Reactivity of C3S and model cement in presence of Na2S2O3 and NaSCN

The impact of NaSCN and Na2S2O3 on the reactivity, microstructure and morphology of C3S and model cement (with a clinker containing 85% C3S and 15% C3A) pastes was systematically investigated. Results concluded that both alkali salts mainly act enhancing the reactivity of the C3S phase while not significant influence on the reactivity of C3A was measured. While both admixtures rose the reactivity of C3S over the studied 7 days of hydration, they only increased the reactivity of model cement pastes up to 14–20 h. NaSCN and Na2S2O3 did not modify the C–S–H stoichiometry but they influenced its morphology. In particular, thicker convergent C–S–H needles were formed in pastes containing Na2S2O3 compared to non-admixed systems, while a higher number of thinner C–S–H needles were formed in presence of NaSCN. Furthermore, greater portlandite clusters and intermixing of AFm and C–S–H were observed in admixed C3S and model cement pastes, respectively, compared to plain systems.


Introduction
Accelerating admixtures are widely used chemicals to enhance the reactivity of cementitious materials [1]. Their addition to concrete formulations enables to shorten the hardening time of precast concrete, the production of ready-mix concrete at low temperatures and a fast setting for shotcrete. Furthermore, accelerating admixtures have become nowadays essential, to increase the early reactivity and mechanical performance of modern sustainable concrete with a high replacement of clinker by supplementary cementitious materials (SCMs) [2][3][4][5][6].
CaCl 2 is the most effective and widely studied accelerator for Portland cementitious systems [7,8]. Juenger et al. [9] concluded from soft X-ray transmission microscopy, that CaCl 2 accelerated the reactivity of C 3 S forming a less dense C-S-H with respect to non-admixed sample that favoured the diffusion of ions and water to the C 3 S particle and its further reactivity. Despite the efficiency of CaCl 2 as an accelerator, its dosage is limited in reinforced concrete to reduce the risk of steel corrosion.
Alkali salts such as NaOH and Na 2 SO 4 have also been shown to increase the early reactivity of alite and Portland cement [10,11]. Kumar et al. [12] reported that the increase of the pH of the solution promoted the faster precipitation of portlandite that enhanced the undersaturation of the C 3 S and increased the reaction rate of alite. However, alkalis may have a negative impact on mechanical properties of cement mortars after 1-3 days [11,12] that it is directly related to the phase assemblage of the admixed mortars and their capillary porosity. At equal alkali concentration and degree of hydration, Mota et al. [11] reported that Na 2 SO 4 lead to similar compressive strengths as for alkali-free-mortars, whereas NaOH decreased 25-35% up to 90 days. These authors showed that the addition of 0.725 M Na 2 SO 4 solution to white Portland cement favoured the formation of ettringite and a decrease of porosity with respect to nonadmixed cement [11], while NaOH inhibited the formation of ettringite and increased the porosity. Furthermore, Na 2 SO 4 led to a divergent C-S-H morphology (needles do not merge to the same point) while NaOH promoted a more planar C-S-H, but these changes did not affect the mechanical strength.
Other soluble inorganic salts such as thiocyanates or thiosulfates are commonly used effective accelerators, however, very few systematic studies on their influence on the reactivity and microstructure of cementitious materials has been reported and their working mechanisms still remain unknown [13][14][15]. Abdelrazig et al. [13] reported that the addition of 1 wt% NaSCN to Portland cement did not modify the extension of the induction period but it enhanced the intensity of the main peak in the calorimetry curve and the reactivity of C 3 S after 6 h of hydration. The addition of 0.2 wt% NaSCN to OPC-fly ash blends enhanced more significantly the reactivity of the C 3 S than the aluminate phases and increased around 15% and 5% the mechanical strength after 2 and 28 days of curing at 20°C, respectively [4].
Understanding which mineralogical phases of Portland cement the accelerating admixtures act on involves a great difficulty due to the complex composition of the material and the simultaneous reactions that take place. For this reason, in this paper the effect of two widely used accelerating admixtures, namely sodium thiosulfate (Na 2 S 2 O 3 ) and sodium thiocyanate (NaSCN) on the reactivity and microstructure of model cementitious materials that contain the most relevant mineralogical phases of cement, C 3 S and C 3 A were examined. By determining which admixtures acts on which mineral phase and at what speed more knowledge-based design of admixtures could be envisaged, something of great importance for enhancing the early mechanical performance of low clinker concrete.

