Expansion mechanisms in calcium aluminate and

0 downloads 0 Views 2MB Size Report
Cement and Concrete Research 56 (2014) 190–202 ... The complete reaction of ye'elimite .... pressure until its chemical potential equals that of the small portion of ...... (Ed.), LEA's Chemistry of Cement and Concrete, Arnold, London, 1998, pp.
Cement and Concrete Research 56 (2014) 190–202

Contents lists available at ScienceDirect

Cement and Concrete Research journal homepage: http://ees.elsevier.com/CEMCON/default.asp

Expansion mechanisms in calcium aluminate and sulfoaluminate systems with calcium sulfate Julien Bizzozero ⁎, Christophe Gosselin, Karen L. Scrivener Laboratory of Construction Materials, Ecole Polytechnique Fédérale de Lausanne (EPFL), 1015 Lausanne, Switzerland

a r t i c l e

i n f o

Article history: Received 25 July 2013 Accepted 27 November 2013 Keywords: Crystal size (B) Pore solution (B) Stability (C) Ettringite (D) Crystallization pressure

a b s t r a c t The long-term expansion of calcium aluminate cement and calcium sulfoaluminate cement in the presence of added gypsum has been studied for samples cured under water. Progressively higher amounts of gypsum were added to the CAC or CSA and it was found that there is a critical amount of gypsum leading to unstable expansion and failure of the samples. The microstructures of systems with gypsum additions just below and above the threshold were similar. Pore solution analyses showed that supersaturation with respect to ettringite increases with the calcium sulfate content, which results in an increase of the crystallization pressure. The supersaturation determines the minimum pore size in which crystals can grow. Therefore with higher supersaturation a larger pore volume is accessible to growing ettringite crystals exerting pressure in the porous skeleton. This could explain the critical amount of gypsum leading to high unstable expansion. © 2013 Elsevier Ltd. All rights reserved.

1. Introduction Alternative cements such as calcium aluminate cement (CAC) and calcium sulfoaluminate cement (CSA) have useful properties when developing concretes and formulated binders with tailored behaviour, for example with fast setting and hardening, or shrinkage compensation. These cements are also interesting from the perspective of lowering CO2 emissions associated with cement production [1,2]. In particular tailored behaviour requires understanding the microstructural development to predict the properties (e.g. strength and dimensional stability). Cementitious materials undergo volumetric changes during their hydration. The porosity is initially saturated with water but when the hydration proceeds, there is self-desiccation due to it being consumed and there can be external evaporation. This leads to capillary stresses causing macroscopic shrinkage and possibly cracking [3,4]. To overcome this, expansive cements which can compensate shrinkage are used. CSA and CAC blended with calcium sulfate are chemically similar. It can be considered that the reactive phase in CSAs, C4A3$1, is chemically equivalent to C$ + 3CA (which is the reactive phase in CAC). Therefore the hydration of CAC blended with calcium sulfate and CSA is similar and both cements can be used as expansive or shrinkage compensating cements [5,6]. The kinetics of hydration and development of mechanical properties depend on the CAC:C$ or CSA:C$ ratio and the type of ⁎ Corresponding author. Tel.: +41 21 69 37786; fax: +41 21 69 35800. E-mail address: julien.bizzozero@epfl.ch (J. Bizzozero). 1 $: SO3. 0008-8846/$ – see front matter © 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.cemconres.2013.11.011

calcium sulfate (anhydrite, hemihydrate or gypsum) [5,7]. The effect of gypsum content in CSA [8–10] and CAC [11] has already been reported, showing that the expansion increases with the gypsum content. Expansion is generally attributed to the formation of ettringite. However, there is a lack of understanding of the exact link between ettringite formation and expansion. In this study we looked at the progressive addition of gypsum to both CAC and CSA, linking the expansive behaviour to microstructural changes, in order to better understand the underlying mechanisms of expansion. 1.1. Hydration in CAC + C$ systems and CSA + C$ systems The hydration of CAC and CSA systems blended with calcium sulfate is comparable. The main phases of CAC and CSA are monocalcium aluminate (CA) and ye'elimite (C4A3$), respectively. The hydration of these phases with calcium sulfate leads to the formation of ettringite and aluminium hydroxide (often poorly crystalline or amorphous), according to reactions (1) and (2).

3CA þ 3CDHx þ ð38‐3xÞH → C3 A:3CD:H32 þ 2AH3

ð1Þ

C4 A3 D þ 2CDHx þ ð38‐2xÞH →C3 A:3CD:H32 þ 2AH3

ð2Þ

with x = 0 for anhydrite, x = 0.5 for hemihydrate or plaster and x = 2 for gypsum. If the amount of calcium sulfate is less than that needed for all the CA or C4A3$ to react according to Eqs. (1) and (2), calcium

J. Bizzozero et al. / Cement and Concrete Research 56 (2014) 190–202

monosulfoaluminate will form after the depletion of the calcium sulfate [6]: 6CA þ C3 A:3CD:H32 þ 16H→3C3 A:CD:H12 þ 4AH3 C4 A3 D þ 18H→C3 A:CD:H12 þ 2AH3

ð3Þ ð4Þ

191

Where K is the ion activity product and Ksp is the solubility product of a given phase. The species forming ettringite are: 6Ca



  − − 2− þ 2AlðOHÞ4 þ 4OH þ 3SO4 þ 26H2 O→Ca6 AlðOHÞ6 2  3SO4  26H2 O

The resulting K for this phase is: Rapid strength development is related to the fast hydration of monocalcium aluminate and ye'elimite. The complete reaction of ye'elimite, gypsum and water to form ettringite and aluminium hydroxide requires a high water/binder ratio (about 0.6 for 30 wt.% of calcium sulfate) [8]. In the present study the water/binder ratio of 0.4 is insufficient for complete hydration. If calcium hydroxide is present, more ettringite can form relative to aluminium hydroxide, according the reactions (5) and (6). ð3‐yÞ CA þ 3 CDHx þ y CH þ ð38‐3x‐4yÞ H→C3 A:3CD:H32 þ ð2‐yÞ AH3 y∈½0; 2 and x∈ð0; 0:5; 2Þ ð5Þ

2  3  26  6  4 K ¼ α Ca2þ  α AlðOHÞ ˙  ðα OH− Þ  α SO2 −  α H2 O 4

4

ð8Þ

and the solubility product of ettringite at 20 °C is Ksp = 10−45.45 [25]. Supersaturation provides the driving force for the development of crystallization pressure (Pc) and expansion according to Eq. (9) (valid for large pores above 0.1–1 μm where size effects are negligible [26]). P c is the maximum pressure that can be achieved with a given supersaturation. RT K ln Pc ¼ vm K sp

! ð9Þ

C4 A3 D þ ð2 þ yÞ CDHx þ y CH þ ð38‐xð2 þ yÞ þ 26=3yÞ H→ ðy=3 þ 1Þ C3 A:3CD:H32 þ ð2‐y=3Þ AH3

