Stability of hydrotalcite (Mg-Al layered double hydroxide) in presence of different anions Cement and Concrete

MgO contained in cementitious materials is experimentally observed to precipitate as poorly crystalline hydrotalcite (Mg-Al LDH). However, the geochemical modelling of hydrotalcite is challenging due to the lack of consistent thermodynamic dataset for this phase. Hydrotalcites with Mg/Al = 2 were synthesised in the presence of NaHCO 3 , NaCl, Na 2 SO 4 , NaNO 3 , and NaOH. Mass balance, XRD and FT-IR indicated the incorporation of the anions in the interlayer together with some carbonate. The crystallinity of the phase increased depending on the anion: SO 42 (cid:0) < Cl (cid:0) < NO 3 (cid:0) < OH (cid:0) < CO 32 (cid:0) . An in-situ increase of temperature monitored by XRD and TGA showed that the stability of the hydrotalcite structure with temperature also depended on the incorporated anion. The solubility products were calculated based on the solution analysis of samples re-equilibrated at different temperatures, while the entropy and heat capacity were obtained from the additivity method or the molar volume. A simple solid-solution model for hydrotalcite containing CO 3 , OH, SO 4 , Cl and NO 3 is suggested.


Introduction
In the context of lowering the 5 to 8% anthropogenic CO 2 emissions from cement industry [1], particularly from the Portland cement manufacture, alternative binders or partial replacement of the clinkers by so called supplementary cementitious materials (SCMs) are studied. The use of reactive aluminosilicate containing SCMs, such as fly ash, blast furnace slag or calcined clays, increases the aluminium and silicate content, which can affect the phase composition in the hydrated cement and increase the Al and Si incorporation into the calcium (alumino-) silicate phases (C-(A-)S-H) [2]. Also, the content of magnesium is expected to raise in future CO 2 -reduced clinkers, either from the use of blast furnace slag or of magnesium silicate minerals. Magnesium from slags or from other MgO sources is observed to precipitate as hydrotalcite, i.e. a Mg-Al layered double hydroxide (LDH) [3][4][5][6][7][8][9], although its exact composition often remains unclear.
Hydrotalcite-like phases are Mg and Al based layered double hydroxide solids (Mg-Al-LDH) with an isomorphic substitution of Mg 2+ by Al 3+ in the brucite-like sheet, that generates positive charges in the main layer. The main layer can have a variable Mg/Al ratio [10,11], Al can be replaced by iron [12,13], and different interlayer anions can be contained in the interlayer [10]. The positive charge of the main layer is compensated by anions and water in the interlayer region [14]. Mg-Al-LDH can be written as [Mg l-x Al x (OH) 2 ] x+ [A x/n n− mH 2 O] x− , with 0 < x < 0.33 [14]; "A" indicates the presence of anions such as OH − , Cl − , NO 3 − , CO 3 2− , SO 4 2− in the interlayer. The affinity is CO 3 2− > SO 4 2− for divalent anions and OH − > F − > Cl − > Br − > NO 3 − > I − for monovalent anions [14]. While crystalline LDHs are widely studied and used as catalysts, anion exchangers and adsorbents [15], LDHs precipitate semiamorphous in cement pastes and the formation of different solid solutions is observed [5]. Not only the incorporation of OH − is detected but also the uptake of chloride and/or carbonate by hydrotalcite which seems to reduce the chloride and carbonation ingresses in hydrated Abbreviations: ATR FT-IR, attenuated total reflectance Fourier-transform infrared; C-(A-)S-H, calcium (alumino-)silicate hydrate; C p • , heat capacity; FWHM, full width at half maximum; K s0 , solubility product; LDH, layered double hydroxide; MAS NMR, magic-angle spinning nuclear magnetic resonance; M-(A-)S-H, magnesium (alumino-)silicate hydrate; Mg-Al-LDH, Mg and Al based layered double hydroxide = hydrotalcite; PC, Portland cement; S • , entropy; SCMs, supplementary cementitious materials; SSA BET , specific surface are obtained by using the Brunauer-Emmett-Teller equation; TGA-FT-IR, thermogravimetric analysis complemented with Fourier Transform Infrared detection of evolved gases; V cell , cell volume; V • , molar volume; XRD, X-ray powder diffraction. cements, thus counteracting the potential degradation of reinforcement in concrete [16,17]. An uptake of SO 4 2− by hydrotalcite at the interface of cement in direct contact with Opalinus Clay might be possible [18,19], as well as the formation of hydrotalcite conceivably containing NO 3 − in the presence of magnesium aluminium silicate hydrates [20].
The variable composition both in terms of Mg/Al and Mg/(Al + Fe) ratio in the main layer and of interlayer anions of the Mg-Al LDHs in cement pastes, hindered the development of adequate thermodynamic data for hydrotalcites, limiting the use of thermodynamic approaches for studying the magnesium precipitation in the cement pastes. In Cem-data2018 chemical thermodynamic database [21] two different Mg-Al LDHs were considered: for Portland cement (PC) a relatively stable OH-hydrotalcite with a fixed composition is suggested, while for alkali activated slags a series hydroxide LDHs [7] with variable Mg/Al and a different stability are suggested. The carbonated LDHs suggested by [12] were not found to be stable under any condition investigated leading to the estimation of revised thermodynamic properties of carbonatehydrotalcites in [22,23]. A recent study suggested variable solubility products for hydrotalcite depending on the pH values used [24]. The use of different models for different situations [21] as well as the definition of pH dependent solubility products [24] suggest a lack of adequate thermodynamic models to capture the variable composition in Mg/(Al + Fe) and anion content.
This study investigates Mg-Al LDHs with a Mg/Al = 2 and different interlayer anions. Mg-Al-LDHs containing NO 3 − , Cl − , SO 4 2− , OH − and CO 3 2− were synthesised, characterised by X-ray powder diffraction (XRD), thermogravimetric analysis complemented with Fourier Transform Infrared detection of evolved gases (TGA-FT-IR), attenuated total reflectance Fourier-transform infrared (ATR FT-IR) spectroscopy and 27 Al and 23 Na MAS NMR spectroscopies. The structure and the stability of the hydrotalcites containing different anions were also studied by insitu XRD in a heater-controlled chamber. The samples were reequilibrated at 7, 23, 40 and 80 • C and for each sample, the solution was analysed and the solubility products were calculated.

