The Production of Rare Earth based Magnesium and Aluminium Alloys – A Review

ABSTRACT

Rare Earth (RE) elements can be used as alloying agents to improve properties such as strength, conductivity, and corrosion resistance of Al or Mg light metal (LM) alloys. The current route for producing Al/Mg-RE alloys is by melting Al or Mg with the required RE metal. However, the high price of RE metals makes the cost of the alloys expensive and, as a result limits, their application. Extensive studies have been conducted to find alternative process routes to making RE alloys with the aim of making the alloys more efficient and less costly. In this study, a comprehensive overview of alternative Al/Mg-RE alloy production methods is provided. Three alternative routes are identified and reviewed: the direct metallothermic reaction of the metal with RE compounds; a multistage reduction of RE compounds; and the electrolysis route. Recommendations for ongoing and future research into Al/Mg-RE alloys as well as industrial implications are also discussed.

KEYWORDS:

• Aluminum
• magnesium
• rare earth
• alloy production
• metallothermic reduction
• molten salt
• electrolysis

1. Introduction

Aluminum and magnesium are classified as light metals (LM) because their specific gravities are relatively low (2.70 and 1.74, respectively) compared to other commonly used metals such as Fe (7.87) and Cu (8.96) (Rumble 2021). The high strength to weight ratio of light metals is also considered beneficial for their use in applications where low weight and high strength are important, such as in transportation (Bray 2020, 2022). Compared to commonly used steels, however, there are a number of properties of Mg that limit its application such as poor ductility, limited creep resistance, and high susceptibility to corrosion (Bian et al. 2022; Esmaily et al. 2017; Mordike and Ebert 2001; Tekumalla et al. 2014). Similarly, pure Al also has limited corrosion resistance, for example where chloride is present in significant quantities such as in coastal regions or in chemical plants, has relatively low strength, and is very flexible (Megson 2012; Schofield 2001). To improve the mechanical and corrosion properties of Al and Mg, different elements are often added as alloying agents. The most common alloying elements with Mg include Al, Zn, Mn, Si, Cu, and Zr (either alone or in combination) while Al is typically alloyed with Cu, Mg, Zn, Sn, Si, and Mn.

Alternative alloying elements that are used to improve the properties of Al and Mg are sourced from the Rare Earth (RE) elements series. The RE series consists of elements, known as lanthanides, that have atomic numbers ranging from 57 to 71 and which belong to the sixth period and IIIA subgroup of the periodic table. Two further elements, Sc (Z = 21) and Y (Z = 39), are commonly included as RE, due to their having similar chemical properties with members the lanthanide series (Connelly et al. 2007). RE elements are typically extracted from sources such as naturally occurring RE-rich deposits (e.g. deposits containing mineral such as bastnaesite (La,Ce)FCO₃, monazite (Ce,La,Y,Th)PO₄, xenotime (YPO₄), or from ion-absorbed clay deposits), as by-products from other processes (e.g. from rutile, alumina, zircon, magnesia, and cassiterite processing); from electronic waste; or from coal and coal-based materials (e.g. fly ash, bottom ash, acid mine drainage, and its treatment products, and coal refuse) (Archambo and Kawatra 2021; Rafique 2023; Talan et al. 2023).

In recent years, addition of RE elements to Al or Mg to form RE-Al/Mg alloys has been the subject of increasing research interest because of their potential engineering applications (Sommer 1992). Beneficial effects of adding RE elements to either Al or Al-rich alloys include: the promotion of dehydrogenation whereby the RE decreases the H₂ solubility in solid Al alloys and instead facilitates the formation of RE-hydrides; the modification of the eutectic Si phase in Al-Si alloys; the strengthening effect of alloys through grain refining, grain boundary strengthening, and solution strengthening; and; improving the corrosion resistance and electrical conductivity of the light metal (Nie et al. 2002, 2003). The addition of RE elements is also known to reduce the deteriorating effects of Cu and Fe impurities in Al alloys (Sims et al. 2021; Zhang et al. 2012a) and some Al-RE (RE=Ce/La) alloys have been shown to have superior castability properties, similar to Al-Si alloys, and can be used with minimal or without any heat treatment (Sims et al. 2022; Weiss et al. 2017). The latter alloys are particularly useful as Ce and La are considered lower-value by-products of the RE supply chain compared to other more expensive and in demand RE elements such as Nd, Pr, Dy and Sm. In addition, Al-Sc alloys are used for making sport equipment such as baseball and softball bats, bicycle frames, lacrosse sticks, and tentpoles because of their superior strength properties (Røyset and Ryum 2005). In more high-tech applications, NASA and Airbus are both evaluating the use of Al-Sc alloys for aircraft components (Jones et al. 2022; Lee and Chen 2004).

For Mg, the effects of adding RE elements to either pure Mg or an Mg alloy include: promoting purification of the alloy melt; improving the alloy castability; refinement of the microstructure; and; improving the mechanical and anti-oxidation properties (Paradiso et al. 2017; Rokhlin 2003; Sivashanmugam and Harikrishna 2020; Tekumalla et al. 2014). In contrast to Al alloys, the use of RE elements as alloying element for Mg alloys is relatively common and a lettering code for Mg alloys developed by the American Society of Testing and Materials (ASTM) is widely used in industry. For example, Mg alloy WE43C-T5 contains approximately 4 wt.% yttrium (W) and 3 wt.% RE (E), the letter C indicates this is the third composition of the alloy that become a standard, and T5 indicates the type of thermal treatment (Calado, Carmezim, and Montemor 2022; Moosbrugger and Marquard 2017). Due to its superior properties, the WE series of Mg alloys has been of particular interest to the automotive and aerospace industries for the development of high-performance car, helicopter, and fighter aircraft components (Calado, Carmezim, and Montemor 2022). Another emerging application for Mg-RE alloys is for use as biocompatible materials. Mg-RE alloys have an elastic modulus similar to the elastic modulus of natural bone, and lower than that of other materials used for implants such as 316 L stainless steel (Calado, Carmezim, and Montemor 2022). Stents and screws made from an Mg-Y-RE-Zr alloy have been shown to have good biocompatibility and to promote bone healing processes (Chen et al. 2018).

Essentially, there are two types of light metal RE alloys, the difference between the two being based on the concentration level of the RE element. The first type is referred to as ‘applied alloys’ and usually contains RE concentrations of 0.1–0.5%. The second type is known as ‘master alloys’ and has RE contents ranging from about 5–12% (Jang et al. 2015). In practice, the master alloys are commonly used as the precursor and feed materials for preparing the applied alloys. The production of LM-RE master and applied alloys is typically carried out by adding the RE metal directly to the molten LM or LM alloy and the RE will either be dissolved into the molten LM or form intermetallics with the LM. However, the requirement of using the RE element in the metallic form provides the biggest barrier toward commercialization of such processes because the pure RE metals are expensive (Savchenkov et al. 2021; Xiao et al. 2020). In addition, significant RE metal losses can occur during alloy preparation, they often require complex melting-casting processes, and the high differences in melting temperatures can result in segregation in the alloy (Liu et al. 2019b; Polat et al. 2018; Shtefanyuk et al. 2016). For example, during addition of scandium in the form of Al-2%Sc into Al to make alloys containing 0.22–0.23% Sc, normally 0.01–0.03% scandium is lost due to evaporation, oxidation, or deposition on the hearth surface (Zakharov and Fisenko 2017). Adding scandium directly to the alloy however, results in higher losses due to its higher melting temperature (~1540°C). In general, existing processes are inefficient.

The aim of the current study is to systematically review major alternative routes to produce LM-RE alloys. The routes explored avoid the use of expensive RE metals, instead employing significantly lower cost RE compounds (e.g. oxides or halides) as a means of introducing the required RE. A direct metallothermic reaction with RE compounds route is first described, followed by the multistage indirect reduction of RE compounds route, and lastly the electrolysis LM-RE alloy route. Finally, industrial implications and future outlooks for LM-RE production are discussed.

2. Direct metallothermic reaction with RE compounds

Rather than adding a RE metal directly to a light metal, an alternative way of producing Al/Mg-RE alloys is by direct addition of an RE compound to the light metal. In this process, the RE compound will be reduced by the Al or Mg. This direct reduction method is commonly used in RE metal production whereby an RE compound, typically a RE oxide, undergoes metallothermic reduction processes using calcium as a reductant to produce the RE metal (Rafique 2023).

In the metallothermic reaction with RE compounds method, a simultaneous process of RE compound reduction by Al/Mg results in alloy formation between the Al/Mg and RE. This is particularly advantageous as adding a RE compound instead of a pure RE metal can significantly decrease the manufacturing costs of the alloy (e.g. the price of a RE oxide is approximately one-third of the price of the RE metal). While several RE compounds could be added to make the alloy (e.g. oxides and halides), the most studied compounds used as precursors to produce LM-RE alloys are that of RE oxides.

The overall reaction for metallothermic reduction of a general metal oxide by Al or Mg is shown in Equations (1)(1) $\frac{6}{\mathrm{y}}\mathrm{A}{\mathrm{l}}_{(\mathrm{l})}+\hspace{0.17em}{\mathrm{M}}_{\mathrm{x}}{\mathrm{O}}_{\mathrm{y}(\mathrm{s})}=\hspace{0.17em}\mathrm{x}\hspace{0.17em}{\mathrm{M}}_{(\mathrm{s})}+\frac{3}{\mathrm{y}}\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_{3(\mathrm{s})}\mathrm{T}>\mathrm{T}{\mathrm{m}}_{\mathrm{A}\mathrm{l}}\left(\approx 660{\hspace{0.17em}}^{\circ }\mathrm{C}\right)$(1) and (2(2) $\mathrm{y}\hspace{0.17em}\mathrm{M}{\mathrm{g}}_{(\mathrm{l})}+\hspace{0.17em}{\mathrm{M}}_{\mathrm{x}}{\mathrm{O}}_{\mathrm{y}(\mathrm{s})}=\hspace{0.17em}\mathrm{x}\hspace{0.17em}{\mathrm{M}}_{(\mathrm{s})}+\hspace{0.17em}\mathrm{y}\hspace{0.17em}\mathrm{M}\mathrm{g}{\mathrm{O}}_{\left(\mathrm{s}\right)}\mathrm{T}>\mathrm{T}{\mathrm{m}}_{\mathrm{M}\mathrm{g}}\left(\approx 650{\hspace{0.17em}}^{\circ }\mathrm{C}\right)$(2) ), where M is the reducing metal:

(1) $\frac{6}{\mathrm{y}}\mathrm{A}{\mathrm{l}}_{(\mathrm{l})}+\hspace{0.17em}{\mathrm{M}}_{\mathrm{x}}{\mathrm{O}}_{\mathrm{y}(\mathrm{s})}=\hspace{0.17em}\mathrm{x}\hspace{0.17em}{\mathrm{M}}_{(\mathrm{s})}+\frac{3}{\mathrm{y}}\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_{3(\mathrm{s})}\mathrm{T}>\mathrm{T}{\mathrm{m}}_{\mathrm{A}\mathrm{l}}\left(\approx 660{\hspace{0.17em}}^{\circ }\mathrm{C}\right)$(1)

(2) $\mathrm{y}\hspace{0.17em}\mathrm{M}{\mathrm{g}}_{(\mathrm{l})}+\hspace{0.17em}{\mathrm{M}}_{\mathrm{x}}{\mathrm{O}}_{\mathrm{y}(\mathrm{s})}=\hspace{0.17em}\mathrm{x}\hspace{0.17em}{\mathrm{M}}_{(\mathrm{s})}+\hspace{0.17em}\mathrm{y}\hspace{0.17em}\mathrm{M}\mathrm{g}{\mathrm{O}}_{\left(\mathrm{s}\right)}\mathrm{T}>\mathrm{T}{\mathrm{m}}_{\mathrm{M}\mathrm{g}}\left(\approx 650{\hspace{0.17em}}^{\circ }\mathrm{C}\right)$(2)

A schematic of the metallothermic reaction with RE compound route is shown in Figure 1.

Figure 1. Generic mechanism involved in the direct metallothermic reaction of RE compounds with a light metal.

A comparison of the Gibbs free energy (ΔG°) of formation of various RE oxides and Al/Mg oxides (LMO) is presented in Figure 2. The graph was constructed using the Reaction module in conjunction with the FactPS database available in the FactSage 8.2 thermochemical software package (Bale et al. 2016). Based on the relative ΔG° values, the metallothermic reduction of RE oxides by Al or Mg (Equations (1)(1) $\frac{6}{\mathrm{y}}\mathrm{A}{\mathrm{l}}_{(\mathrm{l})}+\hspace{0.17em}{\mathrm{M}}_{\mathrm{x}}{\mathrm{O}}_{\mathrm{y}(\mathrm{s})}=\hspace{0.17em}\mathrm{x}\hspace{0.17em}{\mathrm{M}}_{(\mathrm{s})}+\frac{3}{\mathrm{y}}\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_{3(\mathrm{s})}\mathrm{T}>\mathrm{T}{\mathrm{m}}_{\mathrm{A}\mathrm{l}}\left(\approx 660{\hspace{0.17em}}^{\circ }\mathrm{C}\right)$(1) and (2))(2) $\mathrm{y}\hspace{0.17em}\mathrm{M}{\mathrm{g}}_{(\mathrm{l})}+\hspace{0.17em}{\mathrm{M}}_{\mathrm{x}}{\mathrm{O}}_{\mathrm{y}(\mathrm{s})}=\hspace{0.17em}\mathrm{x}\hspace{0.17em}{\mathrm{M}}_{(\mathrm{s})}+\hspace{0.17em}\mathrm{y}\hspace{0.17em}\mathrm{M}\mathrm{g}{\mathrm{O}}_{\left(\mathrm{s}\right)}\mathrm{T}>\mathrm{T}{\mathrm{m}}_{\mathrm{M}\mathrm{g}}\left(\approx 650{\hspace{0.17em}}^{\circ }\mathrm{C}\right)$(2) should not be favored thermodynamically, except in the case of Eu₂O₃, because most of RE oxides are more stable compared to Al₂O₃ or MgO. Note however, even though the LMOs are more stable compared to most of the REOs, the reaction will proceed if the RE metal product is soluble in the liquid Al or Mg, as represented by Equation (3)(3) $\begin{array}{rl}& \frac{6}{\mathrm{y}}\mathrm{A}{\mathrm{l}}_{(\mathrm{l})}/\mathrm{y}\mathrm{M}{\mathrm{g}}_{(\mathrm{l})}+\hspace{0.17em}\mathrm{R}{\mathrm{E}}_{\mathrm{x}}{\mathrm{O}}_{\mathrm{y}(\mathrm{s})}=\hspace{0.17em}\mathrm{x}\hspace{0.17em}{\left[\mathrm{R}\mathrm{E}\right]}_{\mathrm{A}\mathrm{l}/\mathrm{M}\mathrm{g}\hspace{0.17em}(\mathrm{l})}+\frac{3}{\mathrm{y}}\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_{3(\mathrm{s})}/\mathrm{y}\\ & \mathrm{M}\mathrm{g}{\mathrm{O}}_{(\mathrm{s})};\hspace{0.17em}\mathrm{T}\hspace{0.17em}>\hspace{0.17em}{\mathrm{T}}_{\mathrm{m}}\left(\approx 640-650{\hspace{0.17em}}^{\circ }\mathrm{C}\right)\end{array}$(3) :

Figure 2. Comparison of relative Gibbs free energy of formation of Al, Mg, and RE oxides.

