From schwertmannite to natrojarosite: Long-term stability and kinetic approach

Synthesis of iron hydrosulfates

Two sets of precipitates were produced by aging two different aqueous solutions of Fe₂(SO₄)₃ and NaOH (0.12 and 0.2 M for the LC20 samples and 0.20 and 0.4 M for the HC20 samples) at ambient temperature (~20 °C) for various durations from 3 h to 210 days. The pH and composition of the initial aqueous solutions are given in Table 1. The composition of the parent solutions was selected to obtain stable phases faster than in our previous study (Jiménez et al. 2019). Additionally, the effect of temperature on the nature of the precipitates was checked by introducing the starting solutions (Table 1) in a thermostatic reactor preheated at 70 °C (LC70 and HC70 samples for the low and high concentrated solutions, respectively) for 3 h. After preparing the solutions, all the reaction vessels were sealed to avoid evaporation and maintained under constant stirring (100 rpm). All precipitates were produced in triplicate using deionized (MilliQ) water and analytical grade reagents.

The solution pH was measured at the beginning and the end of the aging process with a combination electrode (Ross-Thermo-Orion-810200) and a digital pH meter (Crison Basic20). The solid and liquid phases were separated by centrifugation for 10 min at 3000 rpm in a Rotina 380. The initial aqueous solutions were modeled using the geochemical code PHREEQC (Parkhurst and Appelo 1999) and the WATEQ4F database to calculate the activities of different chemical species and saturation indexes (SI) with respect to the relevant solid phases at the beginning of the experiments. The database was completed with the thermodynamic solubility product of Shm (Ksp = 10^−5.28) (Yu et al. 1999).

Characterization of solid phases

The precipitates were oven-dried at 30 °C and then gently crushed to a fine powder using an agate mortar and analyzed by powder X‑ray diffraction (XRD). The diffraction patterns were collected in the 2θ range between 5° and 60° with a step size of 0.02° on a Philips X’Pert-PRO diffractometer using CuKα radiation. Indexing of the main reflections and calculation of the cell parameters of the solid phases were made using X’Pert HighScore Plus (PANalytical B.V.). The crystallinity of the samples was determined from the full-width at half maximum values (FWHM) of the highest d-spacing and more intense reflections. The FWHM values were established after performing a Kα2 stripping and then applying the Pseudo-Voight profile fitting function. Moreover, the crystallite size of samples was roughly estimated using the X’Pert Plus “Scherrer calculator” tool. Morphology of solids was studied using a scanning electron microscopy (SEM) (JEOL-6610). Transmission Fourier transform infrared spectroscopy (FTIR) spectra of the powdered samples compressed into discs with KBr were recorded from 400 to 4000 cm⁻¹ on a Nicolet Magna IR 560 spectrometer fitted with a DTGS absorbance detector.

The surface of the precipitates was analyzed by ex-situ X‑ray photoelectron spectroscopy (XPS) with a Specs spectrometer, using MgKα or AlKα (30 eV) radiation emitted from a double anode at 50 W. The binding energies of the resulting spectra were corrected by employing the binding energy of adventitious carbon (284.6 eV) in the C1s region or the binding energy of Na⁺ (1071.7 eV) in the Na1s region. The backgrounds were corrected using Shirley’s baselines. All the analyzed regions (C1s, O1s, Na1s, Fe2p3/2, and S2p) were deconvolved by means of mixed Gaussian-Lorentzian functions (90:10). The quantitative analyses were based on atomic sensitivity factors stored in the CasaXPS database (v2.3.12Dev6).

Isothermal experiments

The transformation of the metastable phase (schwertmannite, Shm) to the stable ones (natrojarosite, NaJrs) was studied for the precipitates obtained from the highest concentrated starting solutions (HC experiments). The procedure begins by mixing the starting solutions at ambient conditions (20 ± 1 °C) and then putting them in a thermostatic bath to predetermined temperatures (20, 35, 50, or 70 °C), where the mixture was maintained at isothermal conditions during reaction times varying between 0.5 and 26 h. The transformation extent of Shm into NaJrs with increasing time has been checked by examining the most important reflections of NaJrs (012, 021, and 113) by using X’Pert Viewer. We considered this scenario to establish an extent of transformation up to about 95% (Y ~95%). A more precise quantification of crystalline vs. poorly crystalline phases to assess the transformation extent requires a Rietveld analysis, which is out of the scope of the present work.