Materials and synthesis
Hydrotalcite (652288 Sigma Aldrich CAS Number: 11097-59-9) was used as a source of MgO and Al 2 O 3 , although the composition given on the product was Mg 6 Al 2 (CO 3 )(OH) 16 4H 2 O, a Mg/Al about 2 was found by SEM/EDS. This commercial hydrotalcite (Ht-CO 3 *) was decarbonated and dehydrated for 5 h at 700 • C to remove physically and chemical bound water as well as carbonate resulting in a reactive amorphous solid magnesium aluminate: Mg 4 Al 2 O 9 (M 2 A) as shown in Fig. 1.
The M 2 A was added to different solutions containing NaHCO 3 (VWR chemicals, ACS, Reag, purity >99.5%), NaCl (Merck, for analysis, purity >99.5%), Na 2 SO 4 (VWR chemicals, ACS, Reag, purity >99.5%), NaNO 3 (Merck, for analysis, purity >99.5%), or NaOH (Merck, pellets for analysis, purity >99.5%) with a liquid/solid mass ratio equal to 10. The solutions were prepared according to the amounts summarized in Table S1, using twice the stoichiometric amount of anions to ensure sufficient incorporation of the anion in the interlayer. These Na-based salt solutions can easily be carbonated and this should be avoided. The samples were equilibrated at 80 • C during 20 days to ensure and to fasten the precipitation. The solutions and solids were separated by vacuum filtration using nylon filters (0.45 μm); the solids were washed with 50/50 (volume) water-ethanol to remove dissolved ions. The samples were dried by freezing with liquid nitrogen under vacuum for 2 days in a freeze dryer. The freeze drying minimizes carbonation as it removes free water efficiently. The samples were kept in desiccator over CaCl 2 saturated solution and KOH in order to keep a relative humidity about 34% and to minimise the carbonation until the solid characterisation.
2 g of different hydrotalcites prepared were put back in 50 mL of milliQ-water and equilibrated at 7, 20, 40 and 80 • C for 6 months to measure their solubility. The liquid/solid separation was performed following the method detailed above. However, the samples were not freeze dried to avoid the removal of interlayer water, but dried in a desiccator over silica gel and KOH solution during 2 weeks in order to minimise the carbonation. The solid was characterised and the concentration of the different elements in solution was measured and used to calculate the solubility data at the different temperatures.

Powder X-ray diffraction
Diffraction data were collected from 5 to 75 2θ with a PANalytical X'Pert Pro MPD diffractometer, equipped with a Cu X-ray source (40 kV, 40 mA) with a fixed divergence slit size and an anti-scattering slit on the incident beam of 0.5 • and 1 • and an X'Celerator detector. In-situ hightemperature measurement were carried out on samples re equilibrated at 80 • C on the same diffraction device using a heating chamber HTK 1200 N by Anton Paar. Diffraction data for whole pattern analysis were collected from 5 to 70 2θ with a fixed slit (0.25 • ) and step size of 0.0167 2 • θ. The heating rate applied was 1 • /s. After the target temperature was reached, the sample was given 10 min to equilibrate. Measurement was performed in uneven intervals of 25 • C (25-100 • C), and 20 • C (100-500 • C) plus a final scan at 600 • C. Diffraction data of the 25 • C scan was used to calculate the lattice parameters and cell volume (V cell ) in HighScorePlus using the Pawley method. The R wp -value ranged from 7.4-10.3 from the best to the worst fit.

Evolution of the basal spacing of the hydrotalcite samples with temperature
Measurement settings were 8-16 2θ with a step size of 0.0167 2θ, a scan step time of 17.78 s, and a heating rate of 1 • /s. Diffraction data was collected in steps of 1 • C ranging from 100 to 300 • C. The basal spacing, peak intensity and FWHM of each measurement was extracted with a python script using the savgol_filter and curve_fit functions of the SciPy library. Specimen displacement was corrected using the offset provided by the manufacturer (thermal expansion) and by adjusting the values to the ones obtained from whole pattern analysis.

N 2 -sorption isotherm for specific surface area calculation
Prior to measurements, 40 mg of a sample was dried at 55 degrees overnight in a vacuum using a Microtrac Belprep Vac III. SSA measurements were then carried out using a Microtrac Belsorp Mini X measuring the full isotherm. The specific surface area (SSA BET ) was obtained by using the Brunauer-Emmett-Teller (BET) equation [25], in the pressure range from 0.05 to 0.2 P/P 0 .

Thermogravimetric analysis complemented with Fourier Transform-Infrared gas analysis
TGA measurements were carried out using a Netzsch STA 449 F3 Jupiter TGA apparatus coupled with a Bruker Fourier-transform infrared (FT-IR) spectrometer for the analysis of the exhaust gases. Approximately 40 mg of each sample was heated from 30 to 980 • C with a heating rate of 20 • C per minute in 150 μL alumina crucibles. The

Attenuated total reflectance Fourier Transform-Infrared
Attenuated total reflectance (ATR) Fourier Transform-Infrared (FT-IR) spectra were recorded in the mid-region on a Bruker Tensor 27 FT-IR spectrometer between 600 and 4000 cm − 1 with a resolution of 6 cm − 1 by transmittance on small amounts of powder. Spectra were background corrected and scaled to ease comparison: on the 1520-1700 cm − 1 (H-O-H bending) and 3400-4000 cm − 1 (O-H stretching vibration) from the water loss quantify by TGA between 30 and 250 • C to ease comparison.

27 Al magic-angle spinning (MAS) solid state nuclear magnetic resonance (NMR)
Solid state 27 Al MAS NMR spectra were measured using a 2.5 mm CP/MAS probe on a Bruker Avance III NMR spectrometer. The 27 Al MAS NMR single pulse experiments were recorded at 104.3 MHz applying the following parameters: 25 ′ 000 Hz sample rotation rate, between 2000 and 4000 scans depending of the content of aluminium in the samples, π/2 pulses of 1.5 μs, 0.5 s relaxation delays (identical spectra were obtained when relaxation delays of 0.2, 0.5 & 1.0 s were applied), without 1 H decoupling. The 27 Al NMR chemical shifts of the were referenced to an external sample of Al(acac) 3 . The 27 Al MAS NMR spectra were analysed by the line shape fitting software "DMFIT" [26]. Generally, the fitting of the octahedral sites was performed using i) a Lorentzian shape at 9 ppm (line widths of ca. 270-450 Hz) and ii) a quadrupolar broadened shape using the "Czjzek simple" [27] as detailed in [20].

