Formation of Ceramics from Metakaolin-Based Geopolymers: Part I—Cs-Based Geopolymer


  • Jonathan L. Bell,

    1. Department of Materials Science and Engineering, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801
    Search for more papers by this author
    • *Member, The American Ceramic Society.

  • Patrick E. Driemeyer,

    1. Department of Materials Science and Engineering, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801
    Search for more papers by this author
    • *Member, The American Ceramic Society.

  • Waltraud M. Kriven

    Corresponding author
    1. Department of Materials Science and Engineering, University of Illinois at Urbana-Champaign, Urbana, Illinois 61801
      †Author to whom correspondence should be addressed. e-mail:
    Search for more papers by this author
    • *Member, The American Ceramic Society.

    • **Fellow, The American Ceramic Society.

  • G. Scherer—contributing editor

  • This work was supported by Air Force Office of Scientific Research (AFOSR), USAF, under Nanoinitiative Grant No. FA9550-06-1-0221, through Dr. Joan Fuller.

†Author to whom correspondence should be addressed. e-mail:


The structural evolution and crystallization of a cesium-based geopolymer (Cs2O·Al2O3·4SiO2·11H2O) on heating was studied by a variety of techniques including X-ray diffraction, thermal analysis, dilatometry, pycnometry, specific surface area, and microstructural investigation. The Cs geopolymer gradually crystallized into pollucite (Cs2O·Al2O3·4SiO2) on heating above 900°C. Its low crystallization temperature is believed to be due to the presence of nuclei in the geopolymer precursor, which are formed after curing at 50°C for 24 h. The Cs-based geopolymer was found to be more refractory compared with K- and Na-based geopolymers. Significant shrinkage, due primarily to viscous sintering, did not occur until the samples were heated to above 1200°C. The microstructure of unheated geopolymer had ∼20–30 nm-sized precipitates that coarsened on heating above 1000°C. By 1350°C, the geopolymer surface had a smooth, glassy texture, although large macropores and closed pores remained. After heating to 1600°C, the closed pores were removed, and the geopolymer reached ∼98% of the theoretical density of pollucite. Higher than expected levels of Cs were found near large voids, as seen by scanning electron microscopy and transmission electron microscopy analysis. The presence of this extra Cs was due to Cs left behind in pore water, which was not bound within the geopolymer structure.

I. Introduction

Geopolymers are a class of cementious materials that are formed by mixing aluminosilicate materials (e.g., metakaolin, low-calcium fly ash) with alkali or an alkali-silicate solution.1–4 When cured below 100°C, they are generally X-ray amorphous and consist of cross-linked AlO4 and SiO4 tetrahedra, where charge balance is provided by hydrated alkali metal cations.1 Geopolymers are typically synthesized using K+ or Na+ as the charge-balancing alkali cation, leaving Cs-based geopolymers relatively unexplored, although some attention has been paid to these compositions for treatment of radioactive waste streams.5–7

Geopolymers are increasingly being considered for use in a variety of refractory applications8–10 and as precursors to ceramic formation.11–13 A Cs-based geopolymer of the composition Cs2O·Al2O3·4SiO2·11H2O crystallizes into pollucite (Cs2O·Al2O3·4SiO2) upon heating.14 Pollucite is the most refractory silicate known (Tmelt >1900°C),15 and has exceptional creep resistance between 1400° and 1500°C, comparable to the two most creep-resistant ceramic oxides: mullite and yttrium aluminum garnet.16 The combination of creep resistance, high melting temperature, and low thermal expansion makes pollucite and pollucite glass–ceramics candidates for use in ceramic matrix composites, thermal-shock resistant molds, and in a variety of other refractory applications. In addition, significant industrial interest in pollucite is based around its use as an encapsulant for radioactive 137Cs.17–19

Crystallization of pollucite from a room–temperature-synthesized geopolymer represents a novel approach to forming pollucite-based ceramics. Pollucite has been synthesized previously by a variety of techniques including hydrothermal20,21 or dry oxide-based syntheses,17 ion exchange from leucite,22 and from sol–gel-derived precursors.19,23 Pollucite prepared via melting of mixed oxides may require heat treatments above 1850°C.15 Additionally, high sintering temperatures are required, resulting in volatization of Cs, and high-density pollucite is difficult to attain.16,21 Using geopolymers, it may be possible to form dense pollucite ceramics at a low temperature with minimal Cs volatilization. Additionally, geopolymers can be processed more economically compared with sol–gel precursors as they do not require expensive materials such as tetraethyl orthosilicate (TEOS) or tetramethyl orthosilicate. Moreover, geopolymers can be easily cast into complex monolithic shapes, which can potentially be heated to form intricately shaped ceramics and ceramic composites.

