I. GILA RIVER INDIAN COMMUNITY - Bureau of Reclamation

LA-UR-12-21437Approved for public release; distribution is unlimited.

Title: The Synergy Between Total Scattering and Advanced SimulationTechniques: Quantifying Geopolymer Gel Evolution

Author(s): White, ClaireBloomer, Breaunnah E.Provis, John L.Henson, Neil J.Page, Katharine L.

Intended for: NICOM 4: 4th International Symposium on Nanotechnology inConstruction, 2012-05-20/2012-05-22 (Agios Nikolaos, ---, Greece)

Disclaimer:Los Alamos National Laboratory, an affirmative action/equal opportunity employer,is operated by the Los Alamos National Security, LLC for the National NuclearSecurity Administration of the U.S. Department of Energy under contract DE-AC52-06NA25396. By approving this article, the publisher recognizes that the U.S. Government retains nonexclusive, royalty-free license to publish or reproduce the published form of this contribution, or to allow others to do so, for U.S. Government purposes. Los Alamos National Laboratory requests that the publisher identify this article as work performed under the auspices of the U.S. Departmentof Energy. Los Alamos National Laboratory strongly supports academic freedom and a researcher's right to publish; as an institution, however, the Laboratory does not endorse the viewpoint of a publication or guarantee its technical correctness.

NICOM 4: 4th International Symposium on Nanotechnology in Construction

The Synergy Between Total Scattering and Advanced Simulation Techniques: Quantifying Geopolymer Gel Evolution

C.E. White1,2,3, B. Bloomer1, J.L. Provis4, N.J. Henson2 and K. Page1 1Lujan Neutron Scattering Center, Los Alamos National Laboratory, Los Alamos, USA

[email protected], [email protected] 2Physics and Chemistry of Materials, Los Alamos National Laboratory, Los Alamos, USA

[email protected] 3Center for Nonlinear Studies, Los Alamos National Laboratory, Los Alamos, USA

4Department of Chemical & Biomolecular Engineering, University of Melbourne, Melbourne, Australia [email protected]

ABSTRACT

With the ever increasing demands for technologically advanced structural materials, together with emerging environmental consciousness due to climate change, geopolymer cement is fast becoming a viable alternative to traditional cements due to proven mechanical engineering characteristics and the reduction in CO2 emitted (approximately 80% less CO2 emitted compared to ordinary Portland cement). Nevertheless, much remains unknown regarding the kinetics of the molecular changes responsible for nanostructural evolution during the geopolymerization process. Here, in-situ total scattering measurements in the form of X-ray pair distribution function (PDF) analysis are used to quantify the extent of reaction of metakaolin/slag alkali-activated geopolymer binders, including the effects of various activators (alkali hydroxide/silicate) on the kinetics of the geopolymerization reaction. Restricting quantification of the kinetics to the initial ten hours of reaction does not enable elucidation of the true extent of the reaction, but using X-ray PDF data obtained after 128 days of reaction enables more accurate determination of the initial extent of reaction. The synergies between the in-situ X-ray PDF data and simulations conducted by multiscale density functional theory-based coarse-grained Monte Carlo analysis are outlined, particularly with regard to the potential for the X-ray data to provide a time scale for kinetic analysis of the extent of reaction obtained from the multiscale simulation methodology.

Keywords: Geopolymer, Alkali-activated aluminosilicates, Pair distribution function, Kinetics, Local structure, Amorphous

1. Introduction

Geopolymer cement is a class of alkali-activated cement, commonly synthesized using coal fly ash, ground granulated blast-furnace slag (GGBS) and/or metakaolin as the aluminosilicate precursors to produce a mechanically hard binder [1-3]. The increasing interest in this alternative cement over recent years is due to its low CO2 emissions compared with ordinary Portland cement (80 – 90% less CO2 emitted per ton of binder [2]), coupled with proven mechanical performance [4-8]. However, the success of geopolymer concrete in commercial and industrial settings is dependent on many additional factors [3]. Nevertheless, laboratory-scale research can provide crucial information regarding the nanostructural evolution of the highly heterogeneous and predominantly amorphous geopolymer binders. Analysis of the nanostructural ordering in geopolymers is complicated by the amorphous nature of the precursors, intermediate reaction products and final gel binder [9-14].

