Vernadite is a nanocrystalline manganese oxide, which controls the fate of many trace elements in soils and sediments through sorption and oxidative-degradation mechanisms. This exceptional reactivity directly results from its crystal structure, which may however evolve upon contact with redox-sensitive species. Understanding these changes is a prerequisite to predict and model the geochemical cycle of trace elements in the environment. Here, the structural and morphological modifications affecting synthetic nanocrystalline vernadite (δ-MnO2) upon contact with increasing concentrations of Mn2+ were investigated using wet chemistry, synchrotron X-ray diffraction and transmission electron microscopy. Fresh δ-MnO2 crystals had an Mn oxidation state of 3.94 ± 0.05 and a ∼10 Å layer-to-layer distance. Crystal size was ∼10 nm in the layer plane, and ∼1 nm perpendicular to that. Upon contact with aqueous Mn2+ under anoxic conditions, δ-MnO2 crystals underwent several morphological and mineral evolutions, starting with the stacking, perpendicular to the layer plane, of δ-MnO2 crystals to form crystals ∼10 nm × 2 nm which were then subjected to oriented aggregation both along and perpendicular to the layer plane to form lath-like crystals with dimensions of ∼100 nm × 20 nm. Finally, these laths stacked perpendicular to the layer plane to form synthetic feitknechtite (β-MnOOH) crystals with sizes up to ∼100 nm × 500 nm when the Mn2+ loading reached 31.9 mmol g−1. Structural transformation from δ-MnO2 to synthetic feitknechtite was detected at Mn2+ loading equal to or higher than 3.27 mmol g−1.

These mechanisms are likely to influence the geochemical fate of trace elements in natural settings where Mn2+ is abundant. Firstly, the systematic increase in crystal size with increasing Mn2+ loading may impact the sorption capacity of vernadite and feitknechtite by reducing the density of reactive edge sites. Secondly, the fate of trace elements initially sorbed at the vernadite surface is unclear, as they could either be released in solution or incorporated into the feitknechtite lattice.

Manganese (Mn) oxides are ubiquitous in surficial environments, where they play a key role in the geochemical cycles of major and trace elements, as they are highly reactive towards nutrients (e.g., Xin et al., 2016), metals (e.g., Halbach, 1986; Bellanca et al., 1996; Manceau et al., 2004, 2007, 2014; Marcus et al., 2004; Boonfueng et al., 2006; Fuller & Bargar, 2014), actinides (e.g., Kunzendorf & Friedrich, 1976; Koppi et al., 1996; Duff et al., 1999) and organic matter and organic molecules (e.g., Cheney et al., 1996; Nasser et al., 2000; Barrett & McBride, 2005; Chang Chien et al., 2009; Johnson et al., 2015). Vernadite, the nanocrystalline and turbostratic analogue to birnessite (Giovanoli, 1980), is one of most abundant (Taylor et al., 1964; Taylor, 1968; Ross et al., 1976; Birnie & Paterson, 1991) and reactive Mn oxides, especially in suboxic aqueous systems such as the upper soil. The vernadite structure consists in the turbostratic stacking of layers of (Mn4+O6)8− octahedra connected through their edges and separated from each other by hydrated interlayer cations (e.g., Giovanoli, 1980; Chukhrov et al., 1985; Jurgensen et al., 2004; Villalobos et al., 2006; Grangeon et al., 2008). In natural settings, vernadite mainly has hexagonal layer symmetry (e.g., Marcus et al., 2004; Manceau et al., 2007, 2014; Bargar et al., 2009). These layers may contain vacancies and/or isomorphic substitutions of Mn4+ by cations of lower valence (e.g., Mn3+, Ni2+, Co3+; Manceau et al., 1997, 2014; Peacock & Sherman, 2007; Lanson et al., 2008; Peña et al., 2010; Grangeon et al., 2014) which, together with the minute size of the layers (5–10 nm, e.g., Hochella et al., 2005; Bargar et al., 2009), are responsible for the exceptional reactivity of vernadite. The possible contact with redox-sensitive species such as Mn2+ can induce structural changes whose nature and extent depend on solution pH and Mn2+ concentration. The first step of reaction between aqueous Mn2+ and layer Mn4+, hereafter referred to as Mn2+(aq) and Mn4+(s), respectively, is (Elzinga, 2016):
where Mn3+(s) is layer Mn3+. The 2Mn3+(s) formed may further disproportionate, returning the system to its initial state. Alternatively, they may accumulate in the structure and induce a change in layer symmetry, from hexagonal to orthogonal, as observed experimentally at pH 9 when the molar ratio of Mn2+ to δ-MnO2 is lower than 0.05 (Zhao et al., 2016). At lower pH and/or at higher Mn2+ concentration, a transformation to synthetic feitknechtite (β-MnOOH) is observed, following the reaction:

