CHARACTERIZATION OF THE RHEOLOGICAL AND SWELLING PROPERTIES OF SYNTHETIC ALKALI SILICATE GELS IN ORDER TO PREDICT THEIR BEHAVIOR IN ASR

Transcription

1 The Pennsylvania State University The Graduate School College of Engineering CHARACTERIZATION OF THE RHEOLOGICAL AND SWELLING PROPERTIES OF SYNTHETIC ALKALI SILICATE GELS IN ORDER TO PREDICT THEIR BEHAVIOR IN ASR DAMAGED CONCRETE A Dissertation in Civil and Environmental Engineering by Asghar Gholizadeh Vayghan Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy May 2017

2 The dissertation of Asghar Gholizadeh Vayghan was reviewed and approved * by the following: Farshad Rajabipour Associate Professor of Civil and Environmental Engineering Dissertation Adviser Chair of Committee Carlo G. Pantano Distinguished Professor of Materials Science and Engineering James L. Rosenberger Professor of Statistics William D. Burgos Professor of Civil and Environmental Engineering (Chair of Graduate Program) * Signatures are on file in the Graduate School. ii

3 ABSTRACT Alkali silica reaction (ASR) is a major concrete durability concern that is responsible for the deterioration of concrete infrastructure in the world. The resultant of the reaction between the cement alkali hydroxides and the metastable silicates in the aggregates is a hygroscopic and expansive alkali silicate gel (referred to as ASR gel in this document). The swelling behavior of ASR gels determines the extent of damage to concrete structures and, as such, mitigation of ASR relies on understanding these gels and finding ways to prevent them either from formation, or from swelling after formation. This dissertation focuses on the synthesis and characterization of ASR gels with wide ranges of compositions similar to what has been reported for the filed ASR gels in the literature. The experimental work consisted of three phases as follow. Phase I: Investigation of rheology, chemistry and physics of ASR gels produced through sol method. Inspired from the existing literature, two sol gel methods have been developed for the synthesis of ASR gels. The rheological (primarily gelation time, yield stress, and equilibrium stress), chemical (pore solution ph, pore solution composition, osmotic pressure, solid phase composition, stoichiometry of gelation reactions) and physical (evaporable water, solid content, etc.) properties of synthetic ASR gels have been extensively investigated in this phase. Ca/Si, Na/Si and K/Si, and water content were considered as the main chemical composition variables. In order to investigate the suppressing effects of lithium on the swelling properties of ASR gels, the gels were added with lithium in a part of the experimental program. The results strongly suggested that Ca/Si has a positive effect on the yield stress of the gels and their rate of gelation. Na/Si was found to have a decreasing effect on the yield stress and gelation rate (especially at low Ca/Si levels). K/Si and Li/Si had second order (i.e., polynomial) effects on the yield stress of the iii

4 gels, causing a significant drop in this parameter followed by some increase as they approached their upper values. Na/Si and K/Si were both found contribute to the osmotic potential of the ASR gels, while increase in Ca/Si generally led to a drop in this parameter. The presence of all components (Ca, Na, and K) were found to contribute to the ph of the gels pore solution, and Ca/Si and Na/Si showed a synergistic effect on this parameter. Lithium, on the other hand, was found to be able to drop the OH - concentration of the pore solution by a factor of five in the case of high sodium gels, which could partially explain its ASR mitigating effect. Phase II: Investigation of the free and restrained swelling behavior, hydrophilic potential and viscoelastic properties of ASR gels produced through the paste method. 20 gel compositions were selected (using the central composite design method) with Ca/Si, Na/Si and K/Si molar ratios varying in the ranges ( ), ( ) and ( ), respectively. The gels were produced by batching appropriate amounts of certain precursors containing different chemical components. After curing, the gels were tested for the abovementioned parameters using some innovative test methods as explained in the relevant chapters. The results suggest that increasing the alkali content (Na/Si and K/Si) in ASR gels resulted in an increase in the gels free swelling and water absorption, and a reduction in the equilibrium relative humidity (ERH). However, no significant effect was found for Ca/Si with respect to the ERH. Ca/Si was found to have a multi episode effect on the swelling and water absorption properties of the gels. An increase in Ca/Si up to 0.18 led to a considerable reduction in the swelling strain, followed by a slight increasing effect as it approached 0.4. Further increase in Ca/Si resulted in complete elimination of swelling strain. While Na/Si and K/Si could constantly increase the free swelling strain, their excessive presence was found to have a softening effect on the gels structure, leading to a drop in their swelling pressure. Finally, all gels were found to show viscoelastic behavior that could be best explained via Burger s model. iv

5 The elastic and viscous components have been measured for each gel and related to their composition using regression. Phase III: An Extended Chemical Index Model to Predict the Fly Ash Dosage Necessary for Mitigating Alkali-Silica Reaction in Concrete. In order to have an applied and ready to implement contribution to the realm of alkali silica reaction, a predictive statistical model was developed that determines the optimum fly ash dosage for ASR mitigation depending on the acceptable risk of ASR and structure s importance. The model uses the oxide compositions of portland cement and fly ash, as well as the reactivity of the aggregates. Seventy-six experimental data points (published in the last two decades across the U.S. and Canada) on CPT expansion results for plain portland cement and fly ash-blended concrete mixtures were used to develop and evaluate the model. The model was found to be capable of predicting CPT expansion of concrete incorporating both class F and class C fly ash and reactive aggregates within ASTM C1293 precision criteria. It allows for the use of a wider range of fly ashes compared to what is currently being prescribed by the standards for use in concrete susceptible to ASR. v

6 TABLE OF CONTENTS List of Figures... x List of Tables... xvi Chapter 1: Introduction The nature of alkali-silica reaction Mechanism of ASR Mechanism of gelation Surface vs. intra-particle gel formation Mechanism of swelling Knowledge Gaps Research Plan Organization of contents References Chapter 2: Composition-Rheology Relationships in Alkali-Silica Reaction Gels and the Impact on the Gel s Deleterious Behavior Background The Current State of Knowledge Experimental Methods Gel synthesis process Rheological measurements Gelation time measurements Gel pore solution ph and composition measurements Gel composition measurements Experimental design Results and Discussions Rheological and gel point measurement results Gel pore solution ph and composition measurement results Gel composition measurement results Statistical analyses More discussion on the swelling of the gels and the role of rheology vi

7 2.5. Conclusions References Chapter 3: Characterization of the rheological and chemical properties of alkali silica reaction (ASR) gels containing lithium Background Experimental program Materials Design of experiments (DOE) Gel synthesis method Rheological measurements Chemical characterization techniques Results and Discussion Gelation time and yield stress results Chemical measurements and analyses results Summary and Conclusions References Chapter 4: The influence of alkali silica reaction (ASR) gel composition on its hydrophilic properties and free swelling in contact with water vapor Background Experimental program Materials Gel synthesis method Design of experiments (DOE) Free swelling strain measurements Equilibrium relative humidity (ERH) measurements Results Free swelling results Equilibrium relative humidity results Regression Model Conclusions References vii

8 Chapter 5: Quantifying the free and restrained swelling properties of alkali silica gels as a function of their composition Background Experimental Program Gel synthesis Free swelling experiments on ASR gel disks Swelling pressure experiments on ASR gel disks Design of Experiments Results and Discussion Free swelling results Swelling pressure results Regression analyses Conclusions References Chapter 6: Characterization of mechanical and viscoelastic behavior of synthetic alkali silica reaction (ASR) gels Introduction Experimental program Materials Design of experiments (DOE) Synthesis of ASR gels Compressive strength measurement Viscoelastic properties and Poisson s ratio measurements Closed form analyses for determination of viscoelastic behavior of the gels per Burger s Numerical analyses for determination of viscoelastic parameters of the gels per Burger s model: Regression analyses for modelling the mechanical and viscoelastic parameters to ASR gels chemistry Results Compressive strength Viscoelastic properties and Poisson s ratio results viii

9 6.4. Regression Model Compressive strength Viscoelastic properties Conclusions References Chapter 7: An Extended Chemical Index Model to Predict the Fly Ash Dosage Necessary for Mitigating Alkali-Silica Reaction in Concrete Background Earlier work on ASR predictive models for fly ash concrete The extended chemical index model Configuration of the model Training the model Evaluation of the model Fly ash dosage prediction charts Online application for estimation of fly ash dosage necessary to mitigate ASR Summary and conclusions References Chapter 8: Conclusions and Future Work ix

10 LIST OF FIGURES Figure 1-1. Cracking in a sea defense wall of plain concrete [8]... 3 Figure 1-2. Simplified 2-D schematic diagram of silica structure... 4 Figure 1-3. The solubility limit of amorphous silica in aqueous solution as a function of ph (25 o C)... 6 Figure 1-4. ASR expansion modes observed in different aggregates: (a) bronzite andesite [8], (b) andesite [28], (c) chert (left) and gneiss [32], (d) chert [29], (e) quartzite [29], (f) quartzite [33], (g) calcined flint [31], (h) soda lime glass [30] Figure 2-1. Schematic molecular model of a typical alkali-silica reaction gel Figure 2-2. Example of an expanded ASR gel that has not yet penetrated into the surrounding vacancies, likely because the gel s yield stress has not been exceeded Figure 2-3. Schematic demonstration of the formation of confining pressure and the seepage of the gel into adjacent vacancies Figure D diagram of the ion exchange unit Figure 2-5. The vane rheometry setup and the path of shear stress formed in the material under torsion: (a) the schematic diagrams, (b) the actual setup (the spindle is not completely inserted for the sake of demonstration) Figure 2-6. Shear stress time series plots for two gels at the age of 7 days Figure 2-7. Typical viscosity development time series of the alkali silicate gels used for determining the gelation time Figure 2-8. Gel pore solution extraction setup Figure 2-9. Yield stress development of the gels as a function of time x

11 Figure Equilibrium stress development of the gels as a function of time Figure Shear stress-time response of the gels at early stages of gelation Figure ph variations of the gels pore solution as a function of time Figure Time-dependent variations of solutes concentration and osmotic pressure in the gels pore solutions Figure The interaction plots of Ca/Si and Na/Si on the yield stress results; a) the effect of Ca/Si, b) the effect of Na/Si Figure The interaction plot of Ca/Si and Na/Si on the ph of the pore solution of the gels; a) the effect of Ca/Si, b) the effect of Na/Si Figure 3-1. Schematic illustration of the swelling of ASR gels in concrete, their microstructural and molecular configurations (a) the formation and expansion of ASR gel around the aggregates and the confining pressure from the surrounding cement paste due to the gel expansion; (b) a zoomed in view of the gel near a cement micro crack, the role of water ingress on the formation of confining stress and the hydrostatic pressure driving the gel into the adjacent vacancy, and the role of shear yield stress; (c) a higher magnification zoon in view of the gel microstructure illustrating the different phases in the gels; (d) the molecular structure of the gels showing the sheet like solid structure and the alkaline pore solution of the gel Figure 3-2. Typical viscosity development time series of the alkali silicate sols during the gelation process used for determining the gelation time Figure 3-3. The interaction plot of Ca/Si and Na/Si on the gelation time Figure 3-4. The interaction plot of Li/Si and Ca/Si on the gelation time Figure 3-5. The factorial plots of the average fitted yield stress versus different test variables.. 80 xi

12 Figure 3-6. The influence of water on ASR gels yield stress at different levels of K/Si and Li/Si Figure 3-7. The interaction plot of Ca/Si and Na/Si on the yield stress Figure 4-1. The distribution of the chemical composition of the studied gels in the literature (the x axes of (a) to (e) are the atomic ratios) Figure 4-2. Particle size distribution of the silica powder used for gel synthesis Figure 4-3. Graphical demonstration of RSM data point selection for 3 variables case (α = 1.682) Figure 4-4. Vacuum sealed minibar specimens of each gel composition Figure 4-5. Length change measurements of minibar specimens Figure 4-6. An illustration of the ERH measurements setup: Figure 4-7. The free swelling strain of ASR gel minibars up to 28 days Figure 4-8. The weight change (%) of ASR gel minibars up to 28 days (the error bars indicate two standard errors) Figure 4-9. The free swelling strain (%) versus weight change (%) variations of minibar specimens Figure The equilibrium relative humidity results of gels Figure The contour plots of 28 day free swelling strain (%), 28 day weight change (%) and equilibrium relative humidity (%) of the gels versus Ca/Si, Na/Si and K/Si Figure The pairwise scatter plots of 28 day free swelling strain (%), 28 day weight change (%) and equilibrium relative humidity of the gels xii

13 Figure 5-1. An illustration of synthetic ASR gel disk specimens: (a) three days after casting; (b) vacuum sealed for further curing Figure 5-2. The steel ring molds used for making cement paste disks Figure 5-3. Hydrated, polished, Teflon-taped cement paste disk partially slid back into the steel ring mold Figure 5-4. The 2-D and perspective views of the designed ASR gel disk swelling units: a) a 2-D scheme of the assembly items, b) water penetration and swelling model, c) the schematic 3-D view, d) the actual view of the finalized assembly Figure 5-5. Digital comparator used for measuring the free swelling strain of ASR gel units Figure 5-6. The reversed swelling strain experiment: (a) the loading test setup; (b) typical results of the experiment Figure 5-7. The ASR gel disks free swelling strain test results Figure 5-8. The swelling pressure of ASR gels as a function of their residual swelling strain Figure 5-9. The effect of Ca/Si atomic ratio of the ASR gels on their free swelling strain at different levels of Na/Si and K/Si Figure The effect of Na/Si atomic ratio of the ASR gels on their free swelling strain at different levels of Ca/Si and K/Si Figure The effect of K/Si atomic ratio of the ASR gels on their free swelling strain at different levels of Ca/Si and Na/Si Figure The effect of Ca/Si atomic ratio of the ASR gels on their restrained swelling pressure at different levels of Na/Si and K/Si Figure The effect of Na/Si atomic ratio of the ASR gels on their restrained swelling pressure at different levels of Ca/Si and K/Si xiii

14 Figure The effect of K/Si atomic ratio of the ASR gels on their restrained swelling pressure at different levels of Ca/Si and Na/Si Figure The contour plots of free swelling strain (εg,fr) and swelling pressure (Prs) as a function of Ca/Si, Na/Si and K/Si Figure The composition domain of deleterious ASR gels: (a) the 3D surface plot, (b) the 2D contour plot Figure 6-1. Different linear viscoelastic models examined for characterization of the rheological behavior of ASR gels: (a) Kelvin Voigt, (b) Maxwell model, and (c) Burger s model Figure 6-2. The typical SST variations as a function of the viscoelastic parameters, and their precision intervals Figure 6-3. The compressive strength results of the synthesized ASR gels Figure 6-4. The strain time series of the experimented synthetic ASR gel cylinders under constant stress level (0.4σc) Figure 6-5. The main effects and interactions plots of chemical variables with respect to the compressive strength Figure 6-6. The main effects and interactions plots of chemical variables with respect to the long term elastic modulus of ASR gels (E1) Figure 6-7. The contour plots of the instant and short term elastic response of ASR gels: (a) (c) for the long term elastic modulus: E1, (d) (f) for the short term elastic modulus: E Figure 6-8. The scatter plots of (a) η1 and (b) η2 versus Ca/Si and Na/Si at K/Si = Figure 7-1. Normalized expansion of the concrete prisms versus normalized values of the ASR suppressing components xiv

15 Figure 7-2. Normalized expansion of the concrete prisms versus normalized values of the ASR promoting components Figure 7-3. Normalized expansion of the concrete prisms versus Na2Oeq content of cement and fly ash Figure 7-4. The optimum correlation between the normalized ASR vulnerability index of binders and their CPT expansions. 65% of the expansions fall below the trend line Figure 7-5. Fly ash dosage required to mitigate ASR per ASTM C1293 concrete prism test (E2Yb=0.04%) xv

16 LIST OF TABLES Table 1-1. A summarized list of physicochemical properties of ASR gels studied at each chapter Table 2-1. The chemical composition and coding of the gels (the levels are the molar ratios) Table 2-2. The batched and analyzed composition and of the gels Table 2-3. The ANOVA table of the chemistry of ASR gels with respect to yield stress, (τy), equilibrium stress (τe), ph and osmotic pressure (Π) Table 2-4. The estimated swelling pressures of the gels at the age of 28 days Table 3-1. The studied ranges of the ASR gel composition (input) variables Table 3-2. The coded and natural levels of the input variables Table 3-3. The experimental units representing the gel compositions investigated in this research Table 3-4. The mixture proportions of the synthesized ASR gels (the reported masses are for each 1000 grams of the batched gel) Table 3-5. The gelation time (TG) and yield stress (τy) results of the gels Table 3-6. The chemical measurements and analyses results Table 3-7. The list of gelation reactions of the studied alkali silicate systems Table 4-1. The studied range of the chemical composition of synthetic ASR gels Table 4-2. The experimental design with the coded and natural levels of the composition variables Table 4-3. Mixture proportions of the gels xvi

17 Table 4-4. The shorthand ANOVA table of the free swelling, weight change, and equilibrium RH measurements results Table 5-1. The experimental design with the coded and natural levels of the gel composition variables Table 5-2. Mixture proportions of the gels Table 5-3. The shorthand ANOVA table of the free swelling strain and restrained swelling pressure of the ASR gels Table 6-1. The experimented levels of different chemical components of ASR gels Table 6-2. Chemical compositions and mixture proportions of the synthesized ASR gels Table 6-3. The viscoelastic parameters and Poisson s ratios of synthetic ASR gels along with the precision and confidence intervals Table 6-4. The Pearson correlation coefficients of different viscoelastic parameters Table 6-5. The short hand ANOVA table of ASR gels compressive strength results Table 6-6. The short hand ANOVA table of ASR gels viscoelastic parameters Table 7-1. Fly ash properties obtained from literature [21-24] and used in the model. Oxides are reported in % by mass Table 7-2. Portland cement properties obtained from literature [21-24] and used in the model. Oxides are reported in % by mass Table 7-3. Aggregates and their 1 year CPT expansion (E1Yc) results [21-24], and AASHTO PP- 65 [17] designation Table 7-4. E1Yc and E2Yb values based on the cement, fly ash, fly ash dosage, and aggregate type obtained from the literature xvii

18 Table 7-5. The optimized values of the coefficients in the ASR promoting and suppressing factors Table 7-6. Results on prediction accuracy of the charts for the seven data points with E2Yb=0.04±0.004% xviii

19 CHAPTER 1: INTRODUCTION Together with corrosion of reinforcing steel, alkali-silica reaction (ASR) is the lead cause of concrete infrastructure deterioration in the world. It is widespread across all 50 states of the United States and is costing U.S. taxpayers hundreds of millions of dollars annually to repair and restore the affected structures. ASR is the reaction between the alkalis in concrete (i.e. sodium and potassium hydroxide) and the chemically reactive silicates present in some aggregates. The resultant of ASR is an expansive alkali-silicate gel (conventionally referred to as ASR gel) that forms inside or on the surface of aggregates and can swell and deteriorate the surrounding concrete upon water imbibition within years to decades. The swelling behavior of ASR gels determines the extent of damage to concrete structures and, as such, mitigation of ASR relies on understanding these gels and finding ways to prevent them either from formation, or from swelling after formation. To date, the researchers have been struggling to even find out what properties of ASR gels should be studied and how. Since these gels form on the surface of or within aggregates of concrete, they are impossible to be directly studied without disturbance while they are swelling. Moreover, the indirect in-situ study of the gels has a lot of limitations and difficulties and has not led to any conclusive findings so far. Moreover, the ASR gels behavior varies tremendously as a function of their composition (which is in turn dictated by the aggregate and cement compositions, temperature, gel s age and location in concrete). The only promising alternative is to synthesize these gels under controlled conditions and study them for their important properties. In this chapter, the nature of alkali-silica reaction, ASR gels, the mechanism of swelling and damage are 1

20 introduced, the knowledge gaps are identified and the research plan for bridging the existing knowledge gaps are laid out THE NATURE OF ALKALI-SILICA REACTION Alkali-silica reaction (ASR) is one of the main durability problems of concrete structures and still remains as a challenging topic for many researchers and construction executives [1, 2, 3, 4]. This reaction, which is often followed by expansion and cracking of concrete, occurs between the alkaline pore solution of concrete and the chemically reactive silicates present in some aggregates [5, 6, 7]. Figure 1-1 shows an example of an ASR damaged sea defense concrete structure made from chert and slate aggregates that has developed cracks with 8-12 mm width and up to 33 cm depth [8]. The chemically reactive silicates that have exhibited high potential in undergoing deleterious ASR include opal, cristobalite, tridymite, micro-crystalline silica, strained quartz and some volcanic siliceous minerals [9, 10]. ASR damage is a result of four consecutive processes: (1) dissolution of the metastable silicates into the alkaline pore solution, (2) formation of colloidal silicate particles, (3) gelation of the alkali silicate sol, and (4) swelling of the gel through moisture absorption MECHANISM OF ASR In order to simplify the mechanism of ASR and explain the fundamentals of this reaction, the reactive aggregates are merely represented by their metastable silica components. This assumption 2

21 is commonly made by researchers when explaining alkali-silica reaction [5, 11, 12]. In the interior part of silica (SiO2) network, each silicon atom is surrounded by four oxygen atoms, and each oxygen is connected to two silicon atoms. This renders an overall SiO2 tetrahedral structure known as the short range order unit. Figure 1-1. Cracking in a sea defense wall of plain concrete [8] These units are connected together from the oxygen vertices (i.e. bridging oxygens) and form closed rings of different sizes [13]. The valences of the ions in the interior part of the network are satisfied. The exterior surface region of the network, however, ends with either Si+ or O- ions, which are not readily satisfied [5]. When the siliceous material is exposed to water, hydrolysis occurs at the surface as expressed in reaction 1a. Figure 1-1 shows a simplified 2-D schematic diagram of silica structure that has undergone surface hydrolysis. Many researchers have studied 3

22 the phenomenon of silica dissolution in water in the past decades [14, 15, 16, 17, 18, 19, 20]. Most of the existing literature, however, focuses on moderate levels of alkalinity (ph<11). It is well known that in an alkaline environment, hydroxyl ions (OH - ) progressively attack the siloxane (Si O Si) bonds, resulting in network dissolution of silica. The silanole (Si OH) groups at the surface (attached to the network by three satisfied oxygen atoms) are dissolved into the alkaline solution in the form of Si(OH)4 per reaction 1b. Figure 1-2. Simplified 2-D schematic diagram of silica structure At high ph, reaction 1b is very favorable to the right direction [5], meaning that the silica dissolution is accelerated and extended as the hydroxide concentration increases. This can be concluded by drawing the speciation graph of different silicon species as a function of ph (see Figure 1-3). 4

23 Hydrolysis: ( Si+ + Si O-)s + H2O 2( Si OH)s (1a) Network dissolution: ( Si OH)s + 3(OH-)aq (Si(OH)4)aq (1b) These curves are derived based on the solubility limit of amorphous silica per reaction 2a [ 21 ], and the ionization reactions of the dissolved hydrous silica per reaction 2b and 2c [19]. Note that reaction 2a is an alternative writing for reaction 1b. As the graphs suggest, amorphous silica is not very soluble at neutral ph and can only reach a concentration of 1.93 mm. As the ph increases, the Si(OH)4 species tend to release one and subsequently two protons and convert to H3SiO4 - and H2SiO4 2- groups. The first proton loss is notably initiated at the ph of ~9.5 making H3SiO4 - the dominant species, while H2SiO4 2- becomes dominant only at ph above Network dissolution: (SiO2)s + 2H2O (Si(OH)4)aq Ksp= (2a) Ionization: (Si(OH)4)aq (H3SiO4 - )aq + H + K1= (2b) (H3SiO4 - )aq (H2SiO4 2- )aq + H + K2= (2c) The charge on the ionized silica species can be balanced by alkali ions in the solution (e.g., Na +, K + ) [5]. Also, alkalis can replace hydrogen in any of these products [18]. Reaction 3 is an example of this process, which is usually referred to as ion exchange, where R stands for Na and K. Ion exchange: (HO OH Si OH OH) aq + R + aq + OH - (HO OH Si OH O + R - ) aq +H 2 O (3) 5

24 Silica solubility limit [M] 1.0E E E E E E E E E-12 Total {Si}aq {H2SiO4}aq {H3SiO4}aq {Si(OH)4}aq ph Figure 1-3. The solubility limit of amorphous silica in aqueous solution as a function of ph (25 o C) 1.3. MECHANISM OF GELATION The otherwise satisfied silica species formed per reactions 2 and 3 can join together and undergo condensation per reaction 4, where R usually stands for sodium or potassium [22]. When sufficient siloxane (Si-O-Si) bonds are formed in a region, they respond collectively as recognizable colloidal particles. Colloidal particles are small amounts of matter made of macromolecules or polymer molecules with sizes ranging from 10 to 10,000 angstroms. These particles are distributed within a continuous aqueous solution, which in this case is an alkaline solution. The mixture of these two phases is usually referred to as a sol. Therefore, a sol is a dispersion of colloidal particles in a liquid. 6

25 OH OH OH OH Condensation: RO Si OH + HO Si OR (RO Si O Si OR) sol +H 2 O (4) OH OH OH OH The sol formation step takes place very slowly at high ph, since one of the reactants needs to donate one OH - group, which has a huge barrier due to the high concentration of hydroxyl ions in the aqueous solution. The gelation occurs when the colloidal silica particles link together, which is accompanied by increase in the viscosity of the system. A three dimensional network known as silica gel is formed in this step, which is in turn referred to as polymerization [22]. The polymerization step also takes place slowly at high ph and low concentration of colloidal particles due to the hydroxide barrier and large distance between the colloidal particles. The presence of sodium and potassium also slow down the kinetics of condensation and polymerization since these monovalent cations tend to connect to the oxygen endings of silica species (see reaction 3) and make them unable to cross-link with other species. Orthosilicic species (Si(ONa)4) are the extreme cases that are fully isolated and are unable to undergo sol-gel steps. However, the presence of calcium (and potentially other polyvalent metals; e.g. aluminum) drastically accelerates the condensation and polymerization of the silica species. The preliminary experiments performed by the candidate suggest that the gelation time of a 17% alkali silicate solution (Na/Si=0.5) decreases from several weeks to less than one minute by adding calcium up to a Ca/Si=0.2 ratio. The condensation of silicate species in the presence of calcium ions proceeds as shown in reaction (5) [14]. Further cross linking between colloidal particles results in gelation, which occurs much faster compared to the case of gelation without presence of calcium. 7

26 The role of calcium on suppressing or intensifying ASR remains to be controversial. It has been argued that the presence of calcium is essential for the formation of the gel [23, 24, 25, 26], since in the absence of calcium, the alkali silicate species will remain dissolved in the solution and will not undergo gelation. On the other hand, there are evidences and conclusions that the presence of calcium in the structure of the gel decreases their swelling capacity, and that a calcium-rich gel can be considered as non-swelling gel [5, 27]. OH 2 HO Si O - ( OH ) + Ca 2+ OH OH HO Si O Ca O Si OH ( OH OH ) sol (5) 1.4. SURFACE VS. INTRA-PARTICLE GEL FORMATION One of the main controversial subjects of alkali-silica reaction is the mode of reaction and expansion of the aggregate in the concrete matrix. Some aggregates react with the alkaline solution from their surface and form expansive ASG at their interface with the cement paste [28, 29]. Examples of aggregates reacting from the surface include gneiss, granite, greywacke, andesite, opal and quartzite. Some other aggregates, however, react and swell from the existing microcracks within the aggregates [30, 31]. Calcined flint and recycled soda-lime glass are reported to fall into this category. Both types of expansion are reported for chert and quartzite [8, 28, 29, 30, 31, 32, 33]. Figure 1-4 shows examples of aggregates belonging to each category that have undergone ASR expansion. In the presence of sufficient calcium, the latter category does not 8

27 exhibit gel formation and swelling from the surface. It is argued that they form a non-swelling calcium-rich layer of alkali silicate precipitate that prevents formation of ASG on the surface of the aggregates [30]. In the case of insufficient calcium in the system, surface reaction and expansion might be observed in these aggregates [34]. The formation of the gel is commonly followed by expansion that damages the concrete. This is discussed in the next section. According to many researchers, a reaction rim made of a calciumrich alkali silicate precipitate must form around the aggregates so that due to Gibbs-Donan effect and osmosis, the ASR expansion and pressure is generated [5, 28]. This reaction rim must be permeable with respect to sodium and hydroxide ions so that the aggregate dissolution continues MECHANISM OF SWELLING It is crucial to remember that formation of the gel per se is not dangerous, but it is the swelling capacity and swelling pressure of the gel that determines whether the safety of the concrete is compromised. There are reports regarding the formation of innocuous sodium-rich gels that are flowable enough to diffuse into the pore space or micro-cracks of the cement paste without exerting any significant pressure to their surrounding [31]. The swelling is found to be a function of the gels composition [5, 12, 31]. As the alkali concentration increases, the swelling capacity will generally increase, while the presence of calcium reduces the free swelling of the gel [5]. This idea was postulated by Powers and Steinour in However, Struble and Diamond (1981) later suggested that this could be an oversimplification as they found no significant contribution from calcium in reducing the swelling 9

28 of their synthetic alkali silicate gels [9]. Rodrigues et al. (2001) supported the conclusions of Powers and Steinour by relating the expansion of ASGs to the valence and relative concentration of cations present in the gels [35]. Their work relied on two previous papers published by the same group in 1997 and 1998, where the double layer theory was used to explain the expansion behavior of the gels [36, 37]. According to their work, the expansion of the gel depends on the surface charge density of the reaction products at a given ph. a b c d e f g h Figure 1-4. ASR expansion modes observed in different aggregates: (a) bronzite andesite [8], (b) andesite [28], (c) chert (left) and gneiss [32], (d) chert [29], (e) quartzite [29], (f) quartzite [33], (g) calcined flint [31], (h) soda lime glass [30] 10

