De-Ironing of Aluminium Alloy Melts by High Shear Melt Conditioning Technology: An Overview

: The main problem of recycling aluminium scrap is the gradual accumulation of impurities, especially iron, which tend to form undesired intermetallic compounds that affect the integrity and the mechanical performance of the castings. In this paper, we aim to provide an overview on the topic of iron removal from aluminium melts through primary intermetallic precipitation and the progress made during the LiME Hub project to understand the process and to develop a more efﬁcient procedure. We cover both thermodynamic analysis and experimental validation. We found that high shear melt conditioning technology enhances the typically slow nucleation and growth of the dense primary intermetallics, speeding up their sedimentation and allowing a faster removal of Fe from the melt by simple gravity sedimentation. It also promotes the formation of smaller and more compact Fe-rich intermetallics, allowing an increased volume fraction recovery and mitigating their effect of being present in the ﬁnal castings. The technology is not limited to batch processing, with a 90% efﬁciency, but can also be applied to continuous melt treatment of aluminium scrap, with currently 60% efﬁciency, and could be combined with other solid–liquid separation techniques to increase the puriﬁcation efﬁciency even more.


Introduction
The interest in aluminium alloys from the automotive and aerospace sectors has been continuously increasing in the last decades due to the weight reduction, compared to steel, and the high-performance possibilities that these materials offer [1]. However, the main source of aluminium alloys for industry is still mainly primary production [2,3].
Recycling of aluminium alloy scrap to produce the ingots to be used in foundries is beneficial from an economic and environmental point of view because it requires less energy and produces less CO 2 emissions compared with the manufacturing of primary aluminium ingots [4]. However, the problem of repeated aluminium scrap re-melting is the gradual accumulation of impurities, in particular Fe. The solubility of Fe is high in molten aluminium, but it is very low in the solid state. During solidification, this excess of Fe combines with the Al and other elements in the alloy and tends to form a variety of intermetallic compounds [5]. The shape and size of these intermetallics depends on multiple factors including the alloy composition, solidification process and thermal treatments, but overall, they have a negative effect on the microstructure and the mechanical properties of the aluminium alloy castings [6][7][8].
Typical solutions in industry to overcome this problem include the downgrading of the material to less demanding casting applications or the dilution with expensive primary aluminium, but these are not long-term solutions and none of them solve the problem of iron accumulation [9]. In the last decades, several approaches have been developed and SF = (1 × wt.%Fe) + (2 × wt.%Mn) (3 × wt.%Cr) (1) However, the addition of secondary elements such as Mn or Cr, apart from being an extra step in the process with its associated costs, should be kept as low as possible to maintain a sustainable recyclability of the material and prevent the accumulation of excess in the alloy exceeding the alloy limits or forming other intermetallics [31]. The search for alternatives to improve the method becomes necessary.
The use of high shear melt conditioning (HSMC) technology [32][33][34], developed within the Brunel Centre for Advanced Solidification Technology (BCAST) at Brunel University London, can overcome some of these problems. It can fragment and disperse the large oxide films and clusters into very fine and uniformly distributed individual particles [35,36], thereby enhancing the nucleation of intermetallic phases [37], and at the same time maximize the effect of the alloy chemistry so no additions in excess are required to enhance the formation of Fe-rich phases. In this paper, we present an overview of the progress made during the LiME Hub program to increase our understanding of the P-Fe-IMC precipitation-separation method and also to evaluate the effect of the developed HSMC technology to improve the efficiency of iron removal from aluminium melts, so the process becomes more practical and sustainable to be implemented in the recycling industry.