Despite the potential of geopolymers as refractories and ceramic precursors, the thermal evolution and crystallization in a Cs-based geopolymer has not been explored. The thermal evolution of Na- and K-based geopolymers on heating has been considered only recently.24–26 Therefore, in this study, changes in the structure of a Cs2O·Al2O3·4SiO2·11H2O geopolymer on heating were examined using a variety of methods. The physical evolution and densification on heating was studied using dilatometry, nitrogen absorption/desorption (by the Brunauer–Emmett–Teller—BET method), pycnometry, thermograviometric analysis, and electron microscopy. Crystallization of the geopolymer into pollucite was examined using X-ray diffraction (XRD), differential scanning calorimetery and (DSC), and transmission electron microscopy (TEM).

II. Experimental Procedures

(1) Geopolymer Synthesis

A cesium silicate solution (Cs2O·2SiO2·11H2O) was prepared by dissolving cesium hydroxide hydrate (99.9 wt% metals basis, Alfa Aesar, Ward Hill, MA) in deionized water, followed by addition of amorphous fumed silica (Cabot M-5, T. H. Hilson Company, Wheaton, IL). The cesium silicate solution was allowed to mix for at least 2 days to ensure sufficient proper time for silica dissolution and equilibration. A geopolymer of the composition Cs2O·Al2O3·4SiO2·11H2O (deemed CsGP) was then prepared by mixing metakaolin (Metamax High Reactivity Metakaolin, BASF Corporation, Iselin, NJ) into the cesium silicate solution using an IKA overhead mixer (Model RW 20, Wilmington, NC) equipped with a dispersion blade. The metakaolin used in this study has an average particle size of 1.2 μm, a specific surface area of 13 m2/g, and had >97% purity.

The resultant slurry was cast into 50 mL centrifuge tubes (Corning Inc., Corning, NY), sealed, and cured at 50°C for 24 h. The hardened geopolymer was then removed from the tubes, and fractured into ∼10 g specimens. Samples were prepared for scanning electron microscopy (SEM), BET, XRD, and pycnometry analysis by heating the fracture specimens in a Radatherm high-temperature furnace (Model HT 05/18, Wetherill Park, NSW, Australia) to 1020°, 1040°, and between 900° and 1400°C, using 50°C steps in a platinum crucible. These samples were heated and cooled at 10°C/min with no isothermal soak. An additional sample was prepared, in which the CsGP was heated to 1100°C at 10°C/min, but was held at this temperature for 24 h.

(2) Analytical Techniques

SEM analysis was performed on both fracture and polished surfaces using a Hitachi S-4700 high-resolution SEM (Tokyo, Japan). Polished samples were prepared on a Buehler grinder/polisher (Model Ecomet III, Lake Bluff, IL) and were subsequently etched in 3 wt% HF at 45°C for 15 s. All polished SEM samples were mounted on aluminum stubs and were sputter coated with ∼6 nm of a Au/Pd alloy to facilitate imaging. Additional fracture samples were prepared for energy-dispersive X-ray analysis (EDS) by carbon coating. Flat sample regions were examined using a Jeol 6060 LV SEM (Tokyo, Japan) operated at 20 kV and 10 mm working distance. Copper was used to calibrate the energy, and a minimum of six acquisitions were taken on each sample.

BET analysis was performed on solid fracture samples using a Quantachrome Instruments Surface Area and Pore Analyzer (Model Nova 2200e, Boynton Beach, FL). Samples were degassed and dried under vacuum at 150°C for 24 h before being analyzed using nitrogen adsorption/desorption. Surface areas were calculated using the BET method.27 Additionally, in order to estimate the pore size in unheated CsGP, BET micropore analysis was performed using an Accelerated Surface Area and Porosimetry System (Model ASAP 2010, Micromeritics, Norcross, GA).