Energy dispersive spectrometry (EDS)
The semi-quantification of Mg, Al, Cl, S was done with Zeiss Sigma 500 VP FE-SEM with Oxford Instruments Aztec Energy Advanced Xmax 150 EDS detector. The working distance used for the point analysis was 8.5 mm. To avoid charging artefacts, the powder samples were carbon coated using an evaporative coater. The EDS measurements were calibrated by measuring Co and 50 points were recorded for each sample.

Analysis of the solutions
After the collection of each syringe sample, the pH was measured in a 50 μL aliquot with a Thermo Scientific™ Orion™ PerpHecT™ ROSS™ Combination pH Micro Electrode, and the rest of the sample was kept in refrigerator until further analysis.
Total dissolved inorganic carbon (TIC) was determined by infrared spectrometric techniques using an Analytic Jena Multi N/C 2100S equipped with an infrared NDIR-detector and an APG autosampler and supported by the Software multiWin (aj). TIC is determined directly by oxidation of the dissolved inorganic to CO 2 using 10% phosphoric acid, which is diluted from 85% phosphoric acid (p.a., Merck), injecting the released CO 2 with synthetic air and analysing with the NDIR infrared detector.
The samples were analysed for major cations by ICP-OES on a Varian 720 ES equipped with an autosampler Varian SPS-3 supported by the ICP Expert II Ver. 1.1.2 software and Varian Spectroscopy Database Administrator Ver. 1.6.0.20. Solution samples were diluted with 1% HNO 3 by a factor of 100 before injection. The samples were analysed for the anions (chloride, nitrate, sulphate) by ion chromatography using a Metrohm ProfIC AnCat MCS IC system with automated 5 μL and 50 μL injection loops.

Thermodynamic calculations
The solubility products were calculated from the measured concentrations using the Gibbs free energy minimization software GEMS [28]. GEMS is a broad-purpose geochemical modelling package that computes Table 1 Standard thermodynamic properties (25 • C) and molar volumes of the phases considered in this study with indication of references; some of them are summarized in [21] and directly available in the cemdata18 database, others are estimated from the literature available and ion-exchange constant measured by [14] Calculated using the additivity method, based on the value given in [5], composition corresponds to Mg/Al reported in [14]. c Values given in (Prentice et al., 2020). d Recalculated using the reported concentrations, with the exception of Mg, where [Mg] at equilibrium with brucite was considered, as measured Mg concentrations were more than 10 times higher, indicating analytical problems. e Calculated from additivity based on [14] and OH-hydrotalcite with Mg/Al = 2.35. f Log K calculated from Δ f G • given in [38]. equilibrium phase assemblage and speciation in a complex chemical system from its total bulk elemental composition. The thermodynamic data for aqueous species and for brucite (Mg(OH) 2 ) were taken from the GEMS version of the PSI/Nagra thermodynamic database [29], data for microcrystalline aluminium hydroxide (microcrystalline Al(OH) 3 ) from [21,30], as summarized in Table 1.
Due to the variable composition of hydrotalcite, only few thermodynamic data are available and vary strongly between different studies. The data found in literature are summarized in Table 1. While [5] derived a very stable Mg 4 Al 2 (OH) 14 ⋅3H 2 O (logK s0 = − 56.0), Myers et al. [7] used measured concentrations from short term experiments of a few hours (up to 12 h) [31] to calculate the solubility product for three hydroxide based end-members for MA-OH-LDH Mg 4 Al 2 (OH) 14 [14] using the additivity method [32], which are several log units more stable than the data reported by [13]. Table 1 also contains data derived here from additivity methods from [14]. A recent publication from Prentice et al. [24] gave thermodynamic data for hydrotalcite containing carbonate, hydroxide and sulphate and these authors observed a dependence of solubility products on pH. Their calculations involved magnesium concentrations measured well above the brucite solubility; based on their measured concentrations the log K s0 were recalculated assuming magnesium concentrations at equilibrium with brucite and these are given in Table 1.
The thermodynamic data between 5 and 100 • C at 1 bar were obtained based on the temperature dependence of the apparent Gibbs free energy of formation: where a 0 , a 1 , a 2 , and a 3 designate the empirical coefficients of the heat capacity equation C p o = a 0 + a 1 T + a 2 T − 2 + a 3 T − 0.5 and T 0 the reference temperature of 298.15 K, see [39] and the online documentation of GEMS by Kulik [40]. The apparent Gibbs free energy of formation Δ a G T0 o refers to the Gibbs free energies of the elements at 298 K, and Eq.
(2) is built into the GEMS code.