Feitknechtite, whose structure is built of layers of (Mn3+O6)9− octahedra, may upon ageing convert to manganite (γ-MnOOH) at pH 7.0–7.5 or to hausmannite (Mn3O4) at pH > 8 (Elzinga, 2011; Lefkowitz et al., 2013; Lefkowitz & Elzinga, 2015).

The interplay between these structural transformations certainly influences the geochemical cycle of trace elements by modifying the reactivity of Mn oxides through a change in the nature and density of local crystal charges, but also by modifying the fate of trace elements initially adsorbed at the vernadite surface which, by analogy with the phyllomanganate-to-tectomanganate transformation, could either be incorporated in the structure during transformation or released to solution (Atkins et al., 2014, 2016; Grangeon et al., 2015).

The present study aimed at contributing to a better understanding of the vernadite-to-feitknechtite transformation, and in particular at elucidating the reaction mechanisms at the crystal scale, using wet chemistry, X-ray diffraction (XRD) and transmission electron microscopy (TEM).

2.1. Sample synthesis, sorption experiments and chemical analyses

The δ-MnO2 phase was synthesized using a redox method described elsewhere (Villalobos et al., 2003). The average oxidation state (AOS) of Mn was determined by using a potentiometric method (Grangeon et al., 2012). After synthesis, samples were Na-saturated through ten cycles of centrifugation, supernatant removal and resuspension in a 1 M NaCl solution. A last step of three similar cycles with de-ionized water instead of 1 M NaCl was then performed. The obtained δ-MnO2 paste was immediately freeze-dried, and the resulting powder was introduced in a N2-filled glove box where it was left for ∼7 d. Eight aliquots of ∼20–40 mg powder each were then introduced in previously weighted dialysis membranes. Filled membranes were weighted, filled with a solution containing 0.1 M NaCl and MnCl2 at a concentration ranging 0–0.04 M, sealed, and introduced in 50 mL polypropylene tubes, to which 40.5–45 mL of the solution introduced in the dialysis membrane were added. Initial solution pH was 6.5. Tubes were sealed and left to agitate for 24 h. The dialysis membrane was then separated from the solution whose final pH ranged between 6.5 and 5.8. The solution was acidified to pH ∼3 with HNO3 and analyzed for its Mn2+ content using an ICP-AES (Horiba-Jobin Yvon Ultima 2). The magnitude of Mn2+ retention by δ-MnO2 was quantified using the RD factor (Tournassat et al., 2013) defined as:
where Cinit is the initial concentration of Mn2+, Caqu the Mn2+ concentration in solution after equilibration, and RSL is the solid-to-liquid ratio in the experiment (in kg L−1).

The dialysis membrane was opened and the solid separated by filtration (0.01 μm cut-off) before being washed with bi-distilled water. An aliquot was freeze-dried for TEM analysis, whereas the remaining was sealed in a polyimide capillary for XRD analysis in the wet state within 48 h. Samples were labelled MndBi-XX, where XX is the Mn2+ loading (in mmol Mn2+ per g of dry δ-MnO2).

2.2. Synchrotron X-ray diffraction

The XRD experiments were carried out at station CRISTAL (SOLEIL synchrotron in Orsay, France), using an energy of 28.44 keV (λ = 0.436 Å) and an XPad hybrid pixel detector. Data were recorded over the 1.2–124.5° 2θ range with a total collection time of 30 min and processed with a specific software (Ounsy et al., 2013). To ease comparison with previous studies (e.g., Grangeon et al., 2015, 2017), diffraction data will be expressed relative to the scattering vector q, where q = [4 × π × sin(θ)]/λ.

