1 Introduction

Calcium sulfoaluminate (CSA) cements were developed with the goal of reducing shrinkage-induced cracking [1]. It is a well-known fact that the hydration of CSA cement results in formation of ettringite which can cause expansion [2,3,4,5]. Ye’elimite (4CaO·3Al2O3·SO3) is one of the main phases present in CSA cements along with belite (2CaO·SiO2), calcium sulfate (CaSO4), and calcium aluminoferrite (4CaO·Al2O3·Fe2O3) [6,7,8,9,10]. Various kinds of cements, ranging from rapid-hardening to expansive, can be produced based on the amount of added gypsum, thereby controlling the gypsum/ye’elimite molar ratio [11]. Previous research has shown that the type of sulfate source (i.e., gypsum or anhydrite) affects hydration of CSA cement [12]. A significant number of studies on hydration, phase evolution, property and microstructure development of CSA based cements are available in the literature [13,14,15,16,17,18,19,20].

The use of mineral admixtures such as fly ash along with CSA cement is often adopted for their favorable effects on durability and strength [21,22,23,24]. However, there are limited number of studies on physico-chemical interaction of fly ash with CSA cement [24,25,26]. Garcia et al. [22] studied the sulfate-containing phase dissolution of fly ash-CSA blends and no strong evidence of fly ash reactivity with CSA cement was reported. However, Martin et al. [25] reported that the addition of fly ash to CSA cement led to higher degree of ye’elimite reaction based on heat evolution. Also, the fly ash particles had dissolution rims around them, indicating some degree of reaction. Based on selective dissolution method, Martin et al. [25] reported degree of reaction of fly ash to be around 20% when fly ash replacement level was 45%.

Although there are a few studies on the hydration characteristics of CSA cement and fly ash blends, the studies on the expansion behavior of CSA cement in presence of fly ash are scarce. The addition of fly ash in OPC–CSA blend is expected to reduce the embodied carbon of binder due to replacement of OPC, improve workability of the binder at early-age, and increase the durability due to densification of matrix through pozzolanic reaction. In an earlier study by the authors [26], various OPC-CSA-MA (where MA stands for mineral admixtures such as fly ash and silica fume) ternary cementitious blends were studied to provide a better understanding of the factors affecting early-age expansion of such systems. OPC was replaced with Class C fly ash (CFA) and Class F fly ash (FFA) to understand their influence on expansion characteristics. It was observed that CFA reduced the duration and magnitude of expansion, whereas FFA increased the expansion. Moreover, expansion of OPC–CSA–CFA was complete in 2 days, while expansion of OPC–CSA–FFA continued for 5–7 days. The reduction in the extent and duration of expansion of OPC–CSA–CFA blend was hypothesized to be due to early depletion of sulfates which was attributed to the presence of C3A in CFA.

This work extends the findings of the previous study by the authors [26] in which fly ash type, i.e., CFA and FFA, were shown to influence the expansion characteristics differently. This paper presents a careful examination of the above-mentioned hypothesis, i.e., early depletion of sulfates in presence of CFA ceased the expansion process within 2 days. In order to better understand the influence of sulfates in an OPC–CSA–CFA blend, gypsum was used as an external additive.

2 Experimental program

2.1 Materials

Type I Portland cement (from Essroc Co.) and a commercially available CSA-based expansive cement (sold as KSC Komponent by CTS Cement) were used in this study. Table 1 shows the oxide composition of all raw materials. The phase composition of raw materials was determined using quantitative X-ray diffraction (QXRD) analysis, and is given in Table 2. For QXRD, Rietveld analysis was performed using rutile as an internal standard.