Solid characterisation of hydrotalcite synthesised at 80 • C
The XRD data of the hydrotalcite samples (i.e. Ht-CO 3 , Ht-Cl, Ht-SO 4 , Ht-NO 3 , Ht-OH) synthesised at 80 • C are shown in Fig. 2; the main reflection peaks of hydrotalcite are indexed in Table 2 based on [41]. Richardson gathering numerous Mg-Al layered double hydroxide crystal structure [11] derived linear relationship between the cell parameter a and the Al/(Mg + Al) in the hydrotalcite's crystal structure.
While the reflection peaks are well defined for the Ht-CO 3 , broader and asymmetrical peaks related to the basal plane were observed in the other samples indicate smaller crystallites or less ordered structures. The broadness of these peaks increased depending on the anion incorporated: In agreement with our observations, literature reports an increase of the basal spacing together with a broadening of the peak compared to CO 3 -hydrotalcite, when Cl − , NO 3 − and SO 4 2− is present in the interlayer of hydrotalcite [43,44].
The SSA was calculated from the nitrogen sorption and the results are given in Table S2. The values were between 18 and 33 m 2 g − 1 and no increase with the broadening of the reflection peaks in the XRD patterns was observed. The broadening of the reflection peaks is therefore rather linked to crystallinity of the phases than to their size.
The drying procedure can potentially affect the measured (0 0 l) distance, as strong drying can remove water present in the interlayer. However, although the Ht-CO 3 sample had been freeze-dried before analysis, it showed the same sharp (0 0 3) reflection corresponding to a basal spacing of 7.63 Å which is characteristic of common carbonated hydrotalcite, e.g. [43,45,46] as the Ht-CO 3 * samples, indicating little effect of the freeze drying on the basal spacing. The basal spacing of the Ht-OH sample was measured at about 7.65 Å which is within or slightly higher than the range of previously reported values 7.55-7.67 Å [24,43]. This could be due to a small carbonation contamination (see below, TGA data).
The (0 0 3) reflection of the other hydrotalcite samples (Ht-Cl, Ht- NO 3 and Ht-SO 4 ) is split in two signals: a first signal relatively constant at d ≃ 7.65 Å (11.56 2 • Ɵ, see  [44] and is most likely related to carbonate impurities as observed by TGA (as discussed further below).
In the case of Ht-Cl the second basal spacing of the first reflection agrees with literature (e.g., [43]). The basal spacing for Ht-NO 3 is lower than previously reported values of 8.1-8.8 Å (e.g., [14,47]). This could be related to the harsher drying used, which removed some water from the interlayer and hence, reduced the basal spacing. Another possibility is presence of CO 3 2− impurities in the interlayer of NO 3 − hydrotalcite, which would cause a smaller basal spacing as the CO 3 2− size is smaller than the one of NO 3 − .
For sample Ht-SO 4 , the first peak at 10.39 Å (8.30 2Ɵ) was above and the second peak at 7.75 Å (11.41 2Ɵ) below the numbers found in literature as 8.6-8.9 Å [43,44]. This indicates a segregation due to the presence of sulphate and carbonate anions (from impurities) in the interlayers. The first 003 peak at 10.39 Å (8.30 2Ɵ) contains mostly the SO 4 2− anions layers, as LDH-sulphates with a basal spacing of up to 11 Å have been observed due to the swelling of the interlayers by water [43]. This is supported by the observation of excess water observed by TGA for this sample (see below) and the rapid collapse of the first peak during in situ temperature XRD (see Section 3.2). The second 003-peak at 7.75 Å (11.41 2Ɵ) contains mostly CO 3 2− anions as impurity and a small amount of SO 4 2− , hence the basal spacing is drastically smaller.
The estimated cell volumes of the hydrotalcite samples are presented in Table 2 The FT-IR spectra of hydrotalcite samples measured between 600 and 4000 cm − 1 are shown in Fig. S1; a zoom of the range 600-1500 cm − 1 is presented in Fig. 3. The characteristic bands of the hydroxyl groups between 2600 and 4000 cm − 1 (O-H stretching) and of free water between 1625 and 1650 cm − 1 (O -H bending of water) were found in all the samples. The broad and mixed bands characteristic of the hydroxyl groups were more defined for the Ht-CO 3 and Ht-OH samples with the detection of shoulders at 3100 cm − 1 , while the band areas were broader in the Ht-Cl, Ht-NO 3 , and Ht-SO 4 samples confirming smaller crystallite or less ordered structure for the Ht-Cl, Ht-NO 3 , and Ht-SO 4 samples. Carbonates were also found in all the samples as visible by the asymmetric stretching of CO 3 2− in the range 1400-1500 cm − 1 [48], indicating a carbonate contamination, although limited, in all the samples. The asymmetric stretching of NO 3 − is in a similar range as the CO 3 2− band and overlaps [49,50]. Finally, the M -O bands (where M--Al or Mg) between 660 and 950 cm − 1 were more defined for the Ht-CO 3 and Ht-OH samples and can be related to a better crystallinity as observed in the XRD data. The 27 Al MAS NMR spectra of the hydrotalcite samples are shown in Fig. 4. The 27 Al MAS NMR spectra showed a sharp resonance at 9-11 ppm corresponding to an Al(VI) signal, confirming the predominant presence of aluminium in octahedral sites of the hydrotalcite. In addition a weak signal at 70-80 ppm indicating the presence of some Al(IV) was observed. The later was assigned to the presence of a low amount of asymmetric dehydroxylated Al(IV) in hydrotalcite [46]. The 27 Al NMR chemical shift of the Al(IV) environment depends on the composition of Table 2 Indexing of X-ray diffraction pattern, spacing and intensities for hydrotalcite samples.  [51]. The large variation from 78 ppm for Ht-Cl to 86 ppm for Ht-SO 4 seems to be related to the presence of different anion in the vicinity of the Al(IV). The 27 Al MAS NMR spectra were deconvoluted as detailed in [20] for hydrotalcite; the weak Al(IV) signal was deconvoluted using an asymmetric shape and quantified between 1 and 4% of the total aluminium in the solids. The main Al(VI) signal is better fitted with a symmetric signal at 9.1 ± 0.3 ppm and a second asymmetric signal at 10-11 ppm. The commercial Ht-CO 3 * sample contained only the symmetric Al(VI) signal, while the re-precipitated samples contained about 10% ± 3% of asymmetric Al(VI). This asymmetric Al(VI) resonance might indicate a less crystalline structure. Alternatively, this asymmetric Al(VI) may also indicate the presence of poorly aluminium hydroxide gels [52] although not detected by XRD or TGA. Therefore, the aluminium is considered to be only incorporated in hydrotalcite.
The 23 Na MAS NMR spectra associated to the hydrotalcite samples are shown in Fig. S2 in SI and for each hydrotalcite one signal was observed with a different shift. Signals at 7.4, 1.1, 0.0 and − 8.0 ppm in the 23 Na MAS NMR spectra were present for Ht-Cl, Ht-CO 3 , Ht-SO 4 and Ht-NO 3 samples corresponding to the resonance of NaCl, Na 2 CO 3 Na 2 SO 4 and NaNO 3 [53,54], respectively. Sodium is neither expected to precipitate within the hydrotalcite structure nor to be sorbed on the positively charged main layers. However, the presence of weak broad signal at − 4.6 ppm in the Ht-OH's spectrum indicated the presence of some hydrated NaOH [54] supposedly taken up in the interlayer or near the outer surface.
The thermogravimetric analyses (TGA) of the Ht-CO 3 and Ht-CO 3 * samples are presented in Fig. 5a, the TGA of the other samples are shown in Fig. 5b and a summary of the weight losses are detailed in Table 4. Two loss regions at 30-320 • C and 320-600 • C characteristic for hydrotalcite can be seen [24] related the first loss region between 30 • C and 270 • C to physically and chemically bound water and the second to the interlayer anion. We attributed the first weight loss to water loss, water present as both physical water and as interlayer water and the second peak to the main layer water and water associated with the anion. Similar water losses are commonly observed for calcium aluminate based LDH, so called AFm phases [55].
The Ht-SO 4 sample started to loose water from the beginning of the heating process, indicating a remaining excess of water in agreements with the large basal spacing of 11.39 Å observed by XRD. Although this could also be related to the dehydration of hydrated Na 2 SO 4 impurities remained in the sample. The Ht-Cl samples also started to loose water below 80 • C, while the Ht-CO 3 , Ht-OH and Ht-NO 3 started to loose water around 80 • C.

