Characterization of Aged Slag Concretes

INTRODUCTION

Improved durability has been reported for a wide range of slag and blended cement/slag concretes when compared with ordinary portland cement (OPC).1-7 A detailed review of the literature in the field has recently been presented by Shi et al.7 It is generally believed that the compact structure of calcium silicate hydrates (C-S-H), refined pore size distribution, and low permeability are responsible for improved durability. It is also likely that the low Ca/Si ratio in slag aids resistance to sulfate and acid. Most studies on durability of slag concretes are conducted in the laboratory by immersing samples in specific aggressive solutions, then analyzing at intervals to determine the extent of corrosion. The influence of weathering action on microstructure and on overall durability, however, is not addressed sufficiently in these studies. Extended-duration studies (up to 30 years) of blended OPC/slag concretes under environmental conditions have been reported,2,5,6 but comparable studies on alkaliactivated slag systems are scarce.

Very little detailed work on Soviet slag concretes has been published in English, other than two sets of conference proceedings from 1994 and 1999,8,9 and a recent book reviewing alkali-activated cements and concretes.7 In particular, nuclear magnetic resonance (NMR) and transmission electron microscopy (TEM) studies of these materials have not previously been available, and are presented herein for the first time. By application of these techniques in combination with high-resolution scanning electron microscopy (SEM), allowing elemental mapping of these samples on a level that has not previously been possible, a greater understanding of the mechanisms underlying the durability of these concretes will be developed.

In this study, slag concretes activated by carbonates or by carbonate/hydroxide mixtures in Ukraine between 1964 and 1982, and subjected to service conditions since that time, are analyzed. The materials have retained their original form, quality, and serviceability after being exposed to their service environment-winters below -20 °C (-4 °F), summers above 20 °C (68 °F), with repeated freezing-andthawing cycling-for more than two decades.8 Portlandbased cements placed under identical conditions at the same sites were reported to have rapidly degraded and required replacement after only 2 to 3 years of service.10 This paper aims to characterize these slag concrete samples with an emphasis on the elemental distribution and microstructure of the C-S-H phases, and to link these microstructures to the durability of the samples.

Contradictory information exists in the literature regarding the durability of alkali-activated concretes. Better durability has been noted,2-6 but lower durabilities with poor resistance to freezing-and-thawing,1,11 carbonation, and alkali-aggregate reaction12,13 have also been reported. This poor performance is partly attributed to larger, more open pores and higher porosity,11,12 which conflicts with observations of refined capillary pore size and reduced pore size distribution in slag concretes. Some freezing-and-thawing issues have also been attributed to difficulties with the use of standard air-entraining admixtures in alkali-activated systems rather than being inherent to the slag binder chemistry.1 It is noted that slag concrete performance depends strongly on activator choice; for example, slag concrete activated by sodium silicate has better freezing-and-thawing resistance than that activated by sodium carbonate,7,11 but the converse is true for alkali-aggregate reaction resistance.7 Different activators are expected to activate slag at different rates, giving C-S-H phases with different elemental distribution and microstructure. Moreover, macroscale mechanical and chemical characteristics including chemical resistance, compressive strength, and freezing-and-thawing performance are the net effect of chemical reaction and physical movement on the microscale. The different elemental distributions and microstructures of C-S-H phases may therefore be responsible for the varied durability of slag concretes activated by different activators.

In the 1960s to 1980s, slags used in cement manufacture in the former Soviet Union were collected from metallurgical plants and ground and activated to form slag concrete products without being prerefined or purified. Moreover, the activators used for making slag concretes were industrial by-products collected from production of alumina, nitrogen, and other products. The high impurity levels, both in slag sources and in activators, led to considerable heterogeneity in slag binders, and the wide range of aggregates used further increased this complexity. In spite of these difficulties, significant value may be obtained from the study of these specimens in developing a link between the micro- and nanostructure of the aged slag concretes, the mechanisms involved in slag activation, and long-term durability. In particular, the tendency of these concretes to continue to increase in strength over long periods of time has been noted previously8 and therefore needs to be related to their microstructural development.

RESEARCH SIGNIFICANCE

This is the first NMR and TEM investigation of durable slag concretes from the former Soviet Union. These concretes have served under aggressive conditions for up to 30 years with no sign of degradation, are chemically very different from portland-cement concretes, and their microstructure was previously unknown. Analysis of the mechanisms underlying the observed durability of these samples-significantly exceeding that of portland-cement concretes exposed to the same working conditions-will be important in the future development of durable slag concretes.