Table 1 Oxide composition of raw materials
Full size table
Table 2 Phase composition (wt%) of raw materials obtained using QXRD
Full size table

Cement paste samples were prepared with a constant water-to-cementitious material ratio (w/cm) of 0.44. The mixing was performed based on ASTM C192 [27]. A portion of OPC was replaced by the CSA-based expansive cement (Komponent), fly ash (CFA and FFA) and gypsum. In the mixtures without gypsum, CSA cement and CFA were used as 15% replacement (by weight) of the total cementitious materials. Additional mixtures were prepared by adding 5%, 15%, and 20% (by weight of CSA cement) of additional gypsum while reducing the amount of all other cementitious materials proportionally to maintain the total as 100%. Table 3 presents the various mixtures studied in this study.

Table 3 Mixture proportions (wt%)
Full size table

2.2 Methods

An isothermal conduction calorimeter was used to monitor the heat evolution of CSA-based blends. Small samples for isothermal calorimetry were prepared separately following the mixture proportions mentioned before. The paste samples were mixed manually in glass ampules for a period of 2 min at room temperature (22 °C ± 2 °C) before inserting them in the respective channels. It is noted that the data for the first 45 min was not considered for analysis to avoid the heat associated with mixing and placing the externally prepared samples, and to allow the samples to stabilize at the set temperature of 22 °C. The samples were monitored continuously for 72 h.

Length change measurements for each mixture were performed on three prismatic samples (25 mm × 25 mm × 285 mm). After casting, samples were initially cured for 24 h in sealed condition, and then kept in saturated lime water curing at 22 °C ± 2 °C. Moreover, Dynamic modulus of cement paste in saturated surface dry condition was measured according to ASTM C215 [28]. The samples of size: 38 mm × 38 mm × 150 mm were used for testing.

Capillary porosity of the cement paste samples was determined by solvent exchange method [29]. The slices of 25 mm (diameter) × 1–2 mm (thickness) were cut using a low-speed diamond saw and immersed in isopropyl alcohol for 24 h. Afterward, the sample volume was determined by the buoyancy method using isopropyl alcohol as the liquid medium. Subsequently, the samples were dried in a vacuum desiccator. The weight difference due to drying was attributed to the removal of capillary water. Finally, capillary porosity was calculated by dividing capillary pore volume by the sample volume.

The evolution of ettringite was monitored using QXRD. The dry samples used for the porosity test were ground to the fineness of less than 40 μm before utilizing them for XRD. Siemens X-ray diffractometer operating at 40 kV and 30 mA was utilized for the XRD analysis. Samples were scanned using CuKα radiation (wavelength: 1.5405 Å) with a step size of 0.02° and a speed of 0.5° per minute. After a thorough qualitative analysis using the MDI Jade software (2009) with the objective of identifying all the crystalline phases, the HighScore Plus (Version 3.0e) software was used for Rietveld analysis [30]. Rutile (TiO2) was used as an internal standard (15 wt%) to determine the ettringite content, accounting for the amorphous phase.

Chemical analysis was performed on the extracted pore solution of hardened cement pastes after 1, 3, and 7 days. Samples of 2 inch (50 mm) diameter and 3 inch (75 mm) height were used for pore solution extraction. Pore solution was extracted from OPC–CSA–CFA and OPC–CSA–CFA–G15 samples according to a previously proposed method [31]. After extraction, the pore solutions were filtered using a 0.2 μm filter. Subsequently, the filtered pore solution was treated with 6.3% HNO3 to prevent carbonation, and was subsequently analyzed for the determination of Ca, Al, Si, S, Na and K element concentrations using Inductively Coupled Plasma Optical Emission Spectrometry (ICP-OES). The elemental concentrations have been provided in a supplementary file. The error in the measurements was within ± 10%. After elemental analysis of the pore solution, thermodynamic calculations were performed using a geochemical modeling program: GEMS-PSI [32, 33], which was supplemented with cement specific thermodynamic database [34].