Synthesis of C 3 S and model cement
Pure tricalcium silicate (C 3 S) was synthesized by homogeneizing the stoichiometric quantities of CaCO 3 (Calcium carbonate, reagent grade, Scharlau) and SiO 2 (Silica gel, Merck KGaA) in ethanol. After evaporation of the ethanol, 5 g pellets were pressed and heated at 1500°C in platinum crucibles for 16 h. The sample was afterwards quickly quenched at room temperature in air. Samples were ground and the thermal treatment was repeated a second time.
A model clinker containing 85 wt% of C 3 S and 15 wt% of C 3 A (model clinker) was synthesized by using a similar procedure as described by Marchon et al. [16]. SiO 2 , CaCO 3 , Al 2 O 3 (Aluminium oxide anhydrous, Merck KGaA) and MgO (Magnesium oxide 90%, Panreac) in a molar ratio of 1:2.23:0.198:0.058 were homogenized in ethanol for 1 h. After the evaporation of the ethanol, 5 g pellets were pressed and treated at 1500°C for 16 h in platinum crucibles. Samples were quickly quenched in air afterwards.
Around 50 g of C 3 S and model clinker samples were dry ground in a planetary ball mill (Pulverisette, Fritsch) using a 500 ml jar and agate beads and sieved afterwards through a 45 lm mesh. Table 1 shows the mineralogical composition of both synthetic phases determined by the Rietveld analysis of the XRD patterns (see Figure S1 of Supplementary material). The particle size distribution (see Fig. 1) was measured by laser diffraction (MALVERN MASTERSI-ZER S) using isopropanol as dispersant and the optical model parameters described in [17]. Table 2 shows the characteristic particle diameters (volume based) of the synthetic phases and Fig. 1 the particle size distribution.  Figure S2 in the supplementary material.

Preparation of pastes
C 3 S and model cement pastes were prepared at a liquid/solid ratio of 0.35 and 0.4, respectively. 10 g of solid were mixed with ultrapure water (18.2 MX . cm by a Milli-Q A ? water purification system from Millipore, Merck & Cie) at 200 rpm for 30 s and 800 rpm for 3 min with a 2-bladed propeller stirrer (JANKE KUNKEL IKA-WERK RW 20). 0.3wt% of NaSCN (NaSCN 8 M from Sigma Life Science) and 0.3 wt% and 2 wt% of Na 2 S 2 O 3 (99% purity from Acros Organics) by weight of powder (bwp) was added into the mixing water. These dosages previously showed to effectively accelerate the reactivity of Portland cement pastes by isothermal calorimetry (see Figure S3 in the supplementary material).

Hydration kinetics
The hydration kinetics were determined using an isothermal calorimeter TAM Air (TA Instruments) set Fig. 1 Particle size distribution of C 3 S and model clinker at 25°C. 5 g of the pastes was introduced into the calorimeter. The measurement during the first 30 min were not considered as this is the time required for the equipment to stabilize the temperature after introducing the sample. An enthalpy of hydration of C 3 S and C 3 A of 517 J/g [19] and 1340 J/g [20], respectively, were used to calculate the degree of reaction of the C 3 S and the model cement.

Mineralogical characterization of the solids
The hydration of C 3 S and model cement pastes was stopped after 8 h, 1, 2 and 7 days of reaction by mixing with isopropanol (1 g of paste ? 10 g of isopropanol) for 1 min. The suspension was filtered afterwards through a nylon filter with a pore size of 0.45 lm and the powder was kept in a desiccator at low vacuum (700 mbar) to avoid ettringite dehydration [21].
Phase identification including Rietveld quantitative analysis was carried out on data collected with a XPert PRO MPD (PANalytical) diffractometer (located at SCAI of University of Malaga) in a H-2H configuration using a CuKa 1 (1.5406 Å ) radiation (monochromatized with a primary Ge (1 1 1) monochromator). The samples were scanned between 5°and 70°, with a step size 0.0167 o and with a detection system that consists of a XCelerator RTMS (Real Time Multiple Strip) constituted by 128 Si detectors. The internal standard method was applied to quantify the phase content using quartz (Silicon (IV) oxide, 99.5%, Alfa Aesar) as a standard according to the methodology proposed in the literature [22].
Rietveld Analyses were performed with TOPAS software (Bruker). The overall refined parameters were: phase fractions, zero of goniometer, unit cell, crystal size and strain (in C 3 S) and preferred orientation when appropriated. The Brindley microabsorption correction [23] was applied to the anhydrous and stopped pastes with internal standard by a postanalysis mathematical treatment.
Thermogravimetric analysis (TGA) of the pastes at different hydration times was carried out by using a TGA-DCS-DTA Q600 (TA instruments) equipment. Around 40 mg of sample in an alumina (Al 2 O 3 ) crucible was heated from 25 up to 1000°C with a rate of 10°C/min under a 100 ml/min flow of N 2. Bounded water, portlandite and calcite content were calculated according to [24]. These data were used in conjunction with the quantitative Rietveld analysis of the XRD patterns to determine the phase assemblage of the pastes as a function of time [22,25].