EXPERIMENTAL METHODS

The concretes analyzed in this study were manufactured in Kiev, Ukraine, between 1964 and 1982. Samples were taken from each concrete product in 2000 to 2001. Table 1 lists slag compositions and design parameters and Table 2 lists compressive strengths (from 50 mm [1.97 in.] cubes) and the service conditions of the concretes. The specimens analyzed were obtained as follows: Samples 1 and 4 were segments of larger chunks broken from slabs by a jackhammer and Samples 2 and 5 were obtained as 70 mm (2.8 in) diameter cores. These samples were sawn to size for strength testing. Sample 3 was obtained directly by sawing a wall-block into cubes. All samples were cured in an oven at elevated temperatures (40 to 140 °C [104 to 284 °F]) for 6 to 9 hours before use in service and sealed to prevent water loss.

X-ray diffractometry

X-ray powder diffraction was recorded using CuKa radiation (40 kV, 30 mA) with a scanning rate of 2 degrees per minute from 5 to 70 degrees 2?. Concrete samples were crushed and large aggregate pieces removed before analysis.

Nuclear magnetic resonance

^sup 29^Si magic angle spinning (MAS) NMR spectra were collected at 79.49 MHz, at a spinning rate of 12 kHz. Single pulse NMR was performed using 5 µs pulses with 2-second relaxation delays. Five thousand to 7000 free induction decays were accumulated for each sample. Chemical shifts were reported relative to tetramethylsilane (TMS). ^sup 27^Al MAS-NMR spectra were recorded at 104.26 MHz on the same spectrometer, using 5 µs pulses with 0.2-second delays. Chemical shifts were reported relative to Al(H2O)^sub 6^^sup 3+^. Concrete samples were crushed and sifted to remove pieces of aggregate before analysis.

Electron microscopy

SEM specimens were prepared by impregnating sample slices with epoxy and polishing with diamond lapping film. After sputter-coating with carbon, the specimens were examined in backscattered secondary electron (BSE) mode, with an energy-dispersive X-ray (EDX) spectrometer at an accelerating voltage of 20 kV. Care was taken throughout EDX analysis to avoid the inclusion of calcite, ettringite (AFt,) and monosulfate (AFm) into calcium silicate hydrate analyses.

TEM specimens were prepared by polishing samples to a thickness of 30 to 50 µm then argon ion milling on a liquid nitrogen cooled stage. Samples were carbon coated and analyzed using a Si(Li) X-ray detector for TEM image examination and EDX elemental analysis. Wollastonite, calcium carbonate (AR grade), and kaolinite (CMS KGa-1) were used to obtain Cliff-Lorimer factors K^sub CaSi^, K^sub CaO^, and K^AlSi^, respectively.

EXPERIMENTAL RESULTS AND DISCUSSION

Although these slag concretes were cast as long ago as the mid-1960s, they showed no significant internal or external deterioration. Instead, compressive strengths have increased significantly since casting (Table 2). Table 2 also shows that the carbonation rate of all samples (where recorded) was reasonable, contrary to previous reports of carbonation problems in alkali-activated slag concretes.12 Sample 3 possesses the lowest net strength, but the greatest increase in compressive strength. The aggregate in Sample 3 was bulk slag particles (Table 1). This gives a lower overall strength due to the relatively poor performance of the slag as an aggregate, but the continuing hydration reactions involving all particles in the system provides a high strength increase compared with samples containing less-reactive aggregates. This is also shown in the lower surface pH value of this sample where the continuing reaction of the slag particles has consumed the vast majority of the alkaline components initially added.

It is also notable that the performance of concretes similar to Samples 1 and 4 has previously been tested, following 12 years of exposure to service conditions.10 The concrete from the same site as Sample 1 decreased in permeability by a factor of approximately 500 over 12 years of service, and was not adversely affected by 600 freezing-and-thawing cycles. The concrete from the same site as Sample 4 showed no signs of degradation after 12 years of operation, where, as was noted previously, portland-based cements operating under the same conditions required replacement after 2 to 3 years of exposure. Clinker-free Na^sub 2^CO^3^-activated slag concretes have previously been observed to increase in strength over the entire period of 2 years of wet curing,10 consistent with these observations.

