In ﬂ uence of sodium nitrate on the phases formed in the MgO-Al 2 O 3 -SiO 2 -H 2 O system

silicate hydrate -NaNO 3 -H 2 (8-16 uptake of 8 The incorporation of sodium and aluminium in magnesium silicate hydrate phases (M-S-H), possible binding phases in magnesium silicate cement, was investigated. Magnesium alumino silicate hydrate containing sodium (M-A-S-HN samples) were synthesizedinbatchexperimentswith NaNO 3 ~ 100 mmol/L atMg/Siratios equalto 0.8and 1.2,andAl/Siratios of 0,0.1 and 0.2,and equilibratedat20and 50°C.Thermogravimetric analysis (TGA), X-raysdiffraction(XRD), 29 SiMASNMRdatashowedthatM-A-S-Hphasesformed,withasimilarstructureasM-S-H. 27 Aland 23 NaMASNMRdata showedthat onlylittlesodium wassorbed,whilealuminium waspossiblyin-corporated in both tetrahedral and octahedral sites of M-A-S-H. We found evidence that the presence of sodium nitrate led to the formation of hydrotalcite-like phase probably containing NO 3 − and possibly to the trace formation of hydrated alumino-silicate containing sodium: N-A-S-H gel. These minor phases limited the aluminium uptake by M-S-H at higher Al contents (Al/Si = 0.2).


Introduction
The manufacture of Portland clinker is an energy intensive process and causes 5 to 8% of the anthropogenic CO 2 emissions. The use of reactive magnesium from magnesium silicate minerals is in consideration to decrease the CO 2 emitted from construction [1]. Thus magnesium silicate hydrate phases get increasing attention as reaction product of magnesium-based binders, a potential alternative to Portland cement [2] and cementitious material able to generate high compressive strength [3,4]. The properties of magnesium silicates hydrate (M-S-H) are also of interest as they are observed at the interfacial zone of cement-based materials in contact with clays [5][6][7][8][9] and/or as secondary products from the degradation of cementitious materials by groundwater or seawater [10][11][12][13]. M-S-H is formed from the reaction of magnesium with amorphous silica, released by the decalcification of the Materials and Design 198 (2021) 109391 C-S-H on the surface of the hydrated cement. In the case of primary product, M-S-H is formed directly from reactive MgO-SiO 2 sources. M-S-H phases have an ill-defined structure comparable to hydrated precursors of 2:1 and 1:1 phyllosilicates [8,[14][15][16]. Magnesium phyllosilicates are composed of tetrahedral sheets containing Si 4+ and octahedral sheets containing Mg 2+ . One tetrahedral layer on an octahedral layer corresponds to a 1:1 layer silicate structure while two tetrahedral layers sandwiching an octahedral layer correspond to a 2:1 configuration as detailed in Fig. 1.
The observation of M-S-H formed in situ at the surface of hydrated cement indicated that aluminium could be present either in the magnesium silicate phases [6,8,12,18] and/or in a hydrotalcite [5] which is not well crystalline. Our recent study [17] showed the incorporation of aluminium in M-S-H with Mg/Si equal to 1.1 and 1.7 up to Al/Si~0. 15 [17]. At a pH below 10.5 a negative surface charge of M-S-H with a maximum exchangeable cation/Si~0.05 was observed; magnesium, sodium and as well as other cations present have been observed at such cation exchange sites [19]. The potential uptake of sodium by M-A-S-H can modify the pH values and the phase assemblage and the stability at the interface clay-cement [8,9,20] or in new binders [21].
In the present study, M-A-S-H N samples were synthesized with Mg/ Si = 0.8 and 1.2 and Al/Si = 0, 0.1 and 0.2 in 100 mmol/L NaNO 3 solution. We investigated the stability of M-A-S-H phases at different temperatures (20 and 50°C) with the addition of sodium nitrate and the possible formation of other phases. The aqueous phases were analysed by ion chromatography and pH measurements and the solid phases by thermogravimetric analysis, X-ray diffraction, 29 Si, 27 Al and 23 Na MAS NMR spectroscopy. The experimental investigations were supported by thermodynamic calculations to better understand the MgO-Al 2 O 3 -SiO 2 -NaNO 3 -H 2 O system.

