Modeling the Formation of Alkali Aluminosilicate Gels at the

Article pubs.acs.org/Langmuir

Modeling the Formation of Alkali Aluminosilicate Gels at the Mesoscale Using Coarse-Grained Monte Carlo Kengran Yang and Claire E. White* Department of Civil & Environmental Engineering and Andlinger Center for Energy and the Environment, Princeton University, Princeton, New Jersey 08544, United States S Supporting Information *

ABSTRACT: Alkali-activated materials (AAMs) are currently being pursued as viable alternatives to conventional ordinary Portland cement because of their lower carbon footprint and established mechanical performance. However, our understanding of the mesoscale morphology (∼1 to 100 nm) of AAMs and related amorphous aluminosilicate gels, including the development of the three-dimensional aluminosilicate network and nanoscale porosity, is severely limited. This study investigates the structural changes that occur during the formation of AAM gels at the mesoscale by utilizing a coarsegrained Monte Carlo (CGMC) modeling technique that exploits density functional theory calculations. The model is capable of simulating the reaction of an aluminosilicate particle in a highly alkaline solution (sodium hydroxide or sodium silicate). Two precursor morphologies have been investigated (layered alumina and silica sheets mimicking metakaolin and spherical aluminosilicate particles reminiscent of coal-derived fly ash) to determine if the precursor morphology has an impact on the structural evolution of the resulting alkali-activated aluminosilicate gel. The CGMC model can capture the three major stages of the alkali-activation processdissolution, polycondensation, and reorganizationrevealing that the dissolved silicate and aluminate species, ranging from monomers to nanoprecipitates (100s of monomers in size), exist in the pore solution of the hardened gel. The model also reveals that the silica concentration of the activating solution controls the extent of dissolution of the precursor particle. From the analysis of the aluminosilicate cluster size distributions, the mechanisms of AAM gel growth have been elucidated, revealing that Ostwald ripening occurs in systems containing free silica at the start of the reaction. On the other hand, growth of the hydroxide-activated systems (metakaolin and fly ash) occurs via the formation of intermediate-sized clusters in addition to continual growth of the largest particle. The simulation results indicate that the nature of the gel growth is not influenced by the precursor particle morphology.

■

INTRODUCTION Alkali-activated materials (AAMs) are a class of cementitious binders synthesized by mixing solid aluminosilicate precursor powders (e.g., metakaolin and coal-derived fly ash) with alkaline solutions (e.g., sodium silicate), which are commonly called “activators”. The main role of activators is to accelerate the alkali-activation process because the reactivity of the precursor is generally lower than that of the conventional ordinary Portland cement (OPC). AAMs are widely considered to be potential alternatives for OPC because they possess a lower carbon footprint and similar mechanical performance when compared with OPC-based concrete.1 However, several outstanding questions remain unanswered regarding AAMs, including how the structural morphology of the precipitated alkali aluminosilicate gel evolves during the reaction (at the nano-/mesoscale) and how the development of nanoscale porosity during the alkali-activation reaction is affected by the choice of precursor/activator. Given that the lifetime of a concrete structure is strongly dictated by the durability of the material, it is important to have a deep fundamental © XXXX American Chemical Society

understanding of the formation mechanisms (and subsequent long-term stability) of the material across length scales, with information at the mesoscale (∼1−100 nm) remaining largely untapped until recent years. Significant advances have been made in uncovering the mechanistic and chemical processes of AAMs during the alkaliactivation reaction at the atomic and nanoscale. It is commonly recognized that the alkali-activation reaction mainly consists of three stages that overlap: dissolution, polycondensation, and reorganization.2 At the beginning, the aluminosilicate-rich solid precursors, including but not limited to fly ash, slag, and metakaolin, dissolve in the alkaline solution, where silicate and aluminate species (most likely in the monomeric form) are released. When the dissolution is to an extent such that the solution becomes supersaturated, silicate and aluminate species will undergo extensive condensation (formation of T−O−T Received: July 13, 2016 Revised: October 1, 2016

A

DOI: 10.1021/acs.langmuir.6b02592 Langmuir XXXX, XXX, XXX−XXX

Article

Langmuir

ness in implementing DFT-derived interaction energies in the CGMC technique, where the simulation results agreed semiquantitatively with 29Si nuclear magnetic resonance data.27 Moreover, we have used the DFT−CGMC technique in the past to simulate the alkali-activation of metakaolin (layered amorphous aluminosilicate).28 The major difference between alkali-activation of metakaolin and silica precipitation is that metakaolin consists of not only silicates but also aluminates. After accommodating for this difference in the interactions between aluminate and silicate species in the DFT energy calculations, the CGMC model was able to semiquantitatively replicate the complex alkali-activation process and provide new insights into this process, which cannot be accessed using experimental techniques.28 In this study, the scope of the CGMC simulations is extended to include both metakaolin- and fly ash-based AAMs, where each AAM has been activated by three sodium-based alkali hydroxide solutions with varying silica concentrations, to reveal the impact of (i) precursor composition (Si/Al ratio) and morphology (layered metakaolin vs spherical fly ash) and (ii) activator composition (free silica concentration) on the alkaliactivation reaction. The study aims to assess the structural evolution of the selected AAMs during the alkali-activation reaction and to characterize the reaction product (i.e., hardened AAM gels) from a modeling viewpoint.

linkages, where T denotes tetrahedral silicon or aluminum), signifying the commencement of the polycondensation process. This initial aluminosilicate network (denoted “gel 1”) will gradually undergo reorganization of the structure and eventually form a more interconnected aluminosilicate network (“gel 2”). Provis and van Deventer3 refined this conceptual model by incorporating another reaction path, that is, the possible formation of crystalline zeolitic phases from the initial aluminosilicate oligomers, which has been observed experimentally, especially in an environment of low Si/Al ratio,4−10 prolonged reaction period,11 or elevated temperature.12 Different experimental techniques have been employed to further our understanding of this alkali-activation process at various length scales.4,7,13−17 However, information at the mesoscale (∼1−100 nm) has remained sparse mainly because of the limited availability of suitable experimental techniques. Maitland et al. applied small-angle and ultra-small-angle neutron scattering to characterize the pore structure of metakaolin-based AAMs and found that AAMs with different Si/Al/Na ratios had statistically different micrometer-sized structures but a very similar structure over the 1 μm−60 nm length scale.18 They also observed a small fraction of closed pores in the AAM using the contrast variation method. Nanotomography was applied to a fly ash-based AAM, and a 3D pore network image was produced, showing the existence of “ink-bottle”-shaped pores in the AAM gel.19 However, these investigations were static in nature and do not account for the evolution of the gel and associated formation mechanisms as a result of the alkali-activation process. Steins et al. utilized an in situ small-angle scattering technique to investigate the structural evolution of metakaolin-based AAMs and revealed that the dissolution and gelation occurs more rapidly in an AAM with a smaller alkali in the activator (e.g., Na+ compared to K+).20 Despite the information obtained using the aforementioned experimental techniques, it is still difficult, if not impossible, to fully account for the alkali-activation process quantitatively, mainly because of the multiple chemical processes occurring simultaneously and the difficulty in experimentally tracking individual bond forming/breaking reactions (e.g., hydrolysis and condensation reactions).21 Hence, several theoretical and computational models have been developed in an attempt to clarify the individual processes.3,22 However, most of the modeling approaches focus on either very small scale (i.e., atomic) or relatively large scale (i.e., continuum), and few simulation techniques are available at the mesoscale.23 A multiscale-modeling approach combining atomistic simulations and continuum modeling was adopted by Hansen and Keil to model zeolite-catalyzed chemical reactions.24 Furthermore, a coarse-grained Monte Carlo (CGMC) technique was developed and found to be able to reach much larger time and length scales while retaining a high level of accuracy.25 The CGMC technique was then adopted by Jorge et al. to model the spontaneous formation of silica nanoparticles during the zeolite synthesis.26 However, the effective interaction energy of neutral silica polymerization needed to be calibrated with experimental measurements in the model proposed by Jorge et al., and therefore the accuracy of the model is somewhat dependent on this calibration process. To overcome this hurdle, we have previously used density functional theory (DFT) to calculate the interaction energies between the (neutral and deprotonated) silica species.27 Our simulations on silica precipitation demonstrated the effective-