Evolution of the aqueous pH during precipitation at ambient temperature

Figure 1 shows the evolution of the pH of the aqueous solution as a function of time for the experiments carried out at 20 °C (LC20 and HC20). As can be observed, the pH quickly decreases during the first day of aging and approaches asymptotic values of about 1.89 after 42 days. The variation of pH values is negligible or within the experimental uncertainties in solutions maintained for prolonged reaction times (210 days).

Figure 1

Variation of pH with time during aging at 20 °C. (Color online.)

XRD analysis

The X‑ray powder diffraction patterns of the LC precipitates are shown in Figure 2. The diffractogram of the precipitate obtained after 1 day of reaction at 20 °C shows two humps at 2θ of ~35° and ~61° that match the schwertmannite pattern (PDF-47-1775). Schwertmannite is a poorly crystalline phase, which was recognized as a mineral regardless of its metastability (Bigham et al. 1994; Barham 1997; Acero et al. 2006). The general formula of schwertmannite is Fe₈O₈(OH)_8–2x(SO₄)_x·nH₂O (with x varying from 1 to 1.9). Although the first studies proposed that Shm crystallizes in the tetragonal system (Bigham et al. 1994) with a space group P4/m, a monoclinic structure (space group P1) was finally resolved by combining high-energy X‑ray diffraction and theoretical simulations in which the position of water molecules was disregarded (Fernandez-Martinez et al. 2010). It must be mentioned that 2-line ferrihydrite [Fe₁₀O₁₄(OH)] produces a similar XRD pattern to that of Shm, but it precipitates at neutral pH at ambient temperature (Bigham et al. 1996; Yu et al. 1999), whereas Shm always precipitates under highly acidic conditions (pH < 4 values), which are the prevalent conditions in this work (Fig. 1). The precipitates obtained after aging at 20 °C (LC20-7d and LC20-14d) and 70 °C (LC70-3h) display the typical reflections of natrojarosite (PDF 36-425) [NaFe₃(SO₄)₂(OH)₆], which crystallizes in the trigonal system (R3m space group 160). Other crystalline phases such as hematite or goethite were not identified in the samples aged up to 42 days at 20 °C. In parallel to the evolution of phases, the color of the precipitates obtained 20 °C changes from brown-reddish in the first moments to brown-yellow at the end of reaction (Fig. 2).

Figure 2

X‑ray diffractograms of the solid phases obtained during aging LC experiments. Broad diffraction peaks of schwertmannite (PDF 47-1775) are marked with arrows. The main diffraction peaks of natrojarosite (PDF 36-425) are identified after prolonged times at 20 °C and after 3 h of aging at 70 °C. The evolution of color of the precipitates obtained in experiments at ambient temperature are also included. (Color online.)

The diffraction patterns of the HC20 precipitates (Fig. 3) reveal that schwertmannite precipitates at the beginning of the experiments (HC20-3h). However, incipient reflections at 2θ of ~17.61°, ~28.65°, and ~29.29° observed in the diffraction patterns obtained after 1 day of aging (HC20-1d) indicate the existence of a phase with some degree of crystallinity that becomes more apparent with further aging (Fig. 3). These reflections match the main XRD reflections (012, 021, and 113) of natrojarosite [NaFe₃(SO₄)₂(OH)₆]. No further advance in the mineral transformation is observed after long-period reactions (210 days). Furthermore, the transformation from a poorly crystalline phase to natrojarosite is also evidenced by the change of color of the precipitates for increasing aging times, from dark brown to yellow-brown (Fig. 3). The diffraction patterns obtained for the precipitates produced at 70 °C (Fig. 3) clearly match that of the natrojarosite.