Aluminium Alloy Analysis by Thermodynamic Modelling
The purification of Al melts based on the precipitation of P-Fe-IMC intermetallics is a challenging process because it requires fine control of the temperature for the formation of the primary phases and an understanding of the limits on iron removal associated with each alloy composition. To evaluate whether an alloy with high Fe content will present P-Fe-IMCs, and the temperature range at which they form, the use of computational thermodynamic analysis tools becomes necessary. The Pandat software (Pandat with Pan Al Database Version 2021, CompuTherm LLC, Middleton, WI, USA) [38,39] was used during the investigations.  Figure 1d). The first three, wrought alloys, do not exhibit the presence of primary Fe-rich intermetallic compounds for an Fe level below 1%Fe, while the Fe limits in these alloys are typically much lower ( Table 1). The maximum Fe level in recycled material from these alloys is typically far below 1%, and when it reaches higher values of Fe, it also exhibits the presence of other elements, such as Si, Mn, Cu and Zn, above the specification limits [40]. Therefore, the purification method based on P-Fe-IMC precipitation is not of practical use for wrought aluminium alloys as it is not able to operate at the relatively low Fe levels that these alloys require. The addition of Mn or Cr also does not work in this case because the amounts needed would also exceed the maximum allowed in the alloy specifications. Currently, the focus of our research is also on the applicability of the methods on real aluminium scrap. In a recent publication [49], we covered the use of the de-ironing method on aluminium alloy recovered from the Zorba Cast fraction of shredded end-oflife vehicles, with a composition like the LM27 alloy, and the preliminary results look promising.  In practice, the P-Fe-IMC precipitation method turns out to be effective mainly to purify relatively high Si Al-Si cast alloys. Besides the effect of Si on lowering the temperature at which the α-Al phase appears, which increases the chances of other phases nucleating beforehand, for a given Fe content in the alloy there is a minimum Si level to make it thermodynamically possible to observe the primary β_AlFeSi phases that could settle in the melt [41,42]. Very low values of Si in Al cast alloy, such as, for example, the case of the A201 alloy, leads to the same problem mentioned before for the wrought alloys.
Because of this, the investigations on this method focused on alloys such as A380/LM24 or LM27, for which the Si content is between 6 to 9% and the tolerance to Mn relatively high. Figure 1d shows how the A380-0.2Mn aluminium cast alloy exhibited primary Fe-rich intermetallic compounds at much lower Fe content, which correspond to the α_Al 15 (FeMn 3 )Si 2 phase when the Fe level is between 0.35 and 0.95% and a mixture of the α_Al 15 (FeMn 3 )Si 2 and the β_Al 5 FeSi phases above 1%Fe, The presence of Mn in this alloy plays an important role in the appearance of such P-Fe-IMCs and the effectiveness of the purification method [43,44] as it will be discussed later.
The initial studies of the A380-xMn-yFe alloy system to understand the potential and limitations of the purification method were conducted computationally, although with some experimental validation [31,44]. Later studies involved the use of melt conditioning to study its effect in the separation process [45][46][47] as well as the effect of solidification processing conditions [48]. Table 2 presents an example of two of the alloys studied, and for which we will be presenting some key results later in the paper. These values aim to reproduce typical compositions that could be found in this alloy system when dealing with real scrap [5]. To this end, the Al-80Fe and the Al-60Mn master alloys were used to increase the Fe and Mn levels in the aluminium melts. Currently, the focus of our research is also on the applicability of the methods on real aluminium scrap. In a recent publication [49], we covered the use of the de-ironing method on aluminium alloy recovered from the Zorba Cast fraction of shredded end-of-life vehicles, with a composition like the LM27 alloy, and the preliminary results look promising. Figure 2 shows a schematic illustration of the high shear melt conditioning technology (HSMC) prototype developed at BCAST [34] during the project for the purification of aluminium alloy melts. It consists of three parts. The first part of the setup is the processing chamber and it is the core of the technology. It consists of a heated vessel containing ã 6 kg graphite crucible with a baffle to separate the vessel into two sections, one (~5 kg) for receiving the molten scrap and another (~1 kg) for melt conditioning using a rotor-stator unit [33]. The size and design of the chamber and the processing section account for maximizing the level of melt shear rate and mass flow as these are restricted to the close vicinity of the mixing head [50]. The implementation of the technology on large vessels requires an increase in the size of the unit and still further analysis, by computational fluid dynamic (CFD) modelling [51] and experimental validation [52], on the fluid flow and key features of the high shear processing. At this stage, the oxide agglomerates in the melt are finely dispersed because of the intensive melt shearing [35][36][37] and the composition and temperature are quickly homogenized by the strong macroscopic flow provided by the rotor-stator unit. dynamic (CFD) modelling [51] and experimental validation [52], on the fluid flow and key features of the high shear processing. At this stage, the oxide agglomerates in the melt are finely dispersed because of the intensive melt shearing [35][36][37] and the composition and temperature are quickly homogenized by the strong macroscopic flow provided by the rotor-stator unit.

The High Shear Melt Conditioning Technology Setup for De-Ironing
The melt in the processing chamber is kept at high temperature, above the formation of primary intermetallic phases (TL_P-Fe-IMCs), to prevent early sedimentation. For the case of the A380 alloy (Figure 1d), this implies at least 650 °C for an Fe content up to 2%. The rotor-stator unit is operated at high speed, typically at 3000 rpm and for a period of at least 1 min [53]. The time needed to ensure an effective processing increases with the vessel size [54], as more melt needs to be treated. However, the mass flow through the rotor stator depends linearly on the rotor speed [55] and also changes with the rotor-stator configuration [56], so the minimum processing time can be controlled by varying these parameters. Once the melt is processed, there are two options to operate with the technology. On the one hand, the chamber can be tilted to pour the melt to another crucible in a furnace The melt in the processing chamber is kept at high temperature, above the formation of primary intermetallic phases (T L_P-Fe-IMCs ), to prevent early sedimentation. For the case of the A380 alloy (Figure 1d), this implies at least 650 • C for an Fe content up to 2%. The rotor-stator unit is operated at high speed, typically at 3000 rpm and for a period of at least 1 min [53]. The time needed to ensure an effective processing increases with the vessel size [54], as more melt needs to be treated. However, the mass flow through the rotor stator depends linearly on the rotor speed [55] and also changes with the rotorstator configuration [56], so the minimum processing time can be controlled by varying these parameters.
Once the melt is processed, there are two options to operate with the technology. On the one hand, the chamber can be tilted to pour the melt to another crucible in a furnace where is to be held and further processed as required (batch processing). On the other hand, more aluminium melt can be poured into the inlet of the processing chamber (continuous processing), so there is an outlet flow that goes directly into a cooling launder, where the melt temperature is reduced as it flows down. The function of this second stage of the setup is to decrease the temperature from above the liquidus for the appearance of primary phases down to a temperature above the liquidus for the α-Al phase, i.e., cooling the melt but not allowing it to start solidifying. It is during this stage that the nucleation and growth of the Fe-rich intermetallic compounds are supposed to occur. The design and configuration of this stage need to be carried out specifically for the temperature range that needs to be covered, which is intrinsic to the aluminium alloy composition under consideration. To this end, the use of CALPHAD tools, such as the Pandat software mentioned in previous section, becomes necessary as shown in Figure 1.
From the launder, the melt is transferred into the last stage of the setup, denoted as the separation chamber; a heated vessel made from a graphite crucible like the processing chamber. The purpose of this chamber is to keep the melt at a controlled temperature and prevent the solidification, by operating at least 5 to 15 • C above the liquidus temperature for the formation of the α-Al phase (T L_α-Al ). For the case of the A380/LM24 alloy, this temperature is typically around 600 • C but increases to 610 • C for the LM27 alloy sys-tem [49] due to the lower Si content. Isothermal holding allows the intermetallics formed in the launder to settle directly to the bottom when they reach the separation chamber. The main advantage of this configuration is that "clean" material, with low levels of Fe, can be directly collected from the top of the sedimentation chamber, or from the outlet when operated in continuous mode. The chamber includes baffles to ensure that the incoming particles remain inside, quickly settling, and are not carried out with the outlet flow. The particles accumulated in the bottom of the chamber can be either drained in situ or removed once the processing has finished.  Figure 1d. It can be observed that as Fe increased so did the temperature at which the primary phases appeared. This widens the temperature window for the formation of the primary phases, as the α-Al phase forms at 595 • C independently of the initial %Fe. For example, for an initial 1%Fe in the alloy the primary intermetallics, mostly α_Al 15 (FeMn) 3 Si 2 , are expected to form in a narrow 25 • C range, between 620 and 595 • C, while for an initial 2%Fe, a combination of α_Al 15 (FeMn) 3 Si 2 and β_Al 5 FeSi will form on a wider 51 • C range, between 646 and 595 • C. For an initial level of 3%Fe, a more complex mixture of P-Fe-IMCs (including also the α_AlFeSi phase [41]) is expected to form in a 77 • C range, from 667 to 595 • C. It is clear that as Fe level increases it becomes more likely to observe the presence of P-Fe-IMCs and easier to operate on a sufficiently wide temperature window.

