In this chapter, we will look at microstructures formed in materials in the solid state: especially at phase transformation microstructures, and heterochemical reaction microstructures. As the microstructures of solids play a major role in determining their properties, their formation has been studied in great detail for both ceramic and metallic materials for a long time. The aim here is not to produce an exhaustive or quantitative text on the systematics of solid state microstructures. Excellent treatises on this already exist and the readers are strongly encouraged to consult them (for example the still preeminent book “Phase Transformations in Metals and Alloys” by David A. Porter and Kenneth E. Easterling, CRC Press, 1992. ISBN: 9780748757411; and the ASM series of Handbooks, including Vol. 9, “Metallography and Microstructures”, George F Vandervoort (ed.), 2004, ASM International, ISBN-13: 978-0871707062). Rather, we want to walk through structures based on imagery, explaining the most common types seen but also illustrating the appearance of more uncommon and ‘tricky’ structures, that are harder to recognize. As always, we want to emphasize that microstructures encode information about the state and history of materials. Often, this information is very hard to obtain through any means other than microstructural analysis, which is a cost-effective alternative to expensive or hard to implement direct measuring instrumentation, especially in industrial processes.

As with other microstructures, it should be emphasized that the development of these structures is controlled by three different types of circumstances:

• The chemical composition, which yields a certain set of thermodynamically stable or metastable phases at a given physical condition, such as temperature (T) and pressure (P). All the thermodynamic phase relations apply, such as the phase rule, which determines the number of phases that can stably coexist at a given degree of freedom. Understanding of the chemical composition and thermodynamics of a material is needed to determine whether it is fully equilibrated or not, and whether its microstructure reflects a frozen-in reactive state.

• The rates of processes at ambient conditions. This includes reaction rates in the chemical sense, but most notably transport rates of chemical elements, such as diffusion. These factors often determine the length scales over which structure formation can occur. One important aspect of solids is the importance of deformation rates when subjected to specific stress conditions. This characteristic sets solids apart from liquids or gases, which have no or very little resistance to shear deformation. As solids typically have a crystalline structure, which implies crystal anisotropy, the interaction between deformation and its rates, as well as the transport rates of chemical elements, can lead to a unique way of material differentiation (e.g. pressure shadow precipitation, or pressure solution, as seen in stylolites in geology).

• The rate of change of conditions. If conditions are changing rapidly, the length scales over which structures develop, also change. This can lead to the superposition of microstructures formed at different conditions. In extreme cases, such as rapid cooling (called ‘quenching’), the change might be so fast that structures might entirely not develop. Glasses are the best example of this, where crystal formation that otherwise would occur along a cooling path, is entirely suppressed. Rapid cooling can also preserve microstructures of an earlier state in solids, when structure reformation at cold conditions is too slow. This preservation of microstructures is particularly useful in industrial metallurgy, where properties of products often are determined by metastably preserved deformational microstructures. The influence of the rate of change does not only apply to cooling processes however: An example of this is thermoshock, where rapid heating causes thermal expansion, resulting in the accumulation of stresses. If a material does not have enough time to relieve these stresses through plastic deformation, it may fracture and fail.

Figure 1: Precipitation of BN (boronitride) dendrites in steel. Images are Scanning Electron Microscope (SEM) images in Back-scattered Electron (BSE) mode. Below the images, the typical form of phase diagrams for the exsolution of a new phase (β) of a solution phase (liquid or solid) is shown, in different topological configurations.

Continuous precipitation refers to the nucleation and growth of a phase in a homogeneous medium (solute). Fig. 1 shows the nucleation and growth of BN crystals in steel. As can be seen, continuous crystallization from a solid is morphologically exactly analogous to crystallization from a homogeneous melt. In the case shown here, the matrix steel is a solid solution phase (gamma-iron, austenite) at the temperature of precipitate formation. It carried B and N as dissolved elements at very low concentrations. Nevertheless, during cooling, saturation with BN is reached and thus, BN nucleated. Growth from a nucleation point occurs just as growth from a nucleation point occurs when crystals grow from a liquid. As in the liquid case, the solute elements incorporated in the growing crystal must diffuse to it, and the elements of the solution phase that are rejected by the growing crystal, must diffuse away from it (Fe in this case). The only difference imposed by the circumstance that the solution phase is a solid itself, lies in the potential crystallographical anisotropy of these transport processes, which could enforce preferred growth in specific crystallographic directions.  