(3) $\begin{array}{rl}& \frac{6}{\mathrm{y}}\mathrm{A}{\mathrm{l}}_{(\mathrm{l})}/\mathrm{y}\mathrm{M}{\mathrm{g}}_{(\mathrm{l})}+\hspace{0.17em}\mathrm{R}{\mathrm{E}}_{\mathrm{x}}{\mathrm{O}}_{\mathrm{y}(\mathrm{s})}=\hspace{0.17em}\mathrm{x}\hspace{0.17em}{\left[\mathrm{R}\mathrm{E}\right]}_{\mathrm{A}\mathrm{l}/\mathrm{M}\mathrm{g}\hspace{0.17em}(\mathrm{l})}+\frac{3}{\mathrm{y}}\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_{3(\mathrm{s})}/\mathrm{y}\\ & \mathrm{M}\mathrm{g}{\mathrm{O}}_{(\mathrm{s})};\hspace{0.17em}\mathrm{T}\hspace{0.17em}>\hspace{0.17em}{\mathrm{T}}_{\mathrm{m}}\left(\approx 640-650{\hspace{0.17em}}^{\circ }\mathrm{C}\right)\end{array}$(3)

When the amount of RE metal dissolved into the liquid Al or Mg is low, the thermodynamic activity (α) of RE metal (α_RE) is very small, which results in the overall ΔG of Equation (3)(3) $\begin{array}{rl}& \frac{6}{\mathrm{y}}\mathrm{A}{\mathrm{l}}_{(\mathrm{l})}/\mathrm{y}\mathrm{M}{\mathrm{g}}_{(\mathrm{l})}+\hspace{0.17em}\mathrm{R}{\mathrm{E}}_{\mathrm{x}}{\mathrm{O}}_{\mathrm{y}(\mathrm{s})}=\hspace{0.17em}\mathrm{x}\hspace{0.17em}{\left[\mathrm{R}\mathrm{E}\right]}_{\mathrm{A}\mathrm{l}/\mathrm{M}\mathrm{g}\hspace{0.17em}(\mathrm{l})}+\frac{3}{\mathrm{y}}\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_{3(\mathrm{s})}/\mathrm{y}\\ & \mathrm{M}\mathrm{g}{\mathrm{O}}_{(\mathrm{s})};\hspace{0.17em}\mathrm{T}\hspace{0.17em}>\hspace{0.17em}{\mathrm{T}}_{\mathrm{m}}\left(\approx 640-650{\hspace{0.17em}}^{\circ }\mathrm{C}\right)\end{array}$(3) becoming more negative, as can be seen from Equation (4)(4) $\mathrm{\Delta }G=\mathrm{\Delta }{G}^{\circ }+\hspace{0.17em}RT\hspace{0.17em}ln\frac{{\alpha }_{RE}^2+{\alpha }_{A{l}_2{O}_3/MgO}^{1/3}}{{\alpha }_{Al/Mg}^{2/3}+{\alpha }_{R{E}_2{O}_3}}$(4) , where ΔG° is the standard Gibbs free energy of the reaction, R is the gas constant and T is the reaction temperature.

(4) $\mathrm{\Delta }G=\mathrm{\Delta }{G}^{\circ }+\hspace{0.17em}RT\hspace{0.17em}ln\frac{{\alpha }_{RE}^2+{\alpha }_{A{l}_2{O}_3/MgO}^{1/3}}{{\alpha }_{Al/Mg}^{2/3}+{\alpha }_{R{E}_2{O}_3}}$(4)

Reduced REs may also form intermetallic compounds with Al/Mg which can be more stable compared to their oxides if the melting point of the intermetallic is higher than the reduction temperature (for example, in the Al-Sc₂O₃ system – see Equation (5)(5) $\begin{array}{rl}& 8\mathrm{A}{\mathrm{l}}_{\left(\mathrm{l}\right)}+\mathrm{S}{\mathrm{c}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}=2\mathrm{A}{\mathrm{l}}_3\mathrm{S}{\mathrm{c}}_{\left(\mathrm{s}\right)}+\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)};\\ & \mathrm{T}>{\mathrm{T}}_{\mathrm{m}}\left(\approx 640-{650}^{\circ }\mathrm{C}\right)\end{array}$(5) ).

(5) $\begin{array}{rl}& 8\mathrm{A}{\mathrm{l}}_{\left(\mathrm{l}\right)}+\mathrm{S}{\mathrm{c}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}=2\mathrm{A}{\mathrm{l}}_3\mathrm{S}{\mathrm{c}}_{\left(\mathrm{s}\right)}+\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)};\\ & \mathrm{T}>{\mathrm{T}}_{\mathrm{m}}\left(\approx 640-{650}^{\circ }\mathrm{C}\right)\end{array}$(5)

An early study involving RE oxide addition to light metals was presented in a patent by Tarcy and Foster (1988). In this patent, the RE oxide used was Sc₂O₃ powder that was mixed with Al and pressed to make a compact pellet. The compact was then added to a molten Al bath at temperatures of 750–950°C. A maximum alloy yield of 92.8% was reached with 0.587% Sc present in the alloy (Tarcy and Foster 1988). Since then, further studies have been conducted on the following systems: Al-Sc₂O₃ (Fujii et al. 2003, 2020; Harata et al. 2008), Al-Mg-CeO₂ (Luna A et al. 2011), Mg-Al-La₂O₃ (Lu et al. 2005), Al-Cu-Pr_xO_y (Zhao et al. 2008), Al-Cu-Pr_xO_y-La₂O₃ (Bai et al. 2014; Xia et al. 2012; Zhao et al. 2011), AZ31 Alloy-La₂O₃ (Zhao et al. 2013), Mg-Sc₂O₃ (Kim et al. 2013), and A356-Y₂O₃ (Moussa, El-Hadad, and Khalifa 2019).

To compare the effect of adding the RE as either a metal or an oxide directly to the molten light metal (or light metal alloy), Zhao et al. (2013) compared the difference in microstructure and oxidation behavior of AZ31 Mg alloyed with La or La₂O₃. The results showed that there was no significant difference in the microstructure of the final product when either La or La₂O₃ was used as the precursor, as shown in Figure 3 (Zhao et al. 2013).

Figure 3. Microstructure of AZ31 magnesium alloys prepared using both La and La₂O₃ additions (as indicated in the brackets) (Zhao et al. 2013).

Alternative techniques to make the alloys have also been tested. For example, in one study Sc₂O₃ and Al were mechanically alloyed and then sintered using spark plasma sintering (SPS). It was reported that the Sc₂O₃ was not reduced by the mechanical alloying process but was reduced during the SPS (Fujii et al. 2003). Calcium metal has also been used as a reductant for Sc₂O₃ and then molten Al used to act as a collector to form Al-Sc alloys (Harata et al. 2008). More recently, microwave radiation has been used to produce Al-Sc alloys, with the aim to reduce the overall energy inputs and therefore develop a more environmental-friendly process route (Fujii et al. 2020). Compared to the use of conventional heating, the use of microwave radiation as a heat source often results in high reaction rates, selective chemical reactions, and a smaller reactor requirement.

Although the use of RE oxides as a starting material can potentially reduce the material costs for alloy production, there are some issues which still need to be resolved. For example, alumina (Al₂O₃) and magnesia (MgO), are generated as by-products of the RE oxide reduction and their presence may affect the properties and the purity of the Al and Mg alloys through factors such as: a reduction in the mechanical properties; poor machinability; low surface quality; an increase in porosity and; they may exhibit a tendency to increase corrosion (Li et al. 2017; Lun Sin, Elsayed, and Ravindran 2013). Hence, these impurities present as inclusions in the alloy need to be removed. This is typically done by refining them with chloride salts (Li et al. 2017) although physical separation procedures such as filtration, gas bubble flotation, and settling process could also be employed to remove the inclusions (Damoah and Zhang 2010). It should be noted however, oxide inclusion removal processes may not be necessary particularly in the case when the initial alloy produced is the master alloy – if present the oxide inclusions may be removed when the alloy is used in the subsequent alloy-making process.

In terms of thermodynamics, based on the simulations using FactSage 8.2 (Bale et al. 2016), RE oxide reduction by Mg is most effective for the first half of the lanthanides series, from La up to Sm (except for Promethium for which there is a lack of thermodynamic data). Elements from the second half of the lanthanide series are much more difficult to reduce, as the oxides are more stable compared to the first half (see Figure 2). In comparison, RE oxide reduction by Al does not follow the order of the RE oxide stability. This is a result of the formation of the many intermetallic compounds (which are more stable than RE oxides) between Al and RE that can form during the process. We note however, thermodynamic simulation of these complex systems requires further study and experimental trials are needed to verify any predictions.

The use of RE compounds other than RE oxides as a precursor for allow formation has also been studied. A patent by Duyvesteyn (2019) reported the use of scandium oxalate (Sc₂(C₂O₄)₃.xH₂O) which decomposes to Sc₂O₃ when contacted with the molten metal, releasing vapor and gases. The in-situ Sc₂O₃ generation was suggested to reduce the burning losses (i.e. loss of Sc to the dross) because of better mixing (Duyvesteyn 2019). Scandium chloride may also be used to make Al-Sc alloys (Equation (6))(6) $\begin{array}{rl}& 4\mathrm{A}{\mathrm{l}}_{\left(\mathrm{l}\right)}+\mathrm{S}\mathrm{c}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{s}\right)}=\mathrm{A}{\mathrm{l}}_3\mathrm{S}{\mathrm{c}}_{\left(\mathrm{s}\right)}+\mathrm{A}\mathrm{l}\mathrm{C}{\mathrm{l}}_{\mathrm{s}\left(\mathrm{g}\right)};\\ & \mathrm{T}=700-{1000}^{\circ }\mathrm{C};\mathrm{P}<500\mathrm{m}\mathrm{b}\mathrm{a}\mathrm{r}\end{array}$(6) .

(6) $\begin{array}{rl}& 4\mathrm{A}{\mathrm{l}}_{\left(\mathrm{l}\right)}+\mathrm{S}\mathrm{c}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{s}\right)}=\mathrm{A}{\mathrm{l}}_3\mathrm{S}{\mathrm{c}}_{\left(\mathrm{s}\right)}+\mathrm{A}\mathrm{l}\mathrm{C}{\mathrm{l}}_{\mathrm{s}\left(\mathrm{g}\right)};\\ & \mathrm{T}=700-{1000}^{\circ }\mathrm{C};\mathrm{P}<500\mathrm{m}\mathrm{b}\mathrm{a}\mathrm{r}\end{array}$(6)

The reaction of ScCl₃ with Al to form Al-Sc alloys (Equation (6))(6) $\begin{array}{rl}& 4\mathrm{A}{\mathrm{l}}_{\left(\mathrm{l}\right)}+\mathrm{S}\mathrm{c}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{s}\right)}=\mathrm{A}{\mathrm{l}}_3\mathrm{S}{\mathrm{c}}_{\left(\mathrm{s}\right)}+\mathrm{A}\mathrm{l}\mathrm{C}{\mathrm{l}}_{\mathrm{s}\left(\mathrm{g}\right)};\\ & \mathrm{T}=700-{1000}^{\circ }\mathrm{C};\mathrm{P}<500\mathrm{m}\mathrm{b}\mathrm{a}\mathrm{r}\end{array}$(6) is not favored thermodynamically under 1000°C, but if the partial pressure of AlCl₃ is reduced to 0.01 mbar, the reaction can be reversed (Haidar 2014). Processing at low temperatures is preferred to help with excessive gas production and to prevent powder blowing out of the reaction zone.

Table 1 summarizes these and other previous studies using direct RE oxide and RE compound addition to make RE-Al/Mg alloys.

Table 1. Summary of previous works/patents involving the direct RE compound addition route*.