29 The surface charge density is suggested to be a function of the valence and concentration of the cations. In this work, the expansion is discussed to increase as the ph of the environment and the concentration of the cations increase and their valences decrease. There are other reports, however, that attribute the expansion of the gels to several mechanisms rather than solely their composition. A number of phenomena can be responsible for the significant sorptivity and swelling of some ASR produced gels. The porous network of alkali silicate gels comprises a large number of water accessible O - Na +, O - Ca 2+ O - and O - groups, which are hydrophilic due to Van der Waals forces, and can cause swelling upon water adsorption. In addition, investigations confirm enormous swelling of even dilute gels in contact with water, acid or alkali solutions due to osmosis and Gibbs-Donnan effect [38]. Osmosis is the net movement of solvent between two aqueous media through a semipermeable membrane. The solvent flows from the lower solute concentration region to the higher solute concentration region. This is due to the lower water potential in the higher solute concentration zone compared to the other zone that gives rise to a negative osmotic pressure forcing the water to flow across the membrane. The semi-permeable membrane acts as a molecular sieve allowing water molecules to pass through, while preventing large solutes to cross the membrane. In the case of ASR, the semi-permeable membrane is a calcium-rich alkali silicate precipitate surrounding the gel that allows water molecules as well as diffuse in. However, calcium cations and silica species are mostly hindered to travel across the membrane due to their relatively larger sizes compared to Na + and OH - [5]. Osmosis can generate significant destructive pressures inside an ASR damaged concrete. Assuming an equal ph of 13 between inside and outside of the gel, the concentration of H3SiO4 - and H2SiO4 2- solutes can reach a maximum of 16.4 M and 6.49 M respectively (see Figure 1-3). 11

30 According to Morse s equation (6) the ultimate osmotic pressure that can be generated solely by them is 97.4 MPa, which is far beyond the tensile strength of concrete (T=25 C and assuming ionization constants of 1 and 2 for H3SiO4 - and H2SiO4 2- ). However, calculation of the osmotic pressure cannot be easily done for these systems due to the following reasons: (1) The system is almost never thermodynamically at equilibrium; (2) the equation is for ideal solutions where the osmotic coefficient of the ions are equal to one, which is not true for the silicic ions; (3) due to the Gibbs-Donnan effect some redistribution of the ions across the membrane occurs giving rise to ph change, and (4) the assumption of charge balance on each side of the semi-permeable membrane is also not necessary valid in these systems. Nevertheless, a small fraction of the ultimate theoretical osmotic pressure is sufficient to crack the concrete matrix. Π=-i C R T (6) Where Π is the osmotic pressure, i is the ionization constant, C is the concentration of the solute in moles, R is the gas constant and T is the temperature in Kelvin. The Gibbs-Donan effect is in fact the result of uneven distribution of ions across the two sides of a semi-permeable membrane. Due to unequal concentration of the ions across the membrane, the ions in the highly concentrated region tend to reach equilibrium by diffusing to the less concentrated region. The concentration of Na +, K + and OH - ions in the pore solution is higher compared to the gel, while the siliceous species are more concentrated inside the gel. The semipermeable membrane, however, only allows the small alkali hydroxide ions to diffuse into the gel, while prevents counter diffusion of large siliceous species (i.e. Si(OH)4, H3SiO4 - and H2SiO4 2- ) to the pore solution. The diffusion of Na + and OH - ions into the gel increases the osmotic pressure 12

31 of the gel and causes net movement of the outer solvent towards the gel inside the calcium-rich semi-permeable membrane [25] KNOWLEDGE GAPS In the past few decades, there has been a lot of research on the ASR swelling of reactive aggregates in the context of concrete as a whole, following standard or innovative test methods. There are many factors and mechanisms that may contribute to ASR expansion, while no adequate understanding, control and recording might be available for them in testing concrete samples for ASR expansion. Only a few papers directly measuring the swelling of the gels with a number of compositions have been published [9, 39]. No systematic investigation on the effect of the gel composition (i.e. SiO2.(Na2O)n.(K2O)k.(CaO)c.(Al2O3)a.(Fe2O3)f), on the characteristics of the gels can be found. This is one of the main reasons why the field of ASR remains not fully understood. The general understanding is that as the swelling capacity of ASR gels increases with increase in the alkali content [5, 40]. However, there is research that suggests drastic decline in the swelling pressure of the gels when the sodium content exceeds a certain level [31]. Although the alkalis seem to increase the free swelling of the gels, they may reduce the swelling pressure of the gel. Further research needs to be performed on the effect of alkalis on the rheology and behavior of the gels in restrained concrete. Although calcium is concluded to decrease the expansion of the gel [5, 30], there are reports that suggest no significant contribution from calcium on the swelling suppression. Expansions varying between 0.5% and 81.6% for a given level of Ca/Si are reported, while lower expansion 13

32 was found for the case of higher Na/Si ratio [9]. Some researchers have recognized an alkali recycling effect for calcium, which results in an increase in the ph of the pore solution/gel [5, 14]. On the other hand, some others have recognized calcium contributing in alkali sorption in CaO- SiO2-H2O and CaO-Na2O-SiO2-H2O gels, which leads to a decrease in the ph [41]. The most important characteristics of ASR gels are their free swelling capacity (strain), restrained swelling pressure, viscoelastic and rheological properties (primarily yield stress as quantified in a previous work [42]), and the osmotic pressure of their pore solution. The study of the rheological, chemical and swelling properties of synthetic ASR gels is important as it promotes (a) a better understanding of the gels swelling behavior and the factors that control it, (b) developing new chemical admixtures that suppress the swelling potential of ASR gels, and subsequently, damage in concrete, and (c) developing computer models to simulate ASR damage in concrete for predicting the service-life RESEARCH PLAN The main objective of this work is to characterize the rheological, swelling, mechanical, physical and chemical properties of synthetic ASR gels as a function of their composition (primarily, Ca/Si, Na/Si, K/Si, Li/Si atomic ratios and water content). Different types of statistical design of experiments and experimental setups were developed and the obtained results were analyzed to find the significance of the effects and interactions of the chemical variables on the aforementioned properties of ASR gels. The merit of this research is to find the deleterious compositions (i.e., the combinations of alkali and alkaline earth metals that render ASR gels with excessive swelling 14

33 potential). Such compositions can be further investigated in order find the proper ASR mitigation strategies and inhibiting admixtures, which is not the topic of this research ORGANIZATION OF CONTENTS This document is organized in 7 chapters as follow. A summary of the physicochemical properties of ASR gels studied in each chapter is listed in Table 1-1. Table 1-1. A summarized list of physicochemical properties of ASR gels studied at each chapter Chapter No. Physicochemical properties measured 2 Gelation time Yield stress Pore solution ph Pore solution chemistry and osmotic pressure Theoretical swelling pressure 3 Gelation time Yield stress Pore solution ph, chemistry and osmotic pressure Phase composition (mass ratio of different phases: solid phase, chemisorbed water, physisorbed water and pore solution) 4 Equilibrium relative humidity Free swelling strain in contact with water vapor Weight change in contact with water vapor 5 Free swelling strain in contact with water Restrain swelling pressure 6 Compressive strength Poisson s ratio Viscoelastic parameters per Burger s model Chapter one, as laid out above, was an introduction to the concept of alkali-silica reaction, the nature of ASR gels, their formation and swelling in concrete. The important characteristics of 15

34 ASR gels that govern their behavior in concrete and the existing knowledge gaps were identified and discussed, followed by a brief description of the research plan. Chapter two focuses on the rheological and osmotic properties of some ASR gel compositions as a function of time with a focus on the general effects and interaction of Ca/Si and K/Si atomic ratios. The gels yield stress is introduced and shown to be the major rheological property of ASR gels that affect their movement in concrete and (indirectly) their swelling pressure. The effect of the osmotic pressure of ASR gels pore solution on their swelling pressure is also discussed and studied. Chapter three is a more in depth study on the rheological, physical and chemical properties of ASR gels and their stoichiometry for a much broader range of compositions including K/Si, Li/Si and water content as variables besides Ca/Si and K/Si. This chapter discusses how different chemical components influence the gelation time, yield stress, phase composition (i.e., mass ratios of solid phase, pore water and chemisorbed water), and the stoichiometry of gelation reactions. Regression models have been developed that relate such parameters to the core chemical variables of ASR gels. Chapter four studies the effect of ASR gel composition (Ca/Si, Na/Si and K/Si) on the hydrophilic properties and free swelling capacity of ASR gels in contact with water vapor. The relative humidity at which the gels come to equilibrium with their closed-system surrounding environment was considered as a measure of their hydrophilic potential. ASR gel minibar were produced and their length change and mass change were monitored over time in order to determine their swelling capacity under free setting. The effects and interactions of all variables were statistically evaluated and regression models were developed that are able to adequately predict 16

35 the free swelling strain and the equilibrium relative humidity of ASR gels as a function of composition. Chapter five is a continuation of chapter four adopting the same design of experiments with a focus on the free swelling strain and restrained swelling pressure of ASR gel in contact with liquid water. ASR gel disks were produced and introduced to an innovative test setup that exposed the gel disks to hydrated cement pastes (on both sides) with a water to cement ratio similar to those of conventional concrete. The test units were exposed to water and the time-dependent swelling strains were recorded. Upon completion of the tests, the incurred strains were reversed back to zero in order to estimate the fully restrained swelling pressure of ASR gels as a function of composition. Statistical analyses were performed and regression models for the free swelling strain and swelling pressure of ASR gels were developed. The deleterious ASR gel compositions were then identified by detecting gel compositions (using the regression functions) that met certain criteria. Chapter six studies the mechanical and rheological properties of ASR gels with respect to their chemical compositions. The design of experiments used in chapter four and five was also used in this study. The compressive strength, Poisson s ratio and viscoelastic parameters of ASR gels were measured for the tested gels. Significantly, it was found out that ASR gels viscoelastic behavior can be best explained by the Burger s model. The short-term and log-term elastic moduli and viscosities of the gels were determined through a stress control test using MTS machines. Similar to the previous chapters, the mechanical and viscoelastic properties of the gels were regressed to their chemistry. 17

36 Chapter seven addresses a more practical and ready-to-implement topic; determination of the minimum fly ash dosage necessary to mitigate ASR in concrete made from reactive aggregates. In this chapter, a chemical index model is developed which is capable of determining such fly ash dosage by taking in the chemistry of fly ash, portland cement, the reactivity of aggregates (represented by their 1-year concrete prism test expansion in the absence of fly ash), the significance of the concrete infrastructure and the level of acceptable risk. The developed model eliminates the time and cost associated with the two-year trial and error tests for experimental determination of the optimum fly ash dosage. An online webpage has also been published for the public use in this regard REFERENCES [1] Poole, A.B. (1992). 1 Introduction to alkali-aggregate reaction in concrete. The Alkali Silica Reaction in Concrete, Swamy R.N. (Ed.)., Van Nostrand Reinhold, New York. [2] Hobbs, D. W. (1988). Alkali-silica reaction in concrete. Thomas Telford, London, [3] Ludwig, U. (1980). Durability of cement mortars and concretes. In Durability of Building Materials and Components: Proceedings of the First International Conference: a Symposium Presented at Ottawa, Canada, Aug (No. 691, p. 269). ASTM International. [4] Ueda, T., Baba, Y., & Nanasawa, A. (2013). Penetration of lithium into ASR-affected concrete due to electro-osmosis of lithium carbonate solution. Construction and Building Materials, 39, [5] Powers, T. C., & Steinour, H. H. (1955). An Interpretation of Some Published Researches on the Alkali-Aggregate Reaction Part 1-The Chemical Reactions and Mechanism of Expansion. Journal of the American Concrete Institute, 26(6), [6] Hou, X., Struble, L. J., & Kirkpatrick, R. J. (2004). Formation of ASR gel and the roles of CSH and portlandite. Cement and Concrete research, 34(9), [7] Poyet, S., Sellier, A., Capra, B., Thèvenin-Foray, G., Torrenti, J. M., Tournier-Cognon, H., & Bourdarot, E. (2006). Influence of water on alkali-silica reaction: experimental study and numerical simulations. Journal of Materials in civil Engineering, 18(4),

37 [8] Nishibayashi, S., Okada, K., Kawamura, M., Kobayashi, K., Kojima, T., Miyagawa, T., Nakano, K., and Ono, K. (1992). 10 Alkali-silica reaction Japanese experience. The Alkali-Silica Reaction in Concrete, 270. Swamy R.N. (Ed.)., Van Nostrand Reinhold, New York. [9] Struble, L. J., & Diamond, S. (1981). Swelling properties of synthetic alkali silica gels. Journal of the American ceramic society, 64(11), [10] Diamond, S. (1976). A review of alkali-silica reaction and expansion mechanisms 2. Reactive aggregates. Cement and Concrete Research, 6(4), [11] Dent Glasser, L. S., & Kataoka, N. (1981). The chemistry of alkali-aggregate reaction. Cement and Concrete Research, 11(1), 1-9. [12] Krogh, H. (1975). Examination of synthetic alkali-silica gels. In Symposium on Alkaliaggregate Reaction, Preventive Measures, Reykjavik, Iceland. [13] Shelby, J. E. (2005). Introduction to glass science and technology. Royal Society of Chemistry. [14] Iler, R. K. (1979). The chemistry of silica: solubility, polymerization, colloid and surface pro perties, and biochemistry. Wiley. [15] Hench, L. L., & Clark, D. E. (1978). Physical chemistry of glass surfaces.journal of Non- Crystalline Solids, 28(1), [16] Clark, D. E., & Yen-Bower, E. L. (1980). Corrosion of glass surfaces. Surface Science, 100(1), [17] Molchanov, V. S., & Prikhidko, N. E. (1957). Corrosion of silicate glasses by alkaline solutions. Bulletin of the Academy of Sciences of the USSR, Division of chemical science, 6(10), [18] H. Schloze (1982). Chemical durability of glasses. Journal of Non-Crystalline Solids, 52(1-3), [19] Sjöberg, S. (1996). Silica in aqueous environments. Journal of Non-Crystalline Solids, 196, [20] Bunker, B. C. (1994). Molecular mechanisms for corrosion of silica and silicate glasses. Journal of Non-Crystalline Solids, 179, [21] Walther, J. V., & Helgeson, H. C. (1977). Calculation of the thermodynamic properties of aqueous silica and the solubility of quartz and its polymorphs at high pressures and temperatures. American Journal of Science, 277(10), [22] Buckley, A.M., & Greenblatt, M, (1994). The sol-gel preparation of silica gels. Journal of chemical education 71(7), 599. [23] Sims, I., & Nixon, P. (2003). RILEM recommended test method AAR-1: detection of potential alkali-reactivity of aggregates petrographic method.materials and Structures, 36(7), [24] Hobbs, D.W. (2002). Alkali-silica reaction in concrete. In Structure and performance of cements, Barnes, P., & Bensted, J. CRC Press. 19

38 [25] Bleszynski, R. F., & Thomas, M. D. (1998). Microstructural studies of alkali-silica reaction in fly ash concrete immersed in alkaline solutions. Advanced Cement Based Materials, 7(2), [26] Shafaatian, S. M., Akhavan, A., Maraghechi, H., & Rajabipour, F. (2013). How does fly ash mitigate alkali silica reaction (ASR) in accelerated mortar bar test (ASTM C1567)?. Cement and Concrete Composites, 37, [27] Monteiro, P. J. M., Wang, K., Sposito, G., Dos Santos, M. C., & de Andrade, W. P. (1997). Influence of mineral admixtures on the alkali-aggregate reaction.cement and Concrete Research, 27(12), [28] Ichikawa, T., & Miura, M. (2007). Modified model of alkali-silica reaction.cement and Concrete research, 37(9), [29] Sims, I. (1992). 5 Alkali-silica reaction UK experience. The Alkali-Silica Reaction in Concrete, 122. Swamy R.N. (Ed.)., Van Nostrand Reinhold, New York. [30] Maraghechi, H., Shafaatian, S. M. H., Fischer, G., & Rajabipour, F. (2012). The role of residual cracks on alkali silica reactivity of recycled glass aggregates.cement and Concrete Composites, 34(1), [31] Kawamura, M., & Iwahori, K. (2004). ASR gel composition and expansive pressure in mortars under restraint. Cement and concrete composites, 26(1), [32] Lane, D.S. (1994), Alkali-silica reactivity in Virginia. Virginia Transportation Research Council, Charlottesville, Virginia, USA. [33] Lukschová, Š., Přikryl, R., & Pertold, Z. (2009). Petrographic identification of alkali silica reactive aggregates in concrete from 20th century bridges.construction and building materials, 23(2), [34] Maraghechi, H., Maraghechi, M., Rajabipour, F., & Pantano, C. (2014). Pozzolanic Reactivity of Recycled Glass Powder at Elevated Temperatures: Reaction Stoichiometry, Reaction Products and Effect of Alkali Activation.Cement and Concrete Composites. [35] Rodrigues, F. A., Monteiro, P. J., & Sposito, G. (2001). The alkali silica reaction: the effect of monovalent and bivalent cations on the surface charge of opal. Cement and Concrete Research, 31(11), [36] Prezzi, M., Monteiro, P. J., & Sposito, G. (1997). Alkali-Silica Reaction, Part I: Use of the Double-Layer Theory to Explain the Behavior of Reaction-Product Gels. ACI Materials Journal, 94(1). [37] Prezzi, M., Monteiro, P. J., & Sposito, G. (1998). Alkali-silica reaction; Part 2: the effect of chemical additives. ACI Materials Journal, 95(1). [38] Kunitz, M. (1928). Syneresis and Swelling of gelatin. The journal of general physiology, 12(2), [39] Struble, L., & Diamond, S. (1981). Unstable swelling behaviour of alkali silica gels. Cement and concrete research, 11(4),

39 [40] Berra, M., Faggiani, G., Mangialardi, T., & Paolini, A. E. (2010). Influence of stress restraint on the expansive behaviour of concrete affected by alkali-silica reaction. Cement and Concrete Research, 40(9), [41] Hong, S. Y., & Glasser, F. P. (2002). Alkali sorption by CSH and CASH gels: Part II. Role of alumina. Cement and Concrete Research, 32(7), [42] Gholizadeh Vayghan, A. G., Rajabipour, F., & Rosenberger, J. L. (2016). Composition rheology relationships in alkali silica reaction gels and the impact on the Gel's deleterious behavior. Cement and Concrete Research, 83,

40 CHAPTER 2: COMPOSITION-RHEOLOGY RELATIONSHIPS IN ALKALI-SILICA REACTION GELS AND THE IMPACT ON THE GEL S DELETERIOUS BEHAVIOR 2 Alkali-silica reaction (ASR) continues to be a major challenge to the durability of concrete structures. This is in part because the relationships between the composition, properties, and behavior of ASR gels in concrete are poorly understood. Gels with high pore solution ph, osmotic pressure, and rheological (e.g., yield stress) and swelling properties are the most deleterious. In this paper, the effects of the composition (primarily Ca/Si and Na/Si) of synthetic ASR gels on these characteristics are investigated, and regression analyses are done on the data. The pessimum combination of osmotic and rheological properties was found in the case of gels having intermediate calcium and high sodium contents (i.e., Ca/Si=0.2 and Na/Si 0.85), leading to the highest estimated swelling pressures. These gels also developed the most alkaline pore solutions. While highest yield stresses were observed in the gels with low calcium and low sodium, they showed negligible osmotic and swelling pressures BACKGROUND 2 This chapter was published in the Cement and Concrete Research journal Volume 83 in May 2016 (pp 45 56). 22

41 ASR gel is the product of alkali-silica reaction, and is the expansive phase which is responsible for cracking and damage in concrete. The quantity and nature of the formed gels are in direct relation to the magnitude of the expansive pressures and the resulting damage that will be generated inside concrete during the course of ASR [1]. As such, successful understanding of the reaction mechanisms and mitigation of ASR relies, in part, on advancing our knowledge of ASR gels. There is evidence [2] suggesting that mere formation of ASR gel does not always translate to damage to concrete, and gels of different compositions could behave very differently (i.e., deleteriously or innocuously). For example, it is suggested that while presence of calcium is a prerequisite for gel formation [3, 4, 5], gels with low Ca and high alkali (Na, K) contents could act as a flowable liquid and permeate through the pore structure of the surrounding cement without causing damage [2]. On the opposite side the spectrum, gels with high Ca content (i.e., Ca/Si>0.5 molar ratio) would start to approach the composition and properties of pozzoalnic C-S-H, and could have high stiffness and low expansion [6, 7, 8]. Gels with intermediate Ca content are likely to be both expansive and capable of generating and sustaining high stress levels that cause damage to concrete. Despite these circumstantial evidences and hypotheses, the quantitative relationships between ASR gel composition and its properties have not been systematically studied. This is the subject of the present research THE CURRENT STATE OF KNOWLEDGE 23

42 ASR gels form within the cracks or on the surface of reactive siliceous aggregates. As a result of the OH - ions attack, the metastable silicates in aggregates are depolymerized and dissolved into the concrete pore solution, and the products of this reaction form alkali - alkaline earth silicate hydrogels. ASR gels have a general composition of (SiO2).(Na2O)n.(K2O)k.(CaO)c.(H2O)x, and can be written using cement chemistry notation as N-C-S-H, when N represent alkali oxides. In addition, a small concentration of Mg may be present as a substitution for Ca. Figure 2-1 proposes the schematic molecular structure of a typical ASR gel, which is constructed based on the information from Refs [9, 10], and also the laboratory results on the chemical analysis of the gels pore solution. The model shows two small N-C-S-H clusters surrounded by the gel s pore solution. The concentration of the alkali and alkaline earth metals in field ASR gels typically fall into the range of (Na+K)/Si=0.1~1.2 and (Ca+Mg)/Si=0.0~0.2 (molar ratios) [11], although much higher concentrations of calcium were reported by Šachlová et al. [12]. They conducted a comparative study on a large number of field and laboratory gels, and concluded that age of the gel and composition of the cement paste can considerably affect the morphology and composition of the gels. Struble and Diamond [13] investigated the swelling properties of some sodium-silica and sodium-calcium-silica gels. They concluded that the composition of the gels considerably alters their swelling behavior, and certain gels were found to be innocuous in terms of swelling. However, the free swelling capacity results exhibited large variations (e.g. 0.5% to 81%) even for the gels with almost similar compositions. Due to lack of replication, no solid conclusions with respect to the contribution of calcium, sodium and initial moisture content of the gels on the swelling properties could be drawn. 24

43 Figure 2-1. Schematic molecular model of a typical alkali-silica reaction gel Hou et al. [11] used NMR to investigate the structural properties of synthetic and field ASR gels, and concluded that calcium and sodium reduce the degree of silica polymerization in the gels. The molecular and nanoscale structures of synthetic and field gels were found to be similar, suggesting that synthetic ASR gels can be studied effectively in order to understand the chemical and physical behavior of field gels. Nevertheless, there have been very few attempts towards systematically studying synthetic ASR gels as reliable representatives of field ASR gels. There is a substantial lack of knowledge with regards to the impact of the elemental composition of ASR gels on their key properties that control their deleterious expansion in concrete. The damage induced by ASR gels to the concrete is not only controlled by the hydrophilic potential and free swelling capacity of the gel, but also by how much confining pressure the gel can resist before flowing into adjacent capillary pores and microcracks of the surrounding paste. Such flow would relieve the gel s internal pressure that is responsible for damaging concrete. In other words, gels that are highly hydrophilic and expansive, and at the same time resist flow (due 25

44 to high yield stress as discussed below) are the most deleterious. This latter point underlines the significance of rheological properties of ASR gels in determining their performance inside concrete. It is usually presumed that viscosity is the most and perhaps the only important rheological property of ASR gels. Viscosity, which is associated with the flow of the liquids, is the internal resistance of a moving fluid to flow. Therefore, it only becomes a factor when the liquid is flowing. Many computational models developed to simulate the kinetics and mechanisms of ASR expansion [14, 15, 16] assume that ASR gel is a Newtonian fluid and can flow into the surrounding cement paste s pore structure immediately upon formation. This, however, may not be correct given that there is evidence that ASR gel tends to show a non-newtonian yield stress behavior (i.e., resistance to flow by showing elastic behavior until the stress exceeds a certain limit). There are several examples of SEM images showing that ASR gels are often capable of exerting excessive expansion and cracking to their surrounding concrete without flowing into an open microcrack (see Figure 2-2). Therefore, it is reasonable to consider yield stress as a more important rheological property as it prevents a gel from flowing into adjacent openings, and represents the maximum internal pressure that can be generated by the gel. After the gel starts to flow, its internal pressure decreases due to the additional volume that has become available to the gel (ASR gel s bulk modulus has not been reported but is likely to be higher than that of water; 2.2 GPa [17]). The gel s viscosity determines the rate of flow, which is also dependent on the permeability of the surrounding paste and the residual internal pressure of the gel. As the flow progresses, the gel s internal pressure continues to decrease to fractions of the initial yield strength. It should be noted that based on this argument although yield stress of ASR gels is likely to be more important than 26

45 their viscosity, the two parameters are probably highly correlated; i.e., gels with high yield stress also have a high viscosity. Figure 2-2. Example of an expanded ASR gel that has not yet penetrated into the surrounding vacancies, likely because the gel s yield stress has not been exceeded The above discussion can be better represented by Figure 2-3. The hydrophilic potential of ASR gel (e.g., its osmotic pressure and electrical double-layer repulsive forces) generates moisture absorption, restrained expansion, and a hydrostatic stress built up inside the gel [18]. This hydrostatic stress transforms to shear stress at the gel-vacancy interface (note that Figure 2-3 assumes that the gel forms at the aggregates circumference, although a similar argument can be made for the intra-aggregate gels). The gel s yield stress, however, prevents it from flowing into the vacancies. The yield stress is related to the gel composition and moisture content. The more moisture that is absorbed by the gel, the higher shear stress is generated at the interface. Also, as the water absorption continues, the gel becomes more dilute and its yield stress decreases. The increasing shear stress will eventually exceed the decreasing yield stress (unless the net hydrophilic 27

46 potential of the gel is satisfied prior to flowing). At this point the gel starts to flow through the vacancies. Figure 2-3. Schematic demonstration of the formation of confining pressure and the seepage of the gel into adjacent vacancies Monovalent (i.e., Na and K) and divalent (i.e., Ca and Mg) cations are the main network modifiers of ASR gels that contribute to their rheological properties. Understanding the sole and combined effects of these elements on the yield stress of the gels can be useful in resolving part of the ambiguities about the behavior of ASR gels in concrete. In addition to rheology, the composition of ASR gel s pore solution is also an important factor [6, 19], since it determines their osmotic pressure. This is the negative pressure (suction) in the gel s pore solution that causes imbibition of water and exertion of pressure to the surrounding cement paste. It should be noted that although the gel s pore solution is in contact with cement paste s pore solution, their compositions are not necessarily the same at all times due to the kinetics effects. The ph of the gel s pore solution is also important because it is this solution that is in direct contact with the reactive aggregate (at reaction sites) and therefore controls the rate of ASR [20]. 28

47 In this study, we will investigate the potential effects and interactions of calcium and sodium on the yield stress, pore solution ph, composition and osmotic pressure of synthetic ASR gels. The effects of potassium and magnesium on the gels properties were not separately investigated in this study. Potassium is reported to yield similar gel structures as sodium [11]. Calcium constitutes the majority of the divalent cations in the gels, and magnesium is only present in trace quantities. As such, only sodium and calcium were studied as the monovalent and divalent cations in the gels. In order to better trace the role of Ca and Na on the evolution of the studied properties, time dependent measurements are carried out to study gels from early ages up to four weeks old EXPERIMENTAL METHODS Gel synthesis process A sol-gel method similar to what was first proposed by Krogh [8] was adopted (with certain modifications) to synthesize the ASR gels. Aqueous sodium silicate solution was passed through an ion exchange column to achieve silicic acid (i.e. Si(OH)4(aq)), which was sodium-doped and condensed to a certain concentration. The resulting sol was then batched with adequate dosages of sodium and calcium hydroxide solutions to furnish the desired final concentration and composition. The detailed description of the gel synthesis process is as follows. Sodium silicate solutions with a SiO2/Na2O mass ratio of 3.18 (i.e. Na/Si atomic ratio of 0.61) with a 37.5% solid concentration (wt.%) were diluted to 10.3% and passed through a column of H + -saturated cation exchange resins to obtain aqueous silicic acid with an estimated concentration of 8%. The cation exchange resins had styrene-divinylbenzene matrix with sulfonic acid functional group. Effluents with concentrations exceeding 8% were not workable due to rapid 29

48 gelation before any further steps could be taken. The systematic drop in the solutions concentration from 10.3 to 8 wt% is due to replacement of Na + with H +, which translates to reduction in the mass of solutes. The ion exchange approach taken by Pramer [21] was adopted in this research. A cation exchange unit was set up in the lab capable of exchanging 7.70 moles of Na + with H + (i.e. ~12 liters of 10.3% sodium silicate solution) at a maximum rate of 30 ml/min. Figure 2-4 shows a diagram of the designed ion exchange setup. The first one liter of effluent in each run of ion exchange was discarded due unstable ph readings. After the ph of the effluent was stabilized at , the silicic acid was collected at a rate of ml/min. After collecting the silicic acid, it was doped with a sodium hydroxide solution of the same concentration to furnish a Na/Si molar ratio of 0.2. The sodium doping mainly served to impede the undesired rapid gelation of the silicic acid effluent. The Na-doped sol (i.e. aqueous silicic acid) was condensed down to ~25% in a vacuum Erlenmeyer flask at 60±2 C. Albeit desirable, higher concentrations (i.e., greater than 25%) were not attainable due to gelation. After cooling down, the condensed sol was proportioned with aqueous NaOH and Ca(OH)2 solution containing sufficient amounts of sodium and calcium hydroxides and also water to reach the desired chemical composition while maintaining the 25% concentration. It should be noted that the gels forming in the field probably have higher concentrations. However, reaching higher concentrations in the sol-gel method is extremely difficult and perhaps infeasible due to gelation of the condensed sol before batching. The sols were mixed inside 60-mL polyethylene cylindrical bottles using a highspeed homogenizer (10,000 rpm turbulent mixing) for up to 5 seconds, after which they were stored in the lab at 23±2 C without further disturbance to allow them to gel. The obtained gels could be dried out after formation, as practiced by several researchers [11, 13, 8, 22]. However, 30

49 any disturbance to the structural integrity of the gels (by drying) past the gel point would invalidate the rheological measurement results. Figure D diagram of the ion exchange unit Rheological measurements The yield stress can be measured using a number of different techniques, often categorized as direct or indirect methods. The most common indirect method involves measurement of the shear stress in the material under different shear strain rates and extrapolating the results back to zero shear rate to estimate the yield stress [23]. This method is often criticized due to problems such as slip flow, fracture and expulsion of the sample [23]. The ASR gels do not behave like regular fluids and show brittle behavior under large deformations. Therefore, capturing multiple points at different shear rates and plotting the shear stress-shear strain rate curves is not feasible. There are three direct methods for yield stress measurements; creep/recovery, stress relaxation and stress 31