Morphological and microstructural
characterization of the pastes C 3 S and model cement pastes, with and without admixtures, were prepared according to Sect. 2.2, cast in 1 9 1 9 0.5 cm 3 moulds and cured during 7 days at 25°C and 99% RH. Afterwards the samples were submerged in isopropanol for 5 days to stop the hydration and subsequently kept in a desiccator until constant weight. The microstructure was analyzed by backscattered scanning electron microscopy using a Hitachi S-4800 microscope. Energy dispersive X-ray (EDX) measurements were done with an Oxford Instruments X-Max detector. Hydrated samples at 7 days were previously embedded in epoxy resin, polished and carbon coated.
The surface of hydrated C 3 S samples with and without admixtures, at a degree of reaction of 9% (according to the calorimetry curves), was studied by scanning electron microscopy (SEM) using a JEOL JSM 7600F microscope. The hydration was previously stopped with isopropanol as explained in Sect. 2.4. The dried powder was dispersed on an adhesive carbon tab and coated with a chromium film using a modular coating system for the deposition of electrically conductive films (Quorum Q150T E).
The specific surface area of hydrated C 3 S samples (with a DoR = 9%) were measured by using a BET  Figure 2 shows the evolution of the heat flow and cumulative heat of C 3 S and model cement pastes, with and without admixtures. In admixture-free pastes ( Fig. 2a, b, black lines), a delay of the time of appearance of the peak assigned to the silicates hydration is observed in model cement pastes with respect to C 3 S pastes. This can be explained by the well-known retardation induced by the aluminates released by the aluminium-doped alites (monoclinic polymorphism) that are not present in pure C 3 S (triclinic polymorphism) [27,28].

Hydration kinetics
In C 3 S pastes (Fig. 2a), the addition of NaSCN and Na 2 S 2 O 3 reduces the time of appearance of the main peak in the calorimetry curve and enhances the slope of the acceleration period with respect to the admixture-free paste. At equal dosages, 0.3 wt% of NaSCN and Na 2 S 2 O 3 in C 3 S pastes induce a similar acceleration of around 30 min with respect the plain sample. An increase of the dosage of Na 2 S 2 O 3 up to 2 wt% accelerated its appearance by 1 h and doubled the intensity of the main peak from 5 mW/g C 3 S to 10 mW/g C 3 S. Both, Na 2 S 2 O 3 and NaSCN enhanced the cumulative heat of C 3 S over the 7 days as shown in Fig. 2b. At 8 h of hydration, the addition of 2 wt% Na 2 S 2 O 3 led to the highest values of cumulative heat (with an increase of around 62% with respect to nonadmixed C 3 S) while at 7 days of hydration, both, 0.3 wt% NaSCN and 2 wt% Na 2 S 2 O 3 enhanced up to 12% of the cumulative heat with respect to the plain C 3 S.
The acceleration of the C 3 S reactivity by alkalis have been previously explained by the promotion of portlandite precipitation with the increase of the pH of the solution that reduces the amount of Ca 2? in solution, increases the level of C 3 S undersaturation and consequently enhances its dissolution [3,12]. When we look in further detail the heat flow evolution of the C 3 S pastes over the first 3 h, a slight acceleration of the time of appearance and the intensity of the first shoulder normally assigned to the initial portlandite precipitation is observed in presence of 0.3 wt% NaSCN with respect to plain C 3 S as previously reported for KOH and NaOH in alite pastes [12]. In contrast, the addition of Na 2 S 2 O 3 delayed the time of appearance of this shoulder with respect to the freeadmixtures sample. Similar extension of the induction period has been observed in alite samples containing gypsum [29]. To understand the impact of both alkali salts on portlandite precipitation, a further analysis was done by TGA and is discussed in Sect. 3.2.
In the model cement pastes (Fig. 2c, d), the addition of 0.3 wt% of NaSCN accelerated by 1 h the time of appearance of the main peak associated mainly to the silicate hydration (observed at around 3.5 h) while it delayed and decreased the intensity of the second peak associated with the renewed hydration of C 3 A after sulphate depletion [21]. 0.3 wt% of Na 2 S 2 O 3 did not affect the position and intensity of the aluminates peak, while it accelerated by 1 h the time of appearance of the silicates peak. At dosages of 2 wt% of Na 2 S 2 O 3 , a broad peak is observed and silicates and aluminates hydration peaks cannot be clearly distinguished that infers that the reactions of C 3 S and C 3 A are very close in time. Similar effect has been reported in the literature when NaOH or Na 2 SO 4 were used as accelerators in cement systems [5,11]. Both, NaSCN and Na 2 S 2 O 3 had a positive impact on the cumulative heat of the model cement over the first 14 h, while afterwards the addition of 2 wt% Na 2 S 2 O 3 reduced the cumulative heat with respect plain model cement pastes.
3.2 Chemically bounded water and phase assemblage

