nach oben

Erschienen in:

Open Access 2024 | OriginalPaper | Buchkapitel

6. Temperature-Induced Changes in the Microstructure of the Metal Melt Filters and Non-metallic Inclusions

verfasst von : Martin Thümmler, Milan Dopita, Mykhaylo Motylenko, Anton Salomon, Erica Brendler, David Rafaja

Erschienen in: Multifunctional Ceramic Filter Systems for Metal Melt Filtration

Verlag: Springer International Publishing

Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.

search-config

KI-gestützte Suche

Patentsuche

Aus

Abstract

The functionalization of ceramic foam filters aims typically at the enhancement of the thermal shock resistance and the reactivity of the filters with respect to specific inclusions and impurities. For this purpose, thermodynamically metastable phases are utilized that have a strongly defective crystal structure and/or nanocrystalline character. Such phases possess frequently better or even unique properties in comparison with their thermodynamically stable counterparts. However, the stability of metastable or defect-rich phases is usually impaired by microstructural changes, which occur during the contact of these phases with the metallic melt at high temperatures and which speed up finally the degradation of the functionalized filters. In general, the first step towards the stabilization of the thermodynamically metastable and/or defect-rich phases is the understanding of their microstructure and the microstructure changes accompanying the transition to the thermodynamically stable state. In this chapter, the thermally induced microstructure changes are illustrated on the examples of selected carbon containing binders and metastable alumina phases. In order to be able to describe the crystal structure and microstructure of these compounds in more details, which is required for the targeted development of the functional filter materials, complementary methods of crystal structure and microstructure analysis like X-ray and electron diffraction, X-ray and nuclear magnetic resonance spectroscopy and electron microscopy were combined and further developed.

6.1 Introduction

The most prominent inclusions, which are present in the cast components made of steels or aluminum alloys, are aluminum oxides. These inclusions deteriorate dramatically the mechanical properties of the cast components and reduce significantly their lifetime [1‐3]. Commonly, the inclusions are classified as exogenous or as endogenous according to their origin. Exogenous inclusions are present already in the starting material. Exogenous alumina inclusions appear typically in form of thermodynamically stable corundum (α-Al₂O₃). Endogenous inclusions form during the casting process, for instance as a consequence of a local oversaturation of the melt by oxygen [4]. Endogenous alumina inclusions consist usually of metastable aluminum oxides, mainly of γ-Al₂O₃, δ-Al₂O₃ and θ-Al₂O₃, in particular in the early stages of the formation process [5].

Functional filters must be capable of removing these inclusions from the metal melt either by attaching the oxide particles to the filter surface or by dissolving them in a chemical reaction between the filter surface and the melt. The attachment of the inclusions is facilitated by the epitaxial processes, which occur preferentially at the interfaces of the counterparts having similar crystal structures. For the second process, a selective reactivity of the functionalized filter material with specific elements is needed. In both cases, a high efficiency of the filtration process requires a good adhesion of the non-metallic inclusions and/or reaction products to the functionalized filter surface in order to inhibit their spalling and the contamination of the metallic melt. For these reasons, the filter surface is functionalized by compounds, which have similar crystal structure and similar chemical composition as the non-metallic inclusions [6, 7]. Furthermore, the metal melt filters have to withstand extreme conditions, in particular high thermal shock and the contact with highly corrosive environments at high temperatures [8].

One of the well-established materials used for production of the metal melt filters is the carbon-bonded alumina (Al₂O₃-C). Dudczig et al. [9], Zienert et al. [10] and Salomon et al. [11] have shown that the Al₂O₃-C filters react with liquid iron and with the oxygen solved in the steel melt in a complex reaction, which helps to decrease the oxygen concentration in the steel and to remove the alumina inclusions from the melt. Another positive characteristics of the carbon-bonded alumina filters is their good thermo-shock resistance, which is further improved by the addition of a high melting coal tar resin or pitch [12] to the binder phase. In the current developments, the coal tar resin or pitch are replaced by environmentally friendlier and less toxic substitutes like lactose or tannin [13].

At high temperatures, the carbon binders undergo structural and microstructural changes that are accompanied by changes in the materials properties, which can negatively affect the filtration process. In this chapter, the thermally induced microstructure changes and phase transformations in the coal tar pitch, tannin and lactose are discussed. The other materials under study, namely γ-Al₂O₃, δ-Al₂O₃ and θ-Al₂O₃ are considered for production of functional coatings, because their structure is either identical or closely related to the crystal structure of the metastable alumina phases within the metallic melt, which are present in endogenous inclusions.

6.2 Carbon Binders for Ceramic Filter Bodies

The most prominent carbon-containing binders used in the ‘Collaborative Research Centre 920’ for production of the ceramic foam filters are the high melting coal tar pitch ‘Carbores P’ (Rütgers), the tannin-rich ‘Quebracho extract’ (Otto Dille) and conventional lactose. All of them consist mainly of carbon and hydrogen. Tannin and lactose contain, in addition to C and H, different amounts of oxygen, and are considered as environmental friendly alternative to the high melting coal tar pitches. At high temperatures, these compounds are expected to produce nanocrystalline graphite with a strongly distorted crystal structure, which serves as thermoshock resistant binder.

6.2.1 Characterization of the Microstructure of Carbon Binders

The X-ray diffraction (XRD) experiments have shown that turbostratic carbon is a substantial part of ‘Carbores P’. Turbostratic carbon can be described as a stack of bent graphene layers having different mutual rotations with respect to the $c$ axis, different shifts along the $a$ and $b$ axes, and noticeable fluctuations of the stacking distance as compared to hexagonal (2H) graphite (Fig. 6.1).

In the reciprocal space, a single graphene layer produces infinitely extended rods along the ${c}^{*}$ axis (Fig. 6.2a). The size of the rods in the ${a}^{*}$ and ${b}^{*}$ directions is inversely proportional to the lateral size of the graphene layer. This reciprocity of the size of the objects in the direct and reciprocal space is a generally valid phenomenon, which is described in many textbooks on X-ray diffraction, e.g., in [14]. A three-dimensionally ordered crystal structure of graphite produces reciprocal lattice points, which size is reciprocally proportional to the size of the graphite crystallite (Fig. 6.2a) in the respective direction. The stacking disorder in turbostratic carbon causes a broadening of the reciprocal lattice points along the ${c}^{*}$ axis and the mutual rotations of individual graphene layers a rotation of the rods around this axis.

When the reciprocal space of a powder sample of turbostratic carbon is scanned in an XRD experiment, the ‘scattered intensities’ located at the intersections of the broadened reciprocal lattice points with the Ewald sphere (Fig. 6.2b) are integrated according to the powder pattern power theorem, see, e.g., [15]. This integration reveals diffracted intensity for example as a function of the diffraction angle ($2\theta$), cf. Fig. 6.2c. As the reciprocal lattice points $00l$ of turbostratic carbon form a chain of local intensity maxima, which is located along the ${c}^{*}$ axis (Fig. 6.2a), the Ewald sphere intersects these points always only in the ${c}^{*}$ direction. Consequently, the corresponding intensity maxima in the powder XRD pattern are broadened symmetrically (Fig. 6.2c). Their width is inversely proportional to the size of the turbostratic graphite clusters in the $c$ direction (${L}_{c}$ in Fig. 6.1b) and scales linearly with the magnitude of the diffraction vectors, if the stacking (interplanar) distances vary. Other reciprocal lattice ‘points’ than $00l$ are strongly elongated along the ${c}^{*}$ axis by the shift of individual graphene layers along the $a$ and $b$ directions and by their mutual rotation around the $c$ axis (Fig. 6.1b), which leads to an overlap of the reflections $hkl$ with the same indices $h$ and $k$ but different $l$. These reciprocal lattice points are approached by the Ewald sphere in the $h{{\varvec{a}}}^{\boldsymbol{*}}+k{{\varvec{b}}}^{\boldsymbol{*}}$ direction, which causes an asymmetrical broadening of the diffraction lines with non-zero indices $h$ and $k$ and the formation of strongly asymmetrical $\{hk\}$ bands (Fig. 6.2c).