■

MODEL DESCRIPTION

Details of the CGMC modeling technique have been previously documented in refs 27 and 28, where simulations were used to model (i) silica solubility and precipitation and (ii) the formation of metakaolin-based alkali-activated pastes, respectively. Hence, for the sake of brevity, only a brief summary of the model implementation is described here. Monte Carlo simulations have been conducted using the canonical ensemble (NVT; constant number of particles, volume, and temperature). The size of the cubic lattice simulation box is set at 50 × 50 × 50, under periodic boundary conditions in all three Cartesian coordinates. The rationale for this particular box size is explained in detail in ref 28. Each lattice site can accommodate one silicate or aluminate monomer (or an empty site). It should be noted that the cubic lattice is somewhat limited in accurately replicating the geometry of the silicate/aluminate tetrahedral, so other lattice models, such as bcc cubic lattice,29 should be considered in the future. Once the simulation commences, two events are called alternatively: “swap” and “bond”, where one event is counted as one iteration. In a swap event, an occupied site (silicate or aluminate monomer) is selected at random and its location is swapped with the location of another random site (occupied or unoccupied). The swap event is accepted according to the Metropolis algorithm,27,30 where the energy of the system (Gibbs free energy) is calculated by taking into account all bonds that exist in the lattice, where each bond has an energy given by the dimerization energy between the two units calculated using DFT.31 This is to facilitate the reconfiguration of the system toward lower energy states. In a bond event, an occupied site is selected at random, and a search is performed to see if it has any nonbondedoccupied neighbors. If nonbonded neighbors exist, one is selected at random and their status is updated to “bonded” with a probability determined by the Metropolis algorithm. Over the course of the simulation, the system energy (total Gibbs free energy) is recorded at regular intervals and is used to check the progress of the simulation, as well as to determine if the simulation has reached convergence (see Supporting Information for energy evolution profiles). The configuration of the box contents is also recorded periodically, with analysis of the box contents providing quantitative information that will be subsequently analyzed and discussed. Six AAM gels of different chemical compositions have been investigated, with the aluminosilicate precursors being metakaolin or class F fly ash, each activated by three different sodium-based alkaline B


Article

Langmuir Table 1. Compositions and Stoichiometry of the AAM Gels lattice site percentage (%) system

water + Na

silica in solution

metakaolin

fly ash

stoichiometry

hydroxide-activated metakaolin hydroxide/silicate-activated metakaolin silicate-activated metakaolin hydroxide-activated fly ash hydroxide/silicate-activated fly ash silicate-activated fly ash

76.4 72.2 68.4 76.4 72.2 68.4

0.0 5.6 10.6 0.0 5.6 10.6

23.6 22.2 21.0 0.0 0.0 0.0

0.0 0.0 0.0 23.6 22.2 21.0

NaAlSiO4·5.5H2O NaAlSi1.5O5·5.5H2O NaAlSi2O6·5.5H2O Na2/3Al2/3Si4/3O4·5.5H2O Na2/3Al2/3Si11/6O5·5.5H2O Na2/3Al2/3Si7/3O6·5.5H2O

solutions. The metakaolin precursor particle is constructed as a cube containing alternating aluminate and neutral silicate layers, so that the silicon to aluminum ratio is close to 1, as previously described in ref 28. On the other hand, the fly ash particle is constructed as a sphere with a silicon to aluminum ratio of about 2:1 so as to mimic the experimental elemental ratio of class F fly ash.16 Al−O−Al linkages have been minimized (less than 12% of T−O−T linkages (T denotes tetrahedral Si or Al) of the fly ash sphere are Al−O−Al) during the construction of the fly ash sphere given that the aluminate species are predominately in tetrahedral coordination and known to disfavor this bonding configuration in aluminosilicate glasses.32 As is the case for metakaolin implemented in the CGMC model,28 the fly ash sphere consists of six-coordinated aluminate and neutral silicate sites (as opposed to deprotonated silicate), which greatly simplifies its implementation. Obviously, this bonding environment deviates from the known four-coordinated silicon and aluminum that exists in glassy aluminosilicate particles in fly ash,33 but as shown in our previous investigation on metakaolin-based AAMs using CGMC, this simplification does not appear to adversely influence the conclusions drawn from the CGMC simulations. The total number of monomeric units for both metakaolin and fly ash particles is 26 250 to 29 500 (depending on the type of activator used), rendering the side length of the metakaolin cube at ∼9 nm (given ∼3.1 Å per silicate/aluminate monomer) and the diameter of the fly ash sphere at ∼11.9 nm (see Supporting Information for detailed calculation). It should be mentioned that these dimensions of the precursor particles are significantly smaller compared with the micron-sized particles found experimentally, and therefore future investigations will focus on upscaling to larger-sized precursors. The three alkaline solutions used in the simulations are sodium hydroxide (denoted as H-activated in the subsequent text) and sodium silicate with SiO2/Na2O molar ratios of 1.0 (H/S-activated) and 2.0 (S-activated). The sodium ions and water molecules are not explicitly modeled in the simulations because they have been accounted for in the DFT calculations of the dimerization energies for the different silicate and aluminate species.31 The silicate and aluminate species explicitly modeled in the DFT calculations include neutral, singly deprotonated and doubly deprotonated tetrahedral silicate monomers, and tetrahedral aluminate monomers, where the DFT calculations have been used to compute the dimerization energies of these monomers.31 The pH was set as 11 in the DFT calculations to simulate the high pH environment of alkali-activation, with silica deprotonation pKa values from the literature used to determine and update the extent of deprotonation of silicate sites in the CGMC simulation box (updated every 20 iterations, see ref 28 for more details). The temperature of the MC simulations was set as 298.15 K.28 The compositions of the alkali-activated pastes modeled using CGMC are given in Table 1. The pre-equilibration of silicates in the activating solution (for H/S- and S-activated systems) has been simulated by running the CGMC code only with the solution-based silicates being active in the simulation box, until the system energy reaches equilibrium (after ∼1 000 000 iterations). In the Results and Discussion section, 0 iteration is defined as the commencement of aluminosilicate dissolution, and therefore the pre-equilibration of silicates occurs before 0 iteration.