Thermogravimetry analysis
An increase of the amount of chemically bounded water in plain and admixed pastes was measured with the increase of the hydration time as shown in Fig. 3. This increase in chemically bounded water obtained by TGA ( Figures S4 and S5 in the Supplementary Material), and heat released followed a linear trend for all pastes studied as shown in the Figure S6 in the Supplementary Material. In C 3 S pastes, the addition of 2 wt% Na 2 S 2 O 3 and 0.3 wt% NaSCN led to the highest values of bounded water compared to plain pastes that infers the highest formed amount of hydration products. In particular, the addition of 2 wt% Na 2 S 2 O 3 increased, respectively, 19% and 5% the amount of bounded water at 2 and 7 days of hydration. Moreover, 0.3 wt% NaSCN increased around 6-14% the amount of bounded water at all the studied ages. At 7 days of hydration, C 3 S pastes cointaining 0.3 wt% NaSCN and 2 wt% Na 2-S 2 O 3 had similar values of bounded water contents. The effect of these admixtures on the bounded water in model cement pastes was less marked. While the presence of 0.3 wt% Na 2 S 2 O 3 and NaSCN increased around 1.5-2.5% the bounded water with respect the plain pastes, a decrease of around 3.5% was observed with 2 wt% Na 2 S 2 O 3 after 7 days of reaction. This reduction of the amount of chemically bounded water of the model cement in presence of 2 wt% Na 2 S 2 O 3 agrees with the reduction of the cumulative heat measured by isothermal calorimetry shown in Fig. 2d.
The amount of portlandite with respect to the anhydrous C 3 S and model cement was calculated according to Eq. 1, considering the weight loss between 25 and 550°C measured by TGA [24]: where WL Ca(OH)2 is the water loss due to portlandite decomposition (in percentage); PM H2O is the molecular mass of water, PM Ca(OH)2 is the molecular mass of portlandite and H 2 O bounded is the amount of bounded water (in percentage) shown in Fig. 3.
The presence of both accelerating admixtures enhanced the amount of portlandite formed in C 3 S pastes from the first day of hydration as shown in Fig. 4. In particular, the addition of 2 wt% Na 2 S 2 O 3 led to the highest amount of portlandite at 1 day of hydration, with an increase of up to 30% with respect to non-admixed C 3 S pastes. In model cement pastes, the presence of 0.3 wt% of both alkali salts did not have a significant impact on the amount of portlandite formed, while 2 wt% Na 2 S 2 O 3 decreased up 5-10% the portlandite content over the 7 days of hydration. The linear correlation between the portlandite content and the degree of hydration calculated from calorimetry (see Fig. 5) concludes that the impact of both accelerating admixtures on the amount of portlandite is a consequence of its effect on the reactivity of the pastes. This suggests that both alkali salts enhanced the degree of hydration of C 3 S and consequently also the amount of portlandite formed, while the decrease of portlandite induced by 2 wt% Na 2 S 2 O 3 in model cements is due to the slowdown of its reaction after 14 h in agreement with the results obtained by calorimetry (see Fig. 2b).

