Evolution of Fe Bearing Intermetallics During DC Casting and Homogenization of an Al-Mg-Si Al Alloy

The evolution of iron (Fe) bearing intermetallics (Fe-IMCs) during direct chill casting and homogenization of a grain-refined 6063 aluminum-magnesium-silicon (Al-Mg-Si) alloy has been studied. The as-cast and homogenized microstructure contained Fe-IMCs at the grain boundaries and within Al grains. The primary α-Al grain size, α-Al dendritic arm spacing, IMC particle size, and IMC three-dimensional (3D) inter-connectivity increased from the edge to the center of the as-cast billet; both αc-AlFeSi and β-AlFeSi Fe-IMCs were identified, and overall αc-AlFeSi was predominant. For the first time in industrial billets, the different Fe-rich IMCs have been characterized into types based on their 3D chemistry and morphology. Additionally, the role of β-AlFeSi in nucleating Mg2Si particles has been identified. After homogenization, αc-AlFeSi predominated across the entire billet cross section, with marked changes in the 3D morphology and strong reductions in inter-connectivity, both supporting a recovery in alloy ductility.


Introduction
Due to their high strength and good formability, AA6xxx Al alloys have found widespread applications as structural materials in the transport and building industries [1]. The alloys are direct chill (DC) cast in billet form suitable for subsequent extrusion. Due to the nonequilibrium solidification conditions in DC casting, most of the solute elements (in this system, Fe, Si and Mg) segregate into the interdendritic regions and grain boundaries of the primary α-Al [2]. This solute rich liquid leads to the formation of the inter-dendritic secondary phases, such as Fe-rich intermetallic compounds (Fe-IMCs) and strengthening Mg 2 Si precipitates [3,4]. Previous studies show that α c -AlFeSi and β-AlFeSi are the two dominant Fe-IMCs in 6063 Al alloys [5][6][7][8]. 3D analysis of these Fe-IMCs after extraction from the primary α-Al using an Al dissolution approach revealed that α c -Al(FeMn)Si (from now referred to as α c -AlFeSi) had a dendritic-like morphology while β-AlFeSi had a platelike morphology [9]. Among these IMCs, it has been noted that β-AlFeSi in particular, because of its more planar geometry, reduced the ductility of Al alloy 6063 space [6].
Despite the apparent maturity of AA6xxx alloys there is a significant on-going effort to optimise homogenisation conditions in terms of properties, while also minimising the homogenisation time. Among the factors that govern the homogenisation response, the initial cast microstructure and the alloy chemical composition play key roles. Therefore, in developing new solidification processing routes [15] or advanced solidification technologies [17][18][19][20] that seek to promote more favourable as-cast microstructures, it is important to understand better the link between microstructural evolution in casting and the subsequent homogenisation response of the secondary phases. For example the cooling rate and solid/liquid growth velocity in DC casting significantly affect the proportions of α c -AlFeSi and β-AlFeSi in the final microstructure [8,10,21,22]. But, since the cooling rate and growth velocity vary from position to position in the billet during DC casting [2], [23], the primary Al grain size and proportion of the Fe-IMCs and Mg 2 Si across the cross-section of the billet also varies, which should be accounted for in homogenisation heat treatment optimisation. In this paper, we apply an IMC phase extraction technique, along with a series of other characterisation methods, at different positions in a AA6xxx DC cast billet to track in detail changes in the Fe-rich IMC inter-connectivity, size, morphology, type and proportion during homogenisation. The spatial variation of as-cast microstructure is shown to play a significant role in homogenisation response and the various types of IMCs and their formation and transformation are described in detail.

Experimental details
The chemical composition of the grain refined (Al-5Ti-1B) DC cast, ~180 mm diameter Al-Mg-Si billet is given in Table 1, measured using a Spectrolab LAVFA05A spark analyser optical emission spectrometer. Consistent with industrial practice, the cast billet was homogenised at 580 ºC for ~5 hrs. The billet before and after homogenisation was sectioned horizontally across a diameter at the billet mid-height for microstructural analysis. Samples for metallography were ground with SiC papers and then polished to 0.06 µm colloidal silica finish. Samples were also anodized with Barkers reagent [7 mL HBF 4 (48 pct) and 93 mL H 2 O] for 60 s at 20 V and studied by light microscopy (Zeiss Axiophot2) under polarised light and dark field imaging conditions to assess the primary Al grain size and secondary phases. The mean linear intercept method was used to measure the primary Al grain size.