■

RESULTS AND DISCUSSION In the following sections, qualitative and quantitative data acquired from the alkali-activation simulations of metakaolin and fly ash are presented and discussed. Evolution of the Gibbs free energy of the system throughout the simulation is presented in the Supporting Information for each AAM. Visualization of the Simulation Lattice during AlkaliActivation. The images of the evolution of the six AAM systems outlined in Table 1 during alkali-activation are shown in Figure 1. The initial morphological difference between metakaolin and fly ash (layered vs sphere) can clearly be seen by comparing Figure 1a(i) with 1d(i), whereas the differences in the activating solutions with various amounts of free silica are visible in Figure 1a(i) through 1c(i) for the metakaolin-based AAMs. After commencement of the simulation, the silicate and aluminate species continuously dissolve into the solution (Figure 1(ii)) until the system becomes supersaturated. Then, the AAM gel starts to form via polymerization of the silicate and aluminate species (Figure 1(iii)), together with subsequent reorganization of the gel, until the system reaches equilibrium (Figure 1(iv)). Cluster Size Evolution during Alkali-Activation. The evolution of the clusters (i.e., nonmonomeric species) during the alkali-activation reaction is displayed in Figure 2, where the impact of precursor morphology and activator chemistry on the reaction is apparent. Here, we define a cluster as any nonmonomeric species, and therefore small oligomers (e.g., dimers and trimers) are included in the definition of clusters in this study. At the start of the alkali-activation reaction, free silica is present in the solutions for the H/S- and S-activated systems, and therefore a certain percentage of silicate monomers and oligomers exists in these solutions along with the aluminosilicate precursor particle (∼10.3% and 8.5% of all silicate and aluminate sites exist as monomers, in the H/S- and S-activated systems, respectively). The amount of silicate monomers in the solution reflects the total silicate concentration, with a higher SiO2/Na2O molar ratio (2.0 for S-activated compared with 1.0 for H/S-activated) leading to more silicate clusters (oligomers and nanoprecipitates) at the expense of the monomers.4,13,34 Figure 2 can be approximately divided into two regions: the first region (before ∼104 iterations) is dominated by dissolution of the aluminosilicate precursor particle, whereas the second region (after ∼104 iterations) is dominated by condensation reactions (formation of the AAM gel). It should be noted that the horizontal axis is on a log scale, and therefore the dissolution-dominated region is relatively short compared to the time (iterations) it takes for the condensation processes to reach completion. To better illustrate this point, the evolution of percentage of monomers (i.e., 100% minus percentage of clusters) for the metakaolin-based systems is plotted using a linear iteration (time) axis in Figure S2, where there is a sharp C


Article

Langmuir

Figure 2. Cluster evolution of (a) metakaolin and (b) fly ash activated with H-, H/S-, and S-based solutions. Note that the iterations are plotted on a logarithmic scale. The percentage of clusters measures the amount of monomers that exist in the bonded species (as opposed to nonbonded monomers).

Figure 1. Three-dimensional images of the silicate and aluminate species present in (a) H-activated metakaolin, (b) H/S-activated metakaolin, (c) S-activated metakaolin, (d) H-activated fly ash, (e) H/ S-activated fly ash, and (f) S-activated fly ash at various stages (number of iterations) during the simulation. Yellow spheres represent silicate monomeric sites, and pink spheres are aluminate monomeric sites (bonded and nonbonded). 0 iteration denotes the start of aluminosilicate dissolution, after pre-equilibration of free silica in solution, if present.

the simulation, and therefore ∼5% of aluminate and/or silicate sites exist as monomers in the pore solution of the hardened AAM gel. Pore Solution Chemistry of AAM Gels. As outlined in the previous paragraph, it appears that the chemical composition and/or morphology of the precursor particle (layered metakaolin vs spherical fly ash) influence the extent of gel formation (and therefore pore solution chemistry) of the resulting AAM gel. To explore this phenomenon, the concentration of monomers in the pore solution at the end of the simulation for all systems (H-, H/S-, and S-activated metakaolin and fly ash) is plotted in Figure 3. Here, the different systems are designated by their Si/Al ratios, as given in Table 2. Note that there is a slight offset from Si/Al ratios of 1.0 and 1.5 for the H- and H/S-activated metakaolin systems because of the manner of implementation for the metakaolin precursor particle in the CGMC simulations, where the particle was eroded starting from a perfect cube along an edge surface (nonbasal surface) one silicate/aluminate row at a time. Therefore, incomplete layers of silicate/aluminate exist at the edge surface to keep the total amount of particles the same for both metakaolin and fly ash precursors. It should be noted that even though Figure 3 considers only monomers as an indicator of the dissolved species in the pore solution, it is known that these solutions will also contain silicate/aluminosilicate oligomers and potentially nanoprecipi-

spike at the very beginning of the simulation, followed by the gradual decrease in the amount of the monomers. A similar trend was also observed by Zhang et al. using isothermal conduction calorimetry (ICC),15 except that the ICC data also showed a secondary broad peak after the first spike, which is indicative of the condensation of the aluminosilicate gel. It can be seen from Figure 2a that the activator chemistry affects the percentage of clusters existing at the end of the alkali-activation process for metakaolin, where H-activated metakaolin has the highest percentage of silicate and aluminate sites being incorporated into the clusters (∼99.0%). As the amount of free silica in the initial activating solution increases (increasing Si/Al ratio of the system), the percentage of clusters at the end of the reaction decreases, leading to an increased amount of aluminate and silicate monomeric species in the pore solution of the hardened gel. By contrast, Figure 2b shows that the fly ash-based systems with different activators converge to roughly the same percentage of clusters (∼95%) at the end of D


Article

Langmuir

(see Figure S3, where the aluminate concentration for the FA + H system is 0.3% higher than that for the MK + S system). The situation for the dissolved silicate species is more complex as shown in Figure 3 for metakaolin systems, where as the Si/Al ratio increases the silicate concentration is also seen to increase, from 0.5% to 2.25%. However, for fly ash systems, the silicate concentration seems to be at a stable state (as is the case for the aluminum species). Unlike the case of aluminum, there is a discontinuity in the silicate concentration between the MK + S and FA + H systems (which have the same Si/Al ratio of 2.0), as seen in Figure 3. The occurrence of this discontinuity may be attributed to the different amounts of silicate and aluminate units present in the simulation box for MK + S (approx. 30 000 Si and 15 000 Al) and FA + H (approx. 20 000 Si and 10 000 Al) systems. The higher concentration of species in the MK + S system will lead to an increased probability that a silicate monomer reacts with another species, thereby reducing the number of silicate monomers at the end of the reaction (effectively have a higher concentration of silicate in solution at the end of the reaction). On the other hand, there are fewer units in the simulation box (size of the box remains constant) in the FA + H system, and therefore the effective silicate concentration in the pore solution at the end of the reaction is less, leading to a higher monomer concentration. Aluminate does not follow the same trend because of the highly unfavorable Gibbs free energy associated with the breakage of Si−O−Al linkages (compared with Si−O−Si linkages) combined with an excess of silicate compared with the aluminate concentration (Si/Al ratio ≥1), which leads to the majority of aluminate sites being incorporated into the AAM gel. Hence, the Si/Al ratio of the precursors does influence the concentration of Si and Al species in the pore solution of the aluminosilicate gel. Buchwald et al.37 observed a similar trend in the solution chemistry for synthetic aluminosilicate gels. Figure 5 in ref 37 shows that as the Si/Al ratio increases from 1 to 2, the concentration of the aluminate species in solution remains relatively constant, whereas the concentration of the silicate species increases with increasing Si/Al ratio in an almost linear trend. However, for the fly ash-based systems here with Si/Al ratios ranging from 2 to 3.5, there is a clear difference in the behavior of the solution composition when compared with the evolution reported by Buchwald et al.37 Although the trend for the aluminum species is similar (remains constant), there is no increase in the overall amount of silicate in solution in our data (Figure 3) with increasing Si/Al ratio, whereas Buchwald et al.37 reported a monotonically increasing trend. The discrepancy may be explained by the difference in the extent of reaction between the two scenarios. In the investigation performed by Buchwald et al.,37 the aluminosilicate system was allowed to react for 24 h before the remnant solutions were extracted. However, a previous investigation has shown that the aluminosilicate gel, and therefore the remnant solution composition, changes over a time period of ∼90 h (time limit of the investigation) for high Si/Al ratio systems (>6.66). 38 Hence, the nature (i.e., structure) of the aluminosilicate gel investigated by Buchwald et al. is likely dependent on the duration of the synthesis, with the high Si/Al ratio systems taking longer time to reach equilibrium because of the increased concentration of silicate. Furthermore, in our previous investigation on the formation of AAM gels using in situ X-ray pair distribution function analysis, we observed a continual change in the structure of

