- Email: [email protected]
The interrelationship between surface chemistry and rheology in alkali activated slag paste
Construction and Building Materials 65 (2014) 583–591
Contents lists available at ScienceDirect
Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat
The interrelationship between surface chemistry and rheology in alkali activated slag paste Alireza Kashani a, John L. Provis a,b,⇑, Greg G. Qiao a, Jannie S.J. van Deventer a,c a
Department of Chemical and Biomolecular Engineering, University of Melbourne, Victoria 3010, Australia Department of Materials Science and Engineering, University of Sheffield, Sheffield S1 3JD, United Kingdom c Zeobond Pty. Ltd., P.O. Box 23450, Docklands, Victoria 8012, Australia b
h i g h l i g h t s
g r a p h i c a l a b s t r a c t
Surface chemistry controls rheology
in alkali-activated slag cements. Particle–particle interactions and
gelation influenced by activator nature. Yield stress depends on square of zeta potential. Silicate and hydroxide activators differ notably in behavior. Influence of pH on fresh paste chemistry is indirect.
a r t i c l e
i n f o
Article history: Received 3 February 2014 Received in revised form 10 April 2014 Accepted 26 April 2014
Keywords: Alkali activated slag Rheology Yield stress Isothermal calorimetry Zeta potential pH
a b s t r a c t Ground granulated blast furnace slag can react with an alkaline activating solution to form a cement-like binder based on a calcium–sodium aluminosilicate gel, which is a potential alternative to Portland cement in many applications. This study provides new information regarding the effect of activator type and dosage on rheology by monitoring changes in pH, particle surface charge (zeta potential), and heat evolution in the early stages of the reaction process. Sodium and potassium hydroxide silicate solutions, at two different M2O (M: Na, K) dosages, are used here as activators. Alkali hydroxide activators cause a significant increase in the yield stress of an activated slag paste, especially at higher dosages as reactions take place rapidly, while within the same timeframe, the yield stress of the silicate activated slag remains unchanged. The results imply a direct relationship between a higher reaction rate with the formation of solid products (causing both spatial blockage effects and consumption of free water), and a rapid yield stress increase. However, the dependence of reaction rate on pH for different alkali-activated pastes is, at most, indirect. All activators induce a highly alkaline pH and a concentrated electrolyte solution environment in the fluid paste. As a result of complexation of poorly-hydrated ions on the surfaces of the particles, the magnitude of the zeta potential increases. A direct relationship is observed between the dosage of the activators and zeta potential. A zeta potential further from neutrality generally reduces yield stress by increasing the magnitude of double layer repulsive forces, with the exception of a higher dosage of silicate activator, which shows an indication of some attractive double layer forces. Ó 2014 Elsevier Ltd. All rights reserved.
⇑ Corresponding author at: Department of Materials Science and Engineering, University of Sheffield, Sheffield S1 3JD, United Kingdom. Tel.: +44 114 222 5490. E-mail address: j.provis@sheffield.ac.uk (J.L. Provis). http://dx.doi.org/10.1016/j.conbuildmat.2014.04.127 0950-0618/Ó 2014 Elsevier Ltd. All rights reserved.
584
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591
1. Introduction Ground granulated blast furnace slag (GGBFS) is the by-product of reduction of iron oxides to metallic iron in a blast furnace, and when quenched forms a glassy calcium–magnesium aluminosilicate which granulates and is ground to form a fine powder. GGBFS is often combined with Portland cement as a supplementary cementitious material to reduce cost and attributed CO2 emissions and to enhance mechanical and durability properties [1]. However, slag can also react as a sole precursor in an alkaline environment to form a cement-like binder [2], which is predominantly composed of a calcium–sodium aluminosilicate gel [3]. The commonly used alkali activators include alkali metal hydroxides, silicates and carbonates [2]. Among these activators, sodium silicate provides benefits regarding the ultimate strength of the alkali activated slag (AAS) binder, which can be similar to that of a good quality Portland cement depending on the activator modulus (molar ratio SiO2/Na2O) and dosage [4]; however, its setting rate is not as fast as when alkali hydroxides are used [5,6]. A key aim of many studies discussing and characterizing reaction products in AAS [7–10] has been to clarify the reaction mechanisms which are involved in the setting and hardening process. Upon contact with an alkaline solution, slag undergoes partial dissolution [11], with an exothermic reaction process [12–14] resulting in hardening if the water/solids ratio is sufficiently low. Isothermal calorimetry is used to monitor heat flow changes during alkali-activation reactions, and shows distinguishable differences as a function of alkali activator type and dosage, both in terms of the magnitude and duration of heat release [14–17]. It is also used as a tool to monitor reaction rate based on normalized heat flow curve versus time for Portland cement and blended cements [18,19]. The fluidity of an AAS paste can be monitored through changes in rheological properties such as yield stress and apparent viscosity. A low yield stress paste is generally considered desirable, as it reduces the energy required for pumping a concrete mix, and enhances the ability to fill voids when placing concrete. In addition, concretes are usually premixed for a specific time before placement, so it is also important to avoid much yield stress increase during the early age reaction. A rapid early reaction leading to thickening and solidification of a cement paste affects the workability and fluidity of the binder in transportation and placement. This is controlled in Portland cements by the use of organic plasticizing admixtures [20,21], which has enabled the development of concretes which are workable (or pumpable), and yet show high early strength. These admixtures are designed to control surface interactions, and so can reduce the yield stress of the mix, reduce the likelihood of a sudden yield stress increase during the reaction process, and control the rate of thickening [22]. However, chemical admixtures are not effective in all different environments of cementitious materials, and need to be tailor designed case by case. For example, the typical plasticizers used in cement do not plasticize as effectively in alkali activated binders [23,24]. The development of admixtures for alkali-activated systems is still ongoing [25,26], and requires a much more detailed understanding of the surface chemistry of slag particles within the complex reacting environment of a fresh AAS paste. The work presented in this paper is targeted toward building this understanding. Compared to the chemistry of a Portland cement, the addition of an alkali activator provides a higher pH and different electrolyte environment in the fresh binder paste, leading to differences in surface charges of particles and thus also in the electrostatic interactions between particles. However, the surface chemistry of alkali
activated binders, and particularly its interrelationship with rheology, is not well understood. This paper provides new insight into the effects of activator type and dosage on the rheological properties of an alkali activated slag paste. Yield stress is measured and related to the reactions taking place in this complex particle–fluid system. Zeta potential measurement is used to monitor the changes in electric double layer forces under different pH and electrolyte environments, which determine the sites and signs of the charges on particle surfaces [27]. Measurement of electric double layer forces by means of zeta potential provides insight into key parameters which determine yield stress [28]. The study results in an important new understanding of the relationship between the yield stress and the zeta potential of the activated slag paste. 2. Materials and methods 2.1. Materials Ground granulated blast furnace slag, with chemical composition as displayed in Table 1, and anhydrous sodium metasilicate powder (Na2O 50.9 wt.% and SiO2 49.1 wt.%) were provided by Zeobond Pty. Ltd., Australia. AR-grade NaOH and KOH (Sigma–Aldrich, Australia) were used. All activators were dissolved in MilliQ-grade purified water before mixing with slag to avoid any interference due to solid activator particle dissolution in the heat flow measurements, or the influence of additional particles on the paste solids fraction (and thus rheology). Dosages of 2.25 104 and 4.5 104 moles M2O per gram of slag were used, and specified on a molar basis to give the same amount of alkali ions for different activators at each dosage (for instance, 2.25 104 mole M2O per gram slag is equivalent to 1.80 g NaOH, 2.52 g KOH, and 2.75 g Na2SiO3 per 100 g slag). The samples were formulated by mixing slag with the activator solution at a water to slag weight ratio of 0.45. One sample was formulated without an activator but with the same water to binder ratio as a reference. After hand mixing to combine the components, the paste was mechanically mixed for 2 min or 10 min at 500 rpm before calorimetry and rheology measurements, respectively. 2.2. Rheological tests The rheological behavior of each AAS sample was measured using a Haake VT550 rheometer with a four-blade vane with a length of 50.1 mm and diameter of 15.3 mm, in a sample volume sufficient to simulate an infinite medium [29]. A direct yield stress measurement at a low rotational rate (0.2 rpm) was used because of the accuracy and reproducibility of the measured data [30], and the yield stress was calculated from the applied torque based on the geometry of the vane and the rotation rate according to [30]. Before each measurement, the paste was mixed by hand to avoid settling of particles and to provide a homogenous paste, left to rest for 30 s to dissipate residual stresses induced by mixing, before the 0.2 rpm rotational rate was applied. The first yield stress measurement was taken 12 min after combination of the slag and the activator (this includes 10 min of mixing and 2 min for the residual stress relaxation after high shear mixing). Five consecutive measurements were then taken at intervals of 2 min. The apparent viscosity at a range of higher shear strains was also measured, immediately after yield stress measurement, for each sample in 15 consecutive steps (Fig. 1) and without mixing between measurements.
Table 1 Chemical composition of the slag used, as determined by XRF. LOI is loss on ignition at 1000 °C and d90 was determined using a Malvern Mastersizer instrument. Oxide component
wt.%
SiO2 TiO2 Al2O3 Fe2O3 MnO MgO CaO Na2O K2O SO3 LOI d90 (lm)
33.1 0.6 15.0 0.6 0.3 5.9 41.8 0 0.3 2.3 0.4 39.8
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591
585
Fig. 1. Shear strains applied in 15 consecutive steps to the pastes for the measurement of apparent viscosity at higher shear strain (each step takes 20 s).
2.3. Isothermal calorimetry The activated slag paste was mechanically mixed for 2 min at 500 rpm, then poured into glass vials (about 20 g per sample), weighed, capped, and sealed, then inserted into a TAM Air calorimeter (TA Instruments). As a control sample to normalize the heat flow measurements, an equivalent volume of aged, hardened alkali-activated slag paste was placed in an identical glass vial in the calorimeter. Heat flow was measured at 20 °C for about 48 h for each sample. 2.4. Zeta potential A ZetaProbe analyzer (Colloidal Dynamics) was used to measure the zeta potential (f, mV) of slag particles well-dispersed in MilliQ water (3 wt.% suspension). This method measures the movement of colloidal particles due to the applied alternating electrical field, which depends on the particle size and zeta potential [31]. Na2 SiO3, NaOH and KOH solutions were each used for concentration titration of the slag sample, to monitor the changes in surface charge induced by addition of the activators. At each step, a quantity of alkaline solution equivalent to 2.2 104 moles M2O per gram of slag was added to the slag suspension, and after equilibration, zeta potential was measured based on the initial sample volume and the added volume of the titrant. Titration continued up to 1.1 103 moles M2O per gram slag. Before each set of measurements, the instrument was calibrated using potassium tungstosilicate solution, and all calculations were performed using the ZetaProbe Polar software (v. 2.14). 2.5. pH measurement A glass electrode pH meter was used to measure pH of a 50 wt.% slag suspension with activator doses of 2.2 104 moles M2O per gram of slag, and also for the samples in the ZetaProbe instrument during activator titration.