Within the kinematical X-ray diffraction theory, the total intensity scattered coherently by a cluster of atoms results from the interference of the electromagnetic waves scattered by individual atoms [15]:

$${I}_{{\text{coh}}}\left({\varvec{q}}\right)\propto {\left|F\right|}^{2}={\sum }_{m}{f}_{m} \, \mathrm{ exp}\left(\mathrm{i }{\varvec{q}} {{\varvec{r}}}_{m}\right){\sum }_{n}{f}_{n} \, \mathrm{ exp}\left(-\mathrm{i }{\varvec{q}} {{\varvec{r}}}_{n}\right)$$

(6.1)

In Eq. (6.1), ${\varvec{q}}$ is the diffraction vector and $F$ the structure factor (neglecting anomalous absorption and dispersion) of the cluster. The magnitude of the diffraction vector ($\left|{\varvec{q}}\right|\equiv q$) is related to the diffraction angle ($2\theta$) like $q=4\pi \sin\theta /\lambda$, where $\lambda$ is the X-ray wavelength. Individual atoms are characterized by their atomic scattering factors ($f$) and positions (${\varvec{r}}$). The integration of the scattered intensity over the Ewald sphere, which describes the X-ray scattering on an ensemble of randomly oriented clusters (on a powder sample), leads to the Debye equation [16]:

$$\begin{aligned} {I}_{{\text{coh}}}\left(q\right) & =\sum_{m}\sum_{n}\frac{{f}_{m}{ f}_{n}}{4\uppi {r}_{mn}^{2}}\int_{0}^{\pi} {\int_{0}^{{2\pi }} {\exp}}\left(\mathrm{i }q {r}_{mn} \cos\psi \right) {r}_{mn}^{2} \sin\psi \,{\text{d}}\phi {\text{d}}\psi \\& =\sum_{m}\sum_{n}{f}_{m} {f}_{n}\frac{{\text{sin}}\left(q {r}_{mn}\right)}{q {r}_{mn}} \,\, {\text{with}} \,\, {r}_{mn}=\left|{{\varvec{r}}}_{m}-{{\varvec{r}}}_{n}\right|\\ \end{aligned}$$

(6.2)

In contrast to the general formulation of the X-ray scattering within the kinematical approximation [Eq. (6.1)], the Debye Eq. (6.2) takes into account all possible orientations of the cluster in the direct space and operates with the interatomic distances (${r}_{mn}$) instead of with the atomic positions. A principal drawback of the Debye equation is a high number of the atomic pairs, $\left({N}^{2}-N\right)/2$, , which scales almost quadratically with the number of the atoms in the cluster ($N$). Thus, it was assumed that individual graphene layers have the same structure and the same lateral dimensions, and that they are only randomly rotated around the $c$ direction. In such a case, Eq. (6.2) can be replaced by [17, 18]:

$$\frac{{I}_{{\text{coh}}}}{{f}_{C}^{2} N}=i\left(0,q\right)+2\sum_{k=1}^{P-1}\left(1-\frac{k}{P}\right)i(k,q)$$

(6.3)

A consequence of this assumption is that the clusters of turbostratic carbon are approximated by cylindrical objects having the diameter ${L}_{a}$ and the height ${L}_{c}$ (Fig. 6.1b). In Eq. (6.3), ${f}_{C}$ is the atomic scattering factor of carbon, $P$ the number of graphene layers, $i\left(0,q\right)$ describes the X-ray scattering by a single layer and $i(k,q)$ is the Warren-Bodenstein integral [19, 20]:

$$i\left(k,q\right)=\frac{16N}{\uppi { L}_{a}^{2} P q} {\int_{\frac{1}{2} k c}^{\sqrt{{L}_{a}^{2}+\frac{1}{4}{k}^{2}{c}^{2}}}}\left({\text{arccos}}\left(u\right)-u\sqrt{1-{u}^{2}}\right){\text{sin}}(q r){\text{d}}r$$

(6.4)

With $u=\sqrt{{r}^{2}-0.25{k}^{2}{c}^{2}}/{L}_{a}$, this approximation contains only intuitive physical parameters. The parameter $c=2{d}_{002}$ represents the mean out-of-plane lattice parameter of the turbostratic carbon (Fig. 6.1). Another parameter of the model is the in-plane lattice parameter ($a$) of graphene, which defines the distances between the carbon atoms within the graphene layers. The calculation of the one-dimensional integral from Eq. (6.4) and the one-dimensional sum from Eq. (6.3) is significantly faster than the 'two-dimensional' summation in Eq. (6.2), because both operations [Eqs. (6.3) and (6.4)] scale linearly with the number of graphene layers. Finally, the scattered intensity from Eq. (6.3) was corrected for static displacements of carbon atoms in $c$ direction $({u}_{c})$ using the Debye–Waller factor $D{W}_{00l}={\text{exp}}\left(-\langle {u}_{c}^{2}\rangle {q}^{2}\right)$ [18]. This correction covers the bending of the graphene layers and the fluctuations of the interlayer distances.

6.2.2 Thermally Induced Changes in the Microstructure of Carbon Binders

The capability of the whole powder pattern fitting based on the theoretical background from Sect. 6.2.1 was first tested on ‘Carbores P’ [18, 21] and later applied to the new binder materials that are based on tannin (Quebracho extract) and lactose. In order to examine their temperature behavior, the binders were annealed for 3 h at the temperatures up to 1400 °C in reducing atmosphere (CO/CO₂ + N₂) to prevent their combustion.

After annealing at high temperatures, the XRD patterns of all carbon binders show the presence of the bands {10} and {11} (Fig. 6.3), which are the typical diffraction features of turbostratic carbon with a low number or with poorly ordered graphene layers, cf. Fig. 6.2. In the XRD patterns of ‘Carbores P’, also the diffraction lines $00l$ are visible, which indicate the formation of a turbostratic carbon structure with a higher number of parallel graphene layers.

The ‘Quebracho extract’ (denoted as tannin in Fig. 6.3) that was heat-treated above 600 °C contains crystalline phases, which were identified as Na₂CO₃ and CaS stemming from the impurities present in the natural product. The presence of Na, Ca, O, S (and Mg) was confirmed by energy dispersive X-ray spectroscopy (EDX) in the scanning electron microscope (SEM). Additionally, SEM revealed that the grains containing Na₂CO₃ and CaS are attached to the surface of agglomerated carbon clusters. Lactose was crystalline below 400 °C. It contains α-lactose monohydrate (C₁₂H₂₄O₁₂) at room temperature and the corresponding anhydride (C₁₂H₂₂O₁₁) that forms at temperatures in between 90 and 140 °C. The heat treatment of lactose caused excessive foaming, which was reduced by addition of TiO₂ powder to the samples heated above 400 °C. Consequently, the XRD patterns from Fig. 6.3 contain the diffraction lines from two TiO₂ modifications (anatase and rutile) and from TiN, which was produced in the reaction of TiO₂ with carbon and nitrogen: ${{\text{TiO}}}_{2}+2{\text{C}}+\frac{1}{2}{{\text{N}}}_{2}\to {\text{TiN}}+2{\text{CO}}$.

Most of the parameters of the structure model used for the quantitative description of the turbostratic graphite, i.e., ${L}_{a}$, ${L}_{c}$, $\langle {u}_{c}^{2}\rangle$, $a$ and $c$ (Fig. 6.4a-e), could be determined from the XRD patterns shown in Fig. 6.3. These parameters were complemented by the measurement of the gas pressure in a high-temperature (HT) chamber used for in situ annealing experiments (5 Kmin⁻¹) (Fig. 6.4f), by the ex situ measurement of the mass loss (Fig. 6.4g) and by the chemical analysis using carrier gas hot extraction (CGHE). These measurements revealed the carbon content (Fig. 6.4h) and the [H]/[C] ratio (Fig. 6.4i). A rapid increase of the gas pressure in the constantly evacuated chamber (cf. Fig. 6.4f) points to an intense formation of gaseous substances like H₂O, CO, CO₂, CH₄ etc. at the respective temperature.

In addition, the layer size ${L}_{a}$ was verified by Raman spectroscopy from the intensity ratio of the D (≈ 1355 cm⁻¹) and G (≈ 1575 cm⁻¹) modes [22]. These results (Chap. 5, Fig. 5.4) are in good agreement with the values obtained by XRD (Fig. 6.4a).

Coal Tar Pitch (Carbores P)

In comparison with other binder materials, ‘Carbores P’ has the highest carbon content of $94\left(1\right) {\text{wt}}\%$ (Fig. 6.4h) and the lowest [H]/[C] ratio of 0.53(5) (Fig. 6.4i) already in the initial state. The carbon content and the [H]/[C] ratio stay more or less constant up to approx. 400 °C. The carbonization process, which is generally associated with an increase of the relative carbon content (Fig. 6.4h) in organic materials upon heating, starts in ‘Carbores P’ above 600 °C. The carbonization process is typically accompanied by the growth of individual carbon layers (visible as an increase of L_a, Fig. 6.4a), and by the dehydration reactions and by the elimination of point defects [23, 24], which result in an increase of the lattice parameter $a$ [18]. After the graphene sheets are formed, the lattice parameter $a$ approaches the in-plane lattice parameter of graphite [25] (cf. Fig. 6.4d). An onset of the thermally induced graphitization of ‘Carbores P’ is documented by the formation of the $00l$ peaks (Fig. 6.3, cf. Fig. 6.2c) and by the increase of ${L}_{c}$ above 800 °C (Fig. 6.4b). Since the main components of the coal tar pitch [24] are large polycyclic aromatic hydrocarbons (PAHs), which possess nearly a perfect planar molecular structure, the displacement of carbon atoms in $c$ direction is negligibly small (Fig. 6.4c). However, the lattice parameter $c$ (Fig. 6.4e) is still larger than the corresponding intrinsic lattice parameter of graphite, even after annealing at 1400 °C. Thus, the crystal structure of graphite formed in calcined ‘Carbores P’ stays highly turbostratic in the whole temperature range under study. It should be noted that the typical graphitization temperatures are between 1600 to 3000 °C [23, 24].

