Kaolin clay, quartz and feldspars are used for the production of whiteware porcelains and vitrified ceramic products that are white in colour or artificially coloured, translucent (except for thick pieces) and resonant (Singer & Singer, Reference Singer and Singer1963). Their characteristics have been reviewed recently in the case of vitreous sanitary ware by Bernasconi et al. (Reference Bernasconi, Marinoni, Pavese, Francescone and Young2014). Chen et al. (Reference Chen, Wang and Hon2004) studied the kinetics of mullite formation and the evolution of mullite grain width in kaolin sintering at 1300°C. The phase transformation of kaolinite, as a function of temperature, was reviewed extensively by Chakraborty (Reference Chakraborty2014). The locations of the compositions of various ceramic products in a triaxial diagram (Romero & Perez, Reference Romero and Perez2015) are shown in Fig. 1.

Fig. 1. Triaxial diagram showing compositions of various porcelain products, modified after Romero and Pérez (Reference Romero and Perez2015). Group 5 also includes α-alumina filler.

The raw materials and the processing and microstructure of porcelains were studied by Carty & Senapati (Reference Carty and Senapati1998). The formation of the first liquid happens at temperatures >985°C, related to the disappearance of K-feldspar and subsequent progressive decrease in crystalline SiO₂ until total disappearance. Mullite, formed mainly from kaolinite, is the only crystalline phase in equilibrium with melt at the end of the process, although SiO₂, which is not totally dissolved in the liquid phase, usually remains in porcelains. Mullite compositions may be referred to as the solid solution series Al_4+2xSi_2−2xO_10−x, corresponding approximately to the range 55–90 mol.% Al₂O₃ (Schneider et al., Reference Schneider, Schreuer and Hildmann2008). The two stoichiometric compositions of mullite are 2Al₂O₃·SiO₂ and 3Al₂O₃·2SiO₂ (referred to by many as 2:1 or primary mullite and 3:2 or secondary mullite, with 77 and 72 wt.% Al₂O₃, respectively). Mullite compositions ranging from 2:1 to 3:2 have been observed in clay-derived vitreous ceramics (Lee et al., Reference Lee, Souza, McConville, Tarvornpanich and Iqbal2008).

Mullite crystals show various morphologies as a function of the processing method. In the case of standard porcelains, small (<0.5 μm long) primary 2:1 mullite crystals formed in clay relicts and 3:2 acicular (>1 μm long) mullite crystals formed in feldspar relicts were distinguished by Iqbal & Lee (Reference Iqbal and Lee1999). Iqbal & Lee (Reference Iqbal and Lee2000) and Lee & Iqbal (Reference Lee and Iqbal2001) reviewed the compositions and morphologies of three different types of mullite formed during the firing of porcelains. Type I, 2:1 primary mullite, with a crystal aspect ratio in the range 1:1–3:1, grows in micro-regions of kaolinite clay agglomerates. Type II, 3:2 secondary mullite, with a crystal aspect ratio of 3:1–10:1, crystallizes from feldspar-clay micro-regions. Finally, type III, 3:2, secondary needle-like mullite, with a crystal aspect ratio of 30:1–40:1, forms in micro-ternary regions of fine clay, feldspar and quartz. These types were also observed in industrial blends made of 50% kaolinitic clay, 40% feldspar and 10% quartz of porcelain stoneware (Martín-Márquez et al., Reference Martín-Márquez, Rincón and Romero2010).

The appearance of the various phases in industrial porcelains was studied by scanning electron microscopy (SEM; Carty & Senapati, Reference Carty and Senapati1998; Iqbal & Lee, Reference Iqbal and Lee1999, among others). Microstructural features of porcelains are characterized by various fields such as feldspar melts with well-developed, elongate mullite crystals and continuous or discontinuous areas where small mullite crystals were grown from the kaolinite matrix. Large, partially dissolved quartz crystals surrounded by melt were observed, along with relicts of small quartz crystals almost totally dissolved in the melt. Large quartz crystals were usually surrounded by cracks; no cracks were observed in small grains.