In all three topological situations in the phase diagrams shown in Fig. 1, the mechanics of precipitation and growth are also same. A homogeneous solute phase becomes instable and dissociates in a growing nucleus of a precipitate within the solution phase. In order for nucleation to occur, a nucleus must form, which requires overcoming an energetic barrier in the form of activation energy. Therefore nuclei form faster at a high supersaturation, where the energy gain from precipitation is increased. Thus, when a nucleus is formed, there is a relatively rapid growth that follows in order to relieve the solute medium of the supersaturated elements. As crystal growth is faster at edges where diffusive transport of solute elements to the growth site is easier, the initial growth morphology tends to be dominated by edge forms, leading to the development of dendritic crystal forms (arms extending out into the surrounding). Typically, such dendrites have star-like, bushel-like or tree-like morphologies, from which their nucleation point can be inferred. The degree of crystallographical control on the growth forms depends not only on the energetic surface anisotropy of the growing crystal but also on the properties of the solute medium. Here is a difference between growth from liquid solutes (melts), which displace easily and are isotropic, and growth from a solid solute phase, where the host phase can impose its own anisotropy onto the growth forms. In the case shown here, diffusive iron displacement has to cooperate with the growth of the incipient BN dendrite. Solid-solid dendrites often appear less strictly crystallographic for this reason. In the present case, there is also a very strong hardness difference. BN is a crystallographic analog of graphite and can easily accommodate growth stresses through plastic deformation. Still, the edge dominated growth of the BN dendrites can be observed in its star like shapes.

Figure 2: Heterogeneous nucleation of boronitride (BN) (dark) on an inclusion of MnS (mid grey) in steel (bright matrix). SEM BSE image, same sample as in the previous figure.

Figure 2 gives an example of heterogeneous nucleation and growth. The image shows a different area of the same steel sample that was imaged in Figure 1. It shows bushel-shaped dendrites of BN growing on the surface of a MnS inclusion in the steel. As discussed above, the driving force for the precipitation and growth of the BN is the supersaturation of the solute matrix phase (austenite in this case) with B and N at decreasing temperature. However, nucleation, especially of a phase that is crystallographically as different from austeinte as BN is, involves overcoming an energetic barrier to nucleation. Many preexisting microstructures in a solute phase do lower this energetic barrier. Grain boundaries and phase boundaries are a good example of this. The presence of the crystallographic discontinuity inherent in the phase boundary between MnS and austenite makes it easier for the solute elements to accumulate along this discontinuity as a crystallization nucleus. Nucleation tends to utilize preexisting discontinuities as templates for nucleation. As a result, precipitate crystals grow from these features. The case imaged in Figure 2 shows a high density of BN dendrites covering the surface of the preexisting MnS in the steel. A concentration of nucleation spots along discontinuities is commonly observed in nature. Various structures can serve as templates for nucleation, such as free surfaces or the surfaces of bubbles in a liquid. Conversely, the walls of a liquid container can also act as preferred nucleation sites, as in the common case of bubble nucleation in a soda drink. One consequence of having spatially preferred nucleation sites is that this can distort the homogeneous distribution of elements. In the above case, the growing dendrites of BN are supplied by diffusion from the depth of the surrounding phase, which is relatively free of nuclei. As a result, the elements B and N become quantitatively accumulated around the MnS which acted as the template for dendrite growth. This is one of the fundamental mechanisms by which chemical heterogeneity can develop in originally homogeneous materials.

Figure 3: Exsolution of iron oxide (hematite, Fe2O3, bright) from an ilmenite solid solution ((Ti,Fe)2O3). SEM BSE image from McEnroe et al 2007. Note the presence of multiple generations of exsolutions.