ð6Þ

y∈½0; 6 and x∈ð0; 0:5; 2Þ

1.2. Theories of expansion The expansive behaviour associated with ettringite formation in cement systems has been widely studied [12–17] but the mechanisms of expansion are still not fully understood. Various mechanisms implicated in external sulfate attack of cementitious materials were well reviewed by Brown and Taylor [18]. Most of their arguments have a general validity to systems in which expansion is related to ettringite formation: the main points are summarised here. • One of the simplest ways used to explain expansion is that the precipitation of ettringite leads to an increase of solid volume [19]. However, there is a lack of evidence to link expansion directly to the amount of ettringite [20,21]. Furthermore, the formation of other hydrates, such as C–S–H, also gives an increase in solid volume, but their formation is not generally considered to be expansive. • Mehta [14] linked expansion to the adsorption of water by colloidal sized ettringite crystals. However, he gave little theoretical support for this hypothesis. It is unlikely that crystalline ettringite would show gel-like swelling behaviour and such behaviour has never been reported for synthetic preparations of this phase. A crystalline structure cannot adsorb water molecules within the crystal lattice and therefore it cannot swell by water adsorption. Furthermore he suggested that the “colloidal” ettringite formed only in the presence of lime or calcium hydroxide, whereas expansion also occurs when lime is absent, as in the present study. • Several authors have proposed that expansion is due to the solid state formation of ettringite around the cement grains, but the totally different crystal structure of ettringite from any of its precursors make such theories invalid [18,22]. These arguments leave the theory of crystallization pressure as the most plausible mechanism of expansion. This theory was developed and experimentally validated more than a century ago [23,24]. The first important aspect is that the crystal must grow from a supersaturated solution. The saturation index (SI) is defined by Eq. (7), if SI = 0 the solution is at equilibrium, if SI b 0 the solution is undersaturated and if SI N 0 the solution is supersaturated with respect to a given crystal.

SI ¼ log10

K K sp

! ð7Þ

Where: R = 8.314 J/K/mol is the gas constant, T is the absolute temperature and vm is the molar volume of the crystal (for ettringite vm = 705.8 cm3/mol using a density of 1.78 g/cm3 [27] and a molar mass of 1255 g/mol). In the case of crystals growing in small pores, where size effects are relevant, it is important to consider the interfacial free energy of the crystal–liquid interface which is described by Eq. (10) assuming spherical pores [16]. This pressure opposes the pressure generated by the salt supersaturation. There is a film of solution of thickness δ between the crystal and the pore wall that brings ions for the crystal growth. The radius rc of the crystal is given by rc = rp - δ, rp is the radius of the pore and δ is estimated to be 1–2 nm [16]. P w ¼ γCL K CL ¼ γ CL

2 2 ¼ γ CL rc r p −δ

ð10Þ

Where KCL is the curvature of the crystal and γCL is the interfacial free energy between the crystal and the liquid. The interfacial free energy of salts present in cementitious materials is expected to be highly anisotropic, moreover there are no reliable data for these values. The value used in this study (γCL = 0.1 J/m2) is just an estimate previously used by other authors [17,28]. The pressure Pw which is relevant in the case of crystals or pores below 0.1–1 μm [28] and which has to be taken into account in the determination of the net crystallization pressure exerted by the crystal on the pore wall is given by Eq. (11). ΔP ¼ P c −P w ¼

! RT K 2 ln −γCL vm K sp r p −δ

ð11Þ

Eq. (11) combines the effects of a supersaturated crystal growing in a small pore and exerting positive pressure on it and of the curvature of the crystal–liquid interface generating a negative pressure opposing its growth resulting in a net crystallization pressure ΔP. The equilibrium condition is described by Eq. (12) when the overall pressure is equal to 0 (valid for a spherical pore/crystal) (adapted from [29]). ΔP−P c −P w ¼ 0

ð12Þ

From Eq. (12) the Freundlich Eq. (13) is derived [30]: RT IAP ln vm K sp

! ¼ γ CL

2 rp −δ

ð13Þ

For a given supersaturation there is a corresponding pore radius (rp) where equilibrium conditions are met. To have crystallization pressure,

192

J. Bizzozero et al. / Cement and Concrete Research 56 (2014) 190–202

the minimum pore radius should be larger than rp. Large stresses are localized in nm size pores and that the phase generating crystallization pressures has to be confined [17]. Crystals will always grow preferentially in free space (large pores) at equilibrium conditions, without exerting expansive forces. If we consider a spherical pore with smaller cylindrical pore entrances in which a crystal is growing, the surrounding solution around the larger (low curved) portion of the crystal will have the same concentration as the solubility of the small (highly curved) portion of the crystal growing in the cylindrical pore entrance (see Fig. 1). Therefore the pore entrance determines the equilibrium solubility of the small and highly curved portion of the crystal that can just penetrate the pore. The difference between the curvature at the entrance and in the larger interior of the pore determines the supersaturation of the larger portion of the crystal [28]. The large portion of the crystal will continue to grow generating pressure until its chemical potential equals that of the small portion of the crystal at the pore entrances [31]. In real conditions the crystal growth depends also on the kinetics of diffusion of the ions to the crystal. 2. Materials and methods 2.1. Materials Two commercial cements were used, a calcium aluminate cement (CAC) Secar® 51 produced by Kerneos and a calcium sulfoaluminate cement (CSA). Table 1 shows the oxide composition of the cements (by XRF) and Table 2 the main anhydrous phases (by XRD and Rietveld refinement). These cements were blended with gypsum (Acros Organics, 98 + % purity) in different proportions, from zero to around 50 wt.%. Table 3 indicates the composition of the studied systems. To compare the molar calcium sulfate relative to CA content between the two cements, the molar quantity of ye'elimite in the CSA systems is expressed as the molar quantity of equivalent CA: C4A3$ → 3CA + C$. The calcium sulfate part of C4A3$ is taken into account for the molar percentage of gypsum. The notation adopted for the presentation of the results for CAC systems is %CA–%G and for CSA systems is %CA–%G, see Table 3. 2.2. Methods All the experiments were carried out at 20 °C. Cement pastes were prepared with a constant water/binder ratio of 0.4 (preparation of pastes at higher w/c ratios would lead to bleeding). Gypsum was first added to the mixing water (to ensure a saturated solution) which was then added to the cement and mixed for 2 min using a paddle mixer (1100 rpm).

Table 1 Oxides composition of CAC and CSA cements (wt.%).

