The Application of Advanced Mineralogical Techniques to Coal Combustion Product

Characterisation David French 1 and Colin R. Ward 1,2

1: CSIRO Energy Technology, Menai, NSW, Australia 2: University of New South Wales, Sydney, NSW, Australia KEYWORDS Coal combustion by-products, mineralogy, QEMSCAN, quantitative X-ray diffraction ABSTRACT Quantitative X-ray diffractometry and automated electron beam image analysis have been successfully used to evaluate the nature, mode of occurrence, distribution and relative abundance of the different crystalline phases (minerals) and amorphous (glassy) phase occurring in a range of coal combustion by-products. The automated electron beam image analysis was performed using a QEMSCAN system, an SEM equipped with four integrated energy dispersive X-ray detectors that rapidly acquire X-ray data from numerous individual points in polished sections of the sample. The software then uses the spectral data to identify the mineral or other phase represented at each point in the analysis, drawing on an internal database of known mineral compositions. Such an approach provides comprehensive information on the constitution, morphology, grain size, distribution and phase associations for several thousand particles in an ash sample. In this study, the abundance and average chemical composition of the amorphous or glassy components have been determined using quantitative X-ray diffraction, including tests based on different preparation and processing techniques. QEMSCAN was used to map the form and distribution of mineral phases and characterise the variation in glass composition for a range of ash materials. The calculated average chemical composition of the glassy phase provides information that can be related to particle density and surface area as well as ash behaviour in different utilisation processes. The relative abundances of the different crystalline components and the inferred chemical composition of the amorphous fraction can also be related to the mineral matter in the feed coal.

INTRODUCTION A comprehensive understanding of ash characteristics is essential for effective management of coal combustion by-products, including evaluation of different options for ash utilisation and assessment of short and long-term impacts of ash disposal in different environmental situations. The characteristics of fly ashes are traditionally evaluated from the overall abundance of particular elements based on chemical analysis data. Such data do not, however, indicate the ways in which the

2009 World of Coal Ash (WOCA) Conference - May 4-7, 2009 in Lexington, KY, USAhttp://www.flyash.info/

different elements actually occur within the fly ash, and hence how they might be expected to react under different conditions. Mineralogical analysis of fly ash, involving the determination of the relative abundance of the different phases, their size, shape and association, is an essential complement to the chemical analysis process. It is, however, an aspect that has traditionally been inadequately evaluated, due to a lack of quantitative analytical techniques. Recent technological developments have allowed the application of quantitative X-ray diffraction and automated electron beam image analysis techniques, providing both a basis for more comprehensive fly ash characterisation and an understanding the link between the characteristics of particular ashes and the coals from which they were formed. The data may then be used to predict more specifically the properties and behaviour of ashes associated with particular coals or utilisation processes. ADVANCED MINERALOGICAL ANALYSIS TECHNIQUES From the mineralogical viewpoint, fly ash can be regarded as essentially consisting of three types of components: minerals having a well-defined atomic structure (quartz, mullite, spinel etc), unburnt carbon particles, and non-crystalline aluminosilicate glass. The glass commonly represents the dominant component present in most fly ash samples and, because of its abundance and amorphous nature, is usually regarded as the principal component involved in the chemical reactions associated with fly ash utilisation, such as in the cement and concrete industry or in zeolite and geopolymer production. The glass in fly ash is also a major host within the ash for adsorbed trace elements, some of which may be released into the surrounding environment by leaching processes. Quantitative X-ray Diffraction Analysis Use of classical microscope techniques for the mineralogical analysis of fly ash is constrained by the fact that the crystalline components in fly ash are typically too small, and often too intimately inter-grown with other components, to allow reliable estimation of the glass and mineral contents. X-ray diffraction (XRD) provides a more definitive technique for identifying minerals and other crystalline phases in a wide range of materials [1], including coal ash [2-4], especially where the individual particles are too small to be reliably identified by microscopic techniques. Although traditionally regarded as qualitative, the use of XRD as a quantitative tool has been considerably enhanced in recent years following development of purpose-specific computer processing systems based on the principles of full-profile XRD analysis developed by Rietveld [5]. One of these is the SIROQUANT technique, developed in Australia by Taylor [6]. Based on Rietveld principles, the SIROQUANT method uses full profile fitting routines to generate a synthetic pattern that can be systematically adjusted through a user-friendly interface to match the observed XRD profile of the sample under analysis. The scaling factors developed in this process are then used to evaluate the percentages of the different crystalline phases or minerals present. As well as computation of standard XRD patterns from crystallographic and chemical data, reference patterns have also been developed for use in SIROQUANT directly from measured XRD traces of the mineral in question [8, 9]. These allow phases