Figure 3

X‑ray diffraction patterns of the solid obtained during aging HC experiments. Diffraction peaks for poorly crystalline phase are marked with arrows. The main reflections of natrojarosite (PDF 36-425) are signaled with gray points. The evolution of color of precipitates from the beginning to the end of the experiments is included. The increase of crystallinity of natrojarosite during aging is indicated by italic numbers of the FWHM values (2θ) of reflections 021 and 113. (Color online.)

The unit-cell parameters of natrojarosite calculated from the powder diffraction diagrams show similar values for all the HC20 precipitates (a ~ 7.325 ± 0.010 Å and c ~ 16.842 ± 0.080 Å), regardless of the increase of crystallinity. The unit-cell parameters of the sodium-hydronium jarosite series decrease with increasing Na content because of the lower size of Na⁺ ions in comparison with H₃O⁺ ions (Basciano and Peterson 2008). Jarosite-type compounds often have iron deficiency, with Fe-site occupancies lower than 86%, which seems to be responsible for discrepancies in the unit-cell parameters (Brophy and Sheridan 1965; Drouet and Navrotsky 2003; Basciano and Peterson 2007). A good estimation of the composition of the precipitates can be obtained from the position of reflection 006, being 32.19 (°2θ) for pure natrojarosites, whereas it takes a value of 31.53 for pure hydronium-jarosite. For intermediate compositions, the 006 reflection could broaden or split up into two peaks, the latter indicating the presence of a miscibility gap. Here, the position of the 006 reflections (~32.17, 2θ) is near that of the pure natrojarosite, with no broadening or splitting of this reflection observed in the diffractograms (Fig. 3). Thus, from the unit-cell dimensions and the 006 reflection, HC20 and HC70 precipitates can be identified as Na-rich members of the (Na,H₃O)Fe₃(SO₄) (OH)₆ solid solution (NaJrs).

The crystallinity of the HC20 precipitates was also examined from the broadening (FWHM) of the most intense reflections 021 and 113 of NaJrs. The decrease of FWHM values of both reflections, along with the increase of their intensity observed from XRD (Fig. 3), confirm the increase of crystallinity during aging (Langford and Wilson 1978). Working with the FWHM of the 113 reflections, i.e., the reflections with the highest intensity, the crystallite size can be estimated to be 66 nm for NaJrs. As it is widely known, the crystallite size represents the size of the coherently diffracting domain and not the size of the precipitate particles.

SEM

The electron images of the precipitates obtained at different temperatures (20 and 70 °C) are shown in Figure 4. SEM images of schwertmannite obtained in the LC20 precipitates show that their surface is constituted by an agglomerate of nanospheres (Fig. 4a), formed from the fast precipitation of discrete colloidal particles that prevent the incorporation of growth units to the mineral surface (Sangwal 1999). After 14 days of aging, the agglomerate evolves to crystals of natrojarosite displaying an equidimensional habit (~1 μm size) whose morphology could be similar to pseudo-rhombohedra (Fig. 4b). Similarly, the evolution of the crystallinity with aging is observed in the HC20 precipitates. SEM images show that aggregates of small crystals with shapeless morphology at the beginning of the aging process gain the typical rhombohedral morphology of natrojarosite, showing well-defined faces and edges after 42 days of aging (Fig. 4c).

Figure 4

SEM images of iron hydroxysulfates. (a) The surface of schwertmannite is composed by aggregate of particles. (b) Natrojarosite crystals. (c) Rhombohedral crystals of natrojarosite.