Author	Reaction and parameters	Process details
Fujii et al. (2020)	Sc2O3(s) + Mg(g) + Al(s) → Al-Sc Alloy metal + MgO T for Al/Sc2O3 briquettes = 660 °C. T for Mg ribbon/Zr beads = Mg vaporization temperature (> 650 °C)	Al and Sc2O3 was made into a briquette and Mg ions were used as a reductant for Sc2O3. Only Spinel (MgAl2O4) was detected in the sample as a by-product. By lowering the temperature, spinel formation was reduced. No Mg-Sc alloy were detected. Only a small amount of Sc dissolved in Al at 660 °C, the eutectic point of Al3Sc was 1320 °C and the melting point was 2100 °C. The process is non-thermal equilibrium reaction. Magnesium atoms were excited by microwaves and the generations of plasma radicals and electrons compensate the lack of sufficient reduction reaction energy. 69.8% yield of Al3Sc.
Xiao et al.)(2020)	2ScCl3 + 3Mg + 6Al = 2Al3Sc + 3MgCl2 m(Al):m(Mg):m(ScCl3) = 10:1:1.5 Al and Mg melted first and stirred. Some cover agent is added. After that ScCl3 molten salt are also added and the mmoussolten mixture is stirred. Lastly cover and slag removal agent is added. T= 900-1200 K; Time: 40 min; stirring speed: 8 rpm.	Final alloy content: 2.90% Sc, 5.73% Mg, 0.0058% Cu, 0.29%, 0.029% Ti, 0.13% Fe, 0.075% Zn, 0.025% Na. 96.72 % of scandium was recovered. Al promotes the reaction between ScCl3 with Mg because Sc creates stable intermetallic compounds with Al. Higher temperature results in reduced reaction progress (because of exothermic reaction) and evaporation of molten salts. Optimum ScCl3 molten salts: 15 gr (based on m(Al):m(Mg):m(ScCl3) = 10:1:1.5). On scale up, the Sc yield and content in the alloy were decreased, but after applying a new stirring method (argon blowing method), the results improved but were still below the small scale. After adding another new steps/method (molten salt was pressed into Al-Mg alloy) the results were improving (Sc recovery reached 96.78%, higher than 94.98% in small experiment). The bell jar method added the risk of splashing of the molten alloy. Optimum temperature: 1223 K
Moussa et al. (2019)	2Al + Y2O3 = Al2O3 + 2Y Y2O3 with the concentration of 0, 0.3, 0.6, 1.0, 1.5, 2.0 and 2.5 wt.%. Y2O3 was added in a premixed form with 70% Al to A365 alloy. T= 750 °C; Stirring 600 rpm for 5 min; Cast into sand mold.	Al3Y started to solidify first at 637 °C because the liquidus temperature of A356 is 615 °C. Al3Y particles were segregated at the edge of eutectic Si, hence restricting the growth of the eutectic phase.
Duyvesteyn (2019)^p	Sc2(C2O4)3.xH2O(s) → Sc2(CO3)3(s)+3CO (g)+xH2O(g) Sc2(CO3)3(s) → Sc2O3(s) +3CO2(g) Sc2O3(s)+4Mg(l)→(Mg,Sc)(l)+3MgO(s) Sc2O3(s)+3Al(l)→(Mg,Sc)(l)+Al2O3(s) Temperature: 700 – 750 °C. Maximum % Sc in Al alloy: 2%. Maximum % Sc in Mg alloy: 30%.	Scandium oxalate decomposes into Sc2O3. In situ decomposition was thought to improve the mixing process of the alloys and reduce the entrapment of scandium oxide in the dross. Maximum % Sc in Al alloy: 2% Maximum % Sc in Mg alloy: 30%
Kim et al. (2013)^p	Maximum % Sc in alloy: 30% but only 0.5% Sc2O3 result was shown. Temperature: 600 – 800 °C. Stirring: 1 – 400 min.	Sc2O3 powder was added to Mg melt to make a Mg-Sc master alloy. The Mg-Sc master alloy was added to Al melt to make an Al-Mg-Sc alloy.
Haidar (2014)^p	ScCl3 + (x+1)Al → Sc-Alx + AlCl3(g) Pressure: 0.01 mbar. Temperature: 600 – 800 °C.	Reaction was not favorable under 1000 °C but if the partial pressure of AlCl3 was below 0.01 mbar, the reverse reaction was reduced and the forward reaction maximized. Lower temperatures are preferred to avoid excessive gas production and blowing of the powder out of reaction zone.
Zhao et al.)(2013)	Two different additions were made using La2O3 and a mixture between La 20 wt.% and balanced Mg, to an AZ31 Alloy Practical La content: 0.3, 0.75, 0.95, 1.15% Additions were wrapped in Al foil and added into molten Mg. T= 700 °C; Time= 30 min	Similar microstructures were observed between La and La2O3 addition. Phases present: αMg, Mg17Al12, and Al11La3.
Xia et al. (2012)	PrxOy (mainly Pr6O11) (0 and 6%) and La2O3 with Al-Cu alloy PrxOy concentration: 0 and 6%; La2O3 concentration: 0, 0.2, 0.4, 0.6, 0.8%. Al was melted and then pure Cu was added. Other material was wrapped in Al foil and added to the melt. REO was also wrapped in Al foil and added to the melt. T= 780 °C.	The RE elements La and Pr have little solubility in solid Al. A large amount of the dissociated [Pr] and [La] will be pushed to the front of solid–liquid phase interface during cooling because of their low solubility in the Al melt and forming Al11La3 and Al11Pr3.
Luna et al. (2011)	2Mg+4Al+4CeO2→ 2MgAl2O4+4Ce A powder injection technique was used. Injection parameters: 8 gr CeO2/min, 6 L Ar/min. 8 kg molten alloy. T= 800 and 850 °C. Mg content: 0.5, 3, and 4 wt.%.	A 4% Ce master alloy could be produced with process parameters: 60 min, 850 °C, powder size -140+200 mesh, 4 wt.% Mg, injection time 60 min, 8 gr CeO2/min.
Zhao et al. (2011)	LaxOy and PrxOy were wrapped in Al foil and added into molten Al. T= 750-800 °C; Time 20-30 min.	Original REOs are decomposed to the free state [Pr], [La], and [O] and formed intermetallic compounds with [Al] and [Cu].
Zhao et al. (2008)	1-2 wt.% PrxOy was wrapped in Al foil and added into molten Al-Cu alloy. Compact containing 20% PrxOy and 80% Al was also sintered. Time= 20-30 min. T= 750-800 °C.	Unlike the other experiment results, which usually forming intermetallic between Al-RE, in this study PrxOy react with Al to form AlPrO3. Phases present: AlPrO3.
Harata et al. (2008)	Sc2O3/2ScF3 + 3Ca → 2Sc[Al]+ 3CaF2/CaO T= 1000 °C; Time: 2 h A sealed tantalum crucible was used for the reaction.	Sc2O3/ScF3 was reduced by Ca and Al was acting as a collector. The use of Ca as the reductant produced an alloy with Ca contamination. CaCl2 was used as a flux to improve separation of metals and salts. The alloy contained 9 wt.% Sc.
Lu et al. (2005)	AZ91D-1,2,4% La2O3 La2O3 was added into molten AZ91D in the form of preform. Preform was made from Mg and Al with ratio of 10:1. La2O3 concentration in the preform was 1, 2, and 4%. T= about 700 °C; Holding time: 30 min; stirred uninterruptedly. Cast into metal mold.	Addition of La2O3 to AZ91D alloy resulted in the formation of LaAl4 phase with needle-like and particle-like morphology. Phases present: αMg, LaAl4, Mg17Al12(β).
Fujii et al. (2003)	Al and 10%Sc2O3 were mechanically alloyed and then sintered using SPS under a vacuum of 5 Pa at 600 °C with an applied pressure of 49 MPa for 3.6 ks.	No reaction occurred during mechanical alloying. Sc2O3 was reduced after Spark Plasma Sintering (SPS) processing.
Tarcy and Foster (1988)^p	8Al + Sc2O3 → 2Al3Sc +Al2O3 T= 750 – 975 °C.	Pellet consisting of Al and Sc oxide, with density of 2 g/cm^3, was added to a molten Al bath. Maximum percent reduction: 92.8%; Maximum % Sc in alloy: 0.587%.

^*prefers to patent

3. Multistage indirect reduction of RE compounds using molten salts

In this route, the RE oxides are first dissolved in molten salts to form an intermediate RE-salts solution. The intermediate molten salts containing the RE are then reduced via metallothermic reduction to produce the required LM-RE alloy. The generic steps of this process are presented schematically in Figure 4. Compared to direct reaction between a LM with RE compounds, this process has several advantages such as better oxidation protection because of the use of salts covering the metals and typically produces higher yields in some systems. For example in the case of Er₂O₃ reaction with Al, adding certain salts (chloride, fluorides, etc.) can increase the yield from 0.17% to 57.65% (Kosov, Bazhin, and Kopylova 2019). Three different salts systems are typically used for this process, namely chloride salts, fluoride salts, and chloride-fluoride salts (and mixtures thereof). Figure 5 shows previously studied systems to produce LM-RE alloys through the multistage indirect reduction of RE compounds using molten salts. Detailed discussion regarding these systems is presented in the following subsections.

Figure 4. Generic schematic for the multistage indirect reduction of a RE compound.

Figure 5. Classification of previously studied LM-RE alloy production routes via multistage indirect reduction in molten salts.

3.1. Chloride salt system

Chloride salts are usually used as a cover flux to prevent the molten alloy from oxidation, but recently they have also been used as the media involved in the actual reaction process itself. Systems involving RE chloride salts have been previously studied for the making of Mg-RE alloys however, only one study was found for the making of a Al-RE alloy using RE chloride salts (Min et al. 2019). This is most likely because the use of chloride salts for the Al system provides some complexities due to the possibility of AlCl₃ formation. If formed, AlCl₃ as a by-product must be recovered in gaseous form due to its low boiling point, thus complicating the process and equipment. In addition, the chloride may react with water vapor in the air producing alumina dust and hydrochloric acid. This does not occur in the Mg system because MgCl₂ as the by product is still in the liquid form at the processing temperature used.

The RE oxide starting material is usually chlorinated first to form RECl₃ by reaction with NH₄Cl (Equation (7))(7) $\begin{array}{rl}& \mathrm{R}{\mathrm{E}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}+6\mathrm{N}{\mathrm{H}}_4\mathrm{C}{\mathrm{l}}_{\left(\mathrm{s}\right)}=2\mathrm{R}\mathrm{E}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{s}\right)}+3{\mathrm{H}}_2{\mathrm{O}}_{\left(\mathrm{g}\right)}+6\mathrm{N}{\mathrm{H}}_{3\left(\mathrm{g}\right)};\\ & \mathrm{T}=180-{360}^{\circ }\mathrm{C}\end{array}$(7) (Meyer and Ax 1982). The RECl₃ is then dissolved in a low melting point chloride salt mixture and the RE becomes reduced when it comes into direct contact with the liquid alloy, i.e. the RE³⁺ salt reacts with the Mg and is reduced to RE metal which then dissolves into the molten Mg (Equation (8))(8) $2\mathrm{R}\mathrm{E}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{l}\right)}+3\mathrm{M}{\mathrm{g}}_{\left(\mathrm{l}\right)}=3\mathrm{M}\mathrm{g}\mathrm{C}{\mathrm{l}}_{2\left(\mathrm{l}\right)}+2{\left[\mathrm{R}\mathrm{E}\right]}_{\mathrm{M}\mathrm{g}}$(8) . The ΔG° for Equation (8)(8) $2\mathrm{R}\mathrm{E}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{l}\right)}+3\mathrm{M}{\mathrm{g}}_{\left(\mathrm{l}\right)}=3\mathrm{M}\mathrm{g}\mathrm{C}{\mathrm{l}}_{2\left(\mathrm{l}\right)}+2{\left[\mathrm{R}\mathrm{E}\right]}_{\mathrm{M}\mathrm{g}}$(8) is positive indicating that the reaction is not favorable thermodynamically under standard conditions (i.e. 1 atm and 298.15 K). However, comparable with the process from the previous section involving direct metallothermic reaction with RE compounds, the reaction can still progress forward due to the low activities, i.e. low ${\alpha }_{MgC{l}_2}$and ${\alpha }_{R{E}_{\left[Mg\right]}}$ in Equation (9)(9) $\mathrm{\Delta }G=\mathrm{\Delta }{G}^{\circ }+RTln\frac{{\alpha }_{MgC{l}_2}^3+{\alpha }_{R{E}_{\left[Mg\right]}}^2}{{\alpha }_{Al/Mg}^2+{\alpha }_{R{E}_2{O}_3}}$(9) , respectively (Fu et al. 2022; Zhang et al. 2023). A high ratio between the RE chloride salts and the metal is typically used in this process (i.e. m_salts/m_alloy >1). The reason for this high ratio is to ensure that the protection given by the molten salts to the liquid magnesium alloy is adequate. When only small amounts of salts are present, for example only covering the top of the liquid Mg alloy, the relatively higher density of the molten salts compared to the Mg alloy results in the molten salts passing through the liquid Mg alloy and sinking to the bottom of the furnace. In comparison, for a high ratio, the liquid Mg alloy will be covered on all sides and stays afloat inside the molten salts (Zhang et al. 2023). For example, Zhang et al. reported that during the process, the alloy was reported to be moving both upward and downward in the molten salts caused by density changes through the process. Different chloride salt mixtures could potentially be used in this process, namely NaCl-KCl and NaCl-KCl-CaCl₂.

(7) $\begin{array}{rl}& \mathrm{R}{\mathrm{E}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}+6\mathrm{N}{\mathrm{H}}_4\mathrm{C}{\mathrm{l}}_{\left(\mathrm{s}\right)}=2\mathrm{R}\mathrm{E}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{s}\right)}+3{\mathrm{H}}_2{\mathrm{O}}_{\left(\mathrm{g}\right)}+6\mathrm{N}{\mathrm{H}}_{3\left(\mathrm{g}\right)};\\ & \mathrm{T}=180-{360}^{\circ }\mathrm{C}\end{array}$(7)

(8) $2\mathrm{R}\mathrm{E}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{l}\right)}+3\mathrm{M}{\mathrm{g}}_{\left(\mathrm{l}\right)}=3\mathrm{M}\mathrm{g}\mathrm{C}{\mathrm{l}}_{2\left(\mathrm{l}\right)}+2{\left[\mathrm{R}\mathrm{E}\right]}_{\mathrm{M}\mathrm{g}}$(8)

(9) $\mathrm{\Delta }G=\mathrm{\Delta }{G}^{\circ }+RTln\frac{{\alpha }_{MgC{l}_2}^3+{\alpha }_{R{E}_{\left[Mg\right]}}^2}{{\alpha }_{Al/Mg}^2+{\alpha }_{R{E}_2{O}_3}}$(9)

It has been reported that the yields of RE could reach up to 58%, 84%, and 91%, in the case of Mg-Nd, Al-Mg-Sc, and Al-Mg-Y alloys, respectively (Bazhin et al. 2015; Fu et al. 2022).

This process has also been studied for the process of making WE43A and WE43B magnesium alloys (Zhang et al. 2023). In these studies, a mixture of Y₂O₃, Gd₂O₃, and Nd₂O₃ dissolved in KCl-NaCl was used as the RE source, with the result being an alloy containing 3.92% Y, 2.02% Nd, and 1.47% Gd. The concentrations of impurities such as Na and K were reported to be very low implying the process did not introduce any inclusions to the alloy. It should be noted however, that RECl₃ dissolved in molten chloride salts may be hydrolyzed to REOCl when in contact with water vapor in an ambient atmosphere, and this would decrease the efficiency of the process as REOCl formation would contribute to RE loss and inhibit the reaction process. The RECl₃ hydrolysis could be decreased by lowering the molten salt surface area (interface between molten salts and gas) and increasing the height of molten salt (Zhang et al. 2023). This is because the hydrolysis of RECl₃ was found to be controlled by the diffusion of RECl₃ to the interface between the molten salts and the ambient atmosphere.

After the alloy production process, the used chloride salt mixture will be contaminated with MgCl₂ (Equation (10))(10) $2\mathrm{R}\mathrm{E}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{l}\right)}+3\mathrm{M}{\mathrm{g}}_{\left(\mathrm{l}\right)}=3\mathrm{M}\mathrm{g}\mathrm{C}{\mathrm{l}}_{2\left(\mathrm{l}\right)}+2{\left[\mathrm{R}\mathrm{E}\right]}_{\mathrm{M}\mathrm{g}}$(10) . In order to reuse the chloride salt mixture, the MgCl₂ has to be removed because it can potentially react with RE elements in the Mg-RE alloy, i.e. the reverse of Equation (10)(10) $2\mathrm{R}\mathrm{E}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{l}\right)}+3\mathrm{M}{\mathrm{g}}_{\left(\mathrm{l}\right)}=3\mathrm{M}\mathrm{g}\mathrm{C}{\mathrm{l}}_{2\left(\mathrm{l}\right)}+2{\left[\mathrm{R}\mathrm{E}\right]}_{\mathrm{M}\mathrm{g}}$(10) can occur. MgCl₂ can be removed by reacting the spent chloride salt mixture with REF₃ (Equation 11(11) $2\mathrm{R}\mathrm{E}{\mathrm{F}}_{3\left(\mathrm{l}\right)}+2\mathrm{M}\mathrm{g}\mathrm{C}{\mathrm{l}}_{2\left(\mathrm{l}\right)}=2\mathrm{R}\mathrm{E}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{l}\right)}+3\mathrm{M}\mathrm{g}{\mathrm{F}}_{2\left(\mathrm{l}\right)};\mathrm{T}={800}^{\circ }\mathrm{C}$(11) ). In this case, the MgCl₂ is converted into MgF₃ and removed via a dilute acid leaching and filtration process (Fu et al. 2022; Zhang et al. 2023).