Synthesis
In the following the nomenclature M-S-H x, M-S-H xN, M-A-S-H x y, and M-A-S-H xN y is adopted to describe the samples, where x indicates the Mg/Si ratio, y the Al/Si ratio, and N the presence of NaNO 3 .
Magnesium oxide (Merck, pro analysis, 0.18 ± 0.02 wt% Na 2 O, surface area of 24 m 2 /g [32]) and silica fume (SiO 2 , Aerosil 200, 0.9 wt% HCl, specific surface area of 200 m 2 /g) were chosen as starting materials for the synthesis of M-S-H x and M-S-H xN synthesized in presence of sodium nitrate (NaNO 3 , VWR chemicals, Analar normapur) as detailed in [19,32]. Sodium aluminate (NaAlO 2 , anhydrous, technical from Sigma Aldrich, which contains 6.9 wt% of water as quantified by TGA) was used to synthetize M-A-S-H xN y samples. To avoid the increase of pH value by the addition of sodium aluminate which usually lead to the slower formation of M-S-H due to preliminary formation of brucite [32] a corresponding amount of nitric acid (HNO 3 , Merck, suprapur, 65%) was added. In addition, sodium nitrate was added until a total sodium concentration of 100 mmol/L was reached. Two Mg/Si ratios were studied (0.8 and 1.2) and the Al/Si in the mixtures was 0, 0.1 and 0.2 as indicated in Table 1. The samples were prepared in PE-HD containers using milli-Q water and a water/solid (W/S) ratio of 45 to ensure a homogeneous suspension and sufficient solution for analysis. All sample handling was carried out in a glove box under N 2 to avoid CO 2 contamination. The samples equilibrated at 20°C were placed on a horizontal shaker (100 rpm) and the samples stored at 50°C were shaken once per week.
The suspensions were equilibrated at different temperatures (20 and 50°C) and for different times (up to 1 or 2 years) for kinetic and long-term investigations as mentioned in previous work [32]. The solid and liquid phases were separated by filtration under pressure (4-5 bars N 2 ) using nylon filters (0.45 μm). After filtration, the solids were washed with 50/50 (v/v) water-ethanol and then with 94 wt% ethanol to remove dissolved ions and to prevent the precipitation of salts during drying [33]. The samples were freeze-dried with liquid nitrogen (for approximatively 20 min at −196°C) and kept for 7 days at −40°C under vacuum (pressure of 0.28 mbar). After further equilibration in N 2 -filled desiccators at a relative humidity of~30% (above saturated CaCl 2 solution) over a period of at least 14 days, the solid phases   Fig. S1 (Supporting Information), TGA and 27 Al MAS NMR data are shown in the following for comparison.