CAC CSA

CaO

Al2O3

SiO2

Fe2O3

K2O

Na2O

TiO2

MgO

SO3

37.48 42.5

52.41 26.5

5.07 6.5

1.12 2

0.19 b0.5

0.11 b0.5

2.51 1

0.51 1.5

b0.05 19.5

For the expansion tests, the cement paste was cast in steel moulds of 10 × 10 × 40 mm3 with end pieces for the measurement of the length evolution. The samples were then cured for 1 day in the moulds in a high humidity environment (96% R.H.) and unmoulded after 24 h to start the measurements. The samples (6 prisms) were then submerged into 60–70 g of deionised water. The length of the samples (~40 mm) was measured regularly with an extensometer having a precision of ± 1 μm. The flexural and compressive strength of the paste systems were measured on 2 beams 16 × 16 × 80 mm3 at 7 days. The same curing conditions as in the expansion tests samples were followed. The hydration kinetics were studied by isothermal calorimetry (TAM Air (3114/3236) from TA Instruments). Ten grammes of cement paste (mixed outside the calorimeter) were introduced in a glass ampoule which was then sealed with a cap and introduced in the calorimeter. For SEM, XRD and MIP the samples were cast in polystyrene cylinders (35 mm ∅ × 50 mm). These cylinders were introduced in a water bath at 20 °C to maintain a constant temperature for the first 24 h. Then the samples were demoulded and stored in small quantities of deionised water to minimise leaching. At the required ages three slices of 3–4 mm thickness were cut from the cylinders and then immersed in isopropanol to stop the hydration. After 7 days in the solvent, they were stored in a desiccator under vacuum for 2–7 days and over silica gel to remove the alcohol, prevent carbonation and remove possible moisture. It is important to be aware of the fact that stopping the hydration with isopropanol and storing the samples under vacuum has a detrimental effect on the hydrates and particularly on ettringite [32,33]. For this reason it is important to follow always the same protocol to have reproducible results. X-ray diffraction (XRD) analyses were done with a Philips X'Pert Pro PANalytical (CuKα, λ = 1.54) working in Bragg–Brentano geometry with a 2θ-range of 5°–65°. The analyses were done on the dry slices cut from the cylindrical samples. Quantitative Rietveld analyses were done with the HighScore Plus software and using the external standard method (using rutile Kronos 2300, Titanium dioxide). Thermogravimetric analysis (TGA) was carried out using a Mettler Toledo TGA/SDTA851e. Around 50 mg of ground cement paste were placed in alumina crucibles covered by aluminium lids to reduce the carbonation before the analysis. The temperature ranged from 30 to 1000 °C with a heating rate of 10 °C/min under N2 atmosphere to prevent carbonation. Table 2 Mineralogical composition of the cements.

Fig. 1. Spherical pore with cylindrical pore entrances filled by solution (white). Large crystal (grey) growing with low curved portion (rp) and highly curved portion (re). Adapted from [28].

Anhydrous phase

Cement notation

[wt.%]

CAC Calcium aluminate Gehlenite Belite Ferrite Perovskite Spinel

CA C2AS C2S C3FT CT MgAl2O4

70 20 2 3 2 3

CSA Ye'elimite Belite Anhydrite + Gypsum Calcite Aluminates Brownmillerite Periclase

C4A3$ β-C2A C$ + C$H2 C C3A C4AF M

50 20 22 b3 b8 b3 b1

J. Bizzozero et al. / Cement and Concrete Research 56 (2014) 190–202

193

4.0

Table 3 CAC and CSA systems.

3.5

Sample name

wt.% CAC

wt.% Added C$H2

100 80 70 65 60 55 54 54 53.5 53 52 50 50 48 48 46 46 44 43.8 43

0 20 30 35 40 45 46 46 46.5 47 48 50 50 52 52 54 54 56 56.2 57

100%CA–0%G_ref 80%CA–20%G 70%CA–30%G 65%CA–35%G 60%CA–40%G 55%CA–45%G 54%CA–46%G 54%eqCA–46%G_ref 53.5%eqCA–46.5%G 53%eqCA–47%G 52%CA–48%G 50%CA–50%G 50%eqCA–50%G 48%CA–52%G 48%eqCA–52%G 46%CA–54%G 46%eqCA–54%G 44%eqCA–56%G 43.8%eqCA–56.2%G 43%eqCA–57%G

100.0 84.0 75.4 70.9 66.3 61.6 60.6

0.0 16.0 24.6 29.1 33.7 38.4 39.4

wt.% CSA

100.0 99.3 98.6 58.7 56.7

wt.% Added C$H2

3.0

45.2

52.8

47.2

2.0 1.5 1.0

0.0 0.7 1.4

0.5

41.3 43.3

54.8

46 % CA - 54 % G 48 % CA - 52 % G 50 % CA - 50 % G 52 % CA - 48 % G 54 % CA - 46 % G 55 % CA - 45 % G 60 % CA - 40 % G 65 % CA - 35 % G 70 % CA - 30 % G 80 % CA - 20 % G 100 % CA - 0 % G_ref

2.5 crack threshold

0.0 0 93.1

6.9

91.4

8.6

88.9 85.3 84.6 83.1

11.1 14.7 15.4 16.9

Polished resin impregnated slices were examined by backscattered electrons (BSE) and energy dispersive X-ray (EDS) analyses in a scanning electron microscope (SEM) (FEI Quanta 200) with an accelerating voltage of 15 kV. Mercury intrusion porosimetry (MIP) was carried out on Thermo Scientific Pascal 140 and 440 machines with a pressure capacity of 400 MPa. Five pieces of samples of about 4 × 5 × 5 mm3 were used for the analysis. The main limit of this technique is that it measures the volume of porosity that can be accessed from a given pore entry and not the real pore size [34]. Pore solutions were extracted from cylindrical cement paste specimens 50 mm diameter and 100 mm high. 350 g of cement paste were cast in plastic bottles and stored for 24 h in sealed conditions. Then the samples were cured for 6 days under a small amount of water (40 g of water for 350 g of cement paste). These storage conditions limit leaching, but ensure that pores are filled with pore solution. They also best represent the conditions under which the expansion takes place, where water can enter into the samples. The pore solution was extracted after 7 days of hydration at a pressure of 560 MPa, applied for 3 min to extract 5–10 ml of pore solution. The method has been adapted from the one used by [35,36]. The extracted pore solution was then filtered with a 0.45 μm filter, stored in a plastic vial at 5 °C to avoid carbonation and precipitation before analysis. The ionic composition of the solutions was measured with a Dionex ICS-3000 ion chromatograph for sulfate ion analysis and a Shimadzu ICPE 9000 spectrometer for cation analysis. The standard error of measurement is about 10%.

5

10

As this work was focussed on long-term dimensional stability the expansion was measured from 24 h. Of course dimensional changes will occur before this time [37,38]. Fig. 2 shows the expansion profiles as a function of time for the different CAC systems. Expansion increases with the gypsum content. Macro-cracks appear after 2–2.5% of expansion (Fig. 4a). There is a qualitative change in the expansion behaviour between the systems with a gypsum content lower than 40 mol% and those with a content above 45 mol%. The low expansion systems (solid curves) show an

25

30

increase in expansion at the beginning (swelling when the samples are introduced into water) followed by a plateau. For the high expansion systems (dotted curves), there is rapid initial expansion, which slows to a much lower rate after a few days. Fig. 3 shows the expansion as a function of time for CSA systems. CSA systems are more sensitive to expansion than the CAC systems. In CSA systems, the macro-cracks appear after ~ 0.5% of expansion, leading to fracture of the samples (Fig. 4b). The systems are very sensitive to small additions of gypsum at 56.2 mol% of gypsum and below there is a small expansion followed by a plateau, while for 57 mol% gypsum there is destructive expansion. Fig. 5 compares the expansion at 28 days as a function of gypsum content for the two systems. The expansion increases slowly and linearly up to around 45 mol% gypsum for CAC and 56 mol% gypsum for CSA systems, and then expands much more steeply. The sudden change in expansion behaviour observed suggests some kind of threshold effect. Below this threshold the expansion at 28 days increases gradually. Initial swelling is observed within a few days of immersion of the samples in water; after this the sample dimensions remain stable and the samples stay intact over a long period of time (at least 65 days). Above this threshold the samples continue to expand gradually after the first swelling and eventually fail.

4.0

0.5

3.5

Expansion [%]

3.1. Expansion

20

Fig. 2. Expansion of CAC systems.