Quantitative X-ray powder diffraction analysis
Rietveld quantitative phase analysis (RQPA) of the XRD patterns of non-admixed and admixed C 3 S and model cement pastes were done to establish the impact of the accelerating admixtures on the phase assemblages (type and amount of hydration products). It is worth highlighting that the starting C 3 S had around 10 wt% of amorphous fraction that might be due to its amorphisation during the dry-grinding process [30] but also to the partial prehydration of the sample as shown by the presence of Q o hydroxylated Si units in the { 1 H} 29 Si CPMAS NMR [19] in Figure S7 in the Supplementary Material. The same arguments would   Figures S10 and S11 give, as representative examples, the Rietveld plots of selected patterns. Tables S1 and S6 in Supplementary material give the phase assemblage, including the amorphous phase (ACn) and free water determined by TGA, of all pastes as a function of hydration time. Figures 6, 7 and 8 present the evolution of selected crystalline and amorphous phases (ACn) in C 3 S and model cement pastes over time. All Rietveld analysis had a R WP lower than 10% that confirms good refinements. Due to differences in the linear absorption coefficients among phases [* 310 cm -1 for C 3 S, * 270 cm -1 for C 3 A, * 40 cm -1 for CH and * 25 cm -1 for AFt] and the internal standard [* 90 cm -1 for quartz] the microabsorption effect has a relevance in the RQPA [23]. Consequently, the Brindley correction was performed in a post-refinement spreadsheet calculation. The average particle size of each phase has to be known to apply such correction. For quartz, and phases in non-hydrated C 3 S and model cement particle sizes were determined by laser diffraction (see Fig. 1) while for hydrated C 3 S and model cement pastes, particle sizes of hydrated phases were estimated from scanning electron microscopy. In C 3 S pastes, portlandite and calcite were identified as the main crystalline phases while portlandite, ettringite, calcium hemicarboaluminate and calcite were identified in the model cement pastes. In addition, calcium monosulfoaluminate was found in model cement pastes containing 0.3 wt% NaSCN. The presence of calcite and hemicarboaluminate in the samples confirm their partial carbonation due to reaction with the atmospheric CO 2 during sample preparation and measurements. The quantification of portlandite by XRD is usually underestimated due to the presence nano-portlandite with a crystal size below the size of the coherent scattering domains detectable by X-rays [31]. For this reason, in this paper, we have relied on the TGA measurements for the determination of the portlandite content as already discussed in Sect. 3.2.1. Figures 6 and 7 indicate that 2 wt% Na 2 S 2 O 3 mainly influenced the reactivity of C 3 S in pure C 3 S pastes and model cement pastes while it did not modify the amount of C 3 A in model cement pastes over the studied time. C 3 S pastes containing 2 wt% Na 2 S 2 O 3 showed respectively a decrease of 19% and 12% of C 3 S with respect plain pastes, at 8 h and 1 day,  6 Weight percentage of C 3 S and amorphous phase (ACn) in C 3 S pastes without admixtures, with 0.3 wt% NaSCN and 2 wt% Na 2 S 2 O 3 from Tables S1-S3 in the Supplementary Material while a decrease of around 4% occurred at 2 and 7 days of hydration. After 1 day of hydration, the addition of 2 wt% Na 2 S 2 O 3 to model cement decreased around 17% and 24% the amount of C 3 S compared to non-admixed paste, after 1 day and 2 days, respectively, while it did not have impact on this phase at 7 days of hydration with respect to nonadmixed paste.
The addition of 0.3 wt% NaSCN mainly enhanced the reactivity of C 3 S in the model cement. Around 20% more C 3 S reacted at 1 and 2 days of hydration but no enhancement of the reactivity was observed at 7 days of hydration, while the measured changes in the amount of C 3 A were small and inside the error analysis of the quantification method.
As explained above, the amorphous content determined by XRD corresponds to the remaining amorphous C 3 S, amorphous portlandite and C-S-H. The amount of C-S-H was consequently calculated from the portlandite content determined by TGA (see Fig. 4) and the stoichiometry for C 3 S hydration shown in Eq. 2. The C 1.67 SH 2.5 stoichiometry was established by comparing the theoretical values of the C-S-H water and portlandite with the experimental values determined by TGA [32]. Considering the amount of portlandite determined by TGA after 7 days, the total reaction of the initial amorphous C 3 S (* 10 wt%) has to be considered. Moreover, to estimate the amount of C-S-H formed, a stoichiometry has to be assumed. The Ca/Si rate from 1.6 to 1.8 and the water molecules of C-S-H from 2.1 to 4 were tested, to match the RQPA obtained and given in Tables S1 to S3. Finally, the determined stoichiometric was: C 3 S þ 3:83 H ! C 1:67 SH 2:5 þ 1:33 CH ð2Þ With this stoichiometry the water losses from C-S-H and from portlandite were theoretically calculated and compared to those obtained experimentally by TGA, Table 3. From this, it was inferred that the presence of NaSCN and Na 2 S 2 O 3 did not largely affect to the C-S-H stoichiometry. The morphology and stoichiometry of C-S-H gels are highly dependent on the ionic environment [33], however, because of the overlapping signals assigned to the dehydration of ettringite and C-S-H in TGA, the calculation of the C-S-H stoichiometry for model cement pastes was not possible.
The evolution of the amount of C-S-H formed in C 3 S and model cement pastes over time is shown in Fig. 8. In C 3 S pastes, the presence of both alkaline salts increased the amount of C-S-H up to 7 days of hydration, being more significant with the addition of 2 wt% Na 2 S 2 O 3 . In contrast, while 0.3 wt% NaSCN enhances the C-S-H precipitation over the 7 days of hydration, the addition of 2 wt% Na 2 S 2 O 3 only increases the amount of C-S-H up to 2 days. This would confirm that Na 2 S 2 O 3 only rises the reactivity of model cement pastes at early ages, as already concluded from the amount of chemically bounded water (Fig. 3b) and calorimetry measurements (Fig. 2d).

