Effective coagulation is critical to the production of safe, potable drinking water, but variations in the chemical composition of source water can present challenges in achieving targeted contaminant removal and predicting coagulation outcomes. A critical literature review describes factors affecting the hydrolysis reactions of metal salt coagulants and the resulting precipitates. Properties of two key contaminants, turbidity and natural organic matter (NOM), are explored in the context of removal during coagulation, and the influence of co-occurring ions is described. While it is apparent that NOM character determines the minimum achievable organic carbon residual, the effects of water quality—including pH, NOM character and concentration, and concentrations of synergistic and competitive ions—on overall coagulation efficacy and NOM removal may be underestimated. An experimental research plan was devised to investigate the influence of water quality in coagulation and provide data to support the development of a predictive coagulation model.

NOM is capable of interfering with ferric iron hydrolysis and influencing the size, morphology, and identity of precipitates. Conversely, calcium is known to increase the size and aggregation of Fe3+ precipitates and increase surface potential, leading to more effective coagulation and widening the pH range of treatment. Experiments and modeling were conducted to investigate the significance of the Fe/NOM ratio and the presence of calcium in coagulation. At the high Fe/NOM ratio, sufficient or excess ferric hydroxide was available for NOM removal, and coagulation proceeded according to expectations based upon the literature. At the low Fe/NOM ratio, however, NOM inhibited Fe3+ hydrolysis, reduced zeta potential, and suppressed the formation of filterable Fe flocs, leading to interference with effective NOM removal. In these dose-limited systems, equilibrating NOM with 1 mM Ca2+ prior to dosing with ferric chloride coagulant increased the extent of Fe3+ hydrolysis, increased zeta potential, decreased the fraction of colloidal Fe, and improved NOM removal. In dose-limited systems without calcium, complexation of Fe species by NOM appears to be the mechanism by which coagulation is disrupted. In systems with calcium, data and modeling indicate that calcium complexation by NOM neutralizes some of the negative organic charge and minimizes Fe complexation, making Fe hydrolysis species available for growth and effective coagulation.

Experiments were conducted to investigate the influence of aqueous silica and pH on the removal of natural organic matter (NOM) by coagulation with ferric chloride. Samples with preformed ferric hydroxide were also compared to samples coagulated in situ to assess the role of coprecipitation. The moderate (10 mg/L) and high (50 mg/L) SiO2 concentrations both demonstrated interference with NOM removal at pH 6.5-7.5. In turn, NOM at 2 mg/L as DOC interfered with silica sorption at the moderate silica level and in samples with preformed ferric hydroxide at the high silica level. The combination of NOM and high silica led to decreases in DOC sorption and unexpected increases in silica sorption in the coprecipitated samples. The fraction of colloidal Fe passing a 0.45-μm filter also increased in the coprecipitated samples with both NOM and high silica. It is hypothesized that the combination of NOM and high silica synergistically interfered with Fe precipitation and particle growth processes, with NOM having the greater effect at lower pH and shorter reaction times, and silica exerting greater influence at higher pH and longer reaction time. Direct competition for surface sites and electrostatic repulsion were also influential.

An overall goal for this research was the development of a quantitative coagulation model. Previous attempts to model coagulation have been limited by the inherent complexities of simultaneously predicting ligand sorption, metal complexation, floc surface charge, and particle removal. A diffuse layer (DLM) surface complexation model was formulated to simultaneously predict sorption of NOM and other key species, including silica, calcium, and carbonate alkalinity. Predictions of surface potential were used to estimate zeta potential and resulting regimes of effective aggregation and turbidity removal. The model provided good predictive ability for data from bench-scale experiments with synthetic water and jar tests of nine U.S. source waters. Under most conditions, the model provides excellent capability for predicting NOM sorption, calcium sorption, and particle destabilization and adequate capability for predicting silica sorption. Model simulations of hypothetical scenarios and experimental results help to explain practical observations from the literature. The DLM can be optimized to site-specific conditions and expanded to include sorption of additional species, such as arsenic.