having imperfectly known or poorly developed crystal structures to be incorporated in the analysis, with data derived from the measured patterns refined and quantified along with those for the other reference components using Rietveld processing methods. Such a capability is particularly relevant to the clay minerals, but is also of potential use in evaluation of glassy components in coal ash materials. The validity of the mineral percentages provided by X-ray diffraction and SIROQUANT, including materials containing clay minerals and similar poorly crystalline components, has been tested for coal and other rocks against independent information derived from chemical analysis and in some cases petrographic data [7, 10-13]. The (inferred) chemical composition of the mineral mixture indicated by SIROQUANT, for example, is usually very close to the (actual) chemical composition of the same material as determined by direct chemical analysis, suggesting a high level of consistency and increasing general confidence in use of the technique. Automated Electron Beam Image Analysis Although CCSEM (Computer Controlled Scanning Electron Microscopy) has been widely used in coal research since the 1980’s, the application of QEMSCAN technology has only recently been investigated. QEMSCAN was initially developed for applications in the mineral processing industry and as such has features not available in CCSEM. These include the provision of morphological information in association with textural characteristics and phase associations. This is possible as QEMSCAN collects compositional information on a pixel-by-pixel basis rather than on a particle basis as is done in CCSEM. Originally developed within the Commonwealth Scientific and Industrial Research Organisation (CSIRO) of Australia as QEM*SEM (or Quantitative Evaluation of Minerals by Scanning Electron Microscopy), the system uses a Windows-based PC operating system to control an automated SEM fitted with four energy dispersive X-ray spectrometers (EDS) which have enhanced light element capability (carbon and oxygen) due to the use of thin polymer windows. This capability enables improved mineralogical identification and classification with the ability to discriminate phases such as iron oxides (hematite and limonite) and siderite in mineral matter. The ability to detect the characteristic radiation of carbon and oxygen further provides the potential to analyse the coal matrix (i.e. the macerals), thus providing information on the textural setting of minerals in coal on a particle-by-particle basis. The measurement modes employed in data acquisition for QEMSCAN are fundamentally different from those used in most CCSEM systems in that an X-ray spectrum is acquired at each point of a rectangular grid superimposed on the particle, thus building up a phase composition map of the particle. QEMSCAN uses both backscattered electron and energy dispersive X-ray signals to create digital images in which each pixel corresponds to a mineral species, or phase, in a region under the electron beam. With the use of light element X-ray detectors, and carnauba wax as a mounting medium, coal (maceral) particles and particles consisting of coal-mineral mixtures, can also be identified and classified. The backscattered electron signal is used to discriminate between the mounting medium and the particles of interest. After calibration of the BSE signal, the sample is moved under computer control to a particular field of view (a frame) which is rapidly

scanned by the electron beam to identify the mineral particles from the mounting medium based upon preset threshold values of the BSE intensity. Further discrimination can be made to identify particles that are on the frame boundary, particles that are either too large or too small for consideration, or particles having a particular shape as defined by the operator (Figure 1).

Figure 1: Partial screen capture view, partway through a frame during automated QEMSCAN measurement of a typical pf coal sample. Pixel spacing used was ~2 microns. Note the progressive way in which the system measures individual particles. Particles on the edge of the frame have been discarded in this run, but can be measured if required. After particles have been identified and selected for measurement, an X-ray spectrum is collected from each point of a user-defined grid. The spacing of analytical points can be varied to give three basic scan modes, depending on the type of information and statistical accuracy required to satisfactorily address the particular issue. Modal Mineralogical Analysis (MMA) employs an equi-point rectangular grid for analysis, based upon the point scan described above, which is analogous to optical point counting. Only modal abundances are measured (which may be expressed as either weight or volume per cent), and no information is obtained on size or mineral association. An inferred bulk chemical composition may also be calculated using the chemistry and abundance of the constituent phases.

Bulk Mineralogical Analysis (BMA) is based on line scans, with closely spaced points in the X direction along widely spaced lines. The distance between lines is selected to be greater than the largest particle size, so that each particle in the sample is traversed by no more than one linear transect. Information is provided on modal abundance and bulk composition (as with MMA), but the output can additionally include grain size and association data. Particle Mineralogical Analysis (PMA) is the most detailed form of QEMSCAN analysis and is based on the area scan of closely spaced points in the X and Y direction. This mode of analysis provides the most detailed information on particle composition, shape, size, and association. However, fewer particles are measured than in the MMA or BMA modes, so that more information is obtained at the expense of statistical accuracy. Depending on the point spacing selected, the analysis also takes significantly longer. A field scan also uses an area scan of equally spaced points, but differs from PMA in that an X-ray spectrum is taken at every point within the defined area. It is used to show the texture in large specimens, and contiguous fields can be taken to build up an image of a specimen that is larger than a single field

`

Pyrite

Kaolinite

Coal

500 microns

Figure 2: QEMSCAN PMA analysis mode showing the grid of analytical points superimposed on the particle, the X-ray spectra obtained from selected points and the SIP classification based upon the X-ray spectrum.

A 1,000-count X-ray spectrum (typically collected within 5 milliseconds) is obtained from each point, and the elements present and their X-ray intensities are used to identify the mineral species present at that point. This is done by comparison with a look-up table (the Species Identification Program (SIP)), from which a species number is assigned to the pixel (Figure 2). The X-ray spectrum is matched to a mineral whilst the next spectrum is being acquired, thus allowing approximately 100,000 spectra to be processed and identified in one hour of measurement. The beam is then moved to another point and the process repeated until all selected particles in the frame have been measured. Another frame is then selected and

another set of points measured, the process being repeated until a preset number of particles have been analysed (Figure 3).

Figure 3: Partial screen capture of the resulting file of digital images (unprocessed) that were created following measurement, using Intellection’s iExplorer data-interrogation software. Some of the minerals and phases present are listed on the left-hand side of the figure. The SIP table is one of the most important elements of the QEMSCAN system as it is the primary phase classification table. Potentially it may have several hundred entries that contain information on the BSE response, the elements present and their relative intensities. It may also contain supplementary information in the form of logical operators as to which elements must be present and those which may be present. The SIP is created from the analysis of standards from which reference X-ray spectra are obtained. These reference spectra are then used to generate simulated 1,000-count spectra against which the SIP is tested for correct assignment of each spectrum to a particular mineral phase. To facilitate studies of particular materials, it may be necessary to reduce the size of the SIP to more manageable levels. A list of the mineral species likely to be present in a particular sample group is built up by combining entries in the SIP to produce what is known as a Primary Species list. Secondary and Tertiary mineral lists may then be constructed, each of which contain a more limited number of mineral species

