High-temperature study of basic ferric sulfate, FeOHSO4

We report in this paper a new crystal-chemical study of synthetic basic ferric sulfate FeOHSO4. The structure solution performed by the Endeavour program, from new X-ray powder diffraction (XRPD) data, indicated that the correct space group of the monoclinic polytype of FeOHSO4 is C2/c. Selected Area Electron Diffraction (SAED) patterns are also consistent with this structure solution. The arrangement of Fe and S atoms, based on linear chains of Fe3+ octahedra cross-linked by SO4 tetrahedra, corresponds to that of the order/disorder (OD) family. The positions of the hydrogen atoms were located based on DFT calculations. IR and Raman spectra are presented and discussed according to this new structure model. The decomposition of FeOHSO4 during heating was further investigated by means of variable temperature XRPD, thermogravimetry, and differential thermal analysis as well as IR and Raman spectroscopies.


Introduction
The basic iron hydroxyl sulfate, FeOHSO 4 , has attracted in the last decades significant attention in many research fields such as environmental science, catalysis, electrochemistry, mineralogy, and soil science (Cheng and Demopoulos 2004;Fleming 2010;Xu et al. 2010;Reddy et al. 2009). It is also known to form during the thermal decomposition of hydrates of Fe 2+ sulfate (e.g., Swamy and Prasad 1981;Pelovski et al. 1996) and other secondary iron-sulfate minerals (e.g., Ventruti et al. 2014Ventruti et al. , 2016. The crystal structure FeOHSO 4 was solved by Johansson (1962) in the orthorhombic space group Pnma and described as consisting of linear chains of Fe 3+ octahedra cross-linked by SO 4 tetrahedra. More recently, Ventruti et al. (2005) demonstrated, based on the OD theory, that this compound may crystallize under two different polytypes. The first corresponds to the orthorhombic polytype reported by Johansson (1962). The second is a monoclinic form that was proposed by Ventruti et al. (2005) to have P2 1 /c symmetry. The monoclinic polytype was later synthesized by Reddy et al. (2009), who demonstrated that this material has reversible exchange capabilities for Li, and thus excellent electrochemical performance for energy storage applications. The formation of monoclinic FeOHSO 4 was also observed as a by-product at elevated temperatures in the industrial pressure oxidation (POX) hydrothermal treatment of refractory gold ores (Fleming 2009(Fleming , 2010Adams 2016;Dutrizac and Jambor 2007). It has also been documented that during sub-microscopic gold or copper POX hydrothermal extraction, FeOHSO 4 can incorporate up to 9.8-10 wt% of AsO 4 substituting for SO 4 in the structure (Dutrizac and Jambor 2007;Gomez et al. 2013;Paktunc et al. 2013). The solid solution of basic ferric sulfate (BFS) and basic ferric arsenate sulfate (BFAS) has shown promising features in the environmental remediation of arsenic from POX hydrothermal operations of Au and Cu (Swash and Monhemius 1994;Dutrizac and Jambor 2007;Gomez et al. 2011;Bruce et al. 2012;Paktunc et al. 2013). Paktunc et al. (2013) indicated that a residue from POX hydrothermal gold extraction was in the form of an iron hydroxy sulfate with a 5.42 wt % AsO 4 content, and proposed it precipitated in a triclinic (I1) space group based on TEM-SAED investigation. It is worth noting that Transmission Electron Microscopy (TEM) may be used for sub-micron-sized samples to obtain an approximate structure model (e.g., Li and Sun 2017). However, if a sufficient amount of sample is available, structure refinement is better suited via powder X-ray diffraction because of the detrimental effects of dynamical scattering on electron diffraction intensities (e.g., Gemmi et al. 2010). Recently, good results have been obtained applying dynamical theory to electron diffraction structure refinement (e.g., Palatinus et al. 2017).