3. Results and discussion 3.1. Effect of activator type and dosage on paste rheology The use of alkali hydroxide activators in AAS pastes leads to a higher yield stress, and also a very rapid increase in yield stress within a few minutes after mixing, compared to alkali silicate activators. Fig. 2a shows that at 2.25 104 moles M2O/g slag, both NaOH and KOH activated slags show an increase in yield stress within 10 min after mixing, while Fig. 2b shows the same trend but with higher intensity at 4.5 104 moles M2O/g slag. At this higher alkali dosage, the NaOH-activated paste shows a particularly rapid yield stress increase, rising to 510 Pa after 20 min. This is a very thick paste and far from the fluid state. For self-leveling or pumpable concrete, a paste yield stress value much nearer to zero is more appropriate. The slag paste without activator showed no obvious changes in yield stress within this timeframe, which means that the yield stress changes in the alkali hydroxide activated slag paste are due to the effect of the activator in inducing inter-particle adhesion forces, and/or a dissolution/gel formation reaction process within
Fig. 2. Yield stress of the alkali-activated slag pastes with (a) 2.25 104 and (b) 4.5 104 moles M2O/g slag, and slag paste without activator, as a function of time after initial mixing.
this short time period. The sodium metasilicate activated paste, regardless of the time after mixing within this timeframe, shows a consistent yield stress of around 14 Pa, which seems satisfactory for fluidity of a cementitious binder. The fact that the sodium metasilicate-activated system shows a lower yield stress than the non-activated system is consistent with the known deflocculating activity of alkali silicate solutions in many particulate suspensions [32], and could be related to an increase in repulsive forces at the particle surfaces by surface adsorption of silicate species. This will be discussed in more detail in Section 3.3, using the outcomes of the zeta potential measurement to provide further insight. However, this does also indicate that the increase in yield stress of the hydroxide-containing pastes is likely to be related at least in part to flocculation effects, as the dissolution/gel formation process would be expected to be similar (or even more rapid) in the silicate-containing paste. Fig. 3 shows the apparent viscosity (shear stress/shear strain) of the activated slag pastes at 2.25 104 moles M2O/g slag,
Fig. 3. Apparent viscosity of the activated slag pastes with 2.25 104 moles M2O/ g slag according the shear strain regime in Fig. 1.
586
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591
measured according to the applied shear strain regime described in Fig. 1. The apparent viscosity curve for alkali hydroxide activated slag at 4.5 104 moles M2O/g slag is difficult to measure due to the lack of workability and high stiffness of the paste a few minutes after mixing and so the data here are presented only for the lower alkali dosage. The viscosity of the alkali activated slag pastes drops substantially with increasing shear strains, which means that inter-particle forces can be more easily broken at higher shear rates in the early stages of the alkali-activation process, and constant mixing of the paste at higher shear strain could reduce the viscosity to provide better workability before application. Longer-duration mixing has also been shown to provide enhanced compactness and resistance to drying shrinkage in alkali-activated slag pastes [33], which is consistent with this concept. Alkali hydroxide-activated slag pastes are more shear-sensitive than those activated by sodium silicate. This means that the viscosity reduction in the alkali hydroxide activated paste with increasing shear rate is substantial and reaches the same viscosity level as alkali silicates, although at the lowest shear strain the viscosity difference between them is quite substantial. Although sodium silicate activated slag shows shear thinning behavior, the intensity of the viscosity drop at higher shear strains is lower compared to alkali hydroxide-activated pastes. Activated slag pastes show thixotropic behavior (time-dependent viscosity differences between forward and backward lines in Fig. 3 at the same shear strains). For instance, the first and the final viscosity measurement at the lowest shear strain in Fig. 3 are very different, which shows that the destruction of many of the bonds between particles is non-reversible after 3 min of exposure to high shear strains, especially for alkali hydroxide activated slag. Therefore, mechanical mixing is essential to maintaining a low paste viscosity in alkali-activated slag, which would help to provide better workability for pouring and pumping of a concrete based on this binder. Also, contrary to the general belief that extensive mixing can lower strength by disruption of the setting structure, a recent study showed that increasing mixing time improves matrix cohesion and compactness of AAS, and hence its mechanical strength [34]. For an application scenario in which both higher stiffening rate and lower paste viscosity are needed, possibly such as spray concreting, alkali hydroxide activated slag with high-shear mechanical mixing before application could be an option.
3.2. Effect of activator type and dosage on heat flow and its relationship with yield stress Isothermal calorimetry (Fig. 4) shows that the reaction of alkali hydroxide activated slag is more rapid than sodium metasilicate activated slag. An increase in activator dosage results in faster reaction in all pastes, with the calorimetric data showing a sharper main peak at earlier time. During the first 30 min a local sharp peak is observed for all activated slag pastes (inset of Fig. 4). The heat evolved immediately upon contact between the slag and the activator is attributed to wetting and dissolution of slag particles and formation of dissolved silicate species [16], and is only partially captured in the measurements due to the time taken to load and equilibrate the samples in the instrument following external mixing. After the initial heat release peak within the first 30 min, which is common to all alkali-activated systems as well as most other cements, a difference in behavior is noticeable between the hydroxide-activated and the silicate-activated slag pastes. Heat flow starts to increase again almost immediately for hydroxide systems, but remains constant or slightly decreasing for silicate activated systems, for an extended induction period before the start of another heat flow peak, as has been well documented in the literature for these systems [16,17]. This rapid increase in heat flow for the hydroxide-activated system is in some way comparable to the rapid increase in yield stress for this system, while the long induction period for the silicate activated slag correlates to the observed workability retention. There is a correlation between the point at which the heat release begins to increase after the first sharp peak (often described as the onset of the ‘acceleration period’ in calorimetry of Portland cement), and the trend in yield stress increase, for systems activated by alkali hydroxides. According to Fig. 2, the yield stress of alkali hydroxide activated slag increases rapidly after mixing, while the yield stress of sodium silicate activated slag remains almost constant during the first 20 min. Comparison between the isothermal calorimetry results (normalized heat flow) at the early hours and yield stress changes of activated slag pastes with 4.5 104 moles M2O/g slag are shown in Fig. 5. The same trend was also observed for the lower activator dosage. The trend between samples in the increase in heat flow after 1 h correlates with the increase in yield stress in the same sample set, although this takes place much earlier in the reaction process. The reasons controlling the time difference
Fig. 4. Normalized heat flow of all alkali activated slag pastes with 2.25 104 and 4.5 104 moles M2O/g slag (as marked), measured using isothermal calorimetry.
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591
587
Fig. 5. Relationship between normalized heat flow and yield stress changes for all alkali activated slag pastes with 4.5 104 moles M2O/g slag.
for this correlation between heat evolution and yield stress increase will be discussed later. One of the key factors determining the changes in yield stress after mixing for alkali hydroxide activated slag is the reaction of the slag, where dissolution of slag particles into solution and formation of reaction products, both on the slag surfaces and in the liquid regions [35]. This reaction process will affect inter-particle forces via the formation of new bonds (physical and/or chemical) between new products and other constituents of the system, as well as an increase in the effective volume fraction of solids within the paste as some of the water becomes chemically bound. The first of these phenomena was observed in pastes consisting of metakaolin activated by hydroxides [36], as the formation of sodium aluminosilicate gels as a product between metakaolin grains was responsible for elastic behavior of the paste leading to an increase in yield stress at early age. However, for systems containing slag, the presence of calcium must be taken into account, as this leads instead to the formation of calcium–sodium aluminosilicate hydrate gel as the main product, and this fills space within the binder [37] and leads to an increase in solids volume fraction. Regardless of the exact mechanism, if an increase in the heat flow is an indication of the formation of reaction products, then it is likely that the reaction products are responsible in some way for the observed yield stress increase of alkali hydroxide activated slag. Consumption of water during the reaction process of alkali activated slag, both through the gel hydrate formation mechanism as mentioned above and also through its consumption
in hydrolysis of silicate and aluminate species from the slag glass into solution, could also be another factor changing yield stress because some portion of the water which is necessary to disperse slag aggregates by formation of a liquid layer between slag particles is no longer available because it is now bound. If this ‘‘excess water’’ [38] is consumed during reactions, it would cause reagglomeration of the particles, and thus an increase in yield stress. If the connection between heat release and yield stress changes is correct, then an obvious question is why they are not happening at the same time in Fig. 5; the yield stress increase takes place earlier than the end of the induction period as measured by calorimetry. To clarify this point, additional calorimetry measurements were undertaken for alkali hydroxide activated pastes, under the same measurement conditions, but using samples of paste which were taken directly from the rheometer after yield stress measurement (Section 2.2). The reason for this experiment is to determine how the rate of reaction is influenced by the shear applied during yield stress measurement, compared to when the paste is sealed in the glass vials immediately after mixing, and the results are shown in Fig. 6. It is evident from these results that both the time and intensity of the main heat flow peak depend on the shear history of the material. The alkali hydroxide pastes which were used for yield stress test before calorimetry show significantly faster reaction with sharper peaks compared to the paste which was sealed into the glass vials of the calorimeter after mixing. The reason why the reaction is accelerating during the rheology measurement rather than in the glass vials of the calorimeter could
Fig. 6. Comparison between normalized heat flows (NHF) for alkali hydroxide activated slag pastes immediately after mixing and after yield stress measurement (immediately mixed samples shifted 30 min forward in main plot for comparison) at 2.25 104 moles M2O/g slag. Inset: first 1 h of reaction, without time shift.