(10) $2\mathrm{R}\mathrm{E}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{l}\right)}+3\mathrm{M}{\mathrm{g}}_{\left(\mathrm{l}\right)}=3\mathrm{M}\mathrm{g}\mathrm{C}{\mathrm{l}}_{2\left(\mathrm{l}\right)}+2{\left[\mathrm{R}\mathrm{E}\right]}_{\mathrm{M}\mathrm{g}}$(10)

(11) $2\mathrm{R}\mathrm{E}{\mathrm{F}}_{3\left(\mathrm{l}\right)}+2\mathrm{M}\mathrm{g}\mathrm{C}{\mathrm{l}}_{2\left(\mathrm{l}\right)}=2\mathrm{R}\mathrm{E}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{l}\right)}+3\mathrm{M}\mathrm{g}{\mathrm{F}}_{2\left(\mathrm{l}\right)};\mathrm{T}={800}^{\circ }\mathrm{C}$(11)

3.2. Fluoride salt system

In contrast to the chloride system, the fluoride salt system has been mainly used for making Al-RE alloys. There are only limited studies on the production of Mg-RE alloys using fluoride salt systems. One of the complexities associated with this route relates to the higher operating temperatures (T > 750°C) required which are not suitable for a relatively low boiling point metal such as Mg. The high temperatures are required because MgF₂ forms in the fluoride system and this phase has a high melting point (~1263°C), which in turn increases the melting point of the mixed molten salts.

When fluoride salts are used to make Al-RE alloys, however, an RE oxide can be dissolved directly into the molten salts without the need of a prior (and separate) fluorination process. The RE oxide will react with the fluoride salt (usually a mixture of NaF and KF) to form a complex compound dissolved in the molten fluoride salts, K₂NaREF₆ (Skachkov et al. 2018). This intermediate complex solution then reacts with Al to form the Al-RE alloy. It has been reported that yields of 93% for Sc and 61% for Y, respectively, can be obtained in a system involving Sc₂O₃/Y₂O₃-NaF-KF-Al (Skachkov et al. 2018).

RE fluorides could also be used as the source of RE to be incorporated in the LM alloy. In a previous study, a mixture of KF and NaF (1:1) was mixed with YF₃ (5 wt.%) and then reacted with Al (Yatsenko, Skachkov, and Pasechnik 2020). The resulting Al-Y alloy was reported to contain 30.5 wt.% Yttrium. The study also described a method for the enrichment of an Al-Sc alloy. The use of a mixture of Sc oxide/oxyfluoride/fluoride produced <10 wt.% Sc in the resulting alloy, however, with further processing through filtration, the average Sc concentration in the alloy was increased to 20.3 wt.% (Yatsenko, Skachkov, and Pasechnik 2020).

3.3. Mixed chloride-fluoride salt system

When a mixed chloride-fluoride salt system is used, RE oxides react and dissolve in the molten salts to form intermediate solutions (depending on the specific salt system used). Different molten chloride-fluoride salt mixtures that have been used include, NaF-NaCl-KCl, NaF-AlF₃-KCl, NaF-KCl, and KCl-KF-AlF₃.

In the NaF-NaCl-KCl-Al system, because the diameter of the O^2- ions in the RE oxide and the F ions in molten salts are similar, the oxygen and fluorine may exchange with each other and transform the RE₂O₃ to ${\mathrm{R}\mathrm{E}\mathrm{O}\mathrm{F}}_{\mathrm{x}}^{\mathrm{n}-}$ (Equation (12)(12) $2\mathrm{N}\mathrm{a}\mathrm{F}+\mathrm{R}{\mathrm{E}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}=2\mathrm{R}\mathrm{E}\mathrm{O}\mathrm{F}+\mathrm{N}{\mathrm{a}}_2\mathrm{O}$(12) ) (Jang et al. 2015; Xu et al. 2012). The formation of ${\mathrm{R}\mathrm{E}\mathrm{O}\mathrm{F}}_{\mathrm{x}}^{\mathrm{n}-}$ increases the solubility of RE oxide in the molten salts and reacts with molten Al to form the Al-RE alloy (Equation (13)(13) $3\mathrm{R}\mathrm{E}\mathrm{O}\mathrm{F}+3\mathrm{A}\mathrm{l}=\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3+\mathrm{A}\mathrm{l}{\mathrm{F}}_3+3{\left[\mathrm{R}\mathrm{E}\right]}_{\mathrm{A}}$(13) ). In the case of the Al-La system, the La content in the alloy can reach up to 10.1 wt.% when La₂O₃ is used as precursor (Jang et al. 2015). The solubility of La₂O₃ was reported to increase with an increase in the NaF concentration. In the case of the Al-Sc system, the spent molten salts were tested and able to be reused up to three times (after mixing with fresh Sc₂O₃) with the yield reported to reach up to 91% in the case of Sc₂O₃ being used (Xu et al. 2012).

(12) $2\mathrm{N}\mathrm{a}\mathrm{F}+\mathrm{R}{\mathrm{E}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}=2\mathrm{R}\mathrm{E}\mathrm{O}\mathrm{F}+\mathrm{N}{\mathrm{a}}_2\mathrm{O}$(12)

(13) $3\mathrm{R}\mathrm{E}\mathrm{O}\mathrm{F}+3\mathrm{A}\mathrm{l}=\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3+\mathrm{A}\mathrm{l}{\mathrm{F}}_3+3{\left[\mathrm{R}\mathrm{E}\right]}_{\mathrm{A}}$(13)

In the NaF-AlF₃-KCl-Al system, the RE₂O₃ is initially fluorinated to form KRE₃F₁₀ (Equation (14))(14) $3\mathrm{R}{\mathrm{E}}_2{\mathrm{O}}_3\left(\mathrm{s}\right)+5\mathrm{N}\mathrm{a}\mathrm{F}+7\mathrm{A}\mathrm{l}{\mathrm{F}}_3+5\mathrm{K}\mathrm{C}\mathrm{l}=2\mathrm{K}\mathrm{R}{\mathrm{E}}_3{\mathrm{F}}_{10}+3\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}+5\mathrm{N}\mathrm{a}\mathrm{C}\mathrm{l}$(14) which is then reduced by Al (Equation (15)(15) $3\mathrm{K}\mathrm{R}{\mathrm{E}}_3{\mathrm{F}}_{10}+9\mathrm{A}{\mathrm{l}}_{\left(\mathrm{l}\right)}=9{\left[\mathrm{R}\mathrm{E}\right]}_{\mathrm{A}\mathrm{l}}+{\mathrm{K}}_3\mathrm{A}\mathrm{l}{\mathrm{F}}_6+8\mathrm{A}\mathrm{l}{\mathrm{F}}_3$(15) ). AlF₃ is an important component in the molten salt mixture because it can enhance the fluorinating effect on the RE oxide and it promotes the formation of KRE₃F₁₀ in the presence of KF. The KF forms through reaction between KCl and AlF₃ and by high temperature ion exchange, which likely proceeds via aluminum sub-halides formation.

(14) $3\mathrm{R}{\mathrm{E}}_2{\mathrm{O}}_3\left(\mathrm{s}\right)+5\mathrm{N}\mathrm{a}\mathrm{F}+7\mathrm{A}\mathrm{l}{\mathrm{F}}_3+5\mathrm{K}\mathrm{C}\mathrm{l}=2\mathrm{K}\mathrm{R}{\mathrm{E}}_3{\mathrm{F}}_{10}+3\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}+5\mathrm{N}\mathrm{a}\mathrm{C}\mathrm{l}$(14)

(15) $3\mathrm{K}\mathrm{R}{\mathrm{E}}_3{\mathrm{F}}_{10}+9\mathrm{A}{\mathrm{l}}_{\left(\mathrm{l}\right)}=9{\left[\mathrm{R}\mathrm{E}\right]}_{\mathrm{A}\mathrm{l}}+{\mathrm{K}}_3\mathrm{A}\mathrm{l}{\mathrm{F}}_6+8\mathrm{A}\mathrm{l}{\mathrm{F}}_3$(15)

An incomplete fluorination can occur when the NaF is present in high concentrations. In this case, the NaF can react with KF and AlF₃ to form K₂NaAlF₆, which reduces the available KF and AlF₃ for further reaction with RE₂O₃ to form KRE₃F₁₀. An incomplete RE₂O₃ fluorination was also reported to promote the formation of a binary oxide phase of Al₁₀Er₆O₂₄ (5Al₂O₃.3Er₂O₃) in the system NaF-AlF₃-KCl-Al (Kosov, Bazhin, and Kopylova 2019). Therefore, to avoid incomplete fluorination, molten salts with a small Cryolite Ratio (CR) (i.e. a lower NaF concentration in the molten salt mixture) is preferred. When Er₂O₃ is used in the NaF-AlF₃-KCl-Al system, the yield was reported to reach up to 71.43% with 6.45 wt.% Er content in the alloy (Kosov, Bazhin, and Kopylova 2019). KCl-KF-AlF₃ mixtures could also be used as the molten salt. A yield in oxide reduction of more than 90% was reported when the Sc₂O₃-KCl-KF-AlF₃ system was used (Skachkov, Pasechnik, and Yatsenko 2014).

In the NaF-KCl system, REF₃ will react to form NaREF₄ (Equation (16))(16) $\mathrm{R}\mathrm{E}{\mathrm{F}}_3+\mathrm{N}\mathrm{a}\mathrm{F}=\mathrm{N}\mathrm{a}\mathrm{R}\mathrm{E}{\mathrm{F}}_4$(16) and then either be reduced to the metal form and dissolved in the molten Al (Equation (17)(17) $3\mathrm{N}\mathrm{a}\mathrm{R}\mathrm{E}{\mathrm{F}}_4+3\mathrm{A}{\mathrm{l}}_{\left(\mathrm{l}\right)}=3{\left[\mathrm{R}\mathrm{E}\right]}_{\mathrm{A}\mathrm{l}}+\mathrm{N}{\mathrm{a}}_3\mathrm{A}\mathrm{l}\mathrm{F}+2\mathrm{A}\mathrm{l}{\mathrm{F}}_{3\left(\mathrm{g}\right)}$(17) ), or form intermetallic phases, depending on the thermodynamics of the system (Kosov and Bazhin 2018). RE fluorides could also be used in combination with a NaCl-KCl-CaCl₂ salts mixture (Savchenkov and Beloglazov 2022). In this case, the yield has been reported to reach up to 99.8% for the Mg-Zn-Y system.

(16) $\mathrm{R}\mathrm{E}{\mathrm{F}}_3+\mathrm{N}\mathrm{a}\mathrm{F}=\mathrm{N}\mathrm{a}\mathrm{R}\mathrm{E}{\mathrm{F}}_4$(16)

(17) $3\mathrm{N}\mathrm{a}\mathrm{R}\mathrm{E}{\mathrm{F}}_4+3\mathrm{A}{\mathrm{l}}_{\left(\mathrm{l}\right)}=3{\left[\mathrm{R}\mathrm{E}\right]}_{\mathrm{A}\mathrm{l}}+\mathrm{N}{\mathrm{a}}_3\mathrm{A}\mathrm{l}\mathrm{F}+2\mathrm{A}\mathrm{l}{\mathrm{F}}_{3\left(\mathrm{g}\right)}$(17)

Overall, the use of RE fluorides instead of RE oxides improves the yield of the LM-RE alloy process by 2–7% (Skachkov et al. 2018). This method, however, involves adding steps to the process because RE are usually extracted in the form of oxides. The fluorination of RE can also be done by reacting it with HF solutions.

Table 2 shows the summary of the key previous studies in the production of the Mg/Al-RE alloys through multistage reduction using molten salts.

Table 2. Summary of previous works/patents in multistage indirect reduction in molten salts.