The microstructure and improvements in the properties of porcelain have been considered in the past (Iqbal, Reference Iqbal and Lee2000, among others). In these works, analysis of the mullite microstructure was mainly performed by optical microscopy (OM) and SEM. However, powder X-ray diffraction (XRD) is used increasingly for microstructural analysis of materials (e.g. Mitemeijer & Scardi, Reference Mittemeijer and Scardi2013).

Crystallite-size data for mullite in ceramics have been presented in numerous works (Serrano et al., Reference Serrano, Bastida, Amigó and Sanz1996; Sainz et al., Reference Sainz, Serrano, Bastida and Caballero1997, Reference Sainz, Serrano, Amigó, Bastida and Caballero2000; Kojdecki et al., Reference Kojdecki, Serrano, Clausell and Bastida2001). Sanz (Reference Sanz2015) integrated XRD microstructural analysis with OM and SEM studies to establish the microstructural evolution of mullite on triaxial porcelains. The porcelains were produced from three blends that differed only in terms of the nature of one of the triaxial constituents, considering two types of firing cycles (faster and shorter or slower and longer) and four firing temperatures in a restricted industrial range (1270–1340°C). No significant variance was observed for the (110) crystallite size following analysis of the whole set of samples, considering the types of blends and firing cycles as independent variables. However, a preferential elongation was observed when comparing crystallite sizes (with greater sizes for 001 reflections than for 110 reflections) by the Voigt function method (Langford, Reference Langford1978) with greater lengths of prismatic faces measured using electron microscopy.

XRD microstructural analysis by the method of Warren & Averbach (Reference Warren and Averbach1950) using (001) reflection, which involved average crystallite size and size distributions, was used for deducing the crystal-growth mechanisms of phyllosilicates and other minerals with the Mudmaster program (Eberl et al., Reference Eberl, Środoń, Lee, Nadeau and Northrup1987, Reference Eberl, Drits, Środoń and Nüesch1996, Reference Eberl, Drits and Środoń1998; Lanson & Kübler, Reference Lanson and Kübler1994). Mean crystallite sizes obtained from (001) reflections (estimated from XRD patterns by the Voigt function method and by the Warren–Averbach method) were in agreement with measurements of crystallite thickness by electron microscopy (Clausell et al., Reference Clausell, Bastida, Serrano, Pardo and Huertas2007; Pardo et al., Reference Pardo, Bastida, Serrano, Ibañez and Kojdecki2009).

Sainz et al. (Reference Sainz, Serrano, Bastida and Caballero1997) observed simultaneous and parallel growth of crystallites and crystal aggregates during the thermal transformation of kyanite using SEM and XRD microstructural analysis (Voigt function and Warren–Averbach methods). The method of Kojdecki (Reference Kojdecki2004) was used by Sanz et al. (Reference Sanz, Bastida, Kojdecki, Caballero and Serrano2009) for XRD microstructural analysis of mullite developed in the slow and long firing of two triaxial blends with different feldspar raw materials, with or without petalite. Prismatic crystallite was the dominant form, a preferential growth of crystallites was observed when the petalite-bearing blend was used and greater size was detected at higher firing temperatures. The Warren–Averbach method was also used by Sanz et al. (Reference Sanz, Bastida, Kojdecki, Caballero and Serrano2011) for XRD microstructural analysis of mullite in whitewares produced by fast firing of two triaxial blends differing in the particle size of the quartz raw material. Significant differences in the mean crystallite size of mullite were found at temperatures >1320°C for blends with finer quartz particle sizes.

The crystallite size in powder XRD corresponds to the domain size of coherent diffraction and is not directly comparable to crystal size measured by microscopic methods. However, increasing mean volume-weighted (D₁₁₀) was related to the width of mullite crystals measured by electron microscopy, formed by firing of kyanite at 1350 and 1600°C over long periods (Sainz et al., Reference Sainz, Serrano, Bastida and Caballero1997).