XRD patterns were collected using a Rigaku X-ray powder diffractometer (Model D-max II, Danvers, MA) equipped with a CuKα source (λ=0.1540598 nm). A single-crystal monochromator in the diffracted beam path was used to acquire XRD patterns in Bragg-Brentano geometry, over a 2θ range of 5°–70° with a step size of 0.02°. Before X-ray analysis, fracture specimens were ground to powders and sieved to <325 mesh (<44 μm). X-ray patterns were subsequently examined using Jade 7 software (Minerals Data Inc., Livermore, CA). The density of the sample powders was determined using a helium-based pycnometer (Model 1330, Micromeritics). After being filled with powder, the sample chamber was purged 50 times before analysis to ensure removal of atmospheric gases. A total of 10 measurements were acquired for each sample.

Simultaneous thermal gravimetric analysis (TG) and DSC studies were conducted on 15–20 mg solid fracture chunks, 99–149 μm-sized powders, and <44-mm-sized powders, which were heated up to 1400°C at 10°C/min in a Netzsch DSC/TG (Model STA409 CD, Export, PA). A platinum pan fitted with a lid was used to hold the specimen and as a reference. During the analysis, the sample chamber was purged with He (25 mL/min) and air (50 mL/min). Specimens were prepared for dilatometry by casting CsGP into a 3.97-mm-diameter sealed Tygon tube and curing at 50°C for 24 h. Hardened CsGP cylinders were subsequently cut to ∼10 mm length using a Buehler low-speed diamond saw (Isomet series) to ensure their ends were flat, and were analyzed in a Netzch dilatometer (Model DIL 402 E) up to 1500°C at 10°C/min in air. A single-crystal sapphire cylinder was used as a calibration standard.

TEM and TEM/EDS work was performed on both fracture and thin, ion-milled samples using a Jeol 2010F scanning transmission electron microscope (S)TEM. This TEM was operated at 200 kV and was equipped with an Oxford INCA 30 mm ATW EDS detector (Oxford Instruments Concord NanoAnalysis Sales, Concord, MA). Fracture samples were prepared by grinding CsGP using a mortar and pestle, followed by dispersion in ethanol, and deposition on holey, carbon-coated copper grids. Ion-milled samples were prepared by slicing 300-μm-thick disks from 25.4 mm length × 6.35 mm CsGP cylinders using a Buehler low-speed diamond saw. The disks were then heated to 1100°C for 24 h in a Radatherm (Model HT 05/18) high-temperature furnace, and were subsequently cut into 3 mm-diameter disks using a Gatan ultrasonic disk cutter (Model 601, Warrendale, PA). The 3-mm-disks were thinned to 100 μm using a Buehler Minimet disk polisher (Model Ecomnet III), followed by dimple grinding to 20 μm with a Gatan dimple grinder (Model 656). Finally, samples were ion milled at a low temperature using a Fischione ion mill (Model 2).

III. Results and Discussion

(1) X-Ray Analysis

Crystallization of pollucite from CsGP occurred gradually on heating as shown in Fig. 1. Unheated CsGP had a number of broad diffuse peaks centered near 24.2°, 28.0°, 30.8°, 37.2°, 40.8°, 49.0°, and 52.8°2θ. On heating, these peaks grew in intensity and represent the (321), (400), (332), (440), (611), (721), and (651) reflections in cubic pollucite, respectively. Although pollucite formation was more obvious after heating to ≥1050°C, there was no definitive transformation temperature observable from the X-ray results. Therefore, thermal analysis was deemed necessary to better characterize the details of pollucite crystallization, and is discussed later.

Figure 1.

 X-ray diffractograms for Cs geopolymer, and after being heated to the specified temperatures (in °C) shown to the left at 10°C/min heating and cooling rates, with no soak.

Based on X-ray results, it was concluded that the pollucite phase formed on heating CsGP remained in the cubic state after cooling to room temperature. Pollucite is expected to have a cubic to tetragonal phase transition from I41/a to Ia3d near room temperature.22,23 Tetragonal pollucite has a number of unsymmetrical doublet reflections in the most intense (400) peak, as well as higher order reflections such as the (631) and (732) peaks. These doublets were not observed throughout the entire X-ray pattern for CsGP after being heated to ≥1050°C and cooled to room temperature.