3.0

3. Results

15

Time [days]

Expansion [%]

mol% C$H2

Expansion [%]

mol% CA or eqCA

0.4 0.3 0.2

2.5 CAC : crack threshold

0.1

2.0

0.0

0

2

4 6 Time [days]

8

1.5 1.0

4 days: sample is broken

43 % eqCA - 57 % G 43.8 % eqCA - 56.2 % G 44 % eqCA - 56 % G 46 % eqCA - 54 % G 48 % eqCA - 52 % G 50 % eqCA - 50 % G 53 % eqCA - 47 % G 53.5 % eqCA - 46.5 % G 54 % eqCA - 46 % G_ref

CSA : crack threshold

0.5 0.0 0

5

10

15

20

25

30

Time [days] Fig. 3. Expansion of CSA systems. The insert graph shows the low magnitude expansion to better discriminate the curves.

194

J. Bizzozero et al. / Cement and Concrete Research 56 (2014) 190–202

a)

b)

43%eqCA – 57%G_4d

CAC: Cracks after 30 days of curing in water

CSA: Deterioration after 4 days of curing in water

Fig. 4. Comparison of the deterioration between a CAC and a CSA system.

3.2. Study of low expansion and high expansion systems From the expansion measurements presented in Section 3.1, low expansion and high expansion blends were identified for both CAC and CSA systems. Low expansion ðLEÞ–CAC : 60%CA–40%G; CSA : 44%eqCA–56%G High expansion ðHEÞ–CAC : 55%CA–45%G; CSA : 43%eqCA–57%G The rate of heat evolution is shown in Fig. 6a for CAC systems and in Fig. 6b for CSA systems. There is no difference in kinetics between low expansion and high expansion systems for both CAC and CSA. The CSA systems hydrate much faster than the CAC systems, but the cumulative heat is comparable after 100 h of hydration (not shown in the figure). Fig. 7 shows the quantitative evolution of the phase assemblage from the XRD Rietveld analyses of slices of cement paste cured in water for the four systems. In none of these systems are the anhydrous phases (CA or ye'elimite and gypsum) completely consumed. Table 4 shows the values of ettringite determined by XRD and by TGA at 14 days. It is noted that: • For the low expansion systems, the expansion is stable after 3 days and the ettringite content increases slightly thereafter. • For the high expansion systems ettringite precipitation continues over 10 days but expansion still increases afterwards. The expansion

4.0

Expansion [%]

3.0 2.5

CAC: crack threshold

2.0 1.5 1.0

CSA: Critical sulfate threshold

3.5

CAC: Critical sulfate threshold

CAC systems - 28 days CSA systems - 28 days

CSA: crack threshold

0.5

X

0.0 0

10

20

30

40

50

Calcium sulfate [mol%] Fig. 5. Expansion as a function of gypsum content at 28 days.

60

occurs while there is no measurable increase in the amount of crystalline hydrates. • The absolute amount of ettringite is very similar in the two systems; slightly higher for the high expansive case in both the CAC and CSA systems. However these small differences are within the error of measurement. Fig. 8a and b show the derivative thermogravimetric (DTG) curves of the four studied systems at 1 and 14 days. The main peak position corresponding to the water loss of ettringite is around 140 °C, the peak of gypsum is around 150 °C and AH3 is observed at 270 °C. There is more aluminium hydroxide in CAC systems (Fig. 8a) and slightly higher ettringite content in the CSA systems (Fig. 8b). The amount of ettringite increases slightly between 1 and 14 days in both CAC and CSA systems. The amorphous AH3 increases between 1 and 14 days for CAC systems but seems to be stable in the CSA systems. At 1 day there is more ettringite in high expansion systems. At 14 days no substantial difference between low expansion and high expansion systems is observed in CAC systems, however in CSA systems a slightly less ettringite content is observed in high expansion system compared to the low expansion one. A rough estimation (Table 4) of the ettringite content was made by taking twice the integral of the left half of the peak (from 30 to 140 °C), as gypsum overlaps the ettringite peak in the range 140–170 °C. The AH3 content is calculated by integrating from 200 to 350 °C (crystalline gibbsite, AH3, has 3 mol of water per mole of phase, but the water content may be somewhat different for the poorly crystalline or amorphous aluminium hydroxide formed here). XRD and TGA results show similar trends, TGA overestimates slightly the ettringite content. Fig. 9 shows SEM-BSE images of the low expansion and high expansion CAC and CSA systems after 14 days of hydration. The observations are done on the basis of the analysis of several images per sample but here only one image per sample is shown. The cracks cannot be attributed only to the expansion, they are also a result of the drying during sample preparation. Ettringite is characterised by the small parallel cracks which form on drying. There is a clear difference in the distribution and morphology of the hydrates between CAC and CSA systems. AH3 (dark grey) and ettringite (denoted AFt, light grey) are more closely intermixed and finer in the CSA systems than in the CAC systems, this is probably a result of the faster hydration of the CSA systems. However within each system there is no obvious difference between the low expansion and high expansion systems. It can be noticed (Fig. 9b) that around the gypsum grains there is more AH3 than in the rest of the hydrated matrix. Fig. 10 shows the cumulative porosity curves as a function of the pore entry radius at 1,3 and 14 days. The critical pore radii at 1 day are considered as characteristic of the systems in the ensuing discussion, because this is the age of the samples when immersed in water for the expansion. The critical pore radius is taken at the intersection of the 2 tangents as shown in Fig. 10, i.e. where the mercury starts to intrude the porous matrix. We could also consider the value at the inflection

J. Bizzozero et al. / Cement and Concrete Research 56 (2014) 190–202

a) CAC systems

b) CSA systems 14

60 % CA - 40 % G 55 % CA - 45 % G

12

Normalized heat flow [W/moleq.CA]

Normalized heat flow [W/moleq.CA]

14

195

10

8

6

4

2

0

44 % eqCA - 56 % G 43 % eqCA - 57 % G

12

10

8

6

4

2

0 0

5

10

15

20

25

30

0

1

2

3

4

5

6

7

8

9

10

Time [h]

Time [h] Fig. 6. Isothermal calorimetry of CAC and CSA systems.

b) CSA – low expansion

a) CAC – low expansion 70

0.7

60

0.5 40 0.4 30 0.3 20 0.2 10 0 0

10

20

30

40

50

44%eqCA - 56%G

x

0.5 0.4 30 0.3 20

0.1

10

0.0

0

0.2 0.1 0.0 0

60

10

20

x

60

40

50

60

d) CSA – high expansion

CA Gypsum Ettringite Expansion

43%eqCA - 57%G

0.8

70

0.7

x

60

Ye'elimite Gypsum Ettringite Expansion

0.6 0.5

40 0.4 30 0.3 20

0.2

10 0 20

30

Time [days]

40

50

60

0.8 0.7 0.6

Phase content [wt%]

50

Expansion [%]

Phase content [wt%]

30

Time [days]

55%CA - 45%G

10

0.7

40

c) CAC – high expansion

0

0.8

50

Time [days]

70

Ye'elimite Gypsum Ettringite Expansion

0.6

0.6

50

Expansion [%]

Phase content [wt%]

0.8

Expansion [%]

x

60

CA Gypsum Ettringite Expansion

50 0.5 Break down

40

0.4 30 0.3 20

0.1

10

0.0

0

0.2 0.1 0.0 0

10

20

30

Time [days]

Fig. 7. XRD on slices cured in water for CAC and CSA systems (lines for eye guide only).