Since ‘Carbores P’ contains mainly carbon and hydrogen, the gases, which form during the high-temperature treatment and which were responsible for the increase of the pressure (Fig. 6.4f) in the evacuated chamber used for annealing, are H₂ and hydrocarbons [23]. Additional consequences of the formation of these gases are the mass loss (Fig. 6.4g) and the decrease of the [H]/[C] ratio (Fig. 6.4i). However, whereas the mass loss is almost finished at 600 °C, the [H]/[C] ratio decreases up to the annealing temperature of approximately 1000 °C. As it can be assumed that the formation of H₂ causes mainly a decrease of the [H]/[C] ratio, while the formation of hydrocarbons is also responsible for mass loss, these results suggest that the formations of H₂ and hydrocarbons occur at different annealing temperatures. This is confirmed by the decrease of the absolute carbon content, which also indicated a formation of hydrocarbons only up to 600 °C (Fig. 6.4h). The correlation between the mass loss, the carbon content, the change in the [H]/[C] ratio (Fig. 6.4g-i), the cluster growth (Fig. 6.4a,b) and the consolidation of the lattice parameters (Fig. 6.4d,e) confirms that the carbonization process is associated with the removal of hydrogen from the graphene sheets. Thus, the carbonization process results in the increase of the in-plane lattice parameter a and leads finally to the growth of the graphene sheets in the lateral direction. As already mentioned above, the carbonization of ‘Carbores P’ starts at 600 °C. At this annealing temperature, the main mass loss, which is associated with the production and evaporation of hydrocarbons, is already almost completed.

The formation of volatile phases is finished at approximately 1000 °C. Above this temperature, the relative carbon content in annealed ‘Carbores P’ reaches nearly 100% (Fig. 6.4h). The in-plane lattice parameter of turbostratic graphite approaches the lattice parameter of 2H graphite. In contrast to the carbonization of ‘Carbores P’, which is almost finished above 1000 °C, its graphitization is still in progress. The clusters of turbostratic graphite grow. The out-of-plane lattice parameter ($c$) approaches the lattice parameter $c$ of graphite, but it is still far from its intrinsic value, because the van der Waals bonds between the neighboring graphene layers are not established yet in the turbostratic graphite.

Tannin (Quebracho Extract)

The [H]/[C] ratio measured in the utilized ‘Quebracho extract’ ([H]/[C] = 1.19) is slightly higher than the value of 1.06 expected for the main component profisetinidin (C₁₇H₁₈O₅), cf. Fig. 6.4i. However, as the [H]/[C] ratio decreased already at the annealing temperatures below 200 °C, it can be assumed that the excess of hydrogen in the ‘Quebracho extract’ stems from adsorbed water. This assumption is substantiated by a significant mass loss during the annealing in this temperature range (Fig. 6.4g). The released water vapor is also responsible for the first increase of the pressure in the evacuated annealing chamber (Fig. 6.4f). The second pressure increase is caused by the thermal decomposition of profisetinidin and by the evaporation of catechol, resorcinol and other organic compounds [26] having their boiling points between 200 and 300 °C. The thermal decomposition of profisetinidin is accompanied by cross-linking reactions involving hydroxyl groups (polycondensation). These cross-link reactions are responsible for large variations of the in-plane lattice parameter $a$ observed in samples annealed below 600 °C (Fig. 6.4d). In the early stages of the annealing process, the chaotic cross-link formation does not lead to the formation of carbon layers, which would be larger than the original size of the organic molecules, cf. Fig. 6.4a. Still these cross-links are a prerequisite of the subsequent layer growth and the main carbonization process above 600 °C that proceeds like in ‘Carbores P’. During this heating period, the diffraction bands {10} and {11} become more pronounced.

In contrast to ‘Carbores P’, the parallel arrangement of graphene sheets and the formation of turbostratic graphite in calcined tannin is inhibited even at the highest annealing temperatures, as it can be seen from the absence of pronounced diffraction lines $00l$ (Fig. 6.3). At the positions of these XRD lines, only very broad maxima of the diffuse scattering were observed. This means that the parameters of the microstructure model, which are typically determined from the positions, shape and intensities of the diffraction lines $00l$, i.e., the vertical size of turbostratic carbon clusters (${L}_{c}$), the out-of-plane lattice parameter ($c$) and the displacement of carbon atoms in the $c$ direction ($\langle {u}_{c}^{2}\rangle$), are no reliable quantities, as their values are expected to correlate strongly. Still, the significantly higher values of $\langle {u}_{c}^{2}\rangle$ (Fig. 6.4c) obtained for tannin (and lactose), as compared with ‘Carbores P’, and the consequent decrease of $\langle {u}_{c}^{2}\rangle$ at higher temperatures are consistent with the average molecular structure of the utilized binder materials. In contrast to ‘Carbores P’, the hexagonal rings containing carbon atoms are not planar in profisetinidin and in α-lactose [27, 28], but they flatten with increasing layer size ${L}_{a}$ during the carbonization process.

The results of the XRD analysis of calcined tannin, which are summarized in Fig. 6.4, can only be interpreted in such a way that no real turbostratic graphite is formed. The vertical size of the turbostratic graphite clusters having certain periodic ordering in the $c$ direction stays below 1 nm, which corresponds to 2–3 nearly parallel graphene-like layers. As lactose showed a strong inhibition of the graphitization process as well (see below), the impurities (Na, Ca, O and S) mentioned above cannot be the primary reason for the retarded graphitization of tannin. It seems more likely, that unfavorable cross-links between strongly tilted graphene-like layers prevent a further parallel arrangement up to 1400 °C or higher.

Lactose

The [H]/[C] ratio of 1.99 (Fig. 6.4i) determined using CGHE in the starting sample of α-lactose monohydrate (C₁₂H₂₄O₁₂) is in a good agreement with the expected value, [H]/[C] = 2. The annealing of C₁₂H₂₄O₁₂ leads to a rapid reduction of the [H]/[C] ratio and to the extensive mass loss (Fig. 6.4g), which are accompanied by the production of gaseous compounds (Fig. 6.4f). The first transformation step is the release of crystal water and the phase transition of α-lactose monohydrate to the α-lactose anhydride (C₁₂H₂₂O₁₁). This step is completed at 200 °C, where the [H]/[C] has decreased to 1.80 (Fig. 6.4i).

Upon further annealing, the melt is caramelized. The caramelization process is accompanied by condensation reactions between different hydroxyl groups, which lead to the formation of non-planar carbon layers containing hydrogen, oxygen and other point defects, which resemble small, cross-linked and highly perturbed PAHs [29]. This process is most pronounced around 270 °C and is completed around 400 °C. Consequently, the XRD signal typical for hexagonally coordinated carbon was observed only in the XRD patterns of the lactose samples, which were annealed at 400 °C and above. In samples annealed at the temperatures below 400 °C, no noteworthy amount of carbonized lactose was detected by XRD. The graphitization of lactose proceeds similarly to the graphitization of tannin. The release of hydrogen from the perturbed PAHs and the formation of graphene sheets are documented by the increase of the in-plane lattice parameter $a$ and by the increase of the layer size ${L}_{a}$, which start above 600 °C (Fig. 6.4a, d). In analogy with tannin, the formation of a turbostratic structure in carbonized lactose is strongly inhibited.

In comparison with ‘Carbores P’ and tannin, lactose shows the highest mass loss (~87%), which is accompanied by excessive foaming. These features facilitate formation of a highly porous carbon binder phase, which may negatively affect the mechanical properties of the synthesized filters. In this context, it should be mentioned that the liquid phase sintering in lactose happens at much lower temperatures (Fig. 6.4f) than, e.g., in coal tar pitches. Thus, it cannot contribute significantly to the improvement of the mechanical properties of sintered refractory composites.

Comparison of the Carbon Binders from the Microstructural Point of View

Traditionally, the coal tar pitch is used as a favored source of carbon for production of the carbon binders in thermoshock-resistant refractories. However, as coal tar contains many carcinogenic polycyclic aromatic hydrocarbons, alternatives are sought. Tannin and lactose are considered as possible substitutes. From the microstructural point of view, these alternative binders contain less carbon than coal tar. Additional elements form frequently gaseous phases, which leave the original compound and facilitate the formation of highly porous structures that may negatively affect the overall yield strength of the filter. The graphitization of tannin and lactose is strongly inhibited. As the coefficient of the thermal expansion of graphite is much smaller in the crystallographic direction $a$ than in the crystallographic direction $c$ [30], the lack of the parallel ordering of the graphene sheets and the formation of the graphite structure may improve the thermal shock resistance of the carbon binder.

Therefore, the replacement of certain amount of the coal tar pitch in refractory materials by alternative carbon binders is reasonable, because it reduces the amount of carcinogenic polycyclic aromatic hydrocarbons and improves the thermoshock resistance of the refractories. Still, some problems must be solved. One of them is the absence of a liquid-like phase in the binders with high tannin content. As the presence of a liquid phase usually facilitates the sintering process, the refractories containing binders with a high tannin content must be sintered for a longer time or at higher temperatures. Furthermore, the sources of tannin with a lower amount of impurities and additional phases should be preferred. A principal problem of the carbon binders produced from lactose is the foaming. However, it can be reduced by addition of ceramic powders or fully suppressed by annealing in vacuum.