Analytical techniques
The compositions of the liquid phases were analysed by ion chromatography (IC) immediately after filtration. The concentrations of the dissolved magnesium, sodium, nitrate were quantified using a Dionex DP series ICS-3000 ion chromatography system with a measurement error ≤ 10% in undiluted solutions or in solutions diluted by factors of 10, 100 or 1000. Dissolved silicon was analysed using sodium carbonate/bicarbonate eluent and sodium molybdate, and sodium lauryl sulphate in metasulfonic acid as a post-column reagent using an ion pack AS22 column. Dissolved aluminium was analysed using CS5A Dionex IonPac column with HCl diluted eluent with a post column reagent (ammonium acetate).
All concentrations were determined as duplicates and the results are given as mean values. The pH values (±0.1) were measured at ambient temperature (23 ± 2°C) in an aliquot of the unfiltered suspension and the results were corrected to 20 or 50°C [32]. The composition of the aqueous phase did not change significantly during the 30 min required to cool the solutions from 50°C to ambient temperature [32].
XRD data were collected using a PANalytical X'Pert Pro MPD diffractometer equipped with a rotating sample stage in a Ɵ-2Ɵ configuration applying CuKα radiation (λ = 1.54 Å) at 45 mV voltage and 40 mA intensity with a fixed divergence slit size and an anti-scattering slit on the incident beam of 0.5°and 1°. All samples were scanned between 5°and 75°2Ɵ with a X'Celerator detector.
The 29 Si MAS NMR single pulse experiments were conducted on a Bruker Avance III NMR spectrometer using a 7 mm CP/MAS probe at 79.5 MHz applying the following parameters: 4500 Hz sample rotation rate, minimum of 3072 scans, 30°2 9 Si pulse of 2.5 μs, 20 s recycle delays, RF field strength of 33.3 kHz during SPINAL64 proton decoupling. The 29 Si NMR chemical shifts were referenced to the most intense resonance at −2.3 ppm of an external sample of an octamethylsilsesquioxane (Aldrich No. 52,683-5) which was referenced to tetramethylsilane (TMS, δ 29 Si = 0.0 ppm). For 29 Si MAS NMR data, the 30°flip angle is a compromise between improving signal to noise and quantitative data acquisition. With the applied recycle delay of 20 s, this experimental setup yields a maximum signal for species with T 1 recycle times of 120 s according to the Ernst angle. For a representative sample (M-S-H 0.8 N cured at 50°C during 1 year) we obtained T 1 values in the range of 65 to 80 s applying a 29 Si MAS NMR saturation recovery pulse sequence. Although the maximum signal intensity was not reached with the selected flip angle, the magnetisation in the observed window of T 1 values relaxed similarly under these conditions (ca. 90% of equilibrium magnetisation recovered after each pulse). Assuming that the T 1 values do not change between samples, the changes in relative signal intensities obtained by lineshape analysis of the Q n sites within each sample were evaluated.
The observed 29 Si NMR resonances were assigned using the Q n classification, where one Si tetrahedron is connected to n Si tetrahedra, where n varies from 0 to 4. The lineshapes of the experimental data were analysed by non-linear least-square fits using the "DMFIT" software developed by Massiot et al. [34]. The presence of unreacted silica was confirmed by the resonances at −101 ppm (Q 3 from the surface of the amorphous silica [14]) and of Q 4 at −110 ppm. The procedure used for the lineshape analysis of 29 Si NMR data is described in detail in [32].
The 27 Al NMR spectra were measured on the same instrument using a 2.5 mm CP/MAS probe. 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 on the content of aluminium in the samples, π/12 pulses of 1.5 μs, 0.5 s recycle delays (identical spectra were obtained when recycle delays of 0.2, 0.5 and 1.0 s were applied), without 1 H decoupling. The 27 Al NMR chemical shifts were referenced to an external sample of Al (acac) 3 . The 27 Al MAS NMR spectra were analysed by the above mentioned lineshape fitting software "DMFIT" [34]. The detailed description of the lineshape fitting procedure is given in the Supporting Information section.
The 23 Na MAS NMR data was recorded at 105.9 MHz using a 4 mm CP/MAS probe applying the following parameters: 13′000 Hz sample rotation, 512 scans, 20°pulses of 2.0 μs, 1 s recycle delays, no 1 H decoupling during acquisition. The 23   were analysed by applying Lorentzian lineshapes [34]. Please note that the 23 Na MAS NMR resonances throughout were symmetrical and lineshapes could be simulated by using Lorentzian shapes. All attempts to fit the 23 Na NMR data with 2nd order quadrupolar broadened lineshapes failed, which means that the sodium cations must be quite mobile in the M-S-H N phases.
To obtain uniformly excited NMR resonances, small flip angels of 20 and 8°were applied to record 23 Na and 27 Al MAS NMR data, respectively [36]. The recycle delays applied ensure the recording of quantitative data.