40

50

60

Expansion [%]

70

Phase content [wt%]

60%CA - 40%G

196

J. Bizzozero et al. / Cement and Concrete Research 56 (2014) 190–202

Table 4 Phase content by XRD Rietveld (Error of measurement ± 2%) and by TGA (Error of measurement ± 5%) at 14 days. [wt.%]

CAC systems

Ettringite, XRD Ettringite, TGA AH3, TGA

CSA systems

LE

HE

LE

HE

40.1 45 20

42.8 45 21

48.9 49 17

50.4 50 19

point as a characteristic value. Both values are given in Table 5. CAC systems have a slightly larger critical pore radius and a greater total porosity than CSA systems. Again there is no noticeable difference between the low expansion and high expansion systems. The small differences between the samples in the same system (CSA + calcium sulfate or CAC + calcium sulfate) are typical of the inter sample variation obtained on MIP. Fig. 11 shows the ionic concentration of the pore solutions as a function of the gypsum content after 7 days of hydration for both CAC and CSA systems. The extracted pore solution gives an average composition throughout the sample. Both systems show similar trends; the Ca2+ and SO2− concentrations increase with the gypsum content and 4 + + the Al(OH)− and consequently pH decrease. The pH is 4 , Na , K calculated by charge balance from experimental values of the ionic concentrations. For the CAC system the relative concentrations of the different ions change abruptly between the low expansive and the high expansive system (40 and 45 mol% of gypsum respectively). The concentration of calcium ions increases by almost an order of magnitude and the concentration of aluminate ions drops to very low values. It is difficult to understand the reason for this sudden change. Normally, pore solutions in cementitious systems are in quasi equilibrium with the hydrate phases present [39] and there is no change in the assemblage of solid phases present around the threshold of unstable expansion: solid CA or C4A3$ and gypsum are present in both cases. However, some gypsum remains in the low expansive samples and it seems to be no longer in thermodynamic equilibrium with the solution. This may be because the few remaining grains become isolated from the solution by hydrates, for example AH3, and hence cannot dissolve.

The equilibrium pore solution compositions (Fig. 12) and stable phase assemblage (Fig. 13) of the CA–C$H2 and the C4A3$–C$H2 systems were also calculated by minimisation of the Gibbs free energy with the GEMS-PSI software [40]. The thermodynamic data for aqueous species and other solids is from the GEMS-PSI thermodynamic database [41] and the CEMDATA database for the solubility products of cementitious phases [25]. A degree of hydration of 100% was assumed (i.e. all the CA or C4A3$ and gypsum are free to react completely) and a w/b ratio of 0.65 used to have an excess of water allowing full hydration. Fig. 12a and 12b show the dissolved ions in the CA–C$H2 and C4A3$– C$H2 systems, respectively. For CA-gypsum systems there is a small decrease of Ca2 + and Al(OH)− 4 at 25 mol% of calcium sulfate corresponding to the point where C3AH6 is depleted and ettringite starts to be thermodynamically stable (Fig. 13a). But more importantly there is a threshold at 50 mol% of calcium sulfate, corresponding to the point at which solid gypsum starts to remain in excess, where the Ca2+ and SO2− ions increase suddenly and Al(OH)− 4 4 and pH decrease. The pH values are lower than the ones presented in Fig. 11 but the same trends between experimental measurements and modelling are observed. The same observations are done for the ye'elimite-gypsum systems. The absolute amounts are different to those observed experimentally, due to the absence of alkalis in the thermodynamic model, but the change is similar to that seen in the experimental measurements of the pore solution between 40 and 45 mol% of calcium sulfate. Although this threshold value is somewhat different to that seen experimentally, we know that the hydration is not complete in the experimental systems. Fig. 13b shows similar phase assemblage for ye'elimite-gypsum systems. The plain CSA without gypsum addition (CSA_ref) is indicated by a vertical line at 46 mol% of calcium sulfate. In this system there is a similar point at which solid gypsum remains in excess and a corresponding change in the pore solution composition.

4. Discussion There is a difference in the microstructure of hydrated CAC-gypsum and CSA-gypsum. In the latter, ettringite and amorphous aluminium hydroxide are more intimately intermixed. These differences could explain the greater sensitivity to expansion of the CSA system. However within each system the calorimetry, XRD, SEM, TGA and MIP results do

a) CAC systems

b) CSA systems 0.0

0.0

AH3 gypsum

-2.0x10-3

-3.0x10-3

-4.0x10-3

ettringite 100

200

300

400

Temperature [°C]

500

-3.0x10-3

44 % eqCA - 56 % G_1d 44 % eqCA - 56 % G_14d 43 % eqCA - 57 % G_1d 43 % eqCA - 57 % G_14d

ettringite -5.0x10-3

-5.0x10

0

gypsum -2.0x10-3

-4.0x10-3

60 % CA - 40 % G_1d 60 % CA - 40 % G_14d 55 % CA - 45 % G_1d 55 % CA - 45 % G_14d

-3

AH3

-1.0x10-3

Diff. rel. weight [1/°C]

Diff. rel. weight [1/°C]

-1.0x10-3

600

0

100

200

300

400

Temperature [°C]

Fig. 8. DTG curves at 1 and 14 days for low expansion (solid lines) and high expansion systems (dashed lines).

500

600

J. Bizzozero et al. / Cement and Concrete Research 56 (2014) 190–202

a) CAC – low expansion

197

b) CSA – low expansion

AFt

AH3 60%CA – 40%G 14d

c) CAC – high expansion

d) CSA – high expansion

CAC grain

55%CA – 45%G 14d Fig. 9. BSE images ×5000 (AFt: ettringite, AH3 : amorphous aluminium hydroxide, G: gypsum).

not indicate any clear difference between low expansive and high expansive systems. If we consider that the underlying cause of expansion lies in the crystallization pressure of ettringite, we need to consider: 1. The supersaturation with respect to ettringite. 2. The confinement of ettringite, the size of pores in which it is growing. 3. The mechanical links between the local effects of the growing ettringite and the macroscopic level of the prisms. 4.1. Supersaturation with respect to ettringite The ettringite saturation index was calculated from the activities of the different species, this is plotted as a function of the gypsum content in Fig. 14. (Activities were calculated from the experimentally determined concentrations with the extended Debye–Huckel equation implemented in PHREEQC [42] using the cement-specific CEMDATA database from EMPA [25]). On the right axis of Fig. 14 the maximum crystallization pressure (Pc) is shown calculated from Eq. (9). Although the supersaturations (and so maximum crystallization pressures) corresponding to the high expansion systems are higher than those for the low expansive systems, there is no obvious threshold and the values for the low expansive CSA-gypsum system are very similar to the values for the high expansive CAC-gypsum system. 4.2. Confinement of ettringite crystals In Portland cement systems undergoing external sulfate attack [43] or internal sulfate attack due to high temperature curing (so-called

delayed ettringite formation) [17,20], it has been shown that expansion occurs only when ettringite forms within C-S-H gel. Regions of C-S-H gel contain only small pores around 10 nm or less, which can easily give the constraint needed for the ettringite crystals to exert expansive pressure. In contrast there are no clearly distinguishable microstructural environments identifiable in the case of the systems investigated here where the hydrated matrix is composed of both ettringite and amorphous aluminium hydroxide. Fig. 15 shows the crystallization pressure (ΔP) as a function of the pore radius for the different systems calculated according to Eq. (13) (assuming spherical pores, an interfacial free energy γCL = 0.1 J/m2 [17] and the thickness of the film of solution between the crystal and the pore δ = 1.5 nm). This shows there is a minimum pore radius where ΔP = 0, which corresponds to the size of pore for which the supersaturated solution is at equilibrium, so ettringite crystals cannot grow into pores of smaller size. Vertical lines indicate the critical pore radius of CAC and CSA systems measured by MIP on 1 day samples. The range of pores between the minimum pore radius and the critical pore radius gives the amount of percolated porosity in which ettringite can exert crystallization pressure. It can be noted that MIP in fact measures pore entry radii rather than pore sizes, but this is exactly the same situation as we are concerned with in crystal growth, where it is the entry to a pore which controls whether a crystal can grow into it as shown in [28]. Both low expansive and high expansive systems have a minimum pore radius smaller than the critical pore radius, which means that ettringite can grow through the pore network percolating the microstructure. Only the CAC system with 20 mol% of gypsum has a minimum pore radius above the critical pore radius, indicating that ettringite can grow in a few large isolated pores.