that can be used in specific applications. This capability is illustrated in the mineral maps for Australian coals presented in Figure 4. As well as the mineralogy, textural information such as particle size and association can also be presented to illustrate the different modes of mineral occurrence within and between samples. Given the digital nature of the output from QEMSCAN, attributes of individual particles can be interrogated using a stand-alone desktop analysis system, known as iExplorer, which allows the operator to simultaneously view images, tables and charts. With the use of filters, images can then be sorted based on particle and mineral properties. Areal percentages for each phase are calculated from which weight percentages can be derived if the densities of the phases are known. Particle size is given by calculation of the average length from contiguous pixel groups and association of each species by counting the number of transitions from the pixels of one phase to those of another phase. Particle perimeters and shape are also calculated. The shape factor is a simple one, being largely a measure of the degree of sphericity; a value of 1.0 represents a spherical particle and 0.0 a highly irregular particle. Bulk compositional data can be obtained and elemental deportment between minerals can also be extracted. As shown in Figure 5, the data can be presented in tabular form or graphically, and can be imported directly into reports. QEMSCAN analysis can be performed on a wide variety of samples. These include standard prepared milled samples (-212 μm and -1 mm), sample chips or blocks, and large samples such as ash deposit samples. Granulated samples, chips, and blocks are usually prepared as 30 mm diameter polished mounts using epoxy resin as the mounting medium. Standard 50 mm by 25 mm polished thin sections (as used for electron microprobe analysis) can also be mounted in a special thin section sample holder. Samples are usually polished to avoid potentially misleading BSE and X-ray artefacts that can be produced by surface topography. The use of four X-ray detectors minimises (but does not entirely eliminate) surface topography effects. The prepared sample is coated with a thin layer of carbon to ensure electrical conductivity, thereby preventing charge build-up with consequent beam instability and image distortion. The very low atomic number factor associated with coal macerals is very similar to that of typical epoxy resins used for specimen preparation, resulting in a similar back-scattered electron response that makes coal particles indistinguishable from the mounting medium. A different mounting medium is needed to allow measurement of both the organic and inorganic components in coal samples. Carnauba wax has been found to be acceptable for this purpose due to its good contrast with coal organic matter (Figure 6).

Figure 4: QEMSCAN mineral maps of coal particles, illustrating the range of minerals and textures present. Particle mineral maps of three Australian coal samples are presented to illustrate the range of compositions, textures and associations. Particles are sorted by decreasing area.

Figure 5: Digital, colour coded particle mineral maps generated by QEMSCANTM from measurement of an Australian coal sample under automation, along with associated mineralogical modal and distribution data, and phase association table.

Figure 6: Backscattered electron photomicrograph and line scan showing the back-scattered electron signal brightness variation between carnauba wax, coal and minerals. The dark matrix is wax, the grey particles coal, and the bright matter is mineral.

APPLICATION TO ASH AND COAL SAMPLES Relation of Ash and Feed Coal Mineralogy A range of coal combustion by-products, and also some feed coals have been examined using both quantitative X-ray diffraction and QEMSCAN analysis, to evaluate the effectiveness of the different analytical techniques. Representative samples of the coals were subjected to ashing at around 120°C in an electronically-excited oxygen plasma [14] using an IPC 4-chamber plasma ashing unit, at 35 watts RF power per chamber, and the weight percentage of low-temperature ash (LTA) determined. The coal LTAs and the ash samples were finely powdered, and each powder was subjected to XRD analysis using Cu Kα radiation. Scans were run from 2 to 90° 2θ, with increments of 0.04° and a counting time of 5 seconds per step. Although a mineral spike can also be added to evaluate the proportion of amorphous material present in the ashes [15], the XRD pattern of a poorly crystalline silicate phase (metakaolin) in the SIROQUANT database was used to represent the glass component of each fly ash. The diffractogram of each raw (i.e. unspiked) fly ash sample was processed through SIROQUANT with this phase included in the analysis task. The mineralogy of a range of fly ashes from Australian coals, as determined by this technique, is presented in Table 1. The mineralogy of the LTA from seven of the corresponding feed coals, also determined by X-ray diffraction and SIROQUANT, is given in Table 2, and the percentages of key phases in the coal mineral matter are compared to the proportions of relevant phases in the fly ash materials in Figure 7. Figure 7a shows that the absolute percentage of quartz in the fly ash is less than that in the corresponding LTA, although the relative proportions of quartz in the LTA of the feed coal and in the fly ash appear, with two exceptions (circled), to show a fairly strong correlation to each other. Considered in conjunction with the QEMSCAN results presented below, which typically show quartz occurring as discrete fragments in the fly ash, this suggests that much of the quartz in the coal tends to pass directly into the fly ash, without significant alteration. The lesser proportion of quartz in the fly ash, however, compared to the coal LTA, suggests that some quartz may become incorporated into the glassy components during ash formation, become preferentially incorporated in the bottom ash (not analysed), or possibly be lost as a silica fume into the combustion stream. Incorporation into glass or a fume would be most likely to occur if the quartz in the coal was predominantly of fine particle size, while incorporation into bottom ash might be associated with coarser quartz particles. A similar plot for the iron minerals in the coal and fly ash (Figure 7b) shows a more scattered but still broadly linear relationship, for five of the samples, between the percentage of iron oxide minerals in the fly ash and the percentage of siderite plus pyrite in the LTA of the corresponding feed coal. This suggests that the iron oxide minerals in the fly ash are derived mainly from the iron minerals present in the coals themselves. However, two fly ashes (circled) have high percentages of crystalline iron oxides yet only low percentages of iron minerals in the respective feed coals. The two samples in question are derived from relatively low-rank coals, and the iron

minerals in these particular fly ashes may be derived from iron occurring in some other form, such as separate amorphous phases or incorporated into the organic matter, in these particular coal samples. Table 1: Mineralogy of Australian fly ashes by X-ray diffraction and SIROQUANT

Power Station No 15 16 17 18 19 20 21 22 23 Quartz 18.1 2.3 5.2 25.1 26.3 10.0 9.7 5.3 4.9 Mullite 18.5 8.9 8.9 15.2 20.3 21.3 10.2 18.5 8.5 Cristobalite 0.0 0.0 0.1 0.1 0.1 0.0 0.1 0.0 0.0 Magnetite 0.6 1.9 0.0 1.5 0.6 0.0 0.7 0.3 0.0 Maghemite 1.1 1.3 0.4 1.8 0.4 0.4 0.8 1.1 0.1 Hematite 0.7 0.5 0.0 1.3 2.2 0.0 0.4 0.0 0.0 Metakaolin (= glass) 61.0 85.0 85.4 55.1 50.2 68.2 78.1 74.8 86.4