Using CsGP, only a pure pollucite phase was observed in the X-ray pattern regardless of the calcination temperature over the range of 25°–1400°C. However, this does not exclude the possibility that amorphous phases may have formed on heating, or that additional phases may have formed that are below the detection limits of the instrument (typically 1%–3%). In pollucite produced by both conventional and gel-based methods, Cs2O·Al2O3·2SiO2 and Cs2O·Al2O3·10SiO2 are common impurities that form on heating.17 The formation of these phases may be due to volatization of Cs-rich species or insufficiently mixed alumina-silica sources. However, in pollucite prepared from CsGP, Cs volatilization will be minimized due to the lower formation temperatures required.

The diffuse peaks seen in the X-ray pattern for unheated CsGP (Fig. 1) may suggest that small crystallites of zeolitic pollucite formed under the hydrothermal processing conditions. Because of similarities in the processing, chemistry, and structure, geopolymers have been referred to as the amorphous, metastable equivalents of crystallographically ordered zeolites.1,2,28–30 Compared with zeolites, geopolymers are typically prepared at lower temperatures, using more concentrated solutions, and in the absence of templating agents. However, there is no clear distinction as to what processing conditions are required for zeolite formation to occur. Maclaren et al.21 found that pollucite could be formed hydrothermally in an autoclave at 220°C and a 2-h duration using a mixture of amorphous SiO2, CsOH solution, and Al powder. The CsGP used in this study was cured in a sealed container at 50°C for 24 h. It is unclear from the X-ray results whether this was sufficient to produce small zeolite crystallites within the material. In a recent study, the short to medium-range structural order of unheated CsGP was found to resemble that of pollucite out to ∼9 Å using the pair distribution function (PDF) method.31

It was suggested that on heating, only minor atomic rearrangements were required for long-range order development (i.e., pollucite crystallization). Precursor sol–gels used to fabricate pollucite generally do not contain the ordering shown in the X-ray pattern of unheated CsGP. Hogan and Risbud32 synthesized pollucite from a mixture of TEOS, aluminum di-acetoacetic ester chelate, and cesium acetate. This precursor mixture was found to be X-ray amorphous after calcination at 300°C for 1 h and 750°C for 4.5 h and contained only a broad hump near 25°2θ. Similarly, Xu et al.19 found that the pollucite precursor gel made from TEOS, aluminum tri-sec-butoxide, titanium isopropoxide, and cesium hydroxide was amorphous before calcination.

In some sol–gel materials, the anionic packing and cation distribution in the gel are closely related to that of the crystalline phase, thus allowing complex crystalline phases to form at a low temperature topotactically without the need for diffusion.33 The ordering between CsGP and pollucite may suggest a similar mechanism. The higher level of order for CsGP can be attributed to a number of factors including the setting time, solution viscosity, silicate solution speciation, ion size, and hydration potential, as well as the dissolution of source materials. However, a detailed consideration of these factors is beyond the scope and central focus of this work.

(2) DSC Analysis

As shown in Fig. 2, the DSC pattern for CsGP had a large endotherm from 25° to 300°C. This endotherm had a minimum value at ∼100°C for the three sample types analyzed, which included a solid fracture specimen, 99–149-μm-sized powders, and 44-mm-sized powders. Based on hydration considerations alone, the endothermic minimum for CsGP would be expected to occur below 70°C. Duxson et al.24 observed a similar endotherm in K- and Na-based geopolymers, with a minimum value occurring at 70° and 100°C, respectively, for samples heated at 10°C/min. It was suggested that the Na geopolymer had a higher value due to the increased hydration energy of Na+(aq) compared with K+(aq).

Figure 2.

 Differential scanning calorimetery results for CsGP (a) solid fracture chunk, (b) 99–149 μm sized powders, and (c) <44 mm sized powders. A broad exotherm representative of cubic pollucite crystallization had a maximum value around 1120°, 1165°, and 1175°C respectively. All samples were heated at 10°C/min to 1400°C.