FTIR

Figure 5 shows the FTIR spectra obtained for HC70-3h and LC70-3h samples. The peaks at 970–995 cm⁻¹ (LC70) and 991 cm⁻¹ (HC70) are assigned to the symmetric ν₁ stretching of S O 4 = , whereas the bands at 1088/1093 cm⁻¹ and the shoulders at 1185/1190 cm⁻¹ (HC70/LC70) are attributed to the ν₃ (doublet) vibrations of SO₄⁼ (Powers et al. 1975; Majzlan et al. 2011; Sotiropoulou et al. 2012). The shoulder at 1025 cm⁻¹ may be attributed both to the ν₁ vibration of sulfate and to the Fe-O-H in-plane-bend (δOH) (Powers et al. 1975; Majzlan et al. 2011; Sotiropoulou et al. 2012). The adsorption band at around 613 cm⁻¹ is assigned to the asymmetric ν₄ bending motion of sulfate (Bishop and Murad 2005). The wide band centered at 3310–3345 cm⁻¹ is characteristic of the ν stretching of O-H (Casas et al. 2007). All these bands are characteristic of natrojarosite. However, some other features in the spectra of Figure 5 must originate from a different source. For instance, the conspicuous band at 1635 cm⁻¹ corresponds to the water-bending vibration (Bishop and Murad 2005; Majzlan et al. 2011), but it is present in synthetic jarosite samples where H₃O⁺ replaces some of the monovalent sites (Grohol and Nocera 2002; Majzlan et al. 2011). This band is usually weak in natrojarosite and jarosite (Powers et al. 1975; Majzlan et al. 2011; Sotiropoulou et al. 2012). The peak at 795 cm⁻¹ is also absent in natrojarosite (Casas et al. 2007). On the other hand, its frequency is characteristic of the out-of-plane deformational (γ) mode of hydroxyls in goethite (Prasad et al. 2006). The band at 1635 cm⁻¹ can also be attributed to a bending mode of hydroxyls in goethite, though in natural goethite it is significantly weaker than the peak at 795 cm⁻¹ (Prasad et al. 2006), and therefore the assignation of the 1635 cm⁻¹ frequency to H₃O⁺ prevails. The presence of a small fraction of goethite is further confirmed by the shoulder at around 3130 cm⁻¹, which corresponds to the stretching mode of hydroxyls in FeOOH (Prasad et al. 2006). Finally, two main features can be found in the 400–500 cm⁻¹ region (460–465 and 490–495 cm⁻¹). Both of them are attributed to vibration modes of the FeO₆ lattice. The low-wavenumber mode can originate from goethite (Prasad et al. 2006), hydronium jarosite (Powers et al. 1975), and goethitenatrojarosite mixtures (Casas et al. 2007), whereas the ~495 cm⁻¹ frequency can be attributed to natrojarosite (Casas et al. 2007) and hydronium jarosite (Powers et al. 1975). Added to this, a small fraction of the surface iron remains as goethite. The XPS results will help in the quantification of these surface species.

Figure 5

FTIR spectra for precipitates obtained for HC70-3h and LC70-3h.

XPS

The elemental composition of the different samples obtained from the XPS analyses is shown in Table 2. In all cases, the amount of surface iron is much lower than that corresponding to natrojarosite (15 at%), but the proportion of the remaining elements is more similar to that of this mineral (70% O, 10% S and 5% Na). The amount of iron evaluated with the AlKα source is always higher than that evaluated with the MgKα source. Since the latter produces photoelectrons from a slightly thinner layer of the sample surface, it can be concluded that the iron concentration increases toward the center of the particles, probably to the limit of 15 at% marked by the natrojarosite stoichiometry (in accordance with the XRD results). This fact suggests that the bulk sodium concentration for all the samples is again that corresponding to natrojarosite (5 atoms%). On the other hand, when the surface concentration of sodium is over 5 at% (HC samples) it decreases toward the center of the particles (Table 2), but when the sodium concentration is below 5 at% (LC samples), it increases toward the center of the particles.

With respect to the determination of the different surface compounds, XPS is an analytical tool that has to be used cautiously, and the present analysis is a perfect example of this assessment. Online Materials^[1] Figure S1 shows a region subjected to deconvolution in two different ways (top and bottom), in which only a deconvolution procedure similar to that presented in the top plot permits us to satisfy the charge balance for all the samples without the need of including unlikely species (see the Online Material^[1] for further explanation).

