Where Mn²⁺ _(aq) is found in water supplies, filter media may naturally develop surface coatings bearing MnO_x(s). These may absorb Mn²⁺ _(aq), and in the presence of oxidant, sorbed Mn²⁺* is oxidized to MnO_x(s), regenerating sorption capacity. The filter accomplishes Mn²⁺ _(aq) removal, a process called the "natural greensand effect".

Characterization of naturally coated media showed variation in coating composition and structure. With thicknesses from 1 - 125 μm, primary coating constituents were Al and Mn, with incorporation of minor amounts of Fe, Cu, and Si and trace elements. "Growth ring" features in coating cross-section corresponding to compositional variation were characterized by SEM, electron microprobe, and energy-dispersive x-ray analysis (EDS). Media surface areas of 2 - 135 m² g⁻¹ land microporosity of 15 - 533 cm³ kg⁻¹ were linearly related to extractable Mn content. Diatom remains found in coatings suggest a key role for coating deposition in filtration phenomena. Atomic force microscopy found surface self-similarity over 10 nm - 10 μm. X-ray photoelectron spectroscopy (XPS) confirmed heterogeneous surface composition including C, Al, Si, and Fe.

A method to rapidly deposit up to 4 mg g⁻¹ Mn on media was developed, employing sequential batch and recycle reactors. Mn(IV) was the surface species found by XPS analysis. The Freundlich isotherm described Mn²⁺ sorption on this and the naturally coated media; sorption capacity increased between pH 6.0 and 7.5, and was reduced by [Ca²⁺] = 60 mg L⁻¹. The global Mn²⁺ oxidation rates for all coated media at pH 7.5 were 0.008 - 0.11 mg Mn²⁺ g⁻¹ hr⁻¹: rates increased with flow and decreased with pH.

A numerical process model for sorption and oxidation of Mn²⁺ _(aq) was calibrated with short bed absorber and differential reactor columns. The Freundlich isotherm, film transport, internal diffusion, and hydrodynamic dispersion were included, with sorption capacity apportioned into kinetically available and unavailable sites. The model performed well in calibration, predicting dynamic system response across a range of flow, pH, [Ca²⁺], and reactant levels. Model performance in validation was less satisfactory, probably due to experimental difficulties and the sensitivity of process performance on recent coating history and media regeneration status.