Exsolution is a special case of precipitation where a homogeneous solute phase disproportionates into two phases (a →  a + b transition along a solvus). In this case, the newly formed phase b grows preferentially along crystallographic features of the host phase, so that the entire bi-phase arrangement acquires the a morphology controlled (in large part) by the host phase crystallography. Figure 3 shows an example of such an exsolution structure, in this case an exsolution of hematite (Fe2O3) from a preexisting (Ti,Fe)2O3 (ilmenite) solid solution. As the schematic phase diagrams in Figure 1 illustrate, a solvus between phases a, b typically changes its compositional location as temperatures change. This causes the two phases to continuously react with each other as the temperature changes. Depending on the transport properties of the phases, they could both cooperatively change their compositions through diffusion to reflect changed temperatures. However, if transport (diffusion) is slow, nucleation barriers are low, and/or conditions change rapidly, it can be energetically favoured to re-nucleate a in b and b in a rather than to diffuse compounds out of a and b to their respective adjoining phase. In that case, another generation of exsolution can be formed, wherein both phases again exsolve out of each other as the mutual solubility shrinks. Figure 3 shows this case: one can observe that the original exsolution lamellae of hematite themselves carry small ilmenite lamellae, whereas small hematite lamellae are formed in the surrounding ilmenite lamellae. Exsolution is most frequent when the precipitation nucleation barrier is not high but the crystallography of the host phase a imposes a strong growth anisotropy. Because the solvus between two mutually soluble phases changes continually with temperature, such exsolution microstructures are often used to record temperatures (especially in geology). At high temperature the two phases equilibrate to each other, but at lower temperatures, the shortened reach of diffusion prevents such reequilibration and this means the two phases do not adjust their composition anymore as temperature falls. In this way, such a host-exsolution couple has a 'closure temeprature' recorded by their composition. Moreover, the reintegration of various generations of exsolutions together with their measured phase compositions allows the reconstruction of previously existing compositions on the solvus before exsolution occurred. If combined with detailed knowledge of the dependency of diffusion speeds on T in the phases, such microstructures can even be used to quantitatively reconstruct the rate of change of temperatures (e.g. cooling rates). More often, exsolution microstructures are utilized simply to reconstruct pre-exsolution homogeneous compositions. This is especially useful when these compositions are difficult to observe or measure directly, for example in high temperature industrial processes. However, care must be taken in the interpretation of exsolution microstructures because it is possible that the overall composition of the material does not remain constant. The amount of exsolved phase may then represent a dynamic element influx (or outflux) rather than the decomposition of a preexisting homogeneous bulk solute phase. We will provide an example of this in the following.

Figure 4: “Oxy-exsolution”. This SEM-BSE image shows a pile of rounded magnesiowuestite crystals with remnant traces of solidification of the oxide melt )slag) they were derived from (dark phase: calcium ferrite, the melt minimum phase, traces  an interstitial space and pseudo eutectic intergrowths). The magnesiowuestite was originally a solid solution with a composition of (Fe, Mg, Mn, …)1-xO. Later, reaction-exsolution lamellae of Fe3O4 (bright) formed in this magnesiowuestite. Note the dark diffusive halos around these Fe3O4 lamellae. See text.

Figure 4 shows a case where exsolution of a phase was largely driven by a non-constant bulk composition. This image shows a pile of magnesiowuestite (Mg,Fe,Mn,…)1-xO crystals, formed from the solidification of an oxide melt (slag). Remains of this slag solidification can be seen in the miminum crystallization phase of this slag, calcium ferrite (dark grey in the image) which occupies the interstitial volume between the magnesiowuestite crystals. The magnesiowuestite solid solution has a nonstoichometric composition, as a part of its cations is in trivalent state, especially Fe³⁺ instead of Fe²⁺. Mn too can exist in both trivalent and divalent states, whereas Mg is always divalent. Depending on the value of x, the proportion of Fe³⁺/Fe²⁺ in the phase composition can be calculated: Trivalent iron, expressed on a single-oxygen basis, enters the solution as Fe³⁺2/3O, so that the number of Fe³⁺ = 2 x, and the amount of Fe²⁺ = TotalFe – Fe³⁺. In pure iron wuestites, x can become quite large (see phase diagram in Figure 5), but in most multimetal mixed magnesiowuestites, it does not much exceed 0.06. From such a mixed magnesiowuestite, Fe3O4 can exsolve either isochemically, or by addition of oxygen, that is, oxidation, as shown in Figure 5. The amount of exsolved Fe3O4 in Figure 4 is such that it is likely that in this case, an oxygen influx into the material drove the exsolution. This is also indicated by the visible dark halos around the relatively bright magnetite exsolution lamellae in Figure 4: These are diffusion gradients of Mg around the exsolution lamellae, as Mg is not as easily incorporated into a Fe3O4 spinel structure and preferentially gets rejected to remain enhanced in the surrounding MeO structure. In the example shown, the microstructure preserved a ‘hidden’ trace of an oxidation event. Such information is evidently important for the analysis of industrial process performance.