6.3 Metastable Alumina Phases for Functional Filter Coatings

Metastable alumina phases are promising materials for functionalization of the metal melt filters, because they are supposed to have the same chemical composition and a similar crystal structure like the endogenous inclusions, which form during the casting of steels and aluminum alloys [5]. In general, the alumina phases can be divided in two groups according to their oxygen sublattice [31]. The first group including the thermodynamically stable corundum (α-Al₂O₃) possesses an approximately hexagonal close packed (h.c.p.) sublattice. The second group comprises metastable alumina phases, which crystallize with an approximately cubic close packed (c.c.p.) sublattice.

6.3.1 Structure of γ-Al₂O₃

One of the most popular metastable aluminum oxides is γ-Al₂O₃. Although this phase was extensively investigated in the past, there are still discussions about its crystal structure [32‐36], which are motivated mainly by the presence of crystal structure defects [37‐39]. In an ideal spinel having the space group (SG)$Fd\overline{3 }m$, the anions occupy the Wyckoff positions $32e$, the trivalent cations the octahedral sites $16d$ and the divalent cations the sites $8a$ [40]. In γ-Al₂O₃, the Wyckoff positions $8a$ have also to be occupied by trivalent cations (Fig. 6.5). To preserve the charge neutrality and the cation to anion ratio of 2/3, $2.\overline{6 }$ of the 24 cations sites must remain vacant.

A typical feature of the highly defective crystal structure of γ-Al₂O₃ is a highly anisotropic broadening of diffraction spots and lines, which is observed in selected area electron diffraction (SAED) and XRD patterns (Fig. 6.6). Diffraction lines with $h/2$, $k/2$ and $l/2$ being all even or odd, e.g., 004, 222, 440 and 444, which stem primarily from the fully occupied c.c.p. anion sublattice (Wyckoff sites $32e$), are much less broadened than the other diffraction lines, e.g., 111, 113, 220, 224, 331, 333 and 115, to which solely the scattering on the cation sublattice contributes. This anisotropic line broadening was explained by the presence of non-conservative antiphase boundaries (APBs) having the domain boundaries $\left(00l\right)$ and the domain shift $\frac{1}{4}\langle 10\overline{1 }\rangle$ (Fig. 6.7a). These particular defects keep the c.c.p. oxygen sublattice intact but introduce a disorder on cation sublattice [41].

The kind of the APBs and in particular the shift vector were determined using the phase shift factor

$${A}_{{\varvec{h}}}\left({\varvec{r}}\right)=\exp\left(2\mathrm{\pi i} \, {\varvec{h}}\cdot {\varvec{r}}\right),$$

(6.5)

which must be different from unity for heavily broadened reflections [41]. In Eq. (6.5), ${\varvec{r}}$ is the shift vector and ${\varvec{h}}$ the vector containing the diffraction indices $hkl$. The shift factors calculated for the domain boundary $\left(00l\right)$ and the non-conservative domain shift along the crystallographically equivalent directions $\frac{1}{4}\langle 10\overline{1 }\rangle$ are summarized in Table 6.1. One can see that the reflections 222, 400 and 440 are not broadened by these APBs, while the broadening of the reflections 220, 311 and 333 depends on the shift direction. Furthermore, the reflection 220 is broadened for more equivalent shift directions than the reflection 311 and only in case of non-conservative APBs, which agrees very well with the observation (Fig. 6.6a).

Table 6.1

Phase shift factors [Eq. (6.5)] for APBs on the lattice planes $(00l)$. The reflections are broadened, when the phase shift factor is equal to –1. Corresponding reflections stem solely from the cation sublattice and are highlighted in bold. Other reflections originate from the c.c.p. sublattice. Adopted from [41]

	Conservative $(00l)\frac{1}{4}\langle 110\rangle$		Non-conservative $(00l)\frac{1}{4}\langle 10\overline{1 }\rangle$
${\varvec{h}}{\varvec{k}}{\varvec{l}}$	$\frac{1}{4}[110]$	$\frac{1}{4}[\overline{1 }10]$	$\frac{1}{4}[10\overline{1 }]$	$\frac{1}{4}[\overline{1 }0\overline{1 }]$	$\frac{1}{4}[01\overline{1 }]$	$\frac{1}{4}[0\overline{1 }\overline{1 }]$
220	1	1	−1	−1	−1	−1
311	1	−1	−1	1	1	−1
222	1	1	1	1	1	1
400	1	1	1	1	1	1
333	−1	1	1	−1	1	−1
440	1	1	1	1	1	1

For simulation of the SAED and XRD patterns, the JEMS routine [42] and the Debye equation [Eq. (6.2)] were employed, respectively. The atomic positions were first generated for undisturbed cubic nanocrystallites terminated by the lattice planes $\{100\}$ that had the size of 11 nm. The lattice parameter of γ-Al₂O₃ was 7.942(5) Å. In order to simulate APBs, parts of the nanocrystallites were shifted along the respective shift vector (Fig. 6.8 and Table 6.1) [41]. Near the non-conservative APBs, this shift produces a certain amount of unfavorable nearest neighbors (Fig. 6.7a). Thus, the required vacancies preserving the charge neutrality were located in the vicinity of these APBs – either on the corresponding octahedral or on the tetrahedral sites. The SAED patterns simulated for undisturbed γ-Al₂O₃ and for γ-Al₂O₃ with APBs $\left(00l\right)\frac{1}{4}\langle 110\rangle$ and $\left(00l\right)\frac{1}{4}\langle 10\overline{1 }\rangle$ [38, 43] are shown in Fig. 6.8. In addition to the phase shift factors (Table 6.1), the comparison of the SAED simulations with the measured SAED pattern from Fig. 6.6a confirms that only the non-conservative APBs $\left(00l\right)\frac{1}{4}\langle 10\overline{1 }\rangle$ cause the observed broadening of the diffraction spot 220.

The corresponding XRD patterns for the non-conservative APBs (Fig. 6.7a) on the lattice planes $(00l)$ (green) and on all equivalent planes (red) are depicted in Fig. 6.6b. This comparison reveals that the APBs have to be located on all equivalent planes forming a complex 3D structure of cuboidal nanodomains as depicted in Fig. 6.7b. In order to reduce the long computing time, which is typically required when the Debye equation [Eq. (6.2)] is applied for calculation of the XRD patterns from atomic clusters containing around 1 million atoms, the routine ‘cuDebye’ was developed [44]. The utilization of 2816 parallel operating CUDA cores (GTX 980 TI) in combination with atomic operations, which make storage and post processing of the calculated distances for a frequency evaluation obsolete, reduce the overall computation time to less than 5 min. This is a consequence of the Debye equation [Eq. (6.2)] as the prevention of reoccurring summands significantly increase the calculation speed. Additional line broadening stems from the nanocrystalline character of γ-Al₂O₃ under study, which had a crystallite size of about 11 nm. The crystallite size was determined from the width of the diffraction lines related to the oxygen sublattice. The shoulder of the diffraction line 400 (Fig. 6.6b) indicates a slight tetragonal distortion of the cubic elementary cell. The corresponding $c/a$ ratio was 0.985. This tetragonal distortion is usually described by the SG $I{4}_{1}/amd$ instead of $Fd\overline{3 }m$ [41, 45].

6.3.2 Thermally Induced Phase Transformation of γ-Al₂O₃

At higher temperatures, metastable γ-Al₂O₃ transforms to the thermodynamically stable corundum (α-Al₂O₃) following the phase transformation path [31]:

$$\upgamma \to\updelta \to\uptheta \to \mathrm{\alpha }.$$

(6.6)

Although this transformation path is generally accepted, there are still some uncertainties regarding the transformation process and individual transition temperatures. The main issues are related to the formation of the intermediate phases δ-Al₂O₃ and θ-Al₂O₃ [46‐49] and to their defect crystal structure [38, 39, 41, 50], as it is assumed that the phase transitions $\upgamma \to\updelta \to\uptheta$ lead primarily to a redistribution and possibly to an ordering of crystal structure defects in Al₂O₃. In this context, γ-Al₂O₃ is described as a cubic ($Fd\overline{3 }m$) or a tetragonal ($I{4}_{1}/amd$) spinel [45, 51] and θ-Al₂O₃ as a monoclinic ($C2/m$) crystal structure [52]. The orientation relations between these structures are depicted in Fig. 6.9. The δ-Al₂O₃ phase possesses an intermediate crystal structure between γ-Al₂O₃ and θ-Al₂O₃, which cannot be described unambiguously from the crystallographic point of view [41, 47, 50, 53]. Initially, δ-Al₂O₃ was described as a supercell consisting of three cubic γ-Al₂O₃ unit cells stacked in the $c$ direction with ${a}_{\delta }={a}_{\upgamma }$ and ${c}_{\updelta }=3{a}_{\upgamma }$ [54]. Based on the tetragonal description of γ-Al₂O₃, δ-Al₂O₃ exhibits the space group $P\overline{4 }m2$ [49, 55] with ${a}_{\delta }={a}_{\upgamma }/\sqrt{2}$ and ${c}_{\updelta }=3{a}_{\upgamma }$. In both cases, the orientation relation towards the c.c.p. sublattice remains the same as in the parent γ-Al₂O₃ unit cell. Since the initial tetragonal distortion is quite small (cf. Sect. 6.3.1), the cubic model for γ-Al₂O₃ ($Fd\overline{3 }m$) is utilized for simplicity in the following text.