Results and Discussion
Metals 2022, 12, x FOR PEER REVIEW 7 of 19 falls below the maximum limit of the A380 alloy (Table 2) and opens the possibility of processing contaminated A380 alloy, i.e., Fe > 1.3%, without the need of Mn addition. In practice, the experimental values tended to be a bit higher (~1%Fe) when holding in this temperature range, because the sedimentation process required a long time, but even then, they were still within the alloy specifications [23].

The Effect of Mn Additions on an Fe Contaminated A380 Aluminium Alloy
Despite the possibility of avoiding the use of Mn additions for the purification of the A380-0.2Mn alloy to an acceptable Fe level (0.8-1.0%), if an Fe level below 0.8% is required, then the melt needs the addition of Mn to promote the formation of more Fe-rich intermetallics. To help understand this, Figure 4a shows how the phase diagram in Figure 3a varies with the increment of Mn in the alloy. The variation of Mn does not affect the α-Al phase formation temperature and, therefore, the holding temperature for the formation and separation of the intermetallic phases can be kept the same (600-610 °C).
Mn addition widens the temperature range for the precipitation of the primary phases by increasing the temperature at which they start appearing. For the case of an A380 alloy with 1%Fe, this happened at 620 °C when the Mn level was around 0.2%, at 634 °C for 0.5%Mn, at 652 °C for 1%Mn and at 687 °C when the Mn content was 2%, thus, increasing the temperature window from ~20 °C to ~90 °C. Moreover, increasing Mn in the alloy facilitates the appearance of the α_Al15(FeMn)3Si2 phase to the detriment of the As the primary phases formed, the concentration of Fe in the remaining liquid alloy decreased. Figure 3b shows this variation as a function of the alloy temperature for three different initial Fe contents. Vertical and horizontal lines have been included to mark the key points in the graph. For high temperatures, the differences are clear, but the final Fe level expected in the liquid was almost the same in the three cases when the melt was at a holding temperature below 610 • C. The minimum value possible, at T = T Lα-Al , was 0.66%Fe, although holding at that low temperature would not be practical because small variations would lead to partial solidification of the α-Al and the intermetallic movement would be hindered. For the purification method to be more effective, the melt temperature should be held at least between 600 and 610 • C. In this case, the expected Fe level in the liquid would range between 0.74 and 0.88%, with an average value of 0.81%. This value falls below the maximum limit of the A380 alloy (Table 2) and opens the possibility of processing contaminated A380 alloy, i.e., Fe > 1.3%, without the need of Mn addition. In practice, the experimental values tended to be a bit higher (~1%Fe) when holding in this temperature range, because the sedimentation process required a long time, but even then, they were still within the alloy specifications [23].