2.3. Transmission electron microscopy

The TEM experiments were carried out using a Philips CM20 microscope operated at 200 kV. Prior to observation, samples were embedded in epoxy resin, left to polymerize for 48 h in the dark and cut with an ultramicrotome (Reichert-Jung Ultra-cut E) equipped with a diamond knife. The ∼80 nm thick sections were picked up on lacey carbon films loaded on Cu grids. In addition, to assess possible preparation-induced artefacts (use of an ultramicrotome), a sample identical to MndBi-0.5 (Table 1) was prepared according to the above described protocol, filtered and re-suspended in alcohol. A drop of the obtained suspension was deposited on a Cu grid. Fast-Fourier transform (FFT) analysis of the micrographs was done with ImageJ (Schneider et al., 2012).

3.1. Determination of Mn2+ retention coefficient

An increase in Mn2+(aq) resulted in a decrease in the Mn AOS of the final solids, suggesting retention of Mn2+ on δ-MnO2 (Table 1). The RD values (Table 1) compared well with previously reported values obtained after 8 d of equilibration (Fig. 1). The evolution of RD as a function of Mn2+ concentration and pH (Murray, 1975) has, in some cases, been interpreted in terms of adsorption and quantified in surface-complexation models (Appelo & Postma, 1999; Tonkin et al., 2004). Data interpreted with these models had a Mn2+ loading ranging from 1 to 12 Mn2+ per nm2, depending on the dataset and on the specific surface area considered (value measured by the authors or 600 m2 g−1 as recommended by Tonkin et al., 2004). Such loading roughly corresponds to MndBi-1.5 (1.5 Mn2+ per nm2 assuming a specific surface area of 600 m2 g−1), which means that this sample and MndBi-0.5 could have been interpreted in the frame of these models. For other samples, the density of Mn2+ at the surface of δ-MnO2 would be unrealistically high (up to 32 Mn2+ per nm2 for MndBi-31.9), and interaction of Mn2+ with δ-MnO2 most likely resulted in a structure transformation, which was studied by XRD.
Fig. 1

Evolution of the Mn2+ retention coefficient as a function of Mn2+ concentration. Patterns are indicative of the experiment pH: circles and triangles pointing up, experiments conducted at pH 6 and pH 7, respectively; squares and triangles pointing down, pH 7.5; diamonds, pH 8; triangles pointing left, pH 8.5 (Elzinga, 2011; Lefkowitz et al., 2013). Published data were collected with a contact time of 8 d. Present data (pH 6.5; 1 d contact time) are shown as triangles pointing right. (Online version in colour.)

Fig. 1

Evolution of the Mn2+ retention coefficient as a function of Mn2+ concentration. Patterns are indicative of the experiment pH: circles and triangles pointing up, experiments conducted at pH 6 and pH 7, respectively; squares and triangles pointing down, pH 7.5; diamonds, pH 8; triangles pointing left, pH 8.5 (Elzinga, 2011; Lefkowitz et al., 2013). Published data were collected with a contact time of 8 d. Present data (pH 6.5; 1 d contact time) are shown as triangles pointing right. (Online version in colour.)