588
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591
be because the mixing and strong shear forces result in separation of the flocs forming in the primary calcium–sodium silicate hydrate gel as a consequence of mixing during the early stages of slag activation [34]. This deflocculation would then reduce the mass transport-related hindrance of reaction which is caused by flocculation and gel formation, and mean that the reaction is more intense (due to higher dissolved species concentrations in solution) once it is allowed to occur, when the material is no longer being sheared. Also, the initial (Si-rich) products that are rapidly formed on slag particles could be mechanically removed by mixing, then the new products which form on the slag surface could be different in chemical structure and establish different bonds with adjacent particles. 3.2.1. Effect of AAS pH on the reaction rate For most alkali activated binder systems, an elevated pH is essential to start the dissolution and reaction of the solid aluminosilicate precursors, but will not necessarily result in a faster or more extensive reaction. Also, for concentrated oxide suspensions, it has been shown that the highest yield stress is observed in the pH range in which the zeta potential value is close to zero (i.e., the isoelectric point) [39]. The question of zeta potential will be revisited in Section 3.3, but to guide that discussion, the pH of alkali activated slag pastes with different types and dosages of the activators was first measured. A paste with a high concentration of slag (w/s 0.45) and high alkali hydroxide dosage will be very difficult to analyze by most pH measurement techniques, due to its very high pH (>14), ionic strength and viscosity, which cause sensitivity errors during measurements using a glass electrode. To circumvent this problem, the pH was measured for a more dilute system (3 wt.% slag in water) for different activator types and dosages, as shown in Fig. 7. The activator dosages were held constant with respect to the slag content, giving a dilution factor of approximately 10 compared to the paste specimens for all samples. The measured pH values are very similar to the pH data available in the literature for sodium hydroxide and sodium metasilicate [40], which is consistent with the fact that a slag–water suspension is itself rather alkaline (pH 11.7, data points at Na2O = 0 in Fig. 7) and thus the interaction of the slag with the silicate or hydroxide solutions does not lead to an appreciable reduction in pH. Measurements of pH in the liquids with no addition of slag were in all cases 0.1–0.2 units higher than the suspension pH values shown in Fig. 7. As expected, a higher activator dosage increases the pH. At the same molar ratio of M2O/slag, NaOH and KOH show the same pH at lower dosages, and NaOH slightly lower than KOH at higher doses, but always higher than Na2SiO3 at the same dosage. According to the results of Section 3.2, NaOH activated slag has a higher reaction rate than KOH activated slag, while the pH values are similar or
Fig. 7. Measured pH of 3 wt.% slag suspension in water immediately after addition of different activator dosages.
slightly higher for KOH. Therefore it can be concluded that pH does not directly control the reaction rate, although it certainly plays a role in initiating the reaction. The minimum pH for an effective alkali-activation reaction has not yet been fully defined, but it should be above 11.7 because a suspension of slag in water, which is around this pH, reacts only very slowly. Some low alkali and neutral pH activators have been shown to be able to form reaction products at a low rate for specialized applications such as waste immobilization, where early strength is not critical, but the reacting slag slurry still shows a transient alkaline (>12) pH [41]. It seems that higher pH starts the slag reaction more effectively, but the ongoing reaction process is more dependent on the activator chemistry than pH [42]. Shi and Day [14] also determined that the effect of anions in the activator is more important than pH in characterizing hydration behavior of different Na-alkali activated slags. This could be one reason for the higher early reaction rate in sodium hydroxide activated slag compared to sodium silicate systems, although some other properties also differ between these two activators such as different chemistry of the reactions, speciation of the silicates, and activator viscosity. However, the comparison of NaOH and KOH activated slag in this study also shows the effect of activator cations on the reaction rate. Fig. 8 also shows that the pH of a much more concentrated alkali activated slag slurry (w/s = 1) at 2.2 104 moles M2O/g slag does not change within the first 90 min of reaction. This paste also does not set within this timeframe. The constant pH shows either that OH ions are not significantly consumed during this time period, or that the removal of acidic SiO2 from solution by formation of calcium silicates counteracts OH- consumption and makes the pH constant. This is in accordance with the understanding of the reaction product assemblages in alkali activated slag [43], because unlike in Portland cement hydration where the production of portlandite consumes OH ions from solution, in alkali activated slag no discrete solid hydroxide products are formed at early age. Hydrotalcite, which does consume some OH, is formed mainly at later age [8]. 3.3. Effect of activator type and dosage on zeta potential A suspension of ground granulated blast furnace slag in water has a basic pH (around 11.7 at 3 wt.%, Fig. 7). At this pH some of the silanol groups on the wetted particle surface will deprotonate and induce a net negatively charged surface. Cationic species such as Na+, K+, and Ca2+ will then adsorb on these surface silanol groups and provide some positively charged sites in a double layer structure [44]. According to Fig. 9, the addition of NaOH or KOH to the slag suspension results in a transition from negative to positive zeta potential values, which increase with higher dosages of the
Fig. 8. Measured pH of slag suspension in water (w/s mass ratio = 1) as a function of time for different activators, at 2.2 104 moles M2O/g slag.
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591
Fig. 9. Zeta potential of slag suspensions in water (3 wt.%) after addition of different activator dosages.
hydroxide activators. This runs contrary to the expectation that the silanol groups would become more extensively deprotonated at high pH, and yield a negatively charged surface. This indicates that the release of Ca2+ from the slag particles through partial dissolution at higher pH, and interactions with dissolved M+ cations, leads to a very significant influence on surface chemistry. The positive zeta potential measurements confirm the presence of the cations on the surfaces of slag the particles. A higher positive value is observed for KOH than NaOH at higher dosages because Na+ is a well-hydrated ion and prefers to associate with water molecules rather than slag particles, compared to K+ which is a larger and poorly-hydrated ion with a lower charge density [27,35]. This discussion is in accordance with the higher pH values at higher dosages for the KOH system (Fig. 8). Conversely, sodium silicate addition to slag suspension causes higher negative values of zeta potential, as silica is adsorbed. Addition of silicate species into a slag suspension can result in more negative zeta potential, as negatively charged silicate species from the suspension can adsorb or precipitate on the slag particle surfaces, which results in more negative values of zeta potential. This has been documented to happen when the concentration of silicates is high compared to the availability of Ca2+ [14]. Zeta potential is a key measurement of the electric double layer forces between particles. This force is significant in determining the flocculation or dispersion of the particles in suspension, and hence has a direct relationship to yield stress. 3.3.1. Relationship between the zeta potential of slag suspension and paste yield stress According to the DLVO theory for oxide suspensions [45], there are two main forces between particles: attractive Van der Waals forces, and repulsive electrostatic forces induced via the formation of a double layer of counter-ions. These forces will play a major role in determining the net interaction between particles [39]. Any changes in chemistry which aim to intensify or alleviate each type of force will have a direct influence on the structural evolution of the paste, and therefore on yield stress. Zeta potential is directly affected by the double layer forces, and therefore by measuring the zeta potential, changes in double layer forces can be traced. Scales et al. [28] introduced Eq. (1) to relate DLVO forces to yield stress measurement by the vane method (sy: yield stress, Kstruc: network structural term dependent upon the particle size, the solids volume fraction and the mean coordination number, FVDW: van der Waals force, FEDL: electric double layer forces, f: zeta potential (for the expanded equation refer to [28]).
sy ¼ K struc ½F VDW F EDL ; F EDL f2
589
forces are reduced (i.e. lower yield stress). The assumption of a homogeneous particle charge may or may not be precisely accurate for slag particles, but it will certainly be more appropriate for this material than for other commonly used cementitious components which are either poly-phasic (Portland cement clinker, fly ash), or show differences in particle face and edge charging behavior (calcined clays). Franks [27] studied the zeta potential of silica suspensions in concentrated monovalent electrolytes, and found that at high salt concentration (and also at high pH), when greater quantities of poorly-hydrated ions adsorb on the particle surface, zeta potential increases but yield stress is also higher compared to when zeta potential is zero. It is assumed that the ions take up positions between the surfaces of neighboring particles in order to minimize their free energy, which results in a type of attraction force [27]. This strong bond between particles results in flocculation and increases the yield stress. The same trend was noted here for slag particles at different dosages of the sodium silicate which forms the electrolyte in this study. As shown in Fig. 10, a small addition of sodium silicate increases the magnitude of zeta potential (represented as f2 in Fig. 10), and thus halves the paste yield stress by increasing double layer repulsive forces. However, a further increase in the concentration of Na2SiO3, although raising the magnitude of zeta potential substantially, leads to a yield stress increase. This shows that although the magnitude of the electric double layer force increases at higher electrolyte dosage, it is not purely repulsive any more, but rather shows some signs of attractiveness. Fig. 10 also shows a minimum in yield stress at a specific Na2SiO3 dosage which brings inter-particle forces to a minimum level. Silicate anions adsorb on the slag surface in areas that carry local positive charges due to cationic species such as Na+ and Ca2+ shielding the slag silanol groups [44]. This may result in strong electric double layer repulsive forces between slag particles, as monitored by an increase in the magnitude of zeta potential at lower electrolyte concentration and lower pH (Fig. 7). These repulsive forces also cause particle separation and therefore yield stress reduction at this point. Further addition of sodium silicate induces higher negative values for zeta potential, but according to Fig. 7, it also raises the pH to a higher level, bringing a higher extent of slag dissolution and thus release of cations such as Ca2+ and Mg2+ into the solution. In addition, the liquid phase can rapidly become supersaturated with respect to calcium silicate hydrates, leading to gel formation and development of some electrostatic bonds. As is displayed in Fig. 11, cations such as Ca2+ and Mg2+ are more likely to adsorb on silanol groups than Na+ and K+ because of their double charges,
ð1Þ
According to Eq. (1), yield stress will be at a maximum when the zeta potential is zero. Increasing the magnitude of the zeta potential results in an increase in the repulsive double layer forces between homogeneously charged particles, and therefore net inter-particle
Fig. 10. Relationship between the square of zeta potential of slag suspensions at 0, 2.2, 4.4, and 6.6 104 mole Na2O/g slag (as marked on each point) and the yield stress of sodium silicate activated slag paste at w/c = 0.45.
590
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591
Acknowledgments This work has been funded through an Australian Research Council Linkage Project grant, co-funded by Zeobond Pty. Ltd., and also benefited from support through the Particulate Fluids Processing Centre, a Special Research Centre of the Australian Research Council. The participation of JLP was also supported by the Faculty of Engineering, University of Sheffield.