J. Bizzozero et al. / Cement and Concrete Research 56 (2014) 190–202

a) CAC – low expansion 25

10

25

1 day 3 days 14 days

20

Porosity [%]

15

b) CSA – low expansion

Critical pore radius (8nm)

Porosity [%]

20

60%CA - 40%G

5

0 1E-3

15

10

44%eqCA - 56%G

5

0.01

0.1

1

10

0 1E-3

100

0.01

0.1

Pore radius [μm]

100

20

15

10

43%eqCA - 57%G

1 day 3 days 14 days

Critical pore radius (6nm)

1 day 3 days 14 days

5

0 1E-3

25

Porosity [%]

10

10

d) CSA – high expansion

Critical pore radius (9nm)

Porosity [%]

15

55%CA - 45%G

1

Pore radius [μm]

c) CAC – high expansion 25

20

1 day 3 days 14 days

Critical pore radius (6nm)

198

5

0.01

0.1

1

10

100

0 1E-3

0.01

0.1

1

10

100

Pore radius [μm]

Pore radius [μm]

Fig. 10. MIP cumulative porosity at 1, 3 and 14 days of hydration.

The main difference between low and high expansive systems is that in the later cases there is a larger amount of percolated porosity where ettringite grows exerting crystallization pressure as shown in Table 6. These values are only approximate as the exact values of the interfacial energies are not known and do not account for anisotropy of the ettringite crystals.

growth can generate pressure (“confined ettringite”) is relevant in these cases [17]. The average hydrostatic tensile stress generated is given by Eq. (14) [17]. σ ¼ P  g ðφÞ

ð14Þ

4.3. Mechanical effects

Where P is the crystallization pressure and φ is the volume fraction of confined ettringite. g(φ) is a geometric factor that depends on the shape of the pore. For a cylindrical pore:

As discussed by Scherer [15] a crystal growing in a single pore will not produce expansion. The crystallization pressure should be developed over regions many μm in size for overall expansion of the sample to be observed. The fraction of percolated porosity in which ettringite

  2 φ cyl g φcyl ¼ 3 1‐φcyl

ð15Þ

And for a filled large spherical pore with small entries:

Table 5 Pore size and total porosity measured by MIP on 1 day samples.

Total porosity [% ± 2%] Pore size at inflection point [nm ± 0.5] Critical pore size (tangent intersection) [nm ± 0.5]

!

CAC systems

CSA systems

LE

HE

LE

HE

22.4 5.2 8

20.1 5.5 9

14.5 4.1 6

17.2 4.1 6

  g φsph ¼

φsph 1−φsph

! ð16Þ

These equations are based on spherical and cylindrical pores and so only approximate the situation in cement paste where the pores are irregular in shape and ettringite is highly anisotropic. Thus, Eq. (14)

J. Bizzozero et al. / Cement and Concrete Research 56 (2014) 190–202

CSA systems

CAC systems

14

12 800

Ca2+ Al(OH)-4

600

8

SO24 Na+ K+ pH

400

6

4

Ca2+ concentration [mmol/L] and pH

Concentration [mmol/L]

10

200 2

0 0 15

20

25

30

35

40

45

55

56

57

58

Calcium sulfate [mol%] Fig. 11. Ion concentrations as a function of gypsum content at 7 days. Right y-axis: Ca2+ concentration and pH, left y-axis: all other ions concentrations.

can be used to estimate the amount of confined ettringite needed to cause cracking. It is reasonable to take as average hydrostatic tensile stress the tensile strength of the cement paste estimated as 50% of the measured flexural strength [44], and as crystallization pressure the averaged pressure

over the pore size range (ΔPmean) between the critical pore radius (from MIP) and the minimum pore radius for growth of ettringite. All the values for the calculation and the resulting amounts of confined ettringite needed to cause cracking are shown in Table 6. The results of these calculations are compared to the amount of percolated porosity in which ettringite can grow, discussed in Section 4.2 in Fig. 15. Fig. 16 shows that the range of volume fraction of confined ettringite needed for cracking (φsph: lower limit and φcyl: upper limit) is much higher than the amount of percolated porosity in which ettringite growth can occur in the case of both low expansive CAC and CSA systems. However, in both high expansion systems the volume of confined ettringite needed for cracking is very similar to the volume of percolated porosity in which ettringite can grow. These calculations are only approximate and contain many estimations. In addition to the assumptions of the surface energy for ettringite and the shape of the pores already discussed, it should also be mentioned that the flexural strength is obtained in a conventional test taking a few minutes as opposed to the expansion with takes place over a few days, which would allow deformation at lower stress due to creep effects. Despite these approximations, the analysis indicates that as the supersaturation rises, the volume of percolated porosity in which ettringite can grow increases. This specific volume increases until a threshold is reached above which the confined volume of ettringite generating crystallization pressure is sufficient to cause cracking and unstable expansion. A similar argument emerges from the analysis of ettringite formation after elevated temperature curing (DEF) by Chanvillard and Barbarulo [45]. They used poromechanics [46–48] to analyse a poroelastic skeleton (porous matrix) subject to deformations due to crystallization pressure. The cement paste or concrete is brittle so it cannot store elastic energy beyond a critical threshold [45]. When this threshold is reached, there is irreversible damage (expansion) which releases the stored elastic energy. This threshold is reached when the strain generates stresses above the critical tensile strength of the porous material. The differences in behaviour between low and high expansive systems can be attributed to reaching a critical threshold, with the onset of extensive cracking. Fig. 5 shows two domains, low expansive below the critical sulfate threshold and high expansive above this threshold. For low gypsum contents, the expansion increases gradually as the supersaturation of ettringite, the crystallization pressure and the volume of porosity containing confined ettringite increase. At around

a) CA-C$H2, DH=100%, w/b=0.65

b) C4A3$-C$H2, DH=100%, w/b=0.65 1.0

18

199

1.0

18

CSA_ref

12 0.6

10 8

0.4 6

Ca2+ Al(OH)-4 SO24

4 2

0.2

0.8

14 12

0.6 10 8 0.4 6

Ca2+ Al(OH)-4 SO24

4 2

pH

pH

0

0.0 0

10

20

0.2

Al(OH)-4 concentration [mmol/L]

14

Concentration [mmol/L]

16 0.8

Al(OH)-4 concentration [mmol/L]

Concentration [mmol/L]

16

30

40

50

60

70

Calcium sulfate [mol%]

80

90 100

0

0.0 0

10

20

30

40

50

60

70

80

90 100

Calcium sulfate [mol%]

Fig. 12. GEMS modelling of thermodynamically stable aqueous solution at 20 °C (for b. the calcium sulfate amount is relative to eq. CA). The right y-axis shows the Al(OH)− 4 concentrations.