Figure 5: Phase diagram of a portion of the system Fe-O (based on a FactSage calculation), showing the extent of the wuestite solid solution field towards oxygen. The exsolution of a more oxygen rich phase from a selected starting composition within the field (black circle) can occur by simple cooling (arrow downwards), but also by oxidation, in which case it is not isochemical (oblique arrow). Both lead to the same microstructures.

Figure 6 : Formation of globular oxides below an oxidation front (scale formation) in silicium rich steel

That such seemingly simple microstructures are often not just isochemical transformations but involve complex reactions is also illustrated by the structures in Figure 6. In this SEM BSE image, an oxide scale formation on a Si-alloyed steel is shown. Scale formation is a common and well studied process in metallurgy and involves the reaction of the Fe alloy to Fe1-xO, Fe3O4 and Fe2O3 oxides in an oxidising atmosphere (often air), as seen at the right image edge. Under a more or less coherent scale layer, oxygen access to the steel surface is restricted, which significantly slows down the direct bulk transformation of the exposed metallic iron.  Under these conditions, the diffusion of oxygen into the metallic steel itself is efficient enough to run ahead of the bulk metal consumption front (if the metal surface were directly exposed, bulk metal oxidation would be so rapid that any diffusive gradient would be consumed before it formed). In this oxygen-restricted environment, the limited amount of oxygen available is able to diffuse into the metal base. Here however, it encounters all the metals in solid solution, which includes not only the Fe making up the matrix but all the alloy metals too. In this case this includes Si, but also other minor and tramp metals such as Mn, or Cr, or Cu. The solubility of oxygen in such a metal matrix is not zero (thus allowing diffusion), but still very low, on the order of parts per million (ppm). Continued diffusion of oxygen into the metal therefore quickly reaches the solubility limit of oxygen in that matrix. At this point, an oxide phase will “exsolve”. That is the process that forms the round inclusions seen in the image within a layer beneath the bulk scale oxidation front. The “exsolution” in this case, like in Figure 5, effectively amounts to an oxidation reaction: x Metal Me (Fe, Mn, Si, Cr, …) + y O = MexOy. However, the cationic composition of the exsolved oxide differs significantly from that of the surrounding metal matrix. This is because the chemical energy of the oxidation reaction varies significantly for different metals on an atom-by-atom basis. The Gibbs energy change from the oxidation reaction at a given temperature is highest for metals like Ti, V, then Si, Cr, then Mn, then Fe, then Cu, Ni, and so on. This order is known as the "Ellingham" sequence in metallurgy, named after the "Ellingham diagram" which plots the change in Gibbs free energy (ΔG) of a reaction against temperature for different metal-oxide pairs. 

The metals with the highest energy change are selectively combined with oxygen in this oxy-exsolution reaction. This means that the exsolved oxide will primarily be an oxide of Ti, then Cr, then Mn, only then Fe, and so on, as oxygen is continually added. This can be directly observed in the image: all the exsolved globular oxide 'droplets' are darker in BSE contrast than the iron oxides of the scale, indicating that they contain lower Z metals than Fe, such as Ti, Cr, Mn. In the observed case, the phases are (Mn,Fe)1-xO and galaxite spinel Mn(Cr,Al)2O4. One could express this by stating that as the local oxygen concentration in the multimetal solution increases due to the diffusion of O, it reaches the solvus in the direction of Mn and Cr before reaching the solvus in the direction of Fe. The formation of the sub-scale oxides here reflects the multi-component phase relationships. However, this is not the end of the information in the imaged microstructure.  Pixel counting on the observed area of globular oxides reveals that they constitute approximately 4-5% of the total volume. The densities of FeO and MnO are approximately 5.3-5.7 [t/m3], while that of solid steel (ferrite) is 7.9 [t/m3]. Thus, the globular oxides make up approximately 2.8% of the local mass, with 77 % of it being composed of the element Mn (mass of Mn in MnO). Thus, if the globular oxides are taken to be pure MnO, the image would suggest that the local concentration of Mn alone would be approximately 2.2 wt % Mn. This is significantly higher (by a factor of about 10) than the original Mn content in the steel! The inclusions seen are not pure MnO though; they are a complex mix of oxides (Mn, Cr, Fe, ...) oxide. However, the quantity of inclusions below the scale surface at this locality is so high that the absolute concentration of minor alloy metals Si, Cr, Mn has increased here compared to their initial levels in the steel.