Table 2: Mineralogy of LTA isolated from feed coals associated with fly ash samples, using X-ray diffraction and SIROQUANT Power Station No 16 17 18 19 20 21 23 LTA % 23.3 23.9 13.3 16.8 19.0 27.8 21.7 Quartz 14.3 34.5 37.4 40.8 21.9 28.2 34.3 Kaolinite 63.0 40.6 41.0 48.4 36.6 23.8 31.0 Illite 4.5 8.9 7.9 3.2 0.7 Illite/smectite 12.0 11.1 4.6 39.9 35.3 32.6 Pyrite 1.8 6.2 0.9 0.2 Siderite 2.4 0.1 2.5 0.4 Calcite 1.4 1.9 Dolomite 1.0 0.7 0.2 Bassanite 7.4 0.6 1.6 1.9 1.0 1.2 0.5 Albite 2.1 Sanidine 3.4 0.7 Apatite 1.6 0.5 Anatase 0.6 Rutile 0.6 0.3 The material identified as glass in the present study may include small proportions of crystalline minerals, at concentrations close to or below the limit of detection by the XRD technique. Possible traces of poorly ordered calcite, for example, were identified in the diffractogram of one fly ash (Sample 22), but the material was in proportions too low for reliable quantification by the SIROQUANT technique. The behaviour of some of the ashes on leaching [16] is also consistent with the presence of small proportions of carbonate minerals. Such carbonates, if present, may represent remnants of calcite and/or dolomite in the original feed coals (Table 3.2). Alternatively, they may be the result of interaction between organically-bound Ca in the coals, the presence of which is probably responsible for the formation of bassanite in the LTA samples [17], and CO2 in the furnace atmosphere [18].

QUARTZ

0

10

20

30

40

0 10 20 30 40

Quartz in LTA of Coal (%)

Qua

rtz

in F

ly A

sh (%

)

50

A

IRON MINERALS

0

2

4

6

8

10

0 2 4 6 8

Siderite + Pyrite in LTA of Coal (%)

Fe O

xide

s in

Fly

Ash

(%)

10

B

Figure 7: Relation between (A) quartz and (B) iron minerals in fly ashes and in LTA of feed coal samples, as determined by XRD and Siroquant. See text for discussion of circled data points. Estimation of Glass Composition in Ash Samples The chemical composition of the crystalline mineral components in each fly ash was calculated, based on the weight percentages of each mineral as indicated by the respective SIROQUANT analysis. The aggregate percentage of each oxide for the crystalline phases was based on the stoichiometric composition of each mineral identified and the proportion of that mineral as a percentage of the whole ash sample. These percentages were then subtracted from the bulk chemical composition of the respective ash samples (Table 3), determined independently by X-ray fluorescence spectrometry, to calculate an inferred chemical composition for the glass fraction. Unburnt carbon, although determined separately for the ash samples, was not included in this calculation, and hence the chemistry of both ash and glass represents a composition normalised to a carbon-free basis. It should also be noted that the glass in the ashes is not necessarily of a uniform composition, but may vary within and between the individual ash particles as shown

in the QEMSCAN study presented below. The estimates of glass composition derived from the XRD data therefore represent an average assessment for each ash sample, embracing any compositional variations or unresolvable crystalline phases that may also be present. Table 3: Chemical composition of Australian fly ash samples, inferred composition of mineral and glass fractions [15]. Whole ash chemistry* Power Stn No 15 19 18 16 20 17 21 22 23 SiO2 56.8 57.0 58.3 44.5 62.9 67.0 61.5 57.5 65.9 Al2O3 26.3 25.0 22.2 30.7 29.3 24.8 22.4 28.2 27.6 Fe2O3 9.5 9.9 13.6 14.4 1.8 3.1 7.6 5.6 1.1 CaO 1.4 1.5 1.3 4.2 1.3 1.0 3.3 3.8 0.4 BaO 0.4 0.5 0.4 0.1 0.1 0.0 0.1 0.1 0.0 MgO 0.8 0.7 0.8 1.6 1.1 0.6 1.1 1.2 0.3 Na2O 0.2 0.2 0.2 0.4 0.8 0.6 0.9 0.2 0.2 K2O 0.7 0.5 0.4 0.9 0.5 1.6 1.9 1.1 2.9 TiO2 1.7 1.5 1.7 1.9 1.8 1.0 0.9 1.6 1.3 P2O5 1.9 2.7 1.0 1.0 0.1 0.2 0.2 0.5 0.2 SO3 0.3 0.5 0.1 0.3 0.2 0.1 0.1 0.2 0.1 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 * After Killingley et al [19], corrected for carbon and LOI

Inferred chemistry – minerals Group A Group B

Power Stn No 15 19 18 16 20 17 21 22 23 Minerals % 52.7 55.0 53.9 20.0 39.0 20.9 33.7 34.0 26.3 SiO2 61.3 68.5 65.0 27.8 50.2 53.7 55.1 40.7 54.5 TiO2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Al2O3 33.0 21.5 24.4 42.7 48.1 40.5 33.0 51.9 42.9 Fe2O3 5.7 10.0 10.6 29.5 1.8 5.7 11.9 7.4 2.7 MgO 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 CaO 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Na2O 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 K2O 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 P2O5 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 SO3 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Inferred chemistry – glass Group A Group B

