bonding emerged as the primary mechanisms governing MET adsorption on sediments, particularly those rich in clay minerals. By shedding light on the adsorption mechanism of MET on clay-dominated subsurface sediments, our findings have contributed to a quantitative understanding of MET's adsorption capacity and have highlighted the associated environmental risks.