50 growth [23]. The latter was adopted in this study, due to its more straightforward concept and reliable results. In this method, the material is sheared at a low and constant strain rate, and the resisting shear stress is monitored as a function of time. The yield stress is considered as the maximum shear stress observed on the shear stress time series plot [24] (e.g., see Figure 2-6). The shear stress can be applied to the sample using a variety of different methods (e.g., vane, slotted plate, penetrometer and inclined plane) [24]. The vane spindle method is advantageous in a number of ways. It imposes minimum disturbance upon immersion of the vane into the sample, which is essential for yield stress measurements in thixotropic materials [23, 24, 25]. Yoon and Mohtar [26] investigated the effect of disturbance on the yield stress of bentonite suspensions, in which they referred to the results of vane spindle rheometry as the undisturbed yield stress. Figure 2-5 shows the geometry of vane rheometry and the radial shear stress formed around the spindle. Figure 2-5. The vane rheometry setup and the path of shear stress formed in the material under torsion: (a) the schematic diagrams, (b) the actual setup (the spindle is not completely inserted for the sake of demonstration) 32

51 Note that some shear stress is also created on the sheared surfaces at the top and the bottom of the spindle. The rheometers that are designed for yield stress measurements usually report the maximum amount of torque (Mmax) applied to the sample, from which the yield stress (τy) can be calculated using Eq. 2-1 [24]. τ y = M max πd 3 2 (L D ) Eq. 2-1 where D and L denote the diameter and length of the spindle, respectively. A Brookfield DV3T rheometer (calibrated by the manufacturer) with V-74 spindle (D=5.89 mm and L=11.76 mm) was used for these experiments. The torque was applied to the samples at a constant shear rate as low as 0.5 rpm. Although the gels yielded within the first minute, the test was continued up to five minutes to monitor the residual shear stresses in the gels, which is usually referred to as the equilibrium stress (τe) [27] that is calculated by substituting Mmax with the equilibrium torque (Me) in Eq Me was determined by averaging all of the readings past a point after which the maximum variations in the torque readings did not exceed 20% of the average readings after that point. Figure 2-6 shows the shear stress time series for two gels (C - N - and C + N + as denoted in Table 2-1) at the age of 7 days. Two readings of the yield and equilibrium stresses were made at the top and bottom of each cylindrical gel sample. The average of the two readings is reported as the result for one replicate. 33

52 Figure 2-6. Shear stress time series plots for two gels at the age of 7 days Gelation time measurements The gelation time is the time at which the sol converts to gel via spatial cross-linking of a sufficient number of silicate clusters [10]. Although measurement of the gelation time has little practical implementation in evaluating the behavior of ASR gels in concrete, determining the impacts of sodium and calcium on this parameter is of merit, because it unveils the role of each component in the formation of ASR gels. The gelation time measurements were performed on one replicate of each gel composition using the vane rheometry. The gel point is defined as the instance where the sol starts to show growing resistance against the rotation of the spindle inside the sample, manifested by exponential growth in the viscosity of the sols [28]. A typical viscosity-time plot is shown in Figure 2-7. The gelation time (TG) is indicated in the graph, which for sols with long gelation times is to some extent subject to variations due to involvement of experimenter s judgement. Shear stress-time or torque-time plots may as well be used for determining TG. 34

53 Figure 2-7. Typical viscosity development time series of the alkali silicate gels used for determining the gelation time Gel pore solution ph and composition measurements Two of the replicates that were made for each gel were randomly selected for pore solution ph measurements. The chemical composition of only one replicate of the gels pore solution was analyzed via inductive coupled plasma atomic emission spectroscopy (ICP-AES) using Perkin- Elmer Optima 5300 UV instrument. The objective of ph measurements was to determine the alkalinity of the gels pore solution and, therefore, their corrosiveness for reactive aggregates on a comparative basis. The rationale behind monitoring the composition of the pore solution was to track their chemical evolution with respect to calcium, sodium and silicon in order to better understand the kinetics of gelation reactions. The pore solutions of the gel samples were extracted using a pressurized filtration setup as shown in Figure 2-8. The amount of pressure required to extract the pore solutions depended mainly on the composition and age of the gels, but usually ranged from 4 to 10 MPa. After the pore solutions were collected, their ph was measured using a ph meter (Fisher Scientific accumet model 25) followed by the ICP-AES measurements. 35

54 Figure 2-8. Gel pore solution extraction setup The sum of the molar concentrations of the solutes in the gels pore solution (i.e. Ca 2+, Na +, OH -, H3SiO4 - and H2SiO4-2 ) is plugged into the Morse equation (Eq. 2-2) for estimating the osmotic pressure generated by each gel. The Van t Hoff s factor is considered to be 2 in this equation, which assumes complete dissociation of sodium hydroxide, calcium hydroxide and sodium silicate species. This is a conservative assumption for sodium silicates and leads to slight overestimations of osmotic pressure. However, due to lack of thermodynamic values for sodium silicate species in aqueous solutions, this assumption was inevitable. Π = M i RT Eq. 2-2 where: 36

55 Π: the gel s pore solution osmotic pressure in MPa (assuming an ideal solution) Mi: the molarity of the i th ionic species R= kj.k 1.mol 1 : the gas constant T: the absolute temperature in Kelvin Note that the osmotic pressure of the gels is not the only driving force for water imbibition. The hydrophilic potential of the gels solid skeleton (mostly O -, O - Na +, and O - Ca 2+ O - endings) also contribute to the swelling pressure of the gels, but they are beyond the scope of this paper Gel composition measurements At the age of 28 days, one replicate of the gels was selected at random for chemical analysis using ICP-AES. The objective of the analyses was to compare the batched versus actual compositions of the gels. Ten grams of each gel specimen, after being tested for yield stress, were sampled and dissolved in 100 ml of 1M HCl acid solution for 24 hours. The majority of the alkaline part of the gel along with a portion of the silicates was dissolved in the acid. The undissolved precipitates were filtered and dissolved in 100 ml of 10M KOH solutions for another 24 hours. 5-mL samples were separately taken from the filtered acid and base solutions for ICP-AES analyses. The concentrations of Ca, Na and Si in each sample (i.e. acidic and basic) were measured and summed in order to obtain the total concentration of each element in the gels. The concentrations were converted to molar ratios of Na/Si and Ca/Si to compare with the batched values Experimental design Two levels of molar ratios (denoted as and +) were selected for Ca/Si and Na/Si considering the commonly observed compositions of the field ASR gels. Table 2-1 shows the actual and coded 37

56 levels and the four different possible combinations of the gels compositions. Two to three replicates of each composition were synthesized and tested at 1 hour, 1 day, 7 days and 28 days. Table 2-1. The chemical composition and coding of the gels (the levels are the molar ratios) Ca/Si Na/Si Levels 0.05, , 0.8 Coded levels -, + -, + Gel code Ca/Si Na/Si C - N (-) 0.20 (-) C + N (+) 0.20 (-) C - N (-) 0.80 (+) C + N (+) 0.80 (+) The adopted 2 2 factorial design with replications at four different levels of age allows for monitoring the linear effects of Na/Si and Ca/Si as well as the non-linear effect of age, and all possible interactions of these parameters on the desired response parameters (i.e. yield stress, equilibrium stress, ph and osmotic pressure). After obtaining the data, analysis of variance (ANOVA) was performed on the results with respect to each response, in order to determine which factors (i.e. Ca/Si, Na/Si, and age), have a statistically significant influence on the response parameters. The significance of statistical analysis is to ensure the repeatability of the experiments and the reliability of the conclusions RESULTS AND DISCUSSIONS Rheological and gel point measurement results The yield and equilibrium stress results are shown in Figure 2-9 and Figure The error bars represent the 95% confidence intervals for the average results (i.e., the intervals into which the true average values of the measured parameter fall with 95% confidence). It is observed that both 38

57 age and composition significantly affect the yield stress and equilibrium stress of the gels. C - N - (i.e., the gels with low Ca/Si and Na/Si atomic ratios) specimens had translucent and relatively coarse granular appearance. They showed little to no yield behavior at first (i.e., 1 hour), but then turned into very rigid gels with growing structural properties at later ages. By the age of 28 days, C - N - specimens were extremely rigid gels with the highest yield and equilibrium stress properties compared to the other gels. This could be due to the fact that C - N - gels have considerably more silica and therefore more Si-O-Si bonds compared to other gels, which led to higher mechanical strength. The mode of failure for C - N - specimens at later ages was entirely brittle and the term yield stress might be better replaced with shear strength for them. C + N -, on the other hand, showed rapid strength development at the first few minutes. White rigid gels with very fine granular appearances and highest yield and equilibrium stress values were obtained within the first hour. However, there was a significant drop in the rheological and structural properties of these gels from the first day onwards, followed by some recovery at later ages. Therefore, in the case of low sodium (i.e., N - ) gels, an increase in calcium content results in tremendously lower long-term yield and equilibrium stress values, while leading to greater initial rheological properties compared to low calcium gels. Ca/Si showed an opposite effect on the rheological properties of N + gels (i.e., an increase in Ca/Si led to higher yield and equilibrium stress values for these gels). This can be substantiated by comparing C - N + and C + N + gels at all ages. C - N + showed no yield stress behavior especially at the first hour, because of its slow gelation. Except for this reading, there was no variation in the rheological properties of N + gels over time, especially in the case of C + N +. However, there was a slight decline in the 28-day results for both gels. Both gels had completely homogeneous appearances with translucent white (C - N + ) and opaque white (C + N + ) colors. 39

58 Figure 2-9. Yield stress development of the gels as a function of time Figure Equilibrium stress development of the gels as a function of time 40

59 All gels preserved a portion of their yield stress after failure (i.e., they had measurable equilibrium stress). The magnitude of the shear stress retention factor (defined as the equilibrium to yield stress ratio: τe/τy) was a function of time and more significantly, the composition of the gels. The highest τe/τy ratios were found in the case of C - N - gels at later ages (i.e., 0.38 at the age of 28 days). C + N -, on the other hand, exhibited the lowest τe/τy ratios (i.e., 0.07) at the same age. The residual yield stress of the gel after the original yield point is an important characteristic in terms of the swelling pressure that it can produce after penetrating into the cement paste s pore network. The gels also showed different tendencies towards gelation. Figure 2-11 shows the shear stress-time plot for C - N -, C - N + and C + N +. As mentioned before, torque and shear stress can alternatively be used instead of viscosity in detecting the gel point. C + N - could not be tested for its gelation time due to extremely rapid gelation. It gelled up in less than 3 seconds upon addition of Ca(OH)2 solution. The rest of the sols reached their gel points in a matter of few minutes to more than an hour. The gelation time can be used to identify the influence of different chemical components on the kinetics of gelation reactions. Since formation of gel is a prerequisite to ASR damage, the chemical components promoting gelation can be considered deleterious in terms of ASR. As the graph suggests, an increase in the calcium content results in a tremendous drop in the gelation time, possibly due to rapid reaction of calcium and silicates as well as promoting crosslinking of the silicate species. This can be substantiated by comparing C - N - (TG=204 sec) to C + N - (TG<3 sec), or C - N + (TG=110 min) to C + N + (TG<100 sec). Comparing C - N - to C - N +, and also C + N - to C + N + suggests that the presence of sodium significantly delays gelation, which may be due to occupation of the Si-O - endings of silicate species with the monovalent Na + cations that 41

60 prevents them from crosslinking and forming the gel network. The presence of a peak in the shear stress time series of C - gels is due to the breakage of the gel network due to the continuous application of shear stresses to the gels before gelation, which is an indication of the slower rate of microstructural development (repair) of the C - gels compared to the rate of application of damage (torsion). However, in the case of C + gels, the gel s network develops faster than damage is imposed. Figure Shear stress-time response of the gels at early stages of gelation Gel pore solution ph and composition measurement results Figure 2-12 shows the ph variations of the gels pore solution over the studied time span. The error bars represent the 95% confidence intervals for the average ph readings. It is observed that generally, an increase in Ca/Si and especially Na/Si increases the ph. In the case of N - gels (i.e., low Na/Si), an increase in the Ca/Si does not have much impact on the ph. In the case of N + gels, however, an increase in calcium causes significant increase in the ph. Calcium and sodium seem 42

61 to have a synergistic effect on the ph, meaning that when sodium and calcium are both present at high concentrations, the alkalinity of the pore solution is considerably amplified. Comparing C - N + and C + N + at the age of 28 days confirms that high concentrations of sodium along with relatively high concentrations of calcium increases the ph beyond the level attainable by a saturated Ca(OH)2 solution (i.e., 12.45). This could be attributed to the alkali-recycling effect of calcium in ASR gels [29], but is also caused by the fact that C + N + gels have the lowest silica content of the four gels and silica is the component that promotes acidity while calcium and sodium promote alkalinity. Comparing the ph values obtained for the gels pore solutions with what is common ph values in cement paste s pore solution (i.e., 13~13.5), one finds out lower values in the case of ASR gels. The main reason for this difference is the acidic nature of silica [6]. The dissolved silica contributed by reactive aggregates deprotonate and can act as a buffer, preventing the ph of pore solution from reaching high values. In the cement pore solution, on the other hand, dissolved silica is depleted, due to abundance of calcium, which reacts with silica and form C-S-H. Presence of solid portlandite maintains the ph above 12.45, and alkali sulfates in collaboration with portlandite are able to boost the ph to higher values. The regression analyses of the ph results show that for an N + gel with a Ca/Si of 0.3 at the age of 28 days, the estimated (extrapolated) ph value is with a 95% prediction interval of ( ). Moreover, the ph values increase by age (as the regression model suggests) and also concentration of the gel. Therefore, under certain circumstances, highly alkaline gel pore solutions might be observed. 43

62 Figure ph variations of the gels pore solution as a function of time The mass concentrations of Ca, Na and Si species as well as the estimated osmotic pressure of the gels pore solution are presented in Figure Except for C + N - up to the first week, time (age) has shown no significant effect on any of the results. The calcium contents of the pore solutions are all minimal and almost identical for all different gels after the first week. In the case of Na and Si, however, the composition of the gels affects their concentrations in pore solution of the gels. For both C + and C - gels, the higher the Na/Si, the higher the Na and Si species concentrations in the pore solutions. This is probably due to occupation of Si-O - endings of dissolved silicate species by the sodium cations that prevents them from incorporating into the gels structures, thus retaining them in the pore solution. Na/Si has similar influence on the osmotic pressure. An increase in the Ca/Si content, as long as there is low sodium in the gel (i.e. N - ), makes no difference in the Na, Si and osmotic pressure. Ca/Si showed a reducing effect on 44

63 these parameters in the case of N + gels. Excessive osmotic pressures exceeding the tensile strength of concrete seem to exist in C - N + and C + N + gels pore solution. Although there are other sources of hydrophilic behavior in the gels as discussed before, it should be noted that in actual concrete mixtures, part of this pressure is evened out by the osmotic pressure of the surrounding cement paste pore solution in contact with the gels. Figure Time-dependent variations of solutes concentration and osmotic pressure in the gels pore solutions Gel composition measurement results The ICP-AES results of the gels chemical composition is shown in Table 2-2. The calcium to silicon atmic ratios were almost exactly as batched. However, the measured sodium to silica ratios were in most cases greater than the batched values. The only uncontrollable source of potential 45

64 variations in the composition of the gels is the composition of the silicic acid sol that is supposed to have undergone complete ion exchange. Therefore, the sodium-doped effluent of the ion exchange column, which should theoretically have a Na/Si of 0.20, was analyzed using ICP-AES technique to find out if there was any extra sodium in the sol. Any difference between the actual Na/Si ratio of the sol and the designed value (i.e. 0.20) stems from the incomplete ion exchange. The analysis of the chemical composition of the sol suggests a Na/Si ratio of This explains part of the variations in the actual Na/Si ratios of the gels compared to the batched values. Apparently, the Na + ions are not completely exchanged with H + ions, due to the relatively high affinity between sodium and silicate species; that is a trace amount of sodium content will remain in the silicate species. This property of sodium silicates should be taken into account when synthesizing gels via sol-gel methods. Table 2-2. The batched and analyzed composition and of the gels Gel Label Batched Ca/Si Analyzed Ca/Si Batched Na/Si a Analyzed Na/Si C - N C + N C - N C + N a The values are reported as the total Na/Si present in the gels accounting for the residual sodium from the ion exchange step Statistical analyses In order to confirm the conclusions drawn with regards to the effects of Na/Si, Ca/Si and Age on the rheological, chemical and osmotic properties of the ASR gels and their pore solutions, an analysis of variance (ANOVA) was performed on the data. The factors (i.e., variables) were Ca/Si, Na/Si and Age. The response parameters were the yield stress (τy), equilibrium stress (τe), ph and osmotic pressure (Π). The ANOVA was conducted using the Minitab software. The main output 46

65 of ANOVA includes testing whether a given parameter has a significant influence on the response, whether the effect is positive or negative, and which parameters interact and in what fashion. Table 2-3 shows the summarized ANOVA output for the rheological properties, ph and osmotic pressure. For each response parameter with respect to each input parameter, three outputs are reported: the coefficient of the input parameter (as it appears in the regression equation of the corresponding response parameter), the p-value of the input parameter s effect, and whether it is significant. The p-value is the probability of finding a more extreme value of the statistic under the assumption of the null hypothesis (i.e., no effect). In simpler terms, the p-value is the probability of mistakenly deeming an input parameter significant when it is actually not. The significance level chosen for these analyses is 0.05, which is a common value and implies that any parameter with a p-value greater than 0.05 is deemed insignificant (i.e., does not meaningfully affect the response). With respect to the yield and equilibrium stresses, it is observed that all interaction terms (i.e., Ca/Si Na/Si, Ca/Si Age, Na/Si Age and Ca/Si Na/Si Age) are highly significant. Ca/Si alone does not appear to be significant, while Na/Si seems to meaningfully decrease the yield stress. This is consistent with the findings of Kawamura and Iwahori [2] on the effect of sodium on the flowability of ASR gels, where they concluded that the alkali-rich gels are highly flowable. The interaction between the Ca/Si and Na/Si has the largest coefficient, indicating that the influence of each of these parameters depends on the level of the other parameter. In other words, no single statement can necessarily be made regarding the sole effects of these parameters on the rheological properties of ASR gels. At high levels of Na/Si, an increase in Ca/Si leads to higher rheological properties, while an opposite effect is observed for Ca/Si in low levels of Na/Si. The same conclusion can be reached for the effect of Na/Si on the rheological properties of the gels. 47

66 Table 2-3. The ANOVA table of the chemistry of ASR gels with respect to yield stress, (τy), equilibrium stress (τe), ph and osmotic pressure (Π) 48

67 The effect of interaction between Ca/Si and Na/Si is shown in Figure 2-14 a and b. Note that the yield stress results shown on these plots are the average values across different levels of age. The large confidence interval of C - N - gels is due to significant strength development of these gels over time. Similar plots are obtained for equilibrium stress. These conclusions, however, may not necessarily be always the case, and are only true for the studied ranges of Ca/Si and Na/Si. (a) (b) Figure The interaction plots of Ca/Si and Na/Si on the yield stress results; a) the effect of Ca/Si, b) the effect of Na/Si With respect to pore solution ph, the calcium content alone (i.e. Ca/Si) does not seem to be significant (p-value=0.85). At low levels of Na/Si, an increase in Ca/Si causes no change in the ph of the gels pore solution, while it shows an increasing effect on the ph at high levels of Na/Si (Figure 2-15a). In other words, it is the interaction between Ca/Si and Na/Si (i.e. Ca/Si Na/Si) that contributes to ph. This is what can be referred to as the synergistic effect of calcium and sodium on the ph of the ASR gels pore solution, as previously discussed. Sodium alone (i.e., Na/Si) seems to be increasing the ph especially at high levels of calcium content (Figure 2-15b). The pore solution osmotic pressure seems to be affected by all the parameters except for age. Note that a positive and significant coefficient for Ca/Si does not necessarily translate to 49

68 increasing effect of Ca/Si on the osmotic pressure, because Ca/Si has a strong interaction with Na/Si and Age. In other words, Ca/Si appears in the regression function of the osmotic pressure as the following: ( Na/Si 0.324Age Na/Si Age) Ca/Si. The coefficient of Ca/Si in this term is always negative except for N - gels up to the age of two weeks. Na/Si, on the other hand, shows an increasing effect on the osmotic pressure regardless of Age and Ca/Si. As the regression functions suggest, the increasing effect of Na/Si on the osmotic pressure of the gels pore solution seems to grow as the gels age. (a) (b) Figure The interaction plot of Ca/Si and Na/Si on the ph of the pore solution of the gels; a) the effect of Ca/Si, b) the effect of Na/Si More discussion on the swelling of the gels and the role of rheology As explained before, the confining stress (induced due to restrained expansion of ASR gel) transforms to shear stresses at the gel-vacancy interface. This shear stress is tolerated by the gel up to its yield point. Generally speaking, if the gel experiences a volumetric strain of εvg, the amount of confining stress exerted by the surrounding cement matrix will be σc=k.εvg, where K is the bulk modulus of the gel, assuming that the expansion is initially fully restrained. Due to the stress applied to the gel at the interface, the gel will slightly lean towards the vacancy 50

69 (deflection=x; see Figure 2-3). The amount of deflection is related to the amount of confining stress and the diameter of the cylindrical capillary pore (d). As long as the deflection of the gel is below its maximum shear deformation (measured for each gel during the yield stress experiments), the gel resists yielding. The confining stress, which is assumed to be equal to the hydrostatic stress built up in the gel, can be derived as shown in Eq σ c = 4τ. x d Eq. 2-3 where: σc: confining stress (MPa) τ: Shear stress in the gel at the interface (MPa) x: deflection of the gel (mm) d: capillary pore diameter (mm) The confining stress that a gel can tolerate before yielding can be estimated using Eq σ c,max = 4τ y. x max d Eq. 2-4 For different gels studied in this paper, the xmax is measured during the yield stress experiments. The vacancy aperture, which could be either a cylindrical capillary pore or a microcrack, has a statistical distribution specific to the cement paste used. Therefore, d is a random variable and so are τy and xmax. Therefore, σc,max is also a random variable. Generally, the maximum capillary pore size of the surrounding cement paste determines the maximum swelling stress a gel can tolerate without flowing. For a typical cement paste, the large capillary pore size is around 10 μm [30]. This value is adopted for further calculations. 51

70 It should be noted that a high σc,max does not necessarily translate to a high swelling pressure. σc,max is the amount of pressure that a gel can tolerate without flowing, but such pressure will not be generated if the gel does not exhibit expansive behavior. In other words, the swelling pressure that is generated inside a gel is controlled by the osmotic pressure of the gel s pore solution and is limited by its σc,max. Therefore, the swelling pressure σs can be defined as min{πlong term, σc,max}; note that this ignores the contribution of the gels solid part (i.e., the double layer repulsive forces, and the hydrophilic potential generated by the charged endings such as O -, O - Na +, etc.) in the swelling pressure. The average swelling pressure of the gels at the age of 28 days is calculated using the osmotic pressure, yield stress and maximum deflection results of the gels at this age. The results are shown in Table 2-4, suggesting that the C + N + gels are able, by far, to exert the highest swelling pressures. Therefore, the simultaneous presence of sodium and calcium, as concluded for the pore solution ph, leads to the most deteriorative characteristics. Table 2-4. The estimated swelling pressures of the gels at the age of 28 days Gel Label Πlong term (MPa) τy (MPa) xmax (mm) σc (MPa) σs (MPa) C - N E C + N E C - N E C + N E These calculations are obviously estimative and further research is required to more accurately calculate and/or measure the swelling pressure of the gels as a function of their composition. Nevertheless, the swelling pressure generated in the gel induces tensile stresses in the surrounding concrete matrix and can result in cracking [31]. Such tensile stress can be estimated as a function of the swelling stress generated by the gel, as well as the aggregate radius and the volume fraction of the reactive particle (see work of Nielsen et al. [32] as an example). 52

71 It also needs to be pointed out that the composition of ASR gel in concrete changes gradually as it ages, and especially when it moves away from its formation point (e.g., inside or in contact with aggregate) and into concrete matrix. In concrete structures (e.g., pavements) that have access to moisture but are not in constant contact with liquid water, gel expansion occurs slowly. The gel forms, expands and eventually flows into surrounding cracks or other vacancies, often after inducing some damage to concrete. As the gel flows into the concrete matrix, it starts to calcify as a result of exposure to portlandite, where the alkali ions in the gel are replaced with Ca (this phenomenon has been named alkali recycling [29]). This calcification stiffens the gel and makes it more resistant to further flow, while it is still slowly absorbing water and generating swelling stress. As such, the concrete could experience more damage when moisture absorption is slow, allowing ASR gel to be calcified. Moreover, the alkali-recycling increases the ph of concrete s pore solution, resulting in further ASR. In contrast, where concrete structures are in constant contact with liquid water (e.g., dams, sea walls), ASR gel expansion occurs more rapidly. The gel flows into cracks and vacancies and will continue to absorb water, be diluted, and flow further probably before considerable calcification. Alkali recycling occurs much less and the ph of concrete pore solution decreases as a result of alkali leaching into the body of water. Overall, less ASR damage may occur in such structures. Our unpublished data based on ASTM C1293 concrete prism test supports this argument. We have compared the expansion and moisture uptake of otherwise identical concrete specimens that were wrapped with a vapor permeable membrane versus not-wrapped specimens. The vapor permeable membrane allows maintaining nearly 100% relative humidity (RH) inside concrete prisms, but it blocks access to liquid water that otherwise condensates on the surface of concrete prisms. The results show that the wrapped concrete specimens (i.e., stricter access to moisture) expand slower with time, but at the same level of 53

72 moisture uptake, they expand significantly more than unwrapped specimens. In other words, specimens that expand slower due to stricter moisture access, end up expanding considerably more in the long run. Clearly, further research is needed to better understand the effect of moisture availability on the rate and magnitude of ASR expansion and damage in concrete. The present research has taken a few positive steps towards understanding how the composition of ASR gels affects their rheological and swelling properties as well as alkalinity and osmotic pressure. These parameters are some of the characteristics that govern the expansive behavior of ASR gels in concrete. However, further research on a wider range of chemical composition, focusing on how it affects the actual swelling capacity, swelling pressure and hydrophilic potential of the ASR gels is required CONCLUSIONS The yield stress is an important rheological property of ASR gels as it largely determines the maximum confining pressure that a gel can tolerate before flowing into crevices of the surrounding cement paste. The yield stress of ASR gels is substantially affected by the gel s chemical composition (i.e., Ca/Si and Na/Si) and age. The highest yield stress was observed for gels with low Ca/Si and low Na/Si (since this composition constitutes a high purity silica gel with a high Q4 coordination). The second highest yield stress was observed for gels with moderately high Ca/Si and high Na/Si. The gelation time of the studied ASR gels increased with an increase in Ca/Si, or a decrease in Na/Si ratios. 54

73 Ca/Si and Na/Si have a synergistic effect on the alkalinity of the ASR gels pore solutions. Simultaneous increase of Ca/Si and Na/Si gave rise to ph levels beyond what the sole increase of calcium or sodium in the ASR gels could produce. Calcium showed no significant influence on the osmotic pressure of the gels with low Na/Si, while it reduced that of the gels with high Na/Si, which is favorable in reducing potential ASR damage. Sodium, on the other hand, showed an increasing effect on the osmotic pressure of the gels especially at lower levels of Ca/Si. The swelling pressure generated by ASR gel can be estimated as the minimum of its osmotic pressure and yield stress. The highest swelling pressures, thus estimated, were found in the case of gels with moderately high calcium and high sodium contents REFERENCES [1] Haha, M. B., Gallucci, E., Guidoum, A., & Scrivener, K. L. (2007). Relation of expansion due to alkali silica reaction to the degree of reaction measured by SEM image analysis. Cement and Concrete Research, 37(8), [2] Kawamura, M., & Iwahori, K. (2004). ASR gel composition and expansive pressure in mortars under restraint. Cement and concrete composites, 26(1), [3] Diamond, S. (1989, August). ASR-another look at mechanisms. In Proceedings of the 8th International Conference on Alkali-Aggregate Reaction, Kyoto, Japan (pp ). [4] Iler, R. K. The chemistry of silica: solubility, polymerization, colloid and surface properties, and biochemistry Canada: John Wiley &Sons Inc. [5] Glasser, F.P., (1992). Chemistry of alkali-aggregate reaction, in: The Alkali Silica Reaction in Concrete, R.N. Swamy Edt., Van Nostrand Reinhold, New York. 55

74 [6] Powers, T. C., & Steinour, H. H. (1955). An Interpretation of Some Published Researches on the Alkali-Aggregate Reaction Part 1-The Chemical Reactions and Mechanism of Expansion. Journal of the American Concrete Institute, 26(6), [7] Monteiro, P. J. M., Wang, K., Sposito, G., Dos Santos, M. C., & de Andrade, W. P. (1997). Influence of mineral admixtures on the alkali-aggregate reaction.cement and Concrete Research, 27(12), [8] Krogh, H. (1975). Examination of synthetic alkali-silica gels. In Symposium on Alkaliaggregate Reaction, Preventive Measures, Reykjavik, Iceland. [9] Rajabipour, F., Giannini, E., Dunant, C., Ideker, J. H., & Thomas, M. D. (2015). Alkali silica reaction: Current understanding of the reaction mechanisms and the knowledge gaps. Cement and Concrete Research, 76, [10] Buckley, A.M., Greenblatt, M. (1994). The sol-gel preparation of silica gels. Journal of chemical education, 71(7), 599. [11] Hou, X., Kirkpatrick, R. J., Struble, L. J., & Monteiro, P. J. (2005). Structural investigations of alkali silicate gels. Journal of the American Ceramic Society, 88(4), [12] Šachlová, Š., Přikryl, R., & Pertold, Z. (2010). Alkali-silica reaction products: Comparison between samples from concrete structures and laboratory test specimens. Materials Characterization, 61(12), [13] Struble, L. J., & Diamond, S. (1981). Swelling properties of synthetic alkali silica gels. Journal of the American ceramic society, 64(11), [14] Multon, S., Sellier, A., & Cyr, M. (2009). Chemo mechanical modeling for prediction of alkali silica reaction (ASR) expansion. Cement and Concrete Research, 39(6), [15] Giorla, A. B., Scrivener, K. L., & Dunant, C. F. (2015). Influence of visco-elasticity on the stress development induced by alkali silica reaction. Cement and Concrete Research, 70, 1-8. [16] Xu, W. (2013). Calculation of alkali silica reaction (ASR) induced expansion before cracking of concrete. Journal of Wuhan University of Technology-Mater. Sci. Ed., 28(1), [17] Halliday, D. R. (1997). R. & WALKER, J. Fundamentals of Physics extended. John Wiley & Sons. INC. New York. 56