Impact of Na 2 S 2 O 3 and NaSCN
on the morphology and microstructure of the pastes

Morphology of hydrates on admixed C 3 S pastes
Isothermal calorimetry and QXRD measurements showed the effect of Na 2 S 2 O 3 and NaSCN acted mainly on the reactivity of C 3 S. For this reason, the morphology of the reaction products on these pastes was studied at a degree of hydration (DoH) of 9% (according to the isothermal calorimetry, see Fig. 2). This degree of hydration corresponded to the acceleration period where nucleation and growth of hydrates occurred. Figure 9 shows the surface of the plain and admixed C 3 S pastes covered by hydrates. C-S-H needles that grow outwards from the C 3 S surface and merge together to a same point were observed in all cases, however the presence of the admixtures seemed to influence the thickness of the C-S-H needles. While Na 2 S 2 O 3 induced the formation of slightly thicker needles, thinner C-S-H needles were formed in NaSCN-containing pastes. These results are in agreement with the SSA BET values of the pastes. In particular, SSA BET of 5.4, 4.3 and 9.9 g/m 2 were measured for plain, Na 2 S 2 O 3 -and NaSCN-C 3 S containing pastes, respectively.

Microstructure and hydrates composition
of C 3 S and model cement pastes Figures 10 and 11 show the microstructure of C 3 S and model cement pastes at 7 days of hydration. For C 3 S pastes in absence of admixtures, thin dense C-S-H rims (inner product) are observed around the C 3 S particles while outer C-S-H product with a greater porosity is formed in the initially water-filled space exterior to the original C 3 S grains. In contrast, the thickness of the inner-C-S-H rims is significantly reduced in C 3 S pastes containing both alkali salts, and outer C-S-H is mainly observed. Furthermore, NaSCN-containing C 3 S pastes showed a much greater porosity (black contrast) than the plain samples. The addition of both alkali salts also led to the formation of portlandite clusters. EDX analyses of the matrices (see Table 4) indicate a higher Ca/Si ratio for the admixed C 3 S pastes that would infer that the admixtures favour the intermixing of C-S-H and portlandite as it was  . 9 SEM images of the surface of C 3 S pastes at 9% degree of hydration, a, b plain C 3 S and C 3 S with c, d 2 wt% Na 2 S 2 O 3 and e, f 0.3 wt% NaSCN determined that they did not modify the C-S-H stoichiometry as previously described in Sect. 3.2.2.
In the model cement pastes, the addition of NaSCN tended to decrease the density of the matrix and no C-S-H ring around the anhydrous particles was observed. The addition of both alkali salts led to the precipitation of AFm phases intermixed with the outer C-S-H as it can be also observed in the two dimensional scatter plots of atomic ratio (see Fig. 12) [34,35]. From the BSEM analysis, more AFm seems to be formed in Na 2 S 2 O 3 -containing samples with respect to model cement pastes with NaSCN. The presence of NaSCN and Na 2 S 2 O 3 also lead to the formation of large portlandite clusters, as already described for C 3 S pastes (see Fig. 11).