Table 3
Compositions and quadrupolar parameters of Al (IV) and Al(VI) sites determined by line shape simulation of 27 Fig. 3. FT-IR spectra of the hydrotalcite samples zoomed between 600 and 1500 cm − 1 , full spectra are given in Fig. S1. Fig. 4. 27 Al MAS NMR spectra (scaled to ease the comparison) of hydrotalcite samples (dashed line corresponds to the initial hydrotalcite-CO 3 from Sigma Aldrich). The mean centers of gravity of specific regions are given and the mean isotropic 27 Al MAS NMR chemical shifts are given in Table 3.
The FT-IR analysis of the exhausted gas of the TGA was used to understand the nature of the second and third weight loss for the Ht-CO 3 *, Ht-CO 3 , Ht-OH and Ht-NO 3 samples. The deconvolutions of the TGA-FT-IR are presented in the SI. The Ht-CO 3 * sample was used to quantify the trace of H 2 O and CO 2 in other samples, i.e. the traces were normalised based on the traces obtained for the reference sample. The water loss occurred over a large range from 30 to 600 • C, with carbonates were observed between 400 • C-650 • C, as summarized in Table 5. The theoretical 1 carbonate per 2 Al in hydrotalcite with Mg/Al = 2 was found by the analysis of the TGA data of the Ht-CO 3 sample. Also the other samples contained some CO 2 (see SI) confirming the FT-IR data of the solids: 0.2, 0.4, 0.3 and 0.35 CO 3 2− per 2Al in formula unit were calculated in the Ht-OH, Ht-Cl, Ht-SO 4 and Ht-NO 3 samples, respectively. For Ht-NO 3 , NO 2 was lost between 400 • C and 650 • C arising from nitrate salt decomposition slightly above a temperature observed by [56], who reported a loss starting at 280 • C, but similar to the temperature observed by [44]. 1.1 NO 3 − per 2Al in formula unit were calculated in the Ht-NO 3 . Cl 2 was released progressively between 400 and 1000 • C, which corresponds to data reported in literature [44]. SO 2 seemed to be exhausted starting from 500 • C, differently from literature where [44] reported that sulphate grafted the oxide layers with the increase of temperature and stayed in the structure until SO 2 was released starting at 800 • C. Thus, the contents of Cl − and SO 4 2− were not determined by this method. The content of hydroxyl group in each sample was obtained from the charge balance after taking into account the content of the other anions and then the water molecules were calculated from the left over from the total water loss. Based on the TGA results (when possible), the SEM/EDS data and the mass balance (Table S3) from calculation based on starting and final solution concentrations (i.e. considering the fraction of anions remaining in solution, see below), an estimation of the composition is given in Table 5. The TGA, SEM/EDS and the mass balance data agreed within the error margin.

Anion dependency and thermal stability studied by in-situ heated Xray powder diffraction
The changes in the water content upon heating can also be followed by the changes of the basal spacing. The evolution of the mean basal spacing obtained by in situ X-ray powder diffraction experiments during heating from 100 to 300 • C are shown in Fig. 6 (a video clip of the 003peak evolution in SI). Every sample showed a slight reduction (0.1-0.2 Å) of the basal spacing between 100 to ~150 • C. Ht-SO 4 additionally lost its first diffraction peak (at 10.39 Å, left shoulder of the main peak), confirming the excess of water in the sample. At around 150-155 • C the peak of Ht-CO 3 and Ht-OH collapsed rapidly and emerged as a new sharp peak at ~6.6 Å in a rapid transition. In the case of Ht-Cl the main peak remained while a new second peak (at ~7.4 Å) was formed. Ht-SO 4 showed a similar behaviour although its basal spacing was reduced by ~0.3 Å. The only sample not forming a second peak was Ht-NO 3 . However, its spacing was also reduced (~0.3 Å) and the peak became increasingly broad. The second peaks of Ht-SO 4 (~7 Å), Ht-CO 3 (~6.5

Table 4
Weight losses observed in the TGA curves for the different hydrotalcite samples.

Table 5
Estimation of the composition of the hydrotalcite samples (adapted from the TGA, EDS data and mass balance results, see Table S3). Å), Ht-OH (~6.5 Å) and Ht-Cl (~7.1 Å) collapsed at ~240-255 • C followed by the main peak of Ht-Cl shortly afterwards. The third and final peak of Ht-SO 4 (at ~6.6 Å) remained even to 300 • C, although it became a broad and small hump at the very end. A similarity of all samples is the slight increase of the basal spacing towards the end of their existence, mainly from the broadening of the main diffraction peak. The main difference is mostly related to the dehydration shown by the TGA (Fig. 5). The dehydration of Ht-CO 3 and Ht-OH was a clean transition, due to the evaporation of the interlayer water molecules. Ht-Cl encapsulated in what appears to be an intermediate stage which allows the coexistence of a fully hydrated and dehydrated phase. However, Ht-Cl had no continuous shift to the second 003 peak as one would expect, rather the hydrated peak is barley outlasting the dehydrated one (which can also be observed in the full pattern XRD measurement). One proposed explanation is the existence of a strong electrostatic interaction between the layers of high charge density and the anion because the interlayer distance (2.88 Å) is smaller than the anion itself (Cl − = 3.62 Å). Hence, the basal spacing remains unaffected by the release of water molecules [43]. However, it is unclear why this effect does not occur in the other samples as their basal spacing is also smaller than their anion. The occurrence of the second peak might then be explained by the CO 3 2− impurities ( Table 5) which form their own anion layer or mix only partially [57] and thus dehydrate more like Ht-CO 3 . The evolution of the 003 peak in Ht-NO 3 is similar although a second peak is only observed in the full pattern XRD measurement (see SI) which is in accordance with previous observations [47]. This indicates that the structure of Ht-NO 3 requires more time to form. Peaks of Ht-NO 3 and Ht-Cl become increasingly broad which may indicate differences in the atomic arrangement in the interlayer region, possibly related to the incorporation of CO 3 2− (Table 5). Finally, the behaviour of the peak evolution of the Ht-SO 4 samples was similar to the ones of Ht-OH and Ht-CO 3 samples apart from its broader peaks due to its poor crystallinity. The first peak (~8.5 Å) that collapsed at ~150 • C and was interpreted as already reduced from the 10.39 Å peak measured under ambient conditions. The major difference is the persistence of this peak during dehydration despite Ht-SO 4 being thermodynamically less last stable than Ht-OH and Ht-CO 3 (see below, 3.4). This is probably due to the grafted SO 4 2− anions which form crosslinks on the adjacent cation layers above 250 • C [43].