3 Results: effects on early-age properties

3.1 Hydration

One of the first indication that a ternary mix with CFA behaves differently from the other mixes with CSA cements came from early-age hydration study using isothermal calorimetry. The authors have reported earlier that the calorimetric curves, as shown in Fig. 1a, with CFA and FFA are different [26]. The results are incorporated briefly in this section of the current paper in order to put the rest of the results section in perspective. As evident from the graph, the presence of CFA delayed the hydration of OPC–CSA blend and resulted in a shoulder peak at around 44 h. The appearance of the shoulder peak in the calorimetric curve of OPC–CSA–CFA blend was hypothesized to be due to the early depletion of sulfates. In order to test this hypothesis, heat of hydration of OPC–CSA–CFA blend was monitored in the presence of external gypsum, as shown in Fig. 1b. Since there was significant time difference between the experiments of this work (Fig. 1b) and the previously published work (Fig. 1a), the occurrence of shoulder peak in OPC–CSA–CSA blend has shifted by a few hours (the new position is around 38 h). This difference can be attributed to compositional changes in the raw materials. However, it is evident from Fig. 1b that addition of 5% external gypsum delayed (i.e., right shift) the appearance of the shoulder peak. Further increase in the dosage of the external gypsum to a 15% and 20% level caused complete disappearance of the peak. This confirms that the occurrence of shoulder peak is associated with the depletion of gypsum in presence of CFA, leading to the formation of AFm phase [35]. As shown in Table 2, CFA contains 7% C3A which also competes for gypsum apart from ye’elimite in CSA cement. Since there was no C3A in FFA, there was more gypsum available for ye’elimite hydration in presence of FFA than in CFA.

Fig. 1
figure 1

Rate of heat evolution of: a OPC–CSA–CFA and OPC–CSA–FFA blends (adapted from Ref. [26]), and b OPC–CSA–CFA blend with different dosages of gypsum

Full size image

3.2 Unrestrained deformation

Figure 2 shows the influence of fly ash type and gypsum addition on early-age expansion behavior of OPC–CSA blends, respectively. The main difference between the two types of fly ashes (FFA and CFA) was reflected in the duration and magnitude of expansion. The expansion ceased earlier in OPC–CSA–CFA blend than in OPC–CSA and OPC–CSA–FFA blends. Figure 2 shows that the early age expansion of OPC–CSA–CFA blend increased as 15% and 20% gypsum was added. In addition to the two dosages of gypsum presented in Fig. 2, higher dosages were also studied. For example, 70% gypsum addition exhibited up to 3% expansion which eventually led to appearance of visible cracks on the specimen (data not shown in the Fig. 2). Therefore, it is clear that gypsum addition progressively increases the magnitude of expansion in OPC–CSA–CFA blend. The results highlight a need for adjusting sulfate content in a ternary mix with fly ash and CSA cement. To achieve similar level of expansion as that offered by a mixture without fly ash or with FFA, the mixture with CFA may require addition of external gypsum. Of course, the optimal dosage of gypsum will depend on the C3A content of the specific CFA.

Fig. 2
figure 2

Influence of fly ash type and external gypsum on the unrestrained expansion of OPC–CSA and OPC–CSA–CFA blends, respectively (Note: length change is plotted from 1 day.)

Full size image

4 Discussion: altered expansion behavior due to physico-chemical changes

4.1 Sulfate availability for ye’elimite hydration

In Portland cement, the gypsum content is optimized for desired setting time, dimensional stability and strength. It is well-known that OPC does not contain enough gypsum for its complete reaction with C3A. Once free sulfate in the pore solution is consumed, C3A is known to react with AFt phase and convert it into AFm phase. Since the optimal gypsum content in OPC ensures dimensional stability, calcium sulfate-to-C3A ratio of Portland cement was used to calculate available sulfates for ye’elimite reaction in CSA-based binders. For example, the calcium sulfate-to-C3A ratio was 0.38 for the OPC used in this study. In this calculation, the calcium sulfates include gypsum, anhydrite, and hemihydrate. It is acknowledged that the dissolution kinetics of these phases are different [12], and will therefore influence the availability of sulfate in the pore solution. However, for simplicity, all forms of calcium sulfates were taken into account to determine the total availability of sulfates. It is understandable that the addition of CFA will create an imbalance in available sulfates for ye’elimite hydration due to the presence of C3A in CFA. In order to calculate the available sulfates for ye’elimite, it was necessary to determine the fraction of C3A in each mixture. Now, based on the calcium sulfate-to-C3A ratio for OPC, the calcium sulfate used by C3A was estimated by multiplying the C3A fraction in each mixture and calcium sulfate-to-C3A ratio (i.e., 0.38). Hence, the available sulfates for ye’elimite was obtained by subtracting the sulfates used by C3A from the total sulfates (i.e., calcium sulfate) available in the mixture. The amount of sulfates calculated this way was assumed to have reacted with ye’elimite to form ettringite and result in expansion.