The Effect of Mn Additions on an Fe Contaminated A380 Aluminium Alloy
Despite the possibility of avoiding the use of Mn additions for the purification of the A380-0.2Mn alloy to an acceptable Fe level (0.8-1.0%), if an Fe level below 0.8% is required, then the melt needs the addition of Mn to promote the formation of more Ferich intermetallics. To help understand this, Figure 4a shows how the phase diagram in Figure 3a varies with the increment of Mn in the alloy. The variation of Mn does not affect the α-Al phase formation temperature and, therefore, the holding temperature for the formation and separation of the intermetallic phases can be kept the same (600-610 • C).   Figure 4b shows the expected effect of Mn addition on the Fe and Mn levels in the liquid phase when the alloy is maintained at 610 °C. In this particular case, the Fe level can be theoretically reduced down to 0.42%Fe when Mn is added up to 1%, with experiments showing good agreement with an Fe decrease down to 0.47%. In the process, Mn is also consumed in the formation of the α_Al15(FeMn)3Si2 phase and the final Mn level in the liquid is also much lower than the initial addition level. However, experimental trials with these additions showed that, in practice, the remaining Mn level in the liquid phase tended to be underestimated by the software predictions. Experimental content after purification of a A380-1Mn-1Fe alloy showed a Mn content of 0.32% [44] when the Pandat software predicted 0.24%Mn. This highlights the importance of an adequate thermal analysis consideration and not using Mn in excess to maintain a sustainable iron removal process. Furthermore, it has been reported that the increase in Mn in the alloy balances the content of Fe and Mn in the α_Al15(FeMn)3Si2 phase towards more Mn and less Fe [61], thus limiting the Fe removal efficiency [58].

Effect of HSMC on the Nucleation Kinetics
The nucleation of the primary Fe-rich intermetallics exhibits a high undercooling, which facilitates the suppression of their formation if the cooling rate is high enough [62]. Mn addition widens the temperature range for the precipitation of the primary phases by increasing the temperature at which they start appearing. For the case of an A380 alloy with 1%Fe, this happened at 620 • C when the Mn level was around 0.2%, at 634 • C for 0.5%Mn, at 652 • C for 1%Mn and at 687 • C when the Mn content was 2%, thus, increasing the temperature window from~20 • C to~90 • C. Moreover, increasing Mn in the alloy facilitates the appearance of the α_Al 15 (FeMn) 3 Si 2 phase to the detriment of the β_Al 5 FeSi phase. When the Mn/Fe ratio in the alloy is below 0.5 we can expect both α and β phases with dominance of the β phase. When Mn/Fe exceeds 0.5 the α phase becomes more dominant and when Mn/Fe > 1 the β_Al 5 FeSi phase is suppressed and only the α_Al 15 (FeMn) 3 Si 2 phase is expected to appear during the solidification [57,58]. The α_Al 15 (FeMn) 3 Si 2 phase is more compact and rounded than the β phase and, therefore, an adequate Mn addition can be used to control the microstructure of the castings and improve the mechanical properties [58][59][60]. However, the change of morphology of the P-Fe-IMCs with the increased Mn/Fe ratio also facilitates their separation as the compact α particles become larger and can settle faster. Figure 4b shows the expected effect of Mn addition on the Fe and Mn levels in the liquid phase when the alloy is maintained at 610 • C. In this particular case, the Fe level can be theoretically reduced down to 0.42%Fe when Mn is added up to 1%, with experiments showing good agreement with an Fe decrease down to 0.47%. In the process, Mn is also consumed in the formation of the α_Al 15 (FeMn) 3 Si 2 phase and the final Mn level in the liquid is also much lower than the initial addition level. However, experimental trials with these additions showed that, in practice, the remaining Mn level in the liquid phase tended to be underestimated by the software predictions. Experimental content after purification of a A380-1Mn-1Fe alloy showed a Mn content of 0.32% [44] when the Pandat software predicted 0.24%Mn. This highlights the importance of an adequate thermal analysis consideration and not using Mn in excess to maintain a sustainable iron removal process. Furthermore, it has been reported that the increase in Mn in the alloy balances the content of Fe and Mn in the α_Al 15 (FeMn) 3 Si 2 phase towards more Mn and less Fe [61], thus limiting the Fe removal efficiency [58].