Thermal analyses and decomposition studies of FeOHSO 4 (e.g., Mahapatra et al. 1990;Pelovski et al. 1996 and references therein), are present in the literature and have been aimed at characterizing the stability of this compound, the reaction pathway of its thermal decomposition and at investigating the environmental variables that control the process under temperature treatment. Although, in most cases, the investigations have been done by combining different techniques and methods (derivatographic, thermogravimetric, powder X-ray phase analysis, and Mössbauer spectroscopy), the results have been subjected to considerable controversy. The inconsistency of some of these results is mainly due to both the order-disorder (OD) character of the FeOHSO 4 structure and to the difficult identification of possible intermediate unstable decomposition products, thus making the process complicated to follow in detail. Another source of uncertainties in the study of these compounds is the extremely complex and variable molecular state of sulfur for different experimental (RH, pH, and T) conditions. Given the importance of these phases and close analogs (e.g., BFAS) in industrial processing/environmental compliance and potential energy applications, the present work was aimed at characterizing the structure of synthetic FeOHSO 4 via XRPD and TEM-SAED techniques. Furthermore, we investigate the decomposition of FeOHSO 4 during thermal treatment using a combination of in situ techniques.

Experimental methods
The FeOHSO 4 monoclinic phase was prepared following the method described by Reddy et al. (2009). Accordingly, the starting product consisting of reagent-grade ferrous sulfate heptahydrate, FeSO 4 ·7H 2 O, was ground in an agate mortar to a fine powder, transferred to an alumina boat, and dehydrated in the air at 280 °C for 7 days.
XRPD data were collected in reflection mode from room temperature (RT) to 900 °C using a Panalytical Empyrean diffractometer equipped with a PIXcel3D RTMS detector and an Anton Paar HTK1200N furnace. The sample was ground in an agate mortar, passed through a 400-mesh sieve, and loaded into an alumina sample holder. The sample was irradiated with X-rays generated by a copper long-fine focus tube operated at 40 kV and 40 mA. A 0.125° divergence slit, a 0.25° anti-scattering slit, and a Soller slit (0.02 rad) were mounted in the incident beam pathway. The diffracted beam pathway included a Ni filter, a Soller slit (0.02 rad), and an anti-scatter blade (7.5 mm). XRPD data were collected in the angular range 3-70° (2θ) using a virtual step scan of 0.026° and a counting time of 360.0 s/ step. The program ENDEAVOUR (Putz et al. 1999) was used for ab initio structure determination. The GSAS (Larson and Von Dreele 2000) software was used for the unit-cell parameter refinement, the Rietveld analysis of the structure, and the quantitative phase determination. DFT calculation using the CRYSTAL09 (Saunders et al. 2003) program package was used to locate the hydrogen position.
For transmission electron microscopy (TEM) observation, an aliquot of the untreated sample used for XRPD analyses was dispersed in isopropanol and ultrasonicated, and few microliters of the suspension were pipetted onto a holey-carbon, 300-mesh, Cu-grid. TEM observations were performed at the Department of Physical Sciences, Earth and Environment of the University of Siena with a Jeol JEM 2010 instrument. The instrument operates at 200 kV, and is equipped with an Oxford Link energy dispersive system (EDS) and an Olympus Tengra 2.3 k × 2.3 k × 14-bit slowscan CCD camera. Semi-quantitative EDS analyses were obtained with the standardless method using the approximation of Cliff and Lorimer (1975).
Simultaneous thermal gravimetric analysis (TGA) and differential thermal analysis (DTA) were conducted in an N 2 atmosphere (flow rate 60 mL/min) with an STA7200 analyzer (HITACHI). The powdered sample (~ 22 mg) was put into an alumina pan and heated from room temperature (RT) to 900 °C with a heating rate of 10 °C/min. Sintered α-Al 2 O 3 powder was the thermally inert reference material used. To simultaneously follow the evolution of the gaseous decomposition products over the temperature range investigated, the thermobalance was connected to a Fourier transform infrared (FTIR) spectrometer through a heated line. The balance adapter, the transfer line, and the FTIR gas cell were kept to about 220 °C to avoid possible condensation of the evolved gases. N 2 was used as a carrier gas, because it is inert and difficult to ionize and will not interfere with the infrared spectra of the evolved product gases. FTIR spectra of the studied FeOHSO 4 compound were collected with the conventional pelletized sample using a Nicolet iS50 spectrometer equipped with a Globar IR source and a KBr beamsplitter. In situ HT-FTIR measurements were done using pure mineral powders employing a Linkam T600 heating/freezing stage fitted in a Bruker Hyperion 3000 FTIR microscope attached to a Vertex 70 V optical bench and equipped with a 15X Schwarzschild objective and an MCT N 2 -cooled detector. The sample powder was finely dispersed on a CaF 2 sample holder; the nominal resolution was set at 4 cm −1 and 128 spectra were co-added for both mineral and background. Although a few crystals were illuminated with a beam diameter of 30 μm, the signal in the principal OH-stretching region was saturated, and thus, the dehydration of the sample was monitored using the combination band at 4470 cm −1 (see below).