Liquid analysis
The dissolved concentrations and pH values measured in solution after the synthesis of the different hydrotalcites are reported in Table 6. The Al concentration for the Ht-CO 3 is 0.36 mmol/L and increased to 11 mmol/L for the Ht-OH. Al concentrations were even higher for Ht-SO 4 , Ht-Cl and Ht-NO 3 with 35, 57 and 67 mmol/L, respectively. All measured Al-concentrations were below the solubility of microcrystalline Al(OH) 3 except for the Ht-NO 3 , where the Al concentrations correspond to the solubility of microcrystalline Al(OH) 3 within the measurement error. In each sample, some dissolved CO 3 was measured confirming some carbonate contamination in the samples. The concentrations of the CO 3 2− , Cl − , SO 4 2− and NO 3 − were used to calculate its incorporation and the results are given in Table S 3.

Solubility products
Based on the analysis of the solutions after the syntheses, the "oversaturated" solubility products of the different hydrotalcites were calculated based on the dissolved concentrations and pH values measured in solution, and are given in Table 7 as oversaturated solubility products. The concentrations of Mg were below the detection limit and concentrations were thus estimated from the solubility of brucite at the measured pH values. For each sample, the solubility products (log K s0 ) were calculated based 1) on theoretical "ideal" formula (maximum uptake, no CO 2 contamination) and 2) based on the estimated "effective" composition ( Table 5). All the oversaturated log K s0 were in the range of − 44.1 and − 52.5, with the most stable oversaturated log K s0 for the Ht-CO 3 and Ht-OH with − 49.9 and − 49.7 respectively; followed by Ht-SO 4 > Ht-Cl > Ht-NO 3 , where Ht-NO 3 was the least stable. The log K s0 calculated for the effective compositions were lower than the one of the ideal composition indicating that the incorporation of carbonate in the Fig. 6. Mean basal spacing evolution vs. temperature on heating. The peak positions are displayed by the black lines. The peak intensity is displayed by the colour (the more transparent the lower the intensity). The width of the stripes represents the full-width at half maximum (FWHM) of the peaks. structure stabilises the phase, but it did not change the relative order of stability based on the anion type. In literature, carbonated hydrotalcites are also reported as more stable (see Table 1).
A comparison with calcium-aluminium based-LDH (AFm phases) showed a somewhat similar behaviour. The stability of such calciumaluminium based-LDH is higher for bivalent anions than for monovalent anions, with CO 3 -AFm being more stable than SO 4 -AFm, followed by NO 3 AFm > Cl-AFm > OH-AFm [21], indicating that not only the charge of the anion but also the structural arrangement of both the anion and the main layer plays an important role in stabilising the different Mg-Al and Ca-Al LDH phases.
The different hydrotalcite samples were re-equilibrated in water at 7, 23, 40 and 80 • C to determine their solubility at different temperatures. After re-equilibration the composition of the solid samples was checked by FT-IR, and in all cases the presence of the added hydrotalcite containing the specific anion was observed as the main solid (see SI, FT-IR). Based on the measured concentrations and the pH values (see SI, Table S4), "undersaturated" solubility products were calculated for the different hydrotalcites based on the effective composition (bold values in Table 7). In most cases, the magnesium concentrations were below the detection limit of 0.002 mmol/L, such that an upper limit for magnesium concentrations was estimated from the solubility of brucite. The results are given in Table 7 and the oversaturated and undersaturated solubility products versus pH are compared in Fig. 7a. The undersaturation data confirmed the sequence observed above, with Ht-CO 3 being the most stable phase, followed by Ht-OH > Ht-SO 4 > Ht-Cl > Ht-NO 3 .
Final solubility products at 25 • C were derived taking into account all the undersaturation experiments at the different temperatures (as discussed below), they are summarized in Table 7. These solubility products were further verified by calculating magnesium and aluminium concentrations as a function of the pH and compared it to the experimental data in Fig. 7b-f. The calculated solubility agrees in all cases well with the measured Al concentration, confirming that the estimated magnesium concentrations were in fact near the brucite solubility line. It is interesting to note that in the case of nitrate and chloride hydrotalcite, the calculated concentrations where only slightly lower than the solubility of brucite and microcrystalline Al(OH) 3 , while for OH and carbonate-hydrotalcites much lower Al concentrations were calculated, indicating the Ht-OH and Ht-CO 3 are the most stable hydrotalcites. In general, the solubility products from over and undersaturation agree well, with the exception of Ht-CO 3 , where the solubility from oversaturation was 2 to 3 log units higher than the solubility derived from undersaturation, which could indicate a kinetic hindrance. Note, that also the relatively high solubility data (less negative log K s0 values) from Rozov et al. [12,13] are based on oversaturation experiments. Fig. 8 shows that the temperature has a minor effect on the solubility of hydrotalcite as the solubility products measured at the different temperatures were comprised within the range of 1-2 log unit for the CO 3 , Cl, SO 4 and OH-Ht and 5 log units for the NO 3 -Ht. Such a limited effect of temperature has also been observed for Ca-Al-based LDH (AFm phases) containing sulphate, carbonate, chloride, nitrate or hydroxide in the interlayer [21,58].