Author	Metal	Reaction and parameters	Process details
Fu et al. (2022)	Mg	First step: Nd2O3 + 6NH4Cl = 2NdCl3 + 3H2O + 6NH3 Chlorination of Nd2O3 using NH4Cl as chlorinating agent NdCl3 is mixed with chloride salts (NaCl:KCl=50.6:49.4 mol%) and heated in the furnace at 993 K for 1.5 hours to guarantee the complete removal of NH4Cl. Second step: 2RECl3 + 3[Mg] = 3MgCl2 + 2[RE] Mg solid immersed into the liquid salt and then heated for 18 h. Cooled down in air after furnace cooling up to 773 K. Third step: 2NdF3 + 3MgCl2 = 2NdCl3 + 3MgF2 NdF3 is added to the used molten salts and then heated to 1073 K for 3 h, NdF3 is used to remove MgCl2. Mg–Nd binary alloys via the magnesiothermic reduction method in liquid NaCl–KCl–xNdCl3	Mg solid initially sinks to the bottom because its density is higher than that of molten salts. When the Mg is completely melted, it rises to the top of the molten salts and becomes spherical in shape but remains completely immersed. As the RE begins to dissolve in the alloy, the density rises, and the alloy again sinks in the bottom. 6-18 wt.% of Nd content in the alloy could be made by varying the amount of NdCl3 and MgCl2 in the molten salts. A maximum of ≈58% conversion rate of NdCl3 was achieved
Savchenkov and Beloglazov (2022)	Mg	YF3-35NaCl-35KCl-30CaCl2 wt.% Temperature: 650 to 700 °C Mg:Zn = 1:2 25 min, ratio of chlorides to yttrium fluoride of 6:1, periodic stirring. 1^st: 2Zn + Mg = MgZn2 2^nd: YF3 + MgZn2 = MgxZnyYz + MgF2	Y yield up to 99.1-99.8% at 700 °C Digestibility of yttrium in the produced master alloy ranged from 94-98% while making Mg-6Zn-1Y-0.5Zr alloy. Master alloy composition: Zn-25.6Mg-23.9Y
Ricketts (2021)^p	Al	Sc2O3(s)+3*Al(l)→2Sc[Al]+Al2O3(s) Dual flux layer system 1^st flux consists of KF and AlF3 or KF and K3AlF6. The 1^st flux is mixed with scandium oxide and placed at the bottom of the crucible. It is believed that 1^st flux which have more fluidity and will give higher Sc recovery to the alloys. 2^nd flux is a cover flux which consists of a cheap chloride salts mixture. Temperature: 850 – 900 °C. Reduction time 90 – 120 minutes.	Al-2%Sc alloy could be easily produced after 60 min processing time. 74% Sc recovery.
Savchenkov et al. (2021)	Al	Same as (Ya I. Kosov, Bazhin, and Kopylova 2019)	Adding Er to 8% and above (over the eutectic region) changed the microstructure greatly. Al3Er inclusions were present. Applying faster cooling rates ~100 °C, primary crystal of individual Al3Er were absent thus help the higher rate of dissolution of the master alloy during alloying with Al alloy because of the higher dispersion and uniformity in the phase distribution of the modifier.
Zhang et al. (2023)	Mg	RE2O3 (Y, Gd, Nd) 1^st step: RE2O3 + 6NH4Cl = 2RECl3 + 3H2O + 6NH3 2RECl3(l) + H2O = REOCl(s) + HCl(g) nRE2O3:nNH4Cl = 1:18 were ground KCl-NaCl-RECl3 molten salts, KCl:NaCl = 1:1, RECl3 10-20 wt.% NaCl-KCl-(15.4 wt.%) YCl3-(2.7 wt.%) NdCl3-(1.0 wt.%) GdCl3 then 600 gr of NaCl and KCl was placed on top. Heated at 750 °C for 0.5 h. 2^nd step: 2RECl3 + 3[Mg] = 3MgCl2 + 2[RE] Mg rod was placed horizontally for magnesiothermic reduction at 750 °C.	Suitable for commercial alloy production. The reaction is easily controlled by the content of MgCl2 in the molten salt. The obtained alloy went up and down as a whole in molten salts which improve process safety without resulting a lot of chloride inclusion. Chloride inclusions were only 0.0019 wt.% in the Mg alloy. RECl3 in the molten salts are hydrolyzed to REOCl and this hydrolysis contributed to the RE loss and inhibited the reaction process via adhering on the surface of the liquid alloy. The hydrolysis could be reduced by increasing the H/D of molten salts. Results: Mg-(3.92 wt.%) Y-(2.02 wt.%) Nd-(1.47 wt.%) Gd. Most of the RE were dissolved in the Mg matrix because of the high cooling rate and no intermetallic phases were found on XRD but present in SEM-EDX results.
Yatsenko et al. (2020)	Al	For element heavier than Al: KF-NaF salt mixture with a ratio of 1:1 and adding alloying compound based on their solubility: up to 5-10 wt.% of fluorides or oxyfluorides or 5 wt.% of oxides. Remelting the mixture into homogenous mass up to 800 °C. Inserting granulated Al at 750-780 °C. Holding for 30 min with occasional stirring. Removal of the salt melt.	Very high content of Sc in the alloy (>20%). 30.5 wt.% Y content in the alloy. 20.3 wt.% Sc content in the alloy.
		Filtering Process: Exposing the resulting alloy melt at 750±25 °C for 15 min to form an agglomerated intermetallic compound. Centrifugation of the alloy with 1000-2500 rpm rotor speed for 10 min with a specially designed filter. Removal of sediment mechanically from the bottom along the outer boundary.
		For Al-Y, use salt melts of NaF and KF with a ratio of 1:1 and 5% YF3, 40.5 gr KF, 40.5 gr NaF, and 4.3 gr YF3 and salt melt temperature of 750 °C. Exposing the alloy for 775±25 °C for 20 min and centrifuge at 2000 rpm for 10 min.
Min et al. (2019)^p	Al-Ca alloy	Sc2O3 - CaCl2 flux. Al-Ca alloy is used as the starting alloy and Ca is acting as a reductant. Content of Ca as a reducing agent is 6-20 wt.% Sc content in alloy is 1.2 to 2% and 3 to 4.5% when Ca content in the original alloy is 10 and 20 wt.%, respectively. Temperature: 1000 °C.	The flux used can reduce the activity of the reducing metal oxide which is produced as a by-product of the reduction reaction thus increasing the power of the reducing agent. The Sc activity (or concentration) in the alloy could be controlled by adjusting the amount of the reducing agent oxide in the flux.
Kosov et al. (2019)	Al	Er2O3-NaF-AlF3-KCl with Al Fluorination: 3Er2O3 + 5NaF + 7AlF3 + 5KCl = 2KEr3F10 + 3Al2O3 + 5NaCl Reduction: 3KEr3F10 + 36Al = 9Al3Er + K3AlF6 + 8AlF3	The use of only chloride salts barely has any effect on Er2O3 reduction. Adding NaF to chloride salts improved Er yield and the yield was higher compared to adding AlF3. Adding both NaF and AlCl3 to the salts gave the best results. Best reaction parameters: Cryolite ratio: 1.56; T: 900 °C, melt stirring. Results: 71.43 % Erbium recovery with 6.45 wt.% Er content Higher temperatures and lower CR increases the yield.
Kosov and Bazhin (2018)	Al	ErF3, NaF, KCl -> chloride fluoride melt (flux) Temperature: 800 °C. ErF3 + NaF = NaErF4 2NaErF4 + 12Al = 3Al3Er + Na3AlF6 + 2AlF3	Increasing of the Er content in master alloy from 3.4 to 4.2 wt.% was followed by an increase of Er in eutectic from 12.5 to 16 wt.%. But further increases to 6.1 wt.% of Er in the master alloy decreased the wt.% of Er in eutectic to 8.6 through coarsening of intermetallic particles of Al3Er by diffusion resistance and reduction in the dissolution of Er in Al. Up to 6.1 wt.% Er in master alloy.
Skachkov (2018)	Al	0.4NaF + 0.6KF Sc2O3 or Y2O3, NaF, and KF react to form K2NaScF6 and K2NaYF6 Salt/metal ratio: Sc = 0.8; Y = 1.0 Temperature: 730-750 °C. Injected by mixture of CO2.	Replacing oxides with fluorides increased the yield from 2% to 7%. Sc2O3 → 0.91 wt.% Sc content and 93% direct yield. ScF3 → 0.98 wt.% Sc content and 96% direct yield. Y2O3 → 0.95 wt.% Y content and 61% direct yield. YF3 → 1.15 wt.% Y content and 69% direct yield. Oxides or fluorides are ground with fluorides salts. The ground powder is very hygroscopic and needs to be used immediately or stored in a suitable manner.
Kulikov et al. (2018)	Al	Sc2O3 + 6HF = 2ScF3 + 3H2O. ScF3 + 4Al = Al3Sc + AlF3 AlF3 + 3NaF = Na3AlF6 ScF3 + 3NaF = Na3ScF6 Briquette of ScF3 28.56 g, NaF 35.28 g, and Al 10 g were immersed into molten Al beneath flux of 39.0 g KCl and 39.0 g NaCl. Temperature: 860 ± 10 °C. Time: 30 min. Periodically stirred.	ScF3 reduction by Al is not favorable, but adding NaF to the system, shifts the reaction to the right. This is because AlF3 will react with NaF to form cryolite. Master alloy weight: 586 g Sc content 1.90 ± 0.04 wt.% Sc extraction 88.5% These results increased from Sc extraction of ≈78% without Al powder addition into the briquette. This is because adding high specific surface Al powder resulted in better contact of the reacting phases. Loss of Sc during the process were attributed to pyro-hydrolysis of ScF3 with the formation of Sc2O3 and mechanical losses of Sc compounds because of incomplete separation of salts and metal phases due to increases in fluoride-chloride slag viscosity because of dissolution of Al2O3.
Bazhin et al. (2015)	Al-Mg	Al–2% Sc–0.5% Y master alloy 0.5Sc2O3 + 3Al + 1.5Mg → Al3Sc + 1.5MgO (–80 kJ/mol) 0.5Y2O3 + 3Al + 1.5Mg → Al3Y + 1.5MgO (–67 kJ/mol) Al-Mg alloy was mixed with flux and then heated in a vacuum. Sc2O3 and Y2O3 were prepared from red mud and the concentration reached up to 80-120 and 2.5-4.5 g/t, respectively. Add alkali metal halide salts, manganese oxide, Sc2O3 and Y2O3 Temperature: 850-900 °C. Sc2O3 and Y2O3 are fed constantly. Mixing rate 10 cm/s. Time: 1 h.	Enlarged experiment (18 dm^3 in volumes) resulting in 84 and 91 % yield for scandium and yttrium, respectively. 2.5 wt.% of Sc in the alloy could be reach when using Al alloy containing 15-17 wt.% of Mg.
Jang et al. (2015)	Al	Certain content of NaF, mixture of NaCl and KCl (44:56), and La2O3 heated at 720-780 °C. And then Al was added. Time: 10, 20, 30, 40, 50, and 60 min. Temperature: 750, 800, 850, 900 °C. 2NaF + La2O3 = 2LaOF + Na2O 3LaOF + [3Al] = Al2O3 + AlF3 + [3La]	Solubility of La2O3 increased with the increase in NaF concentration. Kinetics analysis showed that the reaction agreed with a first order reaction model. La content in Al-La master alloy was 10.1 wt.% after reduction at 800°C ± 5°C for 120 min in molten salts of 50 wt.%NaF-22 wt.%NaCl-28 wt.% KCl with 7 wt.% of La2O3.
Skachkov et al. (2014)	Al or Al alloys	50–60% KCl, 15-35% KF, 5–10% AlF3 – 10-25% oxide(s) mixture. Powder mixture of salts and oxides were injected into the molten metal through CO2 gas. Gas blowing was carried out for 1-5 min after complete powder injection.	0.3-0.6 wt.% alloying elements in the alloy. Yield both at individual and joint introduction of alloying element at > 90%. Uniform distribution of dopants in the volume.
Skachkov et al. (2013)	Al	Temperature 780-800 °C.	Replacing alkaline components by calcium ones in halogenides mixture improved the purity in the final alloy.
Xu et al. (2012)	Al	Melting Sc2O3, NaF, KCl, NaCl, ScF3 and Na3AlF6 mixture under a mass ratio of 3:5:10:10:2:30, followed by solidifying and crushing. The mixture was added to aluminum melt with a mass ratio of 60:100Temperature: 1150 K. Time: 60 min.	Adding NaF to the salt mixture promote the formation of $Al{F}_x^{n-}$and made the salts enter the Al melt easily. Because of the ionic diameter of O^2- and F^− is close to each other, 0.14 nm and 0.133 nm, respectively, they may exchange for each other and form $ScO{F}_x^{n-}$ Reusing molten salts residue for 3 times could make the Sc recovery rate to 91%
Varchenya et al. (2009)	Al	Time: 15-20 min. Highest yield (ScF3:96%, 2.6% Sc in alloy, with AlF3 and NaF addition; Sc2O3:80.5%, 1.5% Sc in alloy, with AlF3 and NaF addition). Addition of ZrOF6 and/or Hf2OF6 increases the yield of Sc2O3 extraction.	ScF3 gives higher yield (~96%) and lower temperature (800 – 850 °C) compared to Sc2O3 (~80%; 850 – 900 °C). Yield of scandium decreases without AlF3 addition. KHF2 fosters the reaction.

4. Electrolysis route for production of Mg/Al-RE alloy

The process of making RE alloys using electrolysis may be divided into several types based on the form of the cathode used, the molten salt system used, and the deposition/co-deposition of the alloying element. These processes have previously been summarized by Krishnamurthy and Gupta (2001) however unlike the two previous methods which depends on the differences of chemical stability between RE compound and Al/Mg compound, the electrolytic process is not limited by chemical stability.

In the process of RE alloy production using electrolysis, the RE compounds are dissolved in the molten salts electrolyte and reduced with the application of an electrical current. The RE compound is reduced at the cathode (Equation (18))(18) $\mathrm{R}{\mathrm{E}}^{3+}+3{\mathrm{e}}^{-}\to \mathrm{R}\mathrm{E}$(18) and carbon dioxide gas is formed at the anode (Equation (19)(19) ${\mathrm{O}}^{2-}+\mathrm{C}\to \mathrm{C}{\mathrm{O}}_2+4{\mathrm{e}}^{-}$(19) ). The overall reaction of this electrolysis process, when a RE oxide is used, is provided in Equation (20)(20) $\left(\mathrm{R}{\mathrm{E}}_2{\mathrm{O}}_3\right)+3/2\mathrm{C}=2\mathrm{R}\mathrm{E}+3/2\mathrm{C}{\mathrm{O}}_2$(20) .

An RE alloy could also be made by reducing the RE compound simultaneously with a LM compound. When a LM compound is present in the system, this compound will also be reduced at the cathode and an example of overall reaction involving a LM oxide is provided in Equation (21)(21) $\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3/3\mathrm{M}\mathrm{g}\mathrm{O}+3/2\mathrm{C}=2\mathrm{A}\mathrm{l}/3\mathrm{M}\mathrm{g}+3/2\mathrm{C}{\mathrm{O}}_2$(21) . This co-reduction of RE and LM, especially when using Al, could potentially be conducted using a modified Hall-Héroult cell while producing Al from alumina.

(18) $\mathrm{R}{\mathrm{E}}^{3+}+3{\mathrm{e}}^{-}\to \mathrm{R}\mathrm{E}$(18)

(19) ${\mathrm{O}}^{2-}+\mathrm{C}\to \mathrm{C}{\mathrm{O}}_2+4{\mathrm{e}}^{-}$(19)

(20) $\left(\mathrm{R}{\mathrm{E}}_2{\mathrm{O}}_3\right)+3/2\mathrm{C}=2\mathrm{R}\mathrm{E}+3/2\mathrm{C}{\mathrm{O}}_2$(20)

(21) $\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_3/3\mathrm{M}\mathrm{g}\mathrm{O}+3/2\mathrm{C}=2\mathrm{A}\mathrm{l}/3\mathrm{M}\mathrm{g}+3/2\mathrm{C}{\mathrm{O}}_2$(21)

A general schematic of the electrolysis route for production of Mg/Al-RE alloys is presented in Figure 6. The electrolysis process can be performed using an inert or a reactive cathode. If an inert cathode is used, the formed alloy will sink to the bottom of the electrolysis cell because the wettability of the alloy with the cathode is poor. In comparison, when a reactive cathode is used, the product will stick to the cathode surface when the processing temperature is below the melting point of the cathode, or the product will dissolve into the cathode (when the processing temperature is above the melting point of the cathode).

Figure 6. Generic schematic of the electrolysis route for production of a Mg/Al-RE alloy.

The LM-RE electrolysis processes may be differentiated on the basis of the molten salt system used. Chlorides, fluoride, and mixed chlorides-fluorides salts could all be used for the electrolysis. The molten salts system selection is usually based on the electrochemistry and techno-economics of the system. The electrodeposition energy of the metal halides (Mg, Al, or RE) should be lower than the salts halides.