Saturation indices
The calculations of the saturation indices were carried out using the Gibbs free energy minimization program GEMS [37]. GEMS is a broadpurpose geochemical modelling code which computes 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 [38]. The data for the M-S-H solid solution and for amorphous SiO 2 originate from [32,39] and M-A-S-H phases added to the solid-solution from [17], for microcrystalline aluminium hydroxide (microcrystalline Al (OH) 3 ) from [40,41], hydrotalcite from [42], and the zeolite phases from [43] as summarized in Table 2.
The saturation indices (SI) of the different solids were calculated based on the experimentally determined ion concentrations in solution according to eq. (1): where IAP is the ion activity product calculated from the measured concentrations, while Kso is the theoretical solubility product of the solid (as indicated in Table 2).

TGA and XRD data
The  [14,15] in the XRD data of the M-A-S-H xN y samples can be assigned to M-S-H (Fig. 3, Table 3). Additionally, the characteristic water losses between 30 and 250°C (poorly bound water) and 250°C and 800°C (chemically bound water: hydroxyl groups) [3,14,45,46] of M-S-H were found ( Fig. 2a-d).
For the M-A-S-H xN y samples prepared at 20°C, the formation of brucite (Fig. 2a, b: water loss centered at~400°C [3] and Fig. 3a, b, reflection peaks at 18.6, 32.7, 38.0, 50.9, 58.7, 62.0°2θ [47,48]) was observed. In addition, a poorly crystalline hydrotalcite-like phase (reflection peaks at 11.39, 23.02, 61.35°2θ [49] associated with water loss at 200 and 350°C [44,49,50]) was identified in the M-A-S-H xN 0.2 samples cured at 20°C and at 50°C. Some poorly crystalline hydrotalcite-like phases and/or amorphous aluminium hydroxide may also have formed in M-A-S-H xN 0.1, but in small amounts difficult to identify by TGA and XRD. Finally, the presence of semi crystalline or even amorphous aluminium hydroxide (water loss at~100°C (amorphous) or 250-300°C (semi-crystalline) [44,51]) is possibly observed in samples that were aged at all temperatures. All the solids detected are summarized in Table 3.
Comparing the samples equilibrated for 1 year at 50°C, i.e. the samples which were in equilibrium as discussed in [32], the M-A-S-H xN 0.2 samples showed clearly the presence of a hydrotalcite-like phase, which is present only in trace amounts in M-A-S-H 1.1 0.2 sample [17], indicating tentatively an important role of sodium nitrate in hydrotalcite formation.
Finally, for M-A-S-H xN 0.2 samples some additional broad humps in XRD can be observed at~14, 28 and 44°2θ, which could not be associated with any of the above discussed solid phases. There are two possible explanations for these reflections: the presence of amorphous aluminium hydroxide with broad humps in XRD data at~14, 29, 39 and 49°2θ, usually attributed to amorphous aluminium hydroxide/ boehmite [52]; or the formation of hydrated alumino-silicate containing sodium (N-A-S-H gels) as precursors of zeolites that could present pattern with broad reflections assigned to e.g. sodalite 14.1, 24.6, 35.0, 43.2°2θ [53] or zeolite Y(Na) 6.2, 10.1, 11.9, 15.6, 20.3°2θ [54]. 29 Si MAS NMR data (Fig. 4) confirm the presence of M-S-H in all M-A-S-H xN y samples; about 2/3 of the signal intensity is attributable to Q 3 species between −93 and − 97 ppm and approximately 1/3 to Q 2 at −85 ppm [14,[55][56][57][58][59]. The results from lineshape analysis (see experimental part) for M-A-S-H xN y samples are presented in Table 4. At 20°C, all samples excepted M-A-S-H 1.2 N 0.1 sample showed  samples cured at 20 and 50°C showed a Q 2 /Q 3 ratio of 0.6-0.7, indicating a slightly higher polymerization degree than in M-S-H where a ratio of about 0.9 is expected [32].