Effect of HSMC on the Nucleation Kinetics
The nucleation of the primary Fe-rich intermetallics exhibits a high undercooling, which facilitates the suppression of their formation if the cooling rate is high enough [62]. This is obviously not desired as the aim is to produce the P-Fe-IMCs as early as possible, so they have more time to settle. Applying HSMC enhances the formation of the primary intermetallic by reducing the undercooling and facilitating their earlier formation [36].
To study the solidification behavior, batches of melts (~1 kg) were held at 720 • C and then processed with the HSMC at 3000 rpm for 10 min. The conditioned melt was then poured into a set-up consisting of a preheated small crucible (~0.3 kg) covered with thermal insulation to control the cooling rate [63]. A 0.5 mm K-type thermocouple was immersed in the melt right after pouring and fixed at the center of the melt. The temperature was acquired through a datalogger at 100 Hz. Sampling without HSMC was also collected as a reference. Figure 5 shows some examples of cooling curves for the A380 alloy with different Fe and Mn contents and solidified and different cooling rates by varying the thermal insulation around the crucible. The cooling rate in each case was calculated from the slope of the curve at the beginning of the process, in the range from 700 to 650 • C. It can be observed that the curves present an inflex point somewhere between 650 and 600 • C, which corresponds to the nucleation of the primary intermetallics [61,64].
For the A380 alloy with 2%Fe and 0.2%Mn (Figure 5a), the primary phases are expected to nucleate at 646 • C. However, the natural undercooling moves this point to 619 • C for the reference material when cooled at −0.15 K/s, thus exhibiting a 27 • C difference. The nucleation temperature for the primary intermetallic phase after HSMC increases to 635 • C, with only 11 • C undercooling. Similar low undercooling can be found in the reference melt but when using a much lower cooling rate [61], which highlights the difficulties for practical implementation and the need for enhancement methods.
These differences become clearer as the cooling rate increases, so the formation temperature for the P-Fe-IMCs decreases [57,58,62]. For the A380 alloy with 1%Fe and 1%Mn, the primary α_Al 15 (FeMn) 3 Si 2 phase was expected to form at 652 • C but it was observed at 624 • C for the reference melt when cooled at −0.3 K/s (Figure 5b) and at 603 • C when cooled much faster at −3.2 K/s (Figure 5c). After the use of HSMC, the nucleation temperatures increased to 642 and 622 • C, respectively, thus reducing the undercooling more than 15 • C regardless of the cooling rate, making it more practical for industrial implementation. An earlier formation gives the Fe-rich intermetallics more time to grow and start settling in the melt during the cooling process before reaching the holding temperature.
pected to nucleate at 646 °C. However, the natural undercooling moves this point to 619 °C for the reference material when cooled at −0.15 K/s, thus exhibiting a 27 °C difference. The nucleation temperature for the primary intermetallic phase after HSMC increases to 635 °C, with only 11 °C undercooling. Similar low undercooling can be found in the reference melt but when using a much lower cooling rate [61], which highlights the difficulties for practical implementation and the need for enhancement methods. These differences become clearer as the cooling rate increases, so the formation temperature for the P-Fe-IMCs decreases [57,58,62]. For the A380 alloy with 1%Fe and 1%Mn, the primary α_Al15(FeMn)3Si2 phase was expected to form at 652 °C but it was observed at 624 °C for the reference melt when cooled at −0.3 K/s (Figure 5b) and at 603 °C when cooled much faster at −3.2 K/s (Figure 5c). After the use of HSMC, the nucleation temperatures increased to 642 and 622 °C, respectively, thus reducing the undercooling more than 15 °C regardless of the cooling rate, making it more practical for industrial implementation. An earlier formation gives the Fe-rich intermetallics more time to grow and start settling in the melt during the cooling process before reaching the holding temperature.  Figure 6 shows the effect of HSMC on the microstructure of the A380 alloy containing 0.2%Mn and 2%Fe (Alloy 1) in the solidified samples before and after HSMC. For the reference material, without HSMC, the micrograph in Figure 6a shows both types of intermetallics, but more fractions of the large and elongated β-phase (309 ± 109 µm) and only a few branched α-phase intermetallics (32 ± 16 µm) were formed. After the application of HSMC (Figure 6b), the β-phase became refined in size (224 ± 113 µm) and became less dominant and more compact and bigger α-phase intermetallics (53 ± 25 µm) could be observed in the microstructure [37]. The circularity of the particles (Circ.) can be calculated by Equation (2), where A is the area of the particle and P its perimeter. For the case of Figure 6, the circularity changed from 0.10 ± 0.04 to 0.17 ± 0.05 for the β-phase and from 0.71 ± 0.26 to 0.74 ± 0.17 for the α-phase.

Effect of HSMC on the Size and Shape of Fe-IMC
Circ. = 4π A/P 2 (2) a few branched α-phase intermetallics (32 ± 16 μm) were formed. After the application of HSMC (Figure 6b), the β-phase became refined in size (224 ± 113 μm) and became less dominant and more compact and bigger α-phase intermetallics (53 ± 25 μm) could be observed in the microstructure [37]. The circularity of the particles (Circ.) can be calculated by Equation (2), where A is the area of the particle and P its perimeter. For the case of Figure 6, the circularity changed from 0.10 ± 0.04 to 0.17 ± 0.05 for the β-phase and from 0.71 ± 0.26 to 0.74 ± 0.17 for the α-phase.

Circ. = 4 /
(2) Figure 7 shows the effect of HSMC on the microstructure of the A380 alloy containing 1%Mn and 1%Fe (Alloy 2) in the solidified samples before and after HSMC. In this case, for the reference material without HSMC (Figure 7a), the presence of high level of Mn ensured the formation of only the primary α_Al15(FeMn)3Si2 phase, with an average size of 137 ± 15 μm. The application of HSMC increased the number density of the intermetallics (Figure 7b). It also reduced their size to 102 ± 21 μm but increased their circularity from 0.72 ± 0.15 before HSMC to 0.81 ± 0.12 after HSMC. This supports the idea of the dispersed oxide films acting as nucleation sites for the Fe-rich intermetallics [65]. The  Figure 7 shows the effect of HSMC on the microstructure of the A380 alloy containing 1%Mn and 1%Fe (Alloy 2) in the solidified samples before and after HSMC. In this case, for the reference material without HSMC (Figure 7a), the presence of high level of Mn ensured the formation of only the primary α_Al 15 (FeMn) 3 Si 2 phase, with an average size of 137 ± 15 µm. The application of HSMC increased the number density of the intermetallics (Figure 7b). It also reduced their size to 102 ± 21 µm but increased their circularity from 0.72 ± 0.15 before HSMC to 0.81 ± 0.12 after HSMC. This supports the idea of the dispersed oxide films acting as nucleation sites for the Fe-rich intermetallics [65]. The nucleation of the P-Fe-IMCs occurs progressively in bursts during cooling [66] due to the low content of active nuclei available in the melt, which leads to large and irregularly shaped particles. The dispersion of the oxide biofilms naturally occurring in the melt facilitates the presence of more active nuclei, which results in a more explosive nucleation, with a lower growth but leading to a more compact final shape [67]. Figure 8 shows the level of Fe along the vertical profile for the A380-0.2Mn-2Fe alloy, with and without HSMC, solidified inside a crucible from 700 • C directly to room temperature at a similar cooling rate (−0.15 K/s). A dashed vertical line has been included to indicate the initial Fe content and a dotted line and shadowed region to indicate the expected level of Fe in the melt at 605 ± 5 • C, according to the values shown in Figure 3b. A gradient was observed in both cases, which highlights the formation and partial sedimentation of the primary phases during the cooling. This gradient was clearly higher on the HSMC material because the nucleated particles were less elongated (Figure 6b), and their earlier formation (Figure 5a) gave them time to settle during the cooling process while the alloy was still in the liquid. nucleation of the P-Fe-IMCs occurs progressively in bursts during cooling [66] due to the low content of active nuclei available in the melt, which leads to large and irregularly shaped particles. The dispersion of the oxide biofilms naturally occurring in the melt facilitates the presence of more active nuclei, which results in a more explosive nucleation, with a lower growth but leading to a more compact final shape [67].