75 [18] Baz ant, Z. P., & Steffens, A. (2000). Mathematical model for kinetics of alkali silica reaction in concrete. Cement and Concrete Research, 30(3), [19] Hudec, P. P., & Banahene, N. K. (1993). Chemical treatments and additives for controlling alkali reactivity. Cement and Concrete Composites, 15(1), [20] Molchanov, V. S., & Prikhidko, N. E. (1957). Corrosion of silicate glasses by alkaline solutions. Bulletin of the Academy of Sciences of the USSR, Division of chemical science, 6(10), [21] Pramer, D. (1957). The influence of physical and chemical factors on the preparation of silica gel media. Applied microbiology, 5(6), 392. [22] Mostavi, E., Asadi, S., Hassan, M. M., & Alansari, M. (2015). Evaluation of self-healing mechanisms in concrete with double-walled sodium silicate microcapsules. Journal of Materials in Civil Engineering, 27(12), [23] Liddel, P. V., & Boger, D. V. (1996). Yield stress measurements with the vane. Journal of non-newtonian fluid mechanics, 63(2), [24] Nguyen, Q. D., Akroyd, T., De Kee, D. C., & Zhu, L. (2006). Yield stress measurements in suspensions: an inter-laboratory study. Korea-Australia Rheology Journal, 18(1), [25] Gunasekaran, S., & Ak, M. M. (2002). Cheese rheology and texture (p. 73). CRC press. [26] Yoon, J., & El Mohtar, C. (2013). Disturbance Effect on Time-Dependent Yield Stress Measurement of Bentonite Suspensions. Geotechnical Testing Journal,36(1), [27] Barnes, H. A., & Nguyen, Q. D. (2001). Rotating vane rheometry a review.journal of Non- Newtonian Fluid Mechanics, 98(1), [28] Nitta, S. V., Jain, A., Wayner Jr, P. C., Gill, W. N., & Plawsky, J. L. (1999). Effect of sol rheology on the uniformity of spin-on silica xerogel films. Journal of applied physics, 86(10), [29] Thomas, M. (2001). The role of calcium hydroxide in alkali recycling in concrete. Materials Science of Concrete Special, [30] Mindess, S., Young, J. F., & Darwin, D. (2003). Concrete (p. 75). Prentice Hall. 57

76 [31] Multon, S., Sellier, A., & Cyr, M. (2009). Chemo mechanical modeling for prediction of alkali silica reaction (ASR) expansion. Cement and Concrete Research, 39(6), [32] Nielsen, A., Gottfredsen, F., & Thøgersen, F. (1993). Development of stresses in concrete structures with alkali-silica reactions. Materials and structures,26(3),

77 CHAPTER 3: CHARACTERIZATION OF THE RHEOLOGICAL AND CHEMICAL PROPERTIES OF ALKALI SILICA REACTION (ASR) GELS CONTAINING LITHIUM 3 Synthetic ASR gels incorporating silicon, calcium, sodium, potassium and lithium with broad ranges of composition were synthesized using a sol gel method. The rheological and chemical properties of the gels (primarily, the gelation time, yield stress, gelation reactions, and phase composition) were characterized as a function of their chemical composition. The objective of this research, besides characterizing the chemistry of ASR gels, was to find out which chemical components promote gelation and yield stress of the gels, which is a governing factor in terms of deleteriousness of ASR gels. The results suggest that calcium is the predominant factor promoting the gelation and yield stress of the gels. Moreover, lithium was found to reduce the hydroxide concentration in the gels pore solutions by a factor of seven and the yield stress by a factor of two, which partly explains the suppressing effect of this element on ASR in concrete. 3 This chapter will soon be submitted to the Cement and Concrete Research journal for publication. 59

78 3.1. BACKGROUND ASR damage is the result of reaction between portland cement alkali hydroxides and the meta stable silicate in aggregates, followed by formation and swelling of a hygroscopic alkali silicate gel either within the aggregate micro cracks or exterior surfaces. ASR is a major concrete durability issue that continues to damage outdoor concrete structures. Despite decades of investigation on ASR, the nature and behavior of ASR gels remain poorly understood. The scientific community has yet to conclusively determine (1) how the concrete proportions and the compositions of reactive aggregates and cementitious materials impact the composition of ASR gel that forms inside concrete, (2) how such gel compositions evolve over time as ASR gel matures inside reactive aggregates or inside the cement paste matrix, and (3) how ASR gel composition determines its rheological and swelling properties and deleterious behavior inside concrete. The present study focuses on the third knowledge gap with a deep emphasis on the chemistry of ASR gels, and investigates the influence of lithium on ASR gels rheology, chemistry and swelling potential. The rheological and chemical properties of synthetic ASR gel can together explain the swelling behavior of ASR gels in a restraining porous environment such as concrete. Figure 3-1a is a simplified view of the formation of ASR gel around a reactive aggregate, which upon water imbibition encounters confining stresses from the surrounding cement paste [1]. The gel is in contact with a permeable media (hydrated cement paste) with random micro cracks in its vicinity. As shown in Figure 3-1b, the confining stresses create a hydrostatic pressure in the gel s bulk, driving the gel into the immediate vacancies. The gel can either resist the confining stresses, or flow into these vacancies dissipating the generated swelling pressure. ASR gels are comprised of two distinct phases; a layered solid structure, and the pore solution. The mechanical resistance 60

79 against flowing into the adjacent vacancies is provided by the solid structure. The pore solution, which can bear no shear stress, is the main source of the gel s tendency toward water imbibition through the act of osmosis and the Gibbs Donnan effect. The hydrophilic nature of the solid surfaces also contributes to the gel s hygroscopic nature. The otherwise high surface charge density of the solid surfaces promotes the adsorption of water molecules (usually referred to as the physically absorbed or physisorbed water) by creating Van der Waals bonds. Once the hydrophilic potential of the solid surfaces is satisfied by adsorbing one or few layers of water molecules (see the molecular structure of the layered or sheet like solid structure of the gels in Figure 3-1d, the pore solution will be the main source of the gel s tendency for water ingress. The level of interconnectivity of the solid layers determines the magnitude of the gel s resistance against yielding under shear forces at the vacancies; a rheological property referred to as the yield stress (τy). The yield stress is the amount of shear stress that the gel can tolerate before yielding (i.e., starting to flow). The relation between the confining stress (σc) and shear stress generated in the gel at the vacancy was demonstrated in a previous work to be: σ c = 4τ. x d Eq. 3-1 where: σc: confining stress (MPa) τ: Shear stress in the gel at the interface (MPa) x: deflection of the gel (mm), see Figure 3-1b. d: capillary pore diameter (mm) 61

80 (a) (b) (c) (d) Figure 3-1. Schematic illustration of the swelling of ASR gels in concrete, their microstructural and molecular configurations (a) the formation and expansion of ASR gel around the aggregates and the confining pressure from the surrounding cement paste due to the gel expansion; (b) a zoomed in view of the gel near a cement micro crack, the role of water ingress on the formation of confining stress and the hydrostatic pressure driving the gel into the adjacent vacancy, and the role of shear yield stress; (c) a higher magnification zoon in view of the gel microstructure illustrating the different phases in the gels; (d) the molecular structure of the gels showing the sheet like solid structure and the alkaline pore solution of the gel The confining stress that a gel can tolerate before yielding can be estimated as follows: σ c,max = 4τ y. x max d Eq. 3-2 Studying the stoichiometry of gelation reactions helps us in determining phase composition (i.e., mass ratios of pore solution and solid phases) and understanding how and to what extent do 62

81 the different elements affect the macroscopic properties of ASR gels. This is very useful in unraveling the complicated behavior of such systems, which ultimately enable us in engineering the composition of ASR gels and converting them into less deleterious ones (e.g., via addition of chemical admixtures). Lithium is one such potential admixture that is investigated in this paper to see its effects on the gels chemical and rheological properties. The objective of this paper, then, is to investigate the relationships between the batched composition of lithium added ASR gels and their rheological properties (i.e., gelation time and yield stress), phase composition, and their pore solutions and solid parts chemical compositions. The main test variables in this research are the batched Ca/Si, Na/Si, K/Si, and Li/Si atomic ratios as well as the batched water content of the gels EXPERIMENTAL PROGRAM Gels with various chemical composition (i.e., Ca/Si, Na/Si, K/Si, Li/Si and water content) were synthesized and tested to determine their gelation time, yield stress, pore solution composition and ph, phase composition, and stoichiometry of gelation reactions Materials Reagent grade calcium hydroxide powder, sodium and potassium hydroxide pellets, and lithium hydroxide monohydrate (LiOH.H2O) granules were used as raw materials in synthesizing ASR gels. For the source of silicon, colloidal silicon dioxide IV (50% H2O) with an average particle size of 20 nm was used. It should be noted that all of the above source materials contained certain 63

82 percentages of chemically bound water (24.3%, 22.5%, 16.0%, 64.4% and 50% water, respectively), which were accounted for in calculation of the water content of the batched gels Design of experiments (DOE) Five gel composition (input) variables (i.e., Ca/Si, Na/Si, K/Si, Li/Si and water content) were studied in the ranges shown in Table 3-2. Due to the highly varying chemical composition of ASR gels and the stated aim of this research as to understanding how composition affects their chemical and rheological properties, a careful design of experiments is needed. The designed experiments shall be capable of capturing the linear and non linear effects of the input variables, along with their interactions, in the entire studied ranges of the input variables, with a reasonable number of tests. Table 3-1. The studied ranges of the ASR gel composition (input) variables Variable Note Lower bound Upper bound C: Ca/Si Atomic ratio N: Na/Si Atomic ratio K: K/Si Atomic ratio L: Li/Si Atomic ratio W: water content wt./wt. of gel% Similar to a previous work [2], the concepts of response surface methodology [3, 4] were implemented in designing the experiments. The selection of experimental units was made using the central composite design (CCD) approach [3]. Five distinct levels were defined for each input variable; referred to as Low (L), Intermediate Low (IL), Intermediate (I), Intermediate High (IH), and High (H). These levels were calculated for each input variable and shown in Table 3-2, along with the coded levels. Please refer to [2] for more details regarding the implementation of the CCD concepts in design of similar experiments. 64

83 Coded levels Table 3-2. The coded and natural levels of the input variables (α=2) Intermediate Intermediate Intermediate Categorical Low (L) Low (IL) (I) High (IH) Numerical α α High (H) Test variables C o r r e s p o n d i n g n a t u r a l l e v e l s Ca/Si: C Na/Si: C K/Si: K Li/Si: L Water content: W CCD creates the experimental program by combining three types of experimental units: axial points, cube points and center points. While a normal factorial DOE needs hundreds of experiments to yields similar results, the adopted CCD requires only 32 as explained below. For 5 variables, 10 axial points are needed which associate with the experimental units in which one of the test variables is either at its high or low level while all others are at their intermediate levels. Regularly, 2 5 = 32 cube points (i.e., the ones in which all test variables are either at their IL or IH levels) are needed to fully capture the interactions of the test variables. However, = 16 points can be selected out of the 32 cube points in order to reduce the number of experiments without compromising the capabilities of the final statistical model. This will allow the estimation of up to the two way interaction effects of the test variables. Moreover, 6 center points (all variables at their intermediate levels) are assumed as replicates in order to have an orthogonal design. Hence, 32 experimental units have to be run. The test variables Ca/Si, Na/Si, K/Si, Li/Si and water content are denoted as C, N, K, L, W, respectively. The gels are labeled by assigning the corresponding coded categorical levels (i.e., L, IL, I, etc.) of the test variables as lowercase suffixes to them and concatenating the resultants into one label. For instance, a gel with intermediate low Ca/Si (i.e., 65

84 0.1375), intermediate high Na/Si, K/Si and Li/Si (i.e., 0.8 and and 0.375) and intermediate low water (i.e., 0.675) is labeled CILNIHKIHLIHWIL, which represents one of the cube points. Table 3-3 shows the full list of the gel compositions that were synthesized and tested in this research. The water contents of the gels are reported in both mass and molar ratios. Table 3-3. The experimental units representing the gel compositions investigated in this research Run order Gel label Ca/Si Na/Si K/Si Li/Si Water content (wt./wt.) (H 2 O)/Si Molar ratio 1 CIHNILKIHLIHWIL CIHNIHKILLILWIH CINIKILIWH CINIKILLWI CINIKLLIWI CIHNIHKIHLILWIL CILNIHKILLILWIL CINIKILIWI CILNIHKILLIHWIH CIHNILKILLIHWIH CIHNIHKIHLIHWIH CILNILKILLIHWIL CINIKILHWI CILNIHKIHLILWIH CIHNILKIHLILWIH CHNIKILIWI CILNILKIHLILWIL CINLKILIWI CINHKILIWI CIHNIHKILLIHWIL CLNIKILIWI CINIKILIWI CILNILKIHLIHWIH CINIKILIWI CILNIHKIHLIHWIL CILNILKILLILWIH CINIKILIWI CINIKILIWI CINIKHLIWI CINIKILIWL CIHNILKILLILWIL CINIKILIWI

85 The designed experiments enable us in developing regression models for each response parameter (e.g., gelation time, yield stress, pore solution solutes concentrations, etc.) that provide a polynomial relationship between the response parameter of interest and the test variables per Eq In this equation, y is the response parameter, ε is the prediction error, and Xi is the i th test variable (among the 5 variables). Xi Xj is the interaction term between the i th and j th variables. The statistical significance of each regression term (i.e., any term containing one or more variables in the regression models) is determined by testing the significance of their coefficients using a t test. The mean to standard error ratio of each coefficient is known to be a random variable with a t distribution. The probability of such ratio being zero (which translates to the insignificance of that coefficient) is denoted by P Value. A P Value smaller than 0.05 indicates that the corresponding term is significant. The regression model is constructed by compiling all significant and necessary terms into an equation similar to Eq Besides the regression analyses, the Pearson s correlation coefficients between the test variables (Ca/Si, Na/Si, etc.) and the response parameters (e.g., yield stress, chemically bound calcium), and also between different pairs of the response parameters are calculated to find out possible correlations, which help us in better understanding the chemistry of ASR gels. k k y = β 0 + β i x i + β ij x i x j i=1 i<j j=1 k 2 + β ii x i + ε i=1 Eq

86 Gel synthesis method Table 3-4 shows the masses of different components needed to be batched for rendering gels with the target compositions as listed in Table 3-3. A sol gel method was developed at Penn State and deployed for synthesizing the gels. The colloidal source of silicon dioxide as the main precursor of the gels contained colloidal grains with average particle size of 20 nm. Our studies suggest that such source of silica, if added with other components (i.e., Ca(OH)2, NaOH, etc.), will undergo flash gelation and convert to large clusters of silica without allowing for properly mixing the ingredients. The resultant will have a heterogeneous chemistry, very slow gelation, and inferior mechanical properties compared to the gels whose precursor is a fully dissolved silicic acid as prepared in a previous study [1]. One might consider using synthetic silicic acid as the precursor, but the targeted concentrations (as shown in Table 3-4) would be unattainable due to the rapid, unwanted gelation of the prepared silicic acid prior to the addition of the alkaline sources. Therefore, a protocol was developed for successful production of ASR gels at high concentrations using the selected source of colloidal silicon dioxide such that all components are properly introduced and incorporated to the gel s structure. The synthesis of ASR gels in this study is somewhat similar to what was adopted in Chapter two. The source of silica, however, was switched from silicic acid to colloidal silica, which primarily was for allowing the synthesis of gels at higher concentrations. It can be assumed that a sodium-doped silicic acid behaves more or less similar to a sodium-doped colloidal silica with a similar chemistry at equilibrium. Lithium hydroxide monohydrate (LiOH.H2O) was first dissolved in 20 grams of the distilled mixing water. NaOH and KOH pellets were then weighed and added to the lithium hydroxide solution, followed by sufficient mixing on a shaking table for complete dissolution of all alkalis and dissipation of the generated heat. The introduction of calcium hydroxide to the system was 68

87 decided to be carried out towards the end of the procedure, since due to its divalent nature and high field strength, it causes fast and permanent gelation upon addition to the silicate solutions. Table 3-4. The mixture proportions of the synthesized ASR gels (the reported masses are for each 1000 grams of the batched gel) LiOH.H 2 O Colloidal silica Gel label Ca(OH) 2 (gr) NaOH (gr) KOH (gr) (gr) (gr) CIHNILKIHLIHWIL CIHNIHKILLILWIH CINIKILIWH CINIKILLWI CINIKLLIWI CIHNIHKIHLILWIL CILNIHKILLILWIL CINIKILIWI CILNIHKILLIHWIH CIHNILKILLIHWIH CIHNIHKIHLIHWIH CILNILKILLIHWIL CINIKILHWI CILNIHKIHLILWIH CIHNILKIHLILWIH CHNIKILIWI CILNILKIHLILWIL CINLKILIWI CINHKILIWI CIHNIHKILLIHWIL CLNIKILIWI CINIKILIWI CILNILKIHLIHWIH CINIKILIWI CILNIHKIHLIHWIL CILNILKILLILWIH CINIKILIWI CINIKILIWI CINIKHLIWI CINIKILIWL CIHNILKILLILWIL CINIKILIWI H 2 O (gr) 69

88 The alkalis, on the other hand, are monovalent cations which usually do not cause permanent gelation. As such, the colloidal silica was weighed and added to the alkaline solution containing lithium, sodium and potassium. Upon contact, a rapid release of heat and temporary gelation occurs during which the system turns into a murky gel like material. However, if enough time is provided, the system goes back to the sol state under constant stirring on a shaking table. The resulting alkali silicate mixtures were stirred for five days for complete reversal of the incurred gelation and conversion of the formed large chains of alkali silicate complexes into monomers or short polymeric units. Despite its differences with the in-field ASR gels in terms of formation, synthesis of sound alkali-silicate gels called for dissociation of silicates prior to addition of other chemical components. Otherwise the obtained gels would have heterogeneous chemistries. The obtained sol in almost all cases appeared to be fully transparent. As the final step of procedure, the required amount of Ca(OH)2 and the remainder of the mixing water were to be added to the sol so that the target compositions are obtained and the permanent gelation takes place. As such, calcium hydroxide was dissolved in the remaining amount of mixing water, and the obtained limewater was added to the sol and the system was mixed using a hand held homogenizer at 7000 rpm for 3 5 sec to ensure complete blending of the two phases. The appearance of the resultant was found to range from light translucent to solid white instantly after the addition of limewater. The transparency of the mixture inversely correlated with the calcium content, while the alkalis tended to shift the gel s appearance towards translucent. For instance, solid white colors were found in the case of gels with low alkalis and high calcium, while the ones with opposite combination of alkalis and calcium appeared to be translucent. 70

89 Rheological measurements Gelation time measurements Gelation time is an approximate measure of the kinetics of reactions and it helps us understand how different components promote or delay formation of gels in concrete. Lithium (as a modifier) in particular can be investigated to see whether its ASR suppressing effect [5,6] has to do with how it alters the gelation behavior of ASR gels. The gels were tested for their gelation time upon completion of the synthesis process. The gelation time (or the gel point) is the time at which the sol (for the most part) coverts to gel through spatial cross linking of a sufficient number of silicate clusters [7]. On a macroscopic level, the gel point is considered to be the point at which the sol starts to show growing resistance against shear deformations. As such, the gel s viscosity starts to rapidly increase around the gel point, the rate of which depends on the chemistry and solid concentration of the gel. A Brookfield DV3 T rheometer with a V 73 spindle (D=12.67 mm and L=25.35 mm) was used for monitoring the evolution of viscosity in the gels. Figure 3-2 shows a typical viscosity time series of the alkali silicate sols (as they undergo gelation) and how the gelation time is estimated. The plastic bottles containing the sols were mounted to the rheometer and the spindle was inserted to the sol and started for rotating at 0.5 rpm. The torsional moment generated by the sol was monitored over time and the gelation time was determined as shown in Figure 3-2. The instrument was calibrated by the manufacturer prior to use in the experiments, and the repeatability of the measurements were assessed by the 6 replicates (per Table 3-3). 71

90 Figure 3-2. Typical viscosity development time series of the alkali silicate sols during the gelation process used for determining the gelation time Yield stress measurement The yield stress of ASR gels is a measure of their resistance against flow into the pores and micro cracks of the surrounding hydrated cement paste [1]. ASR gels with low yield stresses, once experience confining pressure from their surrounding as they tend to swell, easily flow into such vicinities without sustaining detrimental swelling pressures [1,8]. In other words, highly alkaline gels might have high tendency towards swelling, but they do not have the capability of causing damage. However, ASR gels with high yield stress can damage their surrounding given they have sufficient tendency towards swelling. The yield stress of the gels were measured and compared in order to see the effect of each chemical component and lithium in particular on this parameter. Due to its high reliability and straightforwardness, the stress growth concept was used for testing the gels for their yield stress [9]. In this approach, the shear stress is applied and increased in the gels at a constant (slow) rate and the maximum tolerated stress is considered to be the yield stress of the material. Such shear 72

91 stress can be applied to the media via different methods, and the vane geometry (adopted here) is known to be the most reliable one, due to its straightforward concept and minimum disturbance that it imposes on the media [10,11]. Therefore, same instrument used for the measurement of the gelation time of the specimens was used for yield stress measurements. The spindle used for applying the shear stress was either V 73 (for gels with low yield stress) or V 74 (D=5.89 mm and L=11.76 mm) otherwise. Please refer to [1] and [10] for detailed information about this experiment. In the case of certain gels with yield stresses greater than 40,000 Pa, the required torsional moment to successfully test the gels using the rheometer exceeded beyond its capacity. In the case of those gels, the V 74 spindle was attached to a high precision torque driver model TD24 (THORLABS, N.mm) via a manufactured connector. The yield stress of those gels were measured using such torque meter via inserting the spindle into the gel and slowly applying torque by hand. For both methods, six to eight measurements were taken from each gel specimen and the results were found to have very high reliabilities in both cases (R 2 >0.98) Chemical characterization techniques The chemical characteristics of the gels investigated in this research are the alkalinity (i.e., ph) and chemical composition of their pore solutions and solid phases (i.e., the chemically bound minerals and water, which together constitute the gels solid phases), the phase composition, and the evaporable and chemisorbed water of the gels. The pore solution extraction was performed by applying hydraulic compressive pressure on representative samples of the gels inside a pressurized stainless steel capsule equipped with a 73

92 multilayer filtering system. The extracted pore solutions were further filtered using 0.2 μm syringe filter discs. The ph of the gels pore solutions was measured by the acid titration method using a 0.1M HCl acid solution. The chemical composition of the pore solutions was determined using Perkin-Elmer Optima 5300 UV inductively coupled plasma atomic emission spectroscopy (ICP AES) technique on the extracted pore solutions of the gels. The detection limits of Ca, K, Li, Na and Si of the instrument were 0.01, 0.2, 0.01, 0.01, and 0.01 μg/ml, respectively. The evaporable water content of the gels was measured by merely drying gel samples inside a muffle furnace at 110 C for 48 hours. The mass ratio of the evaporable water was determined, and the remaining water (determined by subtracting the total admixed water from the evaporable water) was considered to be the chemisorbed water, which is chemically incorporated in the gels solid parts. The mass ratios of the gels pore solutions were calculated as the sum of the masses of the evaporable water and those of different solutes in the pore solutions. Moreover, the mass and chemical composition of the gels solid parts were calculated by subtracting those of their pore solutions from the gels bulk masses and chemical compositions. This enabled us in estimating the gelation reactions for each gel RESULTS AND DISCUSSION Gelation time and yield stress results Table 3-5 shows the gelation time (TG) and yield stress (τy) results of the 32 experimented gels. Notice how variations in composition results in drastic changes in these two parameters. The gelation time changes across four orders of magnitude (comparing rows 14 and 15 as the maximum and minimum observations), and the yield stress shows near two orders of magnitude variation 74

93 (comparing rows 16 and 21). In both cases, the effects of Ca/Si, Na/Si and K/Si are evident. Ca/Si seems to be significantly decreasing the gelation time and increasing the yield stress of the gels, while the opposite is found for Na/Si and K/Si. Table 3-5. The gelation time (TG) and yield stress (τy) results of the gels No. Gel label T G (min) τ y (Pa) No. Gel label T G (min) τ y (Pa) 1 CIHNILKIHLIHWIL CILNILKIHLILWIL CIHNIHKILLILWIH CINLKILIWI CINIKILIWH CINHKILIWI CINIKILLWI CIHNIHKILLIHWIL CINIKLLIWI CLNIKILIWI CIHNIHKIHLILWIL CINIKILIWI CILNIHKILLILWIL CILNILKIHLIHWIH CINIKILIWI CINIKILIWI CILNIHKILLIHWIH CILNIHKIHLIHWIL CIHNILKILLIHWIH CILNILKILLILWIH CIHNIHKIHLIHWIH CINIKILIWI CILNILKILLIHWIL CINIKILIWI CINIKILHWI CINIKHLIWI CILNIHKIHLILWIH CINIKILIWL CIHNILKIHLILWIH CIHNILKILLILWIL CHNIKILIWI CINIKILIWI Short gelation times were observed in the case of gels with intermediate to high Ca/Si and low Na/Si (and/or K/Si). Further, gels with such compositions were found to have more solid 75

94 texture and higher yield stresses. The opposite properties were observed in the case of gels with low Ca/Si and high such alkalis. The effects of lithium and water are more subtle and need statistical analyses. As such, regression functions per Eq. 3-3 were fit to the data to statistically evaluate the effects and interactions of each variable on these two rheological parameters. The regression function of gelation time is shown in Eq The regression coefficient of determination (R 2 ) was found to be 93.28%. Moreover, the prediction R 2 of the developed regression model, which represents the accuracy of the model in predicting new observations, is 87.47%. A Box Cox transformation of the response (λ = 1 /3) was needed in order to normalize the residuals of regression and stabilizing their variations across the fitted values (which are two core assumptions of regression that should not be violated). The terms included in the regression model are the significant predictors of gelation time. Other terms (e.g., W, (Ca/Si) 2, (Na/Si) (K/Si) etc.) did not prove significant. Two pairs of variables were found to have interactive effects on gelation time: Ca/Si and Na/Si, and also Ca/Si and Li/Si, for which interaction plots are provided. The regression model suggests that Ca/Si is by far the most important factor (P Value < ) promoting the gelation of the studied ASR gels (compare the coefficients), while Na/Si and K/Si showed delaying effects (P Values = and < , respectively). Although Na/Si seemed to have a P Value greater than the designated significance level (i.e., 0.05), it has a significant interaction with Ca/Si (P Value = 0.033) that justifies its inclusion in the model. As shown in Figure 3-3, the reducing effect of Ca/Si on TG is more evident at higher levels of Na/Si. The increasing effect of sodium on the gelation time of ASR gels is due to the fact that sodium as a monovalent cation makes bonds with the non-bridging oxygens preventing them from making molecular chains with the adjacent silicate molecules. Calcium, on the other hand, is a divalent cation with a high affinity with oxygen and bonds silicates together 76

95 thus promoting gelation. Due to its higher field strength, it can replace sodium at the O - Na + endings and form bonds with the adjacent silicates, which is why it has a more pronounced decreasing effect on gelation time of high-sodium gels. Similarly, Li/Si showed a marginally significant effect on gelation time (P Value = 0.074). However, it was found to have a significant interaction with Ca/Si (P Value = 0.028). (T G (min)) 1/3 = Ca Na Si Si K Si 5.30 Ca Si Li Si Li Ca Si Si Na Si Eq. 3-4 Figure 3-3. The interaction plot of Ca/Si and Na/Si on the gelation time 77

96 Figure 3-4 shows the interaction plot of Li/Si and Ca/Si. It is observed that at a low Ca/Si level (0.05), an increase in Li/Si leads to a drastic reduction in the gelation time, while it shows minor increasing effect on gelation time at high Ca/Si (0.4). Therefore, it can be concluded that lithium generally has an accelerating effect on the gelation of ASR gels. Interestingly, the water content did not prove significant in the studied range, which could be due to the following reason: the gelation is mostly controlled by the properties of the solid phase and it is not affected by the relatively small variations in water content (65% to 75%). Therefore, the studied range of water content was not wide enough to cause significant variations in the gelation time. Figure 3-4. The interaction plot of Li/Si and Ca/Si on the gelation time 78

97 Eq. 3-5 shows the regression model of the gels yield stress with respect to their chemical composition. It is observed that the effect of chemistry of the yield stress of the gels is more complicated compared to the gelation time, and there are many terms involved in the variations of this parameter. The regular and prediction coefficients of determination for the developed regression model were calculated as 96.46% and 73.28%. ln(τ y (Pa)) = Ca Na Si Si K Si ( K Si ) ( Li Si ) Ca Si Na Si 78.4 K Si Li Ca W ( Si Si )2 Li Ca W 41.8 W+262.4( Si Si )3 Eq. 3-5 While all test variables proved to be significant, the patterns of their influence on yields stress cannot be easily realized due to the presence of non linear terms and interactions. As such, the factorial plots of the average fitted yield stress versus different test variables were prepared and are shown in Figure 3-5. The factorial plots are obtained by plotting the fitted yield stress (per Eq. 3-5) as a function of each test variable while other variables are all at their intermediate levels. For instance, the factorial plot for Ca/Si is obtained by replacing Na/Si, K/Si, Li/Si and W with 0.6, 0.15, 0.25, and 0.7 respectively in the regression equation and plotting τy as a function of Ca/Si, which is the only remaining variable in the resulting function. It is observed that an increase in Ca/Si from 0.05 to 0.4 results in 78 times increase in the yield stress. The increasing effect of Ca/Si is more pronounced once it exceeds 0.3. Na/Si, on the other hand, has a drastic reducing effect on yield stress, causing a reduction in this parameter by a factor of 16 as it changes from 0.2 to 1.0. The effect of potassium appears to 79

98 be complicated. It reduces the yield stress by a factor of 3 as it changes from 0.0 to 0.18, but it leads to a 58% increase as it increases from 0.18 to 0.3. This pattern of behavior is not only suggested by the regression function, but it is also clearly observed in the raw data. Therefore, the impacts of K/Si and Na/Si as two alkali elements are not similar, which might have to do with their different atomic sizes and field strengths. Figure 3-5. The factorial plots of the average fitted yield stress versus different test variables 80