XRD analysis

All samples, as seen in Fig. 1, had quartz and calcite crystalline phases attributed to the quartz aggregates and carbonate activators used. No apparent crystalline C-S-H phase was detected, indicating that all C-S-H phases formed are amorphous. Sample 3 contained no quartz and only a very small amount of calcite, consistent with the absence of a quartz aggregate and a very high slag content. Samples 4 and 5 also contain anorthite from the granite aggregate.

^sup 27^Al and ^sup 29^Si MAS-NMR analysis

Figure 2 shows ^sup 27^Al MAS-NMR spectra of all samples. The majority of the Al in slag possesses a highly disordered tetrahedral coordination that, during hydration, separates to an increasing octahedral and a sharper tetrahedral chemical shift.14 The increasing tetrahedral Al is attributed to aluminumsubstituted calcium silicate hydrates (C-A-S-H).15-17 Figure 2 shows that all aged samples have tetrahedral (~50 ppm) and octahedral (~0 ppm) Al, with the former being dominant. Sample 4 shows the strongest octahedral peak, attributed to the formation of monosulfate (AFm) due to the relatively high levels of both Al and S in this sample. The remainder of the octahedral Al in the hydrated phases of all samples may be present as substituents in the interlayer spaces of C-S-H or in calcium aluminate hydrates.18 Sample 3, which contains very little S and very high slag content, shows almost no octahedral Al. Samples 1, 2, and 5 show very similar ^sup 27^Al spectra, consistent with the use of similar activating solutions. Of these three samples, Sample 5, activated with a mixture of NaOH and Na^sub 2^CO^3^, was synthesized under the most alkaline conditions, and also shows the highest proportion of tetrahedral Al present, concomitant with the tendency of Al to adopt tetrahedral coordination under high alkalinity.19

Figure 3 presents ^sup 29^Si MAS-NMR spectra of all samples. The sharp peak at -108 ppm in Samples 1, 2, 4, and 5 is due to quartz in the aggregates, and is therefore absent in Sample 3. Samples 1, 2, and 4 show strong peaks at approximately -85 ppm corresponding to Q^sup 2^ (chain) silicate structures in the C-S-H phase,20 superimposed on the broad slag peak. The chemical shift range of this peak is also consistent with some tetrahedral Al substitution into chains in place of Si (that is, Q^sup 2^(1Al) sites).21 The slag peak dominates the spectrum of Sample 3. Sample 5 shows a high degree of silicate polymerization, with peaks in the region (-90 to -95 ppm) corresponding to Q^sup 3^ (branching) silicate sites with a low degree of Al substitution.21 This may again be attributed to the highly alkaline conditions under which this sample was activated, as well as the low Al^sub 2^O^sub 3^ content of the slag used in this sample (Table 1). It is also notable that the surface pH of Sample 5 has decreased to below that of Samples 2 or 4, due most likely to alkali consumption during silicate polymerization.

The primary differences between Samples 4 and 5 are the presence of some Ca(OH)^sub 2^ in the activator used for Sample 4, and the lower Al^sub 2^O^sub 3^ content of the slag in Sample 5. Both of these factors will contribute to a higher extent of silicate polymerization in Sample 5. Dissolved Ca2+ in the activating solution means that less slag dissolution was required before the formation of the first C-S-H phases in Sample 4. The extent of silicate polymerization in solution depends strongly on concentration19 and so the higher concentration of dissolved silicate released by slag dissolution before C-S-H formation in Sample 5 will mean that the silicate will be more polymerized when it combines with Ca2+ during gelation.

SEM analysis

Figure 4 shows BSE SEM images of all samples. All samples comprise large unreacted slag particles (S), aggregate (Q), and hydrated C-S-H phases (H). The hydrated rims of the large slag particles are ~1 to 2 µm in width as indicated by white arrows, corresponding to the inner product (Ip), whereas the remainder of the hydrated phase comprises the outer product (Op).22 Quantitative EDX analyses were conducted in the unreacted centers of slag particles, the Ip, and the Op. Twenty-five to 50 points were analyzed for each region of each sample. Ca, Si, Al, O, Na, K, Mg, Mn, Fe, S, and Ti were quantified, with Ca, Si, and Al being reported in Fig. 5 through 9 and Table 3.