Figure 3 compares the length change of all paste samples (uncracked) against the available sulfates (i.e., calcium sulfate) for ye’elimite hydration. Even though this is a simplistic calculation that neglects the rate of dissolution of different forms of sulfate and also the difference in the rate of reaction of C3A and ye’elimite, it is evident from the graph that the available sulfates for ye’elimite hydration is strongly correlated with the measured expansion. As the amount of sulfates available for ye’elimite hydration increases, there is a consequent increase in expansion. The difference in the amounts of available sulfates in OPC–CSA–CFA and OPC–CSA–FFA blends is also evident, which explains the observed difference in expansion. Although these are the chemical factors, the role of physical factor such as early-age stiffness cannot be underemphasized. A previous study by the authors [35] has shown that the reduced material stiffness in presence of FFA increases the expansion potential compared to CFA. This explains why the data point representing OPC–CSA–FFA blend in Fig. 3 appears to be an outlier with higher expansion than what would be expected based on available calcium sulfate for ye’elimite hydration.

Fig. 3
figure 3

Expansion of CSA-based blends plotted against the available calcium sulfate (all forms) for the ye’elimite hydration

Full size image

4.2 Saturation level of ettringite

The main driving force for expansion in OPC–CSA blends is the supersaturation with respect to ettringite which gives rise to the crystallization stress [36, 37]. Hence, an attempt was made to estimate saturation level of ettringite OPC–CSA–CFA blend with/without gypsum. Chemical analysis (see supplementary file for the detailed elemental analysis) of the extracted pore solution was carried out using ICP-OES. Figure 4a shows the evolution of pH in OPC–CSA–CFA and OPC–CSA–CFA–G15 blends. It is evident that the addition of 15% gypsum resulted in a drop of pH of the pore solution, confirming the results of a previous study [38]. As expected, Fig. 4a shows an increase in the pH of the pore solution with time due to continuous release of alkalis during Portland cement hydration.

Fig. 4
figure 4

a Pore solution pH, b saturation level of ettringite, and c ettringite content in OPC–CSA–CFA and OPC–CSA–CFA–G15 blends

Full size image

Using the elemental concentrations of Ca, Na, K, S, Al, and Si, saturation index with respect to ettringite was calculated using GEMS-PSI. Figure 4b compares the saturation indices of ettringite in OPC–CSA–CFA and OPC–CSA–CFA–G15 blends. The presence of additional gypsum led to an increase in saturation levels of ettringite, which emphasizes the role of sulfate concentration on the saturation index of ettringite, and hence, on expansion characteristics [38]. Higher supersaturation should lead to increased precipitation of ettringite in the OPC–CSA–CFA–G15 blend. This was verified by measuring ettringite content at the age of 1, 3, and 7 days using Rietveld analysis, as shown in Fig. 4c. The powdered samples, after solvent exchange and drying, were intermixed with 15% rutile as an internal standard to determine the amorphous phase (i.e., C–S–H), and ettringite content was calculated after accounting for the amorphous content in the samples. OPC–CSA–CFA–G15 blend had higher ettringite content than OPC–CSA–CFA blend, particularly after 3 and 7 days. Higher supersaturation with respect to ettringite and ettringite content in presence of external gypsum explain the observed increase in the expansion of blends with external gypsum.