For making Mg-RE alloys, most previous studies have been conducted using chloride salts systems. This is because in order to use fluoride salts, cheap fluoride salts, such as NaF or KaF, cannot be used because their electrodeposition energies are lower than that of MgF₂ (Figure 7) (Vignes 2013). Instead, more expensive LiF molten salts must be used.

Figure 7. Reversible decomposition voltages for sulfides, oxides, chlorides and fluorides at 1,000 K (from Vignes 2013).

For making Al-RE alloys, both chloride and fluoride salt system can be used however the trend of previous studies has been toward fluoride systems because of its similarity to the Hall-Héroult system and the alloy manufacturing process can potentially be integrated into existing Hall-Héroult cells. Also, for making Al-RE alloys in chloride salt systems, the Al₂O₃ needs to first be converted into AlCl₃, unlike for the process for making the Al-RE alloys in the fluoride system route in which Al₂O₃ can be used directly. The conversion process needs Cl₂ gas as a reactant which will increase the cost and complexity of the system.

The RE compounds used in the electrolysis process could also be reacted with the molten salts (the same process compared to metallothermic reduction of RE compound dissolved in the molten salts route) before being electrolyzed. When a RE oxide is used as the starting material, it will react to form an RE fluoride or a RE chloride based on Equations (22)(22) $\mathrm{R}{\mathrm{E}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}+\mathrm{A}\mathrm{l}{\mathrm{F}}_{3\left(\mathrm{l}\right)}\to \mathrm{R}\mathrm{E}{\mathrm{F}}_{3\left(\mathrm{l}\right)}+\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}$(22) to (24(24) $\mathrm{R}{\mathrm{E}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}+3\mathrm{M}{\mathrm{g}}_2\mathrm{C}{\mathrm{l}}_{2\left(\mathrm{g}\right)}\to 3\mathrm{M}\mathrm{g}{\mathrm{O}}_{\left(\mathrm{s}\right)}+2\mathrm{R}\mathrm{E}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{l}\right)}$(24) ).

(22) $\mathrm{R}{\mathrm{E}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}+\mathrm{A}\mathrm{l}{\mathrm{F}}_{3\left(\mathrm{l}\right)}\to \mathrm{R}\mathrm{E}{\mathrm{F}}_{3\left(\mathrm{l}\right)}+\mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}$(22)

(23) $\mathrm{R}{\mathrm{E}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}+\mathrm{A}{\mathrm{l}}_2\mathrm{C}{\mathrm{l}}_{6\left(\mathrm{g}\right)}\to \mathrm{A}{\mathrm{l}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}+2\mathrm{R}\mathrm{E}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{l}\right)}$(23)

(24) $\mathrm{R}{\mathrm{E}}_2{\mathrm{O}}_{3\left(\mathrm{s}\right)}+3\mathrm{M}{\mathrm{g}}_2\mathrm{C}{\mathrm{l}}_{2\left(\mathrm{g}\right)}\to 3\mathrm{M}\mathrm{g}{\mathrm{O}}_{\left(\mathrm{s}\right)}+2\mathrm{R}\mathrm{E}\mathrm{C}{\mathrm{l}}_{3\left(\mathrm{l}\right)}$(24)

The Al-Sc system is the most studied electrolysis route to produce LM-RE alloys, because Sc is the most effective alloying element strengthener for Al alloys (Dorin et al. 2018). It has been reported that the current efficiency for the production of a Al-Sc alloy could reach up to 63% for a NaF/AlF₃ molten salt system with a graphite crucible acting as the cathode (Martinez et al. 2021) and 65.2% for KF/AlF₃ with a pre-wetted tungsten cathode (Nikolaev, Suzdaltsev, and Zaikov 2019). An example of an electrolytic cell set up for producing LM-RE alloys using a reactive cathode is presented in Figure 8 (Liu et al. 2019b). Previous studies reported segregation of intermetallics in the alloy product (Harata et al. 2009). Segregation in the product is also reported in the direct metallothermic reduction form RE compound route (Xiao et al. 2020). For the direct metallothermic reduction form RE compound route, common mixing methods such as stirring or gas blowing could be employed to reduce segregation, while for the electrolysis route, these methods cannot be employed because it will disturb the electrolysis process.

Figure 8. Typical electrolytic cell for making Mg/Al-RE alloys using a liquid Al metal cathode (Liu et al. 2019ba.

To improve the homogeneity of the alloy, ultrasound could potentially be applied during the process to facilitate redistribution of the intermetallics, as shown in Figure 9 (Liu et al. 2019a). However, an ultrasound method has not been commercially used. The electrolysis process may also be used to simultaneously separate an individual RE from a RE element mixture while also producing a LM-RE alloy (Wang et al. 2015; Zhang et al. 2013). For example, ErCl₃ can be successfully separated from a ErCl₃-PrCl₃ mixture with a separation efficiency up to 95.3% while also producing a Mg-Er alloy (Wang et al. 2015). In comparison, for a SmCl₃-DyCl₃ mixture, DyCl₃ was reported to reduce and form Mg-Dy alloy, while SmCl₃ was not (Zhang et al. 2013). Table 3 summarizes previous studies on the electrolysis route to produce various LM-RE alloys.

Figure 9. The mechanism of refining of the Al₃Sc phase by ultrasound (Liu et al. 2019a).

Table 3. Summary of previous works on electrolysis production of LM-RE alloys.