3.2. Mineralogical evolution

Two sets of reflections could be observed in the XRD pattern of MndBi-0 (Fig. 2), which was not in contact with Mn2+, both being consistent with a nanocrystalline turbostratic phyllomanganate structure. The first set corresponded to broad and asymmetric reflections at 2.58 Å−1, 4.42 Å−1 and 5.11 Å−1 (d-spacing of 2.44 Å, 1.42 Å and 1.23 Å). The reflection at 4.42 Å−1 was shown to be indicative of layer symmetry: approximately symmetric when layer symmetry is hexagonal, but split in two distinct maxima when layer symmetry is orthogonal (Drits et al., 2007). Here, it was symmetric despite the presence of a shoulder on the low-q side, related to a basal reflection (see below), and layer symmetry was hexagonal. For consistency with previous studies (e.g., Lanson et al., 2000, 2008; Webb et al., 2005; Drits et al., 2007; Grangeon et al., 2010, 2012; Zhu et al., 2010; Manceau et al., 2013), bands were indexed with an orthogonal symmetry system (a = √3 × b, γ = 90°) and the three maxima were indexed as (11, 20), (31, 02) and (22, 40) hk band, respectively. Reflections of the second set were sharp and symmetric and formed a rational series of 00l reflections at 0.62 Å−1, 1.24 Å−1 and 1.85 Å−1 (10.15 Å, 5.08 Å and 3.39 Å). The layer-to-layer distance, inferred from the position of the 003 at 1.85 Å−1, was 10.17 Å. When increasing Mn2+ loading up to 1.5 mmol g−1 (MndBi-1.5), the intensity of the 00l reflections decreased, but the position remained essentially unchanged and hk bands were unaffected. The shoulder on the low-q side of the (31, 02) band weakened, supporting its indexation as the 006 reflection. Note the presence of a weak peak at ∼1.36 Å−1 (∼4.6 Å) in the XRD pattern of MndBi-1.5, which could be attributed to synthetic feitknechtite (see below). Weakening of 00l reflections can be interpreted as a loss of parallelism between adjacent layers and was attributed to the incorporation of Mn3+(s) according to the comproportionation mechanism described above. Indeed, Mn3+ octahedra are distorted by the Jahn–Teller effect and their presence within the octahedral layer induces strains that ultimately lead to layer kinking (Webb et al., 2005; Atkins et al., 2014; Grangeon et al., 2014) and subsequent loss of parallelism between layers. In addition to hk bands of δ-MnO2, the XRD pattern of MndBi-3.3 (3.3 Mn2+ per nm2) had additional reflections, e.g., at 1.36 Å−1, 2.34 Å−1, 2.64 Å−1 and 3.40 Å−1 (4.63 Å, 2.69 Å, 2.38 Å, and 1.85 Å), assigned to synthetic feitknechtite (Bricker, 1965). As the latter reflections became more intense with increasing Mn2+ loading, up to 31.90 mmol g−1 (MndBi-31.9), the proportion of synthetic feitknechtite increased with Mn2+ loading.
Fig. 2

From top to bottom, XRD patterns of MndBi-0, MndBi-0.5, MndBi-1.5, MndBi-3.3, MndBi 28.1 and MndBi-31.9. Dashed lines and dashed-dotted lines point out to the presence of diffraction maxima of δ-MnO2 and synthetic feitknechtite, respectively.

Fig. 2

From top to bottom, XRD patterns of MndBi-0, MndBi-0.5, MndBi-1.5, MndBi-3.3, MndBi 28.1 and MndBi-31.9. Dashed lines and dashed-dotted lines point out to the presence of diffraction maxima of δ-MnO2 and synthetic feitknechtite, respectively.

Such an evolution was consistent with previous studies (Elzinga, 2011, 2016; Lefkowitz et al., 2013; Lefkowitz & Elzinga, 2015). In the following section, the mechanisms of transformation were studied at the crystal scale, hypothesizing that an increased concentration of Mn2+ led to an increased conversion of δ-MnO2 to synthetic feitknechtite without significant dissolution of δ-MnO2. In other words, it was hypothesized that the structural and morphological evolution of samples following their contact with a given concentration of Mn2+(aq) underwent intermediate recrystallization steps similar to those observed during post-mortem analysis of samples in contact with lower concentrations of Mn2+(aq). This hypothesis is supported by previous TEM observations, from which it was concluded that there was a lack of significant dissolution/reprecipitation during the recrystallization of δ-MnO2 when in contact with aqueous MnSO4 (Tu et al., 1994). Absence of δ-MnO2 dissolution is further substantiated by the study of Elzinga & Kustka (2015) showing that the first steps of transformation involves solely comproportionation and disproportionation reactions induced by interfacial electron transfer between adsorbed Mn2+ and Mn4+(s). Experimental validation of this hypothesis would require performing a kinetic experiment at high Mn2+(aq) concentration to check whether recrystallization steps observed as a function of Mn2+ concentration are similar to those occurring as a function of time. Such a procedure was applied to the study of the δ-MnO2 to synthetic todorokite transformation (Atkins et al., 2014, 2016), but was not possible in the present case because of reaction kinetics. The RD values obtained in the present study after 1 d of contact time were identical to those reported previously for a contact time of 8 d (Fig. 1). Steady-state was thus reached in less than 1 d of contact time, consistent with previous observations showing that feitknechtite is formed in less than 20 min (Johnson et al., 2016) and that sorption of Mn2+ at the δ-MnO2 surface is completed within less than 1 s (Fendorf et al., 1993).