References Fig. 11. A schematic sketch of electrostatic inter-particle forces at a high dosage of sodium silicate in slag paste.
and so electrostatic bonds between silanol groups on different slag particles and/or dissolved silicate species may form. Conversely, for alkali hydroxide activated slag, although the increase in zeta potential is a sign of repulsive electric double layer forces which could reduce the yield stress, the fast slag dissolution and reaction, with consumption of water, seem to be more influential in determining yield stress changes. The dissolution/reaction process thus causes important rheological changes in alkali activated binders during the early stages of reaction, but the exact nature of these changes depends on the chemistry of the activating solution. 4. Conclusions Alkali hydroxide activators cause a significant increase in the yield stress of an activated slag paste, especially at higher dosages as reactions take place in the fresh paste. Products such as C–(N)– A–S–H gel are rapidly formed in the liquid between slag particles and/or on slag particle surfaces. The rate of the reaction, as measured by heat evolution, shows a correlation with the time at which the yield stress increases, and therefore the rapid yield stress increase observed with hydroxide activators can be related to the consumption of water and formation of gel reaction products which introduce new inter-particle forces. Silicate activated slag shows limited reaction at early age during the induction period, and hence no yield stress increase during this time. Addition of sodium silicate gives a lower yield stress than non-activated slag due to the plasticizing and deflocculating effects of the silicate anions, which adsorb on the particle surfaces, increase the magnitude of repulsive double layer electric forces, and hence reduce the yield stress. Although a higher dosage of the activator induces a higher pH and ionic strength, which results in some attractive dipole forces between ions at the surface and an increase in yield stress, this is still below the yield stress of a suspension of the same volume fraction of slag in water with no activator. The effect of activator pH on reaction rate of alkali activated slag is at most indirect, as long as the pH is high enough to initiate the dissolution/reaction process. All activated slag specimens tested show shear thinning and thixotropic behavior. This means that inter-particle forces can be easily broken under the application of higher shear stresses at early age, and constant high-shear mixing of the paste could lower the viscosity for better workability before placement. This paper has provided new insight into the effects of activator type and dosage on rheological properties of an alkali activated slag paste, and relates these parameters to reaction/dissolution processes and changes in the zeta potential. However, a deeper understanding of the reaction mechanisms which cause changes in the type and magnitude of inter-particle forces is still needed if the full potential of alkali-activated binder systems is to be unlocked.
[1] ACI Committee 226. Ground granulated blast-furnace slag as a cementitious constituent in concrete. ACI Mater J 1987;84:327–42. [2] Bernal SA, Provis JL, Fernández-Jiménez A, Krivenko PV, Kavalerova E, Palacios M, et al. Binder chemistry – high-calcium alkali-activated materials. In: Provis JL, van Deventer JSJ, editors. Alkali-activated materials: state-of-the-art report, RILEM TC 224-AAM. Dordrecht: Springer/RILEM; 2014. p. 59–91. [3] Myers RJ, Bernal SA, San Nicolas R, Provis JL. Generalized structural description of calcium–sodium aluminosilicate hydrate gels: the crosslinked substituted tobermorite model. Langmuir 2013;29:5294–306. [4] Krizan D, Zivanovic B. Effects of dosage and modulus of water glass on early hydration of alkali–slag cements. Cem Concr Res 2002:1181–8. [5] Ben Haha M, Le Saout G, Winnefeld F, Lothenbach B. Influence of activator type on hydration kinetics, hydrate assemblage and microstructural development of alkali activated blast-furnace slags. Cem Concr Res 2011;41:301–10. [6] Provis JL. Activating solution chemistry for geopolymers. In: Provis JL, van Deventer JSJ, editors. Geopolymers: structure, processing, properties and industrial applications. Cambridge (UK): Woodhead; 2009. p. 50–71. [7] Duxson P, Provis JL. Designing precursors for geopolymer cements. J Am Ceram Soc 2008;91:3864–9. [8] Gruskovnjak A, Lothenbach B, Holzer L, Figi R, Winnefeld F. Hydration of alkaliactivated slag: comparison with ordinary Portland cement. Adv Cem Res 2006;18:119–28. [9] Wang SD, Scrivener KL. Hydration products of alkali activated slag cement. Cem Concr Res 1995;25:561–71. [10] Li C, Sun HH, Li LT. A review: the comparison between alkali-activated slag (Si plus Ca) and metakaolin (Si plus Al) cements. Cem Concr Res 2010;40:1341–9. [11] Chen W, Brouwers HJH, Shui ZH. Three-dimensional computer modeling of slag cement hydration. J Mater Sci 2007;42:9595–610. [12] Brough AR, Atkinson A. Sodium silicate-based, alkali-activated slag mortars: Part I. Strength, hydration and microstructure. Cem Concr Res 2002;32:865–79. [13] Krizan D, Zivanovic B. Effects of dosage and modulus of water glass on early hydration of alkali–slag cements. Cem Concr Res 2002;32:1181–8. [14] Shi C, Day RL. A calorimetric study of early hydration of alkali–slag cements. Cem Concr Res 1995;25:1333–46. [15] Ravikumar D, Neithalath N. Reaction kinetics in sodium silicate powder and liquid activated slag binders evaluated using isothermal calorimetry. Thermochim Acta 2012;546:32–43. [16] Bernal SA, Provis JL, Rose V, Mejía De Gutierrez R. Evolution of binder structure in sodium silicate-activated slag–metakaolin blends. Cem Concr Compos 2011;33:46–54. [17] Fernández-Jiménez A, Puertas F. Alkali-activated slag cements: kinetic studies. Cem Concr Res 1997;27:359–68. [18] Double DD, Hellawell A, Perry SJ. Hydration of Portland cement. Proc Roy Soc A 1978;359:435–51. [19] Rahhal V, Talero R. Calorimetry of Portland cement with metakaolins, quartz and gypsum additions. J Therm Anal Calorim 2008;91:825–34. [20] Banfill PFG. Rheology of fresh cement and concrete. Rheol Rev 2006:61–130. [21] Hanehara S, Yamada K. Rheology and early age properties of cement systems. Cem Concr Res 2008;38:175–95. [22] Hamada D, Hamai T, Shimoda M, Shonaka M. Development of new superplasticizer providing ultimate workability. In: Malhotra M, editor. 8th CANMET/ACI international conference on superplasticizers and other admixtures in cement, Sorrento, Italy; 2006. p. 185–90.. [23] Palacios M, Houst YF, Bowen P, Puertas F. Adsorption of superplasticizer admixtures on alkali-activated slag pastes. Cem Concr Res 2009;39:670–7. [24] Palacios M, Banfill PFG, Puertas F. Rheology and setting of alkali-activated slag pastes and mortars: effect of organic admixture. ACI Mater J 2008;105:140–8. [25] Kashani A, Provis JL, Xu J, Kilcullen AR, Qiao GG, van Deventer JSJ. Effect of molecular architecture of polycarboxylate ethers on plasticizing performance in alkali activated slag paste. J Mater Sci 2014;49:2761–72. [26] Marchon D, Sulser U, Eberhardt A, Flatt RJ. Molecular design of comb-shaped polycarboxylate dispersants for environmentally friendly concrete. Soft Matter 2013:10719–28. [27] Franks GV. Zeta potentials and yield stresses of silica suspensions in concentrated monovalent electrolytes: isoelectric point shift and additional attraction. J Colloid Interf Sci 2002;249:44–51. [28] Scales PJ, Johnson SB, Healy TW, Kapur PC. Shear yield stress of partially flocculated colloidal suspensions. AIChE J 1998;44:538–44. [29] Nguyen QD, Boger DV. Measuring the flow properties of yield stress fluids. Annu Rev Fluid Mech 1992;24:47–88.
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591 [30] Dzuy NQ, Boger DV. Direct yield stress measurement with the vane method. J Rheol 1985;29:335–47. [31] O’Brien RW, Jones A, Rowlands WN. A new formula for the dynamic mobility in a concentrated colloid. Colloid Surf A – Physicochem Eng Asp 2003;218:89–101. [32] Weldes HH, Lange KR. Properties of soluble silicates. Ind Eng Chem 1969;61:29–44. [33] Palacios M, Puertas F. Effectiveness of mixing time on hardened properties of waterglass-activated slag pastes and mortars. ACI Mater J 2011;108:73–8. [34] Palacios M, Puertas F. Effectiveness of mixing time on hardened properties of water glass activated slag pastes and mortars. ACI Mater J 2011;108:73–8. [35] Steins P, Poulesquen A, Diat O, Frizon F. Structural evolution during geopolymerization from an early age to consolidated material. Langmuir 2012;28:8502–10. [36] Favier A, Habert G, d’Espinose de Lacaillerie JB, Roussel N. Mechanical properties and compositional heterogeneities of fresh geopolymer pastes. Cem Concr Res 2013;48:9–16. [37] Provis JL, Myers RJ, White CE, Rose V, van Deventer JSJ. X-ray microtomography shows pore structure and tortuosity in alkali-activated binders. Cem Concr Res 2012;42:855–64. [38] Wong HHC, Kwan AKH. Rheology of cement paste: role of excess water to solid surface area ratio. J Mater Civ Eng 2008;20:189.
591
[39] Johnson SB, Franks GV, Scales PJ, Boger DV, Healy TW. Surface chemistry– rheology relationships in concentrated mineral suspensions. Int J Miner Process 2000;58:267–304. [40] Bacon LR, Wills JH. pH-composition relations in the system Na2O–SiO2–H2O at 20 °C. J Franklin Inst 1954;258:347–69. [41] Rashad AM, Bai Y, Basheer PAM, Milestone NB, Collier NC. Hydration and properties of sodium sulfate activated slag. Cem Concr Compos 2013;37:20–9. [42] Bai Y, Collier NC, Milestone NB, Yang CH. The potential for using slags activated with near neutral salts as immobilisation matrices for nuclear wastes containing reactive metals. J Nucl Mater 2011;413:183–92. [43] Provis JL, Bernal SA. Geopolymers and related alkali-activated materials. Annu Rev Mater Res; 2014 [in press], http://dx.doi.org/10.1146/annurev-matsci070813-113515. [44] Labbez C, Jonsson B, Pochard I, Nonat A, Cabane B. Surface charge density and electrokinetic potential of highly charged minerals: experiments and Monte Carlo simulations on calcium silicate hydrate. J Phys Chem B 2006;110:9219–30. [45] Bostrom M, Deniz V, Franks GV, Ninham BW. Extended DLVO theory: electrostatic and non-electrostatic forces in oxide suspensions. Adv Colloid Interf Sci 2006;123:5–15.