200

J. Bizzozero et al. / Cement and Concrete Research 56 (2014) 190–202

a) CA-C$H2, DH=100%, w/b=0.65

b) C4A3$-C$H2, DH=100%, w/b=0.65

100

100

90

90

60

C3AH6

50

sul fa

te

40

C$H2

no

30

ettringite

mo

20 10

70

AH3

60

te

AH3

sul fa

70

water

80

50

C$H2

no

water

40

mo

80

Phase content [cm3/100g solid]

Phase content [cm3/100g solid]

CSA_ref

30

ettringite

20 10 0

0 0

10

20

30

40

50

60

70

80

90

0

100

10

20

30

40

50

60

70

80

90

100

Calcium sulfate [mol%]

Calcium sulfate [mol%]

Fig. 13. GEMS modelling of thermodynamically stable hydrate assemblages at 20 °C (for b. the calcium sulfate amount is relative to eq. CA).

5. Summary and conclusions Systems of calcium aluminate cement and calcium sulfo aluminate cement, both with added gypsum, show unstable expansion when gypsum is added above a given threshold (around 45 mol% of gypsum for CAC and around 55 mol% for CSA systems). Below this threshold there is a gradual increase in the amount of expansion at 28 days. Within this period, the expansion occurs as a fairly rapid swelling within a few days of immersion of the samples in water, after which the sample dimensions remain stable and the samples stay intact over long periods

of time (at least 65 days). Above this threshold the samples continue to expand gradually after the initial swelling and eventually fail. There is a marked difference in the strain to failure between the two systems (about 2.3% for the CAC-gypsum systems and 0.5% for the CSAgypsum systems). The measured expansion represents high strains for a brittle material like cement paste. As the deformations and therefore the stresses develop slowly, there will be a certain amount of creep helping to dissipate the mechanical energy generated by the expansive phase. In both cases ettringite and poorly crystalline aluminium hydroxide are the only hydrates formed. The intermixing of these different hydrates was much finer in the CSA-gypsum samples. A detailed microstructural investigation was made of systems above and below the

80 70

8

60 50

6

40

Low expansion 4

30 20

2 10

CAC: HE CSA: LE

70 60 50

CAC: LE

40 30 CAC: 20 mol%

20 10

0 Minimum pore radius Critical pore radius, MIP 1 day

-10 -20 1

0

0 15

20

25

30

35

40

45

50

55

60

Calcium sulfate [mol%] Fig. 14. Ettringite saturation index as a function of gypsum content at 7 days.

CSA: HE

80

ΔP [MPa]

High expansion

10

Ettringite saturation index [-]

90 90

Max crystallization pressure [MPa]

12

100

100

CAC+Gypsum CSA+Gypsum

CSA+G

110

CAC+G

45 mol% of gypsum (in the CAC systems) and 55 mol% (in the CSA systems) the slope of the curve becomes considerably steeper indicating cracking above the critical threshold of elastic deformation.

10

100

1000

Pore radius, rp=rc +δ [nm] Fig. 15. Crystallization pressure as a function of the pore radius. Based on saturation indexes at 7 days of hydration, assuming δ = 1.5 nm. CAC systems: plain lines, CSA systems: dotted lines.

J. Bizzozero et al. / Cement and Concrete Research 56 (2014) 190–202 Table 6 For each system are presented the critical pore radius (rcrit) measured with MIP, the minimum pore radius (rmin) calculated from pore solution analyses and the volume fraction of percolated porosity in which ettringite growth gives expansion pressure (φep). The values necessary for the calculation of the volume fraction of confined ettringite in cylindrical (φcyl) and spherical (φsph) pores (Eq. (14)) are the tensile strength (σt) and the crystallization pressure averaged over the pore size range ΔPmean. CAC systems

rcrit [nm] rmin [nm] φep [vol.%] σt [MPa] ΔPmean [MPa] φcyl [vol.%] φsph [vol.%]

CSA systems

LE

HE

LE

HE

8 5.9 4.0% 3.9 ± 0.2 7.1 45% 36%

9 4.0 12.0%

6 4.2 5.5% 2.5 ± 0.4 15.0 20% 14%

6 3.6 9.0%

30.8 16% 11%

26.8 12% 9%

threshold for unstable expansion. These studies showed no obvious difference between the systems: • Comparable total amount of ettringite. • Comparable pore size distributions. • Comparable microstructures in the SEM. On the other hand there was a distinct change in the pore solution compositions at the threshold, with a drop in the concentration of aluminate ions, a strong increase in the concentration of calcium ions

Confined ettringite range

201

and small increase of sulfate ions. The solid phases present are the same, both gypsum and anhydrous phases (monocalcium aluminate or ye'elimite) are always in excess. However thermodynamic modelling indicated that by increasing the gypsum content there is a sudden change when the gypsum is in excess. This indicates that even if solid gypsum is present in the low expansive system, it remains isolated from the pore solution by the hydrates. Despite the strong changes in the composition of the pore solutions, the supersaturation with respect to ettringite increases fairly uniformly with the gypsum content. Therefore we conclude that the onset on unstable expansion is a threshold effect, which can be explained by the combination of 2 mechanisms: • The supersaturation of ettringite determines the range of pore sizes where it can precipitate and the magnitude of pressure. It is related to the ability of ettringite growth to penetrate an interconnected pore network – below the critical sulfate threshold ettringite can only exert pressure in a small number of isolated pores, above the critical sulfate threshold ettringite can form in a bigger fraction of porosity leading to a higher total volume of ettringite generating pressure and subsequent onset of unstable expansion. • The formation of confined ettringite exerts pressure on the system. In the unstable high expansion systems, this pressure exceeds the elastic limit of the system and cracking leads to unstable expansion and failure. The macroscopic expansion is thus strongly related to both the supersaturation of ettringite and the distribution and confinement of ettringite crystals. Acknowledgements

Measured percolated porosity The authors would like to thank Kerneos France for the financial support.

50

References 40

[vol%]

30

20

10

0 CAC

CSA

Low expansion

CSA

CAC

High expansion

Fig. 16. Volume of measured percolated porosity and confined ettringite range needed to cause cracking.