Identification of δ-Al₂O₃

A complementary technique used for the phase identification in defect-rich alumina phases is the nuclear magnetic resonance spectroscopy (NMR) (frequently in conjunction with the magic-angle spinning, MAS), as it is sensitive to the local coordination of the ²⁷Al cations [56, 57]. As pointed out in Sect. 6.3.1, the crystal structure of γ-Al₂O₃ ($Fd\overline{3 }m$) requires $8/3 = 2.\overline{6}$ cation vacancies per unit cell, which must be distributed over the tetrahedral ($8a$) and octahedral ($16d$) spinel sites. When all vacancies are located on the eight tetrahedral sites, then $\left(8/3\right)/8=1/3$ of these sites stay vacant. Thus, $1-1/3=2/3$ of the tetrahedral sites are occupied by Al³⁺ ions, while all octahedral positions are fully occupied by Al³⁺. As for γ-Al₂O₃ the NMR resolves the relative site occupancy and not the occupancy of individual sites, the minimum fraction of the tetrahedrally (AlO₄) coordinated Al³⁺ ions in γ/δ-Al₂O₃ is equal to$\left(2/3\times 8\right)/\left(2/3\times 8+1\times 16\right)=25\%$. When all vacancies are located on the sixteen octahedral sites, then $1-1/6=5/6$ of these sites are occupied by Al³⁺. Thus, the maximum fraction of the tetrahedrally (AlO₄) coordinated Al³⁺ ions in γ/δ-Al₂O₃ is$\left(1\times 8\right)/\left(1\times 8+5/6\times 16\right)=37.5\%$, s. Table 6.2. In ideal θ-Al₂O₃ (SG $C2/m$, structure type β-Ga₂O₃), which has the same number of tetrahedral and octahedral positions (both sites are $4i$), the fraction of tetrahedrally (AlO₄) coordinated Al³⁺ ions seen by NMR must be 50% [52]. The remaining 50% of the Al³⁺ ions are located at the octahedral sites. Therefore, during the phase transformation γ-Al₂O₃ → θ-Al₂O₃, at least 12.5% (= 50–37.5%) of the Al³⁺ ions in the parent γ-Al₂O₃ phase must move from the octahedral to the tetrahedral interstitial (non-spinel) sites. Due to the structural similarity of δ-Al₂O₃ with γ-Al₂O₃, only the intermediate states having the fraction of tetrahedrally coordinated cations below or equal to 37.5% (and showing superstructure reflections in the XRD pattern) are classified as δ-Al₂O₃ [53].

Table 6.2

Fractions of tetrahedrally (AlO₄) and octahedrally (AlO₆) coordinated Al³⁺ ions determined by ²⁷Al MAS NMR as function of the annealing temperature $(T)$. The corresponding amounts of vacancies per unit cell residing on the Wyckoff sites $8a$ and $16d$ of γ-Al₂O₃ that are summarized on the right-hand side of the table satisfy the equations

$\left[{{\text{AlO}}}_{4}\right]=\left(8-{N}_{8a}^{{\text{v}}}\right)/\left(24-{N}_{8a}^{{\text{v}}}-{N}_{16d}^{{\text{v}}}\right); \,{N}_{8a}^{{\text{v}}}+{N}_{16d}^{{\text{v}}}=8/3$
${\varvec{T}}$[°C]	AlO₄ [%]	AlO₆ [%]	${{\varvec{N}}}_{8{\varvec{a}}}^{\mathbf{v}}$	${{\varvec{N}}}_{16{\varvec{d}}}^{\mathbf{v}}$
Ideal	37.5	62.5	0.00	2.67
500	33.7	66.3	0.81	1.86
600	34.7	65.3	0.60	2.07
700	35.8	64.2	0.36	2.31
800	37.3	62.7	0.04	2.63
900	37.3	62.7	0.04	2.63

Transformation of γ-Al₂O₃ to δ-Al₂O₃

The relative distributions of the Al³⁺ cations and the respective vacancies over the Wyckoff sites $8a$ and $16d$ of γ-Al₂O₃ that were determined from the ²⁷MAS NMR data are summarized in Table 6.2. As expected, the amount of vacancies on octahedral sites ($16d$) is significantly larger than the amount of vacancies on tetrahedral sites ($8a$). Still, about 30% of the vacancies are located on a tetrahedral site ($8a$) in the initial stage of the γ/δ-Al₂O₃ formation (at 500 °C). With increasing annealing temperature, some octahedral cations (AlO₆) migrate to the tetrahedral sites (AlO₄). When the AlO₄ fraction approaches 37.5% (around 800 °C), almost all vacancies are located on the octahedral sites $16d$ [58‐60]. As the measured fraction of the tetrahedrally coordinated Al³⁺ (Table 6.2) does not exceed 37.5%, it can be concluded that no θ-Al₂O₃ has formed up to 900 °C. Still, the diffraction patterns in Fig. 6.10 showing superstructure reflections were labeled as θ-Al₂O₃ because of the unambiguous structure of δ-Al₂O₃ [50, 53].

According to the SAED patterns (Fig. 6.10a, b), the $\upgamma \to\updelta$ transformation starts at about 700 °C. At 800 °C, a partially ordered superstructure forms. The superstructure reflections become much more pronounced above 800 °C. The average distance between these satellite reflections ($\Delta ={c}_{\delta }^{*}=0.42 \, {{\text{nm}}}^{-1}$, Fig. 6.10b) indicates the formation of a supercell with the lattice parameter ${c}_{\delta }$ of approx. 23.8 Å. In the c direction, this elementary cell is approximately tripled in comparison with γ-Al₂O₃ $Fd\overline{3 }m$) having the lattice parameter ${a}_{\gamma }$ = 7.942 Å, which is consistent with the initially published structures of δ-Al₂O₃ [49, 54, 55].

In γ-Al₂O₃, the APBs are distributed more or less randomly. The vacancies are preferentially located near the planar defects, which prevents an unfavorable proximity of Al³⁺ cations. With increasing temperature, the planar defects start to rearrange, and order periodically. This process is facilitated by the migration of Al³⁺ cations, which results in the shift of the vacancies from the tetrahedral sites to the octahedral sites that are close to APBs (cf. Fig. 6.7a). A perfect ordering of the APBs (and thus a perfect ordering of the vacancies) would resemble the formation of ideal (tetragonal) δ-Al₂O₃, which could be distinguished from γ-Al₂O₃ and θ-Al₂O₃ clearly. A simplified example of such ordering is schematically depicted in the inset of Fig. 6.10b. This model shows two elementary cells of δ-Al₂O₃ (red). Each δ-Al₂O₃ supercell has the lattice parameter of c ≈ 24 Å, consists of three elementary cells of γ-Al₂O₃ (blue) with the lattice parameter a_γ ≈ 8 Å and contains two equidistantly spaced APBs (gray). The distance of the APBs (≈ 12 Å) is close to the distance between APBs, which was determined from the pronounced anisotropy of the XRD line broadening [41]. However, the probability that such an ideal state is observed in an XRD (or SAED) experiment is low, because the same periodic rearrangement of APBs in all diffracting nanocrystallites is rather improbable. Furthermore, the equivalent shift vectors (Table 6.1) and the fact that the APBs can be also found on the equivalent lattice planes (cf. Figs. 6.6b and 6.7b) suggest a more complex arrangement of APBs than it is depicted in the inset (Fig. 6.10b). The variety of the APB arrangements is also a reason for the large number of structural approximations of δ-Al₂O₃ that can be found in literature [48, 49, 53‐55].

The ongoing ordering of APBs and vacancies leads to a more pronounced tetragonal distortion of the c.c.p. oxygen sublattice. This can be seen on the further splitting of the diffraction line 004_γ with increasing annealing time and on the appearance of the diffraction lines $11\overline{2 }$ _θ and 600_θ at the temperatures around 900 °C. The latter does not imply the existence of θ-Al₂O₃ as an additional phase to γ/δ-Al₂O₃ (see Sect. “Identification of δ-Al₂O₃” and Fig. 6.10).

The above findings are consistent with the results of the multi-quantum magic-angle spinning (MQMAS) NMR, which show clearly the splitting and sharpening of the AlO₄ peak for γ-Al₂O₃ calcinated at 900 °C (Fig. 6.11) [56, 61]. The presence of at least two sharp AlO₄ peaks can be associated with the presence of differently coordinated Al³⁺ cations on the tetrahedral sites. Unlike in an ideal spinel structure, where the Al³⁺ coordination is always the same, in a structure with partially ordered vacancies, the Al³⁺ cations located in near APBs can have different coordination than the Al³⁺ cations, which are far from these boundaries. This can also apply for Al³⁺ cations in nanocrystallites with a more pronounced tetragonal distortion of the c.c.p. anion sublattice.