Contents lists available at ScienceDirect
Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat
The interrelationship between surface chemistry and rheology in alkali activated slag paste Alireza Kashani a, John L. Provis a,b,⇑, Greg G. Qiao a, Jannie S.J. van Deventer a,c a
Department of Chemical and Biomolecular Engineering, University of Melbourne, Victoria 3010, Australia Department of Materials Science and Engineering, University of Sheffield, Sheffield S1 3JD, United Kingdom c Zeobond Pty. Ltd., P.O. Box 23450, Docklands, Victoria 8012, Australia b
h i g h l i g h t s
g r a p h i c a l a b s t r a c t
Surface chemistry controls rheology
in alkali-activated slag cements. Particle–particle interactions and
gelation influenced by activator nature. Yield stress depends on square of zeta potential. Silicate and hydroxide activators differ notably in behavior. Influence of pH on fresh paste chemistry is indirect.
a r t i c l e
i n f o
Article history: Received 3 February 2014 Received in revised form 10 April 2014 Accepted 26 April 2014
Keywords: Alkali activated slag Rheology Yield stress Isothermal calorimetry Zeta potential pH
a b s t r a c t Ground granulated blast furnace slag can react with an alkaline activating solution to form a cement-like binder based on a calcium–sodium aluminosilicate gel, which is a potential alternative to Portland cement in many applications. This study provides new information regarding the effect of activator type and dosage on rheology by monitoring changes in pH, particle surface charge (zeta potential), and heat evolution in the early stages of the reaction process. Sodium and potassium hydroxide silicate solutions, at two different M2O (M: Na, K) dosages, are used here as activators. Alkali hydroxide activators cause a significant increase in the yield stress of an activated slag paste, especially at higher dosages as reactions take place rapidly, while within the same timeframe, the yield stress of the silicate activated slag remains unchanged. The results imply a direct relationship between a higher reaction rate with the formation of solid products (causing both spatial blockage effects and consumption of free water), and a rapid yield stress increase. However, the dependence of reaction rate on pH for different alkali-activated pastes is, at most, indirect. All activators induce a highly alkaline pH and a concentrated electrolyte solution environment in the fluid paste. As a result of complexation of poorly-hydrated ions on the surfaces of the particles, the magnitude of the zeta potential increases. A direct relationship is observed between the dosage of the activators and zeta potential. A zeta potential further from neutrality generally reduces yield stress by increasing the magnitude of double layer repulsive forces, with the exception of a higher dosage of silicate activator, which shows an indication of some attractive double layer forces. Ó 2014 Elsevier Ltd. All rights reserved.
⇑ Corresponding author at: Department of Materials Science and Engineering, University of Sheffield, Sheffield S1 3JD, United Kingdom. Tel.: +44 114 222 5490. E-mail address: j.provis@sheffield.ac.uk (J.L. Provis). http://dx.doi.org/10.1016/j.conbuildmat.2014.04.127 0950-0618/Ó 2014 Elsevier Ltd. All rights reserved.
584
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591
1. Introduction Ground granulated blast furnace slag (GGBFS) is the by-product of reduction of iron oxides to metallic iron in a blast furnace, and when quenched forms a glassy calcium–magnesium aluminosilicate which granulates and is ground to form a fine powder. GGBFS is often combined with Portland cement as a supplementary cementitious material to reduce cost and attributed CO2 emissions and to enhance mechanical and durability properties [1]. However, slag can also react as a sole precursor in an alkaline environment to form a cement-like binder [2], which is predominantly composed of a calcium–sodium aluminosilicate gel [3]. The commonly used alkali activators include alkali metal hydroxides, silicates and carbonates [2]. Among these activators, sodium silicate provides benefits regarding the ultimate strength of the alkali activated slag (AAS) binder, which can be similar to that of a good quality Portland cement depending on the activator modulus (molar ratio SiO2/Na2O) and dosage [4]; however, its setting rate is not as fast as when alkali hydroxides are used [5,6]. A key aim of many studies discussing and characterizing reaction products in AAS [7–10] has been to clarify the reaction mechanisms which are involved in the setting and hardening process. Upon contact with an alkaline solution, slag undergoes partial dissolution [11], with an exothermic reaction process [12–14] resulting in hardening if the water/solids ratio is sufficiently low. Isothermal calorimetry is used to monitor heat flow changes during alkali-activation reactions, and shows distinguishable differences as a function of alkali activator type and dosage, both in terms of the magnitude and duration of heat release [14–17]. It is also used as a tool to monitor reaction rate based on normalized heat flow curve versus time for Portland cement and blended cements [18,19]. The fluidity of an AAS paste can be monitored through changes in rheological properties such as yield stress and apparent viscosity. A low yield stress paste is generally considered desirable, as it reduces the energy required for pumping a concrete mix, and enhances the ability to fill voids when placing concrete. In addition, concretes are usually premixed for a specific time before placement, so it is also important to avoid much yield stress increase during the early age reaction. A rapid early reaction leading to thickening and solidification of a cement paste affects the workability and fluidity of the binder in transportation and placement. This is controlled in Portland cements by the use of organic plasticizing admixtures [20,21], which has enabled the development of concretes which are workable (or pumpable), and yet show high early strength. These admixtures are designed to control surface interactions, and so can reduce the yield stress of the mix, reduce the likelihood of a sudden yield stress increase during the reaction process, and control the rate of thickening [22]. However, chemical admixtures are not effective in all different environments of cementitious materials, and need to be tailor designed case by case. For example, the typical plasticizers used in cement do not plasticize as effectively in alkali activated binders [23,24]. The development of admixtures for alkali-activated systems is still ongoing [25,26], and requires a much more detailed understanding of the surface chemistry of slag particles within the complex reacting environment of a fresh AAS paste. The work presented in this paper is targeted toward building this understanding. Compared to the chemistry of a Portland cement, the addition of an alkali activator provides a higher pH and different electrolyte environment in the fresh binder paste, leading to differences in surface charges of particles and thus also in the electrostatic interactions between particles. However, the surface chemistry of alkali
activated binders, and particularly its interrelationship with rheology, is not well understood. This paper provides new insight into the effects of activator type and dosage on the rheological properties of an alkali activated slag paste. Yield stress is measured and related to the reactions taking place in this complex particle–fluid system. Zeta potential measurement is used to monitor the changes in electric double layer forces under different pH and electrolyte environments, which determine the sites and signs of the charges on particle surfaces [27]. Measurement of electric double layer forces by means of zeta potential provides insight into key parameters which determine yield stress [28]. The study results in an important new understanding of the relationship between the yield stress and the zeta potential of the activated slag paste. 2. Materials and methods 2.1. Materials Ground granulated blast furnace slag, with chemical composition as displayed in Table 1, and anhydrous sodium metasilicate powder (Na2O 50.9 wt.% and SiO2 49.1 wt.%) were provided by Zeobond Pty. Ltd., Australia. AR-grade NaOH and KOH (Sigma–Aldrich, Australia) were used. All activators were dissolved in MilliQ-grade purified water before mixing with slag to avoid any interference due to solid activator particle dissolution in the heat flow measurements, or the influence of additional particles on the paste solids fraction (and thus rheology). Dosages of 2.25 104 and 4.5 104 moles M2O per gram of slag were used, and specified on a molar basis to give the same amount of alkali ions for different activators at each dosage (for instance, 2.25 104 mole M2O per gram slag is equivalent to 1.80 g NaOH, 2.52 g KOH, and 2.75 g Na2SiO3 per 100 g slag). The samples were formulated by mixing slag with the activator solution at a water to slag weight ratio of 0.45. One sample was formulated without an activator but with the same water to binder ratio as a reference. After hand mixing to combine the components, the paste was mechanically mixed for 2 min or 10 min at 500 rpm before calorimetry and rheology measurements, respectively. 2.2. Rheological tests The rheological behavior of each AAS sample was measured using a Haake VT550 rheometer with a four-blade vane with a length of 50.1 mm and diameter of 15.3 mm, in a sample volume sufficient to simulate an infinite medium [29]. A direct yield stress measurement at a low rotational rate (0.2 rpm) was used because of the accuracy and reproducibility of the measured data [30], and the yield stress was calculated from the applied torque based on the geometry of the vane and the rotation rate according to [30]. Before each measurement, the paste was mixed by hand to avoid settling of particles and to provide a homogenous paste, left to rest for 30 s to dissipate residual stresses induced by mixing, before the 0.2 rpm rotational rate was applied. The first yield stress measurement was taken 12 min after combination of the slag and the activator (this includes 10 min of mixing and 2 min for the residual stress relaxation after high shear mixing). Five consecutive measurements were then taken at intervals of 2 min. The apparent viscosity at a range of higher shear strains was also measured, immediately after yield stress measurement, for each sample in 15 consecutive steps (Fig. 1) and without mixing between measurements.
Table 1 Chemical composition of the slag used, as determined by XRF. LOI is loss on ignition at 1000 °C and d90 was determined using a Malvern Mastersizer instrument. Oxide component
wt.%
SiO2 TiO2 Al2O3 Fe2O3 MnO MgO CaO Na2O K2O SO3 LOI d90 (lm)
33.1 0.6 15.0 0.6 0.3 5.9 41.8 0 0.3 2.3 0.4 39.8
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591
585
Fig. 1. Shear strains applied in 15 consecutive steps to the pastes for the measurement of apparent viscosity at higher shear strain (each step takes 20 s).
2.3. Isothermal calorimetry The activated slag paste was mechanically mixed for 2 min at 500 rpm, then poured into glass vials (about 20 g per sample), weighed, capped, and sealed, then inserted into a TAM Air calorimeter (TA Instruments). As a control sample to normalize the heat flow measurements, an equivalent volume of aged, hardened alkali-activated slag paste was placed in an identical glass vial in the calorimeter. Heat flow was measured at 20 °C for about 48 h for each sample. 2.4. Zeta potential A ZetaProbe analyzer (Colloidal Dynamics) was used to measure the zeta potential (f, mV) of slag particles well-dispersed in MilliQ water (3 wt.% suspension). This method measures the movement of colloidal particles due to the applied alternating electrical field, which depends on the particle size and zeta potential [31]. Na2 SiO3, NaOH and KOH solutions were each used for concentration titration of the slag sample, to monitor the changes in surface charge induced by addition of the activators. At each step, a quantity of alkaline solution equivalent to 2.2 104 moles M2O per gram of slag was added to the slag suspension, and after equilibration, zeta potential was measured based on the initial sample volume and the added volume of the titrant. Titration continued up to 1.1 103 moles M2O per gram slag. Before each set of measurements, the instrument was calibrated using potassium tungstosilicate solution, and all calculations were performed using the ZetaProbe Polar software (v. 2.14). 2.5. pH measurement A glass electrode pH meter was used to measure pH of a 50 wt.% slag suspension with activator doses of 2.2 104 moles M2O per gram of slag, and also for the samples in the ZetaProbe instrument during activator titration.