4.3 Crystallization stress-induced expansion

After estimating the saturation level of ettringite, the crystallization stress was calculated in order to predict the expansion. The measured strains, as determined through corrugated tube tests, were compared against the calculated strains resulting from the crystallization of ettringite. Maximum crystallization stress (σc) resulting from supersaturation of ettringite crystals was estimated using the following equation [39]:

$$\sigma_{\text{C}} = \frac{RT}{\nu }\ln \left( {\frac{K}{{K_{\text{sp}} }}} \right)$$
(1)

where R is the ideal gas constant: 8.314 J/(K mol), T is the absolute temperature (°K), and ν is the molar volume of ettringite crystal. In order to estimate the longitudinal strain from the crystallization stress (σc), following relationship was used [40]:

$$\epsilon = \frac{{\sigma_{\rm c}}}{3}\left({\frac{1}{K} - \frac{1}{{K_{\rm s}}}} \right)$$
(2)

The bulk modulus (K) was calculated from the measured dynamic modulus (E) and assumed Poisson’s ratio of 0.2 for isotropic and homogenous matrix using the relation: K = E/3(1 − 2ν), and Ks was assumed to be 48 GPa [41]. Figure 5 compares the measured strains with the calculated strain from 1 to 7 days for OPC–CSA–CFA and OPC–CSA–CFA–G15 blends. The effect of gypsum addition on calculated strains is evident which also matches with the trend in measured strains. However, in both cases, the experimentally measured strains are 3–4 times higher than the calculated strains due to the assumption that crystallization induced strains are elastic. This assumption is necessary to come up with a simplistic method for strain prediction, however, authors acknowledge that the assumption is not completely satisfactory at early-age when the influence of tensile creep is expected to be high.

Fig. 5
figure 5

Comparison between experimental and calculated elastic strain for OPC–CSA–CFA and OPC–CSA–CFA–G15 blends

Full size image

4.4 Evolution of porosity and dynamic modulus

As shown earlier, the expansion of OPC–CSA–CFA blend increased with the addition of gypsum. Therefore, it becomes important to ensure that there was no internal cracking due to higher degree of expansion. Furthermore, it is important to verify that the increased expansion is due to chemical factors discussed earlier such as increased supersaturation with respect to ettringite and higher degree of ettringite precipitation, not simply due to the opening of cracks. Figure 6 compares the dynamic modulus and porosity of OPC–CSA–CFA and OPC–CSA–CFA–G15 blends with time. As shown in Fig. 6, OPC–CSA–CFA–G15 blend showed higher porosity and reduced modulus at early age (up to 3 days) than that of OPC–CSA–CFA blend. After 7 days, the modulus and porosity of the samples with/without external gypsum was similar. Moreover, the decrease in porosity and the increase in dynamic modulus of OPC–CSA–CFA–G15 blend with age indicate that the structural integrity of the sample was maintained during the expansion process. It is also important to note that the difference in early-age modulus of the samples is expected to influence the deformation characteristics, and therefore, needs to be accounted for.

Fig. 6
figure 6

Effect of external gypsum on porosity and dynamic modulus of OPC–CSA–CFA paste

Full size image

5 Conclusions

In this study, the effect of Class C fly ash on expansion characteristics of OPC–CSA blend was examined by varying the dosage of external gypsum. Early depletion of gypsum in OPC–CSA–CFA blend reduced the duration and magnitude of expansion. The calorimetric peak in OPC–CSA–CFA blend, at around 2 days, disappeared when sufficient amount of external gypsum was added. This particular behavior of Class C fly ash was dependent on its C3A content which was found to be competing for sulfates, resulting in reduced amount of sulfates available for ye’elimite hydration. The addition of gypsum increased the expansion due to higher crystallization stress derived from increased supersaturation and amount of ettringite. Furthermore, by monitoring the evolution of elastic modulus with time it was verified that the increased expansion in the samples with external gypsum was not simply due to the opening of cracks.