The larger than expected value for CsGP can be attributed to various factors including: the higher degree of ordering, the characteristic pore size and surface area, and cation–water, cation–framework, and framework—water interactions. Even though Cs has a lower hydration energy, its large size makes it more tightly bound and strongly attracted to the tetrahedral framework in CsGP compared with Na- and K-based geopolymers. These trends have been observed in the synthesis of zeolite beta using different alkali cations (Na, K, Rb, and Cs).34 In more compact zeolite structures such as analcime, ion exchange of Na+ with larger K+ or Cs+ leads to the creation of an anhydrous structure.35 Geopolymers set rapidly and do not have sufficient time to form the open zeolitic structures, so that it is unclear how water will be structurally bound within the geopolymer framework. However, 2H MAS NMR analysis has shown that water is more tightly bound in geopolymers synthesized with a high Si/Al ratio near 2, such as in the composition studied here.36

Evaporation is expected to occur more easily in materials with less ordered crystal structures. For example, evaporation in a low-silica (SiO2/Al2O3=2.3) amorphous K-based geopolymer occurred over a lower temperature range compared with the crystalline Na-based geopolymer.24 In crystalline zeolites, the presence of an endotherm is often observed on heating and is related to the heat needed to dehydrate cations within the zeolite framework and water molecules within channels and voids.37,38 In the more ordered zeolites, the minimum value for this endotherm generally occurs at a higher value than what is observed for geopolymers. For example, in sodium-based zeolite A synthesized from fly ash, the minimum of this endotherm was at 155°C.38 In two natural analcime zeolites, the minimum value of this endotherm was above 300°C in samples heated at 5°C/min.39 However, geopolymers contain the bulk of their water as free water rather than structurally bound water, as in zeolites. This can shift the minimum of the endotherm to a lower value.

Consistent with X-ray results, the DSC pattern for CsGP had a broad exotherm, indicating that the conversion to crystalline pollucite occurred gradually (Fig. 2). The onset of this exotherm occurred around 900°C, and extended until 1200°–1250°C. However, the characteristic peak shape, position, and peak crystallization temperature (Tp) were dependent on the type of CsGP sample analyzed. Mazza and LuccoBorlera40 synthesized pollucite via sol–gel and did not observe pollucite crystallization until 1360°C as evidenced by a sharp DSC exotherm. The broadness of the exotherm for CsGP could be related to the presence of nuclei at a low temperature. Ray et al.41 reported that an increase of nuclei in glass can cause a decrease in the DTA peak temperature.

For these three sample types, Tp was at 1120°C for the solid sample, 1165°C for 99–149-μm-sized powders, and 1175°C for <44-μm-sized powders. The crystallization exotherm for powders had a more symmetric shape compared with that for the solid fracture chunk. These various sample types were examined in order to help elucidate the crystallization mechanism. For example, Reynoso et al.42 observed two crystallization peaks in the DSC pattern of powdered phosphate glass, while only one peak was observed in a bulk sample. It was suggested that when two peaks are present, surface nucleation and crystallization were responsible for the first peak, while bulk nucleation and crystallization were associated with the second peak.

Surface nucleation and crystallization may not play a role if the nuclei responsible for pollucite crystallization from CsGP are already present after curing. This would explain the additional ordering observed in the X-ray pattern for unheated CsGP, and the similarities in the short- to medium-range order between CsGP and pollucite observed using the PDF method.31 In addition, only one peak was observed in the DSC pattern, regardless of the sample type. Finally, the presence of more nuclei (both bulk and surface) has been shown to lower Tp.43 For CsGP, Tp was found to increase with decreasing powder size. If surface nucleation was prevalent, it is expected that Tp would have decreased with the powder size.

Over the entire temperature range examined (25°–1400°C), no well-defined glass transition temperature (Tg) for CsGP was observed. Geopolymers are expected to be similar to sol–gel glass precursors, which require densification at an elevated temperature before a well-defined Tg can be observed. If left undensified, the conversion to a glass structure on heating sol–gel glass precursors continually evolves toward a more highly polymerized state, thus obscuring a definitive Tg.44 Similarly, geopolymers undergo condensation and polymerization on heating due to elimination of water from silanol and aluminol groups.24,26 Rahier et al.45 observed a Tg at 650°C in a predensified Na-based geopolymer using thermal expansion and dynamic mechanical analysis. However, CsGP is expected to be more refractory compared with Na- and K-based geopolymers. In the crystalline phase, larger ions form a more refractory material such that the melting point of CsAlSi2O6 (Tmelt>1900°C)15>KAlSi2O6 (Tmelt∼1693°C)46>NaAlSi2O6 (Tmelt∼1040°C).47 Hogan and Risbud32 found that the Tg for a sol—gel-derived pollucite glass–ceramic occurred at 945°C. In this case, the precursor was densified at 750°C for 4.5 h before being analyzed via DSC. For the samples shown in Fig. 2, the CsGP was not densified at an elevated temperature before being analyzed. However, even if the samples were initially densified before DCS analysis, it is likely that Tg would be obscured by the broad pollucite crystallization exotherm.