Figure 6 shows the results of applying the appropriate deconvolution procedure to all spectra in the O1s region, whose binding energies were corrected by employing the binding energy of adventitious carbon (284.6 eV). The vertical lines indicate the average values (530.1 ± 0.1, 531.4 ± 0.1, and 532.0 ± 0.2 eV) for the three peaks needed in the deconvolutions. The O1s spectra obtained with the AlKα source (right plots in Fig. 6) were affected by the Na KLL region in the form of a wide peak at 535.8 eV. The small peak at 530.1 eV is assigned to O⁼ in FeOOH (Jin et al. 2020), whereas the peak at 531.4 eV includes the contribution of oxygen atoms in the sulfate anion (Gard et al. 2020) and hydroxyls (Grosvenor et al. 2004; Liu et al. 2014; Khalid et al. 2017; Gard et al. 2020). Finally, the peak at 532.0 eV is assigned here to hydronium ions (H₃O⁺), consistent with the FTIR results. This peak might be well enclosed within the hydronium jarosite standard spectrum reported by Parker (2008) (Fig. 6). Furthermore, assignation of this peak to adsorbed water molecules, as is often the case (Diao et al. 2018; Cheng et al. 2020) would make it impossible for charge balance to be satisfied. As can be observed in Figure 6, the maxima of the different peaks present some deviation with respect to the average values (dashed lines). These spectra were taken by assuming that the binding energy for the maximum of the C1s region (284.6 eV) was a valid reference. However, during the XPS analyses, the samples suffer intense surface charging, evidenced by a high-binding energy shift of 2.8 ± 0.2 eV, meaning that the samples are non-conducting. Within this frame, the referencing method involving adventitious carbon in the C 1s region is not reliable (Greczynski and Hultman 2020). The XPS regions produced by Na⁺ and sulfur from SO₄⁼ display single-peak spectra (Fig. 7) whose characteristic binding energies should be, in principle, independent of the type of sample. Therefore, we decided to change the C1s referencing method and use the binding energy of Na1s instead, averaged for all samples. This is, in fact, the standard binding energy for sodium ions (Na⁺) (Feliu et al. 2013). As can be observed in Table 3, with the new referencing method, the standard deviation values for the maxima of O1s and S2p peaks are reduced. As expected, the peak at 168.9 eV corresponds to sulfur from sulfates (Sandström et al. 2002; Gard et al. 2020). The Fe2p3/2 region is formed by two peaks (Fig. 7) at ~711 and ~712 eV. These binding energy values are typically assigned to Fe(III) in jarosite-type minerals (Xu et al. 2013; Wang et al. 2021).

Figure 6

Deconvolution procedures in the O1s region for all samples (adventitious carbon as reference). The data for the hydronium jarosite standard plot were extracted from the work by Parker (2008). (Color online.)

Figure 7

XPS results in the S2p, Na1s, and Fe2p3/2 regions for HC70 sample (AlKα source, C1s referencing method). (Color online.)

Finally, Table 3 shows how the binding energy values of the peaks confirming the Fe2p3/2 region are more scattered than the B.E. values for the other regions. There seems to be a dependence of the peak 1 binding energy with the amount of hydronium (w), as shown in Figure 8, which might arise from the effect of the H₃O⁺ cations on the ionic distance between the Fe³⁺ cations and the SO₄⁼ and OH^– anions.

Figure 8

For the Fe2p3/2 region, variation of the binding energy of peak 1 with the amount of hydronium ions (w).

From the B.E. assignations described above, the fittings shown in Figure 7 are the result of minimizing the sum of the curve fitting errors and the errors from the charge balance equation 2s + w + x + 3y – 2 –z = 0, in which the different variables are the stoichiometric coefficients of the generic formulas (FeOOH)_v H 3 O + w N a x + F e v 3 + S O 4 = O H − z . During the deconvolution +process, the coefficients are continuously re-evaluated with the values of the different peak areas. This procedure +guaranteed perfect fits (Fig.6) while keeping the electrical neutrality of the sample surface. Table 4 shows the results.