6.3.3 Stabilization of γ-Al₂O₃

As illustrated above, γ-Al₂O₃ is a metastable phase, which undergoes a phase transition to thermodynamically stable corundum at high temperatures (Fig. 6.10c). The transition temperature is typically between 1000 and 1200 °C [54], but it depends strongly on the kind and on the microstructure of the starting material. For special technical applications, an extension of the temperature stability range of γ-Al₂O₃ is desired. As shown in [50], the use of a reducing environment, e.g., the presence of a carbon binder in the ceramic foam filters, has a positive effect on the improvement of the thermal stability of the metastable alumina phases. It postpones the formation of α-Al₂O₃ of about 50 K. Another approach, which can be applied to stabilize γ-Al₂O₃ to higher temperatures, is based on the inhibition of the migration of Al cations and thus on the hindering of the vacancy ordering. The migration of the Al cations and the ordering of the cation vacancies, which are the fundamental mechanisms accompanying the phase transformations in alumina, can be hindered, when Al₂O₃ is doped with elements having a different valence and coordination. One of the most promising dopants is silicon [62].

In order to quantify the impact of the doping of Al₂O₃ by silicon on the thermal stability of the metastable alumina phases, the nanocrystalline boehmite, which was used as a starting material also in this part of the study, was synthesized using a sol–gel process [63]. In the doped sample, approx. 5% of the Al³⁺ cations were replaced by Si⁴⁺ cations through the addition of tetraethyl orthosilicate (TEOS) [64]. The differential thermal analysis (DTA) (Fig. 6.12) revealed that the formation of the thermodynamically stable α-Al₂O₃ in the doped sample is shifted to a higher temperature (approx. 180 °C) as compared with the non-doped sample. The first endothermal peak in the DTA curve at approx. 150 °C (Fig. 6.12) corresponds to the loss of adsorbed water, whereas the second endothermal peak at approx. 450 °C indicates the structural collapse of boehmite and the subsequent formation of γ-Al₂O₃ [50, 65].

The XRD measurements confirmed the same phase transformation path (boehmite → γ-Al₂O₃ → δ/θ-Al₂O₃ → α-Al₂O₃) for non-doped and Si-doped samples (Fig. 6.13a). However, these measurements have shown that in the doped sample the phase transformations γ-Al₂O₃ → δ/θ-Al₂O₃ and δ/θ-Al₂O₃ → α-Al₂O₃ occur at a higher temperature than in the non-doped one. The results of the XRD analysis were complemented by the results of the ²⁷Al MAS NMR analysis (Fig. 6.13b), which revealed that the fraction of tetrahedrally coordinated Al³⁺ ions (AlO₄) increased and the fraction of octahedrally coordinated Al³⁺ ions (AlO₆) decreased after annealing at higher temperatures, i.e., when γ-Al₂O₃ transformed to δ-Al₂O₃ (see Sect. “Transformation of γ-Al₂O₃ to δ-Al₂O₃”).

The AlO₅ coordination visible in Fig. 6.13b is related to the surface states, which are typically reported for nanocrystalline materials [61]. In the samples synthesized using the sol–gel process, the sizes of the γ-Al₂O₃ crystallites were 3.1 nm (in the doped sample) and 4.3 nm (in the non-doped sample), as estimated using the Scherrer equation [67] from the integral width of the diffraction line 400_γ. The net relative fractions of the AlO₄ and AlO₆ coordinations that were calculated by neglecting the AlO₅ fraction are summarized in Table 6.3. The fractions of tetrahedrally coordinated Al³⁺ ions, which are higher than the maximum (ideal) values calculated for γ/δ-Al₂O₃, indicate the presence of a mixture of δ-Al₂O₃ and θ-Al₂O₃. The second part of Table 6.3 displays the concentrations of corresponding vacancies that are located on the Wyckoff sites $8a$ and $16d$ of γ-Al₂O₃. Negative concentrations of $8a$ vacancies refer to the cations (Al³⁺ and/or Si⁴⁺) located on the interstitial (tetrahedral non-spinel) sites.

Table 6.3

Fractions of the tetrahedrally (AlO₄) and octahedrally (AlO₆) coordinated Al³⁺ ions determined using the ²⁷Al MAS NMR (cf. Fig. 6.13b and Table 6.2). The non-highlighted values were calculated assuming that Si⁴⁺ ions occupy the spinel sites $8a$ (${x}_{8a}^{{\text{Si}}}=1$). The values printed in bold were calculated for Si⁴⁺ located on the tetrahedral non-spinel sites $8b$ and $48f$ [41] (${x}_{8a}^{{\text{Si}}}=0$). The concentrations of corresponding vacancies in γ-Al₂O₃ that are located on the Wyckoff sites $8a$ and $16d$ (second part of the table) were calculated analogously to the vacancy concentrations given in Table 6.2. In the Si-doped sample, the relationship between the amount of vacancies and the fraction of tetrahedrally coordinated Al³⁺ ions takes the form

$\left[{{\text{AlO}}}_{4}\right]=\left(8-{N}_{8a}^{{\text{v}}}-{x}_{8a}^{{\text{Si}}}{N}_{\upgamma }^{{\text{Si}}}\right)/\left(24-{N}_{8a}^{{\text{v}}}-{N}_{16d}^{{\text{v}}}-{x}_{8a}^{{\text{Si}}}{N}_{\upgamma }^{{\text{Si}}}\right); \, {N}_{8a}^{{\text{v}}}+{N}_{16d}^{{\text{v}}}=24-{N}_{\upgamma }^{{\text{Al}}}-{x}_{8a}^{{\text{Si}}}{N}_{\upgamma }^{{\text{Si}}}$
${{\varvec{x}}}_{\mathbf{S}\mathbf{i}}$	0%	5%	0%	5%	0%	5%	0%	5%
${\varvec{T}}$[°C]	AlO₄ ${x}_{8a}^{{\text{Si}}}=1/0$ [%]		AlO₆ ${x}_{8a}^{{\text{Si}}}=1/0$ [%]		${{\varvec{N}}}_{8{\varvec{a}}}^{\mathbf{v}}$		${{\varvec{N}}}_{16{\varvec{d}}}^{\mathbf{v}}$
Ideal	37.5	34.9/40.1	62.5	65.1/59.9	0.00	0.00/0.00	2.67	3.02/4.06
600	32.1	32.8 36.1	67.9	67.2	1.16	0.42/1.47	1.51	2.60
1000	39.1		60.9	63.9	−0.34	−0.25/0.80	3.01	3.27
1250	0.3	38.3	99.7	61.7		−0.68/0.37		3.70

After annealing at 1000 °C, the non-doped Al₂O₃ contained θ-Al₂O₃ domains. This can be seen in the XRD pattern (Fig. 6.13a) and concluded from the fraction of the tetrahedrally coordinated Al³⁺ ions, which exceeds 37.5% (AlO₄ in Table 6.3). Consequently, the amount of the $8a$ vacancies becomes negative. Still, no significant amount of α-Al₂O₃ has formed at this temperature. First at 1225 °C, Al₂O₃ in the non-doped sample transformed almost completely to corundum (α-Al₂O₃). As the Al³⁺ ions in α-Al₂O₃ are octahedrally coordinated, the NMR recognized the AlO₆ coordination only (Table 6.3, 1250 °C).

In the Si-doped γ-Al₂O₃, the positions of the Si⁴⁺ ions must be clarified first. The ²⁹Si MAS NMR revealed that Si⁴⁺ ions are located exclusively on a tetrahedral site ($8a$, $8b$ or $48f$ [41]) and that they are surrounded by the Al³⁺ cations (as second nearest neighbors), which is in a good agreement with the results of Mardkhe et al. [64]. If the Si⁴⁺ ions occupy the spinel sites $8a$ (${x}_{8a}^{{\text{Si}}}=1$), then the maximum fraction of AlO₄ is 34.9% (non-highlighted values in Table 6.3). If the Si⁴⁺ ions are located on the non-spinel sites $8b$ or $48f$ (${x}_{8a}^{{\text{Si}}}=0$), then the maximum fraction of AlO₄ is 40.1% (highlighted values in Table 6.3). Both fractions were calculated for 5 at% Si, i.e., for one Si⁴⁺ ion (${N}_{\upgamma }^{{\text{Si}}}=1$) allotted to twenty Al³⁺ ions (${N}_{\upgamma }^{{\text{Al}}}=20$). If one Si⁴⁺ ion replaces one Al³⁺ ion on the Wyckoff sites $8a$, the charge neutrality is achieved when the unit cell of γ-Al₂O₃ contains three vacancies (${N}_{8a}^{{\text{v}}}+{N}_{16d}^{{\text{v}}}=9/3$) instead of $8/3=2.\overline{6 }$ for non-doped γ-Al₂O₃ (s. Sect. 6.3.1). The main consequence of the Si⁴⁺ distribution over the spinel and non-spinel sites is that the formation of θ-Al₂O₃ can be concluded without doubt only when the amount of tetrahedrally coordinated Al³⁺ exceeds 40.1%, which was observed in none of the Si-doped samples under study.