Power Stn No 15 19 18 16 20 17 21 22 23 Glass % 47.3 45.0 46.1 80.0 61.0 79.1 66.3 66.0 73.7 SiO2 52.3 43.8 50.8 48.8 71.1 70.5 64.9 66.2 66.0 TiO2 3.7 3.3 3.7 2.4 3.0 1.2 1.4 2.5 1.3 Al2O3 19.2 29.5 19.8 27.8 17.4 20.6 17.0 16.0 27.6 Fe2O3 13.9 9.8 17.5 10.6 1.8 2.4 5.4 4.7 1.1 MgO 1.6 1.6 1.9 2.0 1.8 0.8 1.7 1.7 0.3 CaO 2.9 3.4 2.8 5.3 2.1 1.2 5.0 5.8 0.4 Na2O 0.4 0.3 0.3 0.4 1.4 0.7 1.3 0.2 0.2 K2O 1.4 1.2 0.9 1.1 0.8 2.1 2.9 1.7 2.9 P2O5 3.9 6.0 2.2 1.2 0.2 0.3 0.2 0.7 0.2 SO3 0.6 1.1 0.2 0.3 0.3 0.1 0.2 0.4 0.1 Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0

Use of stoichiometric compositions for the different minerals may represent a slight oversimplification. Electron microprobe analyses carried out by the authors show that significant concentrations of iron are often present in the mullite of Australian fly ash. A small but significant proportion of Mg, along with traces of Mn, Ca and Si, may also occur in the magnetite component, an observation supported by unpublished data on magnetite compositions for Australian fly ashes, which often contain aluminium in addition to the elements mentioned above. Similar data on mineral chemistry were not available for the fly ashes in the present study, but even so, the differences relative to stoichiometric compositions would be expected to be relatively small. They would probably not significantly affect the broader comparisons indicated in the present study for the relatively diverse suite of Australian fly ash materials. Based on the glass percentages indicated by SIROQUANT analysis, the percentage of glass in the Australian fly ashes studied ranges from 45 to 80% (Table 3). The inferred glass composition data suggest that the fly ash samples fall into two very distinct groups. One such group (Group A in Table 3) has glass with a relatively low percentage of SiO2 (45-60%) and a high percentage (more than around 10%) of Fe2O3. The other (Group B) has glass with apparently high SiO2 (60-70%) and low (typically <5%) Fe2O3 contents. Although specific details of the respective sites are confidential, the ashes of Group A were derived from coals that are essentially of sub-bituminous rank, while those of Group B were derived from combustion of bituminous coal feedstocks. QEMSCAN Analysis of Pulverised Fuel Fly Ash Four other Australian pf fly ash samples were subjected to QEMSCAN examination, to examine the capacity of the system to evaluate further any variations in chemistry and mineralogy. Two samples were taken from Station B at different time intervals when different coals were being combusted to investigate the effect, if any, of variation in feed coal characteristics. All samples were taken from Zone 1 of the electrostatic precipitators used for ash collection. The silica contents of the ashes from Stations A and B are similar, but the ash from Station A has the highest iron content and an intermediate calcium content (Table 4). Alumina contents are similar for all ashes. In spite of the use of different coal feeds, the two ashes from Station B are similar with respect to their chemistry. The ash from Station C is distinctive in its high silica content and low contents of all other elements, except for alumina. Amorphous aluminosilicate is the dominant phase identified by QEMSCAN in all of the ash samples, with mullite and quartz being the most abundant crystalline phases (Table 5). In spite of the high silica values obtained for the ash from Station C, this ash has an intermediate quartz content; the highest quartz is in the ash from Station A which has the lowest silica content (Table 4). The ashes from stations A and C have the highest mullite contents. The ash from Station C is distinctive in lacking any crystalline iron bearing phases, reflecting its extremely low iron content. Hematite and magnetite are the dominant iron-bearing phases in the Station A and B fly ashes, with maghemite and ferrian spinel also being present in the ashes from Station B. Minor amounts of calcite, anhydrite and aluminium sulphate are also present in the ashes from Station B. Although the chemistries of the two ashes from Station B are similar, the ash from Phase 2 has a lower quartz content, higher mullite, and a lower content of iron minerals.

Table 4: Chemistry of fly ash samples analysed by QEMSCAN Locality Station A Station B Station C

Description ESP Zone 1 Phase 1 ESP Zone 1

Phase 2 ESP Zone 1 ESP Zone 1

SiO2 51.77 56.00 53.18 71.81 TiO2 1.37 0.92 1.17 1.25 Al2O3 23.24 21.70 24.81 22.77 Fe2O3 14.44 10.29 8.92 0.59 Mn3O4 0.13 0.13 0.09 0.01 CaO 1.65 3.05 3.99 0.05 MgO 1.09 1.66 1.18 0.11 Na2O 0.27 0.50 0.40 0.07 K2O 0.95 1.54 1.53 0.25 P2O5 1.63 1.30 1.65 0.05 SO3 0.30 0.25 0.59 0.03 SrO 0.34 0.06 0.07 0.00 BaO 0.46 0.20 0.24 0.02 LOI 2.12 2.85 2.96 1.84 Total 99.75 100.43 100.78 98.86

Note: LOI = Loss on Ignition at 1050oC Table 5: Mineralogy of fly ash samples as indicated by QEMSCAN

Locality Station A Station B Station C

Description ESP Zone 1

Phase 1 ESP Zone

1

Phase 2 ESP Zone

1 ESP Zone

1

Quartz 19.9 7.4 4.8 12.2 Cristobalite 0.1 Mullite 14.9 7.6 11.0 18.1 Hematite 2.0 0.9 0.7 Magnetite 1.9 1.0 0.7 Maghemite 0.8 0.6 Spinel ferrian 0.5 Calcite 0.3 0.3 Aluminium sulphate 0.6 Anhydrite 0.1 Amorphous 61.3 81.5 81.4 69.6 Total 100.0 100.2 100.1 99.9

A particle mineralogical analysis was carried out on polished grain mounts prepared from the four fly ash samples, with a minimum of 3,000 particles being counted at a pixel spacing of 1μm. Immediately apparent from the particle maps (Figures 8 and 9) are the variations in grain size, particularly the coarse grained nature of the ash from Station C. The ash from Station C also shows a distinct relationship of morphology to composition, silica-rich particles derived from quartz being either angular or highly irregular in contrast to the rounded nature of aluminosilicate particles derived from clays. Although less abundant, silica-rich angular grains are also present in the Station A ash.