3.3. Morphological evolution during δ-MnO2 to synthetic feitknechtite transformation

Sample MndBi-0 consisted of crystals ∼10 nm in their largest dimension (presumably the layer plane) and ∼1 nm perpendicular (Fig. 3), henceforth denominated type-1 crystals. The crystal size in the layer plane was in good qualitative agreement with the XRD data, in which broad hk bands could be observed, indicative of crystallites ∼5–10 nm in the layer plane (Lanson et al., 2008; Grangeon et al., 2012; Manceau et al., 2013). On the other hand, the small crystal size perpendicular to the layer plane contrasted with the sharp 00l reflections. This apparent contradiction could be concealed assuming that TEM sample preparation procedure has delaminated δ-MnO2 crystals.
Fig. 3

Transmission electron microscope (TEM) observation of type 1 crystals from MndBi-0.

Fig. 3

Transmission electron microscope (TEM) observation of type 1 crystals from MndBi-0.

Observation of MndBi-0.5 evinced that a Mn2+ loading as low at 0.49 mmol g−1 led to the formation of crystals with contrasting morphologies and sizes, ranging ∼10–200 nm in the layer plane (Fig. 4a). In addition to type-1 crystals, three types of such crystals were identified:

  • Type-2 crystals had a size of ∼10 nm in the layer plane, ∼2–3 nm perpendicular, and were bent (Fig. 4b), probably as a result of the incorporation of Mn3+(s) in layers containing mainly Mn4+. The layer-to-layer distance was ∼5.5 Å, typical for dehydrated phyllomanganates (Cygan et al., 2012; Wegorzewski et al., 2015). Dehydration was possibly induced by sample freeze-drying or by the secondary vacuum in the TEM. For type-1 and type-2 crystals having similar lateral dimensions, it is proposed that the latter formed from the coherent stacking of the former;
    Fig. 4

    TEM observation of MndBi-0.5. (a) Low-magnification view; (b) type-2 crystal; (c) and (d) type-3 crystals; (e) type-4 crystal; (f) crystal built from the stacking of type-4 crystals; (g) and (h) type-3 crystals deposited on the Cu grid from a drop of suspension instead of microtome preparation. Insets in (b), (c), (d), (f), (g) and (h) are enlarged views of regions delimited with dotted squares. In (d) and (h), arrows point to the presence of dislocations. The inset in (e) is a fast-Fourier transform (FFT) analysis of the region shown with a dotted square.

    Fig. 4

    TEM observation of MndBi-0.5. (a) Low-magnification view; (b) type-2 crystal; (c) and (d) type-3 crystals; (e) type-4 crystal; (f) crystal built from the stacking of type-4 crystals; (g) and (h) type-3 crystals deposited on the Cu grid from a drop of suspension instead of microtome preparation. Insets in (b), (c), (d), (f), (g) and (h) are enlarged views of regions delimited with dotted squares. In (d) and (h), arrows point to the presence of dislocations. The inset in (e) is a fast-Fourier transform (FFT) analysis of the region shown with a dotted square.

  • Type-3 crystals were ∼50 nm in the layer plane, and were built of ∼10–20 layers stacked parallel to each other, with a layer-to-layer distance of ∼5.5 Å (Fig. 4c, d). As type-2 crystals, they were bent. Images of the lateral terminations of type-3 crystals showed that they consisted of stacked type-2 crystals (Fig. 4c). Layer dislocations were observed also on some crystals (Fig. 4d), suggesting that type-2 crystals connected also within the layer plane to form type-3 crystals. Most likely, the type-3 crystals thus originated from the oriented aggregation of type-2 crystals, both within the layer plane and perpendicular to it;

  • Type 4 crystals had lath-like shape, dimensions of ∼50–100 nm in the layer plane and ∼10–20 nm perpendicular (Fig. 4e), layer-to-layer distance of 5.5 Å (inset in Fig. 4e), and were rarely found as stacks (Fig. 4f). It is speculated that they resulted from the same aggregation mechanisms proposed for the type-3 crystals, but with fewer defects between connected type-2 crystals and/or a specific distribution of Mn3+(s) so as to minimize layer strains and bending. Note that layer bending and dislocations observed in type-3 crystals cannot be related to sample preparation, as crystals deposited from a drop of a δ-MnO2 suspension exhibited similar microstructural features (Fig. 4g, h).