Discussion
The discussion of our results has been divided in two main sections. In the first one, the impact of NaSCN and Na 2 S 2 O 3 on the reactivity and hydrates formation in the main cement phases is discussed based on the isothermal calorimetry, QXRD and TGA results, while the second part focuses on the morphology and microstructure of the admixed pastes.
a. Influence of NaSCN and Na 2 S 2 O 3 on the reactivity Hydration kinetics measurements have shown the faster reactivity of C 3 S and model cement (compound of a model clinker (85 wt% C 3 S, 15 wt% C 3 A) and 5.5%w/w calcium hemihydrate) in presence of 0.3 wt% NaSCN and 0.3-2 wt% Na 2 S 2 O 3 . Kumar et al. [12] reported that alkalis promoted the earlier precipitation of portlandite and end of the induction period due to the increase of the pH of the pore solution. In our study, the intensity of the shoulder assigned to portlandite precipitation in the heat flow curve (see Fig. 2) appeared at a shorter time and with an enhanced intensity in C 3 S pastes containing 0.3wt% NaSCN in agreement with Kumar et al. [12]. In contrast, a delay of the onset of the acceleration period was observed in C 3 S pastes with Na 2 S 2 O 3 , being the delay longer with the increased dosage of this admixture. A similar delay was previously observed in the literature [36] in C 3 S pastes in presence of divalent anions such as SO 4 2and explained by the electrostatic interactions between the sulfate ions and the C 3 S surface. In particular, Nicoleau et al. [36] argued that the formation of neutral CaSO 4 o species reduced the charge screening provided by calcium cations, that led to a more negative surface charge and a decrease of the C 3 S dissolution rate. A similar mechanism could be proposed to be responsible of the extension of the induction period induced by another divalent anion such as S 2 O 3 2-. Furthermore, the greater slope of the acceleration period in C 3 S pastes containing NaSCN and Na 2 S 2 O 3 compared to plain pastes, would indicate that both admixtures favoured (1) either the precipitation of a higher number of C-S-H nuclei during the induction period that grow over the acceleration period or (2) the faster growth of C-S-H [37,38]. However, with the results provided by the current study is not possible to distinguish which mechanism these accelerating admixtures mainly act. Controlled precipitation experiments of C-S-H in presence of  admixtures in combination with kinetic analysis, as described in the literature [39,40], would be required to determine how NaSCN and Na 2 S 2 O 3 influence the nucleation and growth of C-S-H. The hydration kinetics and phase assemblage studies done in C 3 S and model cement pastes concluded that both alkali salts mainly act on the C 3 S phase while not significant impact on the reactivity of the C 3 A phase was determined by QXRD (see Fig. 7b). As a consequence, both admixtures led to a higher amount of C-S-H and portlandite with respect to non-admixed pastes. However, while Na 2-S 2 O 3 and NaSCN enhanced the reactivity of the C 3 S pastes over 7 days of hydration compared to nonadmixed pastes, they only increased the reactivity of model cement pastes at early ages (up to 14-20 h). In particular, at 7 days of hydration, the addition of 2 wt% Na 2 S 2 O 3 decreased 12% the cumulative heat and 4% the bounded water. This is in agreement with previous results of Mota et al. [11] who also reported a decrease of the degree of hydration of white cement (calculated from calorimetry tests) in presence of Na 2 SO 4 and NaOH after 35 h hours although the reason still remains unclear and needs further investigation.
b. Influence of NaSCN and Na 2 S 2 O 3 on the C-S-H morphology and microstructure Na 2 S 2 O 3 and NaSCN both influenced the morphology of C-S-H and microstructure for both, C 3 S and the model cement pastes. At a degree of hydration of 9%, the addition of Na 2 S 2 O 3 to C 3 S pastes led to the formation of thicker convergent C-S-H needles and a slightly lower specific surface area than plain pastes, while a greater number of thinner C-S-H needles were formed in C 3 S-containing NaSCN with a consequent increase of the SSA BET from 5.4 m 2 /g in plain samples to 9.8 m 2 /g in those containing NaSCN. This contrasts with the results obtained by Mota et al. [41], in which divergent C-S-H needles were formed in presence of sulfates in the pore solution and C-S-H with a foil-like structure was formed in alite pastes containing NaOH, however, these authors did not report an impact of the admixtures on the thickness of the C-S-H needles. Further studies on the impact of the changes of the C-S-H morphology, and consequently, the number of contact points, on the mechanical strength are needed.
Both accelerating admixtures increased the Ca/Si of the matrices from 1.72 in the plain C 3 S, to 2.27 and 2.09 for C 3 S pastes containing Na 2 S 2 O 3 and NaSCN, respectively (see Table). This higher Ca/Si ratios could be explained by the intermixing of C-S-H with microcrystalline portlandite as the comparison of the calculated and experimental amount water molecules of C-S-H and portlandite, Table 3, concluded that the stoichiometry of C-S-H in non-admixed and admixed pastes had a constant value of C 1.67 SH 2.5 . However, both admixtures seemed to influence the distribution of portlandite in C 3 S pastes. In general, bigger clusters of portlandite were observed by BSEM in admixed C 3 S pastes after 7 days of hydration. Furthermore, the correlation of the portlandite content versus the degree of hydration (see Fig. 5) confirmed that the admixtures did not favour the precipitation of more portlandite overtime. This means that Na 2 S 2 O 3 and NaSCN changed the portlandite distribution (forming more clusters) but the higher amount of portlandite measured by TGA and XRD was associated with the higher degree of reaction induced by these admixtures.
A greater intermixing of AFm and C-S-H was observed in admixed model cement pastes with respect to the plain samples. While the binary plot concluded the formation of monosulfoaluminate, the QXRD concluded the formation of hemicarboaluminate that would indicate the possible carbonation of the samples during XRD sample preparation and analysis.
Furthermore, the addition of Na 2 S 2 O 3 and NaSCN increased the amount of outer C-S-H product and a greater porosity as qualitatively observed by BSEM, Figs. 10 and 11.
The studies described above have enabled to gain further knowledge on the influence of Na 2 S 2 O 3 and NaSCN on the reactivity and microstructure of model Portland cements with C 3 S and C 3 A as main phases. Na 2 S 2 O 3 has also been previously reported to enhance the reactivity of synthetic glasses with compositions similar to supplementary cementitious materials (SCMs) [42]. Future studies on the working mechanism and the impact of these alkali salts on the reactivity, phase assemblage and mechanical properties of real blended cements are still required. In particular, the impact of these admixtures on the reactivity of clinker phases and SCMs in blends, their synergistic reactions and its contribution to the formation of hydrates should be investigated. Moreover, the effect of these admixtures on the nucleation and growth of C-S-H needs to be better understood as well as their influence on the late hydration period. That will enable to design more effective accelerators of the reactivity of the clinker phases and SCMs to further decrease the cement clinker factor without detriment of the early mechanical properties of concrete.