The negative amounts of the cation vacancies sitting on the sites $8a$ were only obtained assuming that all Si⁴⁺ ions are located at the sites $8a$ (non-highlighted values in Table 6.3 for the samples annealed at 1000 °C and above). In this case, the fraction of the tetrahedrally coordinated Al³⁺ ions measured using ²⁷Al MAS NMR exceeds the predicted ideal value of 34.9%. However, the XRD pattern of the Si-doped sample annealed at 1000 °C (Fig. 6.13a) did not show any indication of the θ-Al₂O₃ formation, which would produce pronounced superstructure satellites like in Fig. 6.10c above 900 °C. As the negative amount of the cation vacancies at the $8a$ sites is related to the formation of θ-Al₂O₃ (s. Sect. “Identification of δ-Al₂O₃”), the non-existence of θ-Al₂O₃ in the XRD pattern reveals that the Si⁴⁺ cations are located also on the interstitial (non-spinel) sites up to 1225 °C. This may hinder the migration of Al³⁺ cations to the non-spinel sites, which is required for the phase transformation of γ/δ-Al₂O₃ to θ-Al₂O₃. In contrast to the Si-doped sample, the ²⁷Al MAS NMR performed on the non-doped sample revealed that the Si-free Al₂O₃ transforms partially to θ-Al₂O₃ already at 1000 °C.

In the Si-doped specimen, also the formation of α-Al₂O₃ is shifted to a higher temperature. At 1225 °C, almost no α-Al₂O₃ was identified by XRD (Fig. 6.13a). As a dominant phase, α-Al₂O₃ was first observed in the doped sample that was annealed at 1400 °C (Fig. 6.13a). This specimen contained mullite (2Al₂O₃⋅SiO₂/3Al₂O₃⋅ 2SiO₂, SG $Pbam$ [68]) as a minor phase that formed during the reaction of alumina with added silicon. The delay of the α-Al₂O₃ formation has primarily kinetic reasons. As the solubility of silicon in corundum is extremely low [69], Si must leave the parent metastable phase (δ/θ-Al₂O₃), before this phase can transform to α-Al₂O₃. This process requires an additional migration or diffusion of silicon, which shifts the formation of α-Al₂O₃ to a higher temperature.

6.4 Conclusions

This chapter was devoted to the description of temperature-induced changes in the microstructure of carbon binders produced from coal tar pitch, tannin and lactose, and in the microstructure of alumina phases produced by calcination of boehmite. For the crystal structure and microstructure characterizations of the carbon binders, a fast computer routine based on the Debye equation was developed. This routine reveals the essential parameters of the turbostratic graphite, namely the lattice parameters a and c, the size of the graphite blocks in the respective direction and the degree of the atomic disorder. It was shown, how these parameters can be used for description of the kinetics of the graphitization process in different carbon-containing materials. For description of the phase transformations in alumina phases, atomistic models were developed that served as a basis for the explanation of the transformation mechanisms and as a basis for the systematic stabilization of the metastable phases through the obstruction of the transformation mechanisms.

Acknowledgements

The authors thank Zdeněk Matěj for the implementation of the Warren-Bodenstein approach into the MStruct routine, Claudia Voigt, Benjamin Bock, Diane Hübgen, Katrin Becker, Astrid Leuteritz and Beate Wahl for sample preparation or measurements as well as Christian Schimpf, Marius Wetzel, Hanka Becker and Andreas Leineweber for fruitful discussions. Thomas Hammer and Wiebke Hadwich supported this work by their student projects. The German Research Foundation (DFG) is acknowledged for funding the Collaborative Research Center 920 (project number 169148856) in the frame of subproject A05.

Open Access This chapter is licensed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

The images or other third party material in this chapter are included in the chapter's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the chapter's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

Vorheriges Kapitel Structural and Thermodynamic Properties of Filter Materials: A Raman and DFT Investigation

Nächstes Kapitel Interface Reactions Between the Metal Melt and the Filter Surface Activated by a Spark Plasma Sintering Process

A. Schmidt, A. Salomon, S. Dudczig, H. Berek, D. Rafaja, C.G. Aneziris, Adv. Eng. Mater. 19, 1700170 (2017). https://doi.org/10.1002/adem.201700170CrossRef

M. Seleznev, S. Henschel, E. Storti, C.G. Aneziris, L. Krüger, A. Weidner, H. Biermann, Adv. Eng. Mater. 22, 1900540 (2020). https://doi.org/10.1002/adem.201900540CrossRef

R. Wagner, A. Schmiedel, S. Dudczig, C.G. Aneziris, O. Volkova, H. Biermann, A. Weidner, Adv. Eng. Mater. 24, 2100640 (2022). https://doi.org/10.1002/adem.202100640CrossRef

C.G. Aneziris, C. Schroeder, M. Emmel, G. Schmidt, H.P. Heller, H. Berek, Metall. Mater. Trans. B 44, 954 (2013). https://doi.org/10.1007/s11663-013-9828-6CrossRef

K. Wasai, K. Mukai, A. Miyanaga, ISIJ Int. 42, 459 (2002). https://doi.org/10.2355/isijinternational.42.459CrossRef

B. Bock, A. Schmidt, E. Sniezek, S. Dudczig, G. Schmidt, J. Szczerba, C.G. Aneziris, Ceram. Int. 45, 4499 (2019). https://doi.org/10.1016/j.ceramint.2018.11.131CrossRef

M. Emmel, C.G. Aneziris, G. Schmidt, D. Krewerth, H. Biermann, Adv. Eng. Mater. 15, 1188 (2013). https://doi.org/10.1002/adem.201300118CrossRef

T. Wetzig, A. Baaske, S. Karrasch, N. Brachhold, M. Rudolph, C.G. Aneziris, Ceram. Int. 44, 23024 (2018). https://doi.org/10.1016/j.ceramint.2018.09.105CrossRef

S. Dudczig, C.G. Aneziris, M. Emmel, G. Schmidt, J. Hubalkova, H. Berek, Ceram. Int. 40, 16727 (2014). https://doi.org/10.1016/j.ceramint.2014.08.038CrossRef

10.

T. Zienert, S. Dudczig, O. Fabrichnaya, C.G. Aneziris, Ceram. Int. 41, 2089 (2015). https://doi.org/10.1016/j.ceramint.2014.08.038CrossRef

11.

A. Salomon, M. Motylenko, D. Rafaja, Adv. Eng. Mater. 24, 2100690 (2022). https://doi.org/10.1002/adem.202100690CrossRef

12.

M. Emmel, C.G. Aneziris, Ceram. Int. 38, 5165 (2012). https://doi.org/10.1016/j.ceramint.2012.03.022CrossRef

13.

C. Himcinschi, C. Biermann, E. Storti, B. Dietrich, G. Wolf, J. Kortus, C.G. Aneziris, J. Eur. Ceram. Soc. 38, 5580 (2018). https://doi.org/10.1016/j.jeurceramsoc.2018.08.029CrossRef

14.

L.V. Azároff, Elements of X-Ray Crystallography. (McGraw-Hill, New York, 1968), isbn: 9780070026674

15.

B.E. Warren, X-Ray Diffraction. (Dover Publications, New York, 1990), isbn: 9780486663173

16.

P. Debye, Ann. Phys. 351, 809 (1915). https://doi.org/10.1002/andp.19153510606CrossRef

17.

M. Dopita, M. Rudolph, A. Salomon, M. Emmel, C.G. Aneziris, D. Rafaja, Adv. Eng. Mater. 15, 1280 (2013). https://doi.org/10.1002/adem.201300157CrossRef

18.

M. Dopita, M. Emmel, A. Salomon, M. Rudolph, Z. Matěj, C.G. Aneziris, D. Rafaja, Carbon 81, 272 (2015). https://doi.org/10.1016/j.carbon.2014.09.058CrossRef

19.

B.E. Warren, P. Bodenstein, Acta Cryst. 18, 282 (1965). https://doi.org/10.1107/S0365110X65000609CrossRef

20.

B.E. Warren, P. Bodenstein, Acta Cryst. 20, 602 (1966). https://doi.org/10.1107/S0365110X66001464CrossRef

21.

Z. Matěj, A. Kadlecová, M. Janeček, L. Matějová, M. Dopita, R. Kužel, Powder Diffr. 29, S35 (2014). https://doi.org/10.1017/S0885715614000852CrossRef

22.

S. Brehm, C. Himcinschi, J. Kraus, B. Bock-Seefeld, C. Aneziris, J. Kortus, Adv. Eng. Mater. 24, 2100544 (2022). https://doi.org/10.1002/adem.202100544CrossRef

23.

I. Mochida, S.-H. Yoon, W. Qiao, J. Braz. Chem. Soc. 17, 1059 (2006). https://doi.org/10.1590/S0103-50532006000600002CrossRef

24.

C.A. Rottmaier, Dissertation, Universität Erlangen-Nürnberg, 2007, urn:urn:nbn:de:bvb:29-opus-11781

25.

P. Trucano, R. Chen, Nature 258, 136 (1975). https://doi.org/10.1038/258136a0CrossRef

26.

Z. Sebestyén, E. Jakab, E. Badea, E. Barta-Rajnai, C. Şendrea, Zs. Czégény, J. Anal. Appl. Pyrolysis 138, 178 (2019). https://doi.org/10.1016/j.jaap.2018.12.022

27.

National Library of Medicine (PubChem). cid: 101373967, https://pubchem.ncbi.nlm.nih.gov. Accessed 1 March 2022

28.

National Library of Medicine (PubChem), cid: 84571. https://pubchem.ncbi.nlm.nih.gov. Accessed 1 March 2022

29.

I.F. Myronyuk, V.I. Mandzyuk, V.M. Sachko, V.M. Gun’ko, Nanoscale Res. Lett. 11, 508 (2016). https://doi.org/10.1186/s11671-016-1723-z

30.