99 Li/Si has a more or less similar (but less intense) influence on the yields stress as K/Si. It causes 58% decrease in the yield stress as it changes from 0.0 to 0.33, followed by a 25% increase as it increases from 0.33 to 0.5. The promoting effect of Li/Si on the yield stress as it exceeds 0.33 probably has to do with its high field strength and tendency to bond adjacent alkali-silicate gel solid sheets. Finally, a monotonic reducing effect on the yield stress was found for the water content (at intermediate levels of other variables). Note that due to its interactions with K/Si and Li/Si (as suggested by the regression equation), its effect on the yield stress is affected by the levels of these two variables. Figure 3-6 shows the effect of water content on the yield stress fitted values (per Eq. 3-5)) at some combinations of high, low and intermediate levels of K/Si and Li/Si. It is observed that in most combinations of K/Si and Li/Si, water has a reducing effect on the yield stress of the gels (which is simply due to the dilution effect). In the case of K/Si = 0.0 and Li/Si = 0.25, as well as K/Si = 0.15 and Li/Si = 0.0, however, the yield stress appears to be increasing with increase in the water content. The interactive effect of water with these alkalis can be understood noting the fact that the water needed to complete the gelation reactions is greater at lower concentrations of alkalis. Therefore, in the two aforementioned cases of gels with low alkalis, the increase in water can probably result in more gelation reactions taking place, which will in turn lead to better structural development and higher yield stresses despite the dilution effect of water. It can be observed in Figure 3-6 that the slope of the variations in the yield stress is correlated by the alkali content; the more alkalis are present in the system, the more pronounced the reducing effect of water is. On the other hand, when the alkalis are scarce, the dilution effect of water is overshadowed by its chemical influence (in promoting the gelation reactions) and an increasing effect on the yield stress is found for water. 81

100 According to Eq. 3-5, Ca/Si and Na/Si also have an interactive effect on yield stress, which is plotted in Figure 3-7. It is observed that at levels of Na/Si (e.g., 0.2), Ca/Si has a generally increasing effect on yield stress with an episode of reducing effect in the range of 0.16 to However, the Ca/Si has a monotonic increasing effect on yield at high levels of Na/Si. Figure 3-6. The influence of water on ASR gels yield stress at different levels of K/Si and Li/Si 82

101 Figure 3-7. The interaction plot of Ca/Si and Na/Si on the yield stress Chemical measurements and analyses results The ICP chemical analyses results of the gels pore solutions are shown in Table 3-6 (under Step one ; the first step taken in analyzing the gels). These results reveal how much of the different species have remained in the dissolved state, as opposed to being chemically incorporated in the gels. As mentioned before, by measuring the free water content of the gels (in mass ratios as reported in Table 3-6 under Step two ), one can calculate the total mass and then the volume of the pore solution (by having the concentrations and densities of different solutes). Since the solutes concentrations are reported in moles per liter of the pore solution, the number of moles as 83

102 well as the masses of different solutes can be determined as shown in Table 3-6 under Step three for each mole of batched gel (e.g., SiO2.(CaO)0.225.(Na2O)0.3.(K2O)0.075.(Li2O) (H2O)13.2, which corresponds to CINIKILIWI). Finally, the compositions of the gels solid parts can be calculated by subtracting the molar compositions of the batched gels from the molar compositions of their pore solutions (as shown in Table 3-6 under Step four ). These calculations enable us in developing the gelation reactions of the gels based on the initial composition of the ingredients, which are presented in Eq. 3-6 to Eq. 3-36). It should be noted that these reactions, though obtained through analytical calculations, rely on experimental measurements and are, as such, subject to variations. Although good agreement is found among them, comparing the gelation reactions of the six replicates of CINIKILIWI (presented in Eq. 3-12, Eq. 3-26, Eq. 3-28, Eq. 3-31, Eq. 3-32, and Eq. 3-36) shows how much variations should be expected in other reactions. It is observed that the initial proportions of the ingredients radically affect both the amount (see Table 3-6) and the chemical composition (see the reactions) of the resulting gel solid phase and the pore solution. As suggested by Table 3-7, depending on the initial batching proportions, the gel s solid phase can account for 10.15% to as high as 43.51%. It can be readily seen that the lowest gel solid content corresponds to the gel with the lowest Ca/Si batched ratio. Also, the highest gel content is observed for the gels with intermediate high Ca/Si and intermediate low Na/Si, followed by the gel that has intermediate Ca/Si but low Na/Si. These suggest that the initial Ca/Si ratio might have a direct correlation with the amount of gel s solid content. The effects of the gel s batched composition on the phase proportion and composition of the reaction products are discussed in more details in the next section. 84

103 Table 3-6. The chemical measurements and analyses results. (CaO)i (Na 2O)i (K2O)i (Li2O)i (H2O)i [Ca 2+ ] [Na + ] [K + ] [Li + ] [SiTot] [OH ] FW a (wt%) PS b (wt%) SG c (wt%) Ca 2+ a: free water b: pore solution c: Gel solid part OH (CaO)g (Na 2O)g (K2O)g (Li2O)g (SiO2)g (H2O)g (H2O)g/(H2O)i Step one Step two Step three Step four Gel label Batched molar ratios Phase proportaion (mass) Pore solution composition (moles per Gel composition (molar ratios) moles of batched gel) Pore solution composition (moles per liter of pore solution) Na + K + Li + H 2SiO4 2 C IHNILKIHLIHWIL C IHNIHKILLILWIH C INIKILIWH C INIKILLWI C INIKLLIWI C IHNIHKIHLILWIL C ILNIHKILLILWIL C INIKILIWI C ILNIHKILLIHWIH C IHNILKILLIHWIH C IHNIHKIHLIHWIH C ILNILKILLIHWIL C INIKILHWI C ILNIHKIHLILWIH C IHNILKIHLILWIH C HNIKILIWI C ILNILKIHLILWIL C INLKILIWI C INHKILIWI C IHNIHKILLIHWIL C LNIKILIWI C INIKILIWI C ILNILKIHLIHWIH C INIKILIWI C ILNIHKIHLIHWIL C ILNILKILLILWIH C INIKILIWI C INIKILIWI C INIKHLIWI C INIKILIWL C IHNILKILLILWIL C INIKILIWI

104 Table 3-7. The list of gelation reactions of the studied alkali silicate systems C IH N IL K IH L IH W IL C IH N IH K IL L IL W IH C I N I KIL I W H C I N I K I L L W I C I N I K L L I W I C IH N IH K IH L IL W IL C IL N IH K IL L IL W IL C I N I K I L I W I C IL N IH K IL L IH W IH C IH N IL K IL L IH W IH C IH N IH K IH L IH W IH C IL N IL K IL L IH W IL Reaction not available (CaO)+0.4(Na 2 O) (K 2 O) (Li 2 O)+SiO (H 2 O) 0.502{(CaO) (Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.562OH 0.225(CaO)+0.3(Na 2 O)+0.075(K 2 O)+0.125(Li 2 O)+SiO 2 +17(H 2 O) 0.563{(CaO) (Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) 7.29 } Ca Na K Li H 2 SiO H 2 O+0.519OH 0.225(CaO)+0.3(Na 2 O)+0.075(K2O)+SiO (H2O) 0.632{(CaO) (Na 2 O) (K2O) 0.06.(SiO 2 ).(H2O) } Ca Na K Li H 2 SiO H 2 O+0.399OH 0.225(CaO)+0.3(Na 2 O)+0.125(Li 2 O)+SiO (H 2 O) 0.489{(CaO) 0.46.(Na 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.422OH (CaO)+0.4(Na 2 O) (K 2 O) (Li 2 O)+SiO (H 2 O) 0.87{(CaO) (Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.548OH (CaO)+0.4(Na 2 O) (K 2 O) (Li 2 O)+SiO (H 2 O) 0.6{(CaO) (Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.539OH 0.225(CaO)+0.3(Na 2 O)+0.075(K 2 O)+0.125(Li 2 O)+SiO (H 2 O) 0.749{(CaO) (Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.45OH (CaO)+0.4(Na 2 O) (K 2 O) (Li 2 O)+SiO (H 2 O) 0.413{(CaO) (Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.525OH (CaO)+0.2(Na 2 O) (K 2 O) (Li 2 O)+SiO (H 2 O) 0.741{(CaO) (Na 2 O) 0.16.(K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.266OH (CaO)+0.4(Na 2 O) (K 2 O) (Li 2 O)+SiO (H 2 O) 0.494{(CaO) (Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.378OH (CaO)+0.2(Na 2 O) (K 2 O) (Li 2 O)+SiO2+10.3(H 2 O) 0.644{(CaO) (Na 2 O) (K 2 O) (Li 2 O) 0.18.(SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.337OH Eq. 3-6 Eq. 3-7 Eq. 3-8 Eq. 3-9 Eq Eq Eq Eq Eq Eq Eq

105 C I N I K I L H W I C IL N IH K IH L IL W IH C IH N IL K IH L IL W IH C H N I K I L I W I C IL N IL K IH L IL W IL C I N L K I L I W I C I N H K I L I W I C IH N IH K IL L IH W IL C L N I K I L I W I C I N I K I L I W I C IL N IL K IH L IH W IH C I N I K I L I W I 0.225(CaO)+0.3(Na 2 O)+0.075(K 2 O)+0.25(Li 2 O)+SiO (H 2 O) 0.37{(CaO) (Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.487OH (CaO)+0.4(Na 2 O) (K 2 O) (Li 2 O)+SiO (H 2 O) 0.303{(CaO) (Na 2 O) (K 2 O) 0.21.(Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.58OH (CaO)+0.2(Na 2 O) (K 2 O) (Li 2 O)+SiO (H 2 O) 0.519{(CaO) (Na 2 O) 0.22.(K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.236OH 0.4(CaO)+0.3(Na 2 O)+0.075(K 2 O)+0.125(Li 2 O)+SiO (H 2 O) 0.404{(CaO) 0.99.(Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.444OH (CaO)+0.2(Na 2 O) (K 2 O) (Li 2 O)+SiO (H 2 O) 0.535((CaO) (Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) ) Ca Na K Li H 2 SiO H 2 O+0.353OH 0.225(CaO)+0.1(Na 2 O)+0.075(K 2 O)+0.125(Li 2 O)+SiO (H 2 O) 0.74{(CaO) (Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) }+ 0Ca Na K Li H 2 SiO H 2 O+0.091OH 0.225(CaO)+0.5(Na 2 O)+0.075(K2O)+0.125(Li2O)+SiO (H2O) 0.246{(CaO) (Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.928OH (CaO)+0.4(Na 2 O) (K 2 O) (Li 2 O)+SiO (H 2 O) 0.601{(CaO) 0.52.(Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) }+ 0Ca Na K Li H 2 SiO H 2 O+0.192OH 0.05(CaO)+0.3(Na 2 O)+0.075(K 2 O)+0.125(Li 2 O)+SiO (H 2 O) 0.048{(CaO) (Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.766OH 0.225(CaO)+0.3(Na 2 O)+0.075(K 2 O)+0.125(Li 2 O)+SiO (H 2 O) 0.396{(CaO) (Na 2 O) (K 2 O) 0.12.(Li 2 O) 0.2.(SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.457OH (CaO)+0.2(Na 2 O) (K 2 O) (Li 2 O)+SiO (H 2 O) 0.364{(CaO) (Na 2 O) 0.24.(K2O) (Li2O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.467OH 0.225(CaO)+0.3(Na 2 O)+0.075(K 2 O)+0.125(Li 2 O)+SiO (H 2 O) 0.43{(CaO) (Na 2 O) (K2O) (Li2O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.457OH Eq Eq Eq Eq Eq Eq Eq Eq Eq Eq Eq Eq

106 C IL N IH K IH L IH W IL C IL N IL K IL L IL W IH C I N I K I L I W I C I N I K I L I W I C I N I K H L I W I C I N I K I L I W L C IH N IL K IL L IL W IL C I N I K I L I W I (CaO)+0.4(Na 2 O) (K 2 O) (Li 2 O)+SiO (H 2 O) 0.502{(CaO)0.273.(Na 2 O)0.574.(K 2 O) (Li 2 O) 0.36.(SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.284OH (CaO)+0.2(Na 2 O) (K 2 O) (Li 2 O)+SiO (H 2 O) 0.66{(CaO) (Na 2 O) 0.09.(K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.275OH 0.225(CaO)+0.3(Na 2 O)+0.075(K 2 O)+0.125(Li 2 O)+SiO (H 2 O) 0.573{(CaO) (Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.517OH 0.225(CaO)+0.3(Na 2 O)+0.075(K 2 O)+0.125(Li 2 O)+SiO (H 2 O) 0.346{(CaO) (Na 2 O) (K 2 O) (Li 2 O) 0.07.(SiO2).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.595OH 0.225(CaO)+0.3(Na 2 O)+0.15(K 2 O)+0.125(Li 2 O)+SiO (H 2 O) 0.547{(CaO) (Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) 4.75 } Ca Na K Li H 2 SiO H 2 O+0.635OH 0.225(CaO)+0.3(Na 2 O)+0.075(K 2 O)+0.125(Li 2 O)+SiO (H 2 O) 0.68{(CaO)0.274.(Na 2 O) 0.12.(K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.789OH (CaO)+0.2(Na 2 O) (K 2 O) (Li 2 O)+SiO 2 +11(H 2 O) 0.762{(CaO) 0.41.(Na 2 O) 0.19.(K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.129OH 0.225(CaO)+0.3(Na 2 O)+0.075(K 2 O)+0.125(Li 2 O)+SiO (H 2 O) 0.526{(CaO) (Na 2 O) (K 2 O) (Li 2 O) (SiO 2 ).(H 2 O) } Ca Na K Li H 2 SiO H 2 O+0.525OH Eq Eq Eq Eq Eq Eq Eq Eq The effects of the gel s batched compositions on the phase composition and stoichiometry of gelation In order to better understand how the initial batched molar ratios and water content affect the proportions and compositions of the reaction products solid phase and pore solution, separate regression analyses need to be carried out for each response. Since providing detailed information about each regression model might be frustrating for the readers, the important outcomes of each 88

107 model are briefly mentioned in Table 3-8. As illustrated in the table, the gel s solid phase content was found to be mainly affected by the batched calcium to silicon atomic ratio (denoted as (Ca/Si)i) as well as the molarity of the batched water (H2O)i. Calcium content has a linear increasing effect on the amount of gel s solid part formed during the gelation reactions. Also, the batched water content has a decreasing effect followed by a slight increasing effect after it exceeds certain limit. Each factor explains of the total variations observed in the gel s solid content results as listed in Table 3-6 under step two. These two factors were both found to be significant at a significance level of The calculated P Values represent the probability of deeming the corresponding variables significant when indeed they are insignificant. Therefore, such low P Values observed for these two factors (i.e., (Ca/Si)i and (H2O)i) confirm their significance. The CaO content incorporated in the gels (i.e., (CaO)g) was found to be solely affected by (Ca/Si)i and no other factor showed any effect on this parameter. Interestingly, the regression analyses results suggest that an increase in (Ca/Si)i leads to more incorporation of sodium in the gels (see the (Na2O)g row). This was further confirmed by examining the effect of (Ca/Si)i on the portion of the batched sodium that becomes chemically integrated in ASR gels (i.e., (Na2O)g/(Na2O)i). See the 9 th row on Table 3-8; (Na2O)g/(Na2O)i where (Ca/Si)i shows a linear increasing effect on this parameter. The importance of this observation is that it goes against the theory of alkali recycling originally proposed by Thomas [12]. The alkali recycling effect suggests that the calcium existing in the residual portlandite can replace alkalis in the gels and recycle them back to the cement pore solution, which can ultimately have some ph buffering effect and help ASR to further progress. There are two possible explanations for such disagreement: 1) the alkali recycling occurs at Ca/Si dosages greater than those investigated in this research (i.e., ), and 2) the recycled (i.e., exchanged) alkalis are only slightly displaced and are present 89

108 in the interlayer space of the gels solid parts and cannot be drained via pore solution extraction. Therefore, they are considered to be incorporated in the gels solid parts according to the calculations. However, the latter hypothesis does not explain why the increase in calcium results in an increase in the otherwise chemically bound sodium, unless we take into account the silicate binding effect of calcium (see the (SiO2)g and [Si]Tot), during which some alkalis attached to the non bridging oxygens of silicon tetrhedra are also bound into the gel s network along with the silicates. Unlike the case of chemically bound sodium, no effect was found for calcium on the chemically bound potassium or lithium (i.e., (K2O)g and (Li2O)g). The molar concentration of potassium incorporated in the gel was found to be largely controlled by the initial potassium to silicon molar ratio (i.e., (K/Si)i). Moreover, the sodium showed a slight promoting effect on (K2O)g. Similar to the case of potassium, the chemically bound lithium is primarily controlled by the batched molar concentration of the corresponding alkali (i.e., (Li/Si)i). The chemisorbed water is mainly affected by (H2O)i, (Ca/Si)i and (K/Si)i. (H2O)i explains 10 15% of variations in the free (capillary) water content of the gels and has proven to be fairly significant. A significant interaction was found between (Ca/Si)i and (K/Si)i as long as the chemisorbed water is concerned. For the gels with no batched potassium, calcium has a slight decreasing effect, while at high potassium levels it has a clear increasing effect on the chemisorbed water. 90

109 Table 3-8. Semi quantitative description of the factors influencing the chemistry of the ASR gels Response Important variables Type of effect Contribution (%) P Value Gel solid content (Ca/Si)i a Linear ascending (H2O)i Parabolic convex (CaO)g (Ca/Si)i Linear ascending 99% < (Na2O)g (Na/Si)i Convex ascending (Ca/Si)i Linear ascending (K2O)g (K/Si)i Convex ascending < (Na/Si)i Linear ascending (Li2O)g (Li/Si)i Convex ascending < (SiO2)g (Ca/Si)i Concave ascending (H2O)i Linear descending (Na/Si)i Linear descending (H2O)g : Gel water (Ca/Si)i at low (K/Si)i Slight linear desc. N.A (i.e., chemisorbed (Ca/Si)i at high (K/Si)i Convex ascending N.A. water) (K/Si)i at low (Ca/Si)i Slight convex desc. N.A. ~10 (K/Si)i at high (Ca/Si)i Convex ascending N.A. (H2O)i Parabolic convex (H2O)g/(H2O)i (Ca/Si)i at low (K/Si)i Almost none N.A : The chemisorbed to (Ca/Si)i at high (K/Si)i Convex ascending N.A. batched water ratio (K/Si)i at low (Ca/Si)i Slight convex desc. N.A (K/Si)i at high (Ca/Si)i Convex ascending N.A. (H2O)i Parabolic convex (Na2O)g/(Na2O)i (Ca/Si)i Linear ascending (Na/Si)i Parabolic convex [OH ] (Ca/Si)i Linear descending (Na/Si)i at low (Li/Si)i Linear ascending N.A (Na/Si)i at high (Li/Si)i Linear descending N.A. 91

110 (Li/Si)i at low (Na/Si)i Concave ascending N.A (Li/Si)i at high (Na/Si)i Concave descending N.A. [Si]Tot (Ca/Si)i Convex descending (Na/Si)i Linear ascending (H2O)i Linear ascending Table guide: Linear ascending: Parabolic concave: Concave ascending: Linear descending: Convex ascending: Concave descending: Parabolic convex: Convex descending: a: The i" stnads for initial or batched. (Ca/Si)i, (Na/Si)i and (Li/Si)i are the three important variables controlling the ph of the gels pore solutions. While calcium was found to have a marginally significant decreasing effect on ph, sodium and lithium showed a strong interactive effect. In the case of gels with no lithium, (Na/Si)i was found to have a linear increasing effect on ph. However, (Na/Si)i seems to actually have a decreasing effect on the ph in the case of gels with (Li/Si)i greater than Also, lithium has a reducing effect on ph for the gels incorporating average to high sodium contents. The immediate effect of lithium on ASR gels pore solution ph and its altering influence on the impact of sodium on ph might be one of the contributing factors explaining why lithium has suppressing effects on ASR. As shown in Figure 3-8, gels with high Na/Si experience an over seven fold drop in the OH concentration of their pore solutions when lithium is introduced up to a Li/Si of 0.5. Such drop in the ph could drastically reduce the rate of ASR reaction. It can be argued that once a layer of ASR gel is formed around or within a reactive aggregate, the presence of lithium in considerable concentrations results in a significant decrease in the ph of the surrounding ASR gel s pore 92

111 solution, leading to substantial decline in the rate and extent of alkali attack on the remaining unreacted aggregates. Figure 3-8. The interaction plot of Na/Si and Li/Si on the hydroxide concentration of the gels pore solutions Further statistical analyses were performed on the data in order to detect correlations between chemical characteristics of the gels. It was realized that the molar concentrations of the chemically bound sodium and lithium are correlated with a Pearson correlation coefficient of (P Value = 0.001; suggesting the almost certain presence of correlation). Moreover, the chemisorbed water 93

112 (i.e., (H2O)g) was found to be in direct correlations with the chemically bound calcium and silicon, with correlation coefficients of 0.42 and 0.48 and P Values of and , respectively. The findings of this research can be used at the footstone of the future studies more specifically focusing on the rheology and thermodynamics of ASR gels. Moreover, the detected ph reducing effect of lithium can be instrumental in developing new lithium based ASR inhibiting admixtures as it provides better quantitative understanding as to how this element suppresses ASR damage SUMMARY AND CONCLUSIONS In the present study, synthetic lithium added ASR gels with a broad range of compositions were produced using a new sol gel method. The effects of synthetic ASR gels chemistry on their rheological and chemical properties were investigated, which led to the following findings: The composition of the gels changes their gelation time by near four orders of magnitude. Ca/Si molar ratio is the prominent influencing factor reducing this parameter, which was more pronounced at higher levels of Na/Si. Increases in K/Si and Na/Si result in increases in the gelation time, where the latter showed significant interaction with Ca/Si, and its delaying effect was stronger at lower dosages of Ca/Si. The gels yield stress results also change across nearly two orders of magnitude depending on the chemical composition. Ca/Si showed a significant increasing effect on yield stress causing 78 time increase in this parameter as it increased from 0.05 to 0.4. Na/Si showed the opposite effect on this parameter, which was found to experience a 15 fold decrease upon increase in Na/Si from 94

113 0.2 to 1.0. K/Si and Li/Si had more or less similar effects on the yield stress. Increase in K/Si resulted in a drop in yield stress by a factor of three as it changed from 0.0 to 0.18, followed by 58% increase in this parameter at it approached 0.3. In the case of lithium, a 58% drop in the yield stress was experienced once Li/Si increased from 0.0 to 0.33, followed by a 25% increase as it approaches 0.5. While no conclusions can still be drawn on the effects of chemistry on the minerology of ASR gels, the chemical and analytical methods deployed in this research enabled us in estimating the chemical reactions taken place during the gelation of different experimented ASR gels as well as the proportions of different phases within the gels (i.e., free water, pore solution, and solid phase). Ca/Si showed a linear increasing effect on the gels solid part. It was also found to be the sole influencing factor on the amount of the chemically bound calcium in the gels solid parts. Moreover, it showed a silicate binding effect (by reducing the dissolved silicates in the pore solution and increasing the chemically bound SiO2). However, it showed no alkali recycling effect in the studied range as far as the concentrations of the alkalis in the pore solution are concerned. Sodium had a linear reducing effect on the chemically bound SiO2, and an increasing effect on the ph of the gels pore solution in the absence of lithium. However, it showed a slight decreasing effect on the ph once Li/Si exceeded At high sodium levels (i.e., Na/Si=1.0) an increase in Li/Si from 0.0 to 0.5 was found to be associated with a seven fold drop in the OH concentration in the gels pore solution, which is a substantial change in terms of aggregate dissolution rate and extent. 95

114 3.5. REFERENCES [1] Gholizadeh-Vayghan, A., Rajabipour, F., & Rosenberger, J. L. (2016). Composition rheology relationships in alkali silica reaction gels and the impact on the Gel's deleterious behavior. Cement and Concrete Research, 83, [2] Gholizadeh-Vayghan, A., Rajabipour, F. (2016). The influence of alkali silica reaction (ASR) gel composition on its hydrophilic properties and free swelling in contact with water vapor. Cement and Concrete Research, (under peer-review of the revised version). [3] Montgomery, D. C. (2008). Design and analysis of experiments. John Wiley & Sons. [4] Feilizadeh, M., Rahimi, M., Zakeri, S. M., Mahinpey, N., Vossoughi, M., & Qanbarzadeh, M. (2017). Individual and interaction effects of operating parameters on the photocatalytic degradation under visible light illumination: Response surface methodological approach. The Canadian Journal of Chemical Engineering. [5] American Association of State Highway and Transportation Officials (AASHTO), Standard practice for determining the reactivity of concrete aggregates and selecting appropriate measures for preventing deleterious expansion in new concrete construction, PP 65 11, [6] Kawamura, M., & Fuwa, H. (2003). Effects of lithium salts on ASR gel composition and expansion of mortars. Cement and Concrete Research,33(6), [7] Buckley, A.M., Greenblatt, M. (1994). The sol gel preparation of silica gels. Journal of chemical education, 71(7), 599. [8] Kawamura, M., & Iwahori, K. (2004). ASR gel composition and expansive pressure in mortars under restraint. Cement and concrete composites, 26(1), [9] Liddel, P. V., & Boger, D. V. (1996). Yield stress measurements with the vane. Journal of non Newtonian fluid mechanics, 63(2), [10] Nguyen, Q. D., Akroyd, T., De Kee, D. C., & Zhu, L. (2006). Yield stress measurements in suspensions: an inter laboratory study. Korea Australia Rheology Journal, 18(1), [11] Yoon, J., & El Mohtar, C. (2013). Disturbance Effect on Time Dependent Yield Stress Measurement of Bentonite Suspensions. Geotechnical Testing Journal, 36(1), [12] Thomas MDA (2001). The role of calcium hydroxide in alkali recycling in concrete. In: J. Skalny, J. Gebauer, I. Odler (Editors), Materials Science of Concrete Special Volume on Calcium Hydroxide in Concrete, American Ceramic Society, Westerville, Ohio:

115 CHAPTER 4: THE INFLUENCE OF ALKALI SILICA REACTION (ASR) GEL COMPOSITION ON ITS HYDROPHILIC PROPERTIES AND FREE SWELLING IN CONTACT WITH WATER VAPOR 4 Synthetic ASR gels with 20 different chemical compositions similar to those found in field concretes were produced, and the effects of composition (i.e., Ca/Si, Na/Si and K/Si) on their hydrophilic and swelling behavior were investigated. Minibar gel specimens were cast and tested for time dependent measurement of the free swelling strain and weight change of the gels when exposed to 95% RH. The equilibrium relative humidity (ERH) of the gels was also measured for assessing their hydrophilic potential. The results suggest that higher Na/Si and K/Si increase the free swelling and water absorption of the gels, and reduce their ERH. Ca/Si showed a multi episode effect on the swelling and water absorption of the gels, while showing no significant effect on their ERH. Discussions on the observed effects are provided and regression models for predicting the swelling and hydrophilic properties of the gels are developed. 4 This chapter was published in the Cement and Concrete Research journal Volume 94 in April 2017 (pp 49 58). 97

116 4.1. BACKGROUND Alkali silica reaction (ASR) damage is induced by the attack of hydroxide ions on metastable silicate phases in aggregates, and subsequent formation and expansion of ASR gels. As such, the volume and swelling properties of ASR gels formed in concrete play a significant role on the severity of ASR damage. Although research on ASR is not new, systematic studies on characterization and evaluation of ASR gels are scarce, especially with regards to the impact of the gels chemistry on their behavior in concrete. The most important characteristics of ASR gels are their free swelling capacity (strain), restrained swelling pressure, viscoelastic and rheological properties (primarily yield stress as quantified in a previous work [1]), and the osmotic pressure of their pore solution. The study of the rheological, chemical and swelling properties of synthetic ASR gels is important as it promotes (a) a better understanding of the gels swelling behavior and the factors that control it, (b) developing new chemical admixtures that suppress the swelling potential of ASR gels, and subsequently, damage in concrete, and (c) developing computer models to simulate ASR damage in concrete for predicting the service-life. The free swelling capacity (strain) of a gel is a measure of how much the gel can swell if there is no resistance against its expansion. The capacity for freely swelling in humid environments is a prerequisite for deleteriousness of the gels. ASR gels are found in a variety of different compositions in concrete. They are comprised of complex hydrous combinations of silicon (the network former of the gel), alkalis (i.e., sodium and potassium acting as network modifiers/breakers), and alkaline earths species (i.e., calcium and magnesium that serve to crosslink alkali-silicate chains to form complex units) [2]. A survey of 100 ASR gel compositions measured by SEM/EDS and reported in the literature [3,4,5,6,7,8,9,10,11,12] is summarized in Figure 4-1. It is observed that sodium (Figure 4-1a) is usually present in the gels in larger quantities 98

117 compared to potassium (Figure 4-1b). Magnesium (Figure 4-1d) exists in very small amounts compared to calcium (Figure 4-1c) and, as a divalent cation, may conceivably behave similar to calcium. As such, Mg is not directly studied in this research, and the divalent cations are combined and presented as (Ca+Mg)/Si in Figure 4-1e. Sodium and potassium, however, are studied separately in the present work. The Al/Si atomic ratio in the studied ASR gels was in most cases below 0.05 and, as such, was not considered as an influencing variable in this work. For each gel s oxide composition, a Difference from 100% (weight percent) is reported in the studied literature (Figure 4-1f), which is mainly attributed to oxygen in chemically absorbed water (and OH groups), and possibly carbon. Using all of this information, the composition of ASR gels can be represented in the general form of SiO2.(Na2O)n.(K2O)k.(CaO)c.(H2O)x. The important parameters in the chemistry of the gels, therefore, are the sodium to silica atomic ratio (Na/Si), as well as K/Si and Ca/Si as pointed out by other researchers [13,14]. The gels water content is also a potentially important parameter, but it is not investigated in this research. The objective of this research is to study the effects of the ASR gels composition (i.e., Ca/Si, Na/Si, and K/Si) on their swelling behavior at high relative humidity (e.g., RH=95%). Moreover, the hydrophilic behavior of the gels is assessed by determining their equilibrium relative humidity (ERH) in contact with air in a closed system. The findings of this research are used for developing statistical regression models for predicting the swelling and hydrophilic properties of ASR gels as a function of their composition. The developed models can be used to find the pessimum combinations of these elements inside gels that lead to the most deleterious behavior of ASR gels in concrete (i.e., highest swelling and water absorption capacity and lowest ERH). It should be noted that in concrete, ASR gel composition can change as the gel ages and as it moves away from a reactive aggregate to within cement paste. It is common that different gel 99