XRD microstructural analysis may provide information on various directions and is particularly useful when a particular direction of diffraction corresponds with the predominant crystal faces. Kaolinite is an excellent example of this case, as mean 001 XRD crystallite sizes are directly comparable (being of a similar order of magnitude) to direct measurements of crystallite thickness performed by SEM (Clausell et al., Reference Clausell, Bastida, Serrano, Pardo and Huertas2007; Pardo et al., Reference Pardo, Bastida, Serrano, Ibañez and Kojdecki2009). However, this approach may not be applied to mullite and thus complementary observations of crystals by electron microscopy are necessary in order to establish a relationship between the growth mechanisms in crystallites and crystals (Sainz et al., Reference Sainz, Serrano, Bastida and Caballero1997; Yoon et al., Reference Yoon, Sarin and Kriven2008). The D₁₁₀ size is measured in a direction perpendicular to the 110 planes, corresponding to that with the lowest crystal-growth velocity in mullite (Serrano, Reference Serrano, Bastida, Amigó and Sanz1996) and thus it is particularly useful for comparing the crystallite growth achieved by the mullite crystals in various conditions.

The aim of the present work was to study the influence of two different industrial kaolins, which do not influence significantly the chemical composition of the triaxial blend used, in the production of porcelains with respect to the mullite content, the average composition of mullite, the crystallite size using the Warren–Averbach method by analysis of the 110 diffraction direction and the relationship with the width of mullite prismatic faces {110} observed by SEM.

MATERIALS

Raw materials

The raw materials used were kaolin clays (kaolin B and kaolin M), feldspar flux (S) and quartz (G). The feldspar S (100% <50 μm) comes from Quaternary deposits of silica-feldspar sands in Segovia (Spain) and is K₂O-rich. The quartz (G) is an industrial washed sand (100% <50 μm) from Segovia (Spain) with a very low level of impurities.

Kaolin B is sedimentary kaolin (100% <60 μm) from Arguisuelas, Cuenca (Spain) with kaolinite microstructural characteristics similar to that of Poveda de la Sierra, Guadalajara (Spain) (Aparicio & Galán, Reference Aparicio and Galán1999), while kaolin M is an halloysitic industrial kaolin (100% <10 μm) of hydrothermal origin from Matauri Bay, New Zealand (Wilson & Keeling, Reference Wilson and Keeling2016), wet processed and dried to obtain a halloysite content of ~88%. Crystal morphologies of halloysite provide rheological properties different to those of high-quality kaolins made of platy crystals (Clarke, Reference Clarke2008), which affect casting.

Additional details for the raw materials (particle-size distribution, method for chemical analysis by X-ray fluorescence, thermal analysis) and processing can be found in the survey of Sanz (Reference Sanz2015). The mean particle size of kaolin B is 6 μm, while that of kaolin M is 0.3 μm.

The chemical compositions of kaolin B and kaolin M are listed in Table 1. The kaolins have comparable SiO₂, Al₂O₃ and Na₂O contents. Kaolin B is richer in K₂O, Fe₂O₃, CaO and MgO and has a smaller LOI value than kaolin M. These differences in chemical composition are only significant for K₂O in the blends studied here (BGS with kaolin B and MGS with kaolin M), as shown below. The greater K₂O content in kaolin B is mainly related to the more abundant muscovite (cf. Fig. 2).

Fig. 2. XRD patterns of kaolins M and B. The reflections are labelled to match those in the PDF2 database of the International Centre for Diffraction Data (ICDD): in sample M, there is overlap of the first peaks of halloysite (labelled H) and illite (black lines with upwards-pointing triangles) and coincidence of kaolinite (red lines with crosses) with 7 Å halloysite. There is overlap of muscovite (labelled Mu) and illite in sample B. Quartz peaks are indicated by blue lines with inverted triangles.

Table 1. Chemical analyses of kaolins provided.

LOI = loss on ignition at 1000°C.

