Fly Ash Utilisation in Mullite Fabrication: Development of Novel Percolated Mullite

: Fly ash is an aluminosilicate and the major by-product from coal combustion in power stations; its increasing volumes are major economic and environmental concerns, particularly since it is one of the largest mineral resources based on current estimates. Mullite (3Al 2 O 3 · 2SiO 2 ) is the only stable phase in the Al 2 O 3 -SiO 2 system and is used in numerous applications owing to its high-temperature chemical and mechanical stabilities. Hence, ﬂy ash offers a potential economical resource for mullite fabrication, which is conﬁrmed by a review of the current literature. This review details the methodologies to utilise ﬂy ash with different additives to fabricate what are described as porous interconnected mullite skeletons or dense mullite bodies of approximately stoichiometric compositions. However, studies of pure ﬂy ash examined only high-Al 2 O 3 forms and none of these works reported long-term, high-temperature, ﬁring shrinkage data for these mullite bodies. In the present work, high-SiO 2 ﬂy ashes were used to fabricate percolated mullite, which is demonstrated by the absence of ﬁring shrinkage upon long-term high-temperature soaking. The major glass component of the ﬂy ash provides viscosities suitably high for shape retention but low enough for ionic diffusion and the minor mullite component provides the nucleating agent to grow mullite needles into a direct-bonded, single-crystal, continuous, needle network that prevents high-temperature deformation and isolates the residual glass in the triple points. These attributes confer outstanding long-term dimensional stability at temperatures exceeding 1500 ◦ C, which is unprecedented for mullite-based compositions.


Introduction
Owing to rapid industrialisation, there has been an increase in power generation, with numerous coal-powered electricity stations being used currently at high to maximal capacities. This has resulted in increased amounts of by-products being produced from the coal combustion process in these power stations; these products include fly ash, cenospheres, and bottom ash [1]. Of these, there has been increasing focus on the utilisation of fly ash since it is the major component, which is collected in electrostatic precipitators [2]. Recent estimates indicate that the annual global production of fly ash is~800 mT [3]. Unused fly ash generally is stored on-site at power stations in containers or in a slurry form in collection ponds [4]. The former strategy creates problems in terms of space requirements for storage while the latter can be subject to leaching of toxic elements from the fly ash [4]. Dumping of fly ash into water resources is known to occur in some countries, leading to major environmental issues [5]. These volumes illustrate the need for novel strategies for the utilisation of fly ash since only 20-50% is used commercially [3], with the major application being for the cement industry as pozzolan/aggregate [6]. However, higher utilisation in some European countries is the case as they produce very low volumes, although the utilisation levels by major producers in Asia and North America are low [3].
Fly ash consists of whitish spherical particles of size in the range 1-200 µm, with average sizes being 20-40 µm [4]. Fly ash may contain remnants of unburnt carbon from the coal, which makes it appear grey to black in colour. Moreover, it may contain varying amounts of iron oxide, which can generate colours ranging from light to dark brown [4]. The proportions of the different oxides in fly ash vary with the source of the coal and the combustion process. On the basis of composition, the major oxide components are Al 2 O 3 and SiO 2 , which result from the fact that clay and sand are the principal mineral phases of coal. While SiO 2 is the major component of the glassy spherical particles, undissolved α-quartz (SiO 2 ) generally is the principal crystalline phase. Another significant crystalline species that form during coal combustion is mullite (3Al 2 O 3 ·2SiO 2 ) while lower levels of magnetite (Fe 3 O 4 ) and/or hematite (Fe 2 O 3 ) also are observed commonly [7]. While Class F fly ashes are low in CaO, Class C fly ashes have higher CaO contents [8], which have been observed by the authors to be >10 wt% CaO. Mullite is generally seen as a reaction product from reactions of Al 2 O 3 -and SiO 2 -containing materials [9].
The refractoriness of mullite in principle is rated at 1850°-1865 °C (PCE = 38-39) [22] but the actual performance is somewhat lower owing to glass softening and resultant firing shrinkage. In practice, pure single-phase mullite exhibits short-term creep resistance at temperatures ≤1400 °C and commercial mullites perform similarly only to ≤1000 °C [13]. This degradation in practice highlights the importance of the achievement of direct bonding between the mullite grains in contrast to chemically (glass) bonded refractories, the thermal properties of which are dominated by the glassy matrix [23].