The hydronium to sulfate molar ratio (w) is always slightly lower on the most external area of the particles (MgKα source).

This is probably due to some water elimination from the external surface of the particles in the vacuum prevailing in the XPS chamber. According to the XRD results, which show that the bulk is mainly composed of natrojarosite, the w and v values must diminish to 0 toward the center of the particles, while the x, y, and z values must become 0.5, 1.5, and 3, respectively.

Kinetics

The method called “time to a given fraction” (Putnis 1992) was used to determine the transformation kinetic from Shm into NaJrs and its apparent activation energy (E_a). This method establishes that the transformed fraction (Y) and time (t) are related by an exponential function in which t is the dependent variable instead the kinetic constant (K) that is used in standard empirical kinetic models. When the reaction mechanism does not change over the temperature range studied, t_Y can be calculated by:

where A is a fitting constant, R the gas constant (8.3144 J mol^–1 K^–1), and T the temperature in Kelvin (K). Here, the time (t) taken to reach 95% of transformation (ty, y ~ 0.95) has been estimated from a series of isothermal experiments (20, 35, 50, and 70 °C). At the beginning of the experiments, no matter what temperature is chosen, the first precipitate always corresponds to a poorly crystalline phase identified as Shm, which evolves to a crystalline phase identified as NaJrs with time. The evolution is comparable in all the series at different temperatures, but the reaction rate increases at a higher temperature. A good estimation of the transformation extent with increasing time can be checked by examining the first occurrence of the most important reflections (012, 021, and 113) of NaJrs. When these reflections are distinguished (see Fig. 3), crystals of NaJrs are predominant and the aggregates of the poorly crystalline phase have nearly disappeared. This can be considered an approximation when almost complete transformation has been accomplished. We have assumed this scenario to establish a transformation extent of about 95% (Y ~95%). Online Materials^[1] Figure S2 shows XRD diagrams obtained with increasing temperature at times when the transformation rate is considered to be 95%. The most important reflections (012, 021, and 113) of NaJrs can be identified and the broad bands of Sch disappeared, which represented the transition to a crystalline phase. Obviously, with extended reaction times, the crystallinity degree of NaJrs increases as deduced from the decreases of FWHM values similar to that observed in Figure 3 for samples aged at 20 °C.

The experimental data of the time (t) taken to reach 95% of transformation (ty, y ~0.95) have been plotted in Figure 9a, where lnt_0.95 is plotted vs. 1000/T, and the apparent activation energy has been determined from the slope of the straight line fitting these data (E_a/R). Here, a value E_a = 52.1 kJ/mol, was obtained for ~95% of transformation from schwertmannite into natrojarosite. Previous studies have determined apparent activation energies for the NaJrs precipitation, whose values vary from 35 to 106 kJ/mol (Dutrizac 1996), which are a range of values expected given the different experimental conditions applied, such as the presence or absence of sodium jarosite seeds or the range of temperature used to calculate the activation energy.

Figure 9

(a) Linear fitting of the time (lnt = 0.95) for a 95% fraction of transformed natrojarosite vs. the reciprocal of temperature. (b) TTT graph corresponding to a fraction Y = 0.95 of the transformed natrojarosite.

Another suitable way to describe the progress of a trans formation is assessing experimentally a time-temperature-transformation (TTT) diagram, which indicates the time at which the transformation occurs when a sample is kept under isothermal conditions (Putnis 1992). Figure 9b displays the TTT diagram determined for the Shm to NaJrs transformation (Y = 95). TTT diagrams are commonly used in mineral science to deal with phase transformations (Putnis 1992; Putnis et al. 2007; Di Lorenzo et al. 2014; Jiménez and Prieto 2015). At low temperatures, the transformation begins after 1 day, and a slight increase of temperature dramatically reduces the time over which the transformation takes place. Given that the transformation rate increases with increasing temperature, the diagrams cannot be expected to have the typical “C” shape. The shape of the curve is similar to the diagrams of transformations that occur with rising temperature (Di Lorenzo et al. 2014; Jiménez and Prieto 2015).