The XRD traces of the kaolins are shown in Fig. 2. Kaolin B contains kaolinite and minor illite and quartz, while kaolin M contains a mixture of kaolinite and 7 Å halloysite and minor quartz. The predominant polytype is kaolinite because the ratio of intensities of the peaks at 7.0 and 4.4 Å is ~20 (instead of <2, which is typical of 7 Å halloysite; Brindley, Reference Brindley and Brown1961).

The XRD patterns of kaolins B and M differ mainly in terms of the full width at half maximum (FWHM) of the reflection near 7 Å (which was related to the crystalline thickness in the 001 diffraction direction (Amigó et al., Reference Amigó, Bastida, Sanz, Signes and Serrano1994). Kaolin M has a larger FWHM value than its kaolin B counterpart (0.220° and 0.164°, respectively) in the XRD patterns (Fig. 3), where profiles are represented at normalized height (d ₀₀₁ spacing of both peaks at 7.16 Å). Both kaolins were recognized by FESEM (see Fig. 4) using the method of Clausell et al. (Reference Clausell, Bastida, Serrano, Pardo and Huertas2007). The predominant particle shape of both kaolins is platy, but in kaolin M, flattened cylinders were also found in which unclosed pores were noted. Note the presence of a flattened pore at the bottom right of Fig. 4a. The statistics of >100 measurements for each type of kaolin particle are included in Table 2.

Fig. 3. Profiles of 001 peaks of the B and M kaolins at normalized height.

Fig. 4. Field emission scanning electron microscopy images. (a) Halloysitic kaolin M. A longitudinal opening along a particle oblique to the plane of the image is indicated by an arrow. (b) Kaolin B. Scale bars: 500 nm.

Table 2. Statistical analysis of >100 FESEM measurements of crystallite thicknesses.

Production of whitewares

The formulation of whitewares consisted of 53% kaolin, 33% feldspar and 14% quartz. Two blends (BGS, MGS) were prepared from the two kaolins (B and M). Both blends were fired in an electric kiln at four different temperatures with fast/short (5°/min up to firing temperature held for 90 min) or slow/long (2°/min up to firing temperature held for 180 min) firing cycles, with subsequent unforced cooling. The samples were labelled by adding R or L to the composition label and adding 1, 2, 3 or 4 according to the firing temperature (1270, 1300, 1320 or 1340°C, respectively). The complete processing route was shown in the survey of Sanz (Reference Sanz2015). The chemical compositions of both blends are shown in Table 3 (data from Sanz, Reference Sanz2015).

Table 3. Chemical compositions (wt.%) of BGS and MGS. One significantly different value is marked in bold. Note that the difference for Na₂O + K₂O is not significant considering errors.

RESULTS

X-ray diffraction

Figure 5 shows a set of XRD patterns recorded with fast data collection (BGSL samples). Figure 6 shows sample data for the 110 and 220 diffraction peaks collected using the slow mode, which was used for microstructural analysis (by Warren–Averbach method), as well as of the 220 and 111 diffraction peaks used for estimating average Al₂O₃ contents in mullite. The ratio of mullite to (quartz + mullite) was determined using 120 and 100 reflections of mullite and quartz, respectively. The small amount of cristobalite present was not considered in the semi-quantitative analysis.

Fig. 5. Examples of XRD patterns collected with the fast mode (BGSL samples). Key to the symbols: mullite (squares), quartz (diamonds), cristobalite (triangles).

Fig. 6. Selected diffraction peaks obtained by slow data collection. The 110 and 220 peaks belong to mullite. The 220 and 111 peaks were used to estimate average Al₂O₃ content in mullite. The 100 peak of quartz was used as an internal intensity standard of crystalline constituents.

Table 4 lists the average Al₂O₃ content in mullite and the semi-quantitative analysis of mullite and quartz in the crystalline fraction of the porcelains. The calculated values for Al₂O₃ content in mullite were very similar and near the stoichiometric 3:2 composition (71.8 wt.% Al₂O₃), without significant variations considering errors were in the range of 5.7–6.9%. These values were similar to those reported by Serrano et al. (Reference Serrano, Bastida, Amigó and Sanz1996) for mullites formed in stoneware from illite–kaolinite compositions by long firings (50–150 h at 1150°C) of kaolinite KGa-1 of the Source Clays Repository of The Clay Minerals Society and were also similar to sintered mullite compacts fired with a heating rate of 1°/min up to 950°C, then with 5°/min up to 1300°C and held at 1300°C for 30 min from a powdered Malaysian kaolin (Chen et al., Reference Chen, Wang and Hon2004).