While the thermal resistance is a critical factor in refractory performance, chemical resistance often is an essential quality in metallurgical applications. Mullite-containing refractories are used commonly as linings for molten aluminium-alloy holding furnaces [24]. In these refractories, the molten metal wets and corrodes the refractory, which leads to the reduction of the siliceous component of mullite and the resultant formation of Al2O3 as a by-product. However, the corrosion resistance can be improved by the addition of non-wetting additives that react with mullite to form complex oxides that increase the alloy wetting resistance of the refractory surface [25][26][27][28][29]. However, the applicability of fly ash as a source of mullite-based refractory products is challenged by its high glass content, the softening of which is well known as the principal source of high-temperature deformation of many oxide ceramics [13]. As shown in Figure 1b, the heating of highly vitreous bodies such as fly ash, which generally contaiñ 60-70 wt% glass [14], would be expected to result in sigmoidal firing shrinkage kinetics. That is, initial slow glass softening at high glass viscosities would enhance particle rearrangement and firing shrinkage, followed by slumping owing to rapid deformation at low glass viscosities.
The refractoriness of mullite in principle is rated at 1850 • -1865 • C (PCE = 38-39) [22] but the actual performance is somewhat lower owing to glass softening and resultant firing shrinkage. In practice, pure single-phase mullite exhibits short-term creep resistance at temperatures ≤1400 • C and commercial mullites perform similarly only to ≤1000 • C [13]. This degradation in practice highlights the importance of the achievement of direct bonding between the mullite grains in contrast to chemically (glass) bonded refractories, the thermal properties of which are dominated by the glassy matrix [23].
While the thermal resistance is a critical factor in refractory performance, chemical resistance often is an essential quality in metallurgical applications. Mullite-containing refractories are used commonly as linings for molten aluminium-alloy holding furnaces [24]. In these refractories, the molten metal wets and corrodes the refractory, which leads to the reduction of the siliceous component of mullite and the resultant formation of Al 2 O 3 as a by-product. However, the corrosion resistance can be improved by the addition of non-wetting additives that react with mullite to form complex oxides that increase the alloy wetting resistance of the refractory surface [25][26][27][28][29].
More broadly, the corrosion resistance of mullite to some transition metals, such as Ti, V, Fe, Zn, and Zr is poor [12]. However, its corrosion resistance to Ni is good, which suggests the possibility of that mullite may be suitable for containment of Ni-based superalloys. There do not appear to be any studies of the corrosion resistance to Co-based superalloys. Table 1 summarises the published literature on the production of mullite from fly ash. These data reveal that the foci of much of this work were the modification of the Al 2 O 3 :SiO 2 ratio to shift it nearer to the stoichiometric 3:2 mullite ratio (~71.8 wt% Al 2 O 3 , 28.2 wt% SiO 2 ) and to produce porous mullite products. The production of porous bodies generally involves the use of AlF 3 , which tends to produce gaseous SiF 4 upon reaction with the SiO 2 -rich glass present in these compositions, which removes the glass and creates porosity. More critically, heating was done for a maximal time of only 4 h in all of these studies (except for one where they formed whiskers at a relatively lower temperature) and none reported the long-term firing shrinkages. The combination of these factors represents the principal difference between these works and the present work, which reports the development of unique, dense, mullite microstructures revealed by firing shrinkage measurements and that are generated using high-SiO 2 compositions and longterm heating. This novel composition and microstructure, which consists of percolated mullite, exhibits unparalleled high-temperature stability owing to the achievement of a continuous, direct-bonded, single-crystal network of mullite needles [30]. • Alkali-activation treatment adjusted the chemical composition to that of stoichiometric mullite. • 100% fly-ash-based ceramic tiles sintered at 1300 • C exhibited optimal postsintered densification. [45]

Percolated Mullite
Highly dense 3:2 mullite (<2% apparent porosity) with microstructures completely percolated throughout the entire sample volumes were fabricated using either fly ash alone or fly ash mixed with an oxide addition at a low concentration [30,56]. These intergrown microstructures developed through mullite nucleation (small equiaxed precipitated grains of primary mullite) and growth (elongated grains of secondary mullite from primary mullite and pre-existing mullite needles). Although percolation can be achieved using fly ash alone, the kinetics of percolation can be enhanced through the addition of different additive types and amounts. A key but counter-intuitive feature of the generation of percolated mullite is the presence of excess SiO 2 -rich glass. This glass acts as both a medium for the diffusion of ions for mullite growth at high temperatures but also provides a deformable medium that is compliant with the axial and diametral growth of the mullite needles. Once these needles have grown to sufficient lengths, they form direct-bonded intergrowths that completely exclude any glass between the grains [57], thereby forming a non-shrinking single-crystal network that exhibits the intrinsic properties of pure mullite on a full volumetric basis [56,57]. Critically, this percolated microstructure isolates any residual glass within the triple points, effectively forming a 0-3 composite [58,59] whose thermomechanical properties, including high-temperature strength and creep [56], are unaffected by the glass. The thermal expansion would be expected to be increased by the glass but this effect would be mitigated by resistance from the rigid, continuous, mullite skeleton.