Effect of temperature on the solubility of hydrotalcite
To describe the solubility at different temperature, entropy and heat capacity values are needed as detailed in Eq. (1). As no entropy or heat capacity data of the different hydrotalcites could be measured, they were estimated i) based on the values of other hydrotalcites and ii) based on their volume as detailed below.
The entropy (S • ) of each hydrotalcite was calculated based on the measured entropies given in [59] using the additivity method assuming Δ r S ≃ 0 according to Eq. (62) in [32] using Mg(OH) 2 and, MgCl 2 , Mg (NO 3 ) 2 , MgSO 4 and MgCO 3 and zeolitic H 2 O as constituents as given in [33]. The data are compiled in Table S5. The heat capacity (C p • ) was similarly calculated assuming Δ r C p ≃ 0 following Eq. (78) in Helgeson [32] from the hydrotalcite data given in [59] for each hydrotalcite sample (data presented in Table S5 and Fig. S6).
Recent publications [60,61] highlight the close relationship between the molar volume, V • , the entropy and the heat capacity. Thus, S • was also calculated from the relationship between V • and S • given in [61]. The C p • was calculated similarly using the averaged linear equation derived for C p • for hydrotalcite phases in [24]. The additivity and the volume-based values were compared in Table S5 (SI); generally, the values were similar except for the NO 3 -Ht and SO 4 -Ht. This might due to the approximate values of the entropies of Mg(NO 3 ) 2 and MgSO 4 .
Both the volume and the additivity based data were used to calculate changes of the solubility product with temperature as shown in Fig. 8, where "V • -based" corresponds to the fitting with S • calculated based on the equation given in [61]. For Ht-CO 3 , Ht-OH and Ht-Cl the volume based S • and C p • values or the additivity based data described the observed changes in solubility well, and no further fitting step was conducted. The additivity based data were selected as they rely on Table 6 Measured concentrations of dissolved species, pH values of solutions after the synthesised hydrotalcite samples at 80 • C. Concentrations given in mmol/L and pH was measured at 22 • C. Note that the Mg concentration were measured in largely diluted aliquots due to the pH and the large amount of Na to preserve the IC device, therefore the detection limit and the dilution give Mg <100 mmol/L. [ Table 7. Based on the solubility products, magnesium and aluminium concentrations were calculated at equilibrium with the specific hydrotalcite: b) Ht-CO 3 , c) Ht-OH, d) Ht-Cl, e) Ht-SO 4 , f) Ht-NO 3 , as a function of the pH and compared to the experimental data; solubility of brucite and microcrystalline Al(OH) 3 were added for comparison. For b) the experimental data given in [13] were added for comparison. Fig. 8. Solubility products (given in Table 7) as a function of temperature compared to relevant literature (Table 1) and the modelled solubility products are modelled with the S • and C p • given in Table S 4 ; Prentice 2020 data were recalculated from the measured concentrations in [24] assuming Mg concentrations which correspond to brucite solubility and Mg/Al = 2; * = no physical meaning. measured data [59]. For the Ht-SO 4 and the Ht-NO 3 samples, however, the volume based data were selected due to the approximate nature of the entropies of Mg(NO 3 ) 2 , MgSO 4 given in literature.
However, the changes in the solubility with temperature for the Ht-NO 3 phase were still underestimated (see Fig. 8). Thus, the S • values for this hydrotalcite were fitted based on our experimental solubility values following the procedure described in [58,62].
The log K s0 calculated here are comparable to other data reported in literature for Ht-OH, Ht-CO 3 and Ht-SO 4 . Only in the case of Ht-CO 3 , the solubility products recalculated based on the oversaturation experiments of [12,13] or [24] are considerably higher than the values obtained here from undersaturation experiments, which could indicate a kinetic hindrance in the precipitation of Ht-CO 3 . For the Ht-SO 4 phase, the recalculated experimental solubility from [24] assuming Mg in equilibrium with brucite and our data are in good agreement.
From our knowledge, the temperature effect on the solubility of Ht-NO 3 has not been measured before and it showed a steeper decrease than expected. The V cell of 314 Å 3 estimated from the fitted S • is large for hydrotalcite and would correspond to a basal spacing of ~13 Å and thus and thus to the presence of 2 or 3 additional water molecules per 2Al in formula in the interlayer. In fact, some studies reported a bigger basal spacing for NO 3 -hydrotalcite than in the SO 4 -hydrotalcite [14], such that we suspect the presence of CO 3 2− in the interlayer (i.e. a higher CO 2 contamination with a higher temperature). The final data given by this study are in bold in Table S5.