Figure 7: Detail of the globular inclusion zone of the steel sample in figure 6, closer to the scale formation surface. Note here that local µO2 in the steel base has risen to the point that Fe is also “oxy-exsolved” to form Fe oxides (brighter) along the alloy metal oxides in the inclusions (darker part of the inclusions). Note that though the oxides are labelled as “Fe” oxides, they are complex solid solutions involving (Fe, Mn, Cr, Si, ..) cations reflecting the µO2 in their microenvironment.

The reason for this is the oxy-exsolution process itself. As oxygen diffusion into the matrix at that spot consumes the alloyed minor metals, their local concentration in the surrounding Fe matrix gets lowered (the Fe matrix becomes purer Fe). This means that a concentration gradient in Mn, Cr, Si and so on is set up within the metal phase, compared to the alloy concentration in the inner regions of the metal. As soon as that happens, the minor alloy metals begin to diffuse down this concentration gradient, towards the spot where the oxy-exsolution takes place. The diffusion speed of such intermetallic diffusion at process conditions is significantly faster than the inward diffusion of oxygen through a Fe matrix from the surface. Thus, the oxy-exsolved globular inclusions form at the spot where two diffusive streams are in balance: Namely the inward stream of oxygen and the outward stream of the minor alloy metals consumed by oxidation in Ellingham order. In the layer where the oxy-exsolved inclusions form, these metals “pile up” (as oxides), because the growing reaction product is continually supplied from both sides. This has consequences, as it changes the local bulk composition. As the external conditions change (usually by cooling), the changed local composition can become the new effective system composition, because a shorter diffusive length scale prevents equilibration over the scale of the process at higher temperature. Subsequently, the alloy metal oxide enriched composition can further adjust to form phase assemblages that were not initially expected based on the original steel alloy composition.

This demonstrates that apparently simple and ‘common’ microstructures like the one observed in Figure 6 depend on a complex interplay between all three of the initially mentioned classes of parameters: thermodynamic equilibria, process rates, and their rate of change. The point of microstructural analysis is to understand microstructure in order to extract this information, which is critical for industrial or natural process analysis.

Figure 8: Top: SEM BSE image of the same sample as shown in Figure 6. Below are higher magnified images of two spots in this sample, showing the presence of Si3N4 precipitates along the ferrite grain boundary in depth (bottom left) and in higher amount along the base of the globular inclusion layer (bottom right).

The detail in which the reactions forming the microstructures can be traced back is further illustrated, using the same steel sample shown in Figure 6, by taking a higher-magnification look at it. This is shown in figure 8. At high magnification in the steel base, tiny (submicrometric) precipitates are observed the steel base. EDS analysis has revealed that these precipitates are Si3N4, not an oxide phase. This shows that not only oxygen, but also nitrogen diffused into the steel base, which is to be expected when the scale formation is driven by reaction with air. As with oxygen, nitrogen also has a small but nonzero solubility in the alloy but it does not react with the Fe matrix at its exposed surface, and so forms a diffusion gradient into the steel base, while oxygen creates the oxidation reactions mentioned above.  Eventually, nitrogen too builds up in the steel until a solvus is reached, in this case towards a nitride phase. For the given, highly Si alloyed composition (about 3% Si), Si3N4 is the first solvus phase to be encountered, and Si3N4 precipitates develop. The images show that this exsolution preferentially takes place at locations such as ferrite grain boundaries in the steel base where the energetic nucleation barrier is lowered. However, closer to the surface, under continued nitrogen supply, homogeneous precipitation also occurs.

Figure 9: Confocal scanning laser microscopy image of a massive delta ferrite to Gamma transformation in peritectic steel (from Yin et al. Acta Materiala 1999, 47:1523). Gamma phase, bright, grows incoherently in finger pattern but massively into the preexisting Delta grains (polygonal grain structure).

Massive transformations are widespread in metallurgy and are extensively discussed in established literature, both qualitatively and quantitatively. It is sufficient, therefore, to just show a few examples here because such microstructures should be well known to the reader. Figure 8 provides a specific example of the well-known peritectic transformation from delta-ferrite to gamma (austenite) phase in a low carbon steel, as mentioned in the literature cited. This is an example of an incoherent transformation, where there is no direct crystallographic lattice equivalence along the grain boundary. It should be noted that this is not an isochemical transformation but one involving a chemical change (peritectic reaction), where carbon in the new phase (austenite) is higher than in the replaced phase (d ferrite).