The percolated mullite was developed and patented over the period 2012-2016 from a range of Class F fly ashes from Australia and China; typical compositions for these are shown in Table 2. These fly ashes contain glass as the major component, with minor crystalline phases of mullite and α-quartz plus trace amounts of magnetite and α-hematite.  The pure fly ash compositions were uniaxially pressed into cylindrical pellets (15 mm diameter × 5 mm height), heated to 1500 • C (heating rate 120 • C/h), and soaked for 1-96 h. Figure 2 shows the changes in mullite needle dimensions (lengths and width) with increasing heat treatment time. All of the microstructures were etched with HF and then characterised using scanning electron microscopy (SEM). The samples were not carbon coated since this could have impacted on the imaging resolution between the individual needles. However, the result of this was some charging of the samples, as indicated by the lighter coloured regions in the images. It can be seen that the well-formed but discrete mullite needles formed at only 1 h but these developed into a continuous interconnected network by 4 h. At the longer time points, the main difference was needle coarsening, increasing aspect ratio, and some exaggerated grain growth.
Importantly, the diametral firing shrinkages were determined at each sintering time. Figure 3 shows the diametral firing shrinkage trends of the pure fly ashes (1, 2, and 3) and fly ash 3 with α-Al 2 O 3 added in order to shift the overall composition toward that of mullite (~46 wt% α-Al 2 O 3 ). These data clarify that that percolation, indicated by the cessation of further volumetric firing shrinkage, can be achieved at a soak time of only 2 h at 1500 • C. Importantly, the diametral firing shrinkages were determined at each sintering time. Figure 3 shows the diametral firing shrinkage trends of the pure fly ashes (1, 2, and 3) and fly ash 3 with α-Al2O3 added in order to shift the overall composition toward that of mullite (~46 wt% α-Al2O3). These data clarify that that percolation, indicated by the cessation of further volumetric firing shrinkage, can be achieved at a soak time of only 2 h at 1500 °C.  Importantly, the diametral firing shrinkages were determined at each sintering time. Figure 3 shows the diametral firing shrinkage trends of the pure fly ashes (1, 2, and 3) and fly ash 3 with α-Al2O3 added in order to shift the overall composition toward that of mullite (~46 wt% α-Al2O3). These data clarify that that percolation, indicated by the cessation of further volumetric firing shrinkage, can be achieved at a soak time of only 2 h at 1500 °C. It is somewhat surprising that, despite the differences between the fly ashes in terms of amount and type of glass and the amount of mullite, percolation could be achieved consistently at soak times of~2-4 h. The major differences were manifested in the extents of firing shrinkage for pure fly ashes and in the kinetics of firing shrinkage for the fly ashes with α-Al 2 O 3 added. The principal driving force for these differences is shown in Table 2, which reveals significantly different (1) Al 2 O 3 /SiO 2 wt% ratios, (2) amounts of fluxing impurities (e.g., Na 2 O, K 2 O, CaO, and Fe 2 O 3 ), (3) resultant glass viscosity, and (4) amounts of unburnt carbon (shown by the loss on ignition, LOI). The first of these variables would affect both the glass viscosity but also the ultimate mullite content. The second of these would affect the glass viscosity but also the tendency to favour the formation of equiaxed or needle-like mullite [12]. The last of these would affect the firing shrinkage upon oxidative gas emission.