Author	Metal	Reaction and parameters	Process details
Wang (2022)	Al	72 h electrolysis constant at 250-270 A. An industrial Hall-Héroult cell bath was used containing 5 wt.% CaF2 and 3 wt.% Al2O3. Industrial grade AlF3 and analytical grade NaF were used to adjust the bath composition. Anode: Graphite. Cathode: Metal pool. T: 955-965 °C. Bath composition target: Ratio 1.12. Initial bath: metal pad = 18:3. 15-20 kg bath. Al-Sc Alloy: Given the high level of Sc in the bath, Sc should have been converted into fluorides based on the reaction: Sc2O3 + AlF3 → ScF3 + Al2O3 7.3 kg of Al-Sc alloy was tapped in the end of the 1^st run with the composition of 1% Sc. Final bath contain 4.5% Sc. 1^st run: 250 A. 2^nd run: 270 A. Al-RE alloy: (RE: Ce, La, Nd, Pr). In the RE run, instead of using RE2O3, RE2(CO3)2 was used. It will decompose in 400-500 °C based on this reaction: RE2(CO3)2 → RE2O3 + 3CO2 and will be dissolved in the electrolyte and reduced during electrolysis based on this reaction: 2RE2O3 + 3C = 4RE + CO2 High concentration of RE in the bath leads to the formation of RE fluorides. RE2O3 + AlF3 → REF3 + Al2O3 To maintain the stability of the electrolyte, fluoride RE compound need to be added.	A 3.5% Sc master alloy was obtained. The bath needs to be modified/optimized with ScF3 and the oxide content may have to be strictly controlled. Total oxygen in the bath may have to be controlled below 1.5% and Sc2O3/Al2O3 molar ratio maintained above 1.0. Appropriate way of feeding Sc2O3 is needed due to the non-flowability and powdery nature of Sc2O3. RE element in the alloy is proportional to their composition in the oxide. RE concentration in the alloy could be controlled by controlling the ration of RE2(CO3)3 to Al2O3. The partition of RE in the bath and the metal pool is a function of cell voltage, bath temperature, and RE concentration in the alloy and the bath. [REO]bath/[RE]Al = f(V, T,CREIAt,CREOlbath,- ..) 4% concentration of RE in the bath had to be maintained to get 8% of RE in the alloy.
Martinez et al. (2021)	Al	NaF/AlF3 molar ratio of 2.2 at 980 °C, 10 wt.% Sc2O3 All were runs at galvanostatic conditions with constant applied current at 0.79-0.93 A cm^−2 and 0.45 A cm^−2, for small lab scale and large scale, respectively. Discarded dross material from Al-Sc manufacturing was used and after calcination the composition is: 0.7 wt.% Al. nitride; 6 wt.% Sc2O3; 82 wt.% Al2O3; 1 wt.% Al; 11 wt.% aluminum oxide nitride. Sc2O3(diss) + 2AlF3(diss) = Al2O3(diss) + 2ScF3(diss) Lab scale electrolysis: steel and TiB2 cathode and, CR: 2.2, 10 wt.% Sc2O3 at 980 °C, no Al2O3 was added. Long term lab scale electrolysis: bottom of graphite container acted as a cathode, graphite with Si3N4 bounded SiC tube as side-lining as anode, 24 hr, 220 A during first 9 h and then 120 A, Al2O3 was continuously fed, Sc2O3 and electrolyte was added not continuously.	Product with good wettability formed in the case of using steel as the cathode, but not in the case of inert TiB2 as the cathode. Metal drops of different size sink to the bottom of the crucible Lab scale electrolysis: Current efficiency was 77-85 %. The amount of Sc ranging from 1.87 – 2.6 wt.% with current density 0.79 – 0.93 A/cm^2, respectively. 0.6-2.6 wt.% Sc content in the alloy depending on the current density applied. Oxide saturation value is 11 wt.%. Long term Electrolysis: CF4 emission is equal to those of commercial Al electrolysis cell. CO/CO2 ratio increased with electrolysis time probably because of reaction between Al and CO2 combined with Boudouard reaction: 2Al(diss) + 3CO2(g) = 3CO(g) + Al2O3(diss) CO2(g) + C(s) = 2CO(g) Total metal product in the end of the process was ≈750 g, ≈0.6 wt.% Sc, current density 0.45 A/cm^2, and current efficiency ≈63%. It is possible to co-deposit Sc in an Al matrix because of underpotential deposition of Sc at activities lower than 1. Sc content in the alloy depends on the current density applied.
Li et al. (2022)	Mg-Mn	LiCl-KCl-MgCl2 LaCl3-MnCl2 LiCl:KCl 45:55 wt.%. cyclic voltammetry (CV), square wave voltammetry (SWV) and chronopotentiometry (CP) method was used. Mo electrode. T: 773 K.	Electrochemical formation of Mg–La–Mn alloys by co-reduction of Mg^(II), La^(III) and Mn^(II) in LiCl+KCl molten salts. Reduction/oxidation potential of Mg^(II) observed at -1.76 V/-1.53 V in LiCl-KCl-1.2%MgCl2. Underpotential deposition also observed at -1.42 V. Mg^2+ +2e^− = Mg Reduction/oxidation potential of La^(III) observed at -2.05 V/-1.93 V in in LiCl-KCl-1.2%LaCl3. La^3+ +3e^− = La
			UPD of La in Mg deposited on the cathode by this reaction. xMg^2+ + 2xe^− = xMg La^3+ + 3e^− + xMg → LaMgx
Liu et al. (2020)	Al	KF-AlF3-CaF2-Na3AlF6 -Sc2O3 Liquid Al electrode. Temperature: 800 °C. 1 A/cm^2, 0.5-2 h. Cathode: liquid aluminum. Anode: graphite.	Ultrasound transform the AlSiSc intermetallic to short rod (40 μm) to coarse rhombic tubes (≈205 μm); AlCuSc from ≈100 to ≈30 μm. The effect of ultrasonic refining mainly acts in the solidification process.
Nikolaev et al. (2019)	Al	KF-AlF3-Al2O3-Sc2O3 Prepared melts only contain 0.06 ± 0.01 wt.% oxygen content. Sc2O3 was added using corundum tube. Melts mass was ≈20 g. Electrochemical measurement: KF/AlF3 = 1.5 mol/mol, 1 wt.% Al2O3 at 800 °C Tungsten working electrode with Al reference electrode and a graphite cell auxiliary electrode. Temperature: 750-800 °C. Electrolysis test: 20 A 500-520 g of KF-AlF3-(Al2O3, Sc2O3) melt. Tungsten plate cathode with 30 cm^2 side surface. Fine dense graphite anode with 40 cm^2 side surface. Sc2O3 was periodically added to the melt. Sc2O3 is started to add into the melts after 30-60 min with the rate of 8-12 g/hr (depends on cathode current density). Al2O3 + 3/2C = 2Al + 3/2 CO2(g) Sc2O3 + 3/2C = 2Sc + 3/2 CO2(g) 3Al2O3 + Sc2O3 + 4C = 2Al3Sc + 4CO2(g) 5.7% Sc2O3 addition leads to increasing in the cathode limit current of the Al and Sc co-electrowinning from 0.45 to 0.81 A cm^−2 at temperature of 800 °C.	1^st trial: use of bare tungsten cathode in 800 °C resulting in carbon foam sedimentation and electrolyte superheating because of poor adhesion of the reduced phase and the foreign material of cathode. 2^nd trial: the use of pre-wetted tungsten cathode with Al resulting in stable electrolysis process. Current efficiency: 60.3 – 65.2 % for Al and 10.8 – 14.5 % for Sc for current density: -0.4 – 0.7 A cm^−2 and Temperature: 750-800 °C. Al electrowinning from KF-AlF3-Al2O3 melt occurs in the potential range -0.1 – 1.3 V. Electrowinning from KF-AlF3-Al2O3-Sc2O3 resulting in co-electrowinning of Al and Sc with the same kinetics parameters (cathode current, potential condition). Increasing Sc2O3 addition to 5.7% leads to increasing in the cathode limit current of the Al and Sc co-electrowinning from 0.45 to 0.81 A/cm^−2 at T = 800 °C. Al3Sc is the predominant cathode product. Replacing Sc2O3 with ScF3 resulted in complicated voltammograms because of the formation of Al-Sc intermetallic with different composition. Proposed method: co-electrowinning of Al and Sc and then dissolving the deposited layer on the cathode in liquid Al.
Guo et al. (2019)	Al	Na3AlF6-19 wt.% KF-29 wt.% AlF3-2 wt.% CaF2-4 wt.% Sc2O3 Temperature: 1073 K. 120 min, 1 A/cm^2 and 30 mm anode cathode distance. Cathode: liquid aluminum. Anode: graphite.	Using ultrasound increase the average Sc content from 1.13 to 1.51 wt.% Ultrasound transform AlSi2Sc2 from lamellar and network-like to short-bar shapes. Ultrasound also increased the Vickers hardness.
Liu et al. (2019b)	Al	Na3AlF6-19 wt.% KF-29 wt.% AlF3-2 wt.% CaF2 Current density: 1 and 2 A/cm^2. Temperature: 1023 and 1073 K. Electrolysis time: 0.5, 1, and 2 h. 2 and 4 wt.% Sc2O3 addition. Cooling rates: 0.5, 30, 100 K/s. Cathode: liquid aluminum. Anode: graphite 2Sc2O3(l) + 3C(s) → 4[Sc] + 3CO2(g)	Sc content 0.23-1.38 %. Maximum Sc content reach in 1073 K, 2 A/cm^2, 1h, and 4 wt.%. High current density, temperature, and addition level of Sc2O3 are in favor of increasing the Sc content in the alloy. The Sc content at the edge can be as high as 1.09 wt.% and only 0.24 wt.% in the central area (average 0.75 wt.%). Electrodeposition begins only when the voltage is over 1.67 V. A near spherical discharge between liquid Al and electrolyte may be forming in the electrolysis process, and Sc^3+ is supplied from the top to the bottom of electrolyte in the gap between the graphite anode to liquid Al. Limited diffusion rates of Sc^3+ resulting in accumulation of Sc in the edge of Al. Effect of cooling rates: 0.5 K/s: star-like and fine size. 30 K/s: cusped and clustered. 100 K/s: dendritic and cuboid. Cooling rates have a great impact on the morphology and intermetallic particle size.
Yang et al. (2019)	Al	Na3AlF6-K3AlF6-AlF3 Current density: 1.50 A/cm^2. Al2O3 concentration: 1.0 wt.%. Sc2O3 concentration: 2.0-6.0 wt.%. Cathodic current density: 1.0 – 2.0 A/cm^2.	Sc content in the alloy increases from 0.81 to 3.87 wt.% by increasing the ratio of Sc2O3 to Al2O3 from 2 to 6. Sc content in the alloy rises by increasing the cathodic current density from 1 to 1.5 A/cm^2 but decreased by increasing the cathodic current density further on the concentration of Al2O3 and Sc2O3 of 1.0 and 3.0 wt.%, respectively and 90 min electrolysis time.
Liu et al. (2019a)	Al	Na3AlF6-19 wt.% KF-29 wt.% AlF3-2 wt.% CaF2 and Sc2O3 Parameters: 1073 K, 60 min, 1 A/cm^2. Cooling rates: 0.5 K/s. Anode: Graphite. Cathode: Liquid Al. 90 g of electrolyte (4 wt.% Sc) and 10 g of liquid Al.	Sc content increases up to 1.43 wt.% to 74.4% by applying ultrasound on electrolysis and cooling process. Ultrasound can help the elimination of Sc accumulation in discharge interface (molten salts/liquid Al interface). Huge Al3Sc particles transform into fine cubic particles with the average size reduction from 77 ± 36 μm to 21 ± 8 μm. Refining of the phases because of facilitated nucleation by cavitation and fragmentation of primary phases.
Jang et al. (2019)	Mg-Al	KCl-MgCl2-6 wt.% AlF3- 1 wt.% La2O3 molten salts. KCl:MgCl2 ratio= 50:50 (in mol. %). Mo working electrode. Pure graphite rod as counter electrode Temperature: 800 °C. Cathode current density: 6.92 A/cm^2. Molten salts reaction: KMgCl3 → K^++MgCl3^− → K^++MgCl^++2Cl^− → K^++Mg^2++3Cl^− MgCl2 + La2O3 = 2LaOCl + MgO LaOCl + MgCl2= LaCl3+MgO Anodic reaction: 2Cl^− + 2e^− → Cl2 And the formed Cl2 can react with LaOCl: LaOCl + C + Cl2 LaCl3 + CO Cathodic reaction: Al^3+ + 3e^− → Al Mg^2+ + 2e^− → Mg xAl + La^3+ + 3e^− → AlxLa xAl + Mg^2+ + 2e^− → AlxMg	Co-reduction proceeds when current densities are more negative than -0.29 A/cm^2. Al deposition is at -0.95 V. AlxLa formation is at -1.55 V. Current efficiency ranging from 43.45 to 71.2 %. La content could reach up to 7.38 wt.%, but under the most optimum parameters used (which give 71.2 % current efficiency), the content of the alloy was Mg-27Al-3.17La. Higher temperature increases the La and Al contents while decreases the Mg content. Increasing the cathode current density and electrolysis time increases the Mg content and decreases the La content in the alloy.
Suzdaltsev et al. (2018)	Al	KF–AlF3, NaF–AlF3, and KF–NaF–AlF3 oxide–fluoride melts Temperature: 750-980 °C depending on melt composition. Graphite anode and aluminum cathode, 0.25 – 1 A/cm^2.	Zr extraction was higher compared to Sc extraction. The fraction of the aluminothermic reduction of the oxides during electrolysis at 50–65% for ZrO2 for a cathodic current density of 0.25–0.5 A/cm^2 and at 80–85% for Sc2O3 for a cathodic current density of 0.5–1.0 A/cm^2. Sc and Zr extraction average are 20-75% and 40-100%, respectively, depending on the synthesis parameters. The fraction of the aluminothermic reduction of the oxides during electrolysis at 50–65% for ZrO2 for a cathodic current density of 0.25–0.5 A/cm^2 and at 80–85% for Sc2O3 for a cathodic current density of 0.5–1.0 A/cm^2. 10 wt.% Sc and 15 wt.% Zr content in the alloy. the discharge of scandium-containing ions occurs at a potential from –0.20 to –0.25 V relative to the Al formation potential.
Wang et al. (2022)		Mo electrode LiCl-KCl 50:50 mass ratio; HoCl3 (1.0 wt.%)-MgCl2 (0.3 wt.%). Mo working electrode, spectral pure graphite rod counter electrode. T=777K. Li^+ + e^− = Li Ho^3+ + 3e^− = Ho Ho^3+ + 3e^− + 24/5Mg = 1/5Ho5Mg24 Ho^3+ + 3e^− + 1/7 Ho5Mg24 = 12/7HoMg2 Ho^3+ + 3e^− + HoMg2 = 2HoMg Mg^2+ + 2e^− = Mg	Deposition/dissolution of Ho metal at-2.03/-1.87 V (vs Ag/AgCl)
Polat et al. (2018)	Al	CaCl2 based electrolyte containing 4 wt.% ScF3 (ScF3, NaScF4 and NaSc3F6). Temperature: 800 °C. 1050 Al alloy cathode. 3.2 V applied.	Al-Sc alloy could be applied by applying 3.2 V for electrolysis. Al3Sc intermetallic are formed evenly throughout the Al matrix.
Martinez et al. (2018)	Al	FFC Process. CaCl2 with a small addition of NaCl (typically 5 mol%), some oxide ions were added (Ca 1 mol%) in the form of CaO. Temperature: 900 °C. Sc2O3 or Sc2O3-Al2O3 pellet as cathode. Graphite anode. Al2O3(s) + 3Sc2O3(s) + 24e^− → 2AlSc3(s) + 12O^2− 2O2 → O2 +4e^− (inert anode) or 2O^2−+C → CO2(g) + 4e^− (graphite anode) Overall reaction: Al2O3(s) + 3Sc2O3(s) → 2AlSc3(s) + 6O2(g) (inert anode). Al2O3(s) + 3Sc2O3(s) + 6C → 2AlSc3(s) + 6CO2(g) (graphite anode).	Scandium oxide precursors were reduced at −2.3 and −2.4 V versus the AgCl/Ag reference electrode system for about 6 h. anodic current density 0.12 and 0.05 A cm^−2 for obtained current 1 and 2 A. Pure Sc cannot be produced unless Al metal is present (in the form of Al2O3 mixed with Sc2O3 as a cathode). Solubility of CaCO3, formed by reaction between CO2 and CaO, increased with increasing CaO activity and reduced temperature. Carbonate ions diffuse towards the cathode and undergoing de-oxidation process: $C{O}_3^{2-}+4e^{-}\to C\left(s\right)+3O^{2-}$ The carbon formation caused a short circuit in the cell. This phenomenon implied that the use of an inert cathode could potentially solve the problem.
Jiang et al. (2016)	Mg	LiCl-KCl (50:50 wt. %) MgCl2-LuCl3 1073 K, 2 h, -2 A, W electrodes. LuCl3 concentration: 2 wt.%. Lu^3+ + 3e^− →Lu	From CV: Lu deposition at -2.41 V. MgCl2 concentration: 5, 7, 9, 11, 13, and 15 wt.% and Lu content in the alloy: 23.10, 25.35, 49.86, 55.69, 6.48, 5.35, respectively. MgLu intermetallic formed in the galvanostatic electrolysis sample.
Tian et al. (2016)	Al	Na3AlF6-K3AlF6-AlF3 Temperature: 950 °C. Graphite anode and cathode. 500 gr electrolyte. 1.5 A/cm^2, 2 h, 1223 K. Electrolysis reaction: Sc2O3 — 2Sc + 3/2O2 Al2O3 — 2Al + 3/2O2	Wettability of the product wat the graphite cathode was very poor and the products existed in the shape of balls with holes in some of the product. Ratio of Sc2O3 to Al2O3 should be controlled more than 2:1: to obtain the master alloy. Concentration of Al is maintained above 2 wt.%, therefore Sc concentration should be also maintained above 4 wt.%. Sc was unevenly distributed in the alloy because of the difference in melting point between Sc and Al. Segregation of Sc in the alloy: 1.18, 1.23, and 1.22 wt.% in the alloys edge and only 0.26 wt.% in the center.
(Shtefanyuk et al. 2016)	Al	NaF-AlF3 or KF-AlF3 CR: range of 1.3-2.3. Sc2O3 concentration was kept constant at 1 wt.%. Electrochemical measurement: Dense graphite as counter electrode, spectral pure graphite cased in corundum tube as working electrode, graphite rod surrounded by CO+CO2 gas mixture as reference electrode. Electrolysis: Graphite anode, aluminum cathode. Current density 0.03-0.20 A/cm^2. Temperature: 750, 800, and 850 °C.	Cathode peaks for Al-Sc and K+Al+Sc at -1.55 and -1.82 V, respectively. Al-Sc Alloy could be formed by co deposition of Al and Sc (at current density less than its limiting value). Al-Sc alloy could also be formed by co deposition Al, Sc, and alkali metal (at current density increasing) and secondary reaction of Sc2O3 by alkali metal will proceed. Also, reduction of Sc2O3 dissolved in oxide-fluoride (Sc-O-F) by Al could also occur. Sc content 0.1 to 0.5 wt.%. Current efficiency: 1.5 – 40.3 % depending on cathode current density, temperature, and cation composition of the melt. Sc was uniformly distributed due to mechanical agitation to liquid Al.
Wang et al. (2015)	Mg	LiCl-KCl 50:50 wt.% with 1. (2 wt.%) PrCl3 and (2 wt.%) ErCl3 for galvanostatic electrolysis 2. (2.7 wt.%) PrCl3 - (2.7 wt.%) ErCl3 for separation electrolysis spectrally pure graphite rod as counter electrode and W and Mg as working electrodes. Ag/AgCl2 as reference electrode. Temperature: 823 K. On W electrode the reaction follows one step process: Pr^3+ + 3e^−→ Pr,Er^(III) + 3e^−→Er On Mg electrode: Ln^3+ + 3e^‐ + xMg → MgxLn	Reduction potential of Pr(III)/Pr and Er(III)/Er on Mg electrode are more positive compared to on W electrode because of intermetallic formation. Underpotential deposition of Ln(III) on reactive Mg electrode is because of the formation of Mg-Ln intermetallic. Alloys contain Mg. Applying -1.9 V on potentiostatic electrolysis on LiCL-KCl-2 wt.% PrCl3-2 wt.% ErCl3 on Mg electrode successfully separate Pr and Er with separation efficiency of 95.3%.
Wang et al. (2015)	Al	CaCl2-2 wt.% Sc2O3 Temperature: 860 °C. current density of 0.1, 0.3, 0.5 A/cm^2 or 2.2NaF/AlF3-3 wt.% Al2O3-5 wt.% LiF-2 wt.% Sc2O3 Temperature: 950 °C. 0, 0.5, 1 A/cm^2. Aluminum cathode and high purity graphite rod anode.	CaCl2-2 wt.% Sc2O3 produced 0.025-0.21 wt.% Sc in Al-Sc alloys, with current efficiency of no more than 15%. 2.2NaF/AlF3-3 wt.% Al2O3-5 wt.% LiF-2 wt.% Sc2O3 produced 0.14-0.41 wt.% Sc with current efficiency of 88.7% and 79.27% for cathodic current density of 0.5 and 1 A/cm^2, respectively. By applying zero density, alloy with 0.14 wt.% Sc content could be obtained based on this reaction: Sc2O3 + [with F salt] → ScOFx^−n ScOFx^n- + A; → Sc AlOFx^n- Cryolite is the better molten salt.
Liu et al. (2014)	Al	LiCl-KCl-AlCl3(2.5 wt.%)-Er2O3(1.5 wt.%) melts LiCl: KCl = 58.8: 41.2 mol%. Temperature: 773 K. Al plate cathode and graphite rod anode for electrolysis. -1.6, -1.8, -2.0 and -2.2 V vs. Ag/AgCl for 4 h. Er2O3(s) + 2AlCl3(s) → 2ErCl3(s) + Al2O3(s)	Series of redox signal was measured in CV, SWV, and CP. which correspond to different kinds of Er-Al alloys. ErAl3 was the only intermetallic found in -1.6 and -1.8 V electrolysis samples, but in -2.0 and -2.2 V electrolysis sample, ErAl2 was also found.
Castrillejo et al. (2014)	Al	LiCl-KCl-ScCl3-AlCl3 W and Al electrode for electrochemical measurement Al foils and glassy carbon rod for electrolysis Temperature: 723 and 773 K	Preferential parameters for forming Al3Sc layer were I = -100 mA and the ratio ti=0/telec = 4 and for forming a mixture of Al3Sc, Al2Sc, AlSc, and AlSc2 is I = -175 mA and the ratio ti=0/telec = 2
Su et al. (2014)	Al	LiCl-KCl-Dy2O3-AlCl3 Ln2O3(s) + Al2Cl6(g) → Al2O3(s) + 2LnCl3(l) Tungsten Electrode. Temperature: 773 K One step reaction: Dy^(III) + 3e^− → Dy^(0)	The concentration ratio of Dy^(III) to Al^(III) have a great effect on the co-reduction process. When the ratio was high, three signals appear related to three intermetallic formations. Meanwhile lower ratio registered only two signals.
Yan et al. (2013)	Al	Two types of molten salts: LiCl-KCl-NdCl3 and LiCl-KCl-AlCl3-NdCl3 Mo electrode. LiCl-KCl-NdCl3 Temperature: 723 K The reduction of Nd(III) involving two steps: Nd^3+ + e^− → Nd^(III) Nd^3+ +2e^− → Nd LiCl-KCl-AlCl3-NdCl3 Temperature: 873 K	Underpotential deposition at -1.32 and -1.43 V for two kinds of Al-Nd intermetallic. Al2Nd, Al3Nd, Al and Nd phase were present in the alloy product. Extraction efficiency reached 99.25% after potentiostatic electrolysis for 30h.
Tang et al. (2013b)	Al	LiCl-KCl-Pr6O11-AlCl3 melts Tungsten wire electrode Temperature: 773 K and 903 K. Pr6O11(s) + 3Al2Cl6(l) → 3Al2O3(s) + 6PrCl3(s) + O2 (1) yAl^3+ + 3ye^− = yAl; (2) xPr^3+ + 3xe^− = xPr; (3) xPr + yAl = PrxAly (4) xPr^3+ + yAl^3+ + 3(x + y)e^− = PrxAly	Temperature: 773 K → Al2Pr was the main phase. 903 K → 40 wt.% Pr concentration Main phase was Al2Pr. Pr concentration in the Al-Li-Pr alloy exceeded 40 wt.% when the concentration of AlCl3 reached 8 wt.% in the LiCl-KCl-Pr6O11(1 wt.%).
Liu et al. (2013a)	Al	LiCl-KCl-Gd2O3-AlCl3 Molybdenum and aluminum electrode. Temperature: 773 K. Firstly: Gd^3+ + 2Al^3+ + 9e^− → GdAl2 Then, Gd diffusion into Al electrode: GdAl2 + Al → GdAl3 Overall: Gd^3+ + 2Al^3+ + 9e^− → GdAl3	Extraction efficiency: 89.7% for potentiostatic electrolysis at -1.5 V and 96.5% for galvanostatic at -30 mA, on Al electrode.
Zhang et al. (2013)	Mg	KCl-LiCl-MgCl2 SmCl3-DyCl3 Separation of Sm and Dy -2.20 V (vs. Ag/AgCl reference electrode) and Dy-Mg alloys was formed. Mo electrodes. LiCl-KCl (58:42 mol. %); 0.12 mol/L MgCl2, 0.062 mol/L DyCl3 and 0.062 mol/L SmCl3 823, 873 and 923 K.	Separation coefficient between Sm and Dy decreased at higher temperature. Sm was not co-extracted during the process.
Tang et al. (2013a)	Mg-Li	LiCl–KCl–MgCl2–PrCl3 system. Tungsten wire electrode. Spherical bulk alloys were obtained by potentiostatic and galvanostatic electrolysis at −1.85 V for 3 h at 893 K and −6.00 A/cm^2 for 2 h at 913 K, respectively.	Pr and Li contents in the alloys were regulated by changing the concentration of PrCl3 and MgCl2 in the melts. The proportion of Pr was controlled from 9.73 to 19.93 wt.% in Mg–Pr alloys and the ratio of Li changed from 0.10 to 37.27 wt.% in Mg–Li–Pr alloys.
Wang et al. (2013)	Mg	MgCl2−LaCl3−KCl with MgO, La2O3, and CaF2 T=750 °C. Tungsten rod cathode. The mass ratio of KCl to LaCl3 to MgCl2 was fixed at 13:4:3.	Current efficiency decreases with the increase in MgO or La2O3 concentration. MgO content have a higher effect to current efficiency, compared to La2O3 content. Formation of LaOCl was confirmed and decreased the current efficiency and mass of the alloy.
Qian et al. (2016)	Al	Na3AlF6- 4 wt.% MgF2 - 2 wt.% CaF2 - 2 wt.% Sc2O3 - (0.05-1 wt.%) ZrO2 Al cathode, Graphite anode. 1 A/cm^2 for 2 h. Temperature: 960 °C.	Al3(Sc,Zr) particles were formed in the alloy. Zr^4+ reduction proceed in a two steps reaction. Maximum Sc content is 0.32 wt.%.
Liu et al. (2012)	Al	Potassium cryolite (CR= 1.22) melts. KF-AlF3- 2 wt.% Sc2O3 melts. Al liquid cathode. Temperature: 1023 K.	Sc content were 0.44 and 0.73 wt.% with current density of 0.5 and 1.9 A/cm^2, respectively. Reduction of Sc takes place at -0.9 V on Al liquid cathode and Al3Sc formation is electrochemically irreversible. Ultrasound applications are increasing Sc content up to 1.1 wt.% from 0.73 wt.%
Guan et al. (2012)	Al	NaF/AlF3-Al2O3-LiF-Sc2O3 Liquid Al as cathode. Temperature: 950 °C. 2.2NaF/AlF3-5 wt.% LiF and 3 wt.% Al2O3 and Sc2O3	Without applying any current to the system, alloy having 0.14 wt.% of Sc could be produced by this reaction: Sc2O3 + [F containing salt] = ScOFx^n- ScOFx^n- + Al = Sc + AlOFx^n- Sc + Al = (Al-Sc) 0.2-0.5 wt.% of Sc in the alloy could be produced and Sc are uniformly distributed in the alloys. Sc concentration in the alloy is related to the cathode current density. Ultrasound could create shaking at the interface between liquid Al cathode and the electrolyte. This is increasing the concentration of Sc in the alloys.
Liu et al. (2013b)	Al	KF-AlF3 (CR:1.3), 2%Sc2O3 Liquid Al cathode. Temperature: 1023 K.	Sc content in the alloy was 0.44-0.73 wt.% on current density 0.5-2.0 A/cm^2, respectively.
Zhang et al. (2012b)	Mg-Li	LiCl-KCl-KF-MgCl2-La2O3 Mo electrode. Temperature: 923 and 943 K. Current density: 6.37, 9.55, 12.74 and 15.92 A/cm^2	α+Mg17La2, α+β+Mg17La2 and β+LaMg3 Mg-Li-La are obtained depending on the lithium and lanthanum content β phase dissolves more La than α phase.
Cao et al. (2011)	Mg-Li	LiCl-KCl-MgCl2-K3ErCl6 Molybdenum electrodes. Temperature: 600 °C.	Er5Mg24 phases confirmed by XRD. Er are located on the grain boundary. the electroreduction of Er (III) proceeded in a one-step process involving three electrons.
Guo et al.(2011)	Al	LiF-ScF3-ScCl3 – Sc2O3 Current 2-4 A. Temperature: 850 °C.	Maximum of 3.437% of Sc content in the alloy.
Yang et al. (2010)	Mg	Electrolyte: LiF-YF3 Tungsten and platinum electrode. MgO-Y2O3 with weight ratio of MgO/Y2O3 = 4/1 Temperature: 1050 °C. Tungsten electrode.	Deposition of Mg and Y are all one step reduction. System controlled by diffusion of ions from electrolyte to electrode interface. Magnesium and Yttrium will be decomposed at -0.45 and 0.63 V, respectively.
Chen et al. (2010)	Mg	LiCl-KCl-YbCl3 (2 wt.%) Temperature: 500 °C. Mg rod electrode.	V: -1.85 (vs Ag/AgCl) for 12 hr → ~200 nm Mg2Yb film thickness. V: -2.50 (vs Ag/AgCl) for 2.5 hr → ~450 μm Mg2Yb film thickness The reduction process of Yb^3+ is stepwise reactions which are single-electron and double-electron reversible charge transfer reactions. The speed control step was a diffusion-controlled step which is caused by concentration polarization.
Harata et al. (2009)	Al	CaCl2–Sc2O3 melt on liquid Al electrode. Temperature: 1173 K. When Sc2O3 is dissolved in molten CaCl2 Cathode: Sc^3+ +3e^−→ Sc(l, in Al liq.) Anode: C(s)+xO^2−(in salt) → COx(g) + 2xe^− When Sc2O3 is insoluble in molten CaCl2, it sinks into the bottom of the crucible and in contact with the liquid Al cathode. Cathode: Sc2O3(s) + 6e^− → 2Sc(l, in Al liq.) + 3O^2− Anode: C(s) + xO^2−→ COx(g) + 2xe^− Calciothermic reduction is also could occur when Sc2O3 is insoluble. Cathode: Ca^2+ +2e^−→ Ca(l) Anode: 2Cl^−→ Cl2(g) + 2e^− ; C(s) + xO2−→ COx(g) + 2xe^− Sc2O3(s) + Al(l) + 3Ca(l) → Al–Sc(l) + 3CaO(l)	Al-Sc alloys produced has 99 wt.% purity and contain less than 0.65 wt.% Ca. Sc concentration in the alloy was 0.81 – 32.3 wt.%.
Castrillejo et al. (2006).	Al	LiCl–KCl with ErCl3 at W and Al electrodes. Temperature: 673 – 823 K. Er^3+ + 3e^− → Er on W cathode. Er^3+ + 3Al + 3e^− → ErAl3 on Al cathode	The usage of inert W material results in different behaviors compared to using Al as the electrode. Er3Al confirmed by SEM-EDX and XRD.

