The in situ study of soil solutions in gleysols requires a specific procedure of sampling and measurement: soil waters are sampled without any contact with oxygen of air, to prevent oxidation, and in the dark, to prevent photoreduction, in chemically inert devices, placed at different depths, that act as zero-tension lysimeters [14,27]; this system allows us to sample free soil solutions. The non-conservative parameters are measured in the field: temperature, pH, redox potential, FeIIaq. concentration by colorimetry with DPKBH (see [32], modified by [9,20]), nitrite and sulphide by colorimetry. In the laboratory, after filtration under nitrogen atmosphere, major cations and anions, metals, dissolved organic and inorganic carbon and alkalinity are measured. These data allow us, on the one hand, to establish the seasonal dynamics of the chemistry of soil waters, and, on the other hand, to check mineral–solution equilibriums in thermodynamic models providing for activity coefficients corrections and ion-pair formation, such as EQUIL(T) [17] or PHREEQC [31]. An originality of our approach is the following: total dissolved Fe(II) is used, with pH, major cations and anions concentrations to compute the activity of Fe2+aq.; this value is not influenced by particulate or colloidal Fe(III), which would result in a large overestimation of Fe2+aq., as dissolved Fe(III) cannot be separated from colloidal Fe. From the preceding value of the concentration of Fe2+aq., the activity of Fe2+aq. is obtained. All mineral–solution equilibriums are checked by writing all reactions with the activity of Fe2+aq., pH and pe as master variables [9] and by using a consistent set of thermodynamic data of the different iron species [9,36].

Two main patterns of seasonal dynamics can be distinguished, depending on the location of hydromorphic soils in landscapes (Fig. 2) [8]. The vertical movements of the watertable control the first pattern, which is generally observed in colluvio-alluvial systems, as follows: (i) when the level of the watertable is high, soils are waterlogged during a long time; reductive conditions occur and significant release of Fe^II in soil solutions is observed; (ii) when the level of the watertable gets lower, the environment becomes oxidant by entry of air, and Fe^II concentrations in soil solutions decrease drastically to zero (Fig. 3).

The recharge of oxygen by rainwater controls the second, generally observed in hydromorphic soils located at mid-slopes or over lithological discontinuities, as follows: (i) during the rainy season, enough dissolved oxygen is present, though the soils are completely waterlogged, so that the environment stays oxidant and no Fe^II is released in solution; (ii) when the rainflow decreases, the environment gets confined, becomes reductive and Fe^II concentrations increase strongly, until dissolved oxygen is supplied and Fe precipitates in situ.

These two patterns result in two different geochemical behaviours from a same initial condition, i.e. reductive and waterlogged. In the first case, there is a continuous flow of water when Fe is mobile. This results in the formation of bleached horizons, with exportation of Fe from the horizon. In the second case, there is no outflow when Fe is mobile, and Fe is globally conserved in the horizon, though locally it is segregated.

Check of Fe minerals–solutions equilibriums showed that all soil solutions are: (i) oversaturated with respect to Fe(III) oxides, haematite, goethite and lepidocrocite; (ii) undersaturated with respect to Fe(II) hydroxide; (iii) undersaturated with respect to GR1(Cl), GR1(CO₃), GR2(SO₄); (iv) close to equilibrium with a compound having the chemical formula FeII1−xFeIIIx(OH)x+2, with 0.42<x<0.86 [9]. This result is consistent with previous conclusions proposed in pioneering works by Ponnamperuma [33] in Philippines under rice crop or by Lindsay [26] in gleysols in Colorado. However, at this stage, the question remained whether this formula corresponds to a mean compartment of ferrosoferric complexes or to a specific mineral. Thus additional investigations of the Fe solid fraction of the soils were made.

To characterize the Fe solid fraction in soils, the principal experimental difficulties are the following: (i) the control of the redox state of soil, (ii) the heterogeneity of distribution of iron features, and (iii) the lability of the Fe^II–Fe^III minerals.