118 compositions are found within the same concrete sample. However, if and when these gel compositions are determined (e.g., through SEM/EDS), one can use the outcomes of the present study to estimate the swelling properties and deleteriousness of such gels. It will be an interesting topic for a future research to estimate the composition of ASR gels that form inside a concrete member as a function of the aggregate and cement paste compositions Na/Si 80 K/Si (a) (b) (c) Ca/Si Frequency of observance Mg/Si (Ca+Mg)/Si Difference from 100% (d) (e) (f) Figure 4-1. The distribution of the chemical composition of the studied gels in the literature (the x axes of (a) to (e) are the atomic ratios) 100

119 4.2. EXPERIMENTAL PROGRAM Materials Among a number of commercially available sources of silica (SiO2), dried pulverized colloidal silica (Alfa Aesar) was found to be the most suitable for gel synthesis. Colloidal silica (50% H2O) was dried at 110 C for two days, followed by 48 hours of grinding in a porcelain jar mill, which produced silica powder with an average particle size of 6.37μm. The particle size distribution of the silica powder is shown in Figure 4-2. Reagent grade NaOH and KOH pellets (99% essay) were used as alkali sources. Also, reagent grade Ca(OH)2 powder (98.3% essay) was used as the source of calcium. Figure 4-2. Particle size distribution of the silica powder used for gel synthesis The range of ASR gel compositions studied The ASR gel compositions reported in the literature are usually associated with gel samples extracted from aged concrete structures and gels that were not necessarily taken from the vicinity of aggregates. The composition of ASR gel is known to change with time; increasing in Ca/Si and 101

120 decreasing in Na/Si [15]. Also, gels around or within aggregates are often richer in sodium and incorporate less calcium. Thomas [15] argued that ASR gels tend to undergo an ion exchange process; a phenomenon which he referred to as alkali recycling, where the sodium ions in the gels are replaced with calcium ions, dissolving from portlandite in the cement paste [16]. This process maintains high alkalinity of the concrete pore solution and possibly reduces the swelling capacity of the gels. Many of the reported gels in the literature are aged and they may be in their post-ion-exchange state, where they show little tendency towards expansion. According to Hou et al. [13], alkali and alkaline earth metals in field ASR gels typically appear in the ranges (Na+K)/Si=0.1~1.2 and (Ca+Mg)/Si=0.0~0.2 (atomic ratios). As such, a lower range of Ca/Si and a higher range of Na/Si and K/Si, compared to what Figure 4-1 suggests, was studied here to better represent younger gels that are more likely responsible for concrete expansion and deterioration. Using all of the given information and in order to maintain the generality of the research, a wide range of chemical compositions were studied in this research as shown in Table 4-1. The water content of the studied gels was held constant at 40%. Table 4-1. The studied range of the chemical composition of synthetic ASR gels Variable Lower bound Upper bound (atomic ratios) Ca/Si Na/Si K/Si Gel synthesis method In our previous work [1], a sol gel method was adopted and further developed for synthesizing ASR gels. However, sol gel methods do not usually allow for producing gels with low water content, especially at low levels of Na/Si and/or high levels of Ca/Si. The maximum practical solid phase concentration for gels having intermediately high calcium (e.g., Ca/Si=0.2) and low sodium 102

121 (e.g., Na/Si=0.2) was found to be around 25% (i.e., water content = 75%), while gels forming in field concrete well exceed this solid content level. As such, a new gel synthesis method was deployed in this study. Dry silicon dioxide powder was batched and mixed, using a 3 L Hobart mixer, with appropriate amounts of NaOH, KOH, Ca(OH)2 and water to render the target Na/Si, K/Si, Ca/Si, and a constant water content of 40% (the water contained in the metal hydroxides was taken into account when calculating the total water content of the gels). The mixing protocol adopted for gel synthesis was as follows. First, the alkalis (i.e., NaOH and KOH) were dissolved in the mixing water in an airtight HDPE bottle, and the silica was weighed and dry-mixed with an appropriate amount of Ca(OH)2 powder. It is worth noting that mixing high ph alkaline solutions with the silica-lime mixtures can occasionally generate excessive amounts of heat (depending on the composition), which in turn causes boiling of the mixture seconds to minutes after mixing. In order to avoid such event, the rate of initial reactions was controlled by storing the alkaline solutions in refrigerator at 1 C, and the dry powders in a freezer at 18 C. After cooling down, the alkaline solution was transferred to the mixing bowl, followed by addition of the silica-lime powder. The mixer was started at low speed for 30 sec, followed by scraping and mixing at medium speed for 90 sec. If necessary, the mixture was mixed up to 15 seconds at high speed to break up any residual lumps of powder. The resulting mixtures were molded in minibar molds and 60-mL HDPE cylindrical bottles, and allowed to undergo natural silica dissolution and gelation similar to what occurs in concrete. This new ASR gel synthesis method provides several advantages over conventional solgel methods [1,13,14]. (1) Gels with lower water content (and more representative of what forms inside concrete) can be synthesized. (2) No post-synthesis processing, such as drying to reach the required gel water contents is needed. Drying is a common practice in sol-gel synthesis of ASR 103

122 gels and causes substantial disturbance and potentially cracking in the gels. (3) Since the new process does not rely on the silicic acid sol, there is no risk of unwanted gelation during the synthesis process. (4) The new process allows for better control of the target composition of the gels. In the sol-gel method, the concentration of the silicic acid effluent (the main precursor of the gels) is highly affected by the level of H + saturation of the ion exchange resins and the flushing cycles of the ion exchange column with water. This, in turn, can lead to deviation of the gel composition from the target values, as pointed out by previous researchers [13]. (5) The new method reduces the extent of carbonation during the synthesis process due to the shorter time and fewer steps of processing. (6) The new method produces more homogenous gels. This is in part because the risk of rapid reaction between aqueous silica (gel precursor in the sol-gel method) and calcium is eliminated since both silica and portlandite are introduced as fine solids powders and will only gradually react with each other over the gel s curing period. In the present study, regardless of the target gel composition, homogenous gels were produced and this was verified both visually and by SEM imaging of the gels Design of experiments (DOE) The complex composition property relation of ASR gels calls for employing a DOE that allows for building robust and statistically-defensible regression models that help us unravel such relation and quantify the effects of composition on the properties of ASR gels. To this end, the concepts of response surface methodology (RSM) were implemented to determine linear and non-linear (i.e., higher orders) effects and interactions between the three gel composition variables, namely Ca/Si, Na/Si and K/Si [17]. The resulting regression function has the form of Eq. 4-1 where ŷ is the estimated response parameter (e.g., free swelling of the gel), X1, X2 and X3 are the three 104

123 composition variables, and β i s are the multiplying factors for each term. X1 X2 is the interaction term between Xi and Xj. ŷ=β 0 +(β 1 X 1 )+(β 2 X 2 )+(β 3 X 3 )+(β 12 X 1 X 2 )+(β 13 X 1 X 3 )+ (β 23 X 2 X 3 )+(β 11 X 1 2 )+(β 22 X 2 2 )+(β 33 X 3 2 ) Eq. 4-1 There are two available designs in RSM: Box Behnken design and central composite design (CCD) [17]. The latter approach was adopted in the present study. Following the CCD design, a 3D demonstration of combinations of different levels of the gel composition variables (i.e., Ca/Si, Na/Si and K/Si) is presented as three orthogonal axes in Figure 4-3. CCD, which was first introduced by Box and Wilson [18], builds the regression model based on the results of laboratory tests performed at axial, cube, and center points on the surface of the DOE pseudosphere (Figure 3). In other words, we synthesize gels with compositions represented by the axial, cube, and center points, test them in the laboratory to determine their swelling properties, and analyze the results using the analysis of variance (ANOVA) techniques to build the regression models. The axial points are the points where the pseudo-sphere intercepts each axis. Thus, 6 axial points are defined in this design where one composition variable is set at its high or low level, while the other two variables are at their intermediate levels. For example, the axial points on the Ca/Si axis have Na/Si = 0.55 (intermediate level), K/Si = 0.15 (intermediate level), and Ca/Si at either 0.05 (low level) or 0.50 (high level); see Table 4-2. The cube points (which enable estimation of the two way interactions between composition variables) are the points at the vertices of the cube encircled by the DOE pseudo-sphere in Figure 4-3. As such, there are 8 cube points. The 105

124 center point represents the center of the pseudo-sphere, where all composition variables are at their intermediate levels. Usually experiments are repeated several (4 to 7) times at the center point to determine the repeatability of the test. Here, we will use 6 center points, that along with 6 axial points and 8 cube points, yield 20 experiments that will be sufficient for estimating the linear and quadratic effects of the three composition variables as well as their two way interactions on the hydrophilic and swelling properties of the gels. Figure 4-3. Graphical demonstration of RSM data point selection for 3 variables case (α = 1.682) Table 4-2 contains the coded and natural levels of each composition variable, along with their corresponding gel labels, for the cube, axial, and center points. The coded dosage levels of 1.682, 1, 0, +1 and represent the low (L), intermediate low (IL), intermediate (I), intermediate high (IH) and high (H) levels of each composition variable. Also, in the gels labels, C, N, and K represent Ca/Si, Na/Si and K/Si, respectively. For instance, C IL N IL K IH is a gel with intermediate low level of Ca/Si (=0.141), the intermediate low level of Na/Si (=0.282), and the 106

125 intermediate high level of K/Si (=0.239). Experiments on these 20 gels are run in a random order (Table 2) in order to eliminate the effects of time and examiner s learning curve. Table 4-2. The experimental design with the coded and natural levels of the composition variables Variable Ca/Si Na/Si K/Si Run order Point type Label Coded Natural Coded Natural Coded Natural 5 C IL N IL K IL C IH N IL K IL C IL N IH K IL C IH N IH K IL C IL N IL K IH C IH N IL K IH C IL N IH K IH C IH N IH K IH C L N I K I C H N I K I C I N L K I C I N H K I C I N I K L C I N I K H , 8, 12, 13, 14, 17 Cube points Axial points Center points C I N I K I (1) to (6) respectively The mixture proportions for the 20 gels are shown in Table 4-3. The source of silica contained an initial Na/Si of 0.022, which was taken into account in calculating the required NaOH content. Moreover, the water embedded in Ca(OH)2, NaOH and KOH was subtracted from the total water content listed in the last column of Table Free swelling strain measurements For each gel mixture, four 12.7 mm 12.7 mm 127 mm (½" ½" 5") bars (referred to as minibars ) were cast and demolded after 24 to 72 hours (depending on the setting time of each gel). The specimens were vacuum sealed and cured at 25±1 C for two months to ensure that gelation reactions were completed and a microscopically homogenous gel was formed before any 107

126 swelling or RH measurements (Figure 4-4). The vacuum seal prevented carbonation and moisture loss (or gain) during curing. Length measurements were not performed during the sealed curing period. It is possible that small autogenous shrinkage (or expansion) might have occurred during curing, which was not measured. After curing, the specimens were stored in a nitrogen purged chamber at 95% RH for monitoring their expansion over time using a comparator with an accuracy of mm (Figure 4-5). The recorded length change values were then converted to free swelling strain (εg,fr %) values. The weight changes (Δw %) were also monitored using a gr accuracy balance. The measurements were taken at 1, 4, 7, 14, 21 and 28 days. Table 4-3. Mixture proportions of the gels Run order Label Ca(OH)2 NaOH KOH Silica powder H2O (gr/1000 gr) (gr/1000 gr) (gr/1000 gr) (gr/1000 gr) (gr/1000 gr) 1 C IL N IH K IH C IL N IL K IH , 8, 12, C I N I K I (1) to 13, 14, 17 (6) C I N L K I C IL N IL K IL C H N I K I C I N H K I C IH N IL K IH C IH N IH K IH C I N I K H C L N I K I C IH N IH K IL C IL N IH K IL C IH N IL K IL C I N I K L

127 Figure 4-4. Vacuum sealed minibar specimens of each gel composition Figure 4-5. Length change measurements of minibar specimens Equilibrium relative humidity (ERH) measurements Upon completion of curing, the gel specimens cast in 60-mL HDPE cylindrical bottles were extracted and crushed into <5mm particles while sealed in plastic bags. The entire crushed gel specimens were transferred into 120-mL bottles equipped with RH meter sockets (yellow sockets in Figure 4-6a), and stored inside an environmental chamber at 25 C (Figure 4-6b) for one week before testing. The RH sensors were routinely calibrated, and the repeatability of the ERH 109

128 measurements was assessed by performing measurements on 6 duplicate CINIKI gels (center point composition) that were batched and synthesized separately. (a) (b) Figure 4-6. An illustration of the ERH measurements setup: (a) crushed gel specimens stored inside a plastic bottle for ERH measurement, (b) specimens stored inside envirnmental chamber at 25 C 4.3. RESULTS Free swelling results The results of free swelling experiments (expansion strain and weight change) are plotted in Figure 4-7 and Figure 4-8 (each split into two charts for better visibility of the results). Figure 4-7 shows the time dependent variations of εg,fr (i.e., free swelling strain) of the ASR gel specimens. Note that depending on their composition, some gels showed shrinkage at 95% RH, indicating that their ERH are probably greater than 95%. The greatest ultimate εg,fr values were observed in the case of C IH N IH K IH (εg,fr=8.84±0.22%), followed by C I N H K I (7.87±0.67), C L N I K I (7.65±0.27%), C IL N IH K IH (3.96±0.44%) and C IH N IH K IL (3.75±0.15%). Similarly, the greatest shrinkage (i.e., negative εg,fr) values were found in the case of C IL N IL K IL (εg,fr= 6.37±0.04%) followed by C IL N IL K IH ( 3.20±0.05%). 110

129 Figure 4-7. The free swelling strain of ASR gel minibars up to 28 days (the error bars indicate two standard errors) 111

130 Figure 4-8. The weight change (%) of ASR gel minibars up to 28 days (the error bars indicate two standard errors) 112

131 A number of gels showed some expansion in the early ages followed by shrinkage. C I N I K I (1) and (3) (6), C I N I K L and C IH N IL K IL gels showed this behavior. Visual inspections of the gels between the measurements suggested that some of the gel specimens had condensation water droplets formed on their surfaces. This may be partly due to strong water affinity of such gels and partly due to small temperature and humidity variations inside the chamber during the experiment. The condensation water can, over time, leach alkalis out of the gels (similar problem as in ASTM C1293 testing). Such alkali leaching results in gradual loss of water affinity and swelling capacity of the gels and eventually leads to shrinkage. Figure 4-8 shows the weight change (%) of the gel specimens over time. It can be noted that the order of gels showing greatest Δw results, is somewhat different than that that of εg,fr results. The greatest weight gain (i.e., moisture uptake) is observed in the case of C L N I K I (Δw=18.08±0.19%), followed by C I N H K L (15.25±3.18%), C IH N IH K IH (12.56±1.38%), C IL N IH K IH (8.01±1.47%) and C IH N IH K IL (2.05±0.34%). Similar to εg,fr results, the greatest weight loss was found in the case of C IL N IL K IL (Δw= 14.77±4.90%). It is worth noting that the issue of alkali leaching has manifested itself in the weight change behavior of at least nine (out of twenty tested) gels as shown in Figure 4-8. It is worth noting that the relation between free swelling and weight change in minibars is not always linear. While in certain cases (e.g., C I N L K I, C IL N IL K IL, C L N I K I ), a linear correlation between Δw and εg,fr was found, some other gels (e.g., C I N I K H, C IL N IH K IL ) continued expanding beyond the point where they stopped gaining weight and even started to show weight loss. 113

132 Figure 4-9. The free swelling strain (%) versus weight change (%) variations of minibar specimens Figure 4-9a shows the free swelling strain (%) versus weight change (%) variations of all twenty gels. Twelve such gels (including the six replicates) had relatively small variations and, as such, zoomed in views of their curves near the origin are presented separately in Figure 4-9b and c. The legends on the right hand side of Figure 4-9a show the gel labels as well as the Pearson s linear correlation coefficient between εg,fr and Δw of different gel compositions, along with the p value of the reported coefficient. A p value greater than 0.05 indicates a lack of correlation 114

133 between weight change and strain. This was observed in the case of 8 gels, five of which are associated with the 6 replicated gels with C I N I K I composition. Note that the three cases of negative correlation coefficients (i.e., C I N I K H, C I N I K I (3), C IL N IH K IL ) are due to the regressive behavior of the weight change in these gels and not necessarily indicative of any inherent inverse correlation between weight change and swelling in those gels (judging by the large p values associated with these coefficients) Equilibrium relative humidity results Figure 4-10 shows the ERH results of the gels at the time of testing (i.e., 2 months of curing plus one week for stabilization). The ERH results varied from 70.5% to as high as 98.9%. It is observed that gels containing high sodium and/or potassium contents (e.g., C I N H K I, C IL N IH K IH, C IH N IH K IH, etc.) show the lowest ERH results, while the low alkali gels (e.g., C IL N IL K IL, C IH N IL K IL, C I N L K I ) have high ERH values. Unlike alkalis, Ca/Si showed minor effects on the ERH of the gels. The ERH experiments showed high consistency and repeatability. Notice the small variations among the ERH readings of the six replicates of C I N I K I ranging from % (coefficient of variations: COV=1.1%). These variations are partly due to using three different RH probes (calibrated with K2CO3, NaCl, KCl and KNO3 saturated salt solutions), which introduce some variability in the RH readings. Nevertheless, the observed ERH values for identical gel compositions suggests that the designed experiment is capable of adequately measuring the ERH of synthetic ASR gels. 115

134 Figure The equilibrium relative humidity results of gels (COV of the measurements = 1.1%) 4.4. REGRESSION MODEL The obtained results for the 28 day free swelling strain (εg,fr (%)), weight change (Δw (%)), and ERH (%) are further analyzed in this section using statistical tools. Regression models are developed for each of these response parameters in order to better understand the sole and combined effects of each composition parameter (i.e., Ca/Si, Na/Si and K/Si) of the gel. Technical justifications for the statistical observations are provided. Table 4-4 shows the shorthand ANOVA table of the response parameters (ε g,fr, Δw, and ERH) with respect to different predictors. For each predictor, three quantities are reported; Contribution (%) to the total variations in the response, Coefficient of the predictor as it appears in the regression formula for that response parameter, and 116

135 the p value or the significance level of the predictor. A p value smaller than 0.05 indicates a significant influence of the predictor on the response. For more information regarding the regression procedure and interpretation of the ANOVA table please refer to [1] and [19]. Table 4-4. The shorthand ANOVA table of the free swelling, weight change, and equilibrium RH measurements results Response 28 day free swelling strain 28 day weight change Equilibrium relative humidity Contribution (%) (%) Coefficients P value Contribution (%) Δw (%) Predictor Constant Ca/Si Na/Si K/Si Ca/Si Na/Si K/Si Ca/Si Na/Si Ca/Si Na/Si K/Si Coefficients P value Contribution (%) ERH (%) Coefficients P value Prediction accuracy Regular Adjusted Predicted Regular Adjusted Predicted Regular Adjusted Predicted R 2 R 2 R 2 R 2 R 2 R % 79.12% 35.15% 82.06% 71.60% 0.00% 99.16% 98.41% 94.96% R 2 R 2 R 2 The response parameters are analyzed against all possible combinations of the variables and the significant/necessary predictors are included in the regression models. It is observed that the selected predictors can overall explain approximately 85% of the variations in εg,fr (see the regular R 2 value). Slightly lower coefficient of determination was obtained for Δw, while that of ERH is exceptionally high. The ERH regression model, proved to be very reliable in terms of repeatability, evidenced by its great predicted R 2 value. The regression functions for εg,fr and ERH 117

136 are provided in Eq. 4-2 and 4-3 below. Based on the developed regression functions, contour plots of εg,fr, Δw, and ERH are drawn in Figure 4-11 as a function of gel composition. Each contour plot shows variations of a response parameter (εg,fr, Δw, or ERH) as a function of two composition parameters (e.g., Na/Si and Ca/Si) while the third composition parameter (e.g., K/Si) is held at its intermediate value. ε g,fr (%) = Ca Na Si Si K Ca +823 ( Si Si )2-937 ( Ca Si )3 Eq. 4-2 ERH (%) = Ca Si Na Si K Si ( Ca Si ) ( Na Si )2 Eq ( K Si ) ( Ca Si Na Na ) ( ) ( Si Si ) ( K Si )3 The general trend of the results suggests that an increase in Na/Si and K/Si leads to increases in εg,fr and Δw of the gels. This is reasonable considering that sodium and potassium are both monovalent ions and are often identified as network modifiers/breakers in alkali silicate structures [ 20 ]. They promote volumetric instability of the gels via reducing the connectivity of silica network [20], increasing the surface charge density of the gel [ 21 ], and increasing the osmotic pressure of the gel s pore solution. The latter two promote the gel s water affinity [1]. Unlike alkalis, the effect of Ca/Si on εg,fr and Δw does not appear to be monotonic. At low levels (i.e., Ca/Si=0.05 to 0.18) and high levels (i.e., Ca/Si > 0.4), an increase in Ca/Si results in decreases in εg,fr and Δw. However, Ca/Si in the range 0.18 to 0.4 shows an increasing effect on εg,fr and Δw. We hypothesize that there are at least three competing actions in effect. (1) At low ranges of Ca/Si (<0.18), an increase in Ca/Si results in a decrease in the water affinity and swelling of the gels by reducing the surface charge density of the gels (by Ca 2+ ions neutralizing the negative surface charge associated with Si-O- endings). (2) At intermediate to high ranges of Ca/Si (>0.18), calcium 118

137 participates in an ion-exchange reaction by replacing sodium in Si-O - Na + endings to form Si-O - Ca 2+ - O-Si links. This ion-exchange reaction was termed alkali-recycling by [15, 16], and results in release of alkali ions into the gel s pore solution, which increases its osmotic pressure and water affinity. (3) At high levels of Ca/Si, the network binding effects of Ca effectively reduces the degrees of freedom of the gel s network for swelling, thus decreasing its swelling capacity. It is a reasonable assumption that at low levels of Ca/Si, Ca does not substitute alkalis within the gel structure, since there is an abundance of readily available O vacancies that calcium can occupy without confronting the alkali occupied O endings. However, as Ca/Si increases (apparently beyond 0.18), the alkali recycling effect starts to become dominant, which boosts the alkali concentrations in the gel s pore solution and increases the osmotic pressure. It is also plausible that calcium provides cross-linking but does not show considerable interlock and network binding effect at low and intermediate Ca/Si levels. However, as Ca/Si exceeds beyond 0.4, it gradually eliminates the capability of the gel to swell. Extrapolations from the regression model suggests that regardless of the alkali content, gels with Ca/Si greater than 0.55 do not show any swelling capacity. More investigation is needed on the pore solution and atomic structure of the gels to evaluate these hypotheses. Alkalis showed a monolithic decreasing effect on the equilibrium relative humidity of the gels (see the ERH contour plots versus K/Si and Na/Si in Figure 4-11). Increasing alkalis increases the osmolarity of the gels pore solution [1] and reduces its equilibrium RH according to Raoult s law [22]: i=1 x i M i ERH(%)= (1 k i=1 x i M i (1 k y i M i k i=1 ) ) 100% Eq

138 Figure The contour plots of 28 day free swelling strain (%), 28 day weight change (%) and equilibrium relative humidity (%) of the gels versus Ca/Si, Na/Si and K/Si Where xi is the number of ions of the i th solute in the solution, Mi is its concentration in moles and yi is the specific molar volume (m 3 /mole) of that solute. As the alkali content of the gel increases, their concentration in the pore solution (i.e., Mi) will also increase leading to the decrease in ERH. It is worth noting that larger ions are more efficient in reducing ERH (e.g., K + versus Na + ), which is in agreement with the obtained regression equation for ERH. Ca/Si was observed to only have a slight effect on ERH of the gels. 120

139 Finally, it should be noted that εg,fr, Δw, and ERH are correlated. Figure 4-12 shows the pairwise scatter plots of these parameters for different gels. The 6 replicated gels with CINIKI composition are averaged and shown in these plots as CINIKI (1 6). For each pair of responses, a regression line has been fitted to the data. It is observed in Figure 4-12a that the free swelling strain has a general direct linear correlation with weight change. The coefficient of correlation is found to be (p value<0.0005). ERH has an inverse correlation with free swelling strain and weight change, with correlation coefficients of and 0.694, respectively (both with p values smaller than ). (a) (b) (c) Figure The pairwise scatter plots of 28 day free swelling strain (%), 28 day weight change (%) and equilibrium relative humidity of the gels 121

140 As mentioned before, swelling is not always accompanied by weight gain. In some cases, the gels initially gained weight and swelled; however, after a period of time, the weight gain plateaued or even turned into a slight weight loss, while swelling continued to increase (Figure 4-9). Figure 4-12a shows that there are a number of cases were the gels showed swelling at 28 days while the ultimate net weight change was negative; i.e., the gel lost weight (please see the data point on the top left of the 0 0 reference lines). One possible explanation for the slight weight loss of the gels is the leaching of alkalis from the minibar specimens during the experiments, as discussed before. Simultaneously, the absorption or loss of water from the gels change the solutes concentrations in their pore solutions, and hence the ERH of the gels. It is remarkable, however, that during this time, the gels continue to swell, even when they are experiencing some weight loss towards the end of the experiment. Similar counterintuitive responses have been observed for other materials such as dragline silk fibers in which water absorption is accompanied by shrinkage, a phenomenon usually referred to as supercontraction [23]. In the present work, some gels continued to swell after losing some water; we refer to this phenomenon as superswelling behavior. The observed superswelling in some ASR gels (during the slight weight loss period) may be caused by an increase in the surface charge density of the gel s solid skeleton. It should be noted that the leaching of alkalis and hydroxide ions from the minibar specimens leads to a drop in the ph of the gel s pore solution. Such drop in the ph can drive the aqueous sodium silicates species in the pore solution towards gelation. This creates new negatively charged solid surfaces that can contribute to the double layer repulsive forces in the gel, and cause further swelling. Further research is needed to verify and more accurately understand the mechanisms behind the superswelling behavior of ASR gels. 122

141 4.5. CONCLUSIONS The following conclusions can be drawn based on the findings of this research: Increasing the alkali content (Na/Si and K/Si) in ASR gels results in an increase in the gels free swelling and water absorption, and a reduction in the equilibrium relative humidity (ERH) of the gels. Ca/Si of ASR gels has a multi episode effect on the swelling and water absorption properties of the gels. As Ca/Si increases from 0.05 to 0.18, swelling and water absorption of gels decrease by 7.75% and 19.4%, respectively. However, when Ca/Si increases from 0.18 to 0.4, the gel s swelling and water absorption increased by 4.8% and 9.2%, respectively. Ca/Si above 0.4 gradually eliminates the swelling capacity of ASR gels and no swelling is predicted for Ca/Si = Calcium content of ASR gels does not significantly affect their equilibrium relative humidity. Some synthetic ASR gels showed slight weight (water) loss after a period of weight gain. This is likely due to alkali leaching that occurred during the minibar experiments. Interestingly, the slight weight loss did not necessarily result in shrinkage and some gels continued to expand. This superswelling phenomenon may be attributed to an increase in the double layer repulsive forces in the gels, due to formation of new negatively charged alkali silicate solid surfaces, as a result of alkali leaching and decrease in the ph of the gel s pore solution. 123

142 4.6. REFERENCES [1] Gholizadeh Vayghan, A. G., Rajabipour, F., & Rosenberger, J. L. (2016). Composition rheology relationships in alkali silica reaction gels and the impact on the Gel's deleterious behavior. Cement and Concrete Research, 83, [2] Powers, T. C., & Steinour, H. H. (1955). An interpretation of some published researches on the alkali-aggregate reaction. Journal of the American Concrete Institute, 51(26), [3] Stanton, T. E. (1942). Expansion of concrete through reaction between cement and aggregate. Transactions of the American Society of Civil Engineers, 107(1), [4] McConnell, D., Melenz, R. C., Holland, W. Y., & Greene, K. T. (1947, October). Cementaggregate reaction in concrete. In ACI Journal Proceedings (Vol. 44, No. 10). ACI. [5] Hester, J. A., & Smith, O. F. (1958). THE ALKALI-AGGREGATE PHASE OF CHEMICAL REACTIVITY IN CONCRETE PART II. Special Technical Publication, (205), [6] Idorn, G. M. (1961). Studies of Disintegrated Concrete-Part 1, Progress Report N 2, Committee on Alkali Reactions in Concrete, Danish Nat. Inst, of Build. Res, and Acad. of Tech. Sci, [7] Buck, A. D., & Mather, K. (1969). Concrete Cores from Dry Dock No. 2, Charleston Naval Shipyard, SC (No. AEWES-MISC-PAPER-C-69-6). ARMY ENGINEER WATERWAYS EXPERIMENT STATION VICKSBURG MS. [8] Poole A.B. (1975). Alkali-Silica Reactivity in Concrete from Dhekelia Cyprus. Symp. on Alkali-Aggregate Reaction, Proc., Reyjavik, Finland, 101. [9] Gutteridge, W. A., & Hobbs, D. W. (1980). Some chemical and physical properties of beltane opal rock and its gelatinous alkali silica reaction product.cement and Concrete Research, 10(2), [10] Oberholster, R. E. (1983). Alkali reactivity of siliceous rock aggregates: diagnosis of the reaction, testing of cement and aggregate and prescription of preventative measures. Alkali in Concrete, Research and Practice, [11] Visvesvaraya, H. C., Mullick, A. K., Samuel, G., Sinha, S. K., & Wason, R. C. (1986, September). Alkali reactivity of granitic rock aggregates. In Proceedings VIII International Congress on the Chemistry of Cement (Vol. 5, pp ). [12] Šachlová, Š., Přikryl, R., & Pertold, Z. (2010). Alkali-silica reaction products: Comparison between samples from concrete structures and laboratory test specimens. Materials Characterization, 61(12), [13] Hou, X, Kirkpatrick, RJ, Struble, LJ, and Monteiro, PJ (2005). Structural investigations of alkali silicate gels. Journal of the American Ceramic Society 88(4):