Unpolarised Raman spectra were collected using a Labram Micro-Raman spectrometer by Horiba, equipped with an Olympus BX40 confocal microscope, and a Peltier cooled (~ 200 K) 1024 × 256 pixel CCD detector. Raman scattering was excited by a He-Ne laser at 632.8 nm (nominal output power 18 mW). The system achieves the high contrast required for the rejection of the elastically scattered component by the use of edge filters. The backscattered light is dispersed by an 1800 line/mm grating, for a nominal resolution of ~ 1 cm −1 . In situ HT Raman spectra were collected by means of a Linkam T600 heating/freezing stage placed on the microscope sample holder.

RT X-ray structure solution and TEM analysis
The XRPD pattern collected at RT for the studied sample ( Fig. 1) was fully indexed using the well-known and popular indexing algorithm TREOR90 (Werner et al. 1985) yielding a monoclinic solution with approximate cell parameters (in Å) a ~ 7.39, b ~ 7.14, c ~ 7.39, and β (°) ~ 120. Subsequent Le Bail refinement was performed in GSAS to refine cell parameters and to extract diffraction lines. A small peak, likely corresponding to a residue of unreacted starting materials used in the synthesis, was observed at about 18° (2θ). This impurity peak was excluded from the refinement. The resulting peak list was imported into Endeavour along with the contents of the unit cell to solve the structure. During the structure solution calculations, the Endeavour software matches the calculated diffraction pattern of each trial crystal structure to the experimental diffraction pattern. All atoms but the hydrogen of the hydroxyl group were successfully determined in the C2/c space group. At this stage, the structure was refined by the Rietveld method using the GSAS program (Fig. 1). During the refinement, the isotropic displacement parameters of O1, O2, and O3 oxygens were constrained to be equal. The details of the Rietveld refinement are presented in Table 1, while the fractional coordinates and selected bond lengths are reported in Tables 2 and 3, respectively. The resulting atomic arrangement of the unit cell is schematically presented in Fig. 2a. The structural details agree with the previous models (Ventruti et al. 2005), showing that the common building block to all members of the FeOHSO 4 OD family is represented by a chain of trans corner-sharing FeO 4 (OH) 2 octahedra. SO 4 tetrahedra alternate along the chain and interconnect adjacent chains by corner sharing all tetrahedral oxygen to form a 3D openframework. The position of the hydrogen atoms (Table 2)  was obtained from electronic structure calculations using density functional theory (DFT). The experimentally determined unit-cell parameters and the heavy atom positions were used as an input geometry, and an approximate starting position for the hydrogen atom was employed. The resulting environment of the hydroxyl group at O2 position (Table 2) is shown in Fig. 2b. The O-H vectors are perpendicular to the ferric chain and oriented towards the O1O1 edge of the SO 4 group. The two symmetric hydrogen-bond distances H..O1 of about 2.21 Å suggest relatively weak interactions around the OH position. The powdered sample used as starting material for XRPD experiments was then examined by TEM. We generally observed that the powder grains have lamellar habitus-with the longest dimension usually confined within some hundreds of nm. Furthermore, the EDS spectrum was found to be consistent with the expected composition (Fig. 3). SAED patterns taken on different grains are consistent with the C2/c space group. It is worth to note that SAED patterns may indicate a C2/m symmetry because of the presence of h0l, l = 2n + 1 reflections, which should be systematically absent from the presence of the c-glide (Fig. 3). However, it is well known that in electron diffraction experiments, such reflections can be present due to dynamical scattering.