NMR data
For M-A-S-H xN y samples, reasonable deconvolutions were also obtained with similar signals. However, main 29 Si MAS NMR resonances of zeolites such as natrolite usually are also observed between −85 and − 100 ppm, at −87.6 and − 95.2 ppm [60] or for zeolite Y, zeolite X or sodalite at −85, −89 and − 94 and − 100 ppm [61]. Therefore, the unambiguous identification of zeolitic precursors in presence of M-A-S-H phases could be prevented, because their signals would be expected at the same chemical shift region and, furthermore, due to the low crystallinity, rather broad signals would be expected for zeolitic material.
The  Table 5. Deconvolution was mainly performed to determine the isotropic 27 Al NMR chemical shifts of the species Al(IV)a, b, c, but the relative amounts shown in Table 5 are associated with errors of about 10-20%.
For the M-A-S-H 1.1 0.1 sample, only one signal Al(IV)b with δ iso of 69 ppm was necessary to fit the chemical shift range of tetrahedrally coordinated Al of the spectra [17]. For most of the M-A-S-H xN y samples (Fig. 5, Fig. S3 and Table 5) three peaks with isotropic chemical shifts of ≈ 80, 68 and 60 ppm were used for the Al(IV) signal, ascribed to three Al(IV) species in different environments named Al(IV)a, Al(IV)b, and Al (IV)c. The isotropic chemical shift at ≈ 80 ppm was attributed to the presence of hydrotalcite which exhibits a minor amount of dehydroxylated Al(VI) (Fig. 5 and [62]). An intensity >5% is observed in samples, where hydrotalcite was confirmed by TGA and XRD.
The Al(IV)b peak is attributed to Al(IV) sites in the M-A-S-H phase. The Al(IV)c signal is characteristic for zeolites [63,64], however it was not readily observed in each sample, particularly in the samples M-A-S-H 1.2 N 0.1 (cured at 20 and 50°C) and in sample M-A-S-H 1.2 N 0.2 (cured at 20°C). For these 27 Al NMR spectra deconvolutions were performed without the Al(IV)c site. Although the presence of zeolite precursor in our data could not be unambiguously confirmed by XRD or TGA, the 27 Al NMR data indicated that poorly crystalline zeolitic precursor may be present in all the M-A-S-H 0.8 N y samples and the M-A-S-H 1.2 N 0.2 at 50°C.
An average proportion of~20% for the asymmetric environment Al (VI)a at 10-11 ppm was determined for all M-A-S-H xN y samples (Table 5), which corresponds very well to the content of Al(VI)a in M-A-S-H samples [17]. Alternatively, the asymmetric Al(VI)a resonance may also indicate the presence of poorly ordered aluminium hydroxide gels [65].
Since the 27 Al NMR chemical shifts of the Al(VI)b sites in M-A-S-H phase and in hydrotalcite are very similar, as in both cases aluminium is completely neighboured by magnesium in an octahedral layer (trioctahedral phyllosilicates or LDH, where all octahedral positions are filled [66] as discussed in detail in [17]), the distinction by 27 Al NMR of the two phases is not possible. However, the Al(VI) fraction in  (Table 5) in the samples compared to the M-A-S-H phases [17].
The 23 Na MAS NMR spectra of M-A-S-H xN y samples at 50°C are presented in Fig. 6 together with the spectrum of M-S-H 0.8 N samples and M-S-H 0.8 N* (synthesized in NaOH solution at equilibrium with a pH of 12.5 [19]). The 23 Na NMR resonances of all M-A-S-H xN y samples showed symmetrical signals with line widths of 700 ± 50 Hz (Lorentzian shapes determined by "DMFIT" [34], data not shown) at a  But, the observed chemical shift of the broad 23 Na MAS NMR signal between −10 -0 ppm corresponds also to the expected range of partially hydrated sodium (Na(H 2 O) x + ) at the surface sites of N-A-S-H gels [67][68][69][70][71].
Therefore, our data cannot indicate the presence of zeolitic precursor, but indicated only the incorporation of some sodium in the solid phase by sorption of poorly hydrated sodium on the deprotonated silanol groups of M-A-S-H and/or of zeolitic precursor.
All the solids detected by TGA, XRD, and NMR data are summarized in Table 3.