Scanning electron microscopy (SEM) was conducted in a JEOL 840A, and a JEOL 5510 equipped with a secondary electron (SE) detector, a backscattered electron (BSE) detector and an energy dispersive spectrometer (EDS). Second phase particle size distributions were measured using Axio vision image analysis software on BSE images at x500 magnification and at least 1000 particles.
Intermetallic extraction apparatus (IEA) was used to extract the intermetallic particles from the as-cast and homogenised samples. IEA uses anhydrous boiling butan-1-ol (butanol) to dissolve the α-Al matrix of the alloy while keeping the intermetallic phases intact. The intermetallics were then collected on a poly-tetrafluoroethene (PTFE) filter membrane (47 mm diameter, pore size of 0.2 µm). The phases were identified in a Philips 1700 X-ray diffractometer (XRD) operating with Cu-K α (λ = 1.54 A) radiation at 35 kV, 50 mA and 0.05 o step size. An in-house developed calibration procedure was used for determining the relative weight fractions of α c -AlFeSi and β-AlFeSi from XRD data, more details of which can be found in references [8,22,24]. The extracted particles were further analysed using a high resolution field emission gun (FEG) SEM (JEOL 840F). The approach used to measure the IMCs inter-connectivity, length of petals and length of longest facet side of platelets of the extracted particles was detailed in previous work [22]. MTDATA studio 5.10 software with NPL aluminium database V6.1 was used to predict phases as a function of temperature under various assumptions (see later).

Primary α-Al grain analysis
The polarised light microscopy images in Fig 1(a) show the dendritic equaixed morphology of α-Al grains across the cast billet cross-section. The grain size and dendrite arm spacing (DAS) increased from the near surface to the centre of the cast billet ( Fig. 1(c)). This suggests change in cooling rate along the billet cross-section. The cooling rate can be calculated from [25] = 3.57 × 10 4 where V is the cooling rate, Ks -1 . Using the observed DAS (μm), the cooling rate at distance of 15 mm, 30 mm, 60 mm and 90 mm from the billet edge towards the centre was estimated as 20.2 Ks -1 , 10.4 Ks -1 , 5.7 Ks -1 and 5.4 Ks -1 respectively, which showed good agreement with measured data [23,26,27]. In general during DC casting, the liquid metal is delivered from the furnace to the water cooled ring mould by a launder [23]. Once the melt comes in to contact with the mould it starts to solidify and forms a thin solid shell around the hot liquid.
Further withdrawal of the base of the mould results in water directly coming into contact with the solidifying shell, and induces cooling rates of 20 K/s (near surface) to 1 K/s (billet centre) [23], similar those estimated here in this study. Because of the inevitably higher cooling rate at the billet surface, finer α-Al grains and a smaller DAS resulted in this region, as shown in After homogenisation, the α-Al grain morphology showed similar features to the as-cast structure (Figs. 1(a) and 1(b)), but the DAS became difficult to quantify due to a reduction in inter-dendritic contrast. However the grain size in the homogenised billet followed a similar trend with respect to the as-cast billet. All error bars in this manuscript indicates the standard deviation of the mean, unless stated.

Secondary phase analysis (A) XRD analysis of extracted particles
In order to evaluate the type and proportion of the secondary phases in these relatively dilute Al alloy systems, where secondary phases fractions are less than 3 wt.%, it is most effective to extract the particles from the α-Al matrix for various types of subsequent analysis. The XRD trace from the extracted particles in Fig. 2 shows cubic α c -AlFeSi (α c -Al(FeMn)Si), monoclinic β-AlFeSi and cubic Mg 2 Si phases along the billet cross-section, taken at distances of 15, 30, 60 and 90 mm from the billet edge towards the centre. Of the Fe-IMCs, α c -AlFeSi showed the highest intensity peaks (> 50%) in both as-cast ( Fig. 2(a)) and homogenised ( Fig.   2(b)) billets. Using calibrated standards [28] to gain phase fractions, Fig. 3 shows that in the as-cast billet, the proportion of α c -AlFeSi relative to β-AlFeSi was almost uniform across the radius. However after homogenisation heat treatment, in all locations, the homogenised billet had a higher fraction (> 90%) of α c -AlFeSi (i.e. less β-AlFeSi). The Mg 2 Si fraction was also lower in the homogenised billet. It is important to note that the pore size of the filter paper used to collect the extracted particles was 200 nm and only individual particles or cluster of particles bigger than the pore size will be present in the analysis.