Heating experiments on FeOHSO 4 : TG and XRPD
Two main endothermic peaks are observed in the DTA curve (Fig. 4a) at 536 and 682 °C, respectively. FTIR spectra corresponding to the evolution of the gas involved during these two events were selected and analyzed. The first endothermic peak corresponds to the loss of the hydroxyl group. Indeed, the observed mass decrease in the thermal range from room temperature up to 530 °C is about 10 wt%, very close to the expected theoretical value for the complete loss of one water molecule. Unfortunately, the very small concentration of evolved water vapor did not allow the detection of a significant FTIR signal. In the FTIR spectrum (Fig. 4b), associated with the second endothermic peak, the absorption bands of SO 2 (at 1374, 1359, 1345, and 1166 cm −1 ) were identified. Therefore, the second mass loss of about 42% observed in the thermal range 570-750 °C is associated with the disruption of SO 4 groups. Our results are in agreement with a combined thermogravimetric-mass spectroscopy analysis carried out by NETZSCH (AZO Materials 2009) that found a mass number of 18 and 64 plus 32 a.m.u. for the first and second mass loss, respectively.
X-ray diffraction data (Fig. 5) show that the crystal structure of FeOHSO 4 is not significantly affected by the thermal treatment up to 400 °C, since the diffraction patterns are virtually identical in both peak multiplicity and intensities.  Inspection of the individual X-ray patterns shows that the FeOHSO 4 phase is the only phase in the assemblage up to at least 400 °C (Fig. 5). The evolution of the lattice parameters and cell volume determined by GSAS as a function of T is displayed in Fig. 6. In general, we observe that the axial expansion is strongly anisotropic. Up to 400 °C, there is a decrease of the a cell parameter, while the b and c cell parameters increase linearly. However, for all cell edges, the trend departs from linearity at T > 400 °C. This is particularly evident for b and the β angle that show a sudden increase up to 500 °C. The evolution of the cell volume (Fig. 6, Table S1), averaging the variation of the individual cell edges, shows a cumulative linear increase up to 400 °C. A discontinuity in the positive variation of cell volume is observed at about 490 °C in correspondence of the phase transition discussed below. The thermal expansion coefficients for FeOHSO 4 calculated by least-squares linear regression analysis between RT and 400 °C are α a = − 2.51 9 10 -6 K −1 , α b = 26.29 10 -6 K −1 and α c = 10.6 10 -6 K −1 , α V = 33.0 10 -6 K −1 . Above 400 °C, the basic iron hydroxyl sulfate undergoes a dehydroxylation process through two different mechanisms.
Initially, at about 450 °C, the most intense diffraction peaks of α-Fe 2 (SO 4 ) 3 (Christidis and Rentzeperis 1976) appear in the pattern. No other phases were detected in the X-ray patterns up to 490 °C. Quantitative phase analysis via Rietveld refinement indicates that FeOHSO 4 partially decomposes to α-Fe 2 (SO 4 ) 3 as represented by the following equation: At around 490 °C, most of the ferric hydroxy sulfate undergoes an abrupt dehydroxylation process. This causes a new compound, the ferric oxysulfate Fe 2 O(SO 4 ) 2 (Sun et al. 2014), to form according to the reaction: The weight fractions of the phases occurring during the whole heating treatment refined employing the Rietveld method are shown graphically in Fig. 7 and indicate that most of FeOHSO 4 decomposes to ferric oxysulfate (Fe 2 O(SO 4 ) 2 ) in a very short temperature range, while only a small fraction of FeOHSO 4 decomposes to the rhombohedral ferric sulfate (Fe 2 (SO 4 ) 3 ) in a slightly more extended range.   c-e SAED patterns along some low-indexed orientations taken on different grains. All patterns can be indexed according to the C2/m space group. Indexing in the C2/c space group is possible, once it is assumed that h0l reflections with l = 2n + 1 that should be absent by symmetry are indeed present because of dynamical scattering (circles). Note that in (e), the [001] and [101] orientations cannot be distinguished (indexing according to the latter) Indeed, the rhombohedral ferric sulfate increases gradually up to 580 °C that corresponds to the complete decomposition temperature of FeOHSO 4 , and then decreases until it disappears at 700 °C (Fig. 5b). At 580 °C, the relative amount of α-Fe 2 (SO 4 ) 3 calculated from the Rietveld analysis of the powder X-ray pattern is around 18%.