Analysis of liquid phase
The composition of the liquid phases was also analysed. At 20°C, a pH of 8.5 was measured in the solution in equilibrium with M-S-H 0.8 sample after 2 years [19] and 0.33 mmol/L magnesium and 1.33 mmol/L silicon. For the M-S-H 1.2 sample, pH increased to 9.9, magnesium decreased to 0.19 mmol/L and silicon to 0.006 mmol/L [19] as indicated in Table 6. In a previous study [17] Table 6). The lower pH and the partial substitution of magnesium by sodium at surface cation exchange sites led to the observed relatively high magnesium concentrations in the presence of sodium nitrate [19]. The pH values of the samples containing sodium and aluminium decrease with time ( Table 6). This pH decrease could indicate an uptake of sodium on M-S-H (or zeolitic precursors), in agreement with the observations for M-S-H in absence of aluminium [19].
The solutions of the M-S-H 0.8, M-S-H 0.8 N and M-A-S-H 0.8 N y samples contained silicon concentrations between 1.2 and 1.7 mmol/L indicating equilibria with amorphous silica [14,32]. We observed no influence on magnesium concentrations in samples with low aluminium content (Al/Si = 0.1). However, at Al/Si = 0.2 much lower magnesium concentrations were detected, related to the formation of a hydrotalcite like phase in these samples [17], while silicon concentrations were increased, consistent with the presence of amorphous silica (Table 6). Most concentrations of aluminium were below the detection limit, Table 4 Assignments of 29 Si NMR chemical shifts and relative amounts of Q n silicon species obtained by simulation of the 29 Si MAS NMR spectra shown in Fig. 4.   (Table 5).
indicating a strong preference of Al for the solid phases. No or little reduction of sodium concentrations in the solution was observed, indicating that the sodium uptake by the solid phases observed by 23 Na MAS NMR is small only, in accordance with observations for M-S-H samples in the presence of calcium or sodium at low pH [19,72]. Similarly, no specific decrease was observed for the nitrate concentration, indicating only possible little uptake by the solid phase. However, due to the relatively high sodium and nitrate concentrations and their low uptake, such measurements are not very accurate [19,73]. A difference of 10 mmol/L in the measurement corresponds to~2 mmol of NaNO 3 in the solids, a significant amount if related to the Al content in the initial mixes (between 5 and 10 mmol).
As discussed for the experiments performed at 20°C (see above), similar trends were observed for the samples stored at 50°C: i.e. highest magnesium concentrations at low Al/Si, decrease of magnesium concentrations with increasing Al/Si and aluminium concentrations at/or below the detection limit.

Discussions
Saturation indices were calculated with respect to M-A-S-H, micro crystalline Al(OH) 3 , OH-hydrotalcite and different zeolites (zeolite X, zeolite Y, natrolite and sodalite, see Table 7); in addition the ion activity products of NO 3 -hydrotalcite (Mg 4 Al 2 (NO 3 ) 2 (OH) 12

(H 2 O) 3 ) and M-A-
S-H phases were calculated (see Table 8). Most concentrations of aluminium were below the detection limit, such that in many cases only maximum saturation indices (SI) could be calculated (using the detection limit of 0.0001 mmol/L as maximum aluminium concentration). A negative saturation index (SI) indicates that the solution is undersaturated and the respective solid should not form or will dissolve if present. Thus the knowledge of SI can contribute together with the different solid phase analysis to assess which solids might have formed as discussed for the different solids in the following part.