Total Scattering is a technique capable of probing the local structure of disordered crystalline, amorphous and nanocrystalline materials [15]. By analyzing atom-atom correlations in real-space, the pair distribution function (PDF) directly probes the local structure of a material. PDF analysis has been previously performed on metakaolin-based and fly ash-based geopolymer binders [9, 12, 13, 16, 17], and in particular, our recent in-situ neutron PDF analysis revealed important local structural changes occurring during the geopolymerization reaction for metakaolin-based systems [12]. However, the in-situ neutron PDF investigation was not able to quantify the extent of reaction due to background contributions and limited data and time resolution. However, in-situ X-ray PDF analysis, using the rapid acquisition detection system at 11-ID-B (APS, Argonne National Laboratory) [18], is capable of probing the local structural changes occurring during geopolymerization, and therefore elucidating kinetic information regarding the reaction.

Here, in-situ X-ray PDF analysis is used to quantify the extent of reaction as a function of time for metakaolin-based and GGBS-based geopolymer systems. The synergies between advanced simulation techniques and total scattering in the form of X-ray/neutron pair distribution function analysis are also outlined, together with the potential to elucidate key nanostructural characteristics of the geopolymer precursors and the evolution of the gel binder. Advanced simulation techniques that will be discussed include quantum mechanics (in the form of density functional theory) and coarse-grained Monte Carlo analysis. Hence, by combining the capabilities of theory and experiment for several length scales (atomistic, nano- and mesoscale), much more can be understood regarding the structural mechanisms responsible for nanostructural ordering and thermodynamic stability in this important cementitious alternative.

2. Materials and Methods

2.1. Geopolymer gel compositions

The geopolymer binders studied in this investigation were synthesized from high-purity metakaolin, previously characterized in references 19 and 20, and ground granulated blast furnace slag (GGBS) (refer to reference 21 for further details). The precursors were activated with sodium-based alkaline solutions hydroxide-activated (“H-activated”) systems using NaOH pellets, and silicate-activated (“S-activated”) systems using anhydrous sodium metasilicate (Na2SiO3) powder. For preparation of each solution, the solid alkali source was dissolved in distilled water. The geopolymer gel compositions studied in this investigation are given in Table 1. The higher water/binder ratio for the metakaolin-based gels is due to the higher water demand of metakaolin, and the ratio is selected based on previous investigations of metakaolin-based geopolymers [9, 13, 22].

2.2. X-ray pair distribution function analysis

The precursors (metakaolin or slag) were mixed for approximately 30 seconds (until of a uniform liquid consistency) with the activating solutions immediately prior to loading into Kapton capillary tubes (via suction). The tubes were then sealed with epoxy at both ends, and mounted on the sample changer on the 11-ID-B beam line at the Advanced Photon Source, Argonne National Laboratory. Mixing of each sample commenced exactly 15 minutes before the start of collection of the first data set. Samples were measured for 10 hours and 8 minutes, with data sets being acquired every 10 minutes for a duration of 2 minutes 50 seconds per data set. Each sample was re-measured for a single scan at a time somewhere between 13 and 77 hours after initial mixing. Further data sets were acquired at 128 days after mixing.

Table 1. Geopolymer gel compositions studied using in-situ X-ray PDF analysis

Sample Name Na2O (wt. %)a Si2O (wt. %)b Water/binder Ratio

High alkali H-activated metakaolin 28.0 0 0.90

High alkali S-activated metakaolin 28.0 27.2 0.90

Low alkali H-activated metakaolin 7.0 0 0.90

Low alkali S-activated metakaolin 7.0 6.8 0.90

High alkali H-activated slag 11.0 0 0.35

Low alkali H-activated slag 7.0 0 0.35

Low alkali S-activated slag 7.0 6.8 0.35 a Activator contains this number of grams of Na2O per 100g of solid precursor. b Activator contains this number of grams of Si2O per 100g of solid precursor.