5. Discussion, implications, and outlook

This article has reviewed research and development of the various process routes to produce LM-RE alloys. Evidence from the literature shows that the production of LM-RE alloys using alternative routes, rather than just mixing the metals of LM and RE, is still an ongoing process. Many of the processes remain at either laboratory and/or pilot scales and further studies are required to scale up the processes to mass production.

Alternative routes to produce LM-RE alloys have the potential to be more economical compared to direct alloying with RE metal. The price of RE compounds, especially RE oxides, is approximately 50% cheaper compared to RE metals which are currently used as the raw materials for LM-RE alloy production. The direct metallothermic reaction with RE compounds appeared to be economically better from a raw material and capital cost perspective, especially if the RE oxide is used as the precursor material. Inclusion of oxides will be unavoidable, but the alloy could be refined through common industry processes such as density-based separation or filtration. The use of RE oxide as the starting material also means that this process is simplest compared to the others, indicating potential for easier scalability. Furthermore, RE elements are widely traded in the form of their oxides, which mean that additional conversion processes will not be required like other processes which require RE halides. The direct metallothermic reaction process can also be conducted without having to use large amounts of salts, which avoids the problem of post treatment of the salt by-product. Even though the processing temperature is like the other two processes (Mg system at 660–800°C and Al system at 700–975°C, respectively), a salt-free processing route means that the process does not require expensive corrosion resistance equipment. In terms of yield of the process, for the most studied system, which is the scandium-LM alloy system, studies have demonstrated that a 96.72% yield could be achieved for making a 2.9 wt.% Al-Mg-Sc alloy when using ScCl₃ as a starting material (Xiao et al. 2020) and a 92.8% yield for making a Al-Sc alloy when using Sc₂O₃ as a starting material (Tarcy and Foster 1988). Compared to the multistage reduction of RE compounds using a molten salts method, the yield is similar with the yield for the direct metallothermic reduction form RE compound method which reaches ~96% when ScF₃ with fluoride salts were used to make a Al-0.91% Sc alloy (Skachkov 2018) and 91% when Sc₂O₃ was used with cryolite for making Al-2 wt.% Sc-0.5 wt.% Y alloy.

The alloy produced from a multistage indirect reduction of RE compound route is expected to contain less oxide impurities compared to the direct metallothermic reduction route because most of the oxide is dissolved in the salts. Unreacted RE in the salts requires further processing either for recovering the RE or conditioning of the salts to be reused in the process again. A high amount of salts is required for this process and failure to reuse the salts would lead to waste problems. Most European countries are prohibiting salt cake to be deposited in landfill. It is classified as toxic and hazardous waste by the European Union (Padilla et al. 2022; Tsakiridis 2012). If disposed improperly, it can pollute ground water and when in contact with water or moisture, it can release explosive, toxic and unpleasant smelling gasses such as CH_4. NH₃, H₂S, PH₃, and H₂. Molten salts at high temperature also tend to be volatile and corrosive. Hence, equipment at larger scale will need to survive those environments.

The electrolysis route will produce the highest purity of the LM-RE alloy compared to the other routes. The composition of the alloy could also be tuned by changing the process parameters, such as the electrolyte composition and the applied current density/voltage. Two different cathodes can be used, inert or reactive. The use of an inert cathode can cause the alloy sink to the bottom of the cell in the form of droplets because of its poor wettability. In contrast, a reactive cathode will react with the deposited RE and form intermetallics, which causes under-potential deposition of RE metal. This results in a lower operating temperature of the molten salts electrolysis process (Han et al. 2016). There is also the possibility for RE segregation in the produced alloy, however, this may potentially be reduced by applying ultrasound to the process (or other means to stir). Currently, there is no industrial size electrolysis equipment specifically tailored for producing LM-RE alloys although the modified Hall-Héroult process is a promising route, especially for Al-RE alloys. The high temperature of the process increases the volatility and corrosivity of the molten salts, thus requiring the cell to be designed to withstand these corrosive environments (Yin et al. 2021). Compared to the other routes, process control for the electrolysis route is much more complex, however implementation in an industrial scale is possible, with the current scale of study being a 15–20 kg bath for making Al-Sc alloy (Wang 2022).

6. Concluding remarks

RE elements are emerging one of the alloying elements that could be used to improve the properties of light metal alloys. Currently, most LM-RE alloys are made by the expensive route of direct melting of a LM and a RE metal. This process is expensive because RE metals are historically more expensive compared to the LM itself. Research is still needed to develop alternative low-cost and high-efficiency process which use different routes and different, lower cost RE sources.

The review has confirmed that there are three main alternative routes capable of producing LM-RE alloys. The direct metallothermic reduction of RE compound route and the electrolysis route (especially a modified Hall-Héroult route) are the most promising alternatives and should be the focus of future studies. The electrolysis route might seem to have several advantages compared to the others through the demonstrated scalability and existing infrastructure associated with Al production, but the Al-RE alloy process is more complex. The metallothermic reduction of RE compound seems to be the simplest but research still need to improve the efficiency of the process. Further studies are also still needed to scale up the process from laboratory. All routes have their own disadvantages that still need to be addressed for them to be successful at a commercial scale.

Successful development of a new process would lower the price of the LM-RE alloy, and potentially expand its application to different areas. For example, Mg-RE alloys are promising materials for making biocompatible materials as their mechanical properties could make them suitable to be used in biomedical applications. In addition, the potential of utilizing low-cost RE supply chain metals such as Ce and La as alloying elements for producing Al alloys that display similar properties to commercial Al alloys – and require minimal heat treatment – could potentially shift the industry into making more Al-RE alloys. Lightweight and strong material such as Mg-RE and Al-RE alloys will have a significant role in improving the efficiency of a process. For example, 10% weight reduction could deliver 6% fuel economy improvement and 14% increase in range for internal combustion vehicle and electric vehicle, respectively (Luo 2021). The increased demand for improved fuel economy will also drive the demand for light alloys.

Acknowledgments

This work is part of a PhD study by main author under the joint PhD program between Swinburne University of Technology, Australia, and the National Innovation and Research Agency (BRIN), Indonesia. The authors are grateful with the joint funding support from Swinburne University of Technology; the National Innovation and Research Agency-Indonesia (BRIN); the Commonwealth Scientific and Industrial Research Organization-Australia (CSIRO); Magontec Ltd.; Grandfield Technology Pty. Ltd.; and Platina Resources Ltd.

Disclosure statement

The author declares that he has no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Additional information

Funding

The work was supported by the Swinburne-CSIRO-Magontec-Grandfield Tech-Platina Resources Scholarship .