Figure 3. Concentration of monomers (silicate and aluminate) in the pore solution of the final AAM gels according to the type of precursors (metakaolin and fly ash). The use of different activators (H-, H/S-, and S-activated) influences the overall Si/Al ratio of the systems studied (Table 2).

Table 2. Si/Al Ratios of Different AAM Systems activator precursor

H-

H/S-

S-

metakaolin

(MK + H) 0.937 (FA + H) 2

(MK + H/S) 1.426 (FA + H/S) 2.751

(MK + S) 2 (FA + S) 3.511

fly ash

tates. These dissolved oligomeric species have a dynamic nature, where the individual silicate/aluminate monomers in the oligomer can detach (and attach) from the entity fairly rapidly, which is different from small nanoparticles (i.e., nanoprecipitates) that do not possess such dynamic properties.35 Another method to distinguish between oligomers and nanoprecipitates is based on the local bonding, where small species containing Q4 (silicate/aluminate connected with four other silicates/aluminates) sites potentially correspond to nanoprecipitates, and species containing Q1, Q2, and Q3 (no Q4) correspond to oligomers.35 The CGMC technique does not distinguish the species in terms of their dynamic nature. Therefore, an assumption has been made that the percentage of monomers in solution is a good approximation for the behavior (concentration) of dissolved species (i.e., monomers and oligomers) in solution. We have also plotted the curve with monomers and oligomers/nanoprecipitates consisting of less than or equal to 20 monomeric units as dissolved species, which can be found in the Supporting Information and shows a similar trend to that visible in Figure 3. The selection of 20 as the cutoff size between oligomers and nanoprecipitates is based on previous research on silicate anions,36 where dissolved oligomers containing as many as 16 silicate monomeric units were identified using 29Si nuclear magnetic resonance (NMR). It can be seen from Figure 3 that dissolved aluminate monomers appear to reach a common equilibrium state for all systems, fluctuating around a value of 0.5%. It is worth noting that at the Si/Al ratio of 2, the two different systems, namely, Sactivated metakaolin (MK + S) and H-activated fly ash (FA + H), have a similar aluminum concentration in their pore solutions, indicating that the level of dissolved aluminate is not affected significantly by the different precursors and activators E


Article

Langmuir

polycondensation of the aluminosilicate gel commences, and aluminosilicate species grows on the remnant precursor particle. It should be noted that in H-activated systems, there is a “kink” at the beginning of the condensation process (∼105 to 106 iterations), indicating that the size of the precursor particle does not increase during this time frame. This is probably due to the different activators, where extra dissolved silica species in the H/S- and S-activated systems readily attach on the precursor particle at the commencement of condensation process, hence expediting the regrowth of the precursor particle. Such a delay in the regrowth of the precursor particle in hydroxide-based solution is also observed in an in situ Fourier transform infrared spectroscopy study.14 Moreover, Figure 4 shows that as Si/Al ratio increases, the final size of the aluminosilicate gel particle also increases. The inclusion of dissolved silicate in the activator thus has a significant impact on the evolution of the precursor particle and the resulting aluminosilicate gel. Figure 5 shows the evolution of the percentage of aluminum in the largest cluster for the metakaolin and fly ash systems. It

metakaolin-based binder samples up to 128 days (time selected as 100% reacted).39 At 24 h, all binders had a degree of reaction of approx. 50% or less, which is in agreement with the recent investigations by Zhang et al. using isothermal calorimetry.15,40 Although it is likely that the results presented by Buchwald et al. would be different if the gels were synthesized using a longer reaction time, it is still to be determined if the results presented in Figure 3 for the silicate monomers (Si/Al > 2) are realistic (constant with increasing Si/Al ratio). Evolution of the Precursor Particles. The CGMC model is able to provide information on each cluster throughout the simulation (size, T−O−T (T = Si or Al) linkages, composition, Qn speciation). Figure 4 shows the size evolution of the largest

Figure 4. Size evolution of the largest particle (initially (a) the metakaolin and (b) fly ash precursor) with different activators.

cluster (initially the metakaolin and fly ash precursor particles) throughout the simulations for different activator chemistries (H-, H/S-, and S-activated). To facilitate comparison between precursors activated with different alkaline solutions, the size of the particle has been normalized with respect to its initial size. Furthermore, direct comparison of different precursors is available in Figure S5. It can be seen that the overall shape of Figure 4 is quite similar to that of Figure 2 (evolution of the percentage of sites contained within clusters), where initially there is a decrease in the particle size followed by regrowth. It is obvious that the decrease in the particle size (Figure 4) is an indication of the extent of dissolution of the precursor particle, where both silicate and aluminate species are released into the solution until the solution becomes supersaturated. Then,

Figure 5. Evolution of percentage of aluminum in the largest particle (initially (a) the metakaolin and (b) fly ash precursor) with different activators.

can be seen that during the major dissolution stage (between 103 to 104 iterations, where the size rapidly decreases, as shown in Figure 4), the percentage of aluminum seems to be relatively stable for all three systems. Hence, during dissolution of the precursor particle in the CGMC simulations, the release of species (aluminate and silicate) is directly proportional to the composition of the particle, and no preferential release of aluminum occurs. At 104 to 105 iterations, there is a major F