Effect of HSMC on the Sedimentation of the P-Fe-IMCs
3.3.1. Effect during Direct Solidification to Room Temperature Figure 8 shows the level of Fe along the vertical profile for the A380-0.2Mn-2Fe alloy, with and without HSMC, solidified inside a crucible from 700 °C directly to room temperature at a similar cooling rate (−0.15 K/s). A dashed vertical line has been included to indicate the initial Fe content and a dotted line and shadowed region to indicate the expected level of Fe in the melt at 605 ± 5 °C, according to the values shown in Figure 3b. A gradient was observed in both cases, which highlights the formation and partial sedimentation of the primary phases during the cooling. This gradient was clearly higher on the HSMC material because the nucleated particles were less elongated (Figure 6b), and their earlier formation (Figure 5a) gave them time to settle during the cooling process while the alloy was still in the liquid. Based on the initial and final values reached from this analysis and the expected results from the Pandat software analysis, we can define a process efficiency. On the one hand we have a purification efficiency with respect to composition (ηComp) that can be defined using Equation (3), where Cinitial is the initial Fe level in the melt, Climit is the expected Fe level at 605 °C and Ctop refers to the Fe level in the upper part of the sample, obviously lower than Cinitial and different in each case (with and without HSMC). This efficiency increases from 26 to 46% after HSMC. have a purification efficiency with respect to composition (η Comp ) that can be defined using Equation (3), where C initial is the initial Fe level in the melt, C limit is the expected Fe level at 605 • C and C top refers to the Fe level in the upper part of the sample, obviously lower than C initial and different in each case (with and without HSMC). This efficiency increases from 26 to 46% after HSMC.
On the other hand, we can define an efficiency based on the volume fraction of material purified, η Vol , by using Equation (4), where V total is the sample volume and V clean is the fraction of material with a Fe level below the initial content. This efficiency increases from 58 to 67% after HSMC. Figure 9a shows the Fe level along the vertical profile when a reference A380-1Mn-1Fe alloy melt was cooled from 700 • C at a −0.3 K/s cooling rate and poured inside a cylindrical temperature c ontrolled vessel to be kept at 605 ± 5 • C for different times to allow the P-Fe-IMCs to settle. The sedimentation was slow, and the process took up to 10 to 15 min to be completed. The refined melt at the top of the vessel showed a level of 0.47%Fe, which is very close to the thermodynamic calculations using Pandat software (C limit = 0.42%Fe). This gives an efficiency in terms of composition of η Comp = 91%. The concentration at the bottom obviously increased as the P-Fe-IMCs accumulated with time. The corresponding vertical profiles of the melt that were subjected to HSMC before being cooled at same rate and poured into the holding vessel are shown in Figure 9b. In this case, the composition gradient from the top to the bottom of the vessel was more pronounced and the sedimentation process was completed much faster. The purification efficiency in terms of volume fraction was similar in both cases, with a η Vol = 75% before and η Vol = 77% after HSMC. of volume fraction was similar in both cases, with a ηVol = 75% before and ηVol = 77% after HSMC.

Effect during Holding at Constant Temperature in the Liquid
The velocity (vs) at which solid spherical particles settle in the liquid alloy can be expressed by Equation (5) (Stoke's Law) [68], where g is the acceleration due to gravity, ρsol and ρliq are the density of the solid particle and the fluid respectively, φ is the diameter of the particle and μ is the viscosity of the liquid Al (1.4 × 10 -3 Pa s [44]).
The larger the form of the intermetallic particles, the faster they settle in the melt, and therefore the refinement of particle size produced by HSMC should in fact decrease the sedimentation if we stick to the estimation from Equation (5). However, the shape of the particle also needs to be considered in this case. For the case of non-fully spherical solid particles, a recent study [69] has shown that the settling velocity (v * ) can be approached by Equation (6), where ge, Kφ, Kf and vL are correction factors accounting for the particle shape and the fluid flow effects on the particle movement. The earlier formation of the intermetallics after HSMC (Figure 5b) and the fact they are more rounded and compact (Figures 6b and 7b), facilitates the sedimentation process and explains the higher settling velocity in the HSMC melts.