(B) Metallography
The homogenised billet showed secondary phase particles at the grain boundaries (A) and within Al grains (S) (Fig. 4(a)). The particles (brighter in Fig. 4(b)) at the grain boundaries had a higher aspect ratio then particles within the grains. The larger aspect ratio Fe-IMCs particles mostly had script-like ( Fig. 5(a)) and needle-like ( Fig. 5(b)) morphologies, and the smaller aspect ratio particles had a rosette-like (Fig. 5(c)) morphology. Morphologies were similar in the as-cast billet [9,22] and it was concluded that the script-like and needle-like Fe-IMC morphologies could be attributed to α c -AlFeSi and β-AlFeSi respectively [28,29].
The rosette-like 2D morphology was also attributed to α c -AlFeSi [28]. However Fe-IMC particles in the homogenised billet were smaller in size with a narrower size distribution when compared to the as-cast samples (Fig. 6), and the equivalent particle size (the diameter of a circle with the same area as the projection area of the irregular shaped particle) increased towards the centre in both the as-cast ( Fig. 6(a)) and homogenised billets ( Fig. 6(b)). The more spherical, minor fractions of particles within grains changed little after homogenisation, and data from these particles is not included in these plots.
Grain boundary Mg 2 Si in the as-cast billet tended to be co-located with the needle-like Fe-IMCs ( Fig. 7(a)), whereas Mg 2 Si within α-Al grains was mostly co-located with the rosettelike Fe-IMCs ( Fig. 7(b)). For example, the EDS line scan across the rosette-like region in Fig.   7(c) showed Mg 2 Si and Fe-IMCs together. There was also individual needle-like Mg 2 Si at the grain boundaries of the cast billet. However the Mg 2 Si in the homogenised billet was coarser with an average equivalent particle size of Mg 2 Si at the centre of the as-cast and homogenised billets of 1.6 ± 0.1 μm and 2.9 ± 0.2 μm, respectively. In the homogenised billet, the Mg 2 Si mostly had rod-like and spherical-like morphologies at the grain boundaries.

(C) SEM analysis of extracted particles
In order to understand better the nature (inter-connectivity and morphology) of the secondary phase particles that form during the solidification and homogenisation of the DC billets, extracted particles were investigated. Critically this approach revealed that the grain boundary IMC particles in the as-cast billet were a network of highly inter-connected Fe-IMCs in 3D ( Fig. 8(a)), and the inter-connectivity ( Fig. 8(c)) increased towards the billet centre. High magnification imaging revealed that the as-cast Fe-IMCs were mostly dendritic/petal-like α c -AlFeSi ( Fig. 9(a)), and platelet-like β-AlFeSi ( Fig. 9(b)), consistent with previous work [9,22]. The length of the petals increased towards the billet centre ( Fig.  10(a)), but there was little difference in the longest facet length of the platelets. Similar to the dendritic/petal-like particles, the platelet-like particles showed an ability to branch and adopt complex, extended 3D geometries, as well as appearing to nucleate other platelets (Fig 9(b)).
Unusually, Mg 2 Si was resolved forming on the flat faces and facet-sides of the β-AlFeSi (Fig.   11).
Based on this Mg 2 Si interaction with β-AlFeSi platelets, confirmed by the investigation of many such instances, the Mg 2 Si/β interaction was classified into three types. Type 1 was dendrite-like ( Fig. 11(a)), octahedral-like ( Fig. 11(b)) and cubic-like ( Fig. 11(c)) with Mg 2 Si on the basal face of the β-AlFeSi; Type 2 was seaweed-like Mg 2 Si on the facet-sides ( Fig.   11(d)); and Type 3 was fine needle-like particles on the basal face with two different directional alignments (Fig. 11(e)).