143 [14] Krogh, H (1975). Examination of synthetic alkali-silica gels. In: Symposium on Alkaliaggregate Reaction, Preventive Measures, Reykjavik, Iceland: [15] Thomas MDA (2001). The role of calcium hydroxide in alkali recycling in concrete. In: J. Skalny, J. Gebauer, I. Odler (Editors), Materials Science of Concrete Special Volume on Calcium Hydroxide in Concrete, American Ceramic Society, Westerville, Ohio: [16] Rajabipour, F., Giannini, E., Dunant, C., Ideker, J. H., & Thomas, M. D. (2015). Alkali silica reaction: current understanding of the reaction mechanisms and the knowledge gaps. Cement and Concrete Research, 76, [17] Montgomery, D. C. (2008). Design and analysis of experiments. John Wiley & Sons. [18] Box, G. E., & Wilson, K. B. (1951). On the experimental attainment of optimum conditions. Journal of the Royal Statistical Society. Series B (Methodological),13(1), [19] Neter, J., Kutner, M. H., Nachtsheim, C. J., & Wasserman, W. (1996).Applied linear statistical models (Vol. 4, p. 318). Chicago: Irwin. [20] Shelby, JE (2005). Introduction to glass science and technology. Royal Society of Chemistry. [21] Rodrigues, FA, Monteiro, PJ, and Sposito, G (2001). The alkali silica reaction: the effect of monovalent and bivalent cations on the surface charge of opal. Cement and Concrete Research, 31(11) [22] Wheeler, M. J., Russi, S., Bowler, M. G., & Bowler, M. W. (2012). Measurement of the equilibrium relative humidity for common precipitant concentrations: facilitating controlled dehydration experiments. Acta Crystallographica Section F: Structural Biology and Crystallization Communications, 68(1), [23] Guan, J., Vollrath, F., & Porter, D. (2011). Two mechanisms for supercontraction in Nephila spider dragline silk. Biomacromolecules, 12(11),

144 CHAPTER 5: QUANTIFYING THE FREE AND RESTRAINED SWELLING PROPERTIES OF ALKALI SILICA GELS AS A FUNCTION OF THEIR COMPOSITION 5 Synthetic ASR gels were produced and tested to investigate the effects of chemical composition (Ca/Si, Na/Si and K/Si atomic ratios) on the gels free swelling strain (εg,fr) and restrained swelling pressure (Prs). The gels were cast into disk-shape molds and exposed to distilled water after curing. Each gel s εg,fr was recorded over a period of 28 days, followed by measuring Prs, defined as the pressure required to fully reverse and eliminate the gel s free swelling under a drained configuration. Regression models were developed linking gels compositions to their swelling properties. The outcomes show that Na/Si and K/Si monotonically increase εg,fr. Increasing Ca/Si up to 0.23 drastically reduces εg,fr, higher Ca/Si has modest effect on free swelling. Prs increases by increasing calcium up to a pessimum Ca/Si level; Prs decreases for higher Ca/Si. The value of (Ca/Si)pess is related to the alkali content of the gel. Prs also increases by increasing the gel s alkali content, while a (Na/Si)pess exists in the range 0.85 to These observations are linked with the roles of alkalis and calcium in modifying the silica gel network. 5 This chapter has been accepted for publication in the Journal of American Ceramic Society. 126

145 5.1. BACKGROUND The magnitude of alkali silica reaction (ASR) expansion and damage in concrete is directly related to the volume fraction and swelling properties (e.g., free swelling strain, restrained swelling pressure) of ASR gels. Systematic studies on characterization of ASR gels have been scarce, especially with respect to the role of the gels chemical composition on their swelling behavior. Such studies can provide important benefits including: (i) a better understanding of the swelling behavior of ASR gels and the factors that control it, (ii) potentially enabling the synthesis of new chemical admixtures to suppress the swelling potential of ASR gels, and (iii) contributing to the development of computer models to simulate ASR damage and predict the service-life of concrete. The composition of ASR gels can be represented by the general form SiO2.(Na2O)n.(K2O)k.(CaO)c.(H2O)x [1]. ASR gels seldom also incorporate alumina in their structure. However, the important compositional parameters are the atomic (or molar) ratios of Na/Si, K/Si, Ca/Si, and the water content of the gels [2,3]. ASR gels are found in concrete structures in a wide range of compositions [4]. Certain characteristic compositions can exhibit deleterious swelling, while other gel compositions may be innocuous [5,6]. In a recent study [7], we statistically analyzed the chemical compositions of one hundred different ASR gel samples reported in the literature [8,9,10,11,12,13,14,15,16,17]. We found that the following ranges of Ca/Si, Na/Si and K/Si (atomic ratios) conceivably cover the entire range of ASR gels formed in field concrete: ( ), ( ), and ( ), respectively. In that study [1], the effects of composition on the equilibrium relative humidity (ERH) of ASR gels, as well as the free swelling and moisture uptake of gel minibars exposed to water vapor (inside RH = 95% environmental chamber) were investigated. However, the swelling pressure of ASR gels was not studied, and neither was their free swelling response in contact with liquid water. These two parameters (free 127

146 swelling strain and restrained swelling pressure, which are the subjects of the present paper) are demonstrably the most important, yet independent factors governing the swelling behavior of ASR gels. For an ASR gel to be deleterious, it needs to have the capacity to swell extensively and also to exert and maintain pressure to its surrounding while swelling. Neither a gel with high free swelling capacity but low swelling pressure, nor a gel with high swelling pressure but low free swelling are ASR deleterious. For example, a dilute alkali-rich ASR gel can freely expand to high strain values; but it is incapable of imposing damaging pressures to its surrounding where its expansion is restrained. This is due to the low viscosity and yield stress of such gels, causing them to easily flow and relieve their stress [1]. On the other side of the spectrum, rigid high-calcium gels, with Ca/Si approaching those of C-S-H, exhibit practically no swelling strain regardless of their hydrophilic nature and high osmotic pressure [18]. In other words, although there is a driving force for swelling, these gels do not swell due to their three dimensionally connected (cross-linked) calcium-silica network, which internally restrains the swelling [19]. There have been few studies in the past to experimentally relate the composition of ASR gels to their swelling behavior. Most notably, Struble and Diamond [20] investigated the effect of composition of synthetic ASR gels (primarily Na/Si and water content at two levels of Ca/Si = 0.0 and 0.17 to 0.18) on their swelling properties. They found the composition to be strongly affecting the free swelling strain and swelling pressure of the gels. Free swelling strains were found to vary from 0.5 to 81%, while the swelling pressure could be as little as 0.1 MPa to as high as 11 MPa. They also identified gels with high swelling strains and low swelling pressures, as well as gels with very low swelling strains but high swelling pressures. However, they noted several complications as their results seemed to be affected by nuance factors such as the testing age, the 128

147 necessity of applying an activating pressure on several gels for the emergence and acceleration of swelling behaviors, and other unknown factors. These complications, together with the lack of replication (as pointed out by the authors), were the reasons why the sole and combined effects of sodium, calcium, silicon and water contents on the swelling properties of ASR gels could not be quantitatively and reliably determined. In the present work, we implemented a robust statistical design of experiments, which is capable of adequately capturing and quantitatively characterizing the interactive effects of ASR gels compositional parameters on their swelling properties. Sufficient replication and randomization was provided to increase the power of the experiments and block the effects of nuance and unknown factors. Also, we developed and employed a new gel synthesis method, which yields consistent gels without imposing any post-batching disturbances on the gels (such as aggressive drying to adjust the moisture content, crushing, consolidation, and excessive handling that were typically used in the past gel synthesis studies). Moreover, we utilized new test methods for measuring the free swelling strain and swelling pressure of the gels. These proved to be viable and repeatable EXPERIMENTAL PROGRAM Gel synthesis ASR gels were synthesized by mechanically mixing fine amorphous silica particles with appropriate amounts of sodium, potassium and calcium hydroxides and distilled water. Dried, pulverized colloidal silicon dioxide IV (50% H2O, procured from Alfa Aesar) was found to be a suitable source of amorphous silica for gel synthesis. The colloidal silica was dried at 100 C for 48 hours, followed by 48 hours of grinding inside a 5-L porcelain jar mill. The resulting powder 129

148 had an average particle size of 6.37μm. NaOH and KOH pellets (99% assay) were used as the sources of sodium and potassium. Ca(OH)2 powder (assay: >95%) was used as the source of calcium. The following ranges of compositions were studied in the present work: Ca/Si = ( ), Na/Si = ( ), and K/Si = ( ) atomic ratios. The gel synthesis method consisted of mixing the gel ingredients (i.e., silica powder, calcium hydroxide, sodium hydroxide, potassium hydroxide, and water) using a 3-L Hobart mixer and allowing the components to undergo dissolution-gelation processes, similar to what occurs in concrete. The following mixing protocol was employed: NaOH and KOH pellets were first dissolved in the mixing (distilled) water in an airtight HDPE bottle. The silica and calcium hydroxide powders were weighed and dry-mixed together, apart from the alkaline solution. To avoid generation of excessive heat upon mixing the high-ph alkaline solution and the dry silica-calcium hydroxide powder mixture, the alkaline solution was refrigerated at 1 C, and the dry powder was cooled in a freezer at 18 C. The purpose of this step was to reduce the rate of initial reactions and avoid abrupt heat generation and boiling of the water. After cooling down the ingredients, the alkaline solution was transferred into the mixing bowl, followed by addition of the dry powder. The mixer was started at the low speed for 30 seconds, followed by scraping and mixing at medium speed for 90 s. The mixtures were mixed as necessary up to 15 s at high speed in order to break up any lumps Free swelling experiments on ASR gel disks The resulting ASR gel mixtures were cast into disk-shape setups, which later allowed for the water absorption and swelling pressure experiments to be performed. The ASR gel mixtures were molded into 8mm-thick circular stainless steel rings (D = 44mm) and quickly covered by a steel cap and wrap sealed using electric tapes. Three days later, the gel disk specimen along with the 130

149 steel ring (Figure 5-1a), were vacuum sealed for curing for an additional two months (Figure 5-1b) at 23±1 C. The rationale behind vacuum sealed curing was to prevent any unwanted water absorption or desorption by the gel, which would result in swelling or shrinkage. (a) (b) Figure 5-1. An illustration of synthetic ASR gel disk specimens: (a) three days after casting; (b) vacuum sealed for further curing In parallel, cement pastes with water-to-cement ratio of 0.47 were produced and cast into 15.3mm-thick stainless steel rings equipped with threaded studs to allow length measurements (Figure 5-2). First, the studs were screwed into the center holes of solid steel baseplates. The steel ring molds were greased (using high vacuum grease) on the bottom of their perimeter, where they would come in contact with the baseplate. The inside perimeter of the steel rings, that would later contact cement paste, was also greased using light petroleum grease. Each steel ring was then placed on top of a baseplate and the two parts were taped to prevent sliding during casting of cement paste. The fresh cement paste was prepared and poured into these ring molds, and the surface was finished using a spatula. 131

150 A thin steel disk was placed on top of the finished cement paste surface to prevent evaporation. The assembly was then stored in the laboratory and wrapped inside plastic bags for 24 hours, after which the cement paste disks were demolded and cured inside a moist room for 2 months. Upon completion of curing, the cement paste disks were dried at 50±5% RH for 4 hours, and then polished on their perimeter and bottom face (which will later contact the ASR gel) using #80 grit sandpaper to remove roughness. The perimeter of each cement paste disk was subsequently Teflon taped, and the disks were inserted back into the steel rings such that they can freely slide out without friction. This setup allowed ASR gels to freely expand while preventing the leakage of water or gel. A finished cement paste disk that is partially inserted back into a steel ring mold is shown in Figure 5-3. Baseplate Steel ring Plastic electrical tape Figure 5-2. The steel ring molds used for making cement paste disks 132

151 Cement disk Figure 5-3. Hydrated, polished, Teflon-taped cement paste disk partially slid back into the steel ring mold After preparing the cement paste disk units, two were placed on the top and bottom of each ASR gel specimen that was prepared previously inside its own steel ring (as was shown in Figure 5-1). The spaces between the stacked up steel rings were sealed with Teflon tape (Figure 5-4). The assembly was then axially compressed with hand pressure and fixed using a 2 stainless steel clamp (Figure 5-4d shows the final setup). These prepared units were then submerged into distilled water baths at 23±1 C for water imbibition and gel swelling. These units were capable of allowing water to penetrate into the ASR gel through the cement paste disks (Figure 5-4b), while the cement paste disks could slide out without friction as the free expansion of gels occurred. 133

152 ( ( ( ( Figure 5-4. The 2-D and perspective views of the designed ASR gel disk swelling units: a) a 2-D scheme of the assembly items, b) water penetration and swelling model, c) the schematic 3-D view, d) the actual view of the finalized assembly The free swelling of the gels was measured using a digital comparator with accuracy of mm (Figure 5-5). The measurements were taken at 1, 4, 7, 14 and 28 days since submerging inside the water bath. 134

153 Figure 5-5. Digital comparator used for measuring the free swelling strain of ASR gel units Swelling pressure experiments on ASR gel disks After the free swelling tests were concluded, the swelling pressure of the swollen gel specimens were determined by reversing the occurred swelling in a drained configuration (Figure 5-6). At its ultimate free swelling strain (εg,fr), the swelling pressure of each gel approaches zero. At this point, application of a compressive stress reverses the swelling in the gel. As the gel continues to be compressed, it builds up resistance against further compression, since the gel s internal solid structure carries a progressively larger portion of the applied compressive stress. The required applied stress is the highest when the entire swelling is reversed (when the residual swelling strain is zero as shown in Figure 5-6b). That maximum stress, shown as Prs, is considered to be an 135

154 estimate of the fully restrained swelling pressure of each synthetic gel at its original water content and dimensions. (a) (b) Specimen Unif orm Figure 5-6. The reversed swelling strain experiment: (a) the loading test setup; (b) typical results of the experiment As such in this study, after the gels reached their ultimate free swelling strains (εg,fr) in the free swelling test described earlier, the swollen units were placed in a MTS load frame and loaded in compression in a deformation-controlled configuration (Figure 5-6a). Using this load frame, the occurred swelling strain in each specimen was slowly reversed, and the required compressive stress (pressure) was recorded at quarter increments of the free swelling strain (i.e., ¼εg,fr, ½εg,fr, ¾εg,fr and εg,fr, which means full reversion). Since drainage of the gel s pore solution through the cement paste pores is a time dependent process, at each quarter increment, the load frame was paused and the pressure was allowed to reach equilibrium before recording. Plotting the obtained 136

155 stress and strain results produced reversed swelling pressure profiles; an example is shown in Figure 5-6b Design of Experiments To obtain consistent results and develop reliable regression models that characterize the swelling properties of ASR gels, a robust statistical design of experiments (DOE) was employed. A DOE algorithm, based on the response surface methodology, was developed that was capable of capturing and quantifying the linear and non-linear effects and interactions between Ca/Si, Na/Si, and K/Si, and their impacts on the swelling behavior of ASR gels. Details of this DOE algorithm are provided elsewhere [1], but the essential information are included below. Table 5-1 lists the coded and natural levels of the 20 gel compositions that was investigated in this study, along with their corresponding labels. The run order of the experiments was randomized to eliminate any systematic effect of the mixing sequence, operator s learning curve, and/or testing time on the results. In Table 5-1, the coded levels of 1.682, 1, 0, +1 and were labeled as low (L), intermediate low (IL), intermediate (I), intermediate high (IH) and high (H), respectively. Such low and high levels were assigned to the low and high natural levels of each chemical composition variable. For instance, the coded levels and were assigned to the natural levels 0.05 and 0.5 (atomic ratio) of Ca/Si. Other natural levels of Ca/Si were calculated for the remaining three coded levels (i.e., 1, 0, +1) via linear interpolation. C, N and K were used to represent Ca/Si, Na/Si and K/Si atomic ratios in the gels. The gels were labeled by assigning the corresponding dosage level acronyms (e.g., I, IL, etc.) to C, N and K as lowercase suffixes. For instance, CINHKI represents a synthetic ASR gel having intermediate levels of Ca/Si and K/Si (= and 0.15, respectively), and a high level of Na/Si (= 1.0). 137

156 Table 5-1. The experimental design with the coded and natural levels of the gel composition variables Variable Ca/Si Na/Si K/Si Run order Point type Label Coded Natural Coded Natural Coded Natural 5 C IL N IL K IL C IH N IL K IL C IL N IH K IL C IH N IH K IL C IL N IL K IH C IH N IL K IH C IL N IH K IH C IH N IH K IH C L N I K I C H N I K I C I N L K I C I N H K I C I N I K L C I N I K H , 8, 12, 13, 14, 17 Cube points Axial points Center points C I N I K I (1) to (6) respectively Table 5-2 shows the mixture proportions of the 20 gel mixtures produced according to the adopted DOE. Notice that the source of colloidal silica used in these experiments contained an initial Na/Si atomic ratio of 0.022, which was taken into account in calculating the required NaOH content. Moreover, the water embedded in Ca(OH)2, NaOH and KOH was subtracted from the total water content listed in the last column of Table 5-2. Table 5-2. Mixture proportions of the gels Run order Label Ca(OH)2 NaOH KOH Silica powder H2O (gr/1000 gr) (gr/1000 gr) (gr/1000 gr) (gr/1000 gr) (gr/1000 gr) 1 C IL N IH K IH C IL N IL K IH , 8, 12, C I N I K I (1) to 13, 14, 17 (6) C I N L K I C IL N IL K IL C H N I K I C I N H K I C IH N IL K IH C IH N IH K IH C I N I K H

157 15 C L N I K I C IH N IH K IL C IL N IH K IL C IH N IL K IL C I N I K L RESULTS AND DISCUSSION Free swelling results Figure 5-7 shows the results of the free swelling test of the ASR gel disks. A total of 20 curves in three different types can be observed in this graph: solid lines, dashed lines, and dash-dotted lines. The 6 dash-dotted lines correspond to the gels at the center points of the experiments according to Table 5-2; CINIKI (1) to (6), which were used to quantify the experiment s repeatability. CINIKI (1) to (6) have Ca/Si, Na/Si and K/Si atomic ratios at the center levels of the investigated ranges of chemical composition (i.e., 0.275, 0.55 and 0.15, respectively). It is observed that this composition does not produce gels with outstandingly high swelling capacity (ε g,fr <1.9% by 90% confidence). The dashed lines in Figure 5-7 correspond to the axial points (see Table 5-1 for their full information), which in comparison to the center points, have one of the three composition variables at its high (H) or low (L) level, while the others remain at their intermediate (I) levels. Comparing the free swelling results of the axial points to the center point reveals the effect of change in any of the three composition variables (from its intermediate level to high or low level) on free swelling capacity of the gels. For instance, comparing the 28-day free swelling strain result of the 15 th experimental point (gel CLNIKI) to the center point (gels CINIKI) and then to the 6 th experimental point (gel CHNIKI) suggests that increases in Ca/Si from 0.05 to and from to 0.5 reduce the free swelling capacity of the gel by a factor of 16.9 and 2.44 in each step. Therefore, the 139

158 presence of calcium in low concentrations in medium-alkali gels renders a gel with large free swelling capacity; and as the calcium content increases, swelling capacity is diminished. This phenomenon can be generally justified by noting that calcium is a divalent cation that forms bonds with two adjacent silica tetrahedrons and therefore, disables swelling of the gel in its vicinity. Figure 5-7. The ASR gel disks free swelling strain test results 140

159 As the calcium content increases, the gel gradually becomes non-swelling. It should be noted that the reduction in free swelling with increased Ca/Si is observed when the other two composition variables (i.e., Na/Si and K/Si) are at their intermediate level. This observation may not necessarily be valid for gels with very low or very high alkali content. As such, a robust multiple regression model, established using the entire DOE output dataset, is needed to account for the complex interactive nature of the composition variables and to reliably estimate the swelling capacity of the gels as a function of variations in their chemical composition. Unlike calcium, sodium and potassium are known to be silica network breakers and they are expected to enhance the swelling capacity of the gels. Comparing CINLKI to CINIKI and CINHKI, it is observed that increasing Na/Si from 0.1 to 0.55 and from 0.55 to 1.0 results in 10.3 and 8.8 times increases (in each step) in the swelling capacity of the gels at 28 days. Similarly, when K/Si increases from 0.0 to 0.15 and from 0.15 to 0.3, swelling capacity of the gels are observed to increase by 1.1 and 3.2 times in each step. Note that this is not an indication of weaker effect of potassium on swelling, since the variations in the levels of K/Si are within a smaller range (0.0 to 0.3) than those of Na/Si (0.1 to 1.0). Finally, the third category of lines on Figure 5-7 belongs to the cube points (per Table 5-1). These experimental points are planted in the response surface design to enable the regression model to estimate the interactions between different pairs of variables (e.g., interactions between Ca/Si and Na/Si). The interpretation of the results obtained for these data points is more complicated than that of the axial and center points. An interaction between two variables, by definition, is the presence of a significant change in the effect of one variable (on the response parameter) as a function of the level of the other variable. Therefore, if the already interpreted increasing or decreasing effects of each of the three composition variables on the free swelling of 141

160 the gels is significantly altered as the level of another composition variable changes, then one can conclude that an interaction between such composition variables exists. The significance of interactions can be determined by examining the regression function and finding out whether the product terms (e.g., Ca/Si Na/Si) have a meaningful influence on the variations in the free swelling strain observations Swelling pressure results Figure 5-8a-d show the summary of the swelling pressure results for the 20 synthetic ASR gels. Figure 5-8a shows the swelling pressure results of 5 of the 6 gels at the center points of the experiments according to Table 5-1; CINIKI (1), and (3) to (6). CINIKI (2) was identified to be an outlier and, hence, removed from the plot. The results obtained in this graph are averaged and presented in Figure 5-8b along with the 90% confidence interval bars of each pair of average strain and stress. It is observed that the average fully restrained swelling pressure (Prs) of this gel composition (with 40% water content) is MPa and it can conceivably be as high as MPa. This is a negligible pressure compared to the tensile strength of concrete, often ranging from 2 to 5 MPa. Figure 5-8c shows the swelling pressure results of the gels at the axial points of the designed experiments (per Table 5-1), where one of the three composition variables is either at its high or low level, while the other two variables remain at their intermediate levels. It is observed that an increase in the calcium content leads to a decrease in the swelling pressure. This can be substantiated by comparing the restrained swelling pressure results of CLNIKI to CINIKI (Prs is reduced by a factor of 1.38) and comparing CINIKI to CHNIKI (Prs is reduced by a factor of 29.0). Also, comparing the gels CINLKI to CINIKI (Prs is increased by a factor of 4.38) and gels CINIKI to CINHKI (Prs is increased by a factor of 11.9) reveals that an increase in the sodium content not 142

161 only increases the free swelling strain, but it also drastically increases the restrained swelling pressure of the gels. Same behavior was found in the case of potassium where increases in K/Si from its low (0.0) to intermediate (0.15), and from intermediate to high (0.3) levels resulted in increase in Prs by factors of 2.13 and 2.24 in each step. Figure 5-8d shows the swelling pressure results of the gels at the cube points of the designed experiments. It is observed that the restrained swelling pressures (Prs) can exceed 2MPa for some gels. It is also worth noting that the swelling pressure increases with reducing the gel s moisture content as the water is squeezed of the gel and the residual swelling strain approaches zero. Given that the gels in this study were synthesized with an initial moisture content of 40%, it is likely that higher swelling pressures would be measured had the gels been synthesized at lower moisture content. In other words, it is reasonable to conclude that the swelling pressure of gels is inversely related to their moisture content. While both CIHNIHKIL and CIHNIHKIH show similar Prs, the one with high potassium seems to be able to maintain its swelling pressure over a wider range of swelling strains, which is also important in terms of continuation of ASR expansion as the gels absorb water. Moreover, although CILNIHKIH shows a slightly less Prs compared to the above two gels, it is able to preserve its pressure over a much broader range of swelling strain. This is likely because of the network binding effect of calcium, which stiffens yet limits the swelling strain of the gels. Further discussions on the effects of chemical composition on the swelling pressure of the gels are provided after the regression model is developed. 143

162 144 Figure 5-8. The swelling pressure of ASR gels as a function of their residual swelling strain

163 Regression analyses Building the regression models Using Minitab software, the obtained test results of the free swelling strain and restrained swelling pressure of the 20 synthetic gels were compiled into two regression models. The response parameters of the two models were the 28-day free swelling strain (εg,fr (%)) and the restrained swelling pressure (Prs (MPa)). These response parameters were analyzed with respect to the three main composition variables of the experiments (i.e., Ca/Si, Na/Si and K/Si). The significance of each composition (input) variable was statistically evaluated by finding out how much of the total variations in the response parameters can be assigned to the variations in each predictors (i.e., the composition variables (i.e., Ca/Si, Na/Si and K/Si) and their interactions such as Ca/Si Na/Si, etc.)). Such variations are also normally compared to the unexplained variations in the response (i.e., the Error term). A test statistic was obtained by dividing the mean of the squared variations due to each predictor (numerator) to that of the Error (denominator). Such test statistic (referred to as the F-Value) can be demonstrated to have an F-distribution having the degrees of freedom of the variable and the Error term as its parameters. The magnitude of that test statistic was assessed according to the F-distribution mass function, and the probability of obtaining a test statistic that large while the numerator is no different than the denominator was obtained. This essentially is equivalent to the chance of mistakenly deeming a predictor significant while it is indeed insignificant. If such probability (which is referred to as the P-Value) is greater than 5%, then it is concluded that the given predictor does not have a significant effect on the response. The sources of variations in the responses can be the linear and nonlinear sole and combined effects of the main three composition variables. Among all possible terms, only the 145

164 following can be estimated based on the adopted design of experiments: Xi, Xi 2, Xi 3, and Xi Xj, where each Xi can be Ca/Si, Na/Si or K/Si. Table 5-3 shows the shorthand regression ANOVA table for εg,fr and Prs with respect to such terms. For each term, five values are reported: Sequential mean of squared (MS) variations, the F-Value, Contribution (%), Coefficient, and the P-Value. The numerical ratio of the sequential MS of each predictor (i.e., the sequential sum of squares of each predictor divided by its degrees of freedom, where the sequential sum of squares is the sum of squared variations in the response due to the variations in the predictor while the effects of the preceding predictors in the regression analysis are already taken into account) to that of the Error term constitutes the F-Value of that predictor, from which the P-Value is calculated. The Contribution of each predictor is indicative of the percentage of variations in the response parameter that is solely explained by variations in that specific predictor. So, it can be seen, for instance, that Na/Si has the greatest contributions for both response parameters and it is the most significant factor in that sense. Finally, the Coefficient of each predictor is the numerical multiplying factor that appears before each predictor in the regression function. For each regression model, the Error term is separated into two categories: the Lack-of-Fit (LoF) and the Pure Error (PE). The PE represents the total sum of squared differences between observed repeats from their average, which here simply means the variance of the observed values for the six replicates of the gel CINIKI. The LoF, however, represents the sum of squared differences between the fitted values (from regression equations) and the average experimental results. Therefore, the total error, which is the sum of the squared differences between the observed values and fitted values, can be expressed in terms of the sum of PE and LoF. If PE is greater than LoF, one could conclude that the occasional inaccurate predictions of the model are 146

165 due to the inherent variations in the data as opposed to a Lack-of-Fit. As such, the F-Value of LoF with respect to PE is calculated to find out the P-Value or the probability of presence of a meaningful lack of fit. It is observed that neither of the two models are showing any sign of lack of fit. Table 5-3. The shorthand ANOVA table of the free swelling strain and restrained swelling pressure of the ASR gels Response 28 day free swelling strain ε g,fr (%) Restrained swelling pressure Prs (MPa) Sequential MS a F Value Contribution (%) Coefficients Predictor Ca/Si Na/Si K/Si (Ca/Si) (Na/Si) 2 b (K/Si) 2 b b Ca/Si Na/Si b Ca/Si K/Si b Na/Si K/Si b (Ca/Si) b (Na/Si) 3 b (K/Si) 3 b b Error Lack-of-Fit c d Pure Error Prediction accuracy Regular R 2 Adjusted R 2 P Value Predicted R 2 Sequential MS a Regular R 2 F Value Contribution (%) Adjusted R 2 Coefficients P Value Predicted R % 93.08% a The sequential mean of squared variations in response due to variations in each term b Not included in the model due to lack of significance. The calculations are hence not available. c Lack-of-Fit degrees of freedom=9 d Lack-of-Fit degrees of freedom=4 147

166 It should be noted that, as presented in Table 5-3, not all predictors showed significant influences on the response parameters. The quadratic and cubic terms of Na/Si and K/Si as well as the interactions of all three terms were found to be insignificant and were removed from the εg,fr regression model. The regression functions for εg,fr and Prs are provided as Eqs. 5-1 and 5-2. The reason why the response terms are transformed into the logarithmic forms (which is called the Box-Cox transformation) is so that the residuals of the regression functions predictions satisfy the core assumptions of the regression analysis, which are: (1) to have a normal distribution, and (2) to have a constant variance across all observations. The last two rows in Table 5-3 show the goodness-of-fit for the regression functions through the Regular, Adjusted and Predicted R 2. The differences and significance of each R 2 type is discussed in a previous work [1]. It is observed that both models are very well capable of predicting the swelling properties of the existing (i.e., experimentally measured gel compositions) and new (i.e., not measured) observations. This suggests that the models can be used for predictive applications by other researchers. Note that the proposed regression equations can only be safely used in the studied ranges of compositions (i.e., inside the DOE sphere space). In other words Eqs. 5-1 and 5-2 yield reliable results as long as the chemical composition of the gels satisfies Eq ln(ε g,fr (%)) = Ca Si Na Si K Si ( Ca Si ) ( Ca Eq. 5-1 )3 Si ln(p rs (MPa)) = Ca Si 45.5 Na Si 2.1 K Si 35.6 ( Ca Si ) ( Na Si )2 + Eq ( Ca ) Si (Na Si ) 61.8 (Ca Si ) (K Si ) (Na Si ) (K Si ) 43.2( Na Si )3 148

167 Ca Si (( ) Na Si + ( ) 0.45 K Si + ( ) ) Eq. 5-3 Outcomes of the regression models In order to better understand the effect of each predictor on the free swelling of ASR gels, the predicted εg,fr is plotted as a function of each composition variable and at different levels of the other two other variables (Figure 5-9 to Figure 5-11). For example, Figure 5-9 shows such a plot of predicted εg,fr as a function of Ca/Si and for different combinations of Na/Si (N) and K/Si (K). Before interpreting the plot, it should be noted that among all possible 5 5=25 combinations of N and K (each having five distinct levels: L, IL, I, IH, and H), only 9 combinations are shown on the plot. This is because only these combinations fall inside the spherical space of the DOE. The remaining 16 (e.g., NLKL, NHKH, etc.) fall outside the DOE s spherical domain, and hence, not explored. For the plotted combinations, the curves are drawn as solid lines with markers in the range of Ca/Si that falls inside the DOE space. For instance, in the case of gels with sodium and potassium at their intermediate-high levels (i.e., NIHKIH), the gels composition is within the experimental space (and satisfies Eq. 5-3) as long as Ca/Si ranges from IL to IH (i.e., to 0.409). Gels with Ca/Si outside this range are not inside the DOE space and are plotted as dotted line with no markers. One should only use the model predictions for such out-of-range compositions with caution. Figure 5-9 suggests that the effect of Ca/Si on the free swelling strain of the ASR gels is highly non-linear. Ca/Si tremendously reduces the free swelling of the gels as Ca/Si increases from 0.05 up to Then, free swelling plateaus or slightly increases as Ca/Si increases from 0.23 to The free swelling is diminished once Ca/Si increases beyond Similar trends 149