Implementation of HSMC De-Ironing in Continuous In-Line Processing
The operation of the technology in continuous or semi-continuous mode is more practical for industrial purposes as it facilitates the use of smaller configurations for a faster processing of the alloy regardless of the quantity of melt required. However, the implementation in continuous process requires fine control of the setup (Figure 2) to guarantee an adequate temperature gradient in the launder to start forming the intermetallics and a precise control of the melt temperature in the sedimentation chamber as the melt flows through. Figure 10 presents an example of the temperature gradient along the different parts of the prototype system during a test with a A380-1Mn-1Fe alloy melt at a melt flow rate The velocity (v s ) at which solid spherical particles settle in the liquid alloy can be expressed by Equation (5) (Stoke's Law) [68], where g is the acceleration due to gravity, ρ sol and ρ liq are the density of the solid particle and the fluid respectively, φ is the diameter of the particle and µ is the viscosity of the liquid Al (1.4 × 10 −3 Pa s [44]).
The larger the form of the intermetallic particles, the faster they settle in the melt, and therefore the refinement of particle size produced by HSMC should in fact decrease the sedimentation if we stick to the estimation from Equation (5). However, the shape of the particle also needs to be considered in this case. For the case of non-fully spherical solid particles, a recent study [69] has shown that the settling velocity (v * s ) can be approached by Equation (6), where g e , K φ , K f and v L are correction factors accounting for the particle shape and the fluid flow effects on the particle movement. The earlier formation of the intermetallics after HSMC (Figure 5b) and the fact they are more rounded and compact (Figures 6b and 7b), facilitates the sedimentation process and explains the higher settling velocity in the HSMC melts.

Implementation of HSMC De-Ironing in Continuous In-Line Processing
The operation of the technology in continuous or semi-continuous mode is more practical for industrial purposes as it facilitates the use of smaller configurations for a faster processing of the alloy regardless of the quantity of melt required. However, the implementation in continuous process requires fine control of the setup (Figure 2) to guarantee an adequate temperature gradient in the launder to start forming the intermetallics and a precise control of the melt temperature in the sedimentation chamber as the melt flows through. Figure 10 presents an example of the temperature gradient along the different parts of the prototype system during a test with a A380-1Mn-1Fe alloy melt at a melt flow rate of 4 Kg/min. Temperatures were taken in the processing chamber, at the beginning and end of the launder and inside the sedimentation chamber. Some fluctuations in temperature along the launder can be observed due to a still ineffective melt flow control in the current prototype that would be solved with an automatic pouring system. Apart from that, the temperature control was quite good in the processing and the sedimentation chambers. It was above the liquidus for the formation of P-Fe-IMCs in the processing chamber, to prevent precipitation of the intermetallics in this stage, and fell within the range of formation range along the launder to finally be maintained in liquid above α-Al formation temperature. end of the launder and inside the sedimentation chamber. Some fluctuations in temperature along the launder can be observed due to a still ineffective melt flow control in the current prototype that would be solved with an automatic pouring system. Apart from that, the temperature control was quite good in the processing and the sedimentation chambers. It was above the liquidus for the formation of P-Fe-IMCs in the processing chamber, to prevent precipitation of the intermetallics in this stage, and fell within the range of formation range along the launder to finally be maintained in liquid above α-Al formation temperature. The in-line configuration allows the collection of cleaned melt as it flows out of the sedimentation chamber and its transfer to another furnace where it can be heated up and further processed or be cast directly into ingots. Figure 11 shows the concentration of Fe in the liquid collected from the outlet as a function of time for tests carried out with and without applying HSMC in the processing chamber. Horizontal dashed lines were included to indicate the initial Fe concentration (Cinitial), the lowest Fe level achieved during batch processing (Cbatch) and the theoretical limit for this alloy (Climit) The melt collected after HSMC exhibited low Fe levels from the moment that there was an outflow, while the reference melt without HSMC started showing high Fe level that slowly decreased with time as the melt continued flowing out of the sedimentation chamber. This result demon- The in-line configuration allows the collection of cleaned melt as it flows out of the sedimentation chamber and its transfer to another furnace where it can be heated up and further processed or be cast directly into ingots. Figure 11 shows the concentration of Fe in the liquid collected from the outlet as a function of time for tests carried out with and without applying HSMC in the processing chamber. Horizontal dashed lines were included to indicate the initial Fe concentration (C initial ), the lowest Fe level achieved during batch processing (C batch ) and the theoretical limit for this alloy (C limit ) The melt collected after HSMC exhibited low Fe levels from the moment that there was an outflow, while the reference melt without HSMC started showing high Fe level that slowly decreased with time as the melt continued flowing out of the sedimentation chamber. This result demonstrates that with HSMC, the primary intermetallics form and grow at an earlier stage, along the launder connecting the two chambers. When the intermetallics reach the holding chamber they are bigger and tend to settle faster to the bottom, allowing the cleaner fraction to flow out. In contrast, when HSMC is not used, the primary intermetallic particles are more likely to form and grow after the launder cooling stage, i.e., once the melt reaches the holding chamber. In this case, the intermetallics are smaller and there is a competition between the growth and the sedimentation while the melt is passing through the chamber. Only with time, the particles get bigger and settle faster and the chamber starts delivering melt with a lower Fe content. During these initial tests, no removal of the accumulated particles in the sedimentation chamber was performed, explaining the gradual increase for the HSMC line as shown in Figure 11. This issue represents a disadvantage for the current technology implementation because it limits the melt flow that can be processed, which is typically high in industry, but it will be solved in future modifications of the setup. Some of the possible solutions would be either increasing the size of the separation chamber or including a siphon system to remove the sedimented particles in situ.
For simplicity, the current in-line system has been only used to assess the separation of the intermetallics by gravity, but its configuration also allows it to be used in combination with other separation techniques. The options include the use of filters at the outlet of the separation chamber, or by coupling it with existing electromagnetic separation equipment [5,10] or even with gas bubbling for particle removal by flotation [68,70]. Using this technological combination, the overall process efficiency could increase significantly and make it more attractive for the requirements of the aluminium recycling industry.