In addition, there were discrete, spherical-like particles that were not connected with the inter-dendritic secondary phase network, which suggested that they were those secondary phases observed inside grains in the earlier 2D sections. These discrete spherical-like particles were ~ 3 μm in size ( Fig. 10(a)) and were either: (i) clusters of Fe-IMCs ( Fig.   12(a)), (ii) clusters of Si particles ( Fig. 12(b)), (iii) cubic-like Mg 2 Si, or (iv) composite agglomerates of one or more of (i) to (iii). For example, Fig. 12(c) shows a composite agglomerate of Mg 2 Si and Fe-IMCs. The Mg 2 Si with the rosette-like Fe-IMC in Fig. 7(b) was revealed to have a more complex octagon-like morphology when viewed after extraction, as shown in Fig. 12 Interestingly, the interconnectivity reduced significantly after homogenisation in all locations (Figs. 8(b) and (c)) and the billet centre retained the longest inter-connectivity length ( Fig.  8(c)). The previously inter-connected Fe-IMCs fragmented during homogenisation into discrete particles, as shown in Fig. 8(b), of the form: (i) partly degraded dendritic-like α c -AlFeSi in Fig. 9(c); (ii) near cylindrical/disc-like α c -AlFeSi in Fig. 9(d); or (iii) spherical-like α c -AlFeSi in in Fig. 9(e). The dendritic/petal-like Fe-IMCs in Fig. 9(c) were similar to the ascast particles in Fig. 9(a), but were now partly dissolved at the corners and at the thin regions of the arms (indicated with arrows). The wrinkle patterns on the surface of as-cast petal-like Fe-IMC particles in Fig. 9(a) were also now absent. The length of the partly degraded/dissolved petals ( Fig. 10(b)) were similar to that in the as-cast billet ( Fig. 10(a)). It was difficult to resolve any needle-or platelet-like Fe-IMC particles after homogenisation, but there was a notable increased incidence of disc-like particles, e.g. Fig. 9(d), which suggested that larger platelet-like β-AlFeSi particles transformed into fine disc-like α c -AlFeSi, and in agreement with the XRD results in Fig. 2(b). The disc-like Fe-IMC particles were mostly connected to at least two or three other discs. The average interconnected length was ~ 20 μm ( Fig.10(b)).
The apparently large proportion of needle-like Fe-IMC particles in the 2D sections of the homogenised billet in Fig. 5(b) was inconsistent with the XRD quantification of the extracted phases and SEM studies. This discrepancy was in part due to the perpendicular preferred orientation of the dendritic-like particles with respect to the sectioned plane that over emphasises their proportion, even when a systematic and monotonic change in particle size distribution is measured (Fig. 6). In other words, these 2D sections not only fail to convey the correct size and morphology of the extended IMCs, but since the particles are generally far from spherical, 2D sections can also lead to incorrect quantification of phase fractions using standard metallographic approaches.
In the homogenised billet and following extraction, the Mg 2 Si particles were mostly type 1, cylinder-like in Fig. 13(a) (arrow marked) and type 2, fine needle-like in Fig. 13 Fig. 14 (a)) and conditions of perfect elemental mixing in the liquid, but no diffusion in the solid (so called "Scheil" conditions) ( Fig. 14(b,c)). Fig. 14(a) shows that the α-Al is the first phase to form on solidification. Because in the experiments and conforming to industrial practice a TiB 2 -based grain refiner was used, the primary Al manifested as equiaxed α-Al grains across the billet section, and as is well-known, the dendritic nature of the grains relates to the solute-induced instability of a planar interface during their growth. further, the nature of which depend critically on the local composition, cooling rate and nucleation density environment. This scenario is also shown schematically in Fig. 15(a) where a Fe-IMC eutectic front develops late in the solidification sequence. Here, "local" refers to the size of the inter-dendritic regions and the length scale of their inter-connectivity.
The resulting secondary phases were clearly resolved in the anodized microstructures and extracted particles (e.g. Fig. 1 and Fig. 8).