The atomic percentages of both Si and Ca generally decrease from the unhydrated slag large particles to the hydrated rims (Ip) and further to the fully hydrated phases (Op), as the hydrated products contain more water per Ca or Si atom. The only significant exception to this trend is Sample 4, where the presence of Ca in the activating solution complicates the analysis. It is not clear, however, why the Op in this sample contains both more Ca and more Si than the slag particles (Fig. 8(a)).

Figures 5(b) through 9(b) show that the distribution of Al between the slag and Ip phases of all samples is largely uniform. There is slightly more variation in the Op Al distributions, with Samples 3 through 5 (Fig. 7(b) through 9(b)) showing slight Al depletion. This is most notable in Sample 4 (Fig. 8(b)), which suggests that the very high extent of hydration observed in this sample due to the slag aggregate means that any chemical segregation effects are amplified. Similarly, the lack of Al depletion in the Op of Samples 1 and 2 may reflect the less alkaline activator in these samples, with a corresponding decrease in the opportunity for chemical segregation. In Sample 3, the formation of the aluminosilicate phase observable in Fig. 4 leaves the surrounding gel Aldepleted. This phase does not produce observable XRD peaks, and it has been observed that crystalline zeolitic phases are unlikely to form in activation of slag,23 which is not unexpected given the tendency of even moderate levels of calcium to interfere with zeolite formation.24 This is therefore tentatively identified as a semicrystalline or precrystalline dense calcium aluminosilicate phase. It is also noted, however, that in the entire process of SEM analysis, this was the only observable distinct aluminosilicate region, which may mean that it is in fact crystalline but present in quantities too small for observation by XRD.

The greater slag heterogeneity in Sample 4 gives isolated (Ca,Fe)-concentrated regions (black arrows in Fig. 4) and C-S-H phases coexisting. Sample 5 demonstrates comparatively thorough hydration with thick Ip rims along large slag particles, and a less heterogeneous C-S-H phase.

Figures 7 and 9 show that, during hydration of Samples 3 and 5 (both activated by Na^sub 2^CO^3^/NaOH), Si increases and Ca decreases from the original slag particles to the Ip, whereas the Op possesses Si and Ca levels generally lower than the original slag. This is different from Samples 1 and 2 (activated by Na^sub 2^CO^3^/K^sub 2^CO^sub 3^), where both Si and Ca decrease in the Ip region. The increased alkalinity of the NaOH-containing activator will enhance Si dissolution while still allowing calcite and C-S-H precipitation. This additional Si, however, may not be particularly mobile due to the formation of the Op, so it will be present primarily in the Ip regions. The absence of similar phenomena in the Al distributions of these samples may be due to the much greater lability19 (and therefore potentially greater mobility in a constrained system over a very extended period of time) of Al compared with Si.

Sample 4 (Fig. 8(a)) had increased Ca and Si levels in both the Ip and Op regions. The increase in Si in the Ip of Sample 4 compared with the remnant slag, however, was higher than in Samples 3 or 5, which were activated by Na^sub 2^CO^3^/NaOH at a higher pH. This indicates that the role of dissolved Ca(OH)^sub 2^ in the activator could be twofold: 1) increasing pH; and 2) providing an extra Ca source to bind to the Si as it is released. This means that a Ca-containing activator will extract Si not only from slag particles but also from aggregates. A hydrated layer around a quartz particle is observed in Sample 4 (white arrow, bottom of Sample 4, Fig. 4). The Ca/Si ratio in the Ip of Sample 4 is significantly higher than that of Samples 3 or 5. In the Op, however, the average Ca/Si ratio (2.28) of Sample 4 is comparable with that of Samples 3 and 5 (2.05 to 2.14). This suggests that the Ca/Si ratio of these Op regions may be controlled by chemical equilibria within the C-S-H phase rather than by the initial kinetics of activation, and is observed despite the clearly different extents of hydration in these samples.

TEM analysis

Figure 10 is a TEM image of the Ip of Sample 2, showing a primarily amorphous C-S-H gel matrix. Very small calcite crystals are observed under TEM in both Op and Ip in Samples 1 and 2, with the calcite in Op regions (not shown) being both larger in crystallite size and greater in quantity. Calcite is not observed in Samples 3 through 5, so it is suggested that the higher alkalinity and lower carbonate levels of these samples favor the formation of gel products over calcite crystallization.