Soil samples must be collected in a large volume with the surrounding soil solution and maintained in anoxic conditions in airtight box, without sieving and air-drying, the standard protocol in soil science. The permanence of the original blue or green-blue colours during all experiments is used as a preservation criterion. The physical separation of concretions or spots of Fe concentrations by microsampling under binocular in airtight box is preferred to a bulk analysis of the samples [45].

Kinetics by selective extractions [7] with citrate–bicarbonate [45] and dithionite–citrate–bicarbonate [21,30,46] define fractions of differential reactivity between microsites and complete the mineralogical and crystallographic analysis obtained by X-ray diffraction (XRD).

On a gleysol (Fig. 4) developed on granite in Brittany (France), seven microsampled soil features containing iron were distinguished from surface towards the depth. In the organo-mineral horizon, they are coatings with a reddish purple colour (2.5YR 3/6), noted LV, located around very fine roots and on the surfaces of aggregates. In the alluvial redoxic horizon, the BTO sample consists of diffuse ochre spots more clayey as the neighbouring white spots. In the structural redoxic horizon, the STO sample is made of diffuse ochre spots (7.5YR 5/8) in the matrix, with a similar texture as BTO sample. In the sandy-clayey saprolite, PO sample consists of large rusty spots (5YR 5/8) located in the most clayey part of the weathering zone, AR sample is the homogeneous bluish matrix (5GB 6/1), which turns into ochre when in contact with air and GRB and GRG are the rust coatings (7YR 5/8) developed directly on old root channels.

Selective extractions [49] show the occurrence of a Fe labile fraction available without reduction whose size and reactivity can be determined by citrate–bicarbonate extractions. This fraction depends quantitatively on the waterlogged and oxido-reduction states of the milieu. In fact, even a short contact (a few hours) with air of the AR sample before CB extraction resulted in the immobilization of a large Fe fraction, e.g., up to 40% for AR sample (Fig. 4), with respect to the amount extractible by CB under N₂ atmosphere. By calibration of the CB extraction method on synthetic minerals, it can be shown that mixed Fe^II–Fe^III compounds, such as green rusts may contribute to the signal [46]. However, these extractions, though selective, are not specific of a mineral. As the total Fe content is small, about 5%, X-ray diffraction is not sensitive enough, and moreover the main GRs' line is very close to the main line of kaolinite. The question merges with the preceding, whether this Fe labile fraction corresponds to a mixed compartment of ferrosoferric complexes or to a specific mineral.

The structure of the mineral was proposed to be GR1, analogous to GR1(Cl) [34], with a symmetry group type R3¯m and lattice parameters a=0.31901 nm and c=2.3856 nm (Fig. 6). The structure of the mineral was studied by X-ray absorption spectroscopy (XAS), both by recording the Extended X-ray absorption fine structure signal (EXAFS) and the X-ray absorption near edge structure signal (XANES) at the Fe K-edge, along with transmission Mössbauer spectroscopy (TMS). The same techniques were applied to synthetic GR1(Cl), GR1(CO₃), and to a synthetic pyroaurite with approximate formula [Mg(II)5Fe(III)2(OH)14]2+[CO32−⋅nH2O]. Two samples of hydroxycarbonates obtained by mixing Fe²⁺, Fe³⁺ and Mg²⁺ salts and coprecipitating hydroxides by addition of NaOH in the presence of Na₂CO₃ were also prepared and analysed by TMS. The x mole ratios were measured by Mössbauer spectroscopy. Powder X-ray diffractograms were performed on the synthetic samples. The global formulae of the compounds obtained were Fe(II)₂Mg(II)₂Fe(III)₂(OH)₁₂ CO₃ and Fe(II)₄Mg(II)₂Fe(III)₂(OH)₁₆CO₃. The main results obtained [37] are briefly summarized.