Since DC casting provides a range of cooling rates depending on casting conditions and position in the billet, the resulting secondary phases may be thermodynamically stable i.e. predictable by phase fraction calculations with appropriate data, or metastable i.e. unpredictable by calculation, or at least only with some manipulation, as in the case for β-AlFeSi here, which is generally considered metastable although frequently observed in practice. Under the assumptions in Fig. 14 L → α-AlFeSi + α-Al L + α-AlFeSi → β-AlFeSi + α-Al L + Al 13 Fe 4 → β-AlFeSi + α-Al L → β-AlFeSi + α-Al L → Mg 2 Si + α-Al (7) L → β-AlFeSi + Mg 2 Si + Si + α-Al (8) L → α-AlMnSi + Si + α-Al The XRD analyses (Fig. 2) and SEM analyses (Figs. 6 to 12) confirmed the presence of α c -AlFeSi, β-AlFeSi and Mg 2 Si. The largest discrepancy between modelling (Fig. 14(b)) and the possible phases described by reactions 1 to 9 was the absence of Al 13 Fe 4 (equation (1)) in experiment and this may arise either because soon after formation it was rapidly transformed to or -AlFeSi (for which Fig. 14 offers little support), or more likely that it was comparatively difficult to nucleate and grow Al 13 Fe 4 at the required rate so that once or -AlFeSi were formed, it was kinetically preferred for the residual Fe-rich liquid to be transformed as further or -AlFeSi, rather than through the nucleation of a new phase [8,33]. This agrees well with the modelling when the Scheil calculations were performed without the Al 13 Fe 4 phase (Fig. 14(c)). The absence of the predicted much smaller fraction of α-AlMnSi (equation (10)) was possibly due its transformation to other phases, its low fraction, or again due to its suppression by more kinetically favourable solidification reactions. Although the SEM investigations revealed discrete Si particles ( Fig. 12(b)), the absence of Si in XRD analysis was probably because the volume fraction was too low for detection. It should also be noted that by comparison with previous work in [18], the present work suggests that thermodynamic predictions of phase fractions only indicate formation of -AlFeSi, consistent with experiment, when non-equilibrium (no solid diffusion) assumptions were used.
From the as-cast and extracted morphology and length scale of α c -AlFeSi, it was suggested that most formed directly from the liquid via eutectic reaction (3), rather than peritectic reaction (2), because the peritectic reaction implies α c -AlFeSi forming around Al 13 Fe 4 or a composite of both, for which no XRD or SEM evidence was found. By similar consideration, β-AlFeSi is suggested to form directly from the liquid through eutectic reactions (6) and/or (8).
The selection of either  c -AlFeSi or -AlFeSi from the cooling inter-dendritic liquid will be governed by their relative ease of nucleation and growth [2]. For nucleation, the phase with the higher nucleation temperature (small nucleation undercooling) should be preferred initially while the phase with the higher stable growth temperature (higher growth velocity) will be preferred subsequently, if its nucleation is possible. The competition between these two criteria for each phase within the local inter-dendritic environment will determine which phase(s) will be seen in practice. However, as shown in Fig. 2 and Fig. 3 in particular, at all places in the billet at the meso-scale, both  c -AlFeSi and -AlFeSi form. However, at the micro-scale  c -AlFeSi and -AlFeSi did not form together within any particular interdendritic (grain boundary) region. Therefore there is a stochastic, or statistical, aspect to the nucleation of AlFeSi based IMCs, depending on the chemical and physical conditions within the final, isolated solute-rich films around evolving dendrites. Even under exactly the same cooling conditions, some films may favour  c -AlFeSi because of the particular local chemistry, or presence of un-used grain refiners (Fig. 16), oxides or other inclusions, etc., while others will favour -AlFeSi.

(B) Primary α-Al IMC inter-connectivity and spherical droplets
All the as-cast grain boundary secondary phases form from solute-rich grain boundary films around -Al grains. These films have a complex 3D shape, surrounding the evolving dendrites. The secondary phases form from this film and are able to extend, branch and become highly inter-connected over large distances, many times the cell or dendrite spacing or even grain size (as shown in Fig. 15(a)). The IMC particles were comparatively finer and less inter-connected nearer to the billet surface because there was a relatively high grain/liquid surface area per unit volume (smaller microstructural scale) here. Nearer the billet centre, for a constant liquid fraction, the channel width must increase because of a decrease in grain/liquid surface area per unit volume, resulting in coarser particles and greater inter-connectivity (Fig. 8).