Fi

gure

8: Q

EM

SC

AN

par

ticle

map

of f

ly a

sh fr

om le

ft) S

tatio

n A

and

righ

t) S

tatio

n B

, Pha

se 1

.

Fi

gure

9: Q

EM

SC

AN

par

ticle

map

of f

ly a

sh fr

om le

ft) S

tatio

n B

, Pha

se 2

and

righ

t) S

tatio

n C

The particle maps also reveal variations in the composition of the amorphous phase, with iron bearing aluminosilicate particles being present in the ash from Station A and the Phase 1 ash from Station B. Such particles are less common in the Phase 2 ash from Station B, and are absent in the Station C ash. These observations are consistent with the chemistry and mineralogy of the ashes. Although the iron oxides identified in the XRD analysis were not found in the QEMSCAN analysis, this is most likely due to the fine-grained nature of these oxides and their intimate association with the glass matrix. Similarly, the failure to identify mullite is most likely due to the compositional similarity of the mullite grains and the matrix glass, combined with the fine grained nature of the mullite crystals. As indicated in the previous examples, QEMSCAN analysis provides information on the distribution and association of the minerals identified in the XRD analysis; it can also be used to identify variations in glass composition that may be relevant to fly ash utilisation options. Particle size and shape can also be determined, as well as the possible relationship of these physical properties to ash chemistry. This may also have relevance to coal or ash utilisation options.

QEMSCAN Analysis of Fluidised Bed Combustion Ashes Samples of cyclone and bottom ash were obtained from a pressurised fluidised bed combustor and of ash from various points in the ash handling system of a conventional fluidised bed combustor. As shown in Table 6 the samples have a complex mineralogy, which may impact on ash handling characteristics and affect behaviour on disposal or utilisation. The purpose of the QEMSCAN examination was to show how the technique could be used to examine the mineral distribution and mineral association, and provide information to assess ash behaviour. Both ashes from the pressurised fluidised bed system (PFBC) are distinguished by their high calcium contents, reflecting the use of limestone as a bed material: those from the atmospheric fluidised bed system contain moderate amounts of calcium and have higher silica and alumina values (Table 7). Potassium contents are also higher in the atmospheric FBC ashes due to the use of shale as a bed material. Sulphur contents are variable; they are most likely a function of the available calcium, as there is a strong correlation between sulphur and calcium concentrations. The mineralogy of the PFBC ashes is complex and is dominated by the occurrence of calcium bearing silicates such as gehlenite, merwinite and anorthite (Table 6). Calcite is abundant in the bed ash due to the use of limestone as the bed material, and lime and portlandite result from the decomposition of calcite. The presence of anhydrite is a consequence of the reaction of lime and portlandite with SO3 present in the flue gas. Amorphous aluminosilicate is abundant in the cyclone ash. In contrast, the atmospheric FBC ashes have a much simpler mineralogy, dominated by the occurrence of abundant amorphous aluminosilicate that is most likely derived from the decomposition of kaolinite. Quartz is the most common crystalline phase. Illite is derived from either the coal feed, the shale bed, or a combination of both, and hematite most likely reflects the breakdown of siderite. The calcium mineralogy is simple, with lime, portlandite, anhydrite and ettringite being the most common phases; calcite is of minor occurrence. The occurrence of the calcium minerals is

largely restricted to the bottom and economiser ashes (Table 6), apart from the presence of ettringite in the silo and reacted fly ashes. The handling difficulties experienced with this ash are most likely due to the formation of ettringite by reaction between the other calcium-bearing phases, such as portlandite and lime, and the fine-grained aluminosilicate amorphous phase. The complexity of the mineral distribution in the PFBC samples is well illustrated in the QEMSCAN particle maps (Figure 10). The predominance of coarse-grained calcite, lime and portlandite is evident in the bottom ash. Also shown is the reaction of aluminosilicate phases derived from clays with calcium, forming a thin rim of anorthite on the aluminosilicate grains. Calcium sulphate occurs as partial rim on the larger grains but is concentrated in the finer size factions as discrete grains. The mineralogy of the cyclone ash is also complex. Large calcium bearing particles are probably derived from the bed material and, as in the bottom ash, calcium sulphate and anorthite can be found rimming aluminosilicate and calcium rich particles. Abundant calcium sulphate is present in the finer particle sizes. Calcium sulphate is also present as complex mixtures in polymineralic grains, suggesting that it may be acting as a cementing agent and thus could lead to potential problems with ash handling. Many of the particles identified as mullite are more likely to be amorphous aluminosilicates of similar composition derived from the breakdown of clays. The particle morphology is also distinctive, most of the PFBC ash particles being angular in contrast to the rounded ash particles commonly found in pf boilers. This is most likely due to the lower temperature of combustion resulting in less glass formation. The bottom ash, economiser ash and fly ash from the atmospheric FBC unit are mineralogically similar and show a decreasing grain size passing from the bottom ash to the fly ash as would be expected (Figures 11 to 13; see Figure 14 for legend). Although calcium sulphate is identified, no distinction is made between the other Ca-bearing phases, portlandite, lime and calcite. While it may be possible to distinguish calcite with development of the SIP, it will not be possible to separate lime and portlandite; XRD is required for this purpose. Quartz and aluminosilicates derived from clays dominate the mineral assemblage, and many of the aluminosilicate grains have an incomplete narrow rim of a calcic phase. In contrast, silo ash B is more complex mineralogically both within and between grains. Calcium bearing phases are common and, unlike the other materials, are often intimately intermixed with other phases, especially in the coarser particles. The particle size of silo ash B is also notably coarser than that of silo ash A (Figure 12). Part of this apparent coarseness may be due to the calcium phases cementing smaller grains to form polymineralic aggregates, thus affecting the ash handling behaviour. The old ash (Figure 13) is unusual in consisting almost entirely of coarse quartz grains; this would explain its unreactive nature. As mentioned above, some of the calcium phases such as portlandite and lime cannot be distinguished in a QEMSCAN analysis, and integration of QEMSCAN and XRD is required for a full mineralogical characterisation. In addition, ettringite was not identified in the QEMSCAN analysis, most likely due to the lack of a suitable SIP entry. However, QEMSCAN is able to identify variations in the chemistry of the amorphous particles, and this can represent useful information that is not otherwise obtainable.