3. Results and discussion 3.1. Effect of activator type and dosage on paste rheology The use of alkali hydroxide activators in AAS pastes leads to a higher yield stress, and also a very rapid increase in yield stress within a few minutes after mixing, compared to alkali silicate activators. Fig. 2a shows that at 2.25 104 moles M2O/g slag, both NaOH and KOH activated slags show an increase in yield stress within 10 min after mixing, while Fig. 2b shows the same trend but with higher intensity at 4.5 104 moles M2O/g slag. At this higher alkali dosage, the NaOH-activated paste shows a particularly rapid yield stress increase, rising to 510 Pa after 20 min. This is a very thick paste and far from the fluid state. For self-leveling or pumpable concrete, a paste yield stress value much nearer to zero is more appropriate. The slag paste without activator showed no obvious changes in yield stress within this timeframe, which means that the yield stress changes in the alkali hydroxide activated slag paste are due to the effect of the activator in inducing inter-particle adhesion forces, and/or a dissolution/gel formation reaction process within
Fig. 2. Yield stress of the alkali-activated slag pastes with (a) 2.25 104 and (b) 4.5 104 moles M2O/g slag, and slag paste without activator, as a function of time after initial mixing.
this short time period. The sodium metasilicate activated paste, regardless of the time after mixing within this timeframe, shows a consistent yield stress of around 14 Pa, which seems satisfactory for fluidity of a cementitious binder. The fact that the sodium metasilicate-activated system shows a lower yield stress than the non-activated system is consistent with the known deflocculating activity of alkali silicate solutions in many particulate suspensions [32], and could be related to an increase in repulsive forces at the particle surfaces by surface adsorption of silicate species. This will be discussed in more detail in Section 3.3, using the outcomes of the zeta potential measurement to provide further insight. However, this does also indicate that the increase in yield stress of the hydroxide-containing pastes is likely to be related at least in part to flocculation effects, as the dissolution/gel formation process would be expected to be similar (or even more rapid) in the silicate-containing paste. Fig. 3 shows the apparent viscosity (shear stress/shear strain) of the activated slag pastes at 2.25 104 moles M2O/g slag,
Fig. 3. Apparent viscosity of the activated slag pastes with 2.25 104 moles M2O/ g slag according the shear strain regime in Fig. 1.
586
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591
measured according to the applied shear strain regime described in Fig. 1. The apparent viscosity curve for alkali hydroxide activated slag at 4.5 104 moles M2O/g slag is difficult to measure due to the lack of workability and high stiffness of the paste a few minutes after mixing and so the data here are presented only for the lower alkali dosage. The viscosity of the alkali activated slag pastes drops substantially with increasing shear strains, which means that inter-particle forces can be more easily broken at higher shear rates in the early stages of the alkali-activation process, and constant mixing of the paste at higher shear strain could reduce the viscosity to provide better workability before application. Longer-duration mixing has also been shown to provide enhanced compactness and resistance to drying shrinkage in alkali-activated slag pastes [33], which is consistent with this concept. Alkali hydroxide-activated slag pastes are more shear-sensitive than those activated by sodium silicate. This means that the viscosity reduction in the alkali hydroxide activated paste with increasing shear rate is substantial and reaches the same viscosity level as alkali silicates, although at the lowest shear strain the viscosity difference between them is quite substantial. Although sodium silicate activated slag shows shear thinning behavior, the intensity of the viscosity drop at higher shear strains is lower compared to alkali hydroxide-activated pastes. Activated slag pastes show thixotropic behavior (time-dependent viscosity differences between forward and backward lines in Fig. 3 at the same shear strains). For instance, the first and the final viscosity measurement at the lowest shear strain in Fig. 3 are very different, which shows that the destruction of many of the bonds between particles is non-reversible after 3 min of exposure to high shear strains, especially for alkali hydroxide activated slag. Therefore, mechanical mixing is essential to maintaining a low paste viscosity in alkali-activated slag, which would help to provide better workability for pouring and pumping of a concrete based on this binder. Also, contrary to the general belief that extensive mixing can lower strength by disruption of the setting structure, a recent study showed that increasing mixing time improves matrix cohesion and compactness of AAS, and hence its mechanical strength [34]. For an application scenario in which both higher stiffening rate and lower paste viscosity are needed, possibly such as spray concreting, alkali hydroxide activated slag with high-shear mechanical mixing before application could be an option.
3.2. Effect of activator type and dosage on heat flow and its relationship with yield stress Isothermal calorimetry (Fig. 4) shows that the reaction of alkali hydroxide activated slag is more rapid than sodium metasilicate activated slag. An increase in activator dosage results in faster reaction in all pastes, with the calorimetric data showing a sharper main peak at earlier time. During the first 30 min a local sharp peak is observed for all activated slag pastes (inset of Fig. 4). The heat evolved immediately upon contact between the slag and the activator is attributed to wetting and dissolution of slag particles and formation of dissolved silicate species [16], and is only partially captured in the measurements due to the time taken to load and equilibrate the samples in the instrument following external mixing. After the initial heat release peak within the first 30 min, which is common to all alkali-activated systems as well as most other cements, a difference in behavior is noticeable between the hydroxide-activated and the silicate-activated slag pastes. Heat flow starts to increase again almost immediately for hydroxide systems, but remains constant or slightly decreasing for silicate activated systems, for an extended induction period before the start of another heat flow peak, as has been well documented in the literature for these systems [16,17]. This rapid increase in heat flow for the hydroxide-activated system is in some way comparable to the rapid increase in yield stress for this system, while the long induction period for the silicate activated slag correlates to the observed workability retention. There is a correlation between the point at which the heat release begins to increase after the first sharp peak (often described as the onset of the ‘acceleration period’ in calorimetry of Portland cement), and the trend in yield stress increase, for systems activated by alkali hydroxides. According to Fig. 2, the yield stress of alkali hydroxide activated slag increases rapidly after mixing, while the yield stress of sodium silicate activated slag remains almost constant during the first 20 min. Comparison between the isothermal calorimetry results (normalized heat flow) at the early hours and yield stress changes of activated slag pastes with 4.5 104 moles M2O/g slag are shown in Fig. 5. The same trend was also observed for the lower activator dosage. The trend between samples in the increase in heat flow after 1 h correlates with the increase in yield stress in the same sample set, although this takes place much earlier in the reaction process. The reasons controlling the time difference
Fig. 4. Normalized heat flow of all alkali activated slag pastes with 2.25 104 and 4.5 104 moles M2O/g slag (as marked), measured using isothermal calorimetry.
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591
587
Fig. 5. Relationship between normalized heat flow and yield stress changes for all alkali activated slag pastes with 4.5 104 moles M2O/g slag.
for this correlation between heat evolution and yield stress increase will be discussed later. One of the key factors determining the changes in yield stress after mixing for alkali hydroxide activated slag is the reaction of the slag, where dissolution of slag particles into solution and formation of reaction products, both on the slag surfaces and in the liquid regions [35]. This reaction process will affect inter-particle forces via the formation of new bonds (physical and/or chemical) between new products and other constituents of the system, as well as an increase in the effective volume fraction of solids within the paste as some of the water becomes chemically bound. The first of these phenomena was observed in pastes consisting of metakaolin activated by hydroxides [36], as the formation of sodium aluminosilicate gels as a product between metakaolin grains was responsible for elastic behavior of the paste leading to an increase in yield stress at early age. However, for systems containing slag, the presence of calcium must be taken into account, as this leads instead to the formation of calcium–sodium aluminosilicate hydrate gel as the main product, and this fills space within the binder [37] and leads to an increase in solids volume fraction. Regardless of the exact mechanism, if an increase in the heat flow is an indication of the formation of reaction products, then it is likely that the reaction products are responsible in some way for the observed yield stress increase of alkali hydroxide activated slag. Consumption of water during the reaction process of alkali activated slag, both through the gel hydrate formation mechanism as mentioned above and also through its consumption
in hydrolysis of silicate and aluminate species from the slag glass into solution, could also be another factor changing yield stress because some portion of the water which is necessary to disperse slag aggregates by formation of a liquid layer between slag particles is no longer available because it is now bound. If this ‘‘excess water’’ [38] is consumed during reactions, it would cause reagglomeration of the particles, and thus an increase in yield stress. If the connection between heat release and yield stress changes is correct, then an obvious question is why they are not happening at the same time in Fig. 5; the yield stress increase takes place earlier than the end of the induction period as measured by calorimetry. To clarify this point, additional calorimetry measurements were undertaken for alkali hydroxide activated pastes, under the same measurement conditions, but using samples of paste which were taken directly from the rheometer after yield stress measurement (Section 2.2). The reason for this experiment is to determine how the rate of reaction is influenced by the shear applied during yield stress measurement, compared to when the paste is sealed in the glass vials immediately after mixing, and the results are shown in Fig. 6. It is evident from these results that both the time and intensity of the main heat flow peak depend on the shear history of the material. The alkali hydroxide pastes which were used for yield stress test before calorimetry show significantly faster reaction with sharper peaks compared to the paste which was sealed into the glass vials of the calorimeter after mixing. The reason why the reaction is accelerating during the rheology measurement rather than in the glass vials of the calorimeter could
Fig. 6. Comparison between normalized heat flows (NHF) for alkali hydroxide activated slag pastes immediately after mixing and after yield stress measurement (immediately mixed samples shifted 30 min forward in main plot for comparison) at 2.25 104 moles M2O/g slag. Inset: first 1 h of reaction, without time shift.