When comparing the firing shrinkage behaviour of samples fabricated using only fly ashes with those fabricated using mixtures of fly ash and calcined alumina, it is clear that the addition of α-Al 2 O 3 has an insignificant effect at low additions (<20 wt%) and a retarding effect at high additions (>20 wt%) on the achievement of percolation. This clearly reflects the role of Al 2 O 3 in increasing the glass viscosity, as suggested in Figure 1a. When α-Al 2 O 3 is added, the diametral firing shrinkage shows a fluctuation (viz., expansion) at 4 h, which was observed in fly ash 1. This expansion is interpreted in terms of the sequence prior to percolation described in Table 3. Table 3. Interpretation of volumetric shrinkage prior to percolation exhibited in Figure 3. It also is possible that the expansion derived from gas exudation upon the achievement of sufficient reduction in glass viscosity and resultant microstructural disruption but these are not consistent with the LOI levels shown in Table 2.
The preceding considerations clarify the reasons for the failure of previous studies to report volumetric percolation. That is, in addition to the use of soak times generally insufficient to achieve percolation and the non-measurement of firing shrinkages, the glasses at issue were of high viscosities such that the necessary diffusion and compliance for mullite percolation were not established because (1) the studies of pure fly ash involved high-Al 2 O 3 types and (2) the studies of fly ash with Al 2 O 3 added were designed to equate to the mullite composition.
Image analysis of the SEM images was conducted to quantify the microstructural changes occurring during heat treatment as a function of soak time. For each SEM image, 500-1000 grains (depending on time) of mullite were measured in order to provide statistically significant quantification of the needle lengths in terms of the values D10, D50, average, and D90, as shown in Figure 4a. The data in Figure 4b show that the needle lengths and areal coverage of the microstructure increased significantly with increasing sintering time up to 24 h, with the most of the significant alterations occurring within~4 h. Soaking for longer times reduced the number of mullite needles owing to grain growth, terminating in areal coverage of~96% and a maximal needle length (for fly ash 2) of~15 µm. The kinetics to achieve percolation also can be manipulated through the addition of oxides that modify the glass viscosity. Figure 5 summarises the impacts of three additives: MgO (1 wt%), CaO (0.5 wt%), and Fe2O3 (2 wt%). Comparison of Figures 3b and 5 shows the general enhancement of the rate of percolation from the oxide additions. These data also reveal that α-Fe2O3 was more effective than MgO or CaO, which reflects a combination of the solidus temperatures of the respective systems and the amounts of the additives: FeO-Al2O3-SiO2 1170 °C (0.5 wt% addition) [60] The data for the equivalent FeO from α-Fe2O3 are included because reduction of the latter to the former is a common occurrence during the heating in air of thick ceramics owing to oxygen starvation of the interior of the compact, as is evidenced frequently from the black coring of Fe-rich bricks [62]. The additions in the present work represent maximal addition levels as higher amounts resulted in excessive vitrification, loss of sample shape, and prevention of mullite percolation. The kinetics to achieve percolation also can be manipulated through the addition of oxides that modify the glass viscosity. Figure [30]).  show that percolation can be achieved in ~2-4 h for fly ashes both without and with oxide addition, resulting in a 0-3 [58,59] glass-mullite composite consisting of a dense, continuous, single-crystal, mullite skeleton with glass isolated in the triple points. These data also show that there is only insignificant microstructural alteration after heating at 1500 °C for ~24 h, so the grains undergo little additional grain growth. These data also show that the conventional logic of addition of Al2O3 to achieve a net mullite composition prevents the detection of percolation owing to high glass viscosity, insufficient glass content, and low ionic diffusion, all of which result in the formation of a porous polycrystalline conglomerate. Further, Figure 5 shows that oxide additions, such as α-Fe2O3, offer The data for the equivalent FeO from α-Fe 2 O 3 are included because reduction of the latter to the former is a common occurrence during the heating in air of thick ceramics owing to oxygen starvation of the interior of the compact, as is evidenced frequently from the black coring of Fe-rich bricks [62]. The additions in the present work represent maximal addition levels as higher amounts resulted in excessive vitrification, loss of sample shape, and prevention of mullite percolation. Figures 3-5 show that percolation can be achieved in~2-4 h for fly ashes both without and with oxide addition, resulting in a 0-3 [58,59] glass-mullite composite consisting of a dense, continuous, single-crystal, mullite skeleton with glass isolated in the triple points. These data also show that there is only insignificant microstructural alteration after heating at 1500 • C for~24 h, so the grains undergo little additional grain growth. These data also show that the conventional logic of addition of Al 2 O 3 to achieve a net mullite composition prevents the detection of percolation owing to high glass viscosity, insufficient glass content, and low ionic diffusion, all of which result in the formation of a porous polycrystalline conglomerate. Further, Figure 5 shows that oxide additions, such as α-Fe 2 O 3 , offer the potential to achieve percolation with higher Al 2 O 3 contents (40 wt%) approaching that of mullite (48 wt% Al 2 O 3 for fly ash 3) to achieve a net mullite composition), albeit with slower kinetics to achieve percolation but at a lower residual glass content.