Article

Langmuir

activated metakaolin) as opposed to six-coordinated in metakaolin and fly ash. The design of the CGMC methodology is a compromise between accuracy and complexity of the model. The apparent contradiction between the results obtained from this CGMC investigation and those from our previous investigation28 in terms of the preferential release of aluminum from the precursor particle can be explained by taking into account the point at which the particle was analyzed. In our previous study,28 a certain point in the simulation (40 000 iterations) was selected, and the percentage of aluminum in the particle at that particular time was extracted and compared with the original composition of the metakaolin particle, revealing that the preferential removal of aluminum had occurred. However, given that the time scale of these two models is relatively the same (as seen by comparing Figure 2a in this investigation and Figure 5 in our previous study28), a similar drop in the aluminum content compared with the beginning value can also be observed at ∼ 40 000 iterations in Figure 5a. Hence, the CGMC model in our previous investigation did not replicate the preferential release of aluminum from the metakaolin particle and instead showed that the remnant particle undergoes a compositional change (change in Si/Al ratio) prior to condensation of the aluminosilicate gel. Although there are limitations in the ability of the model to replicate the preferential release of aluminum from the precursor particle, the model is still able to reproduce the correct trends for the extent of precursor dissolution with different Si/Al ratios, as shown in Figure 4a. This is likely attributed to the saturation level of the solution for different systems before dissolution, where the systems containing dissolved silicate in solution (H/S- and S-activated) require a lower extent of dissolution of the precursor particle to reach local supersaturation levels associated with the precipitation of the aluminosilicate gel. The CGMC model (Figure 4b) shows that the extent of dissolution of the fly ash precursor particle in H- and H/Ssystems is very similar, whereas the particle in the S-activated system dissolves to a lesser extent. According to Lloyd et al.,7 the gel growth mechanism for the fly ash-based systems is quite different from the metakaolin-based ones. For low silicate concentration pastes (H-activated fly ash), the dissolved silicate and aluminate species preferentially precipitate on the surface of the remnant fly ash particle. The precipitates gradually cover the unreacted portion of the particle and further dissolution is controlled by the diffusion of the species through the gel. On the other hand, if the dissolved silicate concentration in the activator is high (S-activated fly ash), these species interact with the silicate and aluminate species released during fly ash dissolution, leading to precipitation through the gel away from the surface of the precursor particle. Because the dissolved silicate in the activator contains a significant amount of monomers (see Figure 2) and dimers, these species generally have lower degree of connectivity (Q0 and Q1) and therefore are more reactive compared with the species on the surface of the precursor particle. Hence, for the case of S-activated fly ash, most of the species released during dissolution of the precursor particle will participate in condensation reactions in solution rather than attaching back onto the ash surface. Starting from this theory, a reasonable prediction would be that the extent of dissolution of fly ash precursors increases with increasing Si/Al ratio. However, Criado et al.42 observed an opposite trend (decreasing extent of precursor dissolution with increasing Si/

decrease in the percentage of aluminum, as shown in Figure 5, but the size of the metakaolin and fly ash particles remains relatively constant (Figure 4). This indicates that right at the beginning of the condensation stage, there is a major compositional change occurring inside of the particle, where a significant proportion of the aluminate species is released into the solution and replaced by a comparable amount of silicate species. However, this phenomenon is possibly due to the concurrent dissolution of silicate and aluminate monomers (see Figure S4), which is not reflecting the true dissolution process (preferential release of aluminum) observed in experiments. After this major structural rearrangement, the particle regrows due to the (re)attachment of the dissolved silicate and aluminate species. Figure 5 also shows that the final aluminum content in the largest cluster decreases with increasing Si/Al ratio because at the end of the reaction most of the silicate species predissolved in the activator will have participated in (alumino)silicate condensation reactions with the particle forming the interconnected aluminosilicate gel network. Many previous experimental studies have shown the existence of unreacted raw materials within the gel phase.14,41,42 Among them, Duxson et al.43 found that with increasing Si/Al ratio, the amount of unreacted phase also increases, as measured by the intensity of Al(VI) in the 27Al MAS NMR spectra for metakaolin-based AAM samples. This behavior is also observed in the CGMC model, as shown in Figure 4a. Provis et al.44 proposed that this phenomenon might be due to the different nucleation mechanisms in solutions with varying silicate content. In alkaline activators with very dilute silicate concentration (i.e., H-activated system in our model), the rapidly released aluminate species will stay in the aqueous phase for a longer time because there are few extra nucleation sites apart from the surface of the precursor particle. However, with relatively high silicate content (S-activated system), the precursor particle will be surrounded by dissolved silicate species, such that when aluminate monomers are released from the precursor, they will readily bond with the silicate species and form the aluminosilicate gel. This gel network will then cover the precursor particle and retard the subsequent dissolution of the particle. This theory has presumed a preferential release of aluminum from aluminosilicate precursor materials, which has been observed in various studies.41,45,46 White et al.28 also found this phenomenon by using a previous version of this CGMC model. In the previous CGMC model, there was a higher effective temperature used in the simulations (higher rate of bond breaking), which altered the results in comparison with the current study. In this study, where the temperature of the simulations is more reflective of the conditions found experimentally, there is no such preferential release in the model, and the drop in the percentage of aluminum in the main particle actually occurs after the major dissolution stage. This deviation from experimental observation is probably due to the coordination state of the individual species in the precursor particles, where they are six-coordinated because of the cubic lattice grid. After polycondensation, however, they become at most four-coordinated sites so as to resemble the real bonding state known to exist in AAM gels. In this case (metakaolin and fly ash), both aluminate and silicate species tend to dissolve simultaneously because it is more energetically favorable for the silicate species to be present as four-coordinated bonded primarily to aluminates (∼99% Si−O−Al linkages in HG


Article

Langmuir Al ratio). Lloyd et al.7 suggested that this was because Criado et al.42 were taking the measurement long after the aforementioned phenomenon, where higher Si/Al ratio induces greater extent of dissolution, had occurred. According to Lloyd et al., even though the S-activated system would dissolve at a faster rate, once the gel has formed and filled up the space between the fly ash particles, convective transport would be greatly impeded because of the small pore sizes in S-activated samples, limiting further dissolution of the precursor particles.7 By contrast, for the case of the H-activated system, the empty space between the precursor particles is filled up by the outward growth through the gel from the precursor particles. Thus, the continued dissolution would not be influenced significantly because the transport of the gel on the particle surface would remain rate limiting. Eventually, the unreacted content of the fly ash precursor would be higher for samples activated with silicate, which is exactly the trend observed by Criado et al.42 and in Figure 4b. This theory of gel growth in fly ash-based systems assumes that the sodium-based aluminosilicate (N-A-S-(H)) gel that forms on the fly ash surface in the H-activated system behaves in a manner similar to the C−S−H gel in OPC-based systems, where movement of the dissolved precursor species (such as from C3S in OPC) can occur through the gel product, potentially through the gel pores (∼2−5 nm).47 However, without the presence of calcium in the system, the resulting aluminosilicate gel product is devoid of gel pores, as was demonstrated by Lloyd et al. using nitrogen sorption and Wood’s metal intrusion (the nitrogen sorption isotherm showed no plateau at high relative pressure, indicating that the majority of the pores are outside of the range measured using this technique).48 Hence, it would be difficult for the species to move through the N-A-S-(H) gel product. Furthermore, the alkali-activation reaction of aluminosilicate precursors possesses similarities to studies on glass corrosion, where the initial dissolution of the glass leads to the formation of a passivation layer (with low porosity) hindering further dissolution of the glass.49 In summary, the simulation shows a lower extent of dissolution of the S-activated fly ash system, which is in agreement with the experimental results. However, further investigation of the N-A-S-(H) gel product in fly ashbased systems is required to assess the permeability of these gels and the plausibility of the growth mechanism proposed by Lloyd et al.7 Cluster Size Distribution. To better understand the evolution of the clusters (monomers, oligomers, nanoprecipitates, and precursor/aluminosilicate gel) in terms of size throughout the reaction process, cluster size distributions have been generated, as shown in Figures 6 and 7. Here, the information on size distribution at certain iterations during the simulation is given, revealing that a similar progress in the cluster distribution occurs for both metakaolin- and fly ashbased systems. For example, Figures 6a and 7a show, for Hactivated metakaolin and fly ash, respectively, that as the reaction proceeds and the precursor particle dissolves, there is an increase in both the amount and the size of the small clusters (e.g., at 20 000 iterations, where the small clusters consist of oligomers with less than 20 silicate/aluminate monomeric units). Note that the maximum value of the y-axis is set as 2%, and therefore the percentage of monomers throughout the reaction is not shown. In fact, at 20 000 iterations, where the precursor particle dissolves to roughly the greatest extent, there are up to ∼40%−50% monomers in the solution, as shown in

Figure 6. Cluster size distribution at selected iterations for (a) Hactivated, (b) H/S-activated, and (c) S-activated metakaolin.