168 Free swelling strain (εg,fr (%)) were observed in the previous work [1] for the effect of Ca/Si on the free swelling of ASR gel minibars in contact with water vapor Ca/Si atomic ratio NIHKIH NIHKIL NIKIH NIKI NIHKI NIKIL NILKIH NILKI NILKIL Figure 5-9. The effect of Ca/Si atomic ratio of the ASR gels on their (predicted) free swelling strain at different experimented levels of Na/Si and K/Si Determining the causes of such multi-episode effect of Ca/Si on the swelling behavior of the ASR gels requires extraction and analysis of the pore solutions of the gels as well as studying the atomic structure of the gels (for example using NMR or Raman spectroscopy). While the latter was beyond the scope of the present study, we made extensive attempts to extract and analyze the pore solutions of the synthesized gels to investigate their composition and osmotic pressure. Unfortunately, our attempts in extraction pore solution of gels were unsuccessful due to the gels low water content, high hydrophilicity, and polymeric nature. As such, the following discussions 150

169 on the observed gel behaviors lend themselves only to the technical speculations based on previously established knowledge on calcium silicate hydrates. In high-ph sodium silicate systems, the majority of the dissolved and colloidal species have Si O and Si O Na + endings. Calcium as a divalent cation is known to have higher oxygen affinity and field strength compared to sodium [21]. Therefore, calcium has a silicate-binding effect by forming Si O Ca 2+ O Si links. One can hypothesize that increasing Ca/Si in the range from 0 to 0.23 leads to a drop in the water affinity of the gels by satisfying Si O endings and reducing the surface charge of the gels, and such, reducing their tendency for swelling. At low Ca/Si levels, calcium does not confront sodium due to the ample presence of Si O endings. However, when Ca/Si increases beyond 0.23, calcium starts to replace for sodium in the Si O Na + endings to form Si O Ca 2+ O Si links. This phenomenon, originally pointed out by Thomas [22] is known as the alkali recycling effect of calcium. Such alkali exchange leads to a slight increase in the osmotic pressure of the gels pore solution (due to release of Na + into pore solution), which will then increase the gels tendency to absorb water and swell. Once Ca/Si increases beyond 0.37, there is an abundance of calcium in the gel network and it starts to completely bind and interconnect the structure of ASR gel and reduce its free swelling. Extrapolations from the regression analyses suggest that regardless of the Na/Si and K/Si level, once Ca/Si exceeds 0.6, the swelling capacity of the gels is practically eliminated. Figure 5-10 and Figure 5-11 show that Na/Si and K/Si monotonically increase the free swelling of ASR gels across all ranges of Ca/Si. Interestingly, these monovalent alkalis, appear with fairly similar coefficients in the regression function per Eq. 5-1) (i.e., and 5.57). Note that the linear appearance of these two terms in the regression function is not indicative of their linear effects on the free swelling, since the free swelling is expressed in logarithmic scale in this 151

170 Free swelling strain (εg,fr (%)) equation. Therefore, these terms also have nonlinear effects on εg,fr; but unlike Ca/Si, their effects are monotonic, as shown in Figure 5-10 and Figure In Figure 5-10, the increasing effect of Na/Si on free swelling is more pronounced in the case of gels with low Ca/Si and high K/Si. This is because of the network-binding and network breaking effects of calcium and potassium, respectively. Similar trends are observed for the effect of K/Si on free swelling in Figure Na/Si atomic ratio CILKIH CIHKIH CIKIH CILKI CIHKI CIKI CILKIL CIHKIL CIKIL Figure The effect of Na/Si atomic ratio of the ASR gels on their (predicted) free swelling strain at different experimented levels of Ca/Si and K/Si Regarding the restrained swelling pressure of the ASR gels, all three variables (i.e., Ca/Si, Na/Si and K/Si) show nonlinear interactive effects on the response. The effects of each composition variables on the swelling pressure of the gels are plotted at different levels of the other two variables. 152

171 Free swelling strain (εg,fr (%)) CILNIH CIHNIH CINIH CILNI CIHNI CINI CILNIL CIHNIL K/Si atomic ratio CINIL Figure The effect of K/Si atomic ratio of the ASR gels on their (predicted) free swelling strain at different experimented levels of Ca/Si and Na/Si Figure 5-12 shows that Ca/Si does not always have a reducing effect on Prs. Especially in the case of gels with intermediate to high sodium content (NI and NIH gels), increasing Ca/Si first increases Prs, followed by a peak (at a pessimum Ca/Si), and subsequent reduction in Prs with increasing Ca. A similar behavior is observed for KI and KIH gels. A possible explanation for this behavior is as follows. Intermediate to high alkali gels have a low structural connectivity and mechanical strength especially at low Ca/Si levels. Such gels will have high free swelling capacities, but cannot exert much pressure to the surrounding environment due to their soft and flowable structure. The introduction of calcium as a divalent cation enhances their mechanical properties, making them capable of resisting restraint against their swelling (i.e., their swelling pressure will increase). However, once calcium exceeds a pessimum level, it will start to suppress the swelling capacity of these gels. This is why as the Figure 5-12 suggests, Ca/Si has an increasing 153

172 Restrained swelling pressure (Prs (MPa)) effect on Prs of moderate to high alkali gels up to a certain point (i.e., the pessimum Ca/Si level), and beyond that a reducing effect is observed. It should be noted that high-alkali gels (NH and KH) and low-alkali gels (NL and KL) are beyond the DOE sphere and were not plotted on Figure Ca/Si atomic ratio NIHKIH NIHKI NIHKIL NIKIH NIKI NIKIL NILKIH NILKIL NILKI Figure The effect of Ca/Si atomic ratio of the ASR gels on their (predicted) restrained swelling pressure at different experimented levels of Na/Si and K/Si Figure 5-13 shows the effect of Na/Si atomic ratio on the swelling pressure of the gels at different levels of Ca/Si and K/Si. It is observed that as Na/Si increases beyond 0.5, the swelling pressure starts to increase considerably. This is mainly due to contribution of Na to increasing the osmotic pressure of the gels pore solution. However, Na is a network breaker and makes the gel s network softer and less capable of bearing loads. This ultimately leads to a decline in the swelling 154

173 Restrained swelling pressure (Prs (MPa)) pressure past a pessimum Na/Si content of 0.85 or larger. This behavior is observed for all gel compositions shown in Figure Na/Si atomic ratio CIKIH CIHKIH CIHKI CILKIH CIKI CIHKIL CIKIL CILKI CILKIL Figure The effect of Na/Si atomic ratio of the ASR gels on their (predicted) restrained swelling pressure at different experimented levels of Ca/Si and K/Si Note that CIHKIH also has a pessimum Na/Si level, but the peak point occurs out of the chart s range. The position of pessimum Na/Si is affected by the Ca/Si level. The more calcium is present in the gel, the higher the pessimum Na/Si level. Finally, it should be noted that the increase in the Prs as Na/Si approaches 0.1 is not conclusive, because such values of Na/Si are outside the domain of the DOE model. Figure 5-14 shows the variations in the swelling pressure of gels as a function of K/Si and at different levels of Ca/Si and Na/Si. It is observed that the swelling pressure steadily increases 155

174 as the potassium concentration in the gels increases. The rate of increase, however, appears to be highly dependent upon the Ca/Si and Na/Si levels of the gels. It can be seen that potassium has much weaker promoting effect on Prs for gels with low Ca/Si level. This justifies the negative coefficient of the interaction term between these two variables (i.e., Ca/Si K/Si) in the regression function of Prs according to. Unlike Ca/Si, Na/Si shows a synergistic effect with K/Si, which means that the increasing effect of potassium on Prs is more pronounced at higher concentrations of sodium. No pessimum K/Si is observed. However, for intermediate-high Ca/Si gels (CIH), swelling pressure remains small despite increasing K/Si. In order to have a better representation of the effect of ASR gel s chemical composition on its free swelling strain and restrained swelling pressure, a number of easy-to-use contour plots were developed as shown in Figure The three contour plots on the left show the variations in the free swelling strains of the gels as a function of different pairs of the composition variables, while the third variable is held at its intermediate level (e.g., the top graph shows free swelling strain as a function of Ca/Si and Na/Si, while K/Si is held constant at 0.15). Similar contour plots are plotted on the right-hand side of Figure 5-15 for the restrained swelling pressure of the gels. The greyed area on each graph represents the area that is outside the DOE domain. 156

175 Restrained swelling pressure (Prs (MPa)) CINIH CILNIH CIHNIH CILNI CINI CIHNIL CIHNI CILNIL K/Si atomic ratio CINIL Figure The effect of K/Si atomic ratio of the ASR gels on their (predicted) restrained swelling pressure at different experimented levels of Ca/Si and Na/Si Using these contour plots, gel compositions resulting in high εg,fr and Prs can be easily detected. For instance, it can be seen that a gel with Ca/Si=0.40, Na/Si=0.90 and K/Si=0.15 will have a free swelling strain exceeding 10%, and a restrained swelling pressure of approximately 3.0MPa. This gel is likely to be deleterious if formed in concrete. The free swelling contour plots show that the free swelling of ASR gel increases monotonically with increasing its alkali content. Meanwhile, increasing Ca/Si up to 0.23 drastically reduces εg,fr, higher Ca/Si has modest effect on free swelling. The swelling pressure is observed to increase with Ca/Si up to a pessimum Ca/Si level; higher Ca/Si results in lower Prs. This pessimum Ca/Si was found to be related to alkali content of the gel according to Eq

176 (Ca/Si) pess = Na Si K Si Eq. 5-4 Swelling pressure generally increases with the alkali content of the gels. Prs is observed to increase with Na/Si and peak at a pessimum Na/Si in the range 0.85 to 0.95, depending on the Ca/Si content of the gel (higher Ca/Si results in higher (Na/Si)pess). Prs is also observed to increase monotonically (without a pessimum) with increasing K/Si for gels with Ca/Si<0.30. For higher calcium gels, increasing K/Si has a small effect on Prs. Hypotheses explaining such compositionproperty relationships for the studied ASR gels were provided earlier in connection with Figures 9 to 14. Another benefit of the contour plots in Figure 15 is that they aid scientists while studying the microstructure of ASR-damaged concrete under electron microscope. By measuring the composition of the formed ASR gels using x-ray microanalysis (EDS or WDS), one can make an in situ estimation of the free swelling strain and the swelling pressure of these gels, and as such their degree of deleteriousness. If we assume that ASR gels with swelling pressures greater than 2MPa and free swelling strains exceeding 10% are deleterious, using the regression functions per Eqs. 5-1 and 5-2, we can show that the gels with compositions falling into the specified region of Figure 5-16 satisfy these two criteria and are, hence, deleterious in terms of ASR damage. 158

177 Figure The contour plots of free swelling strain (εg,fr) and swelling pressure (Prs) as a function of Ca/Si, Na/Si and K/Si 159

178 Figure The composition domain of deleterious ASR gels: (a) the 3D surface plot, (b) the 2D contour plot 5.4. CONCLUSIONS In this study, 20 synthetic ASR gels, spanning the range of compositions of typical field gels, were produced and tested to measure their free swelling strain (εg,fr) and restrained swelling pressure (Prs). The test results were used in combination of a robust statistical design of experiments (DOE) 160

179 model to develop regresstion equations for εg,fr and Prs as a function of the gel composition. These regression functions can adequately predict the existing and new experimental data points and, as such, can be used for future modeling and predictive applications. The following conclusions can be drawn based on the outcomes of testing and regression analysis: Increasing the Na/Si and K/Si of the gels (in the atomic ratio ranges of 0.1 to 1.0, and 0.0 to 0.3, respectively) monotonically increases the free swelling strain of the ASR gels. Such increases in free swelling are more pronounced at lower levels of Ca/Si and higher levels of the other alkali component. Increasing Ca/Si up to 0.23 drastically reduces εg,fr, while higher Ca/Si has modest effect on free swelling. Significant non-linear interactions exist among Ca/Si, Na/Si and K/Si on their effects on the swelling pressure of ASR gels. Swelling pressure increases with Ca/Si up to a pessimum Ca/Si level; higher Ca/Si results in lower Prs. This value of pessimum Ca/Si is related to alkali content of the gel. Sodium (Na/Si) increases the swelling pressure of ASR gels up to a pessimum Na/Si level, followed by reducing Prs. The value of (Na/Si)pess is in the range 0.85 to 0.95, and depends on the Ca/Si content of the gel (higher Ca/Si results in higher (Na/Si)pess). Prs also increases monotonically (without a pessimum) with increasing K/Si for gels with Ca/Si<0.30. For higher calcium gels, increasing K/Si has a small effect on Prs. 161

180 5.5. REFERENCES [1] Gholizadeh Vayghan, A., Rajabipour, F., & Rosenberger, J. L. (2016). Composition rheology relationships in alkali silica reaction gels and the impact on the Gel's deleterious behavior. Cement and Concrete Research, 83, [2] Hou, X., Kirkpatrick, R.J., Struble, L.J., & Monteiro, P.J. (2005). Structural investigations of alkali silicate gels. Journal of the American Ceramic Society 88(4): [3] Krogh, H (1975). Examination of synthetic alkali-silica gels. In: Symposium on Alkaliaggregate Reaction, Preventive Measures, Reykjavik, Iceland: [4] Hou, X., Struble, L. J., & Kirkpatrick, R. J. (2004). Formation of ASR gel and the roles of CSH and portlandite. Cement and Concrete research, 34(9), [5] Farny, J. A., & Kerkhoff, B. (2007). Diagnosis and control of alkali-aggregate reactions in concrete. Portland Cement Association, PCA R&D Serial No. 2071b. [6] Rajabipour, F., Giannini, E., Dunant, C., Ideker, J. H., & Thomas, M. D. A. (2015). Alkali silica reaction: Current understanding of the reaction mechanisms and the knowledge gaps, Cement and Concrete Research, 76, [7] Gholizadeh Vayghan, A., & Rajabipour, F. The influence of the composition of alkali silica reaction (ASR) gels on their hydrophilic properties and free swelling in contact with water vapor, Cement and Concrete Research (in peer review). [8] Stanton, T. E. (1942). Expansion of concrete through reaction between cement and aggregate. Transactions of the American Society of Civil Engineers, 107(1), [9] McConnell, D., Melenz, R. C., Holland, W. Y., & Greene, K. T. (1947). Cement-aggregate reaction in concrete. ACI Journal Proceedings 44 (10). [10] Hester, J. A., & Smith, O. F. (1958). The alkali-aggregate phase of chemical reactivity in concrete Part II: Deleterious reactions observed in field concrete structures, ASTM STP205-EB [11] Idorn, G. M. (1961). Studies of disintegrated concrete-part 1, Progress Report N 2, Committee on Alkali Reactions in Concrete, Danish Nat. Inst, of Build. Res, and Acad. of Tech. Sci, [12] Buck, A. D., & Mather, K. (1969). Concrete cores from dry dock no. 2, Charleston Naval Shipyard, S.C., U.S. Army Engineer Waterways Experiment Station, Vicksburg, Mississippi. [13] Poole A.B. (1975). Alkali-silica reactivity in concrete from Dhekelia Cyprus. Symposium on Alkali-Aggregate Reaction, Proc., Reyjavik, Finland, 101. [14] Gutteridge, W. A., & Hobbs, D. W. (1980). Some chemical and physical properties of beltane opal rock and its gelatinous alkali silica reaction product. Cement and Concrete Research, 10(2),

181 [15] Oberholster, R. E. (1983). Alkali reactivity of siliceous rock aggregates: Diagnosis of the reaction, testing of cement and aggregate and prescription of preventative measures. Alkali in Concrete, Research and Practice, [16] Visvesvaraya, H. C., Mullick, A. K., Samuel, G., Sinha, S. K., & Wason, R. C. (1986). Alkali reactivity of granitic rock aggregates. In Proceedings VIII International Congress on the Chemistry of Cement (Vol. 5, pp ). [17] Šachlová, Š., Přikryl, R., & Pertold, Z. (2010). Alkali-silica reaction products: Comparison between samples from concrete structures and laboratory test specimens. Materials Characterization, 61(12), [18] Renardy, Y. Y. (Ed.). (1996). Advances in Multi-fluid Flows (Vol. 86, p. 45). Siam. [19] Aligizaki, K. K. (2005). Pore structure of cement-based materials: Testing, interpretation and requirements (p. 9). CRC Press. [20] Struble, L. J., & Diamond, S. (1981). Swelling properties of synthetic alkali silica gels. Journal of the American Ceramic Society, 64(11), [21] Shelby, JE (2005). Introduction to glass science and technology. Royal Society of Chemistry. [22] Thomas, M. D. A. (2001). The role of calcium hydroxide in alkali recycling in concrete. Materials Science of Concrete Special,

182 CHAPTER 6: CHARACTERIZATION OF MECHANICAL AND VISCOELASTIC BEHAVIOR OF SYNTHETIC ALKALI SILICA REACTION (ASR) GELS 6 Synthetic ASR gels with varying levels of Ca/Si, Na/Si and K/Si were produced and their viscoelastic behaviors were studied under constant stress values equal to 40% of their ultimate compressive strength. The time-dependent strain results were used to characterize their viscoelastic parameters using Burger s model. The results suggest that for all compositions, the Burger s model can very accurately describe the viscoelastic behavior of ASR gels. The elastic and viscous components of the gels per Burger s model were obtained via regression. Such viscoelastic parameters were then related to the chemical composition of ASR gels and regression models were developed that predict the viscoelastic behavior of ASR gels as a function of their chemistry INTRODUCTION The elastic and rheological behaviors of ASR gels under confining stresses have major impacts on their performance in concrete. Many scientists in the field believe that ASR gel gradually flows 6 This chapter will soon be submitted to the ACI Materials journal. 164

183 into the concrete pore structure and cracks upon formation [1,2,3,4]. However, it was shown in a previous work that ASR gels can tolerate certain amount of shear stress before flowing into the vacancies, a property referred to as the yield stress [5]. Yielding behavior, however, is more clearly observed in the case of rather dilute gels. High solid concentration gels (studied in this paper) do not show any clear yielding or viscous behavior. They are solid materials with viscoelastic behavior and their yield point is where they crush under shear. However, lower shear stresses can also cause viscous deformation, the rate of which has a significant impact on the maximum stress and the rate of stress attenuation in the concrete matrix. Moreover, the elastic modulus combined with the Poisson s ratio (measured using a compressometer) reveals the bulk modulus, which governs the compressibility of the gels under pressure. More compressible gels tend to exert less damage to their surrounding matrix. As such, the study of the Poisson s ratio of the gels is also important. Note that viscoelastic behavior of ASR gels is very diverse depending on composition and also not explainable by most of the viscoelastic models. For instance, in the Kelvin-Voigt model (Figure 6-1a) the deformation plateaus after a certain amount of time while the gels were found to continue deforming and diverged from Kelvin-Voigt model s predictions as the test progressed. Moreover, this model does not allow for any instant deformation under loading, while the opposite was found for the gels. The Maxwell model (Figure 6-1b) also assumes a constant rate of deformation while the gels were found to show decreasing deformation rates. The combination of these two models constitutes the Burger s model (Figure 6-1c), which was found to successfully predict the viscoelastic behavior of all gels for their entire test durations. 165

184 (a) (b) (c) Figure 6-1. Different linear viscoelastic models examined for characterization of the rheological behavior of ASR gels: (a) Kelvin Voigt, (b) Maxwell model, and (c) Burger s model The objective of this study is to (1) examine the suitability of the Burger s model in predicting the viscoelastic response of ASR gels to external loading, and (2) relate the chemistry of ASR gels to their compressive strength, viscoelastic parameters (i.e., E1, E2, η1 and η2) per Burger s model, and Poisson s ratio. E2 represents the instant elastic response of the material to external load and is hence referred to as instant elastic modulus of ASR gels. The elastic component E1 takes effect long after the instant response of the material to stress and is as such denoted as the long term elastic modulus of ASR gels. Similarly, while viscous effect of η1 is dissipated as the long term elastic component takes effect, while the component η2 continues to cause viscous deformation in 166

185 the material. Therefore η1 and η2 are referred to as the short term and the long term viscosities of ASR gels EXPERIMENTAL PROGRAM Materials ASR gels are silica based hydrogels incorporating varying amounts of alkali and alkaline earth species (primarily sodium, calcium, and potassium). Their general chemical composition can be represented as (SiO2).(CaO)c.(Na2O)n(K2O)k.(H2O)x. Synthetic ASR gels can be produced by carefully combining certain precursors containing the different chemical components present in the composition of the gels. The precursors used for the synthesis of the gels are described below. Dried, pulverized colloidal silicon dioxide IV (50% H2O, procured from Alfa Aesar) with an average particle size of 6.37μm was used as the source of amorphous silica. The colloidal silica was dried at 100 C for 48 hours, followed by 48 hours of grinding inside a 5-L porcelain jar mill. The source of silica contained an initial Na/Si of 0.022, which was taken into account in calculating the required NaOH content for synthesizing each gel. Ca(OH)2 powder (assay: >95%) was used as the source of calcium. NaOH and KOH pellets (99% assay) were used as the sources of sodium and potassium. The water embedded in Ca(OH)2, NaOH and KOH was subtracted from the total mixing water needed to reach the target moisture content as described in the next section. 167

186 Design of experiments (DOE) Similar to the previous works [6,7] the ranges of composition studied in this work are as follow: Ca/Si = ( ), Na/Si = ( ), and K/Si = ( ) atomic ratios. These ranges were concluded to cover almost all ASR gel (extracted from damaged concrete members) compositions reported in the literature [6] and are as such studied for getting a full picture of the effect of chemistry on their viscoelastic behavior. The experiments were designed based on the concepts of the central composite design in order to enable statistical assessment of the nonlinear effects and interactions of ASR gels chemical components (Ca/Si, Na/Si and K/Si) on their viscoelastic properties. To that end, each chemical component was studied at five levels as shown in Table 6-1. Coded levels Table 6-1. The experimented levels of different chemical components of ASR gels Categorical Low Intermediate Intermediate Intermediate (L) Low (IL) (I) High (IH) High (H) Numerical Test variables C o r r e s p o n d i n g n a t u r a l l e v e l s Ca/Si: C Na/Si: N K/Si: K ASR gels with 20 different chemical compositions as shown in Table 6-2 were synthesized. The moisture content was held constant at 40%. The gels were labeled based on the levels of their different chemical components. After designing the experiments, the run order was randomized to eliminate the effects of testing time on the results. Following the concepts of the central composite design, six identical gels were produced and tested for C I N I K I as replicates of the experiments and in order to increase the power of the statistical analyses and obtain an estimate of the pure error. 168

187 Table 6-2. Chemical compositions and mixture proportions of the synthesized ASR gels Run order Label Ca/Si Na/Si K/Si Ca(OH) 2 NaOH KOH Silica H 2 O 1 C IL N IH K IH C IL N IL K IH , 8, 12, 13, 14, 17 C I N I K I (1) to (6) C I N L K I C IL N IL K IL C H N I K I C I N H K I C IH N IL K IH C IH N IH K IH C I N I K H C L N I K I C IH N IH K IL C IL N IH K IL C IH N IL K IL C I N I K L * The masses of the chemical components are in grams per each 1000 grams of gel Synthesis of ASR gels A temperature-controlled high-shear mixing protocol was developed at Penn State for batching the gels. In this method, first the alkalis (KOH and NaOH) were dissolved in the mixing water and stored at 1 C. The silica and calcium hydroxide were dry-blended and stored at -18 C. After 3 hours, the alkaline solution and powder mix were removed from the fridge and mixed as follows. 169

188 The alkaline solution was first transferred to a cylindrical PVC container and the powder was immediately added to the solution. Using a hand-held drill attached with a tri-blade paddle, the materials were mixed at 60 to 100 rpm for 15 seconds or until the content converted to a paste. Then, the paste was mixed at 1,000 to 1,500 rpm for 90 seconds to obtain a fully homogeneous paste. The obtained mixture was then transferred into the predesignated molds for curing and testing as described in the next section Compressive strength measurement The rationale behind testing the gels for their strength was to find out the magnitude of the load that should be applied to the gels for the measurement of their viscoelastic properties and Poisson s ratio. Therefore, after the mixing was completed, the gels were poured into 2 4 cylindrical molds and vacuum cured for two months before testing for their compressive strength (σc). It was attempted to apply the loading in such a rate that the specimens break in 1 to 2 minutes Viscoelastic properties and Poisson s ratio measurements A 2 4 cylindrical specimen was cast and cured for testing the gels for their viscoelastic properties. The specimens were instantly loaded up to a stress level of 0.4σc and the load was held constant at that levels for sufficient duration of time to collect enough time-dependent deformation data for modelling their behavior using Burger s model. The specimens were covered with paraffin wax tape to prevent drying. It was crucial to not allow for development of capillary forces, since they could create significant tri-axial restraining stresses which would foul the test results. Also, 170

189 the Poisson s ratios of the gels were measured per ASTM C469 during the monotonic loading of the specimens Closed form analyses for determination of viscoelastic behavior of the gels per Burger s As shown in Figure 6-1c, the overall strain of the gels is the sum of strains occurred in the three parts of the Burger s model (see Eq. (6-1)). The closed form formulas of these strains (ε 1, ε 2, ε 3 ) that are otherwise functions of time are obtained below (under constant stress condition). ε(t) = ε 1 (t) + ε 2 (t) + ε 3 (t) Eq. (6-1) Finding ε 1 (t): According to Figure 6-1c: σ(t) = σ 1 (t) + σ 1 (t) = E 1. ε 1 (t) + η 1. dε 1(t) dt dε 1(t) dt = σ η 1 E η 1. ε 1 (t) By solving the differential equation and applying the boundary condition of ε 1 (t): ε 1 (t) = σ E 1. (1 exp ( E 1 η 1. t)) Eq. (6-2) ε 2 (t) = σ η 2. t Eq. (6-3) ε 3 (t) = σ E 2 Eq. (6-4) ε(t) = σ E 1. (1 exp ( E 1 η 1. t)) + σ η 2. t + σ E 2 Eq. (6-5) 171

190 Numerical analyses for determination of viscoelastic parameters of the gels per Burger s model: The normal load and deflection results were reported by the machine each 0.1 sec. Using the initial length and the cross-sectional area of the specimens, the load and deflection results were converted to normal stress (constant at 0.4σc) and strain. The optimum E1, E2, η1 and η2 values are the one which minimize the discrepancy between the observed strain values and those suggested per Eq. (6-5). It would be beneficial for optimization purposes if not only the total differences between such strain values are minimized, but also the differences between the strain rate values (as suggested by the machine and as per Eq. (6-6)) are minimized. Using MS Excel software s Solver Add-in, the viscoelastic parameters were optimized by minimizing the sum of (1) total squared differences between the strain values (i.e., as measured by the machine and calculated per Eq. (6-5)) and (2) total squared differences between the strain rate values (as calculated based on the machine output and calculated per Eq. (6-6)) for all data points. In other words, if the deformation value reported by the machine at the time ti is equal to δ(ti), the total squared differences for the strain values and for the strain rate values are calculated per Eq. (6-7) and Eq. (6-8), respectively. dε(t) dt = σ η 1. exp ( E 1 η 1. t) + σ η 2 Eq. (6-6) SSD 1 = (ε(t i ) ( σ. (1 exp ( E 1. t E 1 η i )) + σ. t 1 η i + σ )) 2 E 2 i 2 Eq. (6-7) 172

191 SSD 2 = (ε (t i ) ( σ. exp ( E 1. t η 1 η i ) + σ )) 1 η 2 i 2 Eq. (6-8) Where: L 0 ε(t i ) = ln ( L 0 δ(t i ) ) Eq. (6-9) L 0 = initial length of the specimen ε (t i ) = ε(t i+1) ε(t i 1 ) t i+1 t i 1 Eq. (6-10) SSD2 was rescaled (by a proper scaling factor: SF) to have the same order of magnitude as SSD1. They were then added up and multiplied by a factor to constitute a large number. The Solver was then asked to minimize SST (per Eq. (6-11)) using the Generalized Reduced Gradient (GRG) method by changing E1, E2, η1 and η2 from their arbitrarily selected initial values. SST = (SSD 1 + SF SSD 2 ) Eq. (6-11) After the Solver found such optimum values, the precisions of the obtained values were assessed by finding the deviations from the optimal values needed to induce 10% increase in the SST value. The lower and higher bounds of each variable that are associated with 1.1 SST constitute their precision intervals. In other words, the variable was changed (both reduced and increased) in small increments from each variable s optimal value and the increase SST was monitored. The largest value (smaller than optimal) that was associated with a total sums of squares no more than 10% above the original SST was considered as the lower bound of that variables estimation. Similarly, the smallest value (larger than optimal) that corresponded to no more than 1.1 SST was considered to be the upper bound of the precision interval of that variable. Figure 173

192 6-2 shows the typical pattern of variations in the SST as a function each viscoelastic parameter, and illustrates how the lower and upper bounds are determined. Figure 6-2. The typical SST variations as a function of the viscoelastic parameters, and their precision intervals Regression analyses for modelling the mechanical and viscoelastic parameters to ASR gels chemistry After obtaining the optimal values for the mechanical (compressive strength and Poisson s ratio) and viscoelastic parameters (E1, E2, η1 and η2) of the 20 experimented gel, separate regression models were developed to correlate such parameters with ASR gels chemistry. Minitab software was used for regression analyses. The significant predictors (the linear, 2 nd and 3 rd order terms as well as the interactions between the chemical components) were identified and separate regression equations were developed for each viscoelastic parameter. 174

193 6.3. RESULTS Compressive strength The compressive strength test results are shown in Figure 6-3. It is observed that depending on composition, the compressive strength can be as low as 0.5 MPa (in the case of gel with low Ca/Si) or as high as 11.3 (in the case of high Ca/Si or low Na/Si). For assessing the repeatability of this experiment, the compressive strength results of the five (out of six) available replicates (i.e., C I N I K I (1) (3), (5) and (6)) can be statistically analyzed. C I N I K I was found to have an average of 5.87 MPa with a standard deviation of 1.36 MPa, which translated to a coefficient of variation equal to 23.2%. The 95% confidence interval of the average strength is (4.18 MPa 7.56 MPa). Figure 6-3. The compressive strength results of the synthesized ASR gels 175