Table 4. Al₂O₃ content in mullite and mullite content in the crystalline fraction of porcelain.

REF = reference of porcelain sample; A = wt.% of Al₂O₃ in mullite; M = wt.% of mullite in the crystalline fraction of the sample; RER %< = maximum value for relative error in each column.

Figure 7 shows the advance of mullite formation (wt.%) in the crystalline fraction of the samples. A significant increase in mullite content was observed in fast firing of the MGS blend (MGSR samples) with increasing temperature. Similar behaviour was observed in the fast firing of the BGS blend (BGSR samples) at 1300–1340°C. Under slow firing, the increase in mullite content was evident in MGS and BGS blends treated at 1320°C or 1340°C and at 1300°C or 1340°C, respectively. Under the same firing temperature, differences in mullite contents were significant and positive between MGSR and BGSR sets at 1270–1340°C and between BGSL and BGSR sets at 1270–1320°C (no significant differences were found at 1340°C). Negative differences were observed between the MGSL and MGSR sets heated at 1320 and 1340°C.

Fig. 7. Mullite content (M%) in the crystalline fraction of the two blends (data from Table 4).

Table 5 lists descriptive parameters of the mullite 110 crystallite size distribution in the samples studied and the fired reference kaolinite GW for comparison (with data from Serrano et al., Reference Serrano, Bastida, Amigó and Sanz1996) calculated using the Warren–Averbach method.

Table 5. Main parameters of size distribution of D₁₁₀ of mullite for the various firings at various temperatures (°C). MGSL MGSR, BGSL and BGSR correspond to the MGS and BGS blends with long, slow firings (L) and short, fast firings (R). Values in bold represent significant positive differences.

^a Lack of significant difference of the property at subsequent temperatures. Mean = mean crystallite size <D₁₁₀>; Mo-110 = modal value of D₁₁₀; T (°C) = maximum firing temperature; RER%< = maximum relative error; FWHM = full width at half maximum.

The evolution of mean crystallite size <D₁₀₀> with firing temperature is shown in Fig. 8. The mean crystal sizes of mullites of the MGS and BGS blends are comparable at 1270 and 1340°C and in slow firing cycles, but significantly positive at 1300°C (very slight) and 1320°C (see also Table 5). Under increasing firing temperature, significantly higher values were only recognized for MGS, at 1320°C under slow firing (Fig. 8a) and also at 1320°C (with no significant difference in value at 1340°C) under fast firing (Fig. 8b), and no significant differences in the cases of the other firings. Figure 9 shows the distributions of crystallite sizes D₁₁₀ for various sets of samples. In fast firing of the MGS blend, the modal crystallite size value is gradually displaced to larger particle sizes with increasing temperature (Fig. 9b). The FWHM of distribution and the mean <D₁₁₀> values follow the same trend, reaching a maximum value at 1320°C. The lower modal values of crystallite sizes observed in the BGS2L (1300°C) and BGS3L (1320°C) samples are probably related to artefacts produced by the size-calculation program, as the peak profiles in the XRD traces do not show particular anomalies (cf. Fig. 5).

Fig. 8. Evolution of mean crystallite size <D₁₁₀> with firing temperature. (a) Slow firing, (b) fast firing. See text for discussion.

Fig. 9. Frequency of crystallite size distributions (%) vs. size value <D₁₁₀> (nm). Distributions 1L, 2L, 3L and 4L and 1R, 2R, 3R and 4R correspond to firings at 1270, 1300, 1320 and 1340°C for slow (L) and fast firing cycles (R), respectively. (a) MGSL, (b) MGSR, (c) BGSL and (d) BGSR.