D.K.L. Tsang, B.J. Marsden, S.L. Fok, G. Hall, Carbon 43, 2902 (2005). https://doi.org/10.1016/j.carbon.2005.06.009CrossRef

31.

P. Euzen, P. Raybaud, X. Krokidis, H. Toulhoat, J.-L. Le Loarer, J.-P. Jolivet, C. Froidefond, in Handbook of Porous Solids, ed. by F. Schüth, K. S. W. Sing, J. Weitkamp (Wiley-VCH Verlag GmbH, Weinheim, 2002), p. 1591.https://doi.org/10.1002/9783527618286.ch23b

32.

M. Digne, P. Raybaud, P. Sautet, B. Rebours, H. Toulhoat, J. Phys. Chem. B 110, 20719 (2006). https://doi.org/10.1021/jp061466sCrossRefPubMed

33.

A.E. Nelson, M. Sun, J. Adjaye, J. Phys. Chem. B 110, 20724 (2006). https://doi.org/10.1021/jp0616720CrossRef

34.

G. Paglia, C.E. Buckley, A.L. Rohl, J. Phys. Chem. B 110, 20721 (2006). https://doi.org/10.1021/jp061648mCrossRefPubMed

35.

H.P. Pinto, R.M. Nieminen, S.D. Elliott, Phys. Rev. B 70, 125402 (2004). https://doi.org/10.1103/PhysRevB.70.125402CrossRef

36.

M. Sun, A.E. Nelson, J. Adjaye, J. Phys. Chem. B 110, 2310 (2006). https://doi.org/10.1021/jp056465zCrossRefPubMed

37.

J.M. Cowley, Acta Cryst. 6, 53 (1953). https://doi.org/10.1107/S0365110X53000119CrossRef

38.

A. Dauger, D. Fargeot, Radiat. Eff. 74, 279 (1983). https://doi.org/10.1080/00337578308218421CrossRef

39.

V.P. Pakharukova, D.A. Yatsenko, E. Yu., Gerasimov, A.S. Shalygin, O.N. Martyanov, S.V. Tsybulya, J. Solid State Chem. 246, 284 (2017). https://doi.org/10.1016/j.jssc.2016.11.032

40.

K.E. Sickafus, J.M. Wills, N.W. Grimes, J. Am. Ceram. Soc. 82, 3279 (1999). https://doi.org/10.1111/j.1151-2916.1999.tb02241.xCrossRef

41.

M. Rudolph, M. Motylenko, D. Rafaja, IUCrJ 6, 116 (2019). https://doi.org/10.1107/S2052252518015786CrossRefPubMedPubMedCentral

42.

P. Stadelmann, JEMS v3.67 (Saas-Fee, 2012). https://www.jems-swiss.ch. Accessed 2018

43.

S.V. Tsybulya, G.N. Kryukova, Phys. Rev. B 77, 024112 (2008). https://doi.org/10.1103/PhysRevB.77.024112CrossRef

44.

M. Rudolph, cuDebye v1.5 (Freiberg, 2018). https://github.com/Martin-Rudolph/cuDebye. Accessed 2019

45.

G. Paglia, C.E. Buckley, A.L. Rohl, B.A. Hunter, R.D. Hart, J.V. Hanna, L.T. Byrne, Phys. Rev. B 68, 144110 (2003). https://doi.org/10.1103/PhysRevB.68.144110CrossRef

46.

L. Kovarik, M. Bowden, D. Shi, N.M. Washton, A. Andersen, J.Z. Hu, J. Lee, J. Szanyi, J.-H. Kwak, C.H.F. Peden, Chem. Mater. 27, 7042 (2015). https://doi.org/10.1021/acs.chemmater.5b02523CrossRef

47.

L. Kovarik, M. Bowden, A. Genc, J. Szanyi, C.H.F. Peden, J.H. Kwak, J. Phys. Chem. C 118, 18051 (2014). https://doi.org/10.1021/jp500051jCrossRef

48.

I. Levin, D. Brandon, J. Am. Ceram. Soc. 81, 1995 (1998). https://doi.org/10.1111/j.1151-2916.1998.tb02581.xCrossRef

49.

G. Paglia, C.E. Buckley, A.L. Rohl, R.D. Hart, K. Winter, A.J. Studer, B.A. Hunter, J.V. Hanna, Chem. Mater. 16, 220 (2004). https://doi.org/10.1021/cm034917jCrossRef

50.

M. Rudolph, A. Salomon, A. Schmidt, M. Motylenko, T. Zienert, H. Stöcker, C. Himcinschi, L. Amirkhanyan, J. Kortus, C.G. Aneziris, D. Rafaja, Adv. Eng. Mater. 19, 1700141 (2017). https://doi.org/10.1002/adem.201700141CrossRef

51.

R.-S. Zhou, R.L. Snyder, Acta Cryst. B 47, 617 (1991). https://doi.org/10.1107/S0108768191002719CrossRef

52.

G. Yamaguchi, I. Yasui, W.-C. Chiu, Bull. Chem. Soc. Jpn 43, 2487 (1970). https://doi.org/10.1246/bcsj.43.2487CrossRef

53.

L. Kovarik, M. Bowden, J. Szanyi, J. Catal. 393, 357 (2021). https://doi.org/10.1016/j.jcat.2020.10.009CrossRef

54.

B.C. Lippens, J.H. de Boer, Acta Cryst. 17, 1312 (1964). https://doi.org/10.1107/S0365110X64003267CrossRef

55.

Y. Repelin, E. Husson, Mater. Res. Bull. 25, 611 (1990). https://doi.org/10.1016/0025-5408(90)90027-YCrossRef

56.

L. Kovarik, M. Bowden, A. Andersen, N.R. Jaegers, N. Washton, J. Szanyi, Angew. Chem. 132, 21903 (2020). https://doi.org/10.1002/ange.202009520CrossRef

57.

J.-B. d’Espinose de Lacaillerie, C. Fretigny, D. Massiot, J. Magn. Reson. 192, 244 (2008). https://doi.org/10.1016/j.jmr.2008.03.001CrossRefPubMed

58.

R. Dupree, M.H. Lewis, M.E. Smith, Phil. Mag. A 53, L17 (1986). https://doi.org/10.1080/01418618608242816CrossRef

59.

R. Prins, Angew. Chem. Int. Ed. 58, 15548 (2019). https://doi.org/10.1002/anie.201901497CrossRef

60.

C. Wolverton, K.C. Hass, Phys. Rev. B 63, 024102 (2000). https://doi.org/10.1103/PhysRevB.63.024102CrossRef

61.

C. Vinod Chandran, C.E.A. Kirschhock, S. Radhakrishnan, F. Taulelle, J.A. Martens, E. Breynaert, Chem. Soc. Rev. 48, 134 (2019). https://doi.org/10.1039/C8CS00321A

62.

B.E. Yoldas, J. Mater. Sci. 11, 465 (1976). https://doi.org/10.1007/BF00540927CrossRef

63.

G. Mariotto, E. Cazzanelli, G. Carturan, R. Di Maggio, P. Scardi, J. Solid State Chem. 86, 263 (1990). https://doi.org/10.1016/0022-4596(90)90142-KCrossRef

64.

M.K. Mardkhe, B. Huang, C.H. Bartholomew, T.M. Alam, B.F. Woodfield, J. Porous Mater. 23, 475 (2016). https://doi.org/10.1007/s10934-015-0101-zCrossRef

65.

R. Tettenhorst, Clays Clay Miner. 28, 373 (1980). https://doi.org/10.1346/CCMN.1980.0280507CrossRef

66.

D. Massiot, F. Fayon, M. Capron, I. King, S.L. Calvé, B. Alonso, J.-O. Durand, B. Bujoli, Z. Gan, G. Hoatson, Magn. Reson. Chem. 40, 70 (2002). https://doi.org/10.1002/mrc.984CrossRef

67.

P. Scherrer, Nachr. Ges. Wiss. Gottingen, Math.-Phys. Kl. 1918, 98 (1918). https://eudml.org/doc/59018

68.

S. Ďurovič, Chemical Papers (Chemické Zvesti) 23, 113 (1969), https://www.chempap.org/?id=7&paper=5979

69.

V. Prostakova, D. Shishin, M. Shevchenko, E. Jak, Calphad 67, 101680 (2019). https://doi.org/10.1016/j.calphad.2019.101680CrossRef

Titel: Temperature-Induced Changes in the Microstructure of the Metal Melt Filters and Non-metallic Inclusions
verfasst von: Martin Thümmler
Milan Dopita
Mykhaylo Motylenko
Anton Salomon
Erica Brendler
David Rafaja
Verlag: Springer International Publishing
Buch: Multifunctional Ceramic Filter Systems for Metal Melt Filtration
Print ISBN: 978-3-031-40929-5

Electronic ISBN: 978-3-031-40930-1

Copyright-Jahr: 2024
DOI: https://doi.org/10.1007/978-3-031-40930-1_6

Premium Partner

Marktübersichten

Die im Laufe eines Jahres in der „adhäsion“ veröffentlichten Marktübersichten helfen Anwendern verschiedenster Branchen, sich einen gezielten Überblick über Lieferantenangebote zu verschaffen.

Zur Marktübersicht