Tabl

e 6:

Qua

ntita

tive

X-ra

y di

ffrac

tion

min

eral

ogy

of F

BC a

sh s

ampl

es

PF

BC

FB

C

Phas

e B

otto

m A

sh

Cyc

lone

Fly

A

sh

Bot

tom

Ash

Ec

onom

iser

Si

lo F

ly A

sh

A

Silo

Fly

Ash

B

R

eact

ed F

ly

Ash

O

ld A

sh

Qua

rtz

9.8

3.7

20.3

21.1

25.4

25.7

25.1

78.4

Illite

2.5

2.6

2.2

3.0

Alb

ite

0.

8M

ullit

e

6.1

Geh

leni

te

6.3

8.5

Mel

ilite,

sod

ic

5.7

5.8

Mer

win

ite

5.0

6.6

Ano

rthite

4.

4 11

.8M

aghe

mite

1.0

Hem

atite

1.6

3.0

2.1

2.0

1.3

0.3

Rut

ile

0.

4P

eric

lase

0.7

Lim

e 16

.7

3.0

11.0

Por

tland

ite

2.6

0.6

10.1

Cal

cite

43

.6

3.1

0.6

0.4

2.1

Anh

ydrit

e 5.

8 7.

53.

08.

81.

8E

ttrin

gite

11.7

6.3

Am

orph

ous

41.6

62.1

52.8

68.0

60.6

65.2

17.8

Tota

l 99

.9

100.

010

0.2

100.

199

.910

0.0

100.

099

.9

Tabl

e 7:

Maj

or e

lem

ent c

hem

istry

of F

BC a

sh s

ampl

es.

PFB

C

FBC

Bot

tom

Ash

C

yclo

ne F

ly

Ash

B

otto

m A

sh

Econ

omis

er

Silo

Fly

Ash

A

Si

lo F

ly A

sh

B

Rea

cted

Fly

A

sh

Old

Ash

SiO

218

.71

27.5

555

.00

44.6

252

.41

44.3

737

.87

92.0

0Ti

O2

0.35

0.

970.

590.

500.

80

0.58

0.60

0.26

Al2O

38.

26

19.9

822

.70

19.5

620

.41

15.8

512

.38

4.60

Fe2O

31.

41

3.37

2.72

3.90

3.86

3.

132.

560.

88M

n 3O

40.

03

0.05

0.02

0.05

0.03

0.

030.

030.

01C

aO

46.0

0 31

.77

12.3

018

.11

3.86

8.

9810

.25

0.28

MgO

1.

14

1.91

0.75

0.64

0.60

0.

430.

440.

12N

a 2O

0.

49

0.81

0.17

0.31

0.16

0.

320.

550.

23K 2

O

0.54

0.

451.

391.

221.

06

0.86

0.79

0.60

P 2O

50.

16

0.29

0.13

0.14

0.60

0.

350.

10<0

.005

SO3

4.27

6.

314.

865.

482.

51

3.90

4.47

0.89

SrO

0.

06

0.19

0.04

0.05

0.08

0.

090.

040.

01B

aO

0.02

0.

150.

020.

030.

05

0.06

0.03

0.02

LOI

15.9

8 6.

104.

2011

.93

12.6

0 19

.79

n.d.

3.30

Tota

l 97

.41

99.9

010

4.89

106.

5399

.03

98.7

270

.11

103.

20 N

otes

:

LOI L

oss

on ig

nitio

n at

105

0o C.

A

naly

sis

of re

acte

d fly

ash

is b

y se

mi-q

uant

itativ

e en

ergy

dis

pers

ive

X-ra

y flu

ores

cenc

e.

Fi

gure

10:

QEM

SCAN

par

ticle

map

of l

eft)

PFBC

cyc

lone

ash

and

righ

t) PF

BC b

otto

m a

sh.

Fi

gure

11:

QEM

SCAN

par

ticle

map

of l

eft)

FBC

bot

tom

ash

and

righ

t) FB

C e

cono

mis

er a

sh.

Fi

gure

12:

QEM

SCAN

par

ticle

map

of l

eft)

FBC

silo

ash

A a

nd ri

ght)

FBC

silo

ash

B.

Fi

gure

13:

QEM

SCAN

par

ticle

map

of l

eft)

FBC

reac

ted

ash

and

right

) FBC

old

ash

.

BackgroundAl OxideQuartzMulliteAnorthiteIllite/K-sparCalcium ferritePericlaseTitanium dioxideIron oxideCalcium oxide/carbonateCalcium phosphateCalcium sulphateSi-richSi.KAl.Si. (Minor)OthersAl.Si.Ti.K.Ca.FeSi.(Ca.Ti.Fe)

Figure 14: Legend for the QEMSCAN mineral maps.

CONCLUSIONS Rietveld-based X-ray diffraction analysis provides a useful technique for fly ash characterisation, allowing the proportions of the different minerals and the glass within the ash to be evaluated directly in quantitative terms. Determination of the amorphous content can be based on either an ash sample spiked with a known proportion of a separate mineral, or more directly by using a low-crystallinity silicate mineral such as metakaolin in the Rietveld processing to represent the glass component. An estimation of the overall glass composition can bed calculated, based upon the difference between the composition of the crystalline phases and that of the fly ash as a whole. Although the samples were derived from a diverse range of coals and combustion plants, the crystalline quartz content and, to a lesser extent, the proportions of iron minerals in the pf fly ashes appear to be related to the abundance of equivalent minerals in the parent coal samples. The composition of the glass fractions also suggests that the ashes can be separated into two groups, apparently reflecting the rank of the coal that dominated the original fuel source. Case studies on pf and FBC ashes have demonstrated that QEMSCAN analysis can provide unique information not readily obtainable by other means. The methods of QEMSCAN data acquisition provide advantages over other SEM based techniques in that measurements are made at the individual pixel level rather than averaged over an area. Greater use is also made of the chemical composition for phase identification, rather than relying upon back scattered electron intensity. More particularly, data can be obtained on particle size and shape, phase identification and