The pair distribution function, G(r), is obtained by a sine Fourier transform of the scattering function, S(Q), as depicted in Eq. 1, where the definition of Q is given in Eq. 2.

To quantify the extent of reaction, a methodology was used similar to that outlined by Provis and van Deventer [11, 23], where the initial and final data sets from the experiment were used as the start and end points for quantification. The intermediate data sets were then fitted with a linear combination of the initial and final data sets (Eq. 3, where α is between 0 and 1).

In this investigation, the best fit has been obtained by minimizing Σ(ycalc - yexp)2 with respect

to α for each intermediate data set, giving the α value (or extent of reaction), where α = 0 corresponds to the initial data set, and α = 1 corresponds to the final data set (fully reacted).

Given that the initial data sets were obtained 18 minutes (0.3 hours) after the start of mixing of precursor and solution, the raw α values do not give a truly representative measure of the extent of reaction. In order to account for this, the same method as outlined by Provis and van Deventer was used, where the extent of reaction during the initial stages of the geopolymerization reaction was used to perform a linear extrapolation to the point of mixing, at 0 minutes [11, 23].

(1)

(2)

(3)

3. Results and discussion

3.1. In-situ X-ray pair distribution functions

The in-situ X-ray pair distribution functions (PDFs) of high alkali S-activated metakaolin and low alkali H-activated slag are given in Fig 1a and 1b, respectively. It is clearly visible that for both systems, the local structure initially is amorphous, with no strong correlations past approximately 8 Å. However, it is difficult to assign individual atom-atom correlations to peaks past 3 Å without the aid of a detailed amorphous atomic structural model. For geopolymer binders there are no such detailed amorphous atomistic models, nevertheless, progress in this area is ongoing, as will be discussed in section 3.4. The first and second-nearest neighbor atom-atom correlations are labeled in Fig. 1, where T corresponds to Si and Al. Peaks below the first nearest-neighbor (T-O) at ~1.65 Å are non-structural, and are due to data termination errors and imperfect corrections [15].

Evident in Fig. 1 are the differences in changes over time between high alkali S-activated metakaolin (Fig. 1a), and low alkali H-activated slag (Fig. 1b). The local structure of S-activated metakaolin remains amorphous over the entire extent of reaction (up to 128 days, approximately 3000 hours), whereas for H-activated slag the local structure begins to crystallize after a period of time (at some point between 13.5 and 3000 hours). The commencement of crystallization complicates quantification of the extent of reaction, making it impossible to use the 128 day data set as the end point to define ‘fully reacted’. Nevertheless, in section 3.2, the 10 hour data sets will be used to quantify the extent of reaction, to determine whether reaction kinetic information can be determined by considering only the initial 10 hours of reaction.

3.2. Extent of reaction quantification: Initial ten hours

Fig. 2 displays the calculated extent of reaction for the various systems investigated, using the 10 hour data set as the end point (i.e. assuming the sample has fully reacted by this point). In Fig. 2a, it is clearly visible that the extent of reaction for low alkali S-activated metakaolin is statistically noisy (due to not enough sample being present in the beam), making it difficult to extract quantitative information regarding the kinetics of this reaction. Hence, from here on this sample will not be discussed in detail, however the results have been included for completeness. The remaining systems show that the methodology used for quantification of extent of reaction is able to shed light on the kinetics of reaction over the initial 10 hours. For high alkali and low alkali H-activated metakaolin, the extent of reaction appears to increase in a linear fashion, whereas for high alkali S-activated metakaolin, the extent of reaction is not linear. Nevertheless, the extent of reaction for the metakaolin systems appear to become similar after 6 hours, which is surprising given that previous investigations report significantly different microstructural development and macroscopic properties depending on the activator used (hydroxide or silicate) [4, 24]. However, as will be discussed later, there is a marked difference in the extent of reaction depending on the activator used, when the full reaction (128 days) is taken into account.