The irregularity of the interface indicates that the carbon supply for the new phase does not occur through volume diffusion. Instead, due to the incoherent nature of the transformation, diffusion along the grain boundaries of the transformation is greatly enhanced. As a result, the transformation boundaries themselves serve as pathways for the diffusion of the supplied component. This type of massive transformation is most commonly discussed for metals, but is also observed in ceramic materials. This type of massive transformation should be distinguished from examples where a similar structure results from having a thin layer of a fluid or a melt along the transformation boundary, through which compounds can diffuse faster than through the bulk of the replaced or newly formed phases. Such a grain boundary following fluid becomes a medium into which the old phase dissolves and out of which the new phase is formed. This type of transformation is therefore often referred to as a dissolution-reprecipitation reaction. This is especially common in geologic (that is largely ceramic) materials, where phases are massively replaced by newgrowth phases with very different compositions.

Figure 10: cellular decomposition of a metal alloy, in this case a copper-indium alloy, into a set of two new phases that form a coherent intergrowth (image taken from Literature.) (TEM images)

Both massive transformation and dissolution-reprecipitation transformations can form complex intergrowths when the replaced old phase is replaced by not one, but a set of two (or more) newly formed phases, that together replace the old phase. A “conventional”, metallurgic case of an isochemical massive decomposition of a solute phase into a set of new phases is shown in Figure 10. Here the same conditions apply as in eutectic crystallization of liquids, where the product phases together must fully account for the composition of the parent solution phase. Because of this similarity and the resulting similar microstructures, such a transition is often referred to as an "eutectoid" transition in metallurgy. It results from the same “pinching out” of a solution phase stability fild on a phase diagram as in eutectic liquid crystallization. As the individual phases of the product assemblage individually all have compositions that deviate from the composition of the preexisting solute phases,

Figure 11: SEM BSE image of an oxide inclusion in steel. The inset upper right shows an enlargement of a part of the transformation front. The main phase of the inclusion is a mixed Al-Ti oxide with trivalent Ti: (Al,Ti³⁺)2O3. Along the margins, this phase is transformed into an assemblage of two new phases, which along parts of the transformation front takes on a cellular decomposition microstructure (inset). The two new phases are a hercynite rich aluminate spinel (darker phase) and a rhombohedral oxide (ilmenite structure) rich in ilmenite component (Fe²⁺Ti⁴⁺O3 (slightly darker in BSE contrast than the main inclusion oxide).

there is a necessary rearrangement (diffusion) of compounds along (parallel) to the decomposition front involved. Figure 10 shows the development of this process in detail. Nuclei of one phase along the front require for their growth that those elements which are relatively less incorporated in them, diffuse away. If they would diffuse away from the front into depth of the parent solution phase, the composition of this would change unfavourably to the nuclei resulting in self-inhibition of growth. Instead, sideways diffusion piles up the incompatible elements in between the nuclei, leading to the formation of a phase that incorporate these elements. In this way, a regular succession of nuclei of all the product phases is formed, the length scale of which is dictated by the relative diffusion lengths for the elements in the given solution phase under the specific temperature and pressure conditions. Thus a cooperative growth front is formed that can move through the volume of the parent phase, transforming it into an ordered arrangement of the product phase. This intergrowth is often called ‘cellular’ and the individual diffusion segments along the front are called cells.

While the quantitative treatment of this microstructure has largely been developed in metallurgy, it should not be thought that this microstructure is only limited to metals. The same principles also apply in ceramic (oxidic) as well as in any other compositional group of materials. An example is given in Figure 11, which shows the decomposition of a mixed Al,Ti oxide (Al,Ti³⁺)2O3 which was entrapped in a steel during steel casting. In this case though, as it happened under mildly oxidising conditions with oxygen diffusing in, the cellular decomposition was not strictly isochemical, with hercynite-rich spinel solid solution (darker) and an ilmenite rich rhombohedral oxide solution being formed as product phases. Such cellular, ‘euteoctoid’ decomposition is comparatively rarer in ceramic materials though since the diffusion of compounds through the ionic crystals with complex coordination polyhedral that most oxide phases are, is much more restricted or selective.