Figure 7. Cluster size distribution at selected iterations for (a) Hactivated, (b) H/S-activated, and (c) S-activated fly ash.

Figure 2. After the main dissolution process has occurred, polycondensation begins to govern the alkali-activation process, whereby the newly formed aluminosilicate gel starts to attach onto the remnant precursor particle, indicated by a regrowth of the largest cluster. In general, the growth of the largest cluster is attained at the expense of the smaller clusters, reflected by a H


Article

Langmuir

Figure 8. Percentage of T−O−T (T = Si or Al) linkages in the final AAM gel of H-, H/S-, and S-activated metakaolin and fly ash. Here, the different systems are represented by their nominal Si/Al ratios (see Table 2).

Lloyd et al.7 However, these simulation results are based on a methodology that only crudely accounts for diffusion processes (through the “swap” algorithm), and therefore additional research is required, including extension of the model to include a more accurate description of diffusion, to fully elucidate the gel growth mechanisms in alkali-activated fly ashes. It is observed that the activator chemistry has a large impact on the final distribution of the clusters. The insets of Figures 6 and 7 display the distribution of intermediate-sized clusters, ranging from 100 to 260 monomeric units. Only the Hactivated systems contain clusters of such size (denoted by black stars) at the end of simulation. By contrast, the maximum sizes of the small clusters in H/S- and S-activated systems are around 50 and 30 sites, respectively. Correspondingly, the size of the largest cluster in the H-activated systems is somewhat smaller than those in the H/S- and S-activated counterparts. This observation is in support of the finding reported by many authors that more homogenous AAM gel is formed at higher Si/Al ratios.6,7,13,42 Nearest-Neighbor Bonding Environment. The CGMC model can provide information on the local bonding environment in the gels, specifically the relative percentage of T−O−T linkages (i.e., Si−O−Si, Si−O−Al, and Al−O−Al) in the final AAM gel, as shown in Figure 8. A linear trend can be clearly observed as the Si/Al ratio of the system increases from 1.0 to 3.5, with the percentage of Si−O−Si linkages increases while the Si−O−Al linkages decrease. This result agrees with the findings from Silva et al., who stated that the mutual reaction between the silicate species becomes dominant with increasing silicate content.53 It is worth noting that the trend lines are fitted by the data points of all systems simulated in this model, suggesting that this relationship is not affected by the different precursor morphologies but is affected by the Si/Al ratio of the AAM system. On the other hand, the amount of Al−O−Al linkages is negligible (the maximum percentage is ∼0.08% in Sactivated metakaolin), which is in good agreement with Loewenstein’s rule.54 Furthermore, for the system with the

decrease in both the amount and the size of the small clusters, until the system reaches equilibrium. Such a growth mechanism resembles an Ostwald-ripeningtype process, which is common in many aqueous silicate-based and aluminosilicate-based systems, including silica nanoparticle evolution26,50 and zeolite formation.51,52 White et al. also observed this phenomenon in the S-activated metakaolin using the previous version of this CGMC model.28 The current model shows that Ostwald ripening occurs in the H/S- and Sactivated systems, for both metakaolin and fly ash. However, for the H-activated systems, especially for H-activated fly ash, the intermediate-sized clusters are observed to continually grow in number and size during the simulation, especially between 200 000 and 1 000 000 iterations. This indicates that the nature of the gel growth in the H-activated systems is via the formation and growth of intermediate-sized clusters away from the particle surface, in addition to the reattachment of the species on the surface of the partially dissolved particle. However, this gel growth is not indicative of extensive Ostwald ripening because the largest particle does not grow at the expense of these intermediate-sized clusters. Hence, our original findings in our previous CGMC investigation are valid, where Ostwald ripening was found to occur in S-activated metakaolin but was not apparent in H-activated metakaolin.28 Here, we have also shown that the same behavior according to activator chemistry occurs in the fly ash systems, and therefore these simulations indicate that the particle morphology of the precursor particle does not influence/change the nature of the gel growth. These simulation results are in disagreement with the proposed growth mechanism reported by Lloyd et al.7 for fly ash-based systems. In fact, the trends we report here are opposite to that reported by Lloyd et al., where we observe gel growth throughout the system for the case of the H-activated systems (metakaolin and fly ash) via the formation and growth of intermediate-sized clusters. On the other hand, for H/S- and S-activated systems (both metakaolin and fly ash), the gel grows primarily on the partially dissolved particle via Ostwald ripening, and not throughout the system, as was proposed by I


Article

Langmuir Si/Al ratio of ∼1 (H-activated metakaolin), the percentage of Si−O−Al linkages is approaching 100%, indicating a predominantly Q4(4Al)-type gel structure.5 However, Duxson et al. suggested that non-Loewenstein behavior exists, which may be due to the different cations used.5 Thus, a more in-depth study on the impact of different cations on the population of T−O− T linkages using this CGMC technique is recommended.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: +1 609 258 6263. Fax: +1 609 258 2799.

■

Notes

CONCLUSIONS In this study, the structural evolution of the AAM gel has been investigated at the mesoscale via a purpose-built CGMC modeling technique coupled with DFT calculations. Two precursors have been modeled (metakaolin and class F fly ash), with each precursor activated using three alkaline solutions with varying silicate content to assess the impact of silicate concentration on the resulting AAM gels. It is found that dissolved silicate and aluminate species exist in the pore solution of the hardened AAM gel, with the concentration of the silicate species correlating with the bulk Si/Al ratio. Furthermore, the extent of dissolved species in the hardened gels is in general agreement with experiment results available in the literature. The dissolution and regrowth mechanisms of the metakaolin and fly ash precursor particle are reproduced by the CGMC simulations and have been compared with the growth mechanisms of the AAM gel proposed in the literature. It is found that the local supersaturation state of the solution is the main factor that determines the extent of dissolution of the precursor particle, with free silica in the initial-activating solution leading to a lower extent of dissolution of the precursor, irrespective of the precursor particle morphology (layered metakaolin vs spherical fly ash). For the systems containing free silica at the start of the reaction (H/S- and S-activated), the growth of aluminosilicate gel during the alkali-activation reaction follows an Ostwald ripening process, where smaller clusters are consumed by the larger ones so as to reduce the overall surface energy of the gel. However, for both H-activated metakaolin and fly ash systems, the gel growth occurs via the formation of intermediate-sized clusters in conjunction with the growth of the largest particle. At the end of this process, some small clusters remain in the pore solution of the hardened gel, and their sizes depend on the degree of saturation (predissolved silica concentration) of the system. Finally, the model shows that the distribution of T− O−T (T = Si or Al) linkages in the aluminosilicate gel is dependent on the overall Si/Al ratio of the system, and the distribution generally conforms to Loewenstein’s rule (avoidance of Al−O−Al linkages).