The Op morphologies vary significantly both between and within samples. Samples 1 and 2 have primarily foil-like Op as has been previously observed for alkali-activated slags,25 with polycrystalline particles embedded in an amorphous region. Figure 11(a) shows the structure of Sample 1, and Sample 2 (not shown) appears similar. The elemental distributions of the Op areas of these two samples fluctuate significantly (Table 3), suggesting that the Op areas are nano-heterogeneous. During the TEM/EDX analysis, special care was taken to avoid inclusion of calcite, AFt, and AFm in the statistics of the C-S-H phases. The TEM/EDX results are consistent with the SEM/EDX data, with the greater variability in TEM/EDX Ca/Si ratios attributable to a smaller sampling region and therefore greater sensitivity to nanoscale heterogeneity.

Figure 11(b) depicts a TEM image of the Op region of Sample 3, and Sample 5 also shows a relatively similar microstructure. These samples display a smooth, amorphous Ip region (not shown) containing embedded nanocrystals high in Ca, as well as Fe and Ti from impurities in the slag. The Op is textured and is mainly comprised of an amorphous C-S-H gel (Fig. 11(b) inset). Sample 5 generally shows less Op-like regions and more Ip-like regions, suggesting that ongoing hydration in this sample has led to the collapse of the more porous Op morphology into a denser Ip-like structure.

Sample 4 (Fig. 11(c)) has quite a different Op morphology, showing an amorphous fiber matrix with embedded nanometer-sized crystals. When analyzed under higher magnification (not shown), the fibers are themselves an assembly of many amorphous nanoparticles, approximately 5 nm in size. The diffraction pattern (Fig. 11(c) inset) may therefore represent nanocrystals formed outside the fibers of the Op, but embedded in the fibrous matrix. Small particles of AFt and/or AFm are also found forming in the Ip area of Sample 4, as observed from ^sup 27^Al MAS-NMR.

Microstructure, nanostructure, and durability

It has been reported that the initial products of a slag paste activated by carbonates are alkali metal-calcium carbonate and/or calcium carbonate.26 Alkaline attack on the Ca-O-Si bonds in slag particles then frees Si to bond with Ca and/or alkali metals to form hydrates. In addition to this mechanism, however, M^sub 2^CO^sub 3^ (M = Na,K) plays a second significant role in slag hydration-maintaining a buffered alkaline environment for ongoing hydration.

M^sub 2^CO^sub 3^ + slag [arrow right] CaCO^sub 3^ + 2M^sup +^ + dissolved SiO^sub 2^ [arrow right] C-S-H + M^sub 2^CO^sub 3^

This equation assumes that the slag and C-S-H have similar Ca/Si ratios, which was the case in these samples. The carbonate may be present as CO^sub 3^^sup 2-^ and/or HCO^sub 3^^sup -^ depending on the overall system pH. Both anions, however, are able to participate in continuing hydration, and so the discussion herein will specifically address CO^sub 3^^sup 2-^ with the understanding that HCO^sub 3^^sup -^ participates via similar mechanisms.

The fact that hydration of carbonate-activated slags continues for an extended period after initial setting is seen clearly from the strength data. If supplying an alkaline environment was the only role of the carbonate solution in the hydration process, the lower alkalinity of carbonate compared with hydroxide solutions would give slower slag dissolution, a lower degree of hydration, and poorer mechanical properties. In fact, slag activated by carbonates exhibits superior flexural and compressive strengths over that activated by MOH.25 Additionally, many calcite nanocrystals are found embedded in the Ip regions of Samples 1 and 2 (Fig. 10), which differs from the hydration of C^sub 3^S by water where calcite was found predominantly in the Op and, to a much lesser extent, in the Ip.22 This suggests that calcite may be somehow migrating from the space originally occupied by the activator (that is, the Op region) toward the Ip region as hydration reactions continue for very extended periods of time. Long-term hydration reactions are likely to occur close to the slag particles because the reduced mobility of species through the hardened paste will greatly hinder transport from the slag to the Op, so it is within the Ip that the continued reaction will be most significant.