For the slag-based systems (Fig. 2b), the H-activated samples behave the same, irrespective of alkali content. For low-alkali S-activated slag, the initial stages of reaction progress faster than the H-activated slag systems. This apparent behavior, where the silicate-based slag reaction progresses faster than the hydroxide-based reaction, is the opposite to the hydroxide/silicate effect for metakaolin (where hydroxide-activated appears to progress faster than the silicate system). This difference in reaction progress depending on the activator and precursor used indicates that there are changes in the structural mechanisms controlling

reaction, depending on the precursor and activator used, which will be further discussed in later sections of this paper. Comparing high alkali H-activated metakaolin with slag in Fig. 2c shows that initially the slag system reaction proceeds more quickly than the metakaolin system. However, as will be discussed later, the true nature of this difference is more evident when the 128 day data sets are employed.

Fig. 1. In-situ X-ray pair distribution functions of (a) high alkali silicate-activated metakaolin, and (b) low alkali hydroxide-activated slag, obtained at times as marked.

(a)

(b)

Fig. 2. Calculated extent of reaction, using the 10 hour data set as the ‘fully reacted’ end point. (a) Metakaolin activated with high and low alkali hydroxide and silicate activators. (b) Slag activated with high alkali hydroxide activator, and low alkali hydroxide and silicate activators. (c) High alkali hydroxide activated metakaolin and slag.

3.3. Extent of reaction quantification: 128 days

In section 3.2, it was shown that certain information regarding the kinetics of reaction can be obtained by considering the initial 10 hours of reaction. However, by considering data collected at much longer reaction times (128 days), it is proposed that a more conclusive picture can be obtained regarding the kinetics of the reaction process. As in Fig. 2a, the various metakaolin systems are also displayed in Fig. 3, but here the final data set used was collected at 128 days. Fig. 3a shows that there are marked differences between the total extent of reaction achieved by the use of the various activators during the initial 10 hours of reaction. Fig. 3b displays the same time range as Fig. 2a, and it is evident that it is necessary to take into account structural changes taking place up to 128 days in order to obtain accurate information regarding the extent of reaction over the initial 10 hours. The high alkali S-activated metakaolin system proceeds at a much faster initial rate than the high alkali H-activated system, which was not apparent in Fig. 2a. Nevertheless, as was previously seen in Fig. 2a, there are no significant differences between high and low alkali H-activated metakaolin.

(a) (b)

(c)

Fig. 3. Extent of reaction plots, using the 128 day data set as fully reacted, for metakaolin activated with various activators. (a) Showing the entire extent of reaction. Note that the time scale is logarithmic to aid plotting. (b) Displaying the initial 10 hours of reaction.

(a)

(b)

For slag-based systems, there are also marked differences in the extent of reaction when the reaction is quantified using the 128 day data set as the end point (Fig. 4). As mentioned previously, the low alkali H-activated system is not included in this analysis due to fact the sample crystallized prior to 128 days. In Fig. 4a, it is apparent that the reaction of the H-activated slag system proceeds faster than the S-activated slag system, which is the opposite behavior to that of the metakaolin-based systems (Fig. 3), and contrary to what is seen to occur in the slag systems when only considering the initial 10 hours (Fig. 2b). Fig. 4b displays the extent of reaction for the slag-based systems together with the data for the corresponding metakaolin systems. Evident in this figure are differences in extent of reaction depending on the precursor used. For the high alkali H-activated systems, the slag sample proceeds much quicker towards the end state than the metakaolin sample. Furthermore, the extent of reaction over the initial 10 hours for the metakaolin sample appears to evolve linearly, whereas for the slag system the trend is closer to the kinetics of a pseudo-single step first-order rate expression (1 – e-kt), as was reported by Provis and van Deventer using in-situ energy dispersive X-ray diffraction for alkali activated metakaolin [25].