■

ash particle. Connectivity (Qn, n = 0−6) distributions in the AAM systems (PDF)

The authors declare no competing financial interest.

■ ■

ACKNOWLEDGMENTS The authors thank Ms. Angela Mao for her help with carrying out the CGMC simulations. REFERENCES

(1) Duxson, P.; Provis, J. L.; Lukey, G. C.; van Deventer, J. S. J. The Role of Inorganic Polymer Technology in the Development of “Green Concrete”. Cem. Concr. Res. 2007, 37, 1590−1597. (2) Duxson, P.; Fernández-Jiménez, A.; Provis, J. L.; Lukey, G. C.; Palomo, A.; van Deventer, J. S. J. Geopolymer Technology: The Current State of the Art. J. Mater. Sci. 2006, 42, 2917−2933. (3) Provis, J. L.; van Deventer, J. S. J. Geopolymerisation Kinetics. 2. Reaction Kinetic Modelling. Chem. Eng. Sci. 2007, 62, 2318−2329. (4) Criado, M.; Fernández-Jiménez, A.; Palomo, A.; Sobrados, I.; Sanz, J. Effect of the SiO2/Na2O Ratio on the Alkali Activation of Fly Ash. Part II: 29Si MAS-NMR Survey. Microporous Mesoporous Mater. 2008, 109, 525−534. (5) Duxson, P.; Provis, J. L.; Lukey, G. C.; Separovic, F.; van Deventer, J. S. J. 29Si NMR Study of Structural Ordering in Aluminosilicate Geopolymer Gels. Langmuir 2005, 21, 3028−3036. (6) Hajimohammadi, A.; Provis, J. L.; van Deventer, J. S. J. The Effect of Silica Availability on the Mechanism of Geopolymerisation. Cem. Concr. Res. 2011, 41, 210−216. (7) Lloyd, R. R.; Provis, J. L.; van Deventer, J. S. J. Microscopy and Microanalysis of Inorganic Polymer Cements. 2: The Gel Binder. J. Mater. Sci. 2008, 44, 620−631. (8) Rees, C. A.; Provis, J. L.; Lukey, G. C.; van Deventer, J. S. J. Attenuated Total Reflectance Fourier Transform Infrared Analysis of Fly Ash Geopolymer Gel Aging. Langmuir 2007, 23, 8170−8179. (9) White, C. E.; Provis, J. L.; Llobet, A.; Proffen, T.; van Deventer, J. S. J. Evolution of Local Structure in Geopolymer Gels: An in Situ Neutron Pair Distribution Function Analysis. J. Am. Ceram. Soc. 2011, 94, 3532−3539. (10) Rahier, H.; Simons, W.; van Mele, B.; Biesemans, M. LowTemperature Synthesized Aluminosilicate Glasses: Part III Influence of the Composition of the Silicate Solution on Production, Structure and Properties. J. Mater. Sci. 1997, 32, 2237−2247. (11) Fernández-Jiménez, A.; Palomo, A. Mid-Infrared Spectroscopic Studies of Alkali-Activated Fly Ash Structure. Microporous Mesoporous Mater. 2005, 86, 207−214. (12) White, C. E.; Provis, J. L.; Proffen, T.; van Deventer, J. S. J. The Effects of Temperature on the Local Structure of Metakaolin-Based Geopolymer Binder: A Neutron Pair Distribution Function Investigation. J. Am. Ceram. Soc. 2010, 93, 3486−3492. (13) Duxson, P.; Provis, J. L.; Lukey, G. C.; Mallicoat, S. W.; Kriven, W. M.; van Deventer, J. S. J. Understanding the Relationship between Geopolymer Composition, Microstructure and Mechanical Properties. Colloids Surf., A 2005, 269, 47−58. (14) Rees, C. A.; Provis, J. L.; Lukey, G. C.; van Deventer, J. S. J. In Situ ATR-FTIR Study of the Early Stages of Fly Ash Geopolymer Gel Formation. Langmuir 2007, 23, 9076−9082. (15) Zhang, Z.; Provis, J. L.; Wang, H.; Bullen, F.; Reid, A. Quantitative Kinetic and Structural Analysis of Geopolymers. Part 2. Thermodynamics of Sodium Silicate Activation of Metakaolin. Thermochim. Acta 2013, 565, 163−171. (16) Fernández-Jiménez, A.; Palomo, A.; Sobrados, I.; Sanz, J. The Role Played by the Reactive Alumina Content in the Alkaline

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.langmuir.6b02592. Energy evolution of the CGMC simulations. Evolution of the percentage of monomers as an indication of the extent of reaction. Pore solution chemistry of the AAM gels calculated taking into account monomers and oligomers of 20 sites or less. Evolution of the percentages of Si and Al monomers in H-activated metakaolin. Dissolution rate of metakaolin and fly ash precursor particles. Calculation of the diameter of the simulated fly J


Article

Langmuir Activation of Fly Ashes. Microporous Mesoporous Mater. 2006, 91, 111−119. (17) White, C. E.; Page, K.; Henson, N. J.; Provis, J. L. In Situ Synchrotron X-Ray Pair Distribution Function Analysis of the Early Stages of Gel Formation in Metakaolin-Based Geopolymers. Appl. Clay Sci. 2013, 73, 17−25. (18) Maitland, C. F.; Buckley, C. E.; O’Connor, B. H.; Butler, P. D.; Hart, R. D. Characterization of the Pore Structure of MetakaolinDerived Geopolymers by Neutron Scattering and Electron Microscopy. J. Appl. Crystallogr. 2011, 44, 697−707. (19) Provis, J. L.; Rose, V.; Winarski, R. P.; van Deventer, J. S. J. Hard X-Ray Nanotomography of Amorphous Aluminosilicate Cements. Scr. Mater. 2011, 65, 316−319. (20) Steins, P.; Poulesquen, A.; Diat, O.; Frizon, F. Structural Evolution During Geopolymerization from an Early Age to Consolidated Material. Langmuir 2012, 28, 8502−8510. (21) Schaffer, C. L.; Thomson, K. T. Density Functional Theory Investigation into Structure and Reactivity of Prenucleation Silica Species. J. Phys. Chem. C 2008, 112, 12653−12662. (22) Catlow, C. R. A.; George, A. R.; Freeman, C. M. Ab Initio and Molecular-Mechanics Studies of Aluminosilicate Fragments, and the Origin of Lowenstein’s Rule. Chem. Commun. 1996, 11, 1311−1312. (23) Kupwade-Patil, K.; Soto, F.; Kunjumon, A.; Allouche, E. N.; Mainardi, D. S. Multi-Scale Modeling and Experimental Investigations of Geopolymeric Gels at Elevated Temperatures. Comput. Struct. 2013, 122, 164−177. (24) Hansen, N.; Keil, F. J. Multiscale Modeling of Reaction and Diffusion in Zeolites: From the Molecular Level to the Reactor. Soft Mater. 2012, 10, 179−201. (25) Chatterjee, A.; Vlachos, D. G.; Katsoulakis, M. A. Spatially Adaptive Lattice Coarse-Grained Monte Carlo Simulations for Diffusion of Interacting Molecules. J. Chem. Phys. 2004, 121, 11420−11431. (26) Jorge, M.; Auerbach, S. M.; Monson, P. A. Modeling Spontaneous Formation of Precursor Nanoparticles in Clear-Solution Zeolite Synthesis. J. Am. Chem. Soc. 2005, 127, 14388−14400. (27) White, C. E.; Provis, J. L.; Proffen, T.; van Deventer, J. S. J. Quantitative Mechanistic Modeling of Silica Solubility and Precipitation During the Initial Period of Zeolite Synthesis. J. Phys. Chem. C 2011, 115, 9879−9888. (28) White, C. E.; Provis, J. L.; Proffen, T.; van Deventer, J. S. J. Molecular Mechanisms Responsible for the Structural Changes Occurring During Geopolymerization: Multiscale Simulation. AIChE J. 2011, 58, 2241−2253. (29) Jin, L.; Auerbach, S. M.; Monson, P. A. Modeling ThreeDimensional Network Formation with an Atomic Lattice Model: Application to Silicic Acid Polymerization. J. Chem. Phys. 2011, 134, 134703. (30) Frenkel, D.; Smit, B.; Ratner, M. A. Understanding Molecular Simulations: From Algorithms to Applications. Phys. Today 1996, 50, 1348. (31) White, C. E.; Provis, J. L.; Kearley, G. J.; Riley, D. P.; van Deventer, J. S. J. Density Functional Modelling of Silicate and Aluminosilicate Dimerisation Solution Chemistry. Dalton Trans. 2011, 40, 1348−1355. (32) Zheng, Q.; Smedskjaer, M. M.; Youngman, R. E.; Potuzak, M.; Mauro, J. C.; Yue, Y. Influence of Aluminum Speciation on the Stability of Aluminosilicate Glasses Against Crystallization. Appl. Phys. Lett. 2012, 101, 041906. (33) Fernández-Jiménez, A.; de la Torre, A.; Palomo, A.; LópezOlmo, G.; Alonso, M.; Aranda, M. Quantitative Determination of Phases in the Alkali Activation of Fly Ash. Part I. Potential Ash Reactivity. Fuel 2006, 85, 625−634. (34) Petry, D. P.; Haouas, M.; Wong, S. C. C.; Aerts, A.; Kirschhock, C. E. A.; Martens, J. A.; Gaskell, S. J.; Anderson, M. W.; Taulelle, F. Connectivity Analysis of the Clear Sol Precursor of Silicalite: Are Nanoparticles Aggregated Oligomers or Silica Particles? J. Phys. Chem. C 2009, 113, 20827−20836.