Hydrotalcite formation
Clear hydrotalcite formation was indicated by TGA for the M-A-S-H xN 0.2 samples, while the XRD and 27 Al MAS NMR data tended to indicate the presence of hydrotalcite in all samples containing Al ( Table 3).
The SI data (Table 6) indicate that all the samples were undersaturated with respect to OH-hydrotalcite. The formation of OHhydrotalcite is rather unlikely at pH values <10 (Table 6), i.e. below 0.1 mmol/L hydroxide, while the relatively high nitrate concentration of 100 mmol/L makes the formation of hydrotalcite-containing nitrate instead of hydroxide as counterion in the interlayer probable. Under very similar conditions as in our study, Miyata [26] observed the formation NO 3 -hydrotalcite. The measured concentrations indicate an ion activity product of NO 3 -hydrotalcite (Mg 4 Al 2 (NO 3 ) 2 (OH) 12 (H 2 O) 3 ) of 10 −52 to 10 −56 , which is in the expected range based on the ion exchange experiments of [26]. In addition, also the presence of some carbonates could stabilise hydrotalcite at pH below 10. The formation of hydrotalcite was also consistent with the significant decrease in magnesium concentration from Al/Si = 0 and 0.1 to Al/Si = 0.2 (Table 6), indicating the precipitation of an Mg and Al containing solid phase. It cannot clearly be assessed whether the nitrate concentrations decreased in the presence of Al, which would indicate uptake into hydrotalcite, owing to the inherent error of the liquid phase analysis of ±10%.
The relatively poor crystallinity of the hydrotalcite-like phases possibly present in the M-A-S-H xN y samples compared to hydrotalcite containing NO 3 − (XRD shown in Fig. S1), is due to its low amount, and the possibility of varying composition both in the interlayer (OH − , NO 3 − ) Table 5 Compositions and isotropic 27

Are zeolites forming?
The formation of different crystalline zeolites was observed in a similar system containing amorphous SiO 2 , NaAlO 2 , Mg(OH) 2 and H 2 O, but at higher pH values by Walling et al. [29], while in the present study no crystalline zeolite could be found (see XRD data in Fig. 3). However, the trace formation of poorly crystalline zeolite-like phases (N-A-S-H) seems probable. The 29 Si MAS NMR data of the M-A-S-H 0.8 N y samples at each temperature and the M-A-S-H 1.2 N 0.2 at 50°C sample indicated more Q 3 content than in the other samples (lower Q 2 /Q 3 ), which could indicate the presence of zeolitic precursor in the samples. This high Q 3 content in these samples coincided with the observation of a signal with an isotropic 27 Al NMR chemical shift at~60 ppm in the 27 Al MAS NMR data as typical for Al(IV) in zeolite framework.
The solutions were oversaturated with respect to natrolite and zeolite Y (with exception of M-A-S-H 1.2 N 0.1) and close to saturation for sodalite and zeolite X, indicating that the solids could potentially form. However, we want to stress that this is all indirect evidence, which together make the presence of zeolitic precursor probable, while the actual proof of the presence of N-A-S-H gels is difficult; and the exact nature of the zeolitic precursor remains unclear.

Kinetics of M-A-S-H formation
All solutions were nearly saturated with respect to M-S-H and M-A-S-H phases, in agreement with the solid phase analysis, where M-A-S-H phase was identified as the main product. However, the samples synthesized at 20°C during 2 years show more unreacted silica and brucite than those synthesized at 50°C during 1 year (Table 3). The samples equilibrated only 1 year at 20°C exhibit even higher contents of  amorphous silica and brucite and a higher Al concentration than samples aged of 2 years as shown in Table 3 and Table 6. This indicates that in particular the 20°C samples are not yet at equilibrium after 1 year in agreement with observation of [3,17]. Increasing the temperature accelerated M-S-H formation and no further changes are observed after 1 year at 50°C. The M-S-H formation can be limited by the slow reaction of silica, brucite and Al(OH) 3 . In general, samples with low Mg/Si may contain unreacted SiO 2 and samples with high Mg/Si may contain unreacted brucite. At high Al/Si levels precipitates of Al(OH) 3 can be formed, which are only very slowly incorporated into M-(A-)-S-H at 20°C (increased uptake at 50°C).
The M-A-S-H 0.8 N y samples were under all conditions oversaturated with respect to amorphous silica in agreement with its presence in the samples as observed by 29 Si MAS NMR, the M-A-S-H 1.2 N 0.2 sample only at 20°C (1 and 2 years), indicating an influence of Al on M-S-H formation rate. All other samples were undersaturated, in agreement with the 29 Si MAS NMR data where no unreacted silica was observed ( Table 3).
The M-A-S-H 1.2 N 0.2 at 20°C sample also showed the presence of brucite (Table 3) although the solution was undersaturated with respect to brucite. Brucite formation is generally observed during the early hydration of MgO and SiO 2 , as MgO reacts much faster than SiO 2 [3,17]. The dissolution of the formed brucite, however, proceeds only very slowly as it is kinetically hindered in the presence of Si concentrations in the mmol/L range. The presence of both unreacted silica and brucite in this sample (M-A-S-H 1.2 N 0.2 sample) indicates that equilibrium is not yet reached.
The calculated (maximum) SI of microcrystalline Al(OH) 3 are all negative indicating undersaturation, which is in agreement with its absence in the solid phase analysis. However, it cannot be excluded that traces could be present which continue to react over time.