There are evident differences in the extent of reaction depending on the nature of the activator used. As was mentioned above, the reaction extent of H-activated metakaolin increases linearly over the initial 10 hours, irrespective of alkali content. For high alkali S-activated metakaolin the trend is somewhat different (Fig. 3b), where over the initial two hours the trend is linear, after which the trend of the extent of reaction becomes more akin to a pseudo-single step first-order rate expression, as described previously. It is apparent that the results presented in Fig. 3 and Fig. 4 show that the extent of reaction for various geopolymerization systems can be quantified using in-situ X-ray pair distribution function analysis. It is also evident that by taking into account changes up to 128 days, a more representative picture of the extent of reaction is obtained. Restricting consideration of the extent of reaction to the initial 10 hours of reaction fails to elucidate the true reaction kinetics since there are significant local nanostructural changes occurring in the gel binder past this point in time.

Provis and van Deventer used in-situ X-ray energy dispersive diffraction to quantify the extent of reaction for various metakaolin-based geopolymer systems [11]. Their data showed little change after 3 hours, and therefore the 3 hour diffractograms were used as the final data sets during quantification. Data were obtained at 7 days in that investigation, however, attempts to use these for quantification were unsuccessful due to the appearance of different structural features in the diffraction data (potentially related to medium-range ordering) which rendered the use of a simple linear interpolation such as Eq.3 invalid; this does not appear to be the case for the PDF data used here, which is a further potential advantage of this method. The results from the investigation by Provis and van Deventer revealed that in high-alkali potassium-based geopolymers, the hydroxide system proceeded faster than the silicate system [11]. This result deviated significantly from the observations in the current investigation, however, the influence of the nature of the alkali (potassium vs. sodium), as studied by Provis and van Deventer [11], may contribute to this difference.

Fig. 4. Extent of reaction plots, using the 128 day data set as fully reacted, for metakaolin and slag activated with various activators. (a) Showing the entire extent of reaction. Note that the time scale is logarithmic to aid plotting. (b) Displaying the initial 10 hours of reaction.

(a)

(b)

3.4. Outlook: coarse-grained Monte Carlo simulations

In-situ X-ray pair distribution function analysis has been shown here to be a useful local structure technique to quantify the extent of reaction as a function of time for amorphous systems, providing important information regarding the effects of the various activators and precursors on the geopolymerization reaction. However, the exact structural mechanisms responsible for these different behaviors remain largely elusive without the aid of complementary advanced simulations. Recently, density functional theory-based coarse-grained Monte Carlo analysis (DFT/CGMC) was shown to accurately replicate silicate speciation [26], and to provide important nanostructural mechanistic information regarding the geopolymerisation reaction for sodium silicate and hydroxide-activated metakaolin geopolymers [14].

A possible explanation for why the hydroxide-activated metakaolin system appears to progress much slower than the silicate-activated metakaolin system is related to the large changes occurring in the hydroxide system in terms of cluster size and percentage of silicate and aluminate monomers in solution. In the hydroxide system, the initial metakaolin particle completely dissolves, and forms small aluminosilicate clusters in solution [14]. On the other hand, the silicate system initially consists of a metakaolin particle surrounded by small silicate oligomers. Over the early stages of reaction, the metakaolin particles dissolves, but after a period of time the partially dissolved remnant metakaolin particle begins to grow from the silicate and aluminate monomers in solution, thereby forming large aluminosilicate gel particles indicative of the aluminosilicate geopolymer gel. There is a significant change in the overall number of sites existing in clusters in the hydroxide-activated system as the extent of reaction progresses, while in the silicate-activated system the number of sites existing in clusters remains relatively unchanged. Hence, given that there are significant changes occurring in the cluster size distribution in the hydroxide-activated system which would be expected to be observed in the local structural motifs present in the PDF data, the initial stages of reaction (first 10 hours) may proceed much more slowly than in the silicate-activated system, where the model shows less change in the total number of sites existing in clusters over the time of reaction.

In-situ X-ray pair distribution function analysis may thus provide important experimental validation of the local structural changes occurring during geopolymerization, as simulated using the DFT/CGMC modeling methodology, and may therefore supply an accurate time-scale for the simulation results. However, prior to being able to generate such a time-scale for the DFT/CGMC simulations, the on-lattice coarse-grained models need to be converted into energetically feasible atomistic representations. Only then will it be possible to use in-situ X-ray PDF analysis as a direct validation tool for advanced coarse-grained simulation methodologies such as DFT/CGMC.