588
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591
be because the mixing and strong shear forces result in separation of the flocs forming in the primary calcium–sodium silicate hydrate gel as a consequence of mixing during the early stages of slag activation [34]. This deflocculation would then reduce the mass transport-related hindrance of reaction which is caused by flocculation and gel formation, and mean that the reaction is more intense (due to higher dissolved species concentrations in solution) once it is allowed to occur, when the material is no longer being sheared. Also, the initial (Si-rich) products that are rapidly formed on slag particles could be mechanically removed by mixing, then the new products which form on the slag surface could be different in chemical structure and establish different bonds with adjacent particles. 3.2.1. Effect of AAS pH on the reaction rate For most alkali activated binder systems, an elevated pH is essential to start the dissolution and reaction of the solid aluminosilicate precursors, but will not necessarily result in a faster or more extensive reaction. Also, for concentrated oxide suspensions, it has been shown that the highest yield stress is observed in the pH range in which the zeta potential value is close to zero (i.e., the isoelectric point) [39]. The question of zeta potential will be revisited in Section 3.3, but to guide that discussion, the pH of alkali activated slag pastes with different types and dosages of the activators was first measured. A paste with a high concentration of slag (w/s 0.45) and high alkali hydroxide dosage will be very difficult to analyze by most pH measurement techniques, due to its very high pH (>14), ionic strength and viscosity, which cause sensitivity errors during measurements using a glass electrode. To circumvent this problem, the pH was measured for a more dilute system (3 wt.% slag in water) for different activator types and dosages, as shown in Fig. 7. The activator dosages were held constant with respect to the slag content, giving a dilution factor of approximately 10 compared to the paste specimens for all samples. The measured pH values are very similar to the pH data available in the literature for sodium hydroxide and sodium metasilicate [40], which is consistent with the fact that a slag–water suspension is itself rather alkaline (pH 11.7, data points at Na2O = 0 in Fig. 7) and thus the interaction of the slag with the silicate or hydroxide solutions does not lead to an appreciable reduction in pH. Measurements of pH in the liquids with no addition of slag were in all cases 0.1–0.2 units higher than the suspension pH values shown in Fig. 7. As expected, a higher activator dosage increases the pH. At the same molar ratio of M2O/slag, NaOH and KOH show the same pH at lower dosages, and NaOH slightly lower than KOH at higher doses, but always higher than Na2SiO3 at the same dosage. According to the results of Section 3.2, NaOH activated slag has a higher reaction rate than KOH activated slag, while the pH values are similar or
Fig. 7. Measured pH of 3 wt.% slag suspension in water immediately after addition of different activator dosages.
slightly higher for KOH. Therefore it can be concluded that pH does not directly control the reaction rate, although it certainly plays a role in initiating the reaction. The minimum pH for an effective alkali-activation reaction has not yet been fully defined, but it should be above 11.7 because a suspension of slag in water, which is around this pH, reacts only very slowly. Some low alkali and neutral pH activators have been shown to be able to form reaction products at a low rate for specialized applications such as waste immobilization, where early strength is not critical, but the reacting slag slurry still shows a transient alkaline (>12) pH [41]. It seems that higher pH starts the slag reaction more effectively, but the ongoing reaction process is more dependent on the activator chemistry than pH [42]. Shi and Day [14] also determined that the effect of anions in the activator is more important than pH in characterizing hydration behavior of different Na-alkali activated slags. This could be one reason for the higher early reaction rate in sodium hydroxide activated slag compared to sodium silicate systems, although some other properties also differ between these two activators such as different chemistry of the reactions, speciation of the silicates, and activator viscosity. However, the comparison of NaOH and KOH activated slag in this study also shows the effect of activator cations on the reaction rate. Fig. 8 also shows that the pH of a much more concentrated alkali activated slag slurry (w/s = 1) at 2.2 104 moles M2O/g slag does not change within the first 90 min of reaction. This paste also does not set within this timeframe. The constant pH shows either that OH ions are not significantly consumed during this time period, or that the removal of acidic SiO2 from solution by formation of calcium silicates counteracts OH- consumption and makes the pH constant. This is in accordance with the understanding of the reaction product assemblages in alkali activated slag [43], because unlike in Portland cement hydration where the production of portlandite consumes OH ions from solution, in alkali activated slag no discrete solid hydroxide products are formed at early age. Hydrotalcite, which does consume some OH, is formed mainly at later age [8]. 3.3. Effect of activator type and dosage on zeta potential A suspension of ground granulated blast furnace slag in water has a basic pH (around 11.7 at 3 wt.%, Fig. 7). At this pH some of the silanol groups on the wetted particle surface will deprotonate and induce a net negatively charged surface. Cationic species such as Na+, K+, and Ca2+ will then adsorb on these surface silanol groups and provide some positively charged sites in a double layer structure [44]. According to Fig. 9, the addition of NaOH or KOH to the slag suspension results in a transition from negative to positive zeta potential values, which increase with higher dosages of the
Fig. 8. Measured pH of slag suspension in water (w/s mass ratio = 1) as a function of time for different activators, at 2.2 104 moles M2O/g slag.
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591
Fig. 9. Zeta potential of slag suspensions in water (3 wt.%) after addition of different activator dosages.
hydroxide activators. This runs contrary to the expectation that the silanol groups would become more extensively deprotonated at high pH, and yield a negatively charged surface. This indicates that the release of Ca2+ from the slag particles through partial dissolution at higher pH, and interactions with dissolved M+ cations, leads to a very significant influence on surface chemistry. The positive zeta potential measurements confirm the presence of the cations on the surfaces of slag the particles. A higher positive value is observed for KOH than NaOH at higher dosages because Na+ is a well-hydrated ion and prefers to associate with water molecules rather than slag particles, compared to K+ which is a larger and poorly-hydrated ion with a lower charge density [27,35]. This discussion is in accordance with the higher pH values at higher dosages for the KOH system (Fig. 8). Conversely, sodium silicate addition to slag suspension causes higher negative values of zeta potential, as silica is adsorbed. Addition of silicate species into a slag suspension can result in more negative zeta potential, as negatively charged silicate species from the suspension can adsorb or precipitate on the slag particle surfaces, which results in more negative values of zeta potential. This has been documented to happen when the concentration of silicates is high compared to the availability of Ca2+ [14]. Zeta potential is a key measurement of the electric double layer forces between particles. This force is significant in determining the flocculation or dispersion of the particles in suspension, and hence has a direct relationship to yield stress. 3.3.1. Relationship between the zeta potential of slag suspension and paste yield stress According to the DLVO theory for oxide suspensions [45], there are two main forces between particles: attractive Van der Waals forces, and repulsive electrostatic forces induced via the formation of a double layer of counter-ions. These forces will play a major role in determining the net interaction between particles [39]. Any changes in chemistry which aim to intensify or alleviate each type of force will have a direct influence on the structural evolution of the paste, and therefore on yield stress. Zeta potential is directly affected by the double layer forces, and therefore by measuring the zeta potential, changes in double layer forces can be traced. Scales et al. [28] introduced Eq. (1) to relate DLVO forces to yield stress measurement by the vane method (sy: yield stress, Kstruc: network structural term dependent upon the particle size, the solids volume fraction and the mean coordination number, FVDW: van der Waals force, FEDL: electric double layer forces, f: zeta potential (for the expanded equation refer to [28]).
sy ¼ K struc ½F VDW F EDL ; F EDL f2
589
forces are reduced (i.e. lower yield stress). The assumption of a homogeneous particle charge may or may not be precisely accurate for slag particles, but it will certainly be more appropriate for this material than for other commonly used cementitious components which are either poly-phasic (Portland cement clinker, fly ash), or show differences in particle face and edge charging behavior (calcined clays). Franks [27] studied the zeta potential of silica suspensions in concentrated monovalent electrolytes, and found that at high salt concentration (and also at high pH), when greater quantities of poorly-hydrated ions adsorb on the particle surface, zeta potential increases but yield stress is also higher compared to when zeta potential is zero. It is assumed that the ions take up positions between the surfaces of neighboring particles in order to minimize their free energy, which results in a type of attraction force [27]. This strong bond between particles results in flocculation and increases the yield stress. The same trend was noted here for slag particles at different dosages of the sodium silicate which forms the electrolyte in this study. As shown in Fig. 10, a small addition of sodium silicate increases the magnitude of zeta potential (represented as f2 in Fig. 10), and thus halves the paste yield stress by increasing double layer repulsive forces. However, a further increase in the concentration of Na2SiO3, although raising the magnitude of zeta potential substantially, leads to a yield stress increase. This shows that although the magnitude of the electric double layer force increases at higher electrolyte dosage, it is not purely repulsive any more, but rather shows some signs of attractiveness. Fig. 10 also shows a minimum in yield stress at a specific Na2SiO3 dosage which brings inter-particle forces to a minimum level. Silicate anions adsorb on the slag surface in areas that carry local positive charges due to cationic species such as Na+ and Ca2+ shielding the slag silanol groups [44]. This may result in strong electric double layer repulsive forces between slag particles, as monitored by an increase in the magnitude of zeta potential at lower electrolyte concentration and lower pH (Fig. 7). These repulsive forces also cause particle separation and therefore yield stress reduction at this point. Further addition of sodium silicate induces higher negative values for zeta potential, but according to Fig. 7, it also raises the pH to a higher level, bringing a higher extent of slag dissolution and thus release of cations such as Ca2+ and Mg2+ into the solution. In addition, the liquid phase can rapidly become supersaturated with respect to calcium silicate hydrates, leading to gel formation and development of some electrostatic bonds. As is displayed in Fig. 11, cations such as Ca2+ and Mg2+ are more likely to adsorb on silanol groups than Na+ and K+ because of their double charges,
ð1Þ
According to Eq. (1), yield stress will be at a maximum when the zeta potential is zero. Increasing the magnitude of the zeta potential results in an increase in the repulsive double layer forces between homogeneously charged particles, and therefore net inter-particle
Fig. 10. Relationship between the square of zeta potential of slag suspensions at 0, 2.2, 4.4, and 6.6 104 mole Na2O/g slag (as marked on each point) and the yield stress of sodium silicate activated slag paste at w/c = 0.45.
590
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591
Acknowledgments This work has been funded through an Australian Research Council Linkage Project grant, co-funded by Zeobond Pty. Ltd., and also benefited from support through the Particulate Fluids Processing Centre, a Special Research Centre of the Australian Research Council. The participation of JLP was also supported by the Faculty of Engineering, University of Sheffield.