Initially, Ca and Si dissolve from the slag and form solid CaCO^sub 3^ and dissolved Si species in the Op region. This decreases the system pH. The Si concentration increases to saturation and C-S-H forms. This takes up Ca from CaCO^sub 3^, releasing CO^sub 3^^sup 2-^ back into solution, allowing it to participate in another cycle of hydration. Through control of the pH and the exchange of cations between Na and/or K and Ca, the carbonate solution activates slag gradually and progressively. The relatively low Ca/Si ratio of the slags studied herein compared with OPC clinker means that the present silica will be quite highly polymerized before C-S-H formation, which will also give characteristic micro- and nanostructures.

It is also interesting to note that, despite the very different Ca/Si ratios observed in the C-S-H gels in Samples 1 and 2, there is almost no difference between the Al(IV)/Al(VI) ratios of these samples. Stade and Müller17 found that, for Al-substituted C-S-H gels, the fraction of octahedral Al decreases with increasing Ca/Si ratio. This indicates that the level of Al incorporation into the C-S-H gel phase may be relatively low, meaning that the octahedral Al is likely to be present in other phases such as AFm, AFt, or calcium aluminate hydrates, as was noted previously.

As the Ca/Si ratio in the Op is controlled by diffusion through the cement matrix, the formation of nanocrystalline and short-range ordered regions is the result of localized reactions. At room temperature, short-range diffusion occurs readily over a distance of several nanometers, while long range diffusion is much slower. In the absence of a strong chemical driving force-for instance, a concentration gradient due to a higher dissolution rate in a locally higher pH environment-the coalescence of nanocrystalline regions into a crystalline phase is very slow. This may explain why no crystalline C-S-H needles or fibers are found in any of the Op regions studied herein.

The primary outcome of this study must be an explanation, based on microstructural and nanostructural grounds, for the significantly greater durability of this set of slag cements when compared with portland cements exposed to the same aggressive environment.10 The cyclic hydration process proposed is likely to play a significant role in this, with the regeneration of the carbonate activator providing the opportunity for hydration to continue over an extended period. There is generally a high extent of silicate polymerization due to the low Ca/Si ratio of the samples studied herein (Fig. 3). This high level of cross linking will mean that removal of Si from the C-S-H phase by chemical attack is relatively difficult. Al substitution for Si in a tetrahedral network provides points of potential weakness due to its lability. This may mean that the tendency of the Al to remain close to the slag particles in the Na^sub 2^CO^3^/NaOH-activated samples, leaving Op phases somewhat Al-poor compared with the slag particles (Fig. 7(b) through 9(b)), will improve the durability of the Op in these systems. The similar Ca/Si ratios of the Ip and Op in most specimens will reduce the driving force for diffusional mass transport within the concrete, reducing the likelihood of nanostructural damage by migration of these species via depolymerization reactions. The Op also shows a relatively disconnected pore structure, which will also contribute to durability by preventing the ingress of aggressive agents into the concrete. Strong binding to the aggregates also reduces the permeability of the concretes compared with systems activated under less-alkaline conditions. It will be the combination of all these factors that leads to the very good durability displayed by these samples.

CONCLUSIONS

Durable, aged slag concretes from the former Soviet Union, activated by carbonates and by carbonate/hydroxide mixtures, have been investigated. All samples demonstrate high compressive strengths-significantly higher than when initially cast-and excellent durability over a service life of up to 35 years in aggressive conditions. The C-S-H phases of all samples are X-ray amorphous, and heterogeneous at a nanometer scale. The carbonate activator activates the slag by supplying a buffered, moderately alkaline environment, maintaining a cyclic hydration process whereby the dissolved CO^sub 3^^sup 2-^ level is maintained by its continual release from calcite as the Ca reacts with dissolved silicate to form C-S-H. The C-S-H phases formed show a high polymerization degree, a similar Ca/Si ratio to the slag source, and a disconnected pore structure. A higher polymerization degree in a silicate network with little Al substitution provides more resistance to chemical attack. The similar Ca/Si ratios between the Ip and Op regions reduce chemical potential gradients within the C-S-H, which reduces internal deterioration. A disconnected pore structure inhibits the intrusion of chemicals and the calcite content protects the concretes against further carbonation. The durability of slag concretes is attributed to a combination of these factors, each present to varying extents in each sample depending on composition, synthesis, and service conditions.

ACKNOWLEDGMENTS

The authors thank K. Nairn (CSIRO) and D. McCulloch (RMIT) for assistance and discussions regarding NMR and TEM analysis. The financial support of the Australian Research Council is appreciated.