Spherical particles within the grains, such as the examples in Fig. 7(b) and 12, form from solute-rich liquid droplets that are formed either by pinching off and isolation of the liquid in the final stages of solidification, or are deposited by solute-rich migrating liquid films around the coarsening dendrites [34], as shown in Fig. 15(a)). Consequently, these spherical droplets were not part of the inter-connected IMC network around grains. The fine structure of these solidified droplets further suggest that these droplets may be subject to relatively large undercoolings before they are able to solidify, at temperature lower than that of to the coarser grain boundary Fe-IMCs eutectic, as illustrated in Fig. 15(a).
There were also spherical composites of Si and Fe-IMC particles, similar to those reported in entrained spherical droplet experiments [33,35] but representing only a significant minority of the IMC particles examined. It is worth remarking that if Si and Fe, as inevitable impurities in Al alloys, could be contrived to solidify in increased fractions of these spherical-like particles or clusters, the undermining effects of more usual, extended Fe-rich IMCs on final properties might be reduced.

(C) Mg 2 Si
The presence of the type 1 (dendritic-like, octahedral-like and cubic-like) and type 2 (seaweed-like) Mg 2 Si particles on the surface of β-AlFeSi suggests that they formed either by eutectic reaction (7) or directly from liquid by reaction (9), where pre-existing β-AlFeSi acts as a substrate to nucleate the Mg 2 Si, potentially via an intermediate compound (occasionally a Ca-rich signal was recorded by EDX from these regions). Recent transmission electron microscopy (TEM) analysis showed common orientation relationship between β-AlFeSi and Mg 2 Si in a 6xxx series Al alloy [36], which supports the role of β-AlFeSi as a nucleation substrate.
Once nucleated, further growth of Mg 2 Si depends on factors similar to those already described for the growth of Fe-IMCs, except the residual solute-rich liquid will tend to have ever higher concentrations of Mg and Si, since Mg 2 Si develops later in the solidification sequence. The octahedral-like ( Fig. 11(b)) and cubic-like (Fig. 11(c)) Mg 2 Si are similar to primary Mg 2 Si in the hyper-eutectic Al-Mg 2 Si alloys [37], suggesting they may form directly from solute-rich liquid by reaction (9).
The different morphologies of type 1 Mg 2 Si again may arise from the local solidification conditions in the residual liquid "pockets" because of interfacial instability due to the buildup of elements rejected by the growing solid phases (in this case Al and Fe may accumulate in the liquid) and physical constraints of growth in the highly restricted film geometry, and in the "holes" inside Mg 2 Si particles ( Fig. 11(b,c)). Type 2 seaweed-like Mg 2 Si along the edges of β-AlFeSi suggested they formed by eutectic reaction (7), and had a similar morphology to that in an Al-Mg 2 Si alloy [38].
Finally, the finest-scale (~0.6μm) Mg and Si -rich needle-like particles on the surface of β-AlFeSi are solid state precipitates of Mg 2 Si that formed after solidification at the interface of pre-existing β-AlFeSi and -Al.

Evolution of Fe-IMCs during homogenisation
The higher proportion of α c -AlFeSi in the homogenised billet arose from the β → α c transformation. Kuijpers et al [11] suggested this transformation is solute diffusion controlled, initiated by the α c -AlFeSi nucleation on the basal face of β-AlFeSi platelets; the β-AlFeSi then dissolves and α c -AlFeSi grows by consuming released Fe, Mn and Si. They also observed an increase in Mn content in α c -AlFeSi particles after homogenisation due to longer diffusion times [39]. In the present study, residual β-AlFeSi attached to newly transformed α c -AlFeSi was not observed and it is estimated that more than 90% of the β-AlFeSi transformed (under these particular homogenisation conditions) instead into discrete, fine, cylinder/disc-like α c -AlFeSi. Further, because disc-like particles were often connected to others e.g. Fig. 9(d), more than one α c -AlFeSi crystal may nucleate from the extended area of one β-AlFeSi platelet.
An attempt was made to investigate if transformed α c -AlFeSi might also nucleate on preexisting α c -AlFeSisince there is presumably good lattice matching and low interfacial energy between them. However, no compelling microstructural evidence could be resolved, and only partly degraded eutectic α c -AlFeSi was evident. It is possible that β-AlFeSi dissolves in one place and re-precipitates as α c -AlFeSi in a different region. If such events were resolvable, in-situ X-ray tomography might be used to study these dynamic effects.