References Fig. 11. A schematic sketch of electrostatic inter-particle forces at a high dosage of sodium silicate in slag paste.
and so electrostatic bonds between silanol groups on different slag particles and/or dissolved silicate species may form. Conversely, for alkali hydroxide activated slag, although the increase in zeta potential is a sign of repulsive electric double layer forces which could reduce the yield stress, the fast slag dissolution and reaction, with consumption of water, seem to be more influential in determining yield stress changes. The dissolution/reaction process thus causes important rheological changes in alkali activated binders during the early stages of reaction, but the exact nature of these changes depends on the chemistry of the activating solution. 4. Conclusions Alkali hydroxide activators cause a significant increase in the yield stress of an activated slag paste, especially at higher dosages as reactions take place in the fresh paste. Products such as C–(N)– A–S–H gel are rapidly formed in the liquid between slag particles and/or on slag particle surfaces. The rate of the reaction, as measured by heat evolution, shows a correlation with the time at which the yield stress increases, and therefore the rapid yield stress increase observed with hydroxide activators can be related to the consumption of water and formation of gel reaction products which introduce new inter-particle forces. Silicate activated slag shows limited reaction at early age during the induction period, and hence no yield stress increase during this time. Addition of sodium silicate gives a lower yield stress than non-activated slag due to the plasticizing and deflocculating effects of the silicate anions, which adsorb on the particle surfaces, increase the magnitude of repulsive double layer electric forces, and hence reduce the yield stress. Although a higher dosage of the activator induces a higher pH and ionic strength, which results in some attractive dipole forces between ions at the surface and an increase in yield stress, this is still below the yield stress of a suspension of the same volume fraction of slag in water with no activator. The effect of activator pH on reaction rate of alkali activated slag is at most indirect, as long as the pH is high enough to initiate the dissolution/reaction process. All activated slag specimens tested show shear thinning and thixotropic behavior. This means that inter-particle forces can be easily broken under the application of higher shear stresses at early age, and constant high-shear mixing of the paste could lower the viscosity for better workability before placement. This paper has provided new insight into the effects of activator type and dosage on rheological properties of an alkali activated slag paste, and relates these parameters to reaction/dissolution processes and changes in the zeta potential. However, a deeper understanding of the reaction mechanisms which cause changes in the type and magnitude of inter-particle forces is still needed if the full potential of alkali-activated binder systems is to be unlocked.
[1] ACI Committee 226. Ground granulated blast-furnace slag as a cementitious constituent in concrete. ACI Mater J 1987;84:327–42. [2] Bernal SA, Provis JL, Fernández-Jiménez A, Krivenko PV, Kavalerova E, Palacios M, et al. Binder chemistry – high-calcium alkali-activated materials. In: Provis JL, van Deventer JSJ, editors. Alkali-activated materials: state-of-the-art report, RILEM TC 224-AAM. Dordrecht: Springer/RILEM; 2014. p. 59–91. [3] Myers RJ, Bernal SA, San Nicolas R, Provis JL. Generalized structural description of calcium–sodium aluminosilicate hydrate gels: the crosslinked substituted tobermorite model. Langmuir 2013;29:5294–306. [4] Krizan D, Zivanovic B. Effects of dosage and modulus of water glass on early hydration of alkali–slag cements. Cem Concr Res 2002:1181–8. [5] Ben Haha M, Le Saout G, Winnefeld F, Lothenbach B. Influence of activator type on hydration kinetics, hydrate assemblage and microstructural development of alkali activated blast-furnace slags. Cem Concr Res 2011;41:301–10. [6] Provis JL. Activating solution chemistry for geopolymers. In: Provis JL, van Deventer JSJ, editors. Geopolymers: structure, processing, properties and industrial applications. Cambridge (UK): Woodhead; 2009. p. 50–71. [7] Duxson P, Provis JL. Designing precursors for geopolymer cements. J Am Ceram Soc 2008;91:3864–9. [8] Gruskovnjak A, Lothenbach B, Holzer L, Figi R, Winnefeld F. Hydration of alkaliactivated slag: comparison with ordinary Portland cement. Adv Cem Res 2006;18:119–28. [9] Wang SD, Scrivener KL. Hydration products of alkali activated slag cement. Cem Concr Res 1995;25:561–71. [10] Li C, Sun HH, Li LT. A review: the comparison between alkali-activated slag (Si plus Ca) and metakaolin (Si plus Al) cements. Cem Concr Res 2010;40:1341–9. [11] Chen W, Brouwers HJH, Shui ZH. Three-dimensional computer modeling of slag cement hydration. J Mater Sci 2007;42:9595–610. [12] Brough AR, Atkinson A. Sodium silicate-based, alkali-activated slag mortars: Part I. Strength, hydration and microstructure. Cem Concr Res 2002;32:865–79. [13] Krizan D, Zivanovic B. Effects of dosage and modulus of water glass on early hydration of alkali–slag cements. Cem Concr Res 2002;32:1181–8. [14] Shi C, Day RL. A calorimetric study of early hydration of alkali–slag cements. Cem Concr Res 1995;25:1333–46. [15] Ravikumar D, Neithalath N. Reaction kinetics in sodium silicate powder and liquid activated slag binders evaluated using isothermal calorimetry. Thermochim Acta 2012;546:32–43. [16] Bernal SA, Provis JL, Rose V, Mejía De Gutierrez R. Evolution of binder structure in sodium silicate-activated slag–metakaolin blends. Cem Concr Compos 2011;33:46–54. [17] Fernández-Jiménez A, Puertas F. Alkali-activated slag cements: kinetic studies. Cem Concr Res 1997;27:359–68. [18] Double DD, Hellawell A, Perry SJ. Hydration of Portland cement. Proc Roy Soc A 1978;359:435–51. [19] Rahhal V, Talero R. Calorimetry of Portland cement with metakaolins, quartz and gypsum additions. J Therm Anal Calorim 2008;91:825–34. [20] Banfill PFG. Rheology of fresh cement and concrete. Rheol Rev 2006:61–130. [21] Hanehara S, Yamada K. Rheology and early age properties of cement systems. Cem Concr Res 2008;38:175–95. [22] Hamada D, Hamai T, Shimoda M, Shonaka M. Development of new superplasticizer providing ultimate workability. In: Malhotra M, editor. 8th CANMET/ACI international conference on superplasticizers and other admixtures in cement, Sorrento, Italy; 2006. p. 185–90.. [23] Palacios M, Houst YF, Bowen P, Puertas F. Adsorption of superplasticizer admixtures on alkali-activated slag pastes. Cem Concr Res 2009;39:670–7. [24] Palacios M, Banfill PFG, Puertas F. Rheology and setting of alkali-activated slag pastes and mortars: effect of organic admixture. ACI Mater J 2008;105:140–8. [25] Kashani A, Provis JL, Xu J, Kilcullen AR, Qiao GG, van Deventer JSJ. Effect of molecular architecture of polycarboxylate ethers on plasticizing performance in alkali activated slag paste. J Mater Sci 2014;49:2761–72. [26] Marchon D, Sulser U, Eberhardt A, Flatt RJ. Molecular design of comb-shaped polycarboxylate dispersants for environmentally friendly concrete. Soft Matter 2013:10719–28. [27] Franks GV. Zeta potentials and yield stresses of silica suspensions in concentrated monovalent electrolytes: isoelectric point shift and additional attraction. J Colloid Interf Sci 2002;249:44–51. [28] Scales PJ, Johnson SB, Healy TW, Kapur PC. Shear yield stress of partially flocculated colloidal suspensions. AIChE J 1998;44:538–44. [29] Nguyen QD, Boger DV. Measuring the flow properties of yield stress fluids. Annu Rev Fluid Mech 1992;24:47–88.
A. Kashani et al. / Construction and Building Materials 65 (2014) 583–591 [30] Dzuy NQ, Boger DV. Direct yield stress measurement with the vane method. J Rheol 1985;29:335–47. [31] O’Brien RW, Jones A, Rowlands WN. A new formula for the dynamic mobility in a concentrated colloid. Colloid Surf A – Physicochem Eng Asp 2003;218:89–101. [32] Weldes HH, Lange KR. Properties of soluble silicates. Ind Eng Chem 1969;61:29–44. [33] Palacios M, Puertas F. Effectiveness of mixing time on hardened properties of waterglass-activated slag pastes and mortars. ACI Mater J 2011;108:73–8. [34] Palacios M, Puertas F. Effectiveness of mixing time on hardened properties of water glass activated slag pastes and mortars. ACI Mater J 2011;108:73–8. [35] Steins P, Poulesquen A, Diat O, Frizon F. Structural evolution during geopolymerization from an early age to consolidated material. Langmuir 2012;28:8502–10. [36] Favier A, Habert G, d’Espinose de Lacaillerie JB, Roussel N. Mechanical properties and compositional heterogeneities of fresh geopolymer pastes. Cem Concr Res 2013;48:9–16. [37] Provis JL, Myers RJ, White CE, Rose V, van Deventer JSJ. X-ray microtomography shows pore structure and tortuosity in alkali-activated binders. Cem Concr Res 2012;42:855–64. [38] Wong HHC, Kwan AKH. Rheology of cement paste: role of excess water to solid surface area ratio. J Mater Civ Eng 2008;20:189.
591
[39] Johnson SB, Franks GV, Scales PJ, Boger DV, Healy TW. Surface chemistry– rheology relationships in concentrated mineral suspensions. Int J Miner Process 2000;58:267–304. [40] Bacon LR, Wills JH. pH-composition relations in the system Na2O–SiO2–H2O at 20 °C. J Franklin Inst 1954;258:347–69. [41] Rashad AM, Bai Y, Basheer PAM, Milestone NB, Collier NC. Hydration and properties of sodium sulfate activated slag. Cem Concr Compos 2013;37:20–9. [42] Bai Y, Collier NC, Milestone NB, Yang CH. The potential for using slags activated with near neutral salts as immobilisation matrices for nuclear wastes containing reactive metals. J Nucl Mater 2011;413:183–92. [43] Provis JL, Bernal SA. Geopolymers and related alkali-activated materials. Annu Rev Mater Res; 2014 [in press], http://dx.doi.org/10.1146/annurev-matsci070813-113515. [44] Labbez C, Jonsson B, Pochard I, Nonat A, Cabane B. Surface charge density and electrokinetic potential of highly charged minerals: experiments and Monte Carlo simulations on calcium silicate hydrate. J Phys Chem B 2006;110:9219–30. [45] Bostrom M, Deniz V, Franks GV, Ninham BW. Extended DLVO theory: electrostatic and non-electrostatic forces in oxide suspensions. Adv Colloid Interf Sci 2006;123:5–15.