Possible incorporations in M-A-S-H
The uptake of Al in both the octahedral and tetrahedral sites of M-S-H in the absence of secondary phases such as hydrotalcite and zeolitic precursors has been evidenced in [17,75]. The formation of hydrotalcite and zeolitic precursors in the present paper due the presence of Na and nitrate led to less Al in M-A-S-H phase than observed for the pure M-A-S-H samples from [17], mainly due to the presence of hydrotalcite as evidenced by the narrow and symmetric 27 Al NMR resonance at~9 ppm ( Table 5). The 23 Na MAS NMR data showed that a part of sodium is sorbed on silicate surface consistent with Na sorption on M-S-H phase without Al [19], while the data from solution analysis revealed only a small Na uptake at the pH range of 8-10.
The measured concentrations were used to calculate ion activity products (IAP, Table 8) for possible aluminium containing M-S-H endmembers resulting in log (IAP) = −14.8 ± 1.3 for M 0.75 A 0.10 SH 1.50 , log(IAP) = −18.7 ± 1.9 for MA 0.10 SH 1.75 , and log (IAP) = −26.5 ± 3.1 for M 1.50 A 0.10 SH 1.80 . We use the expression IAP rather than solubility product as we cannot assess whether equilibrium has been reached. These values are in the same order as those of samples produced without sodium nitrate [17], indicating no significant effects of sodium on the solubility of M-A-S-H.

Conclusions
The aim of this work was to investigate the effect of sodium and nitrate on the aluminium incorporation into magnesium silicate hydrate phases at pH values between 8 and 10. The 29 Si MAS NMR data, thermogravimetric analysis (TGA), and X-ray diffraction (XRD) showed that in all cases M-A-S-H was the main hydrate formed.
Our present study demonstrates that the interplay of sodium, nitrate and aluminium in the presence of magnesium and silicate leads to several different phases: M-A-S-H as the main hydrate, hydrotalcite-like phase possibly containing nitrate and a zeolitic gel phase, which limited the Al-uptake into M-A-S-H compared to systems without sodium nitrate. The 23 Na MAS NMR data indicated that minor amounts of hydrated Na + is present at cation exchange sites to compensate the negative charge of M-S-H.
Additionally, in the M-A-S-H xN y samples, the presence of sodium nitrate and/or the lower pH of the solution seems to slow down kinetic of formation of M-A-S-H phase compared to the pure M-S-H [32] and M-A-S-H [17] phases, but the temperature increase from 20°C to 50°C fasten the formation of M-A-S-H phases.
A similar complex mixture of different magnesium and aluminium containing solids can also be expected at the surface of cements exposed to seawater or at the interface with clays, where the presence of alkali sulphate, alkali carbonate and/or alkali chloride may influence the composition of the M-A-S-H phases by the formation of sulphate, carbonate and chloride hydrotalcite-like and zeolite-like phases. The further development of thermodynamic models for aluminium uptake in C-A-S- H, M-A-S-H and hydrotalcite-like solids are very essential to describe changes at the interface of cement pastes with a magnesium containing environment or in cementitious materials based on magnesia-silica.

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.