The selection of the type and amount of additive to enhance the kinetics of percolation will be dependent on the nature of the fly ash (especially its Al 2 O 3 :SiO 2 ratio), the fluxing effectiveness of the oxide addition, the soak temperature, and the soak time. Examination of these variables has the potential to (1) increase the kinetics of percolation by reducing the temperature and/or time required for percolation or (2) reduce the kinetics of percolation while decreasing the amount of residual glass in the triple points.
As discussed previously, the major advantage of percolated mullite over conventional mullite products is the microstructure consisting of a direct-bonded, single-crystal, continuous, needle network that prevents high-temperature deformation and isolates the residual glass in the triple points. These characteristics allow percolated mullite to exhibit unprecedented thermal stability to at least at 1600 • C, as shown by the diametral shrinkage data in Figures 3-5 and the thermal expansion data in Figure 6. In principle, if slip systems are not activated to cause high-temperature deformation, the performance may extend as high as the melting point of mullite (1850 • C).  These data suggest that the thermal expansion is dominated by the [010] and [001] axial directions of the mullite needles. The expansion at 1000 °C, which is higher than the reported bulk expansion of α20-1000 = 6.74 × 10 −6 °C −1 , may derive from the additive effect of the residual glass in the triple points but it also may be due to differences in the mullite samples and/or the measurement techniques.
These microstructures also can be tailored to exhibit specific properties. At high porosities, the microstructures can be engineered to moderate the low thermal conductivities These data suggest that the thermal expansion is dominated by the [010] and [001] axial directions of the mullite needles. The expansion at 1000 • C, which is higher than the reported bulk expansion of α 20-1000 = 6.74 × 10 −6 • C −1 , may derive from the additive effect of the residual glass in the triple points but it also may be due to differences in the mullite samples and/or the measurement techniques.
These microstructures also can be tailored to exhibit specific properties. At high porosities, the microstructures can be engineered to moderate the low thermal conductivities and low compressive strengths. At low porosities, highly dense microstructures containing only closed spherical pores can be engineered [56] for low thermal conductivities but high compressive strengths [63]. Furthermore, manipulation of the microstructure can be done to create highly dense bodies with small closed spherical pores which help to also lower the thermal conductivity.
Consequently, percolated mullite has been used to fabricate dense and porous sintered refractory shapes, aggregates, and porous refractory shapes from sintered aggregates fabricated from fly ash and oxide additions [56]. These refractory shapes exhibited the physical properties described in Table 4. Table 4. Physical properties of refractory shapes fabricated from percolated mullite.

Summary
The present work considers previous work on the fabrication of mullite from fly ash and introduces a novel form of volumetrically percolated mullite with microstructures consisting of direct-bonded, single-crystal, continuous, needle networks, with the residual glass isolated in the triple points. These unique microstructures exhibit unprecedented longterm high-temperature dimensional stabilities upon long-term heating at 1500-1600 • C and possibly higher temperatures. Percolated mullite can be fabricated from both pure fly ash and fly ash with oxide addition, enabling the engineering of the kinetics of percolation to moderate the thermal conductivities and compressive strengths. The application of these novel materials as acidic and neutral refractories has both economic and environmental advantages as they have the potential to utilise high volumes of fly ash, which is a low cost industrial-by product.