4. Conclusions

It has been shown that in-situ X-ray pair distribution function analysis is a suitable experiment tool for probing the kinetics of reaction of metakaolin- and slag-based geopolymer gels. The amorphous nature of the precursors and binders will hinder the ability of conventional techniques to fully probe the progress of reaction. PDF analysis enabled direct quantification of extent of reaction up to 128 days for the geopolymer gels, unless crystallization occurred. Using the 128 day information for quantification was necessary in order to gain an accurate measure of the extent of reaction, since at ten hours significant changes were still occurring in all geopolymer samples. For metakaolin-based systems, the

extent of reaction of the silicate-activated sample proceeded faster than the hydroxide-activated sample. However, in the slag-based systems the influence of the activator nature on the progress of the reaction was reversed (hydroxide was faster than silicate), indicating that there are significant differences in the structural mechanisms occurring during geopolymerization depending on type of precursor and activator used. By combining the in-situ X-ray PDF results with previous conclusions drawn from density functional theory-based coarse-grained Monte Carlo simulations, it was postulated that the slower increase in extent of reaction in the hydroxide-activated metakaolin system (compared with silicate activation) is due to larger changes in number of sites existing in clusters as the reaction progresses. For the silicate-based metakaolin system, less structural changes are measured using in-situ X-ray PDF analysis due to the lower extent of changes occurring in the total number of sites existing in clusters.

Acknowledgments

The authors would like to thank Dr Karen Chapman and Kevin Beyer, ANL, for assistance with sample loading, data acquisition and data reduction on 11-ID-B at the APS, and Dr Dingwu Feng and Prof. Jannie van Deventer, University of Melbourne, for assistance in synthesis of slags and for valuable discussions. The participation of CEW, KP and NJH in this work was supported by Los Alamos National Laboratory, which is operated by Los Alamos National Security LLC under DOE Contract DE-AC52-06NA25396. Furthermore, CEW gratefully acknowledges the support of the U.S. Department of Energy through the LANL/LDRD Program. The 11-ID-B beam line is located at the Advanced Photon Source, an Office of Science User Facility operated for the U.S. DOE Office of Science by Argonne National Laboratory, under U.S. DOE Contract No. DE-AC02-06CH11357. The participation of JLP was supported by the Australian Research Council (ARC), including an ARC Linkage Grant jointly funded by Zeobond Pty Ltd, and through the Particulate Fluids Processing Centre.

References

1. Davidovits J. Geopolymers - inorganic polymeric new Mater.. J. Therm. Anal.. 1991;37(8):1633-1656.

2. Duxson P, Provis JL, Lukey GC, van Deventer JSJ. The role of inorganic polymer technology in the development of `green concrete'. Cem. Concr. Res. 2007;37(12):1590-1597.

3. Van Deventer JSJ, Provis JL, Duxson P. Technical and commercial progress in the adoption of geopolymer cement. Miner. Eng. 2011; in press, DOI: 10.1016/j.mineng.2011.09.009.

4. Duxson P, Provis JL, Lukey GC, Mallicoat SW, Kriven WM, van Deventer JSJ. Understanding the relationship between geopolymer composition, microstructure and mechanical properties. Colloid Surf. A. 2005;269(1-3):47-58.

5. Fernández-Jiménez A, Palomo A. Factors affecting early compressive strength of alkali activated fly ash (OPC-free) concrete. Materiales de Construcción. 2007;57(287):7-22.

6. Fernández-Jiménez AM, Palomo A, López-Hombrados C. Eng. properties of alkali-activated fly ash concrete. ACI Mater. J. 2006;103(2):106-112.

7. Kriven WM, Bell JL, Gordon M. Microstructure and microchemistry of fully-reacted geopolymers and geopolymer matrix composites. Ceram. Trans. 2003;153:227-250.

8. Palomo A, Banfill RFG, Fernández-Jiménez A, Swift DS. Properties of alkali-activated fly ashes determined from rheological measurements. Adv. Cem. Res. 2005;17(4):143-151.