No significant differences in <D₁₁₀> mean crystallite sizes were observed between slow or fast firings of kaolin B and M blends at any firing temperature. Exceptions are the higher <D₁₁₀> values in the fast firing at 1320°C of kaolin B and in the slow firing at 1300°C of kaolin M (Table 5). Significant positive differences of mean crystallite sizes <D₁₁₀> were found for slow firing of blends with kaolin B and kaolin M at any firing temperature, except at 1340°C, in which significant differences were not observed. No significant differences in <D₁₁₀> were observed in the fast firing cycles of both blends at the same temperature.

Higher modal values of D₁₁₀ crystallite size were observed at the same firing temperatures in both slow and fast firing cycles in blends with kaolin M (Table 5). No significant differences in modal crystallite sizes of mullite after slow and fast firing were observed for both blends at any temperature, except at 1300°C (slight significant negative difference). However, the differences for both blends at the same temperature were significantly positive, except at 1340°C (no significant difference). Finally, differences in FWHM were only significant above 1320°C in the slow firing of the blend with kaolin B and in the fast firing of the blend with kaolin M.

EVOLUTION OF MICROSTRUCTURE

The maximum sizes observed in the crystals of the porcelains studied were <20 μm, corresponding to partially dissolved quartz grains. Figure 10 shows a contraction crack around a quartz grain also affecting the grain, which is surrounded by a viscous Si-rich melt with tiny mullite grains. The presence of minute mullite grains may account for the bimodal distributions of mullite determined by the Warren–Averbach method in some whiteware porcelains (Sanz, Reference Sanz2015).

Fig. 10. Scanning electron microscopy image of a coarse quartz grain (light grey) with minute mullite grains growing in the dissolved area (MGSL4 sample, polished etched surface). Scale bar: 5 μm.

FESEM microphotographs of the broken surfaces of samples etched chemically with hydrofluoric acid (at 10% in water) are shown in Fig. 11. Table 6 lists comparative statistics of width measurements obtained from >600 measurements from each sample and the mean values (<D₁₁₀>) of XRD crystallite sizes for the 110 reflection (Table 5). Larger mean widths (MW) of mullite prisms are usually associated with larger values of <D₁₁₀> although exceptions are observed (e.g. MGS1L and MGSL3). Greater MW/<D₁₁₀> ratios were observed in the MGS blends.

Fig. 11. SEM images of the etched porcelain samples. (a) MGS1L, (b) MGS3L, (c) MGS1R, (d) MGS3R, (e) BGS1R and (f) BGS3R. Scale bars: (a–d) 250 nm, (e,f) 500 nm.

Table 6. Comparative results of mean values of measured widths (MWs) of mullite prisms and crystallite sizes. MW/D₁₁₀ values are indicative of the small numbers of crystallites across the prismatic {110} faces of mullite crystals.

W = width; <D₁₁₀> = mean X-ray diffraction crystallite size. σ = standard deviation of W measurements; MoW = mode of W or W value of maximum frequency; Min W = minimum value of W; Max W = maximum value of W.

Figures 11a and 11b show decreases in the MW and modal width of D₁₁₀ after slow firing of MGS at 1270 and 1320°C. Figures 11c and 11d show the increase in MW of mullite prisms (cf. Table 6) after fast firing of MGS at 1270 and 1320°C, concurring with significant changes of the <D₁₁₀> parameter (cf. Table 5). Figures 11c and 11d show the apparent increase in MW from 1270 to 1320°C in fast firing of the MGS blend. Figures 11a and 11c (corresponding to slow and fast firing of the MGS blend at 1270°C, respectively) show the greater value for slow firing at this temperature. Figures 11e and 11f, corresponding to the BGS sample fired at 1270 and 1320°C, also show the increase in MW under fast firing of the BGS blend. An apparent increase in MW of the mullite prism may be observed, which is in agreement with Table 6, concurring with a significant increase in <D₁₁₀> (Table 5). Note that the widths of mullite prismatic faces may be equidimensional if the faces are perpendicular to the plane of the image.