The ferric oxysulfate is stable in a very limited thermal range of 490-550 ºC (Fig. 7). We observed that the ferric oxysulfate fraction increases quickly up to 500 °C, which corresponds with the abrupt dehydroxylation of FeOHSO 4 , and then suddenly drops off. Hence, suggesting this phase is stable in a narrow temperature range under the used experimental conditions. Both phase transformations of FeOHSO 4 are associated with the loss of one water molecule, thus implying that the heating treatment induces the breakdown of the sulfate chain. In particular, the loss of the hydroxyl group shared by adjacent Fe 3+ O 4 (OH) 2 octahedra in the linear chain of the FeOHSO 4 structure causes mainly the formation of dimers of edge-shared Fe 3+ O 6 octahedra present in the Fe 2 O(SO 4 ) 2 structure, and, in lesser part, the building of a 3D framework of Fe 3+ O 6 octahedra sharing all vertices with SO 4 groups to form rhombohedral Fe 2 (SO 4 ) 3 .
For T > 500 °C, the α-Fe 2 O 3 iron oxide (hematite) appears in the assemblage and increases quickly to become the predominant phase for T > 600 °C, and the only component for T > 700 °C. The total mass loss observed in the thermogravimetric plot (Fig. 4) is in agreement with the expected theoretical value 52.7 wt% associated with the water and sulfur dioxide emission according to the overall reaction process:

The RT vibrational (FTIR and Raman) spectra of FeOHSO 4
The RT FTIR and Raman spectra collected for the studied FeOHSO 4 sample are given in Fig. 8. To the best of our knowledge, no complete Raman data exist in the literature for this compound except an RT Raman spectrum in the limited wavenumber range of SO 4 internal modes (Connor and Wang 2014); FTIR spectra, on the other side, have been published in some previous papers (Powers et al. 1975;Majzlan et al. 2011;Gomez et al. 2013). In the principal OHstretching region (4000-3000 cm −1 ), a single and very sharp peak (Fig. 8a) is present in the Raman pattern, at 3447 cm −1 , while for the powder FTIR spectrum, this band occurs at 3454 cm −1 . The FTIR band is, however, broader with respect to the Raman peak, and strongly asymmetric, possibly due to an overlapping component on the lower wavenumber side, at around 3320 cm −1 (Fig. 8b). In the 4000-5000 cm −1 NIR region, a unique band due to the combination of the stretching and libration modes ( OH ν 3 + OH δ) is observed at 4470 cm −1 (Fig. 9). No bands are observed at wavenumbers > 5000 cm −1 , ruling out any presence of H 2 O molecules in the structure (Della Ventura et al. 2009Ventura et al. , 2013Ventruti et al. 2015Ventruti et al. , 2018. Considering a single OH ν 3 at 3454 cm −1 , an OH libration around 1020 cm −1 is expected based on the spectrum of Fig. 9. The spectra in the low-wavenumber (< 1200 cm −1 ) range show the typical modes resulting from vibrations of the SO 4 group, the Fe-O octahedra, and the libration of the OH groups (δ OH ) (e.g., Powers et al. 1975). For sulfate minerals, if the ideal T d symmetry of the SO 4 group is lowered due to the interaction with the surrounding atoms, the degeneracy of the ν 3 mode is lifted, and the vibrations inactive in tetrahedral symmetry become IR active (e.g., Majzlan et al. 2011). As  Evolution of the phase amounts refined from the Rietveld method based on the powder X-ray diffraction patterns in the RT-900 °C range a consequence, each crystallographically distinct SO 4 tetrahedron contributes up to four bands (three ν 3 + one ν 1 ) to the spectra. The IR spectrum of FeOHSO 4 , in particular, has been examined by Powers et al. (1975) who assigned four sulfate bands to these modes in the 1172-1058 cm −1 range, plus the δ OH at 1020 cm −1 . Two peaks at 650 and 638 cm −1 , respectively, were assigned to the ν 4 modes, while several bands in the 400-600 cm −1 range were tentatively assigned to Fe-O modes. The spectrum collected here is very similar in multiplicity and band position to that reported by Powers et al. (1975), except for a few differences in band intensity of the ν 3 -SO 4 higher wavenumber region.