Conclusions
In this paper, the impact of two accelerating admixtures commonly used, NaSCN and Na 2 S 2 O 3 , on the reactivity and microstructure of pastes of model systems, C 3 S and model cement, was investigated.
The hydration kinetics and phase assemblage studies concluded that both alkali salts mainly acted on the C 3 S phase while not significant impact on the reactivity of the C 3 A phase was measured. As a consequence, a higher amount of C-S-H and portlandite in admixed pastes was measured by QXRD and TGA while not great impact of the admixtures on the amount of the hydrated aluminate phases was determined. Na 2 S 2 O 3 and NaSCN enhanced the reactivity of the C 3 S pastes over 7 days of hydration, while they only increased the reactivity of model cement pastes up to 14-20 h.
The addition of both alkali salts did not modify the C-S-H stoichiometry but they influenced its morphology. At an equal degree of hydration of 9% of C 3 S pastes, thicker convergent C-S-H needles were formed in pastes containing Na 2 S 2 O 3 compared to non-admixed systems, while a greater number of thinner C-S-H needles were formed in presence of NaSCN.
Greater portlandite clusters and greater intermixing of AFm and C-S-H were observed in admixed C 3 S and model cement pastes, respectively, compared to plain systems. Furthermore, admixed samples showed a greater porosity, as qualitatively observed by BSEM, compared to plain cementitious systems.
Further research on the influence of these alkali salts on reaction kinetics, microstructure and macroscopic properties of real blended cements are still needed. This will contribute to develop effective accelerators to further increase clinker replacement in blends without detriment to the early mechanical properties of concrete. Moreover, the impact of Na 2 S 2 O 3 and NaSCN on the C-S-H nucleation and growth should be better understood.