[1] E. Gartner, Industrially interesting approaches to “low-CO2” cements, Cem. Concr. Res. 34 (2004) 1489–1498. [2] J.S. Darntoft, J. Lukasik, D. Herfort, D. Sorrentino, E.M. Gartner, Sustainable development and climate change initiatives, Cem. Concr. Res. 38 (2008) 115–127. [3] P. Lura, O.M. Jensen, K. van Breugel, Autogenous shrinkage in high-performance cement paste: an evaluation of basic mechanisms, Cem. Concr. Res. 33 (2003) 223–232. [4] Y. Fu, P. Gu, P. Xie, J. Beaudoin, Effect of chemical admixtures on the expansion of shrinkage-compensating cement containing a pre-hydrated high alumina cement-based expansive additive, Cem. Concr. Res. 25 (1995) 29–38. [5] J.P. Bayoux, et al., Study of the hydration properties of aluminous cement and calcium sulfate mixes, in: R.J. Mangabhai (Ed.), Calcium Aluminate Cements, E & FN Spon, London, 1990, pp. 320–334. [6] F. Winnefeld, B. Lothenbach, Hydration of calcium sulfoaluminate cements —Experimental findings and thermodynamic modelling, Cem. Concr. Res. 40 (2010) 1239–1247. [7] K.L. Scrivener, Capmas, A., Chapter 13: Calcium Aluminate Cements, in: P.C. Hewlett (Ed.), LEA's Chemistry of Cement and Concrete, Arnold, London, 1998, pp. 709–778. [8] F.P. Glasser, L. Zhang, High-performance cement matrices based on calcium sulfoaluminate–belite compositions, Cem. Concr. Res. 31 (2001) 1881–1886. [9] T. Desbois, R. Le Roy, A. Pavoine, G. Platret, A. Feraille, A. Alaoui, Effect of gypsum content on sulfoaluminate mortars stability, Eur. J. Environ. Civ. Eng. 14 (2010) 579–597. [10] I.A. Chen, C.W. Hargis, M.C.G. Juenger, Understanding expansion in calcium sulfoaluminate–belite cements, Cem. Concr. Res. 42 (2012) 51–60. [11] S. Lamberet, Durability of ternary binders based on portland cement, calcium aluminate cement and calcium sulfate, (PhD thesis) Ecole Polytechnique Fédérale de Lausanne, 2005. (N°3151). [12] M.D. Cohen, Modeling of expansive cements, Cem. Concr. Res. 13 (1983) 519–528. [13] M.D. Cohen, Theories of expansion in sulfoaluminate-type expansive cements — Schools of thought, Cem. Concr. Res. 13 (1983) 809–818. [14] P.K. Mehta, Mechanism of expansion associated with ettringite formation, Cem. Concr. Res. 3 (1973) 1–6. [15] G.W. Scherer, Crystallization in pores, Cem. Concr. Res. 29 (1999) 1347–1358. [16] G.W. Scherer, Stress from crystallization of salt, Cem. Concr. Res. 34 (2004) 1613–1624. [17] R.J. Flatt, G.W. Scherer, Thermodynamics of crystallization stresses in DEF, Cem. Concr. Res. 38 (2008) 325–336.

202

J. Bizzozero et al. / Cement and Concrete Research 56 (2014) 190–202

[18] P.W. Brown, H.F.W. Taylor, The role of ettringite in external sulfate attack, Am. Ceram. Soc. (1999) 73–98. [19] M. Polivka, Factors Influencing Expansion of Expansive Cement Concretes, in ACI SP-38, 1973. 239–250. [20] H.F.W. Taylor, C. Famy, K.L. Scrivener, Delayed ettringite formation, Cem. Concr. Res. 31 (2001) 683–693. [21] Y. Yamazaki, T. Monji, K. Sugiura, Early age expanding behavior of mortars and concretes using expansive additives of CaO-CaSO4-4CaO•3Al2O3•SO3 system, The VI International Congress on the Chemistry of Cement, Moscow, 1974. [22] K. Ogawa, D.M. Roy, C4a3s hydration, ettringite formation, and its expansion mechanism.3. Effect of Cao, Naoh and Nacl —Conclusions, Cem. Concr. Res. 12 (1982) 247–256. [23] J. Lavalle, “Recherches sur la formation lente des cristaux à la température ordinaire” (Research on the slow formation of crystals at ordinary temperature), C. R. Acad. Sci. 36 (1853) 493–495. [24] S. Taber, The growth of crystals under external pressure, Am. J. Sci. 41 (1916) 532–556. [25] Available at. Version cemdata07.2, release date 14.08.2008, http://www.empa.ch/ cemdata/(retrieved 20.07.2013). [26] M. Steiger, Crystal growth in porous materials - I: the crystallization pressure of large crystals, J. Cryst. Growth 282 (2005) 455–469. [27] M. Balonis, F.P. Glasser, The density of cement phases, Cem. Concr. Res. 39 (2009) 733–739. [28] M. Steiger, Crystal growth in porous materials - II: influence of crystal size on the crystallization pressure, J. Cryst. Growth 282 (2005) 470–481. [29] D. Min, T. Mingshu, Formation and expansion of ettringite crystals, Cem. Concr. Res. 24 (1994) 119–126. [30] H. Freundlich, Colloid & Capillary Chemistry, Methuen, London, 1926. 154–157. [31] D. Everett, The thermodynamics of frost damage to porous solids, Trans. Faraday Soc. 57 (1961) 1541–1551. [32] Q. Zhou, F.P. Glasser, Thermal stability and decomposition mechanisms of ettringite at b120 degrees C, Cem. Concr. Res. 31 (2001) 1333–1339. [33] L. Zhang, G.W. Scherer, Comparison of methods for arresting hydration of cement, Cem. Concr. Res. 41 (2011) 1024–1036.

[34] A.B. Abell, K.L. Willis, D.A. Lange, Mercury intrusion porosimetry and image analysis of cement-based materials, J. Colloid Interf. Sci. 211 (1999) 39–44. [35] P. Longuet, L. Burglen, A. Zelwer, La phase liquide du ciment hydratatée (The liquid phase of hydrated cement), Rev. Matér Constr. Trav. Publics Cim. Bétons 676 (1973) 35–41. [36] R.S. Barneyback, S. Diamond, Expression and analysis of pore fluids from hardened cement pastes and mortars, Cem. Concr. Res. 11 (1981) 279–285. [37] W. Nocun-Wczelik, Z. Konik, A. Stok, Blended systems with calcium aluminate and calcium sulphate expansive additives, Constr. Build. Mater. 25 (2011) 939–943. [38] C. Evju, S. Hansen, Expansive properties of ettringite in a mixture of calcium aluminate cement, Portland cement and beta-calcium sulfate hemihydrate, Cem. Concr. Res. 31 (2001) 257–261. [39] B. Lothenbach, Thermodynamic equilibrium calculations in cementitious systems, Mater. Struct. 43 (2010) 1413–1433. [40] Available at http://gems.web.psi.ch/(retrieved 20.07.2013). [41] W. Hummel, U. Berner, E. Curti, F.J. Pearson, T. Thoenen, Nagra/PSI Chemical Thermodynamic Data Base 01/01, Universal Publishers, Parkland, Florida USA, 2002. [42] Available at http://wwwbrr.cr.usgs.gov/projects/GWC_coupled/phreeqc/(retrieved 20.07.2013). [43] C. Yu, K.L. Scrivener, Mechanism of expansion of mortars immersed in sodium sulphate solution, Cem. Concr. Res. 43 (2013) 105–111. [44] S. Popovics, Strength and Related Properties of Concrete: A Quantitative Approach, 1998. 125. [45] G. Chanvillard, R. Barbarulo, Stress from confined crystallization occurrence by internal phase change: application to the case of delayed ettringite formation in hardened cement paste, Proceedings of Mechanics and Physics of Porous Solids — A Tribute to Pr Olivier Coussy, Ecole des Ponts ParisTech, 2011, pp. 285–302. [46] O. Coussy, Deformation and stress from in-pore drying-induced crystallization of salt, J. Mech. Phys. Solids 54 (2006) 1517–1547. [47] O. Coussy, Mechanics and Physics of Porous Solids, Wiley, New York, 2010. [48] Z.H. Sun, G.W. Scherer, Effect of air voids on salt scaling and internal freezing, Cem. Concr. Res. 40 (2010) 260–270.