Assignment of the peaks observed in the Raman spectrum can be based on comparison with the IR pattern and with data from the literature on similar compounds (Sasaki et al. 1998;Frost et al. 2006). Accordingly, the peaks at 1186, 1098, and 1061 cm −1 in the Raman pattern are assigned to the ν 3 + ν 1 SO 4 modes. Meanwhile, that at 1028 cm −1 is assigned to the δ OH mode (Sasaki et al. 1998), while the ν 4 SO 4 mode is found at 650 cm −1 and those in the 600-200 range to the ν 2 SO 4 and Fe-O octahedral modes.

Heating experiments on FeOHSO 4 : vibrational (Raman and FTIR) spectroscopies
The Raman spectra collected in situ both in the lattice mode region (< 1200 cm −1 ) and the 3000-4000 cm −1 O-H stretching region are given in Fig. 10. The Raman pattern in the fingerprint region remains almost unaffected at least up to 300 °C. However, the appearance of an intense fluorescence background, due to energy emission in the IR spectral region from the heating stage, strongly interferes, preventing to extend the limit for the basic ferric hydroxyl sulfate stability beyond 400 °C, as identified by TG and XRPD data. However, the OH-stretching band is less affected by thermal treatment, the only appreciable effect being a slight broadening and a small reduction of its intensity. At higher temperatures (> 500 °C), the collected spectra are dominated by fluorescence. To overcome this bottleneck, Raman spectra have been collected at RT after annealing the sample. Figure 11 displays the spectra labelled in terms of the target temperature of each annealing step. It is worth noting that the Raman spectrum collected at RT after the sample was annealed at 400 °C completely recovered its shape, while for T exceeding 500 °C, it started to change. In particular, a broad band appears in the 400-500 cm −1 range, while still showing slightly broadened sulfate modes, a trend that is confirmed at 580 °C where the spectral signature in the framework mode region is completely wrecked and a very broad OH contribution is still visible. Interestingly, this is the T range where XRPD data show the onset of metastable sulfate phases and, because the Raman spectra are collected at RT after the sample was annealed, we can infer that this represents the starting point of irreversibility for Fig. 10 Raman spectra for the studied FeOHSO 4 collected in situ for increasing temperature (°C, indicated on each pattern): a low-frequency mode region and b OH-stretching region the reaction. The Raman spectrum collected after annealing the sample at 650 °C shows, in agreement with XRPD data, the unique presence of hematite Fe 2 O 3 , that is the new stable phase after completion of the dehydration process.
The loss of H during the thermal treatment of FeOHSO 4 was monitored by in situ FTIR micro-spectroscopy. Based on the previous experience on dehydration experiments on different substances (e.g., Della , we collected the data on pure powder deposited on a sample holder. This choice was preferred instead of using the classical technique of embedding the sample within an IR transparent medium (i.e., KBr) to avoid possible ambiguities due to incomplete loss of H across the compressed matrix. Because of the saturation of the detector in the principal OH-stretching region, even for small amounts of powder under the IR beam, we monitored the intensity of the NIR band (Fig. 9). The evolution of the peak intensity as a function of temperature is given in Fig. 12a which shows a sudden loss of the O-H signal for T > 500 °C that is completed in a short T range (ca 50 °C). Interestingly, the data show a significant intensity increase close to the verge of dehydration. Such a feature cannot be due to an increase of the OH content during the thermal treatment; it has been observed in other studies on the thermal behaviour of Fe-rich hydroxy-silicates (e.g., Della ) and has been explained as due to enlarged amplitudes of hydrogen atomic displacement for increasing T. Figure 12b shows that a very slight intensity increase is possibly also visible in the Raman spectra. Figure 12 thus suggests that such a feature is common to different compounds where the H loss is connected to a structural rearrangement involving the Fe electronic properties, as observed for some Fe silicates (Della ).

Fig. 11
Raman spectra for the studied FeOHSO 4 collected ex situ for increasing temperature (°C, indicated on each pattern): a low-frequency mode region and b OH-stretching region Fig. 12 Evolution of normalized intensity (area) as a function of T of a the NIR peak at 4472 cm −1 and b the OHstretching band at 3447 cm −1 in the Raman spectra