These observations demonstrate that the surface of δ-MnO2 is modified at loadings as low as 0.5 Mn2+ per nm2. As a consequence, surface complexation models, which rely on a reversible sorption hypothesis (Tournassat et al., 2013), may not be suited to describe Mn2+ interaction with δ-MnO2.

All types of crystals found in samples with lower Mn2+ loadings were observed in MndBi-1.5 that contained also larger crystals (type-5), ∼100–150 nm in the layer plane and ∼20–30 nm perpendicular (Fig. 5a). In contrast, XRD indicated that MndBi-1.5 consisted mainly of isolated δ-MnO2 nanosheets. However, type-5 crystals were bent (Fig. 5b). Bending did not involve layer disruption (Fig. 5c), but likely led to variable layer-to-layer distance, best quantified from FFT analysis of the images (Fig. 5d). This analysis showed that layer-to-layer distance varied within a given crystal, being equal to 4.4–4.6 Å. Both layer bending and layer-to-layer distance heterogeneity within a given crystal reduced crystallite size within the layer plane and led to vanishing of 00l reflection.
Fig. 5

TEM observation of MndBi-1.5. (a) Low-magnification view with the different crystal morphologies indicated with arrows; (b) type-5 crystals at intermediate magnification; (c) type-5 crystal exhibiting layer bending. The area shown with a dotted square in (c) is enlarged in the inset. (d) A detailed view of (c) and the inset is a FFT analysis of the region shown with a dotted square.

Fig. 5

TEM observation of MndBi-1.5. (a) Low-magnification view with the different crystal morphologies indicated with arrows; (b) type-5 crystals at intermediate magnification; (c) type-5 crystal exhibiting layer bending. The area shown with a dotted square in (c) is enlarged in the inset. (d) A detailed view of (c) and the inset is a FFT analysis of the region shown with a dotted square.

Finally, MndBi-28.1 contained, in addition to all previously mentioned types of crystals, type-6 crystals (Fig. 6) whose selected-area electron diffraction (SAED) patterns were typical for synthetic feitknechtite (Fig. 6c). These crystals were up to ∼100 nm × 500 nm and were built of the stacking of laths resembling type-5 crystals but free of stacking defects and dislocations. These laths were not simply aligned, but were connected (Fig. 6d, e), as shown by both the observation of interference fringes (inset in Fig. 6e) and the FFT image analysis showing a homogeneous layer-to-layer distance (4.4 Å, inset in Fig. 6e). It is proposed that type-6 crystals formed from the oriented aggregation, perpendicular to the layer plane, of type-5 crystals. The presence of dislocations in type-5 crystals is thought to have inhibited their oriented aggregation. Bending was not observed either in type-6 crystals, possibly because all Mn4+(s) was converted to Mn3+(s), allowing for their specific distribution to accommodate their Jahn–Teller distortion. NoMn2+-rich phase could be detected with SAED, despite a molar ratio of sorbed Mn2+ to structural δ-MnO2 Mn of 3 (assuming a defect-free δ-MnO2 and a molar mass of 119 g mol−1, Giovanoli et al., 1970). Consequently, it was concluded that part of the Mn2+ was sorbed by δ-MnO2 and/or synthetic feitknechtite, or precipitated as an amorphous phase.
Fig. 6

TEM observation of MndBi-28.1. (a) Low-magnification view; (b) and (c) SAED patterns collected on the smallest crystals (top right in (a)) and the biggest crystals (centre in (a)), respectively; (d) and (e) type-6 crystal. The FFT analysis shown as an inset at the bottom right of (e) was calculated over the whole image while the inset at the top right is an enlarged view of the area delimited with a dotted square.

Fig. 6

TEM observation of MndBi-28.1. (a) Low-magnification view; (b) and (c) SAED patterns collected on the smallest crystals (top right in (a)) and the biggest crystals (centre in (a)), respectively; (d) and (e) type-6 crystal. The FFT analysis shown as an inset at the bottom right of (e) was calculated over the whole image while the inset at the top right is an enlarged view of the area delimited with a dotted square.