Both Fe²⁺ and Fe³⁺ valence states occur in close neighbourhood, as evidenced from the XANES part of spectrum, which is sensitive to the oxidation state of Fe and to its local symmetry; there are no clusters of Fe²⁺ or Fe³⁺. The mineral is confirmed to be a green rust, a double-layered hydroxide, because its pseudo-radial distribution functions (PRDF) display the same set of peaks (P₁–P₆) and are in agreement with previous studies on similar compounds [38] (Fig. 7). The lattice parameter a, is obtained at about 0.31–0.32 nm, as earlier predicted, from the second P₂ peak of PDRF, i.e. the FeFe distance. The positions of farther – and smaller – peaks of the PRDF correspond to the hexagonal array of cations, second, third…neighbours of Fe, in the sequence a3,2a,a7,3a (Fig. 8).

Mg substitutes for Fe in the mineral. Evidence for this is that the radial distribution function obtained for the mineral is intermediate between those of GR1s and of synthetic pyroaurite, that is between Fe²⁺–Fe³⁺ and Mg²⁺–Fe³⁺ hydroxycarbonates. The peak positions are about the same, which proves that the mineral pertains to the group of GRs. However, the peak intensities observed for the mineral are different from the peak intensities of synthetic GRs and intermediate between those of GR1s and of synthetic pyroaurite. This is not surprising as many natural minerals are mixed Fe(II)–Mg(II) compounds (pyroxenes, micas, amphiboles, clay minerals, etc.).

The local ratio Mg/Fe is about 2±1; this was estimated by fitting a linear combination of the contributions of Mg and Fe to the back-Fourier transform of PRDF curve: in the case of synthetic GR1s, no Mg is present and backscattering atoms are Fe atoms, whereas in synthetic pyroaurite, each Fe atom is surrounded by Mg atoms, so that Mg atoms are the only backscattering atoms. A much better fit is obtained when provision is made for both Mg and Fe contributions than with only Fe contribution.

The global mole ratio Mg/Fe cannot be directly obtained as the mineral is labile and small sized. However, additional data are given by microprobe analyses and by selective extraction techniques [15].

Microprobe analyses performed on 40 volumes from Fougères containing fougerite particles show that Mg_total/Fe_total ratio ranges from 0.18 to 0.62. Each point is the average volume analysis of 2 μm³.

Selective extraction techniques using CB and CBD [45] give another argument. In the saprolite from Fougères, 48% Mg is extracted by CB, and all this Mg is extracted in 6 h, the global mole ratio (Mg/Fe)_CB being close to 1. No additional Mg is extracted by CBD. This shows that Mg is present in the silicated phase (52%), but not in well crystallized Fe oxides.

In the reductic silty horizon in Fougères, 59% Mg is extracted by CB, Mg and Fe being simultaneously released with time.

The kinetics obtained is: Fe/Mg=0.5981+0.000792t (r2=0.975, N=4). The intercept is Fe/Mg= 0.5981, which corresponds to a global Mg/Fe ratio = 1.67. We can thus conclude that Mg present in the CB-extractible compartment is associated with Fe, i.e. originates from fougerite. The values obtained for the global mole ratio by selective extraction techniques in Fougères are thus in the range 0.1–1.7. This range is compatible with the local mole ratio obtained by XAS.

The following structural formula can thus be proposed for fougerite:

The nature of the anion present in the mineral cannot be determined by XAS, because spectra of various GRs proved to be independent of the nature of the interlayer anion. Parameter a is accessible by XAS and determined as 0.31–0.32 nm [37]. This is compatible with the values obtained for pyroaurite as 0.3113 nm [3], for GR(CO32−) as 0.3160 nm [13,23] and for GR(Cl⁻) as 0.319 nm [35].

Parameter c is more sensitive to the nature of the interlayered anion, but it is not accessible for the natural mineral.