An endotherm due to melting also did not occur over the entire temperature range examined in the DSC (25°–1400°C), which is likely due to the formation of a high yield of crystalline pollucite. In order to facilitate pollucite formation, the starting composition of CsGP (Cs2O·Al2O3·4SiO2·11H2O) was equivalent to that of pollucite (CsAlSi2O6). Beall and Rittler15 found that glass formation in the Cs2O-SiO2-Al2O3 system was widespread. The optimum composition range for glass formation was found to be in the area of 25–70 wt% SiO2, 20–50 wt% Al2O3, and 10–35 wt% Cs2O. However, glass compositions close to the pollucite stoichimetry could not be melted, even at 1900°C. Similarly, the composition for CsGP was equivalent to pollucite, and melting of a glassy phase was not observed in the DSC.

(3) Surface Area and Dilatometry Analysis

Consistent with the observations of Duxson et al.24–26 for Na- and K-based geopolymers, the thermal behavior of CsGP can be categorized into four regions. As shown in Fig. 3, Region I extended from approximately 25° to 100°C and involved rapid loss of freely evaporable water and minimal shrinkage. Nearly 50% of the overall weight loss occurred within this region due to water loss. In Region II, weight loss and shrinkage occured concomitantly over the range of 100°–300°C. Shrinkage in this region was much higher than in Region I and ∼94% of the remaining water was lost. Region III involved nominal weight loss and shrinkage from ∼300° to 1200°C. Above 1200°C, there was considerable shrinkage and nominal weight loss in Region IV.

Figure 3.

 Dilatometry and thermal gravimetry (TG) results for CsGP heated at 10°C/min up to 1450°C in air. TG results were collected using solid fracture specimens. Dilatometry specimens were prepared by casting the CsGP into Tygon tubing to produce 3.97 mm diameter ×∼10 mm length cylinders. The characteristic regions are numbered at the bottom of the plot.

The shrinkage in Region I was minimal as only free water from large pores and surfaces was lost in this region.45 Shrinkage in this region accounted for only 0.07% of the overall shrinkage in samples heated up to 1450°C. Over Region II, the amount of shrinkage was much greater and water loss was due to desorption of water from small pores and release of chemically bound water in the form of silanol and aluminol groups at higher temperatures. Shrinkage in this region accounted for 8.78% of the overall shrinkage and arises because of capillary forces created as water evaporates from small pores within the geopolymer structure.45,48 The magnitude of the capillary pressure is proportional to γ/r, where r is the radius of a cylindrical capillary and γ is the surface tension of water (0.073 N/m). In unheated CsGP, the median pore radius was determined to be 1.65 nm using gas absorption micropore analysis. Pores of this size will have considerable capillary pressures on the order of 107 N/m2.

As shown in Fig. 4, the surface area in Region II reduced from 116 m2/g at 100°C to 103 m2/g by 300°C, while the density (measured via pycnometry) increased from 2.65 to 2.93 g/cm3. This substantial increase in the density confirms that there was significant capillary contraction of small pores and shrinkage in this region. However, there was little change in the surface area due to the competing factors of capillary contraction and water evaporation. Despite the high surface area and small gel pore size, the shrinkage for CsGP in Region II was less than that observed for Na- and K-based geopolymers.24,26

Figure 4.

 Apparent density measured on <44 μm sieved sample powders using pycnometry and specific surface area determined using the Brunauer–Emmett–Teller method on solid fracture specimens. Results were collected on CsGP after being heated to the specified temperature and cooled to room temperature at 10°C/min.

The nominal weight loss in Region III was due to removal of water by polycondensation of T–OH groups (T=Si, Al), also known as dehydroxylation. Because there is only a small amount of water present as T–OH groups,49 only nominal weight loss occurred in this region. Shrinkage in this region can conceivably arise from increased packing efficiency and skeletal densification.50 Skeletal densification is due to polycondensation and structural relaxation of the tetrahedral network.50 In region III (from 300° to 1200°C), the density increased from 2.93 to 3.15 g/cm3, while the specific surface area decreased from 103 to 39 m2/g. The shrinkage in this region accounted for 8.3% of the overall shrinkage in samples heated up to 1450°C.