abundance, as well as the mode of occurrence and association of the identified phases. Unlike other techniques such as X-ray diffraction, QEMSCAN analysis can also supply information on variations in chemistry of the amorphous phase, which is relevant to issues such as ash deposition in pf boilers and ash behaviour in fluidised bed systems. Quantitative X-ray diffraction and QEMSCAN analysis are powerful complementary techniques but neither is, in itself, an analytical panacea. The two techniques should be used in conjunction with each other, and with other techniques such as chemical analysis and electron microprobe investigation. QEMSCAN is a beam analysis technique and, as such, has an effective resolution of 1-2μm. Consequently, fine-grained minerals such as dispersed quartz in coal and mullite in fly ash may not be identified; this must be borne in mind, for example, when comparing QEMSCAN data with quantitative XRD mineralogy. Polymorphs such as those of silica (quartz, tridymite and cristobalite) cannot be identified; phases of similar composition, such as hematite and magnetite, also cannot be reliably distinguished from each other due to the short counting times employed in a typical QEMSCAN analysis. Further work on SIP development is still required in specific areas, such as reliable identification of the clay minerals in coals and better discrimination of glass compositions in coal utilisation products. ACKNOWLEDGEMENTS Funding for the work outlined in this paper was provided by the Australian Coal Association Research Program (ACARP) and the Co-operative Research Centre for Coal in Sustainable Development (CCSD); the latter was supported in part under the Co-operative Research Centres Program of the Australian Government. The authors wish to thank Paul Marvig and Christa Pudmenzky for sample preparation, Zhongsheng Li for XRD analysis, and Al Cropp, Jaco van Zyl and Rayni Newberry for their assistance and advice in QEMSCAN data acquisition. REFERENCES 1. Jenkins R, Snyder RL. Introduction to X-ray Powder Diffractometry. John Wiley and Sons,

New York 1996: 403 pp.

2. Vassilev S, Menendez R, Alvarez D, Diaz-Soamarano M, Martinez-Tarazona MR. Phase-mineral and chemical composition of coal ashes as a basis for their multi-component utilization: 1. characterization of feed coals and fly ashes. Fuel 2003; 82: 1793-1811.

3. Vassilev S, Vassileva CG, Karayigit AI, Bulut Y, Alastuey A, Querol X. Phase-mineral and chemical composition of feed coals, bottom ashes and fly ashes at the Soma power station, Turkey. International Journal of Coal Geology 2005; 61: 35-63.

4. Vassilev S, Vassileva CG, Karayigit AI, Bulut Y, Alastuey A, Querol X. Phase-mineral and chemical composition of fractions separated from composite fly ashes at the Soma power station, Turkey. International Journal of Coal Geology 2005; 61: 65-85.

5. Rietveld HM. A profile refinement method for nuclear and magnetic structures. Journal of Applied Crystallography 1969; 2: 65-71.

6. Taylor JC. Computer programs for standardless quantitative analysis of minerals using the full powder diffraction profile. Powder Diffraction 1991; 6: 2-9.

7. Ward CR, Taylor JC, Cohen DR. Quantitative mineralogy of sandstones by X-ray diffractometry and normative analysis. Journal of Sedimentary Research 1999; 69: 1050-1062.

8. Taylor JC, Zhu R, Simultaneous use of observed and calculated standard profiles in quantitative XRD analysis of minerals by the multiphase Rietveld method – the determination of pseudorutile in mineral sands products. Powder Diffraction 1992; 7: 152-161.

9. Taylor JC, Matulis CE. A new method for Rietveld clay analysis: Part 1 – use of a universal measured standard profile for Rietveld quantification of montmorillonites. Powder Diffractometry 1994; 9: 119-123.

10. Ward CR, Spears DA, Booth CA, Staton I, Gurba LW. Mineral matter and trace elements in coals of the Gunnedah Basin, New South Wales, Australia. International Journal of Coal Geology 1999; 40: 281-308.

11. Ward, CR, Matulis CE, Taylor JC, Dale LS. Quantification of mineral matter in the Argonne Premium Coals using interactive Rietveld-based X-ray diffraction. International Journal of Coal Geology 2001; 46: 67-82.

12. Ruan CD, Ward, CR, 2002. Quantitative X-ray powder diffraction analysis of clay minerals in Australian coals using Rietveld methods. Applied Clay Science 2002; 21: 227-240.

13. Ward CR, Gomez-Fernandez F. Quantitative mineralogical analysis of Spanish roofing slates using the Rietveld method and X-ray powder diffraction data. European Journal of Mineralogy 2003; 15: 1051-1062.

14. Standards Australia. Higher rank coal - mineral matter and water of constitution. Australian Standard 1038 Part 22; 2000: 20 pp.

15. Ward CR, French D, Determination of glass content and estimation of glass composition in fly ash using quantitative X-ray diffractometry. Fuel 2006; 85: 2268–2277.

16. Jankowski J, Ward CR, French D, Groves S. Mobility of trace elements from selected Australian fly ashes and its potential impact on aquatic ecosystems. Fuel 2006; 85: 243-256.

17. Ward CR. Analysis and significance of mineral matter in coal seams. International Journal of Coal Geology 2002; 50: 135-168.

18. Bauer CF, Natusch DFS. Identification and quantitation of carbonate compounds in coal fly ash. Environmental Science and Technology 1981;15:783-88.

19. Killingley J, McEvoy S, Dokumcu C, Stauber J, Dale L. Trace element leaching from fly ash from Australian Power Stations. End of Grant Report, Australian Coal Association Research Program, Project C8051, 2000; CSIRO Division of Energy Technology.