Extrapolation to the other Mg/Al ratios
Hydrotalcite phases are part of the LDH group and can have variable M 2+ /M 3+ ratios between 2 and 4, although more crystalline structures are observed for the specific M 2+ /M 3+ ratios of 2, 3 and 4 [10]. To be able to describe the compositional variation an ideal solid solution was assumed, as previously done [7,12,13] based on end-members with Mg/ Al = 2, 3 and 4. The thermodynamic data for the Mg/Al = 2 obtained experimentally in this study were used together with the data for brucite and H 2 O (see Table 1) to estimate the solubility products and entropy data for LDH with Mg 2+ /Al 3+ = 3 and 4 using the additivity method suggested in [7], e.g. for Mg 6 Al 2 (OH) 16 [12,13,24]. The molar volumes V • were estimated from the crystal structure: c was kept constant while the increase of a with the (Al/Mg + Al) change was calculated with the relationship given in [11]. The entropy S • and the heat capacity Cp • were deduced based on equations given in Fig. S6. The full thermodynamic data are given in Table 8. The resulting log K s0 values are given in Fig. 9 and compiled together with S • and C p • values in Table 8. Fig. 9a compares the solubility product of CO 3 -hydrotalcite with Mg/ Al = 2, 3 and 4 with log K s0 values reported in literature [5,12,13,23], and log K s0 values recalculated from [14,24,59]. The log K s0 , calculated in our study are in the same range as most literature values, with the exception of the log K s0 for Mg/Al = 2 and 3 given by [12,13] or recalculated from [24], which are several log units higher (less negative). Those solubility data have been obtained from oversaturation experiments. In fact, in the case of CO 3 -hydrotalcite all log K s0 based on oversaturation experiments [12,13,24] and oversaturation K s0 in Table 7 are higher than those obtained from undersaturation experiments (Table 7) or calculated based on the thermodynamic data of other hydrotalcite [14,[22][23][24]59], which could indicate towards a kinetic hindrance in the precipitation of CO 3 -hydrotalcite.
The solubility products derived for OH-hydrotalcite Mg/Al = 2, 3 and 4 (containing residual CO 3 2− ) agree very well with the log K s0 corresponding to the solid solution given by [7] and reasonably well with the recalculated log K s0 based on liquid composition given in [24] or with the OH-hydrotalcite calculated to precipitate in PC [5] as shown in Fig. 9b. We observed experimentally similar values in over-and under-saturation experiments and the values are similar (±0.5 log10 units) to those calculated in [7] from short term oversaturation experiments, while the data recalculated from [24] were less negative and those from [5] more negative (Table 1). In fact, [21] critically discussed the solubility products suggested for OH-hydrotalcite by Lothenbach et al. (2006) [5] and found that this solubility product for OHhydrotalcite was too low, without being able to suggest a better value due to the lack of experimental data at that time.
In addition, the Fig. 9c compares the log K s0 calculated for the hydrotalcite containing Cl − , SO 4 2− and NO 3 − with the few data found in literature. The values observed here agree well with the log K s0 recalculated for the SO 4 -hydrotalcite from the solution analysis from [24], the log K s0 calculated from [38] for the Cl-hydrotalcite, and the log K s0 for NO 3 -hydrotalcite reported in [37]. Also the log K s0 recalculated from the ion exchange experiments of [14] for the Cl-hydrotalcite and NO 3hydrotalcite agree well.
While anion substitutions in the structure of hydrotalcite are common [10,14,43,44], hydrotalcite containing CO 3 is the most stable and thus together with OH-hydrotalcite the most probable hydrotalcite formed in cementitious materials, i.e. in the presence of NO 3 , Cl and sulphate concentrations in the millimolar concentration range, as the kind of anions present in hydrotalcite depends strongly on the concentrations. The use of this hydrotalcite provisional ideal solid solution based on the data derived in this study predict the formation of a hydrotalcite containing mainly OH-in the interlayer in a carbonate free hydrated Portland cement and of a hydrotalcite containing both carbonate and OH in a Portland cement containing CaCO 3 . This prediction is in good agreement with the experimental observation of hydrotalcite formation in Mg-rich Portland cements [17]. Cl-or NO 3 -containing hydrotalcite may be formed in environment containing high Cl − or NO 3 − concentrations; in fact the formation of chloride containing hydrotalcite has been observed at high chloride concentrations [16].

Conclusions
Magnesium present in cementitious materials is experimentally observed to precipitate as poorly crystalline magnesium-based double hydroxide layer phases (LDH) containing aluminium. Such hydrotalcitelike phases can incorporate different anions in the interlayer to compensate the positive main layer charge. In this study, hydrotalcites with different anions, CO 3 2− , OH − , NO 3 − , Cl − , and SO 4 2− in the interlayer were successfully synthesised at 80 • C. Mass balance, XRD and FT-IR indicated the incorporation of the respective anion, CO 3 2− , OH − , NO 3 − , Cl − , and SO 4 2− , in the interlayer. In addition, some CO 3 2− was present in the interlayer due to the high affinity of hydrotalcite for carbonate. The crystallinity of the hydrotalcites depended on the anion and increased in the order: SO 4 2− < Cl − < NO 3 − < OH − < CO 3 2− .
In-situ XRD and TGA experiments showed that the basal spacing and the amount of interlayer water decreases stepwise with temperature, and that the amount of interlayer water and the temperature at which this water is lost depends on the incorporated anion.
The solubility products, calculated based on the solution analysis of samples re-equilibrated at 7, 23, 40 and 80 • C showed that the CO 3hydrotalcite was the most stable hydrotalcite, followed by OHhydrotalcite, while SO 4 -hydrotalcite, Cl-hydrotalcite and NO 3 -hydrotalcite were less stable. In the case of CO 3 -hydrotalcite a higher solubility was observed in oversaturation experiments than in undersaturation experiments indicating a possible kinetic hindrance in the precipitation of CO 3 -hydrotalcite. Such a kinetic hindrance would also explain the large scatter in literature data in the case of CO 3 -hydrotalcite, where several experimental studies based on oversaturation experiments showed a much higher solubility. For OH-, SO 4 -, Cl-and NO 3 -hydrotalcite no significant difference between the solubility obtained from oversaturation and from undersaturation was observed in our experiments and the data showed reasonable agreement with the few solubility measurements available in literature.
The solubility of the hydrotalcites studied showed only a weak dependence on temperature, similar to the trends observed for Ca-Albased LDH. The thermodynamic data derived experimentally for hydrotalcite with Mg/Al = 2 were extrapolated to Mg/Al = 3 and 4 using a simple additivity method [7], which described the few available literature data well.
Thermodynamic data for hydrotalcite containing OH, CO 3 , SO 4 , Cl and NO 3 are given including, solubility product, molar volume, entropy and heat capacity. This work provides new data that are needed to predict the formation of hydrotalcite in cementitious materials and the use of a provisional ideal solid solution model is suggested. Such a model predicts the formation of hydrotalcite containing carbonate and OH in its interlayer if applied to hydrated Portland cement but can also be used for other cementitious materials such as alkali activated cements.
This work gives basis to work with updated and more homogeneous thermodynamic data for pure Mg-Al LDH phases. However, the iron from Portland cement or from industrial wastes (e.g. slags) is also expected to be incorporated in LDH phases, and further work should be conducted to describe the incorporation of Fe in the LDH by thermodynamic modelling and future research is necessary to construct more rigorous and quantitative solid solution model.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
analyses. Priska Baehler, Christopher Pichler and Dr. Niklaus Waber (UniBern) are acknowledged for the solution analyses. Yiru Yan and Rosa Guidone (Empa) for help with the TGA-FT-IR measurements, Dr. Urs Eggenberger (UniBern) for giving access to measuring setups. The helpful insights from the two anonymous reviewers are thankfully acknowledged.