Conclusions
The purification of molten aluminium alloys by the precipitation of P-Fe-IMCs has been evaluated both from a theoretical and an experimental point of view and the high shear melt conditioning (HSMC) technology developed in BCAST has been used to evaluate its applicability to improve the iron removal process. The key conclusions obtained during the investigations can be summarized as follows: - The precipitation method is only practical for high Si aluminium cast alloys because of their easy tendency to form P-Fe-IMCs within, or slightly above, the alloy specifications. The extension to other alloy systems would first require an increment of the limits for Fe, Si and Mn in the alloy specifications. - The use of CALPHAD tools is essential to accurately predict the potential Fe removal from the contaminated aluminium melts and to estimate the operating temperature windows to achieve the best results. - The Fe level in a contaminated A380 alloy can be theoretically reduced to ~1%Fe without the need of added Mn and regardless of the initial Fe content. However, if lower Fe levels are required, the use of Mn becomes necessary and can easily control the expected final Fe level in the purified melt. With HSMC, the Fe level in the collected melt during the whole duration of the trial was quite stable, with an average Fe level of 0.66%, while for the reference melt the average Fe level in the collected melt was 0.87%. This represents an increase in process efficiency in terms of composition (Equation (3)) from 20 to 57% by using HSMC. The Fe levels during in-line processing are always slightly higher than the results obtained during in-batch processing for long holding time (0.47%Fe) and the process efficiency lower because, in this case, the particle sedimentation must also overcome the vertical melt flow out of the chamber (negative value of v L in Equation (6)).
During these initial tests, no removal of the accumulated particles in the sedimentation chamber was performed, explaining the gradual increase for the HSMC line as shown in Figure 11. This issue represents a disadvantage for the current technology implementation because it limits the melt flow that can be processed, which is typically high in industry, but it will be solved in future modifications of the setup. Some of the possible solutions would be either increasing the size of the separation chamber or including a siphon system to remove the sedimented particles in situ.
For simplicity, the current in-line system has been only used to assess the separation of the intermetallics by gravity, but its configuration also allows it to be used in combination with other separation techniques. The options include the use of filters at the outlet of the separation chamber, or by coupling it with existing electromagnetic separation equipment [5,10] or even with gas bubbling for particle removal by flotation [68,70]. Using this technological combination, the overall process efficiency could increase significantly and make it more attractive for the requirements of the aluminium recycling industry.

Conclusions
The purification of molten aluminium alloys by the precipitation of P-Fe-IMCs has been evaluated both from a theoretical and an experimental point of view and the high shear melt conditioning (HSMC) technology developed in BCAST has been used to evaluate its applicability to improve the iron removal process. The key conclusions obtained during the investigations can be summarized as follows: - The precipitation method is only practical for high Si aluminium cast alloys because of their easy tendency to form P-Fe-IMCs within, or slightly above, the alloy specifications. The extension to other alloy systems would first require an increment of the limits for Fe, Si and Mn in the alloy specifications. - The use of CALPHAD tools is essential to accurately predict the potential Fe removal from the contaminated aluminium melts and to estimate the operating temperature windows to achieve the best results. - The Fe level in a contaminated A380 alloy can be theoretically reduced to~1%Fe without the need of added Mn and regardless of the initial Fe content. However, if lower Fe levels are required, the use of Mn becomes necessary and can easily control the expected final Fe level in the purified melt. - The precipitation and sedimentation of P-Fe-IMCs is a slow process which requires either a very slow cooling rate or long holding times to be effective. - The HSMC technology before cooling enhances the formation of the primary Fe-rich intermetallics by reducing their undercooling, regardless of the cooling conditions. This produces an earlier sedimentation of the intermetallic particles in the melt during cooling and it is more practical for industry because it can reduce processing times and increase productivity. -Moreover, the Fe-rich primary intermetallics produced after HSMC are more compact, something that facilitates their movement in the liquid aluminium and promotes a faster sedimentation. - The HSMC technology is still at laboratory scale and limited to small amounts of melt, but ongoing research is focus on improving the applicability to large melts. -Initial scale-up tests with the technology have demonstrated its feasibility for the continuous melt treatment of contaminated aluminum melts with a~60% purification efficiency, which is below the~90% efficiency obtained in batch. Further development is currently in progress to improve process efficiency. -As a melt conditioning technology, the technology can also be used combined with other existing separation techniques (centrifugal, filtration, EM separation, etc.), therefore increasing the overall iron removal rate and efficiency.

Patents
A patent resulted from work carried out during the LiME Hub regarding the use of HSMC for purifying alloy melts:

Data Availability Statement:
The data presented in this manuscript is available on request from the corresponding author.