The significant shrinkage in Region IV was due primarily to viscous sintering above 1200°C, and accounted for 82.8% of the overall shrinkage (Fig. 4). In viscous flow sintering, atomic movement occurs by a cooperative motion mechanism rather than individual diffusion.33 This mechanism has been well studied in glasses and sol–gel materials, and is most notable in silicate systems. In Region IV (>1200°C), the specific surface area was reduced from 39 to 19 m2/g. However, as shown in Fig. 4, the apparent density measured via pycnometry decreased from 1100° to 1400°C, and increased on heating above 1400°C. By 1600°C, the density reached a value of 3.19 g/cm3 (∼98% of the theoretical density of cubic pollucite).23 The decrease in density over 1100°–1400°C was due to the formation of closed pores. In Na-based geopolymers, a similar trend was observed using BET measurements to estimate skeletal density.26 It was found that the skeletal density artificially reduced above 600°C due to inaccessible porosity created as the gel underwent viscous sintering. The onset of Region IV in Na and K geopolymers was found to occur at around 700°–800°C.24,26 Therefore, CsGP was much more refractory in comparison.

(4) Microstructural Analysis

As shown in Figs. 5(a)–(c), the structure of unheated CsGP contained fine ∼20–30-nm-diameter precipitates, which coarsened to ∼30–40 nm after being heated to 1020°C and ∼50–60 nm after being heated to 1040°C. CsGP developed a smooth, glassy texture after being heated to 1350°C (Fig. 5(d)), although large voids from partially reacted metakaolin were still present. This explains the reduction in density observed in Fig. 4 on being heated from 1100° to 1400°C. The significant coarsening and surface area reduction led to the creation of closed pores above 1100°C that persisted until the sample was heated to 1600°C. As shown in Figs. 6(a) and (b), partially reacted metakaolin particles were surrounded by voids that persisted on heating.

Figure 5.

 High-resolution scanning electron microscopy micrographs of (a) unheated CsGP and after being heated to (b) 1020°C, (c) 1040°C, (d) 1350°C. Micrographs (a)–(d) are fracture surfaces. All Samples were heated and cooled at 10oC/min with no isothermal soak. The initial 20–30 nm sized precipitates observed in unheated CsGP coarsened on heating to 1020°C due to viscous sintering.

Figure 6.

 High-resolution scanning electron microscopy micrographs of (a) unheated CsGP and (b) CsGP after being heated to 1040°C (10°C/min with no isothermal soak) showing a region of microstructure containing partially reacted metakaolin. Micrographs (a) and (b) are both fracture surfaces.

Although pollucite crystals could not be observed in SEM analysis of fracture surfaces in Figs. 5 and 6, their formation was obvious in samples heated at 1100°C for 24 h Figs. 7(a) and (b). The presence of 1–5-μm-sized pollucite grains is shown in Fig. 7(a). Polishing and etching of this sample with 3 wt% HF acid at 45°C for 15 s revealed that pollucite grains were surrounded by glassy grain boundaries, which etched preferentially. EDS analysis of the grains confirmed that they were of pollucite composition (CsAlSi2O6). Regions of the sample near pores were found to contain higher amounts of Cs than expected for pollucite. However, because of the large interaction volume in SEM-EDS, TEM analysis was deemed necessary to further examine microstructural and compositional variations.

Figure 7.

 Scanning electron microscopy micrographs of (a) CsGP fracture surface after being heated to 1100°C for 24 h at 10°C/min and (b) the same sample after being polished and etched with 3 wt% HF acid at 45°C for 15 s.

TEM micrographs for unheated CsGP are shown in Figs. 8(a) and (b). Consistent with SEM observations, the structure of CsGP was comprised of 20–30 nm-sized spherical precipitates. Spherical precipitates have also been observed in Na- and K-based geopolymers, but are generally smaller in size (∼5–10 nm).5,51 As was stated previously, the diffuse reflections in the X-ray pattern of unheated CsGP may suggest the presence of small pollucite crystallites (Fig. 1). It has also been suggested that geopolymers may contain nanocrystalline zeolites compacted in an amorphous gel phase.52 However, the precipitates observed in unheated CsGP via TEM all appeared to be amorphous.

Figure 8.

 Transmission electron microscopy micrographs for CsGP showing (a) low magnification view and (b) high magnification view of amorphous precipitates. Samples were prepared by grinding with a mortar and pestle, dispersing in ethanol, and evaporating the ethanol onto a holey, carbon-coated copper grid.