3.4. The role of oriented aggregation in nanoparticle growth

The present description of the vernadite-to-feitknechtite transformation, from macroscopic to crystal scales, is consistent with our current understanding of this transformation (Elzinga, 2011, 2016; Lefkowitz et al., 2013; Lefkowitz & Elzinga, 2015). One of the key findings is the major role of oriented aggregation in the nucleation and growth of feitknechtite from δ-MnO2 precursors. This process is increasingly recognized as one of the main processes responsible for nanoparticle growth (Penn, 2004; Niederberger & Colfen, 2006; Yuwono et al., 2010): carbonates (Shen et al., 2006), iron (Jia & Gao, 2008a and b) and manganese (Portehault et al., 2007) oxides. It was also shown to be responsible for the phase transformations of nanoparticle precursors (e.g., Hockridge et al., 2009). Recently, oriented aggregation was shown to be involved in the formation of synthetic todorokite and cryptomelane from a δ-MnO2 precursor (Chen et al., 1986; Atkins et al., 2014, 2016; Grangeon et al., 2015) and in the change of δ-MnO2 layer symmetry, from hexagonal to orthogonal (Zhao et al., 2016). It is thus possible to speculate that Mn oxides found in surficial environments were to a large extent formed from vernadite precursors (Bodeï et al., 2007; Xu et al., 2010), possibly through oriented aggregation mechanisms.

Vernadite with hexagonal layer symmetry is a sink for many trace elements in the environment. In natural settings were Mn2+(aq) is abundant, such as in lake and marine water columns and in sediments (Burdige, 1993), interactions between vernadite and Mn2+(aq) may induce morphological and mineralogical changes that could impact its reactivity and the fate of trace elements sorbed at its surface. Two main structural transformations were identified during reaction of δ-MnO2 (synthetic vernadite) with Mn2+(aq). At a low Mn2+ to δ-MnO2 ratio (<3.27 mmol g−1), δ-MnO2 crystal size increased with Mn2+(aq). At a Mn2+ to δ-MnO2 ratio of 3.27 mmol g−1 or higher, δ-MnO2 was converted to synthetic feitknechtite, and the size of synthetic feitknechtite crystals was observed to increase with Mn2+(aq). δ-MnO2 with orthogonal layer symmetry was not detected as a transformation intermediate, possibly because experiments were conducted at slightly acidic pH (Zhao et al., 2016).

When crystal size increases, the specific surface area, and especially the contribution of particle edges to this surface, decreases. Vernadite edge sites being reactive (Simanova et al., 2015), the increase in vernadite crystal size when in contact with Mn2+(aq), even at low concentrations, can be detrimental to its trace-metal scavenging ability. By analogy, the same is expected for feitknechtite at high Mn2+ to δ-MnO2 ratio. The impact of the vernadite-to-feitknechtite transformation on the fate of trace elements remains unclear, however. The lack of any significant dissolution of δ-MnO2 potentially minimizes the release of trace metal elements along the reaction pathway, but their possible incorporation in the newly formed feitknechtite structure remains undocumented. Unlike tunnel structures, feitknechtite has been observed seldom in natural settings, probably because of its limited chemical stability and high sensitivity to fluctuation in redox conditions, which lead to its back-conversion to vernadite (Bargar et al., 2005). Consequently, no data are currently available to evaluate the affinity of trace elements for feitknechtite in natural settings. Similar to tunnel structures (Liu et al., 2004; Kumagai et al., 2005; Cui et al., 2011; Atkins et al., 2014), additional insights into vernadite reactivity in suboxic systems and its impact on the fate of trace metals could probably be sought experimentally as feitknechtite is frequently observed in laboratory experiments mimicking natural conditions (Hem et al., 1982; Murray et al., 1985; Wang et al., 2015).

S.G. acknowledges funding by the French National Research Agency (ANR, grant ANR-14-CE01-0006) and thanks Jacques Deparis, Nicolas Marty and Julie Philibert for fruitful discussions. SOLEIL data were acquired in the frame of proposal 20141260. A. Fernandez-Martinez and T. Conte are respectively thanked for help during XRD data collection and Mn2+ measurement. This article benefited from comments and suggestions by an anonymous reviewer and Alain Baronnet.