9. Bell JL, Sarin P, Driemeyer PE, Haggerty RP, Chupas PJ, Kriven WM. X-Ray pair distribution function analysis of a metakaolin-based, KAlSi2O6•5.5H2O inorganic polymer (geopolymer). J. Mater. Chem. 2008;18(48):5974-5981.

10. Duxson P, Fernández-Jiménez A, Provis JL, Lukey GC, Palomo A, van Deventer JSJ. Geopolymer technology: the current state of the art. J. Mater. Sci. 2007;42(9):2917-2933.

11. Provis JL, van Deventer JSJ. Geopolymerisation kinetics. 2. Reaction kinetic modelling. Chem. Eng. Sci. 2007;62(9):2318-2329.

12. White CE, Provis JL, Llobet A, Proffen T, van Deventer JSJ. Evolution of local structure in geopolymer gels: an in-situ neutron pair distribution function analysis J. Am. Ceram. Soc. 2011;94(10):3532-3539.

13. White CE, Provis JL, Proffen T, van Deventer JSJ. 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(10):3486-3492.

14. White CE, Provis JL, Proffen T, Van Deventer JSJ. Molecular mechanisms responsible for the structural changes occurring during geopolymerization: multiscale simulation. AIChE J. 2011; in press, DOI: 10.1002/aic.12743.

15. Egami T, Billinge SJL. Underneath the Bragg Peaks: Structural Analysis of Complex Materials, Elmsford NY: Pergamon; 2003.

16. Bell JL, Sarin P, Provis JL, Haggerty RP, Driemeyer PE, Chupas PJ, et al. Atomic structure of a cesium aluminosilicate geopolymer: A pair distribution function study. Chem. Mater. 2008;20(14):4768-4776.

17. Meral C, Benmore CJ, Monteiro PJM. The study of disorder and nanocrystallinity in C-S-H, supplementary cementitious materials and geopolymers using pair distribution function analysis. Cem. Concr. Res. 2011;41(7):696-710.

18. Chupas PJ, Qiu XY, Hanson JC, Lee PL, Grey CP, Billinge SJL. Rapid-acquisition pair distribution function (RA-PDF) analysis. J. Appl. Crystallogr. 2003;36:1342-1347.

19. White CE, Provis JL, Proffen T, Riley DP, van Deventer JSJ. Combining density functional theory (DFT) and pair distribution function (PDF) analysis to solve the structure of metastable materials: The case of metakaolin. Phys. Chem. Chem. Phys. 2010;12(13):3239-3245.

20. White CE, Provis JL, Proffen T, Riley DP, van Deventer JSJ. Density functional modeling of the local structure of kaolinite subjected to thermal dehydroxylation. J. Phys. Chem. A. 2010;114(14):4988-4996.

21. Lloyd RR, Provis JL, van Deventer JSJ. Microscopy and microanalysis of inorganic polymer cements. 1: remnant fly ash particles. J. Mater. Sci. 2009;44(2):608-619.

22. Provis JL. Activating solution chemistry for geopolymers. In: Provis JL, van Deventer JSJ, editors. Geopolymers: Structures, Processing, Properties and Industrial Applications. Cambridge, UK: Woodhead; 2009. p. 50-71.

23. Provis JL, van Deventer JSJ. Direct measurement of the kinetics of geopolymerisation by in-situ energy dispersive X-ray diffractometry. J. Mater. Sci. 2007;42(9):2974-2981.

24. Duxson P, Mallicoat SW, Lukey GC, Kriven WM, van Deventer JSJ. The effect of alkali and Si/Al ratio on the development of mechanical properties of metakaolin-based geopolymers. Colloid and Surf. A. 2007;292(1):8-20.

25. Provis JL, van Deventer JSJ. Geopolymerisation kinetics. 1. In situ energy-dispersive X-ray diffractometry. Chem. Eng. Sci. 2007;62(9):2309-2317.

26. White CE, Provis JL, Proffen T, van Deventer JSJ. Quantitative mechanistic modeling of silica solubility and precipitation during the initial period of zeolite synthesis. J. Phys. Chem. C. 2011;115(20):9879-9888.