Fig.14 suggested that as-cast Mg 2 Si formed due to non-equilibrium solute segregation to the liquid during solidification and should dissolve more readily than Fe-IMCs during re-heating and isothermal homogenisation. A previous in-situ hot-stage investigation [28] on a 6xxx series Al alloy showed Mg 2 Si particles that were not in contact with Fe-IMCs dissolved first at a lower temperature followed by the particles that were attached to Fe-IMCs. Thus the dissolution of Mg 2 Si that is attached to β-AlFeSi, which was commonly observed will induce a comparatively high local Si concentration, which can be expected to affect the kinetics of the β → α transformation, and perhaps the transformation mechanism itself.
The relatively coarse, equiaxed Mg 2 Si at grain boundaries suggests that some eutectic Mg 2 Si does not dissolve but firstly tends to spheriodise to reduce the interfacial area [40,41].
Because of the dissolution and degradation of grain boundary secondary phases, mainly Fe-IMCs, IMC inter-connectivity was also notably reduced after homogenisation (Fig. 8(c)), with inter-connectivity only maintained by those particles that retained the larger petals around primary α-Al grains, as shown in Fig. 15(b). Similarly the platelet-like interconnected β-AlFeSi breaks into fine disc-like particles during the β-AlFeSi to α c -AlFeSi transformation leaving finer disc-like particles around grains, as also shown in Fig. 15(b). Fe-IMCs formed from spherical droplets in side primary Al persist after homogenisation, but the as-solidified Mg 2 Si or Si is dissolved.
Based on these insights, summarised schematically in the Fig. 15, it is suggested that because β transformation to α c leads to a dramatic reduction in inter-connectivity, having a relatively high β-AlFeSi and spherical Fe-IMCs fraction in the solidified billet can be advantageous as it will result in fine, less connected α c -AlFeSi particles after homogenisation. A high fraction of inter-connected α c -AlFeSi in the as-cast billet is harder to break-up (more thermodynamically stable), requiring longer homogenisation temperatures and times. So long as the β to α c AlFeSi transformation is completed, finer intermetallic particles in the billet before downstream deformation process will improve both the deformation efficiency and final property of the product [15,18].

Conclusions
The as-cast and homogenised microstructure of a DC cast grain refined 6063 Al alloy was studied across the billet cross-section. The primary Al grain size, dendritic arm spacing (DAS), and secondary phase particle size increased from the billet edge to the centre of the as-cast billet. When the grain size reduced, the Fe-IMCs size interconnectivity also reduced, and vice-versa. Fe was present as majority α c -AlFeSi and minority β-AlFeSi IMCs in the ascast billet, but their relative proportions were not sensitive to position. β-AlFeSi nucleated Mg 2 Si during solidification, and there were additional solidified spherical droplets of further Fe-IMCs as well as Mg 2 Si, Si and composite particulates within primary -Al grains.
The homogenised billet showed little grain/dendrite coarsening and a marked elimination of inter-dendritic secondary phases. The Fe-IMC interconnectivity was reduced everywhere, most markedly at the billet surface. Of the other as-cast particles, only spherical-like Fe-IMCs persisted within the -Al grains after homogenisation. Overall, the results showed that while homogenisation treatments are effective, nonetheless the initial as-cast microstructure, which is influenced by the solidification conditions, exerted a strong influence on the final homogenised microstructure, affecting the size, morphology, type, β to α transformation rate and interconnectivity of the various IMCs.      The Fe-IMC equivalent particle diameter increased from (a) 15mm from the billet edge to (b) 90mm from the billet edge for the as-cast and homogenised billets. Homogenisation resulted in smaller particles with a narrower size distribution. Measurements were made from the BSE-SEM images of the metallography samples, shown as insets.                 As-cast (90mm) Homogenised (90mm) Fig. 6 The Fe-IMC equivalent particle diameter increased from (a) 15mm from the billet edge to (b) 90mm from the billet edge for the as-cast and homogenised billets. Homogenisation resulted in smaller particles with a narrower size distribution. Measurements were made from the BSE-SEM images of the metallography samples, shown as insets.