It has been recently observed that synthetic GR(CO32−) can be partially deprotonated, with FeOFe bonds [19,24] by oxidation. It was argued that this could explain Fe(III) mole ratios larger than 1/3, and that this could be the case for the natural mineral fougerite. However the compound obtained, GR(CO32−)* is not well characterized, and its a and c parameters are smaller than the same parameters for synthetic GR(CO32−), due to the shrinkage of the lattice; specifically, a “shrinks from 0.317 to 0.3014 nm” [19]. Such a shrinkage is not observed in the natural mineral, whose a parameter is larger than 0.31 nm. Moreover, there is a discrepancy between the stability of GR(CO32−) and of other GRs [10], which implies that carbonate is probably mixed with hydrogenocarbonate in the interlayer. It has been demonstrated that either the stoichiometry or the Gibbs free energy of formation of GR(CO32−) must be revaluated. The recent study of GR(CO32−) by FTIR [23] was made at very large concentrations of HCO3− (1 M) and only showed that this technique is not sensitive enough to detect concentrations of hydrogenocarbonate at the levels commonly found in acid brown soils, 10⁻⁴ M.

Anyway, it is likely that the nature of the interlayer anion changes according to the environment. So, we can expect fougerite can accommodate a variety of anions, either inorganic or organic, the same as synthetic GRs and many other layered double hydroxides (LDH) [11]. This is similar to the situation encountered with smectites, which can accommodate a variety of interlayered cations. From the chemical composition of soil waters where fougerite from Fougères occurs, activities of Cl⁻ and CO32− anions are very small, respectively 10⁻³ and 10⁻¹⁰. We suggested [15] that the most probable anion is OH⁻, which can be obtained by a simple deprotonation of a water molecule in the interlayer when Fe(II) is partly oxidized, without the necessity for the pH in the environment to be very high. Check of fougerite/solution equilibriums, based upon the development of a ternary solid solution model with different anions, OH⁻ [10], Cl⁻, SO42− and mixed HCO3−/CO32− [16] shows that the soil solution (i) is always undersaturated when Cl⁻ is the interlayer anion; (ii) always oversaturated when OH⁻ is the interlayer anion. Depending on the Mg mole fraction in the mineral, the solutions are either undersaturated or oversaturated when SO42− or mixed HCO3−/CO32− are the interlayer anion. This can be different in other environments, depending on the ratios of concentrations of the anions and on the Mg content in the layer.

In situ, fougerite transforms into lepidocrocite by oxidation [1] excluding high level of carbonate, dissolved aluminium and silicon in surrounding waters, which are inhibitors of lepidocrocite formation [39]. Moreover, in the experimental study on the electrochemical synthesis of GR(CO32−) [23], the authors noted: “The surface of the iron exhibited a greenish-blue colour after the rinsing and drying procedure. No colour change apparently occurs in air indicating that the oxidation layer is rather stable”. This is contrary to the classical observation that the colour of gleysols readily changes from bluish-green to ochre when exposed to the air. It appears then that the results from those syntheses cannot be straightforwardly extrapolated to natural conditions and to the mineral formation.

It can thus be concluded that for fougerite, the question of the exchangeable character of anions is open. We could speak of CO₃-fougerite, SO₄-fougerite, OH-fougerite or Cl-fougerite as we speak of Ca-montmorillonite, Na-montmorillonite, or of fluoro-apatite, carbonate-apatite, etc. Selectivity constants have not been determined, though qualitative results exist. According to [11], “the following sequence of exchange for both mono- and divalent anions in the LDHs has been derived:

Though the selectivity constants are in favour of divalent anions, which are the same mutatis mutandis for smectites and divalent cations, the exclusive presence of carbonate cannot be established as a rule, and the presence of OH⁻ or Cl⁻ in the interlayer cannot be ruled out. Homologous compounds have been described with OH⁻ in the interlayer, such as meixnerite, Mg₆Al₂(OH)₁₈⋅4 H₂O, and with Cl⁻ in the interlayer, such as iowaite, Mg₆Fe₂(OH)₁₆Cl₂⋅4 H₂O and woodallite Mg₆Cr₂(OH)₁₆Cl₂⋅4 H₂O.