(35) Gebauer, D.; Kellermeier, M.; Gale, J. D.; Bergström, L.; Cölfen, H. Pre-Nucleation Clusters as Solute Precursors in Crystallisation. Chem. Soc. Rev. 2014, 43, 2348−2424. (36) Knight, C. T. G. A. Two-Dimensional Silicon-29 Nuclear Magnetic Resonance Spectroscopic Study of the Structure of the Silicate Anions Present in an Aqueous Potassium Silicate Solution. J. Chem. Soc., Dalton Trans. 1988, 6, 1457−1460. (37) Buchwald, A.; Zellmann, H.-D.; Kaps, C. Condensation of Aluminosilicate GelsModel System for Geopolymer Binders. J. NonCryst. Solids 2011, 357, 1376−1382. (38) Ginter, D. M.; Went, G. T.; Bell, A. T.; Radke, C. J. A Physicochemical Study of the Againg of Colloidal Silica Gels Used in Zeolite Y Synthesis. Zeolites 1992, 12, 733−741. (39) White, C. E.; Provis, J. L.; Bloomer, B.; Henson, N. J.; Page, K. In Situ X-Ray Pair Distribution Function Analysis of Geopolymer Gel Nanostructure Formation Kinetics. Phys. Chem. Chem. Phys. 2013, 15, 8573−10. (40) Zhang, Z.; Wang, H.; Provis, J. L.; Bullen, F.; Reid, A.; Zhu, Y. Quantitative Kinetic and Structural Analysis of Geopolymers. Part 1. The Activation of Metakaolin with Sodium Hydroxide. Thermochim. Acta 2012, 539, 23−33. (41) Provis, J. L.; Duxson, P.; Lukey, G. C.; van Deventer, J. S. J. Statistical Thermodynamic Model for Si/Al Ordering in Amorphous Aluminosilicates. Chem. Mater. 2005, 17, 2976−2986. (42) Criado, M.; Fernández-Jiménez, A.; de la Torre, A. G.; Aranda, M. A. G.; Palomo, A. An XRD Study of the Effect of the SiO2/Na2O Ratio on the Alkali Activation of Fly Ash. Cem. Concr. Res. 2007, 37, 671−679. (43) Duxson, P.; Lukey, G. C.; Separovic, F.; van Deventer, J. S. J. Effect of Alkali Cations on Aluminum Incorporation in Geopolymeric Gels. Ind. Eng. Chem. Res. 2005, 44, 832−839. (44) Provis, J. L.; Lukey, G. C.; van Deventer, J. S. J. Do Geopolymers Actually Contain Nanocrystalline Zeolites? A Reexamination of Existing Results. Chem. Mater. 2005, 17, 3075−3085. (45) Köhler, S. J.; Dufaud, F.; Oelkers, E. H. An Experimental Study of Illite Dissolution Kinetics as a Function of pH from 1.4 to 12.4 and Temperature from 5 to 50 °C. Geochim. Cosmochim. Acta 2003, 67, 3583−3594. (46) Sagoe-Crentsil, K.; Weng, L. Dissolution Processes, Hydrolysis and Condensation Reactions During Geopolymer Synthesis: Part II. High Si/Al Ratio Systems. J. Mater. Sci. 2006, 42, 3007−3014. (47) Bullard, J. W.; Scherer, G. W.; Thomas, J. J. Time Dependent Driving Forces and the Kinetics of Tricalcium Silicate Hydration. Cem. Concr. Res. 2015, 74, 26−34. (48) Lloyd, R. R.; Provis, J. L.; Smeaton, K. J.; van Deventer, J. S. J. Spatial Distribution of Pores in Fly Ash-Based Inorganic Polymer Gels Visualised by Wood’s Metal Intrusion. Microporous Mesoporous Mater. 2009, 126, 32−39. (49) Fournier, M.; Gin, S.; Frugier, P. Resumption of Nuclear Glass Alteration: State of the Art. J. Nucl. Mater. 2014, 448, 348−363. (50) Provis, J. L.; Vlachos, D. G. Silica Nanoparticle Formation in the TPAOH−TEOS−H2O System: A Population Balance Model. J. Phys. Chem. B 2006, 110, 3098−3108. (51) Navrotsky, A. Energetic Clues to Pathways to Biomineralization: Precursors, Clusters, and Nanoparticles. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 12096−12101. (52) Rimer, J. D.; Vlachos, D. G.; Lobo, R. F. Evolution of SelfAssembled Silica−Tetrapropylammonium Nanoparticles at Elevated Temperatures. J. Phys. Chem. B 2005, 109, 12762−12771. (53) Silva, P. D.; Sagoe-Crenstil, K.; Sirivivatnanon, V. Kinetics of Geopolymerization: Role of Al2O3 And SiO2. Cem. Concr. Res. 2007, 37, 512−518. (54) Lowenstein, W. The Distribution of Aluminium in the Tetrahedra of Silicates and Aluminates. Am. Mineral. 1954, 39, 92−96.

K


Modeling the Formation of Alkali Aluminosilicate Gels at the

Recommend Documents