TEM micrographs for CsGP heated to 1100°C for 24 h are shown in Figs. 9(a) and (b). Despite being heated at 1100°C for 24 h, this sample still contained a high degree of porosity. The bright-field micrograph shown in Fig. 9(a) shows a thin section of the sample edge prepared by standard polishing and ion milling. The sample edge contained pores on the order to 100 nm diameter, as well as dark spots that were identified as pollucite grains, as shown by the Moiré fringes in Fig. 9(b). TEM-EDS analysis of the sample edge shown in Fig. 9(a) confirmed that the composition was close to that of pollucite. The average Si/Al and Cs/Al ratios were 2.09±0.09 and 0.79±0.9, respectively. The lower than expected Cs/Al ratio was due to migration of Cs under the 200 kV electron beam. High-resolution examination of hydrous silicate is known to cause rapid amorphization53 and migration of alkali.54 In the EDS results, the Cs content was initially higher, and decreased with beam exposure.

Figure 9.

 Transmission electron microscopy (TEM) micrographs for CsGP after being heated to 1100°C for 24 h showing (a) low magnification view of thinned sample edge containing ∼100 nm sized pores and (b) high magnification view confirming that dark spots in micrograph (a) were pollucite crystallites. Samples were prepared by standard TEM polishing and ion milling techniques.

Consistent with SEM results, TEM-EDS analysis of the Cs content near the edges of large voids was found to be much higher than that expected for pollucite. These voids were left behind by partially reacted metakaolin, and were not removed after heating at 1100°C for 24 h. In metakaolin-based geopolymers, especially those with a Si/Al=2, it is well known that not all of the metakaolin is dissolved before setting.51,55,56 Therefore, because not all the Al was utilized, there will be excess Cs not bound within the structure. Excess alkali has been found to persist in the pore water of some geopolymers.36,57 Additionally, alkali has been shown to leach out when metakaolin-based geopolymers were placed in deionized water.58 On heating CsOH, pore water will be evaporated, leaving behind alkali on pore surfaces. On further heating, the alkali can react with silica and alumina, forming various crystalline or amorphous Cs-silicates or Cs-aluminosilicates.59

V. Conclusions

In this study, the thermal evolution and crystallization of metakaolin-based, Cs2O·Al2O3·4SiO2·11H2O composition geopolymer was studied using a variety of methods. The Cs geopolymer was found to be a novel precursor in the formation of pollucite (Cs2O·Al2O3·4SiO2). The X-ray pattern of the unheated Cs geopolymer had a higher degree of ordering than typically observed in sol-gel or glass-based precursors used to fabricate pollucite. On heating, pollucite crystallization occurred gradually and exhibited an exotherm in the DSC over the range of 900°–1250°C. After heating at 1100°C for 24 h, the geopolymer formed 1–5-μm-sized pollucite grains, which were surrounded by glassy grain boundaries. After being heated to 1600°C, the geopolymer reached a density of 3.18 g/cm3, which was ∼98% of the theoretical density of pollucite.

Consistent with previous observations for Na- and K-based geopolymers, the thermal behavior of CsGP could be categorized into four regions based on dilatometer results. At low temperatures (∼25°–300°C), shrinkage was due to evaporation of free water from surfaces and small pores. Over the range of ∼300°–1200°C, chemically bound −OH groups were removed from the tetrahedral network and shrinkage was minimal. Above 1200°C, there was significant shrinkage due primarily to viscous sintering. Changes in the microstructure were observed in samples heated to ≥1020°C. On heating, spherical precipitates present in the unheated geopolymer noticeably coarsened by 1020°C due to viscous sintering. After being heated to 1350°C, the CsGP attained a smooth, glassy texture. Based on the thermal property observations, Cs-based geopolymers were found to be much more refractory compared with Na- and K-based systems. Cs geopolymers can therefore be used in higher end applications, where thermal performance is the primary concern and cost is less of an issue.


The authors acknowledge the use of facilities at the Center for Microanalysis of Materials, in the Frederick Seitz Research Laboratory at the University of Illinois at Urbana-Champaign, which is partially supported by the U.S. Department of Energy under grant No. DEFG02-91-ER45439. The authors would also like to thank Dr. E. R. (Lou) Vance of